Professional Documents
Culture Documents
A Treatise
Edited by R . K. Willardson
WILLARDSON CONSULTING
SPOKANE, WASHINGTON
Albert C. Beer
BATTELLE COLUMBUS LABORATORIES
COLUMBUS, OHIO
SEMICONDUCTORS
AND SEMIMETALS
VOLUME 25
Diluted Magnetic Semiconductors
Volume Editors
JACEK K. FURDYNA
DEPARTMENT OF PHYSICS
UNIVERSITY OF NOTRE DAME
NOTRE DAME, INDIANA
JACEK KOSSUT
INSTITUTE O PHYSICS
POLISH ACADEMY OF SCIENCES
WARSAW, POLAND
ISBN 0-12-752125-9
PRINTED 1N THE UNITED STATES OF AMERICA
88899091
9 8 7 6 5 4 3 2 1
To the memory of
Professor Jerzy Mycielski
(1930-1986)
Contributors
Numbers in parentheses indicate the pages on which the authors contributions begin.
W. M. BECKER,
Department of Physics, Purdue University, West Lafayette,
Indiana 47907 (35)
Foreword
xiv
FOREWORD
FORE WORD
xv
The work on the book spanned a considerable period of time (as is perhaps
unavoidable in the case of any publication involving a large number of
contributing authors). The last several years have witnessed a particularly
rapid growth of the number of publications concerning DMS. In particular,
several important new contributions emerged when the book was already in
the middle of its preparation process. For this reason, although they are of
unquestionable importance, it was not possible to properly represent them
here. To do them partial justice without delaying the appearance of the book,
we shall call attention to some of the new developments at least in the
Foreword in the form of the list below.
(1) Considerable progress is to be noted in the experimental and
theoretical studies concerning the relative position of Mn-derived states and
various density-of-states features in the valence and conduction bands. The
experimental part has made extensive use of the synchrotron radiation
photoemission technique (Franciosi etal., 1985a,b, and Wallet al., 1986a,b).
The theoretical effort (consisting of a series of papers by Ehrenreich, Hass,
Larson, and collaborators) was recently reviewed by Ehrenreich et al. (1987).
Based on these studies it was, for example, possible to determine that the
dominant mechanism of the Mn-Mn exchange coupling is that of superexchange, and to calculate theoretically the values of the relevant exchange
constants.
(2) Successful growth by molecular beam epitaxy of superlattices and
quantum well systems involving DMS opened a vast field of entirely new
opportunities (for a recent review of the work on DMS superlattices, see
Furdyna et al., 1987).
(3) MaterialsinvolvingMn+2inA1VBV1andA11BV
hosts (e.g., Pbl-,Mn,Te
and (Cdl-,Mn,)2As3, respectively) were prepared in an attempt to enlarge
the family of DMS. Some of the references to these works are given in
Chapter 1 in the section devoted to new materials.
(4) DMS materials containing substitutional Fe ions instead of the
traditional Mn+2(for reviews, see Mycielski, 1986a, and Reifenberger and
Kossut, 1986) proved to possess features absent in Mn-based DMS. A
considerable enhancement of the low-temperature mobility of electrons in
Hgl-,Fe,Se (0.0003 5 x 5 0.01) is the most spectacular example of these
new properties.
( 5 ) Studies of the conductivity in the weakly localized regime and in the
vicinity of the nonmetal-metal transition have shown that the electronic spin
plays an important role in the related processes (see Stankiewicz el al.,
1986a,b, Shapira et al., 1985, Wojtowicz et al., 1986, and Sawicki et al.,
1986).
(6) Investigations of the magnetization induced in DMS by circularly
xvi
FOREWORD
polarized radiation incident on the sample (Krenn et al., 1985) and the
magnetic spectroscopy (Awschalom and Warnock, 1987) open new
possibilities, particularly in the context of time-resolved studies of magnetic
systems.
Expecting that the book would be read at a rate of one chapter at a time,
it was the editors decision to give the contributing authors a free hand in their
choice of units, notation, etc., provided that they were consistent within
their own chapter. For the same reason, a certain degree of overlap between
the material contained in some chapters was not frowned upon inshe hope
that the contributions would, in this way, be more self-contained.
Finally, let us comment on the terminology used in connection with DMS.
Quite often these materials also are referred to as semimagnetic semiconductors. Being aware that this is predominantly a matter of taste, we
nevertheless prefer the term diluted magnetic semiconductor, since this
term conveys the physical nature underlying the properties of the material.
It refers to a magnetic semiconductor (a name already established, as for
EuSe) that has been diluted by the intervening non-magnetic constituent,
giving DMS all their physical characteristics.
The authors of the chapters contained in this volume were, as a rule,
actively involved in the research in the areas which they respectively describe.
One of the authors in particular, Professor Jerzy Mycielski, had a very
special impact on the development of the physics of DMS. Over and above
the contents of his own chapter in this book, his contributions consisted of
formulation of many original and important ideas throughout the entire
period of DMS studies. The papers of his authorship were among the first
that put DMS on the map, while his hypothesis concerning a spatial
ordering of charges within the donor system in Hg, -xFexSe (Mycielski,
1986b) will remain his last. It was with profound sorrow that we learned of
his death in February, 1986, before he could see his contribution in print. To
his memory, we would like to dedicate this volume.
References
Awschalom, D. D., and Warnock, J. (1987). DilutedMagnetic (Semimagnetic)Semiconductors,
edited by R. L. Aggarwal, J. K. Furdyna, and S. von Molnar (Materials Research Society
Symposia Proceedings, Pittsburgh, PA) Vol. 89, p. 71.
Delves, R. T. (1963). J. Phys. Chem. Solids 24, 885.
Delves, R. T., and Lewis, B. (1963). J. Phys. Chem. Solids 24, 549.
Ehrenreich, H . , Hass, K. C., Larson, B. E., and Johnson, N. F. (1987). Diluted Magnetic
(Semimagnetic) Semiconductors, edited by R. L. Aggarwal, J . K . Furdyna, and S. von
Molnar (Materials Research Society Symposia Proceedings, Pittsburgh, PA) Vol. 89, p. 187.
Franciosi, A . , Chang, S., Caprile, C., Reifenberger, R., and Debska, U. (1985a). J. Vac. Sci.
Techno/. A3. 926.
FOREWORD
xvii
Franciosi, A., Reifenberger, R., and Furdyna, J . (1985b). J. Vac. Sci. Technol. A3, 124.
Furdyna, J. K., Kossut, J., and Ramdas, A. K. (1987). OpticalPropertiesofNarrow-GapLowDimensionalStructures, edited by C. M. Sotomayor Torres, J. C. Portal, J. C. Maan, and
R. A. Stradling (Plenum Press, New York) p. 135.
Krenn, H., Zawadzki, W., and Bauer, G. (1985). Phys. Rev. Lett. 55, 1510.
Mycielski, A. (1986a). Diluted Magnetic (Semimagnetic) Semiconductors, edited by R. L.
Aggarwal, J. K. Furdyna, and S. von Molnar (Materials Research Society Symposia
Proceedings, Pittsburgh, PA) Vol. 89, p. 159.
Mycielski, J. (1986b). Solid State Commun. 60, 165.
Reifenberger, R., and Kossut, J. (1987). J. Vac. Sci. Technol. AS, 2995.
Samarth, N. (1986). Ph.D. Thesis, Purdue University, unpublished.
Sawicki, M., Dietl, T., Kossut, J., Igalson, J., Wojtowicz, T., and Plesiewicz, W. (1986). Phys.
Rev. Lett. 56, 508.
Shapka, Y., Ridgley, D. H., Dwight, K., Wold, A., Martin, K. P., Brook, J. S., and Lee,
P. A. (1985). Solid State Cornmun. 54, 593.
Spalek, J., Lewicki, A., Tarnawski, Z., Furdyna, J. K., Galazka, R. R., and Obuszko, Z.
(1986). Phys. Rev. B33, 3407.
Stankiewicz, J., von Molnar, S., and Giriat, W. (1986a). Phys. Rev. B33, 3573.
Stankiewicz, J., von Molnar, S., and Holtzberg, F. (1986b). J. Magn. Magn. Muter. 54-57,
1217.
Wall, A., Caprile, C., Franciosi, A,, Reifenberger, R., and Debska, U. (1986a). J. Vac. Sci.
Technol. A4, 818.
Wall, A., Caprile, C., Franciosi, A., Vaziri, M., Reifenberger, R., and Furdyna, J. K. (1986b).
J. Vac. Sci. Technol. A4, 2010.
Wojtowicz, T., Dietl, T., Sawicki, M., Plesiewicz, W., and Jaroszynski, J. (1986). Phys. Rev.
Letters 56, 2419.
CHAPTER 1
and
J. K. Furdyna
DEPARTMENT OF PHYSICS
UNIVERSITY OF NOTRE DAME
NOTRE DAME, INDIANA
I. INTRODUCTION. . . . . . . . . . . . . . . . . . . .
11. CRYSTAL
STRUCTURE
AND COMPOSITION
OF DMS . . . . . .
1. Overview. . . . . . . . . . . . . . . . . . . .
2 . The A"BV' "Host "Lattice. . . . . . . . . . . . . .
3 . Crystal Structure and Lattice Parameters of DMS . . . .
4. Microscopic Structure of the DMS Lattice . . . . . . .
111. PREPARATION
OF DMS CRYSTALS.
. . . . . . . . . . . .
5. Preparation of Starting Materials. . . . . . . . . . .
6 . Crystal Growth of DMS Alloys . . . . . . . . . . .
IV. THE"NEW"DMS. . . . . . . . . . . . . . . . . . .
1. DMS Films and Superlattices. . . . . . . . . . . . .
8 . A1',FexBV' Alloys . . . . . . . . . . . . . . . .
9. Manganese- and Rare-Earth-Based AIVBV'Alloys. . . . .
10. MisceIIaneous New DMS Systems. . . . . . . . . . .
REFERENCES.. . . . . . . . . . . . . . . . . . . .
3
3
4
7
11
14
14
17
21
28
28
30
31
32
I. Introduction
During the past decade a great deal of attention has been given to semiconductor compounds whose lattice is made up in part of substitutional
magnetic ions (Galazka, 1979; Gaj, 1980; Furdyna, 1982; Brandt and
Moshchalkov, 1984). Most of these materials are based on A"BV' compounds, with a fraction of the group-I1 constituent replaced by manganese.
Cdl-,Mn,Se and Hgl-,Mn,Te are examples of such systems. These ternary
1
Copyright 0 1988 by Academic Press. Inc.
All rights of reproduction in any form reserved.
ISBN 0-12-752125-9
alloys (or "mixed crystals") are of interest for several reasons. Firstly, their
semiconducting and structural properties, such as the energy gap, the lattice
parameter, etc., can be varied in a controlled fashion by varying the
composition (much as in non-magnetic ternary semiconductors, e.g.,
Hgl-,Cd,Te or Gal-,Al,As). Secondly, as dilute magnetic alloys, these
materials are of interest for their magnetic properties, e.g., the spin-glass
transition (Galazka, 1982; Oseroff, 1982), antiferromagnetic cluster
formation (Dolling, 1982), magnon excitations (Ramdas, 1982), and other
magnetic effects of current interest. Thirdly, the presence of magnetic ions
in the lattice leads to spin-spin exchange interaction between the localized
magnetic moments and the band electrons (Gaj, 1980). This interaction
affects the energy band and impurity level parameters of these materials
(e.g., by enhancing the electronic g-factors), resulting in new physical effects.
Such novel and potentially important phenomena as the magnetic-fieldinduced overlap between valence and conduction bands occurring in
Hgl -,MnxTe and Hgl-,Mn,Se (Galazka and Kossut, 1980), extremely large
Faraday rotation in Cdl-,Mn,Te (Gaj et al., 1978), giant negative magnetoresistance associated with hopping conduction in Hgl -,Mn,Te (Mycielski
and Mycielski, 1980), and the magnetic polaron observed in Cdr-,Mn,Se
(Nawrocki et al., 1981; Diet1 and Spalek, 1983) are all consequences of the
spin-spin exchange interaction.
The latter two features-magnetic properties as such, and the exchange
interaction with band electrons-distinguish these materials from other
semiconductors. For this reason, this group of alloys has been collectively
labeled as "semimagnetic semiconductors" or "diluted magnetic semiconductors" (DMS). So far, practically all research on DMS involved A"BV'
semiconductor compounds alloyed with manganese, although recently there
has been a gradual increase of research activity on A1'BV'-Fe alloys (e.g.,
Hgl-,Fe,Te, Cdl-,Fe,Te) and AIVBV'-Mn alloys (e.g., Pbl-,Mn,Te). The
investigations of these latter materials are still at a relatively early stage.
In the present chapter we shall therefore concentrate on the A?,Mn,BV1
alloys.
The material is presented as follows. We first review the crystal structure
and composition of the various A:'_,Mn,BV' alloys. We show that this can
be done in a unified way for both the zinc blende and wurtzite members of
this group of materials. We then present methods of preparation of bulk
DMS crystals, including materials purification, synthesis of starting binary
compounds, and crystal growth. Finally, we discuss-in a much briefer
fashion-the more recent developments in the field of DMS, such as
A:'_,Mn,BV' epitaxial films and superlattices, A:'-,Fe,BV' alloys, and
AIVBV'-based DMS containing substitutional manganese or rare earth
ions.
1.
1. OVERVIEW
The entire family of ternary A:_,MnxBv alloys, along with their crystal
structures, is presented in schematic form in Fig. 1. The heavy lines in the
figure show the composition ranges in which ternary alloys can be formed,
with cub and hex indicating the crystal structure (zinc blende and
wurtzite, respectively) of the stable phases. For example, Cdl-,MnxTe forms
a ternary alloy of zinc blende structure with x up to 0.77, while Z ~ I - , M ~ , S ~
exhibits zinc blende structure for x < 0.30 and wurtzite structure for
0.30 c x c 0.55. At values of x beyond the heavy lines mixed phases occur.
Given the fact that the stable crystal structures of MnTe, MnSe, and MnS
differ from the crystal structure of the ABV compounds (Pajaczkowska,
1978), it is truly remarkable that the range of the A1/-,MnxBV1 solid solutions
CdTe
ZiS
FIG.1. A diagrammatic overview of the A:,Mn,BV
alloys and their crystal structures. The
bold lines indicate ranges of the molar fraction x for which homogeneous crystal phases form.
Hex and Cub indicates wurtzite and zinc blende, respectively. Cd,-,Mn,S and
Cdl-,Mn,Se also form homogeneous crystalline phases near x = 1.0 in the NaCl structure, but
these are not relevant to the present article.
is so wide. For example, MnTe itself crystallizes in the NiAs structure, yet
in the case of Cdl-,Mn,Te the structure of the "host" CdTe (zinc blende)
survives for x as high as 0.77.
Both crystal structures of the DMS-zinc blende and wurtzite-are very
closely related in spite of the differences in symmetry, in that they are both
formed with tetrahedral (s-p3) bonding, involving the two valence s-electrons
of the group-I1 element and the six valence p-electrons of the group-VI
element. Manganese is a transition metal with valence electrons corresponding to the 4s' orbital. Although manganese differs from the group-I1
elements by the fact that its 3d shell is only half-filled, it can contribute its
4s2 electrons to the s-p3 bonding, and can therefore substitutionally replace
the group-I1 elements in the A"Bv' wurtzite or zinc blende structures.
These properties of manganese are apparently a necessary but not a
sufficient condition for forming stable ternary phases over a wide range of
compositions, since other transition metal elements-even those which readily
form divalent ions-are not nearly as miscible with A"BV' compounds. For
example iron (a transition metal with ionic radius comparable to that of
manganese) is about an order of magnitude less miscible, and the other
transition metal elements even less than that. It would appear, therefore, that
the ease with which Mn substitutes for the group-I1 elements in the zinc
blende and wurtzite structures (and the well-known predisposition of this
element for the divalent state) results from the fact that the 3d orbitals of
Mn++ are exactly half-filled. By Hund's rule, all five spins are parallel in this
orbital, and it would require considerable energy to add an electron with
opposite spin to the atom. In this sense, the 3d5 orbit is "complete," and thus
the Mn atom resembles a group-I1 element.
1.
Gb.
(1)
(3)
c = &d.
(4)
Now it is truly remarkable how close is the behavior of the wurtzite Ax*Bvx
compounds to the ideal case. For example, in the wurtzite crystal CdSe,
a = 4.2985 A and c = 7.0150 A (Pajaczkowska, 1978), giving
C
- = 1.632,
a
= f i x 3.825
A,
a = 3.820 A,
c = 6.260 A
fi x 3.833 A .
b = rII
+ rvI.
Such a radius is associated with a specific atom independent of the environment; for example, the same covalent radius of Se predicts rather accurately
the bond lengths (and thus all other lattice parameters) in the zinc blende
HgSe and ZnSe, and in the wurtzite CdSe. We list in Table I the covalent
radii for the group-I1 and group-VI elements. It is remarkable that the bond
lengths-and thus other lattice parameters-for all zinc blende and wurtzite
A1*Bvlcompounds can be immediately generated, to a close approximation,
from Table I. In Table I1 we compare the bond lengths calculated from Table
I with those measured experimentally.
TABLE I
TETRAHEDRAL
COVALENT
R m n (IN
S
Se
Te
Zn
Cd
Hg
Mn
1.127
1.225
1.405
1.225
1.405
1.403
1.326
A)
1970
1970
1970
1970
1970
1970
We shall show next that the covalent radius remains meaningful in the
context of the Ai'-,Mn,BV1 ternary alloys, providing a very useful guideline
to the variation of the lattice parameter with composition in all DMS
compounds.
1.
TABLE I1
Compound
Calculated
Observed*
ZnS
ZnSe
ZnTe
CdS
CdSe
CdTe
HgS
HgSe
HgTe
2.352
2.450
2.630
2.532
2.630
2.810
2.530
2.628
2.808
2.342
2.454
2.631
2.532
2.633
2.806
2.534
2.634
2.797
3. CRYSTAL
STRUCTURE
AND LATTICE
PARAMETER
OF
A:_,Mn,BV DMS
Crystal structure of ternary alloys formed by substituting Mn for the
group-I1 element in the A1BV lattice is indicated in Fig. 1, where cub
refers to the zinc blende, and hex to the wurtzite structure. As a rule, these
DMS retain the crystal structure of the parent A1Bv compound.
Exceptions to this rule are Znl-,Mn,S and Znl-,Mn,Se: these exhibit the
structure of the cubic A1BV host for low Mn content, but above a certain
value of x , they form in the wurtzite structure. A natural upper limit on the
Mn mole fraction x in DMS is imposed by the fact that the MnBVbinary
compounds do not crystallize in the zinc blende or wurtzite structures.
It is interesting to note that, although the stable phase of HgS itself is the
cinnabar structure (HgS is the only ABV compound to crystallize in this
form), the presence of Mn in Hgl-,MnxS appears to stabilize the sphalerite
structure of this system (Pajaczkowska, 1978). At this time, relatively little
is known about this alloy.
The remarkably systematic behavior of lattice parameters observed in the
A1Bv compounds, as described in the preceding section, carries over to their
A:-,Mn,BV1 derivatives. Two aspects are particularly worthy of note.
First, in all DMS crystallizing in the wurtzite structure, the a :c relationship
predicted by the ideal hexagonal close packing of spheres is obeyed. Thus,
for Zno.dsMno.ssSe, the observed value is c/a = 1.635, as compared to the
= 1.633 (Yoder-Short et al., 1985). Since in the ideal case both
ideal
a and c are determined by the nearest neighbor distance d of the anion (or
cation) sublattice, one concludes that d is the one key parameter which
determines the structural dimensions of the ternary DMS lattice.
Second, the lattice parameters for all DMS obey Vegards law very closely.
This is illustrated by Fig. 2, showing the dependence of the three cubic
tellurides on composition (Furdyna ef a/., 1983). It is gratifying that an
extrapolation to x = I-which indicates the lattice parameter for the
hypothetical zinc blende MnTe-is the same for the three alloys. Thus, in
spite of the fact that zinc blende MnTe does not exist in nature, one may
regard Cdl-,Mn,Te as a true pseudobinary alloy of CdTe and MnTe, and
one can write its lattice parameter in the form
a = (1 - x)aIr-vI
+ XaMn-VI.
(5)
0.6
0.8
I .o
Mole fraction X
FIG.2. Lattice parameter as a function of Mn mole fraction x for telluride DMS. Note the
Vegard law behavior, and the convergence of the extrapolated lines at a single point (6.334 A ) ,
which determines the lattice parameter for the zinc blende phase of MnTe. [After Furdyna et
a/.(1983), with Hgl-,MnxTe data taken from the paper by Delves and Lewis (1963).]
0.2
0.4
MnTe
Mn Se
Mn S
MOLE FRACTION X
FIG.3. Mean cation-cation distances d as a function of Mn mole fraction x for A!~,MnxBvl
alloys. Analytic expressions correspondingto the figures are given in Table 111. [After YoderShort et at. (1985).]
10
W. GIRIAT A N D J. K . FURDYNA
TABLE 111
MEANCATION-CATION
DISTANCES
d IN A:'-,Mn,BV1 ALLOYS
Material
Upper limit
of x
Znl-,Mn,S
Cdl-,Mn,S
Hgl-,MnxS
Znl-xMn,Se
Cdl-,Mn,Se
Hgl-,Mn,Se
Znl-,Mn,Te
Cdl-,Mn,Te
Hgl-xMnxTe
0.60
0.50
0.37
0.57
0.50
0.38
0.86
0.77
0.75
Mean cation-cation
distaace d
(A)
3.830
4.123
4.139
4.009
4.296
4.301
4.315
4.587
4.568
+ 0.139~
- 0.151~
-
0.167~
+ 0.164~
- 0.123~
- 0.123~
+ 0.168~
- 0.105~
-
0.080~
obtained via Eqs. (2)-(4), are listed in Table I11 for all A:~,Mn,B'" alloys
(after Yoder-Short et af., 1985).
In analogy with Table 11, we list in Table IV the extrapolated bond lengths
b for zinc blende and/or wurtzite MnTe, MnSe, and MnS, along with the
covalent radius of manganese obtained from these values and.from the data
listed in Table I. The covalent tetrahedral radius for Mn could in principle
be obtained from measurements on zinc blende and/or wurtzite phases of
MnSe and MnS. These tetrahedrally bonded phases, however, are difficult
to prepare, and only limited data are available on the binary compounds. We
believe therefore that the value of d for the three manganese chalcogenides
obtained by extrapolation of the DMS data to x = 1, as illustrated in Figs.
2 and 3, are more reliable than the literature values obtained from measurements made on the binary compounds MnS and MnSe themselves
(Pajaczkowska, 1978). The internal consistency of the extrapolated DMS
values further corroborates the reliability of the values listed in Table IV,
which ultimately provides the covalent tetrahedral radius of Mn,
rMn = (1.326 f 0.018)A (Yoder-Short et al., 1985).
It should be noted that, apart from its fundamental importance, the precise
knowledge of the lattice parameter is of considerable practical interest in that
it provides a convenient determination of crystal composition in ternary
alloys. It is a fortunate circumstance in this respect that in the case of all
A:'_,Mn,Bv' alloys the linear variation of the lattice parameter with x is very
large (contrast this with, e.g., Hgl-,Cd,Te), making this approach to the
determination of composition quite reliable. Typical x-ray techniques yield
the lattice parameter with precision of 1 part in lo4 or better, thus giving the
composition within 1% for most DMS.
1.
11
TABLE IV
BONDLENGTHS
b AND Mn COVALENT
RADIUSrMn IN Mn CHALCOGENIDES
Compound
b (A)
rMn(A)
MnS
MnSe
MnTe
2.432
2.557
2.746
1.305
1.335
1.341
1.326
4. MICROSCOPIC
(LOCAL)
STRUCTURE OF THE
DMS LATTICE
12
W. GIRIAT A N D J. K . FURDYNA
The Cd-Te bond length is approximately unchanged from its value in pure
CdTe, and the Mn-Te bond length is approximately that predicted for the
hypothetical zinc blende MnTe (for which the lattice parameter is predicted
in Fig. 2). The actual bond length values are shown as a function of x in
Fig. 4.
These results are not in contradiction with x-ray diffraction data, which
show Vegards law behavior for the ternary lattice. X-ray diffraction is not
sensitive to the topological details surrounding an individual lattice point,
but determines the lattice parameter averaged over all anion and cation sites.
It is precisely this averaged a(x) which varies linearly with x. The EXAFS
results, however, cannot be reconciled with the interpretation that Vegards
law is obeyed on the microscopic scale by each of the two interpenetrating
fcc lattices (the one occupied by Te atoms, the other bya random distribution
of Cd and Mn). Rather, the two sublattices must adjust locally to accommodate the requirement of keeping the respective Cd-Te and Mn-Te bond
lengths as nearly constant as possible.
While in reality both the anion and the cation lattices are thus distorted
from fcc, it is realistic to represent the microscopic behavior in terms of a
model proposed by Balzarotti et al. (1984) as follows. The cation sublattice,
occupied randomly by Cd and Mn, is to afirst approximation not distorted.
It may be regarded, to first order, as an fcc sublattice, obeying Vegards law
in the spirit of the virtual crystal approximation. The atoms on the anion
sublattice, on the other hand, adjust in order to accommodate the requirement that the Cd-Te and the Mn-Te bond lengths remain as constant as
possible. The anion sublattice, then, is distorted, and is no longer fcc.
2.80t
Cd-Te
2.75 -
Q
Mn-Te
0.1
0.3
0.5
Mole Fraction X
0
0
0.7
FIG.4. Actual Mn-Te and Cd-Te bond lengths as a function of composition, determined by
EXAFS. [After Balzarotti ef al. (1984).]
13
The justification for the above picture (i.e., for the fact that the two
sublattices behave differently) is as follows. In Cdl-,Mn,Te, a given Te site
can be surrounded by 0, 1, 2, 3, or 4 Mn nearest neighbors, the remaining
ions of the tetrahedral quartet being Cd, respectively. When a given anion
is surrounded by four like cations, it will remain in its equilibrium position
(assuming nearest neighbor interactions to be dominant). In the three cases
of a mixed Cd and Mn environment, however, the anion will move from its
equilibrium position so as to adjust its distance as closely as possible to the
Cd-Te and Mn-Te bond lengths. Since the Mn-Te bond length is shorter
than the Cd-Te bond, the Te ion will be displaced from its equilibrium
position towards Mn and away from Cd.
The assumption of the model that, unlike the anions, the cation sublattice
is not distorted is plausible in that each cation is always surrounded by a
tetrad of like atoms (Te in Cdl-,Mn,Te). If a cation were to move to a closer
position with respect to one anion, it would automatically move further away
from other identical anions. In reality, some distortion of the cation sublattice must also occur, but on the basis of the above argument it is likely to
be considerably smaller than the distortion of the anion sublattice and may
be neglected at this stage of our understanding.
Theoretical calculations for the proposed model carried out by Balzarotti
et a/. (1984) are in excellent agreement with observation. The authors show
that if all possible configurations of all possible tetrahedra are considered,
the average anion sublattice obtained in this way turns out to be exactly the
fcc sublattice, with a(x) given by Vegards law. X-ray diffraction, by
averaging over all possible sites existing in the crystal, measures precisely
such an average, Vegard-law-obedient behavior.
The ideas advanced by Balzarotti et al. are expected to apply to other
random ternary zinc blende alloys of the A:-,Mn,BV1 type, as well as to
analogous wurtzite systems. The constant value of the anion-cation distance
that emerges from this picture, gives added strength to the concept of the
tetrahedral radius (see earlier discussion) as a physically meaningful concept.
While important from a directly crystallographic point of view, these
results are also expected to bear on a number of other physical phenomena.
For example, in light of these results, the microscopic environment of atoms
in the zinc blende and wurtzite ternaries are no longer strictly cubic and
hexagonal, respectively, and many optical transitions-particularly those
involving local mode excitations-that are forbidden by the ideal symmetry
(as manifested in the binary parent compound) may be relaxed in these
random alloys. Furthermore, magnetic properties of DMS are expected to
be dominated by super-exchange interaction, i.e., by Mnf+-Mn++ interactions that are mediated through the intervening anion (Larson et al., 1985;
Spalek et al., 1986). The anion-Mn distance, as well as the angle subtended
14
by the Mn++ pair and the anion, are expected to bear directly on this
exchange process, and thus the new insights into the bond length are likely
to play an important role in understanding the magnetic properties of DMS.
111. Preparation of Bulk DMS Crystals
15
FIG.5. Arrangement for vacuum sublimitationof Mn: (1) furnace; (2) external ceramic tube;
(3) inner ceramic tube; (4) manganese before sublimation; (5) manganese after sublimation;
(6) exit to vacuum system.
16
W . GIRIAT A N D J . K . FURDYNA
OF
m.p. ("C)
4A)
Structure
MnTe
1165
a = 4.1475
c = 6.710
NiAs
MnSe
MnS
HgTe
HgSe
HgS
CdTe
CdSe
1510
1530
670
779
825
1098
1260
NaCl
NaCl
zinc blende
zinc blende
cinnabarhinc blende
zinc blende
wurtzite
CdS
1405
ZnTe
ZnSe
ZnS
1293
1526
1722
5.462
5.223
6.460
6.084
5.851
6.481
u = 4.2985
c = 7.0150
u = 4.1368
c = 6.7163
6.1037
5.6687
5.4093
wurtzite
zinc blende
zinc blende
zinc blende
17
6 . CRYSTAL
GROWTH
OF DMS ALLOYS
Although similarities exist between the methods of preparation of specific
DMS alloys, there are also significant differences because of the wide range
of melting points and vapor pressures spanned by this family of materials.
For this reason, it is more convenient-at the risk of some repetition-to
present the preparation procedure for each alloy system separately.
(a) Cdl-,Mn,Te
Cdl-,MnxTe crystallizes in the zinc blende structure in the range
0 < x I0.77 (Triboulet and Didier, 1981). For xexceeding this limit, mixed
phases occur. The phase diagram for this system is shown in Fig. 6 . The
18
I 150-
Il5Ot
Liquid
1100Liquid+solid
?
w 1050-
[r
Solid
904
Oll
1 3
near-coincidence of the solidus and liquidus curves indicates that this system
is ideally suited for growth by the Bridgman method. The melting point,
which for pure CdTe occurs just below llOOC, decreases slightly as the
Mn content increases. For example, for x = 0.77 the melting point of
Cdl-xMnxTeis about 30C below that of pure CdTe.
Crystal growth by the Bridgman method is typically carried out with a
starting temperature of about 1120C,using a temperature gradient of about
50C/cm in a vertical furnace. The charge is lowered at a speed of 4 mm/h.
High quality monocrystals, with typical dimensions of 15 mm in diameter
and 10 cm in length, can be obtained in this manner. The macroscopic quality
and homogeneity of the Cdl-*Mn,Te crystals is, at this stage, the best of the
DMS alloy family, probably because of the unique coincidence of the solidus
and liquidus curves in the phase diagram (Fig. 6). Also, because of this latter
feature, the composition of the Bridgman crystals remains close to the
starting (i.e., nominal) value. In their mechanical properties, single crystals
19
- I500
Liquid
rn
- ,I
---
-1200
-1100
Wurtzite ond
rock solt
Wurtzite
Ya-
-1000
I
F
g
- 900 E
1.0
Cd Se
0.8
0.6
0
I
0.4
0.2
000
I - 700
0.0
MnSe
Mole Fraction X
FIG.7. Phase diagram for CdSe-MnSe. [After Cook (1968).]
20
and liquidus lines in the phase diagram are very close, the effect of segregation is not very serious (although it is somewhat more pronounced than in
Cdl-,Mn,Te), and the crystals display only a small composition gradient
along the direction of growth.
(c) Cdl-,Mn,S
Single crystal growth of this important system is at a much more preliminary stage than that of the preceding two alloys. The phase diagram of
CdS-MnS solid solutions was determined by Cook (1968) and Wiedemeier
and Khan (1968). Finely powdered MnS and CdS samples were thoroughly
mixed in desired molar ratios and compressed into pellets at a pressure of
approximately 13.5 K-bar. The pellets were sealed under high vacuum in
quartz tubes, and annealed for approximately 100 hours at each of the
following temperatures: 1000", 800, 700, and 600C. The tubes were then
taken out of the furnace, and the samples quenched in air. This procedure
proved to be sufficient to maintain the equilibrium established at the high
temperature, resulting in stable sintered polycrystalline specimens. To
ensure that equilibrium was established, selected samples were subjected to
additional repeated annealing cycles at 800, 700, and 600"C, for a total of
several hundred hours at each temperature. Lattice constants of samples
subjected to repeated annealing cycles were identical, within the limits of
error, with values for alloys obtained by the original annealing sequence. The
temperature-composition phase diagram of the CdS-MnS system is presented
in Fig. 8. As can be seen in the figure, the liquidus and solidus are practically
indistinguishible on the CdS-rich side of the diagram.
In the case at hand, the melting points of the constituent binaries are very
high: roughly 1400C for CdS and about 1600C for MnS. Because of this,
silica can no longer be used in the crystal growth of Cdl-,Mn,S from the
melt. Single crystal growth of Cdl-,Mn,S by the Bridgman method was
reported by several authors using carbon crucibles, under high inert gas
pressure (Ikeda et al., 1968; Komura and Kando, 1975). Compressed pellets
were used as starting charges in these investigations. The pellets were first
repeatedly annealed at 1000C in nitrogen/HZS atmosphere. The carbon
crucible containing the annealed powder was then placed in a furnace filled
with dry Ar gas. The pressure of Ar was between 40 and 100 kg/cm2. After
holding the temperature slightly above the melting point of the alloy for
approximately one hour, the crucible was lowered out of the furnace at a rate
of 30 mm/h. The fusing temperature of the mixed powder was set at 1450C
for the Cd-rich samples (0 5 x I 0.30), and at 1620C for the Mn-rich
composition (0.7 Ix 5 1). This procedure demonstrated that Cdl-,Mn,S
forms in the wurtzite structure for Cd-rich compositions, and that the rocksalt phase of this material may form at very high values of x (see Fig. 1).
1.
Liquid
II
,*
..
,.'
.,,*-'__,'- -
21
;-"I600
- 1500
Wurtzite
Wurtzite a n d
700
22
W. GIRIAT A N D J. K . FURDYNA
As can be seen from Table 111, MnSe is soluble in ZnSe for x I0.57. In
the composition range 0 < x < 0.30 this solid solution crystallizesin the zinc
blende structure, and for 0.35 5 x I 0.57 in the wurtzite form. For x > 0.57
mixed phases occur, including various selenides of Mn. Considerable
twinning is observed in the zinc blende phase as x approaches the transition
value of 0.30.
As in most other DMS, pre-reacted binary ZnSe and MnSe are used as
starting materials. Taken in desired proportions, they are ground to a
powder, mixed, pressed into pellets, and reacted for several days in the
temperature range of 1OOO-11OO"C. The resulting sintered compounds are
then used as starting material for crystallization. The phase diagram of
ZnSe-MnSe is not known, but it is clear that the melting point is slightly over
1500C throughout the entire range 0 < x < 0.57.
Chemical transport is the simplest method to grow single crystals of
Znl-,Mn,Se. For this purpose, a tube of inner diameter of about 15 mm and
12to 15 cm long can be used, in which the sintered material is placed, together
with iodine (10 mg/cm2) as the transport agent. The tube is then sealed in
vacuum and put in a furnace, where the temperature is at 1000C and 980"C,
respectively, at the two ends of the tube. Single crystals can be harvested after
one week, size of the crystals being several millimeters in each direction.
1.
23
4
6
7 1
C
C
C
C
C
8-
3
2
FIG.9. Self-sealing graphite crucible used in the growth of Zn, -xMnxSe:(1) graphite crucible;
(2) threaded plug; (3) thread; (4) outside silica tube; (5) induction heating coils; (6) powdered
material; (7) molten zone: (8) crystallized material; (9) exit to dynamic vacuum. [After Debska
et a/. (1984).]
24
W . GIRIAT A N D J. K . FURDYNA
is inverted so that the plug is at the bottom. The crucible is placed at the center
of an induction coil, with the bottom half of the plug extending beyond the
coil. After the system is evacuated to about
Torr, the temperature is
slowly raised beyond the melting point of the charge, the graphite acting as
the susceptor in the heating process. Some of the material then flows down,
and ultimately solidifies in the space between the crucible wall and the plug,
thereby sealing the crucible. The center of the crucible is RF-heated to about
100C above the melting point of the material, and the whole assembly is then
lowered out of the stationary induction coil at the rate of about 10 mm/h.
The whole process is conducted under dynamic vacuum of 10-4-10-6 Torr.
The induction heater in the specific system described by Debska et al. utilized
a 20 KW generator operated at about 300 kHz, and produced a hot zone with
a sharp thermal gradient. The narrowness of the heater zone is important in
view of the fact that graphite has a high thermal conductivity, which can act
to smear out the temperature gradient. Since the alloy itself has a lower
vapor pressure than the constituent elements, it is advisable to pre-synthesize
the material at the desired composition prior to the actual crystal growth. A
number of good quality Znl-,Mn,Se single crystals have been produced over
the entire solid solution range, 0 5 x 5 0.57. Ingots are typically 9 mm in
diameter by about 70 to 90 mm, composed of crystals of about 5 to 10 mm
in length, with occasional monocrystals as long as 5 to 8 cm.
(f) Znl-,Mn,S
Znl-xMn,S alloys form single phase zinc blende crystals for x up to
= 0.1, and wurtzite crystals for 0.1 < x 5 0.60, as indicated in Table Ill. The
phase diagram for this alloy is shown in Fig. 10 (Sombuthawee et al., 1978).
-- -.:--z:--..----
2000@50
2 1600u-
LIQUID
--e--
1610
d
,
A
/
FIG. 10. Phase diagram for ZnS-MnS. [After Sombuthawee et al. (1978).]
1.
25
ZnS has a melting point of 1722"C, the highest of the A"BV' compounds. The
melting point for MnS (rocksalt phase) is 1530"C, also the highest among the
Mn-chalcogenides .
Because of the high melting points involved, chemical transport is
commonly preferred as the crystal growth technique of Znl-,Mn,S. To
grow single crystals of this alloy, sintered material (prepared from prereacted binaries ZnS and MnS, in a manner similar to that described for
Znl-,Mn,Se) is placed in quartz tubes with inner diameter of about 15 mm,
12 to 15 cm long. Iodine(l0 mg/cm3)is used as the transport agent. The tubes
are sealed in vacuum and located in the furnace such that Tm,, = 1000C and
Tmin= 980"C, for a period of one week. For samples with a small Mn content
(x < 0.15), monocrystals of the size of 0.5 cm3 have been obtained in this
manner. For higher values of x the crystals tend to be smaller (several
millimeters in each direction). The composition of the crystals is typically
quite close to that of the starting sintered material.
(g) HgI-,Mn,Te
The preparation of this system over a wide range of compositions has been
first investigated in the classical work of Delves and Lewis (1963). As with
the other tellurides, the Hg, -,Mn,Te system shows that a surprisingly high
fraction of Hg atoms (about 80%) can be replaced by Mn, with the zinc
blende structure retained.
The phase diagram obtained by Delves and Lewis for the HgTe-MnTe
system is shown in Fig. 11. From this figure it is clear that the liquidus-solidus
separation is considerable, and standard crystallization is therefore not
expected to yield highly homogeneous crystals.
In spite of the high segregation coefficient, the majority of Hgl-,Mn,Te
monocrystals prepared to date have been grown by the vertical Bridgman
method, because of its simplicity. Bridgman crystals are typically prepared
as follows. Pre-reacted HgTe and MnTe (or HgTe and elemental Mn and Te
in proper amounts) are ground and placed into a thick-walled quartz
ampoule with an inner diameter of 10 to 15 mm (Kaniewski et al., 1982). For
crystals containing more than 5% Mn, the inner wall of the ampoule is coated
with carbon. The ampoules, evacuated to approximately
Torr and
sealed, are placed in the furnace at ca. 450C. After 24 hours, the temperature is increased to 820C and is held at this value over a period of 48 hours
to achieve complete solution of manganese. The Bridgman growth is then
started. The recommended growth speed is slow, typically about 1 mm/h.
Higher growth rates result in a high degree of polycrystallinity.
As can be expected from the phase diagram, Hgl-,Mn,Te crystals grown
by the Bridgman method have a significant composition gradient along the
growth direction due to a large segregation coefficient. The segregation
26
600
0.0
HgTe
735%
Mole Fraction X
Mn e
FIG.1 1 . Phase diagram for HgTe-MnTe. [After Delves and Lewis (1963).]
1.
950-
I-
Heating
Mn,Hg,-,Se
21
Mn,Hg,-,Se+MnSe
I
I
HgSe
Oll
0
012 013 6!4 d.5 016 017 d.8 019
MnSe
Mole Fraction X
FIG.12. Phase diagram for HgSe-MnSe. [After Pajaczkowska and Rabenau (1977a).]
28
W. GIRIAT A N D J. K . FURDYNA
much less thoroughly than the bulk A:'-,MnxBv' alloys, progress is rapidly
being made, particularly in the layered A:!.,Mn,BV' (e.g., MBE films and
superlattices), in bulk A:t,FexBV' alloys, and in the A:!!,MnxBV' systems. In
concluding this chapter, we therefore wish to mention these new DMS
systems briefly, so as to identify the evolving materials trends and to provide
a brief literature guide for these emerging DMS activities.
7. DMS FILMSAND SUPERLATTICES
One of the most important-and most rapidly evolving-recent developments in the area of new DMS systems is the successful preparation of DMS
films, heterostructures, and superlattices. The first reported DMS film
structures involved the quaternary epitaxial layers of Hgl -x-yCdxMnyTe
prepared by the close-spaced isothermal vapor transport growth technique
(Debska et af., 1981). The field has taken a major step forward when it was
demonstrated that Cdl -,Mn,Te films and Cdl -,Mn,Te/Cdl-,Mn,Te
superlattices could be readily grown by molecular beam epitaxy (MBE) on
a variety of substrates (Kolodziejski et a/., 1984b; Bicknell et a/., 1984;
Kolodziejski et a/., 1984a). Since 1984, the field has made truly remarkable
progress, including successful preparation by MBE of films and superlattices
of Znl-,Mn,Se (Kolodziejski etal., 1984) and of Hgl-,Mn,Te (Harris et al.,
1987; Faurie, 1987). In parallel with these developments, atomic layer epitaxy
(ALE)-an equilibrium growth method, in contrast to MBE, which involves
nonequilibrium growth-has also been shown as a viable technique for the
growth of DMS films (Herman et a/., 1984: Tammenmaa et a/., 1985). It is
clear that the ability to grow high quality DMS films, superlattices and
heterostructures, coupled with the novel properties of DMS materials, opens
a rich spectrum of possibilities in basic and applied science. A survey of these
opportunities is beyond the scope of this chapter. It should be mentioned,
however, that apart from exciting electronic, magnetic, and optical applications (Datta et a/., 1984), epitaxy also opens the way to the growth of entirely
new systems that do not form in the bulk. The recent preparation by MBE
of zinc blende Znl-,Mn,Se for x > 0.30 and, indeed, of zinc blende binary
MnSe layers in thin film form (Kolodziejski et al., 1986)are examples of such
new systems. This aspect alone-i.e., the MBE growth of metastable alloys
and compounds-is extremely important in its potential to advance the
understanding of the crystal growth of magnetic semiconductor alloys, and
of semiconductor compounds in general.
8. A:'_,Fe,Bv'
ALLOYS
1.
29
30
which in turn leads to (at least partial) space ordering. Such an ordered,
periodic array of ionized impurities can then be shown to lead to reduced
scattering rates as compared to a random distribution of ionized impurities
(Pool et al., 1986).
.
The preparation of A!'_,Fe,Bv' materials is generally similar to their sister
Mn alloys (Mizera et al., 1980). The range of x in which Fe can be incorporated into the A"BV' lattice is, however, considerably less than in the case
of Mn alloys. For example, the highest values of x reported so far for various
recently investigated alloys are -0.06 for Hgl-,Fe,Te (Guldner et al., 1980),
0.12 for Hgl-,Fe,Se (Vaziri et al., 1985), and 0.15 for Cdl-,Fe,Se (Lewicki
et al., 1986). The pattern which thus appears to em&+ is that, generally, the
higher the melting point of the A"BV' host, the higher is the amount of Fe
which can enter substitutionally at the group-I1 sites. The important problem
of establishing the limits of miscibility of Fe in specific A1'BV' lattices, and
the determination of the corresponding phase diagrams, still awaits a
systematic investigation.
9. MANGANESEAND RARE-EARTH-BASED
A1"Bv' ALLOYS
Activity has also been recently increasing in the area of A:?,M,Bv' alloys,
where M indicates a substitutional magnetic element (either Mn or a rareearth). Lead salts with a fraction of the Pb-sublattice replaced by Mn are the
most thoroughly studied members of this group (Pbl-,Mn,S: Karczewski
et al., 1982, 1985; Pbl-,Mn,Se: Kowalczyk and Szczerbakow, 1984;
Pbl -,Mn,Te: Niewodniczanska-Zawadzka and Szczerbakow, 1980; Pascher
et al., 1983; Anderson and Gorska, 1984). Crystal growth methods employed
in preparing these alloys were, for the most part, Bridgman growth for bulk
crystals and hot wall epitaxy for single crystal films, with growth parameters
similar to those which proved successful in the preparation of the parent lead
salt binaries. The miscibility of Mn in these systems appears to be smaller
than in the A:'_,Mn,BV' alloys, the highest values being those reported for
Pbl-,Mn,Te (x = 0.2). No comprehensive study of this aspect has as yet
been described for this class of alloys.
In addition to the development of the A!!,Mn,BV'
alloys, there is
increasing activity in lead salt alloys involving rare earths, for example,
Pbl -,Eu,Te and Pbl -,Gd,Te. The methods of successful preparation
include Bridgman growth (Golacki and Gorska, 1985), hot wall epitaxy of
thin film specimens (Krost etal., 1985), and MBE (Partin, 1983,1984). While
most of the non-MBE growth has been restricted to x < 0.10, the MBE films
showed very high rare-earth miscibility in the lead salt lattice (e.g., x = 0.40
for Pbl-,Yb,Te (Partin, 1983)). Much of the interest in these materials is
stimulated by their photovoltaic properties and their potential for diode laser
31
There are two additional cases of DMS alloys that are potentially highly
interesting. (Cdl -,MnX)3As2 resembles, in the absence of an external
magnetic field, its parent Cd3As2: it is understood to be a zero-gap semiconductor, and can exhibit relatively high electron mobilities. The introduction of Mn brings into play spin-spin exchange interaction between the
band electrons and localized magnetic moments (Neve et al., 1981), and in
this respect the system should resemble qualitatively the behavior of doped
Hgl -,MnxSe and Hgl-,Mn,Te. Comparatively little research has been done
on this and related systems (e.g., Zn3Asz-based DMS), but the feasibility of
such alloys has been demonstrated and the initial research (Neve et al., 1981)
indicates a highly interesting line of investigation.
Finally, we wish to point out the very promising DMS opportunities
offered by the chalcopyrite structure. Chalcopyrites are tetrahedrally bonded
semiconductor compounds which, in their structure and many physical
properties, resemble A"BV1 or A"'BV compounds, as discussed in the
excellent monograph by Shay and Wernick (1975). CuInSez and CdGeAsz
are examples of such systems. The chalcopyrite crystal structure is shown in
Fig. 13, side by side with the zinc blende structure, showing the close
relationship of the two crystallographic forms.
Considering the predisposition of Mn to act as a group-I1 atom, and
the close relationship between the chalcopyrite structure and the simple
A"BV' tetrahedrally bonded semiconductors, one is immediately tempted to
think of incorporating Mn (and other magnetic atoms) into the chalcopyrite
lattice. In particular, structures of the form A:'_,Mn,BIVCx (where Mn
occupies substitutionally the group-I1 sites of the original A"B"C; lattice);
or A:-,B:'f,Mn2,Cx1
(where Mn can occupy either the group-I or the groupI l l sites of the parent cation lattice) immediately suggest themselves.
It should also be borne in mind that Fe is readily incorporated into the
chalcopyrite structure. Indeed, the original term "chalcopyrite" refers,
specifically, to the compound CuFeS2 , the mineral prototype from which
the structure class gets its name. Thus Fe, which can be incorporated to
only a limited degree into the A:!-,Fe,Bv' alloys, could play a more important
role in the chalcopyrite DMS that may be prepared in the future. There are
thus vast new possibilities in this family of materials which-like the
"established" A:!-,Mn,BV' alloys-is based on the tetrahedral bond. Initial
32
W . GIRIAT A N D J . K . FURDYNA
Zinc Blende
C hatcopyrite
FIG. 13. Comparison of the chalcopyrite structure (one unit cell, shown on right) with zinc
blende structure (two unit cells, left).
References
Anderson, J. R., and Gorska, M. (1984). Solid State Commun. 51, 115.
Balzarotti, A., Czyzyk, M., Kisiel, A . , Motta, N. Podgorny, M., and Zimnal-Starnawska, M.
(1984). Phys. Rev. B30, 2295.
Bicknell, R. N., Yanka, R. W., Giles-Taylor, N. C., Blanks, D. K., Buckland, E. L . , and
Schetzina, J . F. (1984). Appl. Phys. Lett. 45, 92.
Brandt, N. B., and Moshchalkov, V. V. (1984). Advances in Physics 33, 193.
Cook, W.R. (1968). J. American Ceramic Society 51, 5 1 8 .
Datta, S., Furdyna, J. K., and Gunshor, R. L. (1985). SuperlatticesandMicrostructures 1,327.
Debska, U., Dietl, M., Grabecki, G., Janik, E., Kierzek-Pecold, E., and Klimkiewicz, M.
(1981). Phys. Staf. Solidi (a) 64, 707.
Debska, U., Giriat, W., Harrison, H . R., and Yoder-Short, D. R. (1984). J . Cryst. Growth 7 0 ,
399.
1.
33
Delves, R. T., and Lewis, B. (1963). J. Phys. Chem. Solids 24, 549.
Dietl, T., and Spalek, J. (1983). Phys. Rev. B28, 1548.
Dolling, T.,Holden, T. M., Sears, V. F., Furdyna, J. K., and Giriat, W. (1982). J. Appl. Phys.
53, 7644.
Dziuba, 2. E. (1969). J. Electrochemical Society 116, 104.
Faurie, J. P. (1987). Proc. NATO Advanced Research Workshop on Optical Properties of
Narrow Gap Low Dimensional Structures, St. Andrews (U.K.), July 29-August 1, 1986,
edited by C. M. Sotomayor Torres et al. (Plenum, New York), p. 25.
Furdyna, J. K. (1982). J. Appl. Phys. 53, 7637.
Furdyna, J. K., Giriat, W., Mitchell, D. F., and Sproule, G. (1983). J. Solid State Chem. 46,349.
Gaj, J. A. (1980). J. Phys. SOC.Japan 49, Suppl. A, 797.
Gaj, J. A., Galazka, R. R., and Nawrocki, M. (1978). Solid State Commun. 25, 193.
Galazka, R. R. (1979). Znst. Phys. Conf. Ser. 43, 133.
Galazka, R. R. (1982). Proc. Znt. Conf. on Narrow Gap Semiconductors, Linz, Sept. I981,
Lecture Notes in Physics 152 (Springer-Verlag. Berlin), p. 294, and references therein.
Galazka, R. R., and Kossut, J. (1980). Lecture Notes in Physics 132 (Springer-Verlag, Berlin),
p. 245.
Golacki, Z., and Gorska, M. (1985). Acta Phys. Polon. A67, 379.
Greene, L. H., Orenstein, J., Wernick, J. H., Hull, G. W., and Berry, E. (1986). Bull. Am.
Phys. SOC.31, 383, Abstract EN 15.
Guldner, Y., Rigaux, C., Menant, M., Mullin, D. P., and Furdyna, J. K. (1980). Solid State
Commun. 33, 133.
Harris, K. A., Hwang, S., Burns, R. P., Cook, J. W., Jr., and Schetzina, J. F. (1987). Diluted
Magnetic (Semimagnetic) Semiconductors, edited by R. L. Aggarwal, J. K. Furdyna, and
S. von Molnar (Vol. 89, Materials Research Society Symposia Proceedings, Pittsburgh,
PA), p. 255.
Herman, A., Jylha, 0. J., and M. Pessa, M. (1984). J. Crystal Growth 66, 480.
Ikeda, M., Itoh, H., and Sato, H. (1968). J. Phys. SOC. Japan 25, 455.
Kaniewski, J., Witkowska, B., and Giriat, W. (1982). J. Crystal Growth 60, 179.
Karczewski, G., Klimkiewicz, M., Glas, I. Szczerbakow, A., Behrendt, R. (1982). Appl. Phys.
A29, 49.
Karczewski, G., von Ortenberg, M., Wilamowski, Z., Dobrowolski, W., andNiewodniczanskaZawadzka, J. (1985). Solid-State Commun. 55, 249.
Kolodziejski, L. A., Sakamoto, T., Gunshor, R. L., and Datta, S. (1984). Appl. Phys. Lett.
33, 799.
Kolodziejski, L. A., Bonsett, T. C., Gunshor, R. L., Datta, S., Bylsma, R. B., Becker, W. M.,
and Otsuka, N. (1984). Appl. Phys. Lett. 45, 440.
Kolodziejski, L. A., Gunshor, R . L., Bonsett, T. C., Venkatasubramanian, R., Datta, S.,
Bylsma, R. B., Becker, W. M., and Otsuka, N. (1985). Appl. Phys. Lett. 47, 169.
Kolodziejski, L. A., Gunshor, R. L., Otsuka, N., Gu, B. P., Hefetz, Y., and Nurmikko,
A. V. (1986). Appl. Phys. Lett. 48, 1482.
Komura, H., and Kando, Y. (1975). J. Appl. Phys. 46, 5294.
Kowalczyk, L., and Szczerbakow, A. (1985). Acta Phys. Polon. A67, 189.
Krost, A., Harbeck, B., Faymonville, R., Schlegel, H., Fantner, E. J., Ambrosch, K. E., and
Bauer, G. (1985). J. Phys. C. 18, 2119.
Larson, B. E., Hass, K. C., Ehrenreich, H., and Carlsson, A. E. (1985). Solid State Commun.
56, 347.
Lewicki, A., Mycielski, A., and Spalek, J. (1986). Acta Phys. Polon. A69, 1043.
Mizera, E., Klimkiewicz, M., Pajaczkowska, A., and Godwod, K. (1980). Phys. Stat. Solidi(a)
58. 361.
34
CHAPTER 2
I . INTRODUCTION . . . . . . . . . . . . . . . . . . . .
1 . General Discussion. . . . . . . . . . . . . . . . .
11. VALENCEAND CONDUCTION
BANDSIN WIDE-GAPDMS ALLOYS.
2. Band Structure at Zone Center. . . . . . . . . . . .
(a) Basic Features . . . . . . . . . . . . . . . . .
(b) Variation of EBwith Mn Concentration. . . . . . .
(c) Variation of E8 with Temperature and Pressure. . . .
3 . Band Structure A way From Zone Center . . . . . . . .
111. OPTICAL
PROPERTIES
OF d-ELECTRONS. . . . . . . . . . .
4. General Considerations . . . . . . . . . . . . . . .
5 . Mn Transitions in Specific DMS Alloys. . . . . . . . .
IV. CONCLUDING
REMARKS. . . . . . . . . . . . . . . . .
6 . Summary and Discussion . . . . . . . . . . . . . .
REFERENCES.. . . . . . . . . . . . . . . . . . . .
35
35
36
36
36
38
50
56
51
51
59
68
68
I0
I. Introduction
1. GENERAL
DISCUSSION
35
Copyright B 1988 by Academic Press, Inc.
All rights of reproduction in any form reserved.
ISBN 0-12-752125-9
36
W . M . BECKER
a. Basic Features
For wide-gap DMS, such as Cdl-xMn,Te, Cdl-,Mn,Se, etc., crystallographic studies indicate that single phase regions-either zinc blende or
wurtzite-exist over wide ranges of x [see Giriat and Furdyna, this volume].
All results to date suggest that in these regions of solid solution, the band
2.
31
gap is always direct. Therefore, the zone center band structure of these
materials should be characteristic either of zinc blende or wurtzite crystals,
and experimental data, i.e., optical absorption, excitonic emission, etc., have
been interpreted accordingly.
To understand these data, we first give a brief review of the two types of
band structure to be encountered at k = 0. These can be conveniently
introduced in terms of the bonding of the A"BV1 host binaries. (For an
extended treatment, see Phillips, 1973.) In the A"BV' compounds, the two
valence electrons of the group I1 element and the six electrons of the group
VI elements are distributed according to the so-called s-p3 orbital bonding
configuration. In this picture, the highest-lying valence band is triply
degenerate (in the absence of spin-orbit splitting), and the states at the center
of the Brillouin zone are bonding combinations of functions which are plike about the nuclei. The lowest conduction band is s-like. Figure 1 gives the
results of a band structure calculation using the pseudopotential method; the
structure, here obtained for ZnTe (Cohen and Bergstresser, 1966), may be
taken as representative of the cubic (i.e., zinc blende) Cd and Zn
chalcogenides.
The introduction of the spin-orbit interaction alters the bands, principally
by splitting certain degeneracies. At k = 0 in the zinc blende structure, the
six-fold degenerate (three-fold not counting the two spin states) T1sv state
splits into a four-fold r8(J = %)state and a lower, two-fold r7(J = 3)state,
with the splitting indicated in Fig. 2 by A,, , the spin-orbit splitting energy.
In the "quasi-cubic" model of the valence and conduction band in wurtzite
a4
- 0
W
-I
-2
-3
FIG. 1 . Energy band structure of cubic ZnTe as calculated by the empirical pseudopotential
method (Cohen and Bergstresser, 1966).
38
W. M. BECKER
2 INCBLENDE
ZINCBLENDE
Aso = Acr = 0
Ac,= 0
(a)
(b)
0.
crystals, the valence band states are considered to be derived from the Tlszinc
blende valence band state through the action of two perturbations: crystal
field splitting and spin-orbit splitting. The quasi-cubic model may be viewed
as a reasonable zeroth-order approximation, since the atoms in the zinc
blende and wurtzite structures are in nearly similar positions in the unit cell.
The model predicts the splitting of the TSstate by the hexagonal crystal field
into a rg and r7 state, each state being two-fold degenerate. Figure 3 gives
the details of the resulting band structure around k = 0 and also shows the
selection rules for optical transitions near the band edge.
b. Variation of Eg with Mn Concentration
2.
39
WURTZ ITE
40
W. M. BECKER
3.0
FIG.4. Variation of the energy gap in Cd,Zn,Mn,Te (x + y + z = 1) with manganese concentration z for fixed x : y ratios. Experimental values: 0 x = 0,. x = 3y, 0 3x = y , M y = 0,
A x = y . Line fits ___ are discussed by the authors (Brun Del Re ef al., 1983).
compound, may often not be identified in the DMS. In the wurtzite systems
the transition mechanisms may also be obscure, but the dichroism associated
with the hexagonal symmetry of these crystals does provide a useful caliper
of valence band splittings in the alloys.
Although qualitative similarities in behavior are seen in all DMS alloys,
it is convenient to present the results alloy-by-alloy, a strategy which will be
carried forward through the remainder of this chapter. Table I and Table I1
summarize some representative data for the Cdl-,Mn,BV' and Znl-xMnxBV'
alloys (B = Te, Se, or S), respectively. Comments are given below on these
and related results in the literature. It should be noted that the column in
Tables I and I1 designated as Eg represents the fundamental gap based on
strong experimental features, but the specific electronic transition associated
with the features may be obscure for x # 0. In the wurtzite region, the
(r9, - r7c)and (r7"- r7c)
transition energies (see Fig. 3) are designated in
the tables by the superscripts I and 11, respectively, where they have been
specifically reported in the literature.
(a) Cdl-,Mn,Te
We begin with Cdl-,Mn,Te, the most extensively studied and probably
best understood of the wide-gap DMS. Absorption edge measurements by
TABLE I
OF BANDGAPENERGY
VERSUS MOLEFRACTION
x IN Cdl-xMnxBv' ALLOYS
SUMMARY
Eg in eV
Alloy
1.585 +
1.595 +
1.50 +
1.586 +
1.53 +
1.528 +
1.586 +
1.595 +
Cdl-,Mn,l'x
Cdi -,Mn,Se
1.50~
1.592~
+ 1.08~
+ 1.23~
Ell 1.725 + 1 . 2 3 ~
E l 1.800 + 1.33x
Ell 1.829 + 1 . 3 3 ~
1.807 +'1.318x
1.821 + 1 . 5 4 ~
EL 1.74 + 1 . 1 6 ~
Ell 1.756 + 1 . 1 3 ~
E L 1.807 + 1 . 3 4 ~
Ell 1.833 + 1 . 3 1 ~
E L 1.818 + 1 . 4 2 ~
1.70
E L 1.705
Ell
Cdl -,Mn,S
1.51~
1.587~
1.44~
1.393~
1.26~
1.316~
1.842
+ 1.40~
Range of x
77
LHe
300
76
300
300
80
10
R.
E.M.
E.R.
P.
E.L.
P.R.
300
297
297
86
86
LN2
2
300
300
80
80
10
10
< x c 0.5
0 < x < 0.3
0 < x < 0.3
P.
R. & O.A.
P.R.
O.A.
2.45
298
Minimum at
x 0.03
dE,/dX
0.9 x 2 0.1
Measurement
Methods'
T(K)
x=o
Reference
TABLE I1
SUMMARY
OF BANDGAP ENERGYVERSUS MOLEFRACTION
x in Znl-xMnxBV' ALLOY.
Eg in eV
Alloy
Znl -xMnxTe
2.377
2.381
+ 3.656~
+ 0.68~
2.271
2.365
2.376
Zn,-,Mn,Se
+ 0.518~
+ 0.721~
+ 0.820~
2.80
Minimum at x
2.80
- 0.01
Znl -,Mn,S
3.8 - 3.9
TW)
1.6
2.2
300
80
10
Range o f x
0.01 -c x
< 0.15
x > 0.02
0 5 x I 0.6
0 5 x < 0.6
0 5 x < 0.6
Measurement
Methods'
Reference
M.
E.M.
Twardowski (1982)
Twardowski e? al. (1984)
P.R.
2.2
x=o
M.
6.5
x=o
R. & P.
x > 0.3
x=o
M.R.
2.
43
Nguyen The Khoi and Gaj (1977) provided one of the first indications that
the energy gap increases with x in this material. These authors also showed
that the pinning of the edge occurring near 2 eV at high values of x was
directly attributable to the onset of Mn2+ intra-ion transitions. This same
pinning effect was also seen by Sundersheshu (1980) using photoconductivity
were published by Gaj et
techniques. The first quantitative results on Eg(x)
al. (1978) using reflectivity data. Linear increases in Egus x were recorded
by Twardowski et al. (1979) using magnetoabsorption measurements, but in
this case, the data was confined to low x values, 0 < x < 0.2. Modulated
electroreflectance results by Bottka et af. (1981) provided a survey of Eg(x,
T)
over nearly the complete single phase region of Cdl-,Mn,Te, and confirmed
the linear increase of Eg with x seen by previous investigators. Photoluminescence results by Vecchi et af. (1981) demonstrated that crystal quality
was sufficient to give band edge emissions for all x. Finally, Lee and Ramdas
(1984), using piezo-modulated reflectivity, were able to separate band edge
features from spectra due to the presence of Mn2+ intra-ion transitions (to
be discussed in Part 111). Figure 5 gives their results for various Mn concentrations and temperatures. In the figure, the peak labeled A is identified as
the free exciton. No indications of band gap bowing are seen in the entire
range of solid solution. In particular, there is an absence of the band gap
bowing noted in several other DMS at low x (see below).
A summary of measured variations of Egwith x is given in Table I. Based
on the many experimental results available on Cdl-,Mn,Te, reasonable
estimates (to within k0.02 eV) can be given for Egat x = 1 (Le., for zinc
blende MnTe), and these are listed in Table 111 for various temperatures.
TABLE I11
ESTIMATES
OF ENERGY
GAP OF HYPOTHETICALX = 1 MnB ZINC BLENDE (2.b.) OR
WURTZITE(w.) SEMICONDUCTORS
E* (x = 1)
Alloy
Cdl-,Mn,Te
Temperature
in eV
2.86
3.05
3.18
(2.b.)
Cd1-,MnrSe ( w . )
2.92
3.15
3.24
Znl-,Mn,Te (2.b.)
Znl-,MnxSe (2.b.)
II
2.87
3.13
3.30
2.79
3.09
3.19
-3.5 (2.b)
R.T.
LNz
LHe
R.T.
LN2
LHe
R.T.
LN2
LHe
44
W . M. BECKER
G-4
28
C.-.-
t-
1.4 i
0.0
Cd ,+MnxTe
300K
80 K
I
0.2
0.4
0.6
(b) Cdl-,Mn,Se
2.
45
2.6-
C d , - x MnxSe
10 K
2.4-
z
a
a
u)
w
U
c'a
2.2-
LL
LL
>
(3
5w
2.0-
46
W. M. BECKER
~~~~~~
r7v- rgv
Alloy
AS0
in eV
in eV
2 6 lo-
~
20 x 10-3
0.436
Cd, -,Mn,Se
(29 t 5 ) x 10-3
(27 k 3) x lo-
14 x
x=O
O<x<0.3
17 x
10 x 10-3
x=o
x =0
x = 0.42
X = 0.20
0.30
0.46
- 2 O x lo-
10 x
0.35
0.50
0.440
0.312
Znl-xMnxSe
Range of x
~~~~
Measurement
Method
O.A.
O.A.
E.R.
E.R.
O.A.
O.A.
Reference
Cardona ei al. (1967)
Antoszewski and
Kierzek-Pecold
(1980)
Cardona et at. (1967)
Stankiewicz (1983)
Stankiewicz (1983)
Stankiewicz (1983)
Morales et at. (1985)
Morales et at. (1985)
Morales (1985)
Morales (1985)
(c) Cdl-,Mn,S
Cdl-,Mn,S is the least investigated of the Cd-based family of DMS. In this
case, pronounced bowing was seen by Ikeda et al. (1968) for low values of x
(see Fig. 7), as well as a monotonic increase in Eg vs x for x > 0.1. According
to these authors, alloying effects are ruled out as the mechanism for bowing
near x = 0. Recent work by Bylsma et al. (1986) on Znl-,Mn,Se at low values
of x (see Fig. 8), and by Diouri et al. (1985) on Cdl-,Mn,Te suggest that the
bowing in Cdl-xMn,S may be related to s-d and p-d exchange interactions
with the Mn magnetic moments. Crystal field and spin-orbit splitting energies
are known for CdS, but still remain to be investigated in Cdl-,Mn,S ternary
alloys.
2.
41
2.9
T = 114K
-_
2
-
rn
2.3
K and 298 K
FIG.
(Ikeda et al., 1968).
T-6.5K
- PL energy
3,10-
->
-Other PL
3.001
z
W
.*
,
* *
2.90
2.60/
2.50
0.00 0.10
0.20
0.30
040
0.50
0.60
CONCENTRATION, X
FIG. 8. Band gap energy versus Mn mole fraction x in Znl-xMnxSe, taken from photoluminescence and reflectivity maxima at 6.5 K (Bylsma et al., 1986).
48
W . M. BECKER
2.8 Zn I
Mnx Te
300K
BOK
0.2
"
0.4
0.6
FIG.9. Variation of the energy gap and Mn transition with Mn concentration in Znl-,Mn,Te
for 10 K, 80 K, and 300 K (Lee et a/., 1986a). In the figure, the peak labeled A is identified as
the free exciton. The concentration-independent feature, labeled Mn2+,is associated with the
leading edge of the Mn 6A~(6S) 4T~(4G)absorption band.
-t
2.
49
x for x 2 0.3 (see Fig. 8). Apparent scatter in the data for 0.2 Ix I0.3 was
traced to the presence of various polytypes found in the vicinity of the zinc
blende-wurtzite structural phase transition (x = 0.3). Recently, metastable
zinc blende Znl-,Mn,Se epilayers with (100) orientation, free of polytypes,
have been grown on GaAs substrates over the complete alloy range by
molecular beam epitaxy (Kolodziejski et al., 1986). In these samples (see Fig.
lo), photoluminescence results give both a shallow minimum in Eg near
x = 0, and a smooth transition to a linear dependence on x at higher Mn
concentrations. Extrapolation of Eg to x = 1 gives the same value for both
the bulk crystals and the epilayers.
2.701 ,
0
M n PracLion
FIG. 10. Energy of dominant near-band edge features versus Mn mole fraction for MBEgrown zinc blende Znl-,Mn,Se epilayers (Kolodziejski e t a / . , 1986a). (Cross-hatched points are
data obtained on bulk crystals by Twardowski el a/., 1983b).
50
W. M. BECKER
(f) Znl-,Mn,S
Crystal growth studies show that Znl-xMnxS is zinc blende for
0 5 x 5 0.1 and wurtzite for 0.1 Ix 5 0.43 (Pajaczkowska, 1978). ZnS
itself exhibits both the zinc blende and wurtzite modifications, as well as
rotation twinning and polytypism (Steinberger, 1983), so that sorting out
small band structure changes associated with these crystallographic
anomalies as a function of x may be quite complicated. As yet, such changes
remain unexplored.
c. Variation of Eg with Temperature and Pressure
The study of Egin DMS as a function of temperature and pressure has been
stimulated by (1) the need to distinguish between band-to-band transitions
and Mn2+ intra-ion (d d) or Mn2+-to-band transitions, and (2) the
possibility of observing anomalous thermal shifts which may be related to
the onset of magnetic phase transitions (Sundersheshu and Kendelewicz,
1982). For (I), the problem arises because of the strength of the Mn-related
absorption features near the energy of the fundamental gap. Pronounced
differences in the temperature coefficients of the respective transitions then
help to identify them separately. With regard to (2), in the usual semiconductor, the temperature variations of the band gaps are well described by
an equation given by Varshni (1967),
+
aT2
Eg(T) = Eg(0) - _ _
T + 6
where Eg(T)and Eg(0)are the band gap energies at T and 0 K, respectively,
and a and b are constants of the material. This equation predicts (1) that
dEg/dT 0 as T 0 K, and (2) that JdE,/dTI approaches a constant value
at high temperatures. Anomalous behavior would then consist of departures
from this functional dependence on T near the magnetic ordering temperature. Listed in Table V is a summary of results obtained on several DMS
alloys. In each case, Egdecreases approximately linearly with Tabove liquid
nitrogen temperature, and results are presented in Table V in terms of an
average value of the coefficient dE,/dT. In a number of recent observations,
however, departures from the Varshni-like behavior (namely, an extra shift
to higher energies) below the characteristic spin freezing temperatures have
been reported in DMS. The same effect had been seen earlier in magnetic
semiconductors such as MnS (Chou and Fan, 1974) and MnTe (Kendelewicz,
1980a). Temperature anomalies of this type are also cited in Table V, where
seen in various DMS alloys.
+
2.
51
TABLE V
AVERAGED
TEMPERATURE
COEFFICIENTS
dEg/dTv s x ~ TEMPERATURE
o~
RANGE77
- dEg/dT
Alloy
in
Cdl-rMn,Te
3.7
3.7
Cd,-,Mn,Se
eV/K
8.9
6.6
Range of x
0
0
Measurement
Methods
M.E.
O.A.
0.6
0.4
4.6 - 7.5
6.8
9.1
3.5
8.5
3.3 - 10
0 - 0.4
0.4*
0.73*
0.1
0.7*
0 - 0.7
4.5
< 0.1
O.A.
O.A.
O.A.
P.R.
P.R.
O.A. & R.
Reference
Bottka et al. (1981)
Sundersheshu and
Kendelewicz (1982)
Abreu et al. (1981)
Diouri et a/. (1985)
Diouri et al. (1985)
Lee and Ramdas (1984)
Lee and Ramdas (1984)
Bucker et al. (1985)
3.9
3.9 - 7.3
7.3
5.3 - 8.6
6.8 - 7.9
0.28
x > 0.28
0 - 0.35
0.3 - 0.45
M.E.
M.E.
M.E.
O.A.
O.A.
Antoszewski and
Kierzek-Pecold (1980)
Stankiewicz (1983)
Stankiewicz (1983)
Stankiewicz (1983)
Abreu et al. (1983)
Morales et al. (1985)
Cd,-,Mn,S
4.4 - 9.4
0 - 0.4
O.A.
Ikeda e t a / . (1968)
Znl-xMn,Te
4.14 - 4.61
4.3 - 9.8
6
- 9.4
0 - 0.071
0 - 0.6
0.1 - 0.6*
O.A.
P.R.
O.A.
Znl-,MnxSe
4.7
8.5
9.7
0.066
0.36*
0.55*
R. &. P.
R. & P.
5.5
0.1
R.
0.3
R. &. P
52
W. M. BECKER
E,(P) = Eg(0)+ yP
+ 6Pz,
(2)
where Eg(0)is the band gap at P = 0. Values of the coefficients y and 6 are
given in Table VI for a number of DMS.
We now examine the temperature and pressure dependence of Eg for
specific DMS alloys.
(g) Cdl -,Mn,Te
The temperature dependence of the band gap of Cdl-,Mn,Te was first
investigated systematically by Bottka et al. (1981). These researchers found
that for T > 100 K, (dEg/dTIis constant for x < 0.15,but increases linearly
with x thereafter. The authors suggested that the sharp change in slope at
x = 0.15 is related to lattice mode softening at the percolation threshold of
the Mn spin system. Sundersheshu and Kendelewicz (1982) carefully
examined Eg vs T for anomalous optical behavior expected to accompany a
magnetically induced shift in the band gap. While such effects were not
detected, they found instead a sharp drop in the values of b [see Eq. (I)] above
x = 0.3, and took it as indicative of a magnetically induced widening of the
gap. Later, Bucker et al. (1985)used reflectivity to survey dEg/dTfor all x.
In contrast to the electroreflectance results of Bottka et al. (1981),they found
only a uniform increase in the magnitude of the temperature coefficient with
x. Lee and Ramdas (1984)noted that (dEg/dT(for 10 4 T I80 K is greater
than IdEg/dT)for 80 IT 5 300 K for an x = 0.7 sample, a first indication
of the sought-for anomalous effect presumably associated with the magnetic
phase transition in this material. Stronger evidence for this behavior was
obtained by Diouri et al. (1985)from optical absorption measurements (see
Fig. 11). Working with very thin layers ground and polished from the bulk,
these investigators found shifts of 45 meV and 17 meV to higher energies in
x = 0.7 and x = 0.4 samples, respectively, as compared to Eg(0) values
predicted from higher temperature results using Eq. (1).
Turning now to pressure effects, Ambrazevicius et al. (1984)were able to
observe the pressure variation of Eg by the use of samples thin enough to
allow measurements at the level of absorption coefficients greater than
lo3cm-'. With these precautions, positive values of the pressure coefficient
dEg/dPwere recorded for all x . At lower levels of the absorption coefficient
(smaller than lo2cm-I), the sign of the coefficient becomes negative between
x = 0.4 a n d x = 0.5.This latter result is interpreted in terms of intra-ion d-d
transitions of Mn2+. Similar sign changes were observed by Wei Shan et al.
(1985)when comparing behaviors of x = 0.1 and x = 0.5 samples at low
levels of absorption.
TABLE VI
OF EBAT ROOMTEMPERATURE
FROM ABSORPTION
EDGEMEASUREMENTS
PRESSURE DEPENDENCE
Alloy
Cdt-,Mn,Te
dE,/dP
eV/MPa
8
1.9
1.3
5.9
4.6
Ea(0)'
in eV
y' in
eV/MPa
1.483
1.618
2.161
2.115
8.3
7.1
- 4.9*
-5.1'
Znl-,Mn,Te
2.21
2.265
2.28
2.37
10.4
10
10.6
8.2
Znl-,Mn,Se
2.688
2.648
2.615
2.756
7.12
6.1
6.30
6.34
dt in
eV/GPaZ
- 4.0
- 3.9
- 28
- 26
- 54
- 80
- 15
- 11
- 18
- 27
Reference
0.4
0.52
0.6
0.7
0
0.1
0.5
0.615
Ves et al.
Ves et al.
Ves et al.
Ves ef at.
0.01
0.1
0.3
0.1
0.15
0.25
(1985)
(1986b)
(1986b)
(1986b)
54
W.
M. BECKER
FIG. 11. Absorption edge as a function of temperature for (a) Cdo.Z,Mno.73Te, and
(b) Cdo.sMno.4Te (Diouri et af., 1985). The dotted curves show the extrapolation of Eg obtained
from high temperatures according to Eq. (1).
2.
55
Twardowski et al. (1984) found little variation in -dE,/dT for x < 0.1.
The piezomodulated reflectivity results of Lee et al. (1986a), obtained from
data taken at 80 K and 300 K, show that IdEg/dTI increases by a factor of two
between x = 0 and x 5 0.5. Similar to the observations on Cd,-,Mn,Te and
Cdl-,MnxSe, Morales Tor0 et ul. (1984) showed that in Zn1-,MnxTe
1-1
for 10 < T < 77 K is greater than 1-1
for 77 < T < 300K
at high x, with reverse behavior seen at low x .
Ves et al. (1986a) carried out optical absorption measurements on
Znl-,Mn,Te under hydrostatic pressure in a diamond anvil cell. As expected,
Eg increases with increasing pressure, but strong nonlinearities were
observed. Results were limited to low-x material (x 5 0.3) because of the
deterioration of the absorption feature associated with Eg at high x and P i n
Znl-,Mn,Te, and also because in this material the pressure at which the
phase transition occurs decreases with increasing x.
56
W. M. BECKER
(k) Znl-,Mn,Se
Bylsma et al. (1986) found a monotonic increase in - dEg/dTwith x for
Znl-,Mn,Se samples in the range 0 < x < 0.55. A deviation from Varshnilike behavior, namely, an extra shift of E,(T) to higher energies at low
temperatures and large Mn concentrations, was also seen to occur in this
material.
Ves et al. (1986b) carried out optical absorption measurements at
high pressure on samples with x values lying in the zinc blende region
(0 < x < 0.3). As in Znl-,Mn,Te, Eg was found to increase with pressure.
Pronounced nonlinearities were observed, but values of 6, the quadratic
pressure coefficient [see Eq. (2)], were smaller than in Znl-,Mn,Te.
(1) Znl-,Mn,S
At the time of this writing, no published information on the temperature
and pressure dependence of Eg in Znl-,Mn,S is available.
3. BANDSTRUCTURE
AWAYFROM
ZONECENTER
Reflectivity measurements for energies above the fundamental edge have
proved indispensable in probing the electronic structures of semiconductors
deep into the bands or away from k = 0. In DMS materials, the technique
has been applied to several alloys over the available solid solution regions.
In this work, there was the expectation of being able to detect sharp spectral
features at high x. This was based on earlier findings that these features
persist over large alloy ranges in other ABV mixed crystals, such as
Hgl -,CdxTe and Znl -,Cd,Te.
First indications of reflectivity features in DMS associated with transitions
away from k = 0 were seen by Kendelewicz and Kierzek-Pecold (1978) in
their work on the narrow-gap alloy Hgl-,Mn,Te. In contrast to the behavior
in non-DMS ABV ternary alloys, however, a blurring of the reflectivity
spectral structures occurred with growing Mn concentration. This effect was
also noted in studies on the wide-gap DMS Cdl-,Mn,Te (Kendelewicz,
1980b; Kendelewicz, 1981; Zimnal-Starnawska et al., 1984; Bucker et al.,
1984) and on Znl-,Mn,S and Znl-,Mn,Se (Zimnal-Starnawska et al., 1984).
Except for the persistence of the fundamental edge reflectivity feature, other
details, such as the El and El + A1 transitions, disappear for x z 0.3. The
rate of deterioration of the reflectivity spectra with increasing Mn content
is highest for Znl-,Mn,S and lowest for Cdl-,Mn,Te, according to ZimnalStarnawaka et al. (1984). These authors explain the blurring of the spectra
in terms of localized Mn2+3ds spin down states lying at higher energies
(spin splitting of = 6 eV; see also Podgorny and Oleszkiewicz, 1983).
2.
57
4. GENERAL
CONSIDERATIONS
The introduction of Mn not only changes the band structure of the host
A"BV' semiconductor, but it also introduces new optical transitions. These
include prominent absorption and emission features, which can be correlated
with the expected splittings of Mn multiplets in a crystalline field. Below is
given a brief summary of the origin of the multiplets and their splittings,
leading to the hierarchy of states usually observed in wide-gap A:t,Mn,BV'
alloys. The reader is referred to standard references (e.g., Tanabe and
Sugano, 1970; Abragam and Bleaney, 1970; Griffith, 1971; Ballhausen,
1961; McClure, 1959) for more complete details.
Significant understanding of this aspect of DMS can be gained by considering an isolated Mn ion in an A"BV' lattice. A freeMn atom has the 3d54s2
ground state configuration for the outermost shells. When the atom is put
into a DMS, it substitutes for the cation. The two 4s electrons form the bonds
with the surrounding ions, and the five 3d electrons interact strongly with the
electrical potential of the nearest neighbor anions. The total Hamiltonian is
H = Ho + V + H,,, where Ho is the many-electron Hamiltonian of a free
ion without spin-orbit interaction, Vis the contribution of the crystal field,
and H,, is the spin-orbit interaction. Since the magnitude of V satisfies the
inequalities H,, < V < H,, V can be treated as a perturbation.
We now review very briefly the spectroscopic notation used in describing
the ground and excited states of both the free ion, and the ion when in a ligand
field of an appropriate symmetry. For the free ion, the state is given by ""L,
where S is the total spin, and L is the total orbital angular momentum. In
the ground state of a free Mn2+ ion, the electron spins are aligned due to
Hund's rule, giving S = f; this state is designated as 6S ( L = 0). When the
Mn2+ion is put into a crystal field that is not strong enough to break Hund's
rule, the 'S state remains as the ground state. Since it is spherically symmetric, it transforms as A1 (in group theoretical notation), and is therefore
unsplit by the crystal field.
In the free ion state, spin reversal of a single 3d electron gives rise to four
quartet states corresponding to S = 5; 4G ( L = 4), 4P ( L = l), 4D ( L = 2),
and 4F ( L = 3). The 4Gstate has nine-fold orbital degeneracy (2L + 1 = 9).
This degeneracy, when lifted by a crystal field of cubic symmetry, leads to
four states labeled (in group theoretical notation) 4T1,4T2, 4A1, and 4E. The
degeneracies of the states are as follows: 4Tl-three fold; 4Tz-three fold;
4A~-one fold; 4 E - t ~ o fold. Using the same method, it is found that the 4D
state splits into 4E and 4T2, and the 4P state does not split, but transforms
as 4T1. State 4Fsplits into 4T1, 4T2,and 4A2; 4A2 is non-degenerate, and the
other states have the degeneracies indicated earlier.
58
W. M. BECKER
Using these eigenstates as basis, the total Hamiltonian matrix for the Mn2+
ion in an A"BV1host (without spin-orbit interaction) can be written and
, 4A1, and 4Estates.
diagonalized, yielding all the energy levels of the 4 T ~4T2,
The energy associated with each state is usually given as a function of the
crystal field parameter, Dq, and the Racah parameters B and C (Tanabe and
Sugano, 1954). The parameter Dq is related to the orbital splitting of a single
d electron in a cubic crystal potential, and B and Care combinations of radial
integrals over the interelectronic distance. Tanabe and Sugano (1954) carried
out this calculation for a cubic field, and all the energy matrices were
collected by McClure (1959). The solution of the secular equation of these
matrices gives the appropriate energies of the states. Shown in Fig. 12 is the
energy level diagram appropriate for splitting of states of the d5 configuration in a cubic field (Tanabe and Sugano, 1970). For the usual values of
Dq/B, the 4Gstates are the lowest excited states, and are therefore the first
to become optically accessible in DMS alloys as the band gap widens.
Transitions between the states are governed by strict selection rules. For
example, in the free ion case, transitions between the 6Sground state and the
4Gexcited states are forbidden both by the AS = 0 selection rule and by the
E/B
2::
'A2 'TI
4 T ~
6A,(dc3 d'yz)
4T1( d r 4 d y )
10 -
'Tp(dr5)
6A I
6S
6s
4
Dq/B
FIG. 12. Splitting of states of the dS configuration by a crystal field of oh symmetry (Tanabe
and Sugano, 1970). Similar splitting is expected for a crystal field of Td symmetry.
2.
59
(m) Cdl-,Mn,Te
The first optical effects seen in Cdl-,Mn,Te related to Mn transitions were
those of Nguyen The Khoi and Gaj (1977), who associated an absorption
edge at =2.0eV with the onset of the 6A1(6S)+ 4T1(4G)transition. From
photoluminescence results on thick samples (=1 mm), Tao et af. (1982) and
Moriwaki et af. (1982) suggested that both the 6A~(6S) 4T~(4G)and
6A~(6S) 4T2(4G)excitation features were present at high x . However, it was
not until samples thinned to the micron range were employed that the
-+
TABLE VII
MnZ+ABSORPTION/EXCITATION
BANDMAXIMA(IN eV)
IN
DMS ALLOYS
_______~
Alloy
Cdl-xMn,Te
6Ai
4T1
- 2.2
- 2.2
Cdl-,Mn,S
Znl -,MnxTe
Znl-,Mn,Se
Znl-,Mn,S
-2.5
- 2.43*
- 2.4*
6Ai
4Al, 4E
Other
- 2.63*
-2.63*
- 0.48
- 0.4
-2.3 (Threshold)
2.43
2.58
- 2.3
-2.3
-2.33
2.34
2.38
2.4
2.42
- 2.36
-2.38
2.40
B(cm-l)
720
713
713
~(cm-')
750
749
75 1
2.22(z.p.)
2.35
2.48
2.57
2.5
2.54
2.44(z.p)
2.53
0.4 5 x I0.7
0.4 Ix 5 0.7
0.7
0.7
0.7
0.4 5 x 5 0.7
- 2.6*
- 2.43*
-2.2
Cdl-,Mn,Se
6A1 + 4T2
2.72
-2.576
- 2.605
-2.745
C(cm-')
Dq(cm-l)
2664
743(Td), 330(0h)
1638
750(Td), 333(0h)
2637
913(G), 406(0h)
2.655
2.68
2.70
2.9
2.70
2.93
C(cm-')
Dq(cm-')
2730
836(%), 372(0h)
2772
830(G), 369(0h)
2782
795(G), 353(0h)
2.635 (z.P.)
2.69
T a o et al. (1982)
Moriwaki et al. (1982)
Rebmann et a/. (1983)
Lascaray et at. (1983)
Diouri et al. (1985)
Lee and Ramdas (1984)
Morales et al. (1985)
Ikeda et al. (1968)
-0.1
-0.5
0.6
-0.1
0.5
-0.6
0.001
0.23
0.35
0.5
0.23
0.35
0.50
Morales (1985)
Morales (1985)
Morales (1985)
Reference
2.
61
absorption band features (see Fig. 13) due to Mn were clearly revealed
(Rebmann ef af., 1983, and Lascaray el a[., 1983). Some uncertainty in
separately identifying the bands is encountered at high x because of
broadening and overlap effects, as indicated in Table VII.
As mentioned earlier, the Mn intra-ion transitions in DMS alloys can be
conveniently discussed in terms of the Tanabe-Sugano diagram for the d5
configuration. Based on this diagram, the energy of the 6A1(6S) 4T1(4G)
transition for Mn2+decreases with the increase of the crystal field parameter
Dq. In a cubic field, Dq is proportional to R-' in the approximation of a
point charge ligand, where R is the distance between the transition metal ion
and the ligand ion. Therefore, the application of pressure should give a
decrease in the energy of the 6A1(6S) -+ 4T1(4G) transition, and in the
associated 4T1(2S)-+6A1(4G)emission (that is shifted to longer wavelengths
due to lattice relaxation). Investigation of the pressure shifts of (1) the
2.2 eV Mn absorption edge (Muller et af., 1983, Ambrazevicius et al., 1984)
and (2) the 2.0 eV photoluminescence band (Miiller er al., 1983)gave negative
pressure coefficients, consistent with the above picture of the d-d transitions.
Also, the magnitudes of the coefficients, 3-8 x
eV/GPa, were
explained using known compressibility results for the host A1'Bv' semiconductors and reasonable estimates of the pressure dependence of Dq and
the Racah parameters B and C (Miiller et af., 1983). Piezomodulated
reflectivity studies (Lee and Ramdas, 1984) also yield phase shift data (see
Fig. 14) near the Mn absorption edge, that are consistent with the hydrostatic
pressure results.
-+
62
W. M. BECKER
CMT 10
-I
6000 5000 -
4000
3000 -
2000
l8ooo
2.
63
PHOTON ENERGY ( eV 1
FIG.14. Piezomodulation spectrum of Cd0.35Mn0.35Teat T = 80 K (Lee and Ramdas, 1984).
The opposite signs of the derivative signals of the A and Mn2+ features are consistent with the
direction of energy shifts observed using hydrostatic pressure.
were able to detect three Mn2+absorption bands below the intrinsic edge (see
Table VII), that, for this composition, corresponds to Eg = 2.85 eV.
(P) Znl-,Mn,Te
Zn-based DMS alloys are special in that all of them have gaps Eg > 2.0 eV,
allowing the study of Mn transitions in even the low-xlimit. In Znl-,Mn,Te,
results were first obtained by Skowronski et al. (1978), who observed
absorption bands associated with Mn on a sample with x = 0.1. Morales
Tor0 et al. (1984) extended the alloy range of investigation up to x = 0.6 (see
Fig. 16) and found a total of four Mn absorption bands in samples with the
highest Mn concentrations at liquid helium temperatures. Of these, the three
lowest bands are ascribed to the 4G transitions, while the origin of the
highest-lying band remains in doubt. The values of Dq (listed for both Td and
Oh symmetry) and the Racah parameters given in Table VII show changes
with x that are consistent with the known composition-related parameter
changes.
Ves et al. (1986a) carried out room temperature hydrostatic pressure
measurements on samples with x I0.6. Four absorption features, identified as Mn intra-ion transitions, were shown to have negative pressure
64
W. M. BECKER
Cd,
X
g
a
-,Mn, Se
RT
LN,
rl
LHe
0.45
10'
t
Ioo
1.90 2.00
2.10
2.20
2.30
2.40
2.50
2.60
PHOTON ENERGY ( eV 1
FIG.15. Optical absorption coefficient versus photon energy in Cdl-,Mn,Se for an x = 0.45
sample at RT, LNz, and LHe temperatures (Morales et a / . , 1985). The different polarizations
of the incident light are shown in the figure. Band gap dichroism is seen over the complete range
of a at room temperature, but only appears for a's greater than 2 5 x 102cm-' at low
temperatures.
2.
2.2
2.3
2.4
2.5
2.6
65
2.7
2.8 2.9
hv(eV)
FIG. 16. Optical absorption coefficient versus photon energy in Znl-,MnxTe for an x = 0.6
sample at RT, LN2, and LHe temperatures (Morales Toro et a/., 1984). (See Table VII for
assignment of transitions.)
(4)Znl -,MnXSe
Four Mn absorption bands and a shoulder have been seen (Morales Toro,
1985) at liquid helium temperature in Znl-,Mn,Se for x = 0.5 (Figs. 17 and
18). The absorption band at ~ 2 . 9 eV
3 and the shoulder at ~ 3 . 1 eV
0 involve
multiplets of, as yet, unknown origin. Dichroism associated with band-toband transitions is clearly indicated at all temperatures for absorption
coefficient values greater than lo3cm-', as seen in Fig. 18. The dichroic
effects in the Mn absorption region and, in several cases, distinct energy
shifts with polarization of the Mn bands are unexpected for Td symmetry.
An explanation for this result appears to be outside the limits of ligand
theory.
Ves et al. (1986b) carried out room temperature absorption measurements
as a function of hydrostatic pressure on Znl-,MnxSe for 0 Ix 5 0.25.
66
W. M. BECKER
I
Znl-xMnxSe
3
P
03
X - 0.5
LHe
*PI;
0
11;
ZI
8
lo-''
I.&
I.QO
2.k
2.10
2.10
2.k
PHOTON ENERGY ( e V
2.AO
3.00 32!0
FIG. 17. Optical absorption coefficient versus photon energy in Znl-,MnxSe for an x = 0.5
sample at LHe temperature (Morales Toro, 1985). (See Table VII for assignment of transitions.)
The different polarizations of the incident light are shown in the figure.
-+
-+
-+
2.
67
lo4 -
RT
LN2
LHe
'02
2.k 2.k
3hO
3.k
3.20
FIG.18. Optical absorption coefficient versus photon energy in Znl-,Mn,Se for an x = 0.5
sample at RT, LN2, and LHe temperatures (Morales Toro, 1985). The different polarizations
of the incident light are shown in the figure. Band gap dichroism is seen for absorption
coefficients greater than lo3cm-', but dichroism is also present in the region of Mn absorption.
(r) Znl-,Mn,S
Early work on ZnS doped with Mn (McClure, 1963) showed that five
absorption bands with their associated zero-phonon lines could be seen. The
later work of Neumann (1971) on Znl-,MnxS (reported by Gumlich, 1981)
for 0 c x I 0.2 revealed the characteristic broadening-and ultimate disappearance-of spectroscopic structure, as discussed in the beginning of this
section. As yet, the highest two bands observed in this alloy remain
unidentified.
68
W . M. BECKER
L--
22-
26
2.8
ENERGY (eV)
24
3.0
32
FIG.19. Absorption coefficient versus photon energy at various hydrostatic pressures for an
x = 0.15 sample of Znl-,Mn,Se (Ves ef al., 1986b). The curves for p 2 4.02GPa have an
arbitrary shift of the ordinate scale to allow for a clear display of the results.
2.
A-l
2.5
27
2.9
2.3
O-O2.l
69
3.0
ENERGY (eV1
FIG.20. Absorption coefficient versus photon energy at various hydrostatic pressures for an
> 0.001 GPa have an
arbitrary shift of the ordinate scale.
70
W. M. BECKER
Mn content. Much less certain is the origin of the transitions seen at higher
energy. This problem, that arose early in the study of Mn at very low values
of x (McClure, 1963), remains an active area of theoretical research
(Richardson and Janssen, 1986). More experimental work is certainly
needed, but present techniques, such as optical absorption at high pressure,
seem unlikely to impinge on this problem.
Strong evidence exists for the hybridizing of the ground state of the Mn
ion with valence band states at energies well below the band edge. Such
hybridization does not appear to affect the existence and observation of the
lowest energy localized d-d transitions of the ion. Further experimental and
theoretical work is needed to understand the coexistence of these two effects.
Finally, charge transfer states, such as Mn'+3d6 levels, have been predicted
recently for DMS materials (Ehrenreich et al., 1986), but experimental
evidence for such states is slight. Recent results by Lee et al. (1986b) using
reflectivity at high magnetic fields rule out this mechanism as being responsible for the 2 eV absorption/excitation feature in CdI-,Mn,Te. The
possibility exists that short-lived charge transfer states are located in the gap;
these will probably require investigation by time-resolved spectroscopy to
verify their presence.
References
Abragam, A., and Bleaney, B. (1970). Electron Paramagnetic Resonance of Transition Ions
(Clarendon Press, Oxford).
Abreu, R. A., Stankiewicz, J., and Giriat, W. (1983). Phys. Stat. Sol. (a) 75, K153.
Ambrazevicius, G., Babonas, G., Marcinkevicius, S., Prochukhan, V. D., and Rud, Yu V.
(1984). Solid State Commun. 49, 65 1.
Antoszewski, J., and Kierzek-Pecold, J. (1980). Solid Sate Commun. 34, 733.
Babonas, G. A., Bendoryus, R. A., and Shileika, A. Yu.(1971). Sov. Phys.-Semicond. 5,392.
Ballhausen, C. J. (1962). Introduction to Ligand Field Theory (McGraw-Hill Book Company,
Inc., NY.)
Bottka, N., Stankiewicz, J., and Giriat, W. (1981). J. Appl. Phys. 52, 4189.
Brafman, O.,and Steinberger, I. T. (1966). Phys. Rev. 143, 501.
Brun Del Re, R., Donofrio, T., Avon, J., Majid, J., and Woolley, J. C. (1983). I1 Nuovo
Cimento 2D, 1411.
Bucker, R., Gumlich, H. E., and Krause, M. (1985). J. Phys. C: Solid State Phys. 18, 661.
Bylsma, R. B., Becker, W. M., Kossut, J., Debska, U., and Yoder-Short, D.(1986). Phys. Rev.
B 33, 8207.
Cardona, M., Shaklee, K. L., and Pollak, F. H. (1967). Phys. Rev. 154, 696.
Chou, H . H., and Fan, H. Y. (1974). Phys. Rev. B 10, 901.
Cohen, M. L., and Bergstresser, T. K. (1966). Phys. Rev. 141, 789.
Diouri, J., Lascaray, J. P., and El Amrani, M. (1985). Phys. Rev. B. 31, 7995.
Ehrenreich, H . , Hass, K. C., Johnson, H. F., Larson, B. E., and Lempert, R. J. (1986).
Proceedings of the 18th International Conference on the Physics of Semiconductors,
Stockholm, Sweden.
Gai, J. A., Galazka, R. R., and Nawrocki, M. (1978). Solid State Commun. 25, 193.
Giriat, W.,and Stankiewicz, J . (1980). Phys. Stat. Sol. (a) 59, K79.
2.
71
12
W . M. BECKER
CHAPTER 3
and
Pieter H. Keesom
DEPARTMENT OF PHYSICS, PURDUE UNIVERSITY
WEST LAFAYETTE, INDIANA, USA
INTRODUCTION. . . . . . . . . . .
BACKGROUND. . . . . . . . . . .
RESULTSAND ANALYSIS.
. . . . . .
1. Electron Paramagnetic Resonance.
2. Magnetic Susceptibility . . . . .
3 . Specsfic Heat . . . . . . . . .
4. Remanent Magnetization . . . .
IV. CONCLUDING
REMARKS.. . . . . .
REFERENCES.. . . . . . . . . . .
I.
11.
111.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
. . . . .
. . . . .
. . . . . .
. . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . . .
. . . . . . . . .
13
14
80
80
91
107
112
117
119
I. Introduction
The purpose of this work is to review the magnetic properties of diluted
magnetic semiconductors (DMS). These are ternary compounds of the form
AI-,Mn,C, where A = Zn, Cd, Hg are group-I1 elements of the periodic
table, C = S , Se, Te are group-IV elements and xis the mole fraction of the
magnetic ion MnZ+.The properties of interest will be electron spin resonance,
magnetization, and the contribution to the specific heat due to the magnetic
Mn ions incorporated substitutionally in the lattice (neutron scattering data,
that are also of importance as far as magnetic properties of DMS are
concerned, are discussed elsewhere in this volume). These properties can be
calculated when the energy level scheme of the Mn ions is known. Vice versa,
when the above properties are known from the measurements, they can be
used to deduce the energy level scheme. The divalent manganese ions have
a 3d5 configuration and 6S5/2 ground state. A spherically symmetric S state
cannot be affected by a crystalline electric field. However, it is known that
a crystal field splitting of the ground state of Mn2+ does occur. This is
because it is not a pure 6Sstate, but contains admixtures of higher lying levels.
73
Copyright 0 1988 by Academic Press, Inc.
All rights ot reproduction in any form reserved.
ISBN 0- 1 2 - 7 9 125-9
74
The mechanisms of such mixing involve higher order processes that take into
account crystal fields, spin-orbit coupling, spin-spin interaction, relativistic
effects, etc. (Watanabe, 1957; Gabriel et al., 1961). However, the admixtures
in question are relatively small and often can be neglected. We shall return
to this problem later in this chapter.
11. Background
In this section, we shall try to give a list of general problems that are
encountered when studying magnetic properties of DMS with increasing
number of magnetic ions incorporated in the host semiconductor matrix.
Thus, we start with an examination of the properties of an isolated Mn2+ ion
in the 11-VI lattice. We then proceed to describe the interation of the ion with
other Mn2+ ions. The latter interaction will be the dominant effect determining the magnetic properties of concentrated systems. We shall also give
expressions appropriate to describe various quantities of interest to us in this
chapter for each regime of Mn2+ concentration.
For concentrations of Mn below 0.1 atomic percent, x < 0.001, the interaction between the Mn ions can probably be neglected. Thus, in any analysis,
only isolated paramagnetic Mn ions with a S m ground state need to be
considered. The presence of a crystal electric field will remove part of the sixfold degeneracy of the ground state. A crystal field cannot remove all the
degeneracy of a 6 S ~ / 2state; at most it will split it into three Kramers doublets.
Because of the complexity of the mechanisms involved in the splitting, it is
difficult to perform direct calculations. Therefore, spin Hamiltonian method
is usually used, Abragam and Bleaney (1970). The crystal field Hamiltonian
for Mn2+ can be written as an even polynomial of fourth degree in the
projection of the spin operators &,
and $. For example, for an
orthorhombic lattice, the crystal field Hamiltonian for Mn2+ is:
sy,
3.
75
where the first term accounts for the electronic Zeeman interaction, the
second for the cubic crystal field, and the third for the hyperfine interaction.
In Eq. (2), we have ignored smaller contributions, such as the superhyperfine
interaction and the nuclear Zeeman interaction. For a more complete
discussion on the subject, see Abragam and Bleaney (1970).
When the energies of the Mn levels are known, the magnetization M ,
susceptibility x and specific heat C follow directly. For low concentrations,
the uncoupled Mn ion levels in zero magnetic field will only show up at
temperatures such that the value of k T is smaller or comparable than their
crystal field and or hyperfine splitting, less than 0.1 K. If a sufficiently strong
magnetic field (compared with the crystal field splitting) is applied,
H 2 lOkOe, then the Mn ions give a Schottky contribution to M , x, and C
characteristic of a system with essentially six evenly spaced levels. The
susceptibility for the temperatures above 1 K and low concentration follows
a Curie law:
76
TABLE I
PROBABILITY
THAT
FOREIGN
ATOMIS I N A SINGLE,
DOUBLE,
AND TRIPLE
CLUSTER
ASSUMING
ONLY THE NEAREST-NEIGHBOR
INTERACTIONS
fcc
Cluster Type
hCP
(1 - x)l2
12x(l - X ) l 8
18?(1 - ~)~[5(1- x)
24?(1 - x)
S
D
Tb
z1
(1 12X(1 - X ) 8
i8X2(1 - Xy3[2 + 5(1 - X)I
3x2(1 - ~ ) ~ +
[ 16(1 - X) + (1 - x ) ~ ]
+ 21
TABLE I1
PROBABILITY
THAT A FOREIGN
ATOMIS IN A SINGLE,
DOUBLE,
OR TRIPLE
CLUSTER
ASSUMING
THE
NEAREST-NEIGHBOR
AND NEXT-NEAREST-NEIGHBOR
INTERACTIONS
fcc
Cluster Type
(1 - x)8
12x(1 - X)26
6x(1 - x ) ~
902(1 - x ) ~
92(1 - x)41[4
72u2(1 - x)37[1
24x2(1 - x ) ~
+ (1 - x)]
+ (1 - x ) ~
36x2(1 - x)34
0
(1 - x)l8
12x(l - x)26
6x(1 - x ) ~
9x2(1 - x)[l
9(1 - x)]
9 2 ( 1 - x)39[2(1 - x)2 + 2(1 - X ) 3 + 11
18x2(1 - ~ ) ~ [+3 5(1 - x)]
3x2(1 - xI2[l + 6(1 - x) + (1 - x ) ~ ]
0
36x2(1 - x ) ~ ~
Tables, I, 11, and 111. The corresponding Hamiltonians and eigenvalues for
singles, pairs, and triples are given in Table 1V.
The susceptibility and specific heat C for a system consisting of clusters
of various sizes can be written as:
where xi and Ci are the contributions from each cluster type including
isolated ions, and Ni is the number (per unit volume) of occurrence of clusters
of the i-th type. The susceptibility and the specific heat contributions of each
3.
77
TABLE I11
PROBABILITY
THAT A FOREIGN
ATOMIS IN A SINGLE,
DOUBLE,
OR TRIPLE
CLUSTER ASSUMING
INTERACTIONS UP TO 3RD-NEAREST-NEtGHBORS FOR
fcc LATTICE'
Cluster Type
Probability
(1 - x)42
12x(l - x)5'
6x(1 - x)62
24x(1 182(1 - x)74
9x2(1 - x ) ~ +
* 36x2(1 - x)"
36x2(1 - x)'~+ 144x2(1- x ) ~ '
+ 252x2(1 - x)93+ 180x2(1- x)94
72x2(1 - x ) ~ '
144x2(1 - x)" + 144x2(1- x)" + 216x2(1- x)'4
72x2(1 - x ) ' ~+ 144x2(1- x)*'
24x2(1 - x)~'
24x2(1 - x)'~
36x2(1 - x)~'
72x2(1 - x)"
72x2(1 - x)"
72x2(1 - x)71
72x2(1 - x)'3
Dl
L?
D3
z'
T,'
z3
s%B
3k~T
w + 1)(2S + l)exp(-Es/ksT)
cs(2s + l)exp(-&/kBT)
d ES exp(- E s / k T
~)
Cj(T) = a T exp( - &/kB T ) '
(6)
(7)
where Es is the energy for the case of zero magnetic field. In the presence
of a magnetic field H, Es in Eqs. (6) and (7) should be replaced by
E, = Es - g,uBmH where m is the magnetic quantum number. In that case,
an additional summation over rn (- S 5 m IS) must be performed.
As the concentration increases the mean cluster becomes more extended
until, at a certain critical concentration x = x, , its size becomes comparable
with the size of the sample. The value of xc , known as the critical concentration for the site-percolation problem, depends on the lattice structure and
the number of neighboring magnetic ions for which the exchange interactions
are significant. For example, if interactions up to the lst, 2nd, or 3rd
neighbors are taken into account in an the fcc lattice, we have xc = 0.195,
78
TABLE IV
HAMILTONIANS
AND EIGENVALUES
FOR SINGLE,
DOUBLE,
AND TRIPLE
CLUSTERS'?
Type
Hamiltonian
Eigenvalue
0.136, and 0.061 respectively (see e.g., Domb and Dalton, 1966; Frisch et al.,
1961).
3.
79
TABLE V
CRYSTAL
STRUCTURE
AND RANGEOF COMPOSITION
FOR
THE DMS STUDIED
Material
Znl-,Mn,S
Znl-,Mn,Te
Cdl-,Mn,S
Cdl-,Mn,Se
Cdl -,MnxTe
Hgl-,MnxS
Hgl-,Mn,Se
Hgl-,MnxTe
Crystal
Structure
Range of
Composition
Zinc blende
Wurtzite
Zinc blende
Wurtzite
Zinc blende
Wurtzite
Wurtzite
Zinc blende
Zinc blende
Zinc blende
Zinc blende
x < 0.10
0.10 s x s 0.43
x s 0.30
0.30 < x < 0.55
x s 0.75
x s 0.45
x s 0.50
x s 0.70
x ~0.37
x s 0.30
x 5 0.50
80
This section is divided into several parts, each devoted to a single physically
measureable phenomenon or quantity. We start with electron paramagnetic
resonance, to be followed by magnetic susceptibility, specific heat, and
magnetization. In each case, we first review the available experimental data
in the given field, and then give a detailed theoretical analysis.
PARAMAGNETIC
RESONANCE
1. ELECTRON
a. Experimental Results
The EPR spectra of Mn2+in various II-VI compounds for x < 0.001 have
been studied in detail by several authors (Matarrese and Kikuchi, 1956;
Dorain, 1958; Matamura, 1959; Lambe and Kikuchi, 1960; Hall er al., 1961;
Kikuchi and Azarbayejani, 1962; Schneider et al., 1963; Title, 1963a,b, 1964;
Estle and Holton, 1966; Falkowski, 1967; Deigen et al., 1967; Leibler et al.,
1970a,b, 1971, 1973). The main features of the spectra are: the gyromagnetic
3.
81
factor g is nearly equal to that of the free electron, 1.99 < g c 2.02; a well
resolved isotropic hyperfine structure is observed consisting of six lines,
possibly due to configurational interaction between the 3s23p63d54sconfiguration with the 3s23p63d5configuration (Abragam, 1950); and a fine
structure that results from an anisotropic contribution. The values of the
gyromagnetic factor g, the hyperfine structure constant A , and the crystal
field parameters are obtained by fitting the data to the energy spectrum
corresponding to a spin Hamiltonian analogous to that given by Eq. (2).
As the concentration of Mn increases, the spectrum changes its shape. For
x > 0.005, the hyperfine lines initially broaden, eventually becoming a
structureless single line for x 1 0.02 (Ishikawa, 1966; Deigen et al., 1967;
Leibler et al., 1977; Oseroff et al., 1979, 1980a,b; Andrianov et al., 1980;
Oseroff, 1982; Manoogian et al., 1982; Webb et al., 1983; Sayad and Bhagat,
1985). The EPR linewidth AH,, first narrows as x increases and passes
through a minimum at x = 0.03. Van Wieringen (1955), after calculating the
second and fourth moments of the spectrum, suggested that the variation of
the shape of the spectrum may result from the narrowing effect of the
hyperfine splitting due to the exchange interaction with other Mn2+ions. The
effect of exchange interaction on the linewidth has also been discussed by
Bleaney (1970).
After reaching the minimum at x = 0.03, the line broadens monotonically
with concentration (Leibler et al., 1976, 1977; Oseroff, 1982; Manoogian et
ai.,1982; Webb et al., 1983; Sayad and Bhagat, 1984). Different mechanisms
which contribute to the broadening of the line with concentration have been
discussed by Kittel and Abrahams (1953); Swarup (1959); Deigen et al.
(1967); Abragam and Bleaney (1970); and Sayad and Bhagat (1985). Values
of the gyromagnetic factor g and of the linewidth AH,, at room temperature,
defined as the peak-to-peak in the first derivative of the absorption line, are
given in Fig. 1 for Cdl-,MnxTe as a function of concentration and frequency.
Similar behavior was found for other DMS studied. At high temperatures,
the linewidth and the gyromagnetic factor g were found to be independent
of microwave frequency used (Oseroff, 1982; Webb et al., 1983; Sayad and
Bhagat, 1985).
For concentration above x = 0.03, a significant increase in the linewidth
with decreasing temperature was observed for all the DMS studied so far by
EPR (Leibler et al., 1976, 1977; Grochulski et al., 1979; Oseroff et al., 1979,
1980a,b; Mullin et al., 1981; Oseroff, 1982; Manoogian et al., 1982; Webb
et af., 1983; Sayad and Bhagat, 1985; Kremer and Furdyna, 1985a,b). This
can be understood as an increase (as the temperature is lowered) of the
internal field due to the presence of finite clusters. A typical dependence of
the linewidth AH,, on the temperature can be seen in Figs. 2 and 3 for
Cd,-,Mn,Te and Cdl-,Mn,Se. When the linewidth AH,, is of the order of
82
I"""'
I
"/ 1
40.05
6'
-0.05
the resonance field, H R,special care must be taken to obtain meaningful data
for the width and the position of the line. Because the radio frequency (r.f.)
field is linearly polarized in a typical microwave experiment, the experimental
signal contains, in addition to the resonance signal induced by the resonanceactive circularly polarized component of the r.f. field, also a contribution due
to the resonance-inactive-circular component. Thus, to obtain the true
resonance line shape from the raw data, the unwanted component has to be
subtracted. Such a correction was made when preparing Figs. 2 and 3.
It has been also found that the resonance field H R of the EPR line in DMS
shifted to lower magnetic fields with decreasing temperature (Oseroff, 1982;
Manoogian et al., 1982; Kremer and Furdyna, 1983, 1985a,b). As the shift
is found to be almost independent of the microwave frequency, it is
attributed to be due to an internal field rather than to a change in the
gyromagnetic factor itself (Oseroff, 1982). The observed change in the
position of the resonance line, Hi = H , - HR(whereN, is the magnetic field
corresponding to g = 2) is shown in Fig. 4 for Cdl-,Mn,Te and
Cdl-,Mn,Se.
The difficulty to perform EPR measurements when o n < 1, where o is
the microwave frequency and T is the relevant relaxation time, was successfully overcome by using the microwave Faraday rotation technique based on
3.
83
9000-
7000 n
Q)
P
P
b. Analysis
(a) Low Mn Concentrations (x s 0.03)
For extremely small concentrations of Mn, x < 0.001, the interaction
between the Mn ions is negligibly small. For the wide gap DMS, the gyromagnetic factor and the crystal field contribution were found to be larger for
Mn2+ in the zinc than in the cadmium compounds with the same anion,
whereas the hyperfine structure parameter A was found to be smaller in the
zinc compound (see, for example, Kikuchi and Azarbayejani, 1962, and
Title, 1963b).
Kikuchi and Azarbayejani (1962) noted that the ionic radius of the Mn2+
ion was intermediate in size between the ionic radii of Zn2+and Cd2+.Then
84
j
2/
3.
85
T( K)
FIG. 4. Temperature dependence of the internal magnetic field Hi for Cdl-,Mn,Se, and
Cdl-,Mn,Te measured at 9 and 35 GHz. Solid lines are least-squares fits to the data using Eq.
(12). H
i is defined as Hi = Ho - H E , where Ho is the resonance field corresponding to g = 2,
and H R is the measured resonance field. [After Oseroff (1982).]
86
to x 3 0.03. They found that within experimental error, the values of the
hyperfine structure constant, A , and the crystal field parameter, a, were
independent of concentration, at least forx < 0.008. Van Wieringen (1959,
who studied EPR of ZnS:Mn, calculated second and fourth moments of the
spectrum, and suggested that the variation observed in its shape was due to
the narrowing of the hyperfine splitting by the exchange interaction between
Mn2+ions. Ishiwaka (1966) analyzed the EPR spectra of exchange-coupled
Mn2+ ions in ZnS and CdS using a spin-Hamiltonian of the form:
X = gpBusHSz
+ C JijSi
i,j
Sj,
(8)
zzl
3.
87
and Bhagat (1985) suggested that the broadening of the resonance lines arises
from a distribution of internal fields, or rather their projections along the
applied field.
The observed broadening of the linewidth with decreasing temperature was
analyzed by several authors. Leibler et al. (1977), Grochulski et al. (1979),
Oseroff et al. (1979) and Manoogian et al. (1982) analyzed the temperature
dependence of AH,, in terms of a theory first used for antiferromagnets
above the Nee1 temperature (see Huber, 1972 and Seehra, 1972). The
broadening of the linewidth was described using Huber's expression:
AH,, = A T( - 2T , 7
+ B,
(9)
'AH
IaAH
T - GAH
+ B[; + 11,
88
Using Eq. (lo), the authors obtained values of GAHthat were jn good
agreement with those that follow from the dc susceptibility measurements at
the same field, = 3.3 kOe. The values of T ~ A Hbecame substantially larger for
x > 0.29 than for the low concentration region, and increased (within the
large experimental error, linearly) with concentration. The values of T f a ~
were calculated by using only the low temperature data. Oseroff et al.
(1980a,b) and Oseroff (1982) justified their analysis by arguing that the high
temperature data bear no relation with the critical phenomena, the high
temperature behavior being accounted for by the second term of Eq. (10).
These authors found that the critical exponent increased with concentration
for x < 0.30, and for larger concentration, it became less dependent on it.
It is not clear at this point if the behavior observed for C X A His just an artifact
of the model or not. The solid lines in Figs. 2 and 3 are a least-squares fit
of the experimental results with Eq. (10). Values of &H and CXAHfor a wide
range of concentration are given in Tables VI and VII for Cdl-,Mn,Te and
Cdl -,Mn,Se, respectively.
Webb et af.(1983) and Sayad and Bhagat (1985) used adifferent expression
to adjust the data for AHpp(T).Theirs is an empirical expression used first
by Bhagat et al. (1981) to fit the temperature dependence of AH in metallic
spin-glasses and in re-entrant ferromagnets. They expressed AHpp(T ) as
A H ~ ~ (=T ro
) + rl exp(-T/&),
where To is the high temperature linewidth, and
r
1
(1 1)
TABLE VI
OF THE CURIE
TEMPERATURE
8; THE
EXTRAPOLATED
TEMPERATURE
FOR THE MAXIMUM
OF x, f i ( H
0) AFTER ZERO-FIELD
COOLING;
THE TEMPERATURES
fimAND E H ~AND
, CRITICAL
EXPONENTS
CYANAND C X HOBTAINED
~
FROM THE
TEMPERATURE
DEPENDENCE
OF THE EPR LINEWIDTH
VALUESAS A FUNCTION
OF CONCENTRATION
FOR Cdl-,Mn,Te
2
5
10
15
20
30
40
53
60
3
8
22
35
I0
100
170
230
310
350
A f t e r O s e r o f f , 1982.
e l
el
2 + 0.5
8+1
12 + 2
18 ? 2
23 3
0.01
0.35
0.85
4+2
9 2 2
13 k 3
20 f 3
25 + 3
0.05
0.2
0.3
0.4
0.5
0.6
1.5
1.5
1.6
1.5
1.7
2.2
3.
89
TABLE VII
VALUESAS A FUNCTION
OF CONCENTRATION
FOR Cd,-,Mn,Se OF THE CURIE
TEMPERATURE
0, THE
EXTRAPOLATED
TEMPERATURE
FOR THE MAXIMUM
OF x, E(H
0) AFTER ZERO-FIELD
COOLING;
THE TEMPERATURES
f i AND
~ &,yi,
~ A N D CRITICAL EXPONENTS
CYAHA N D CYH~OBTAINED FROM THE
TEMPERATURE
DEPENDENCE
OF THE EPR LINEWIDTH'
+
(at. To)
-e(+io)
(K)
2
5
11
15
20
23
25
30
35
41
45
10
E{H+O)
(K)
26
<I
80
I20
135
1.8
3.0
0.5
+ 0.5
220
9.3
*1
330
15
+2
EAH
EHi
(K)
(K)
41
e l
41
-0.3
-1
-1
3.5 + 1
6+1
9+1
11.5 t 1
- 0.2
- 0.3
- 0.4
-1
ffAH(k0.2)
0.2
0.3
0.4
0.5
0.6
I .o
1.3
1.8
1.9
2.0
CXH,(+~.~)
2.2
2.9
32
5.4
"After O s e r o f f , 1982.
parameters associated with the freezing of the spins. Figure 5 shows the
agreement between Eq. (11) and the results for Cdo.-loMn0.30Te.Figure 6
shows the agreement between Eq. (11) and the data obtained by other
authors. Webb et al. (1983) and Sayad and Bhagat (1985) interpreted the
increase of AH,, with decreasing temperature as a result of the spins slowing
down when the freezing temperature was approached from above.
Oseroff et al. (1980a,b), Oseroff (1982), and Manoogian et al. (1982)
measured a shift of the resonance field HRwith temperature. Oseroff (1982)
attributed the observed shift to an internal field Hi,because it was found to
be independent, within the experimental error, of the microwave frequency
used. The increase of H
i with decreasing temperature was analyzed with an
expression similar to Eq. (10) in which the second term accounting for Hi at
high temperature was neglected. The expression used for Hi was:
90
1 04,
FIG.5 . Excess linewidth for Cdo.7oMno.3oTeas a function of temperature, fitted with Eq.
( 1 1 ) with To = 650Oe. r1 = 4500Oe, and To = 5OK. [After Webb et al. (1983).)
=0.30,p0.55
I
10
x= 0.20,y=o
40
80
T( K)
FIG.6 . Excess linewidth as a function of T fitted with Eq. ( 1 1) for Cdo.s,,Mno.z,,Te (0).
from Oseroff (1982); Cdo.s~sZno.17~Mn0.30Te
(O), C ~ O . I S Z ~ O , ~ S M ~
(a),
O . J and
OT~
Zn0.70Mn0.3oTe(0)from Manoogian et al. (1982). [After Webb et al. (1983).]
E H ~As
. for the linewidth AHpp,no divergence was observed for Hi at
T = TfHi (Kremer and Furdyna, 1983, 1985b). The solid lines shown in Fig.
4 are least-squares fit to the data using Eq. (12). Values for G H and
~ CYH~are
given in Tables VI and VII for CdI-,Mn,Te and Cdl-,Mn,Se, respectively, As mentioned, Kremer and Furdyna (1983, 1985b) found that for
Cdl-,Mn,Te, there was no anomaly in either AHpp or in the shift of HR at
the temperature where the ZFC dc susceptibility shows a cusp, contrary to
the case of metallic spin glasses (Schultz ef a[., 1980 and Oseroff et al.,
1983).
3.
91
2. MAGNETIC
SUSCEPTIBILITY
a. Experimental Results
At high temperatures the dc susceptibility for the DMS can be approximated by a Curie-Weiss law,
C / ( T - O ) + Xd,
(13)
..
120
1
4
o
m
....
T( K1
FIG.7. The temperature dependence of the inverse susceptibility x - l of Cdl-,MnxTe. The
low-temperature data (open circles) were taken after ZFC for increasing temperature in a 30 Oe
field. The high-temperature data (full circles) were obtained in an 8.5 kOe field. [After Oseroff
(1982).]
92
CdMnSe
100
200
100
T( K)
FIG. 8. Inverse susceptibility x-' of Cdl-,Mn,Se as a function of temperature. The low
temperature data (open circles) were taken after ZFC for increasing temperature in a 30 Oe field.
The higher temperature data (fun circles) were obtained in an 8.5 kOe field. [After Oseroff
(1982).]
3.0r
2.5h
a,
H g,-,Mn,Te
2.0-
a:0.01;
b:0.02:
c:0.03
e0.1 1
f0.14
g:O.16
h:0.22
i:0.25
i0.35
0)
FIG. 9. Inverse susceptibility x-' vs. T for Hgl-,Mn,Te. [After Nagata et al. (1980).]
3.
93
and Acker, 1981; Heidrich, 1981; Oseroff, 1982; Escorne and Mauger, 1982;
McAlister et al., 1984; Oseroff and Gandra, 1985; Spalek et al., 1986) which
can be discussed in the framework of a cluster model using Eq. (4), as in,
e.g., Kreitman et al. (1966); Simpson (1970); Savage et al. (1973); Nagata et
al. (1980; Galazka et al. (1980); Okada (1980); Shapira et al. (1984).
For still lower temperatures, T z 1.5 K, and at concentrations of Mn
x k 0.17, it was found that the susceptibility depends on whether the sample
was cooled in the presence or the absence of an external magnetic field H
(Nagata et al., 1980; 1981; Galazka et al., 1980; Oseroff et al., 1980a,b;
Oseroff and Acker, 1981; Heidrich, 1981; Escorne et al., 1981; Khattak et
al., 1981; Oseroff, 1982; Escorne and Mauger, 1982; Yang et al., 1983;
Amarasekara et al., 1983; McAlister et al., 1984; Oseroff and Gandra, 1985).
When the sample is cooled in zero field (ZFC), and then a magnetic field is
applied at the lowest temperature, the susceptibility measured in an
increasing temperature presents a maximum at a certain temperature r f ,that
depends both on x and the magnetic field H. The susceptibility for
Znl-,MnxTe, Cdl-,Mn,Te, and Cd,-,Mn,Se, showing these irreversible
effects, is given in Figs. 10-12, respectively, as a function of x and the
temperature near r f .
Most of the susceptibility data were taken for T 2 1.5 K. For this region
of temperatures, a cusp was observed only for x 2 0.17. This was the fact
that seemed to support the original idea that a spin glass (SG) transition in
DMS occurred only above the critical concentration derived for the first
nearest-neighbor percolation, x, 0.2 (Nagata et al., 1980; Galazka et al.,
1980; Oseroff, 1982; Yang et al., 1983; Amarasekara et al., 1983; McAlister
1.1,
T( K)
FIG. 10. Comparison of field-cooled (0)and zero-field-cooled (El) magnetizations for
Znl-,Mn,Te with x = 0.37, measured in 200 Oe. The values of magnetization are normalized
to their values at Tr. [After McAlister et 01. (1984).]
94
0.141
O.Ol0L----L---I----I------L--J
4
8
12
16
20
T( K)
FIG. 11. ZFC and FC susceptibility data of Cdl-xMnxSe and Cdl-,Mn,Te as a function of
T near Tr for H = 30 Oe. The arrow (-+) indicates ZFC, (+-) indicates FC. [After Oseroff (1982).]
111
0.
Cdl-xMnxTe
30.225
:I 0.30
30.40
:10.60
*.
3;
10
15
20
25
T( K)
FIG.12. ZFC and FC susceptibility data of Cdl-,MnxTe as a function of Tnear f i observed
in a 15 Oe field. [After Oseroff and Gandra (1985).]
3.
95
T(K)
96
l-
0.5
1.o
1.5
T( K)
FIG.14. Field-cooled susceptibility normalized to the concentration of Mn, for Cdl-,Mn,Te
(open symbols) and Cdl-,Mn,Se (closed symbols). [After Novak et al. (1985).]
FIG. 15. The behavior of the susceptibility near Tr in Cdo.ssMnod e for three different
magnetic fields. Arrows pointing to increasing Tindicate ZFC; arrows in decreasing-Tdirection
correspond to FC susceptibility data. The temperatures of the cusps are indicated by vertical
arrows. [After Oseroff (1984).]
3.
97
1 .o
0.8
-5 0.6
h
k-
0.4
0.2
o
0.53
0.45
HWOd
FIG. 16. The ratio F(H)/Tr(O)as a function of H. F ( H ) is the temperature at which the ZFC
susceptibility exhibits a maximum at a given H , with F(0) being the freezing temperature
obtained by extrapolating to zero magnetic field. [After Oseroff (1982).]
and Ayadi et al. (1986). They found a small shift of the maximum to higher
temperatures when the frequency of measurement was increased. A similar
behavior was also observed in metallic and other insulating SG systems (see,
e.g., Mulder et al., 1981a,b and Goldfarb and Patton, 1981).
b. Analysis
(c) Low Mn Concentrations
The magnetic susceptibility for x < 0.001 and temperatures above 1 K
follow the Curie-Weiss law given by Eq. (13), with a Curie-Weiss temperature @ 6 1 K (Deigen et al., 1967; Andrianov et al., 1976; Oseroff, 1984).
The diagmagnetic susceptibility contribution of the lattice, Xd , is significant
for this range of concentration. Values for the diamagnetic susceptibility for
11-VI compounds are collected in Table VIII. Magnetization measurements
performed down to 10 mK for CdTe with 0.05% Mn by Novak et al. (1981)
showed a departure from the Curie-Weiss law below 0.1 K. At these low
values of x, it seems unlikely that this departure results from the presence of
finite clusters, like pairs, triples, etc. That would require significant long
range interactions, i.e., cluster formation involving Mn2+ ions that are
separated by more than twenty coordination spheres. The deviation is
possibly due to the presence of hyperfine interaction and the existence of a
small but non-zero crystal field splitting.
98
TABLE VIII
DIAMAGNETIC
SUSCEPTIBILITY
FOR
~
THE
~
Compound
- ~ ~ ( 1 0emu/g)
-~
Reference
CdTe
3.5
3.1
3.3
3.3
3.8
3.5
3.6
3.2
3.3
3.6
2.3
2.1
2.3
CdSe
CdS
ZnTe
ZnSe
ZnS
HgTe
HgSe
HgS
3.
99
negative. It is likely that the use of an erroneous value of Xd may have been
responsible for the early interpretation of Sondermann and Vogt.)
From the Curie-Weiss temperature, it is possible to obtain an estimate for
the exchange interaction J. For antiferromagnets and ferromagnets,
@ =
2S(S + l)x
3kB
1K ZKJK,
100
TABLE IX
NEAREST-NEIGHBOR
EXCHANGE CONSTANT FOR THE DMS: (*) DETERMINED
BY
USINGTHE VALUEOF THE FIRSTSTEPOBSERVED
IN THE MAGNETIZATION
DATA
IN THE REFERENCE
QUOTED,(**) DETERMINED
BY USINGTHE CURIE-WEISS
e REPORTEDIN THE REFERENCE
QUOTED.
TEMPERATURES
Compound
CdMnTe
CdMnTe
CdMnTe
CdMnTe
CdMnTe
CdMnSe
CdMnSe
CdMnS
CdMnS
ZnMnTe
ZnMnTe
ZnMnTe
ZnMnSe
ZnMnS
HgMnTe
HgMnTe
HgMnTe
HgMnTe
HgMnSe
HgMnS
- Ji/ke (K)
Reference
-7
-6
-7.5
-1
7.7 5 0.3*
8.3 5 0.7*
-8
-4
-8
15**
-9
- 10*
13*
13
16
15
-7**
-8
8**
13**
Brumage et al. (1964), Savage et al. (1973), Nagata et af. (1980), Galazka
et al. (1980), Oseroff (1982), and Shapira et af. (1984) analyzed their
susceptibility data for low concentration and low temperature using a cluster
model with only nearest-neighbor (nn) interactions. This assumption was
made first by Brumage et a/. (1964) because the next nearest-neighbor (nnn)
interaction between Mn ions was believed to be much smaller than the nn
interaction. This assumption is supported by Kreitman et al. (1966), Bastard
and Lewiner (1979b), Davydov (1980), Escorne and Mauger (1982), and
Keesom (1983, who concluded that the interaction between nearest
neighbors in DMS is about one order of magnitude larger than the interactions between more distant Mn ions.
While it is difficult to calculate the interaction for large (realistic) clusters,
valuable, qualitative insights into the mechanisms governing inter-ion
interactions, and thus the overall behavior of the magnetic susceptibility of
DMS, can be obtained from calculations involving pairs, triples, etc., such
as those described above.
3.
101
where 9 ,P3, and P4 are the probabilities for singles, open triplets and closed
triplets. When the concentration of Mn increases above = 0.05, larger
clusters than triplets should be included. Kreitman et al. (1966) suggested that
an upper limit for the contribution of these larger clusters can be estimated
by assuming that clusters of four or more spins will be replaced by nn
102
o.8ro\...
y....
\I
Y
'. ...
--..+
I2
..._.
..._..
'.....&
0 CdMnTe
0 CdMnSe
0 CdMnS
A ZnMnSe
FIG. 17. Comparison between experimental results and theoretical predictions for R/x in
various materials; (0)Gaj et al. (1979); (0)Heiman et al. (1984); (0)Heiman et al. (1983),
Nawrocki et al. (1984); (+) Twardowski et a/. (1984a); (A) Heiman et al. (to be published);
Twardowski et al. (1984b); (V) Shapira and Foner (unpublished). The solid line represents
Eq. (16), which ignores clusters larger than triples. The dashed line represents Eq. (17), which
includes an estimate for these larger clusters. [After Shapira et a/. (1984).]
- = PI + 4 1 3
X
+ P4/15 + (1 - PI - P2 - P3 - P4)/5,
(17)
where PZis the probability for nn pairs, etc. In Fig. 17, the ratio .Vx vs. x
and its fit to Eqs. (16) and (17) are given for several DMS. The statistical
nature of Mn distribution was also confirmed by Barilero et al. (1986) in
Zn -,MnnTe.
Even though Shapira et al. (1984) and Aggarwal et al. (1984, 1985) were
able to fit the magnetization data by taking into account only interactions
between nearest-neighbors, there is experimental evidence that for temperatures below 2 K the interactions between farther Mn neighbors become
noticeable (see, e.g., Larson et al., 1986). When long range interaction is
assumed, the problem becomes difficult to handle because of the mathematical complications involved. In that case, only the number of isolated Mn
ions can be easily estimated from
Ns = Nx(1 - x)",
(18)
3.
103
x is the concentration, and n is the number of cation sites with which the
104
2l ;$$
Cd,-,Mn,Te
10
Zn
-Y,
20-
t=
10-
/d
40
M n SeB
20
2ol
20
40
x(at.96)
rZn,-,Mn,S
10
SG
//
SG
0
20
40
x( at .%)
1 0
L
x(at.%)
FIG. 18. Paramagnetic (P)-spin glass (SG) phase diagram, Tr vs. x, for several DMS.
from Galazka et al. (1980), (x) from Oseroff (1982) and (0)from Escorne
Cdl-,Mn,Te (0)
and Mauger (1982); Cdl-,Mn,Se (+) from Oseroff (1982) and (0)from Amarasekara et al.
(1983); Cdl-,Mn,S (0)from Yang et al. (1983); Hgl-,Mn,Se (0)from Khattak et al. (1981);
Znl-,Mn,Se (0)from Akbarzadeh et al. (1983); Znl-,Mn,Te (0)from McAlister et al. (1984);
Znl-xMn,S (0)Yang et al. (1983).
3.
105
25
Te
Cd,-,Mn,
20 -
15-
5
c-10-
5-
0.40
0.20
0.60
FIG. 19. Paramagnetic (P)-spin glass (SG) phase diagram, f i vs. x, for Cdl-,Mn,Te with
0.05 < x < 0.60. The solid line is a guide for the eyes. The values of f i were obtained from data
reported by Novak et al. (1984), Oseroff (1982), Galazka et al. (1980), and Oseroff and Gandra
(1985).
106
FIG. 20. Paramagnetic (P)-spin glass (SG) phase diagram, for Cdl-,Mn,Se with
0.05 < x < 0.50. The solid line is a guide for the eyes. The values of Tr were obtained by Novak
e f ai. (1985), Oseroff (1982), and Amarasekara e f a/. (1983).
251
FIG.21. Paramagnetic (P)-spin glass (SG)phase diagram, for Hgl -,Mn,Te. (A) from Brandt
e t a / . (1983), (x) from Mycielski et a/. (1984) and Otto e t a / . (1980), a.c. magnetic susceptibility
data; (0)from Mycielski et a/. (1984), Faraday rotation data. [After Mycielski ef al. (1984).]
3.
107
The specific heat, C , of DMS has been measured in the temperature range
between 0.3 K and 50 K and in magnetic fields up to 30 kOe (Nagata et al.,
1980, 1981; Galazka et al., 1980). In the case of the pure 11-VI host (i.e.,
x = 0), only the lattice vibrations contribute to the specific heat, and an
extrapolation of the low temperature specific heat to T = 0 determines the
Debye temperature. Values of the Debye temperature for different 11-VI
systems are collected in Table X.
TABLE X
VALUESOF THE DEBYE
TEMPERATURE
FOR
Zn
Cd
Hg
DIFFERENT
DMS (IN KELVINS)
S
Se
Te
260
223
202
174
158
140
1i a
144
144
The low temperature specific heat in DMS is substantially greater than that
in non-DMS. The data for Cdl-,Mn,Te at zero magnetic field are given in
Fig. 22 as an example. The excess specific heat (Cex= C - CII-I)has its
origin mainly in the magnetic properties of the Mn ions. Small changes in
the lattice vibrations spectrum can be also expected, but with the lattice
contribution being very small below 2 K (as compared to C,), these changes
can be safely neglected.
As shown in Fig. 23, for Mn concentrations greater than x = 0.2, C e x
behaves very similarly to the magnetic contribution found in metallic spinglasses (Wenger and Keesom, 1975, 1976); that is, C e x is linear in T i n the
108
0.1 0.5 1
3 5 10 30
T( K)
0.25
/.
-0
a20kG
028kG
a',
A'
FIG.23. Specific heat for Cdl-,Mn,Te, x = 0.2, 0.3, and 0.5, and pure CdTe. Dashed lines
show the linear behavior of C vs. T for x = 0.2 and 0.3. [After Galazka et a/. (1980).]
3.
109
Cd,-,MnxSe
0.3-
++
X=O.2O
9 H=O
aH=O(demagnetized)
f '
+ H=POkG
+
.
.
+A
+-
T( K)
FIG.24. Excess specific heat C,, vs. T for Cdo.8oMno.zoSe in fields of 0 and 20 kOe. Also
shown are measurements in zero field after demagnetization from 20 kOe. The insert shows the
zero-field data at low temperature. [After Arnarasekara et af. (1983).]
110
b. Analysis
( f ) Low Mn Concentrations
3.
111
112
was cooled. That is, it depended on whether the cooling took place in zero
field or whether it was done in the presence of a magnetic field and then, after
the field was turned off, the actual measurement was performed. The
differences between the two cases are shown in the inset of Fig. 24. The
authors interpreted this effect as resulting from loose (uncoupled) spins that
were bound to large clusters by the magnetic field and then remained frozen
after the field was turned off. This suggestion is supported by the fact that
a smaller effect was observed as x increased, until for x = 0.40 the effect was
not observed at all, agreeing with an almost complete absence of loosely
bound spins.
4. REMANENT
MAGNETIZATION
One of the characteristic features of SG systems is the presence of a timedependent remanent magnetization ( M R)below a certain temperature TR
(Tholence and Tournier, 1974; Guy, 1978; Gray, 1980; Oseroff et al., 1982).
In metallic SG, TRwas found to be within a few percent of the temperature
Tf ,where a cusp in the low field susceptibility is observed (Chamberlin et al.,
1981; Oseroff et al., 1982, 1983). The value of MR depends on the history
of the sample. If the sample is cooled from T > TR to Tz < TR in the
presence of an external magnetic field, and the remanence is measured after
the magnetic field is turned off, the M R is known as the thermoremanent
magnetization (TRM). On the other hand, if the sample is cooled from
5 > TRto Tz < TRin zero field, and a field is turned on and after a certain
time turned off again, then the MR obtained is known as the isothermal
remanent magnetization (IRM) (see Tholence and Tournier, 1974).
The remanent magnetization M Rin DMS was measured as a function of
time t , magnetic field, temperature and concentration of Mn ions by two
different techniques. One set of data was obtained by the EPR method,
where the magnetization M and the EPR spectrum were measured simultaneously by observing the EPR spectra of two thin p-doped silicon spin EPR
markers glued to the sample (see Schultz et al., 1980 and Oseroff, 1982). In
Fig. 26 the increase and decay of TRM and IRM with time, obtained by the
EPR method, is given for Cdo.6oMno.~oSe.The other set of data for MRin
DMS was obtained by using conventional magnetometers.
While there remains no doubt about the existence of remanence
phenomena in DMS the results of M R reported by different authors are in
considerable, quantitative disagreement. The M R data reported by Oseroff
(1982) and Oseroff and Gandra (1985) show behavior similar to that observed
in other metallic or non-metallic spin-glasses. However, the MR data
reported by Escorne et al. (1981) and Escorne and Mauger (1982) suggest that
the freezing process of the Mn spins in DMS can be analyzed in terms of
3.
113
rr,
(u
0.3
0
L
N
0.2 I
0
L
0.1 d
I 2
C I 0;
10
/Y
L _ _ L
25 50 100
---1
I
500 1000
5000
t(s)
FIG. 26. EPR measurements of the increase (i) and the decreae (d) of TRM and IRM as a
function of time for Cdo.aoMno.40Seat 4.2 K, normalized to the difference in linewidth between
H F C obtained
,
after FC in the resonance field and HZFC(25), measured after ZFC with the field
on for 25 seconds. [After Oseroff (1982).]
clustering, supporting the idea that the DMS behave like mictomagnetic
compounds rather than spin glasses. Figure 27 shows the time dependence
of TRM for Cdo.60Mn~.4~Te.
The evolution with time of the ZFC susceptibility for Cdo.60Mno.40Teis given in Fig, 28. The temperature dependence of
TRM in Cd, -,MnxTe is given in Fig. 29. This figure shows that MRvanishes
at TRwhich is close to r f . The results shown in Figs. 27 and 29 are in contradiction with the temperature dependence of M R reported for Cdl -xMnxTe
by Escorne and Mauger (1982). They observed the remanence well above T f
and found that MR did not depend on time on the scale of one-half hour.
Also, Galazka et al. (1980) did not detect, within spans of 30 min, any time
dependences for the zero-field-cooled magnetization. Possible explanations
for this disagreement will be discussed later in this section.
As mentioned above, the remanent magnetization depends also on the
strength of the magnetic field applied. An example of such dependence is
presented in Fig. 30 for Cd0.6Mn0.4Te.
The study of the remanent magnetization MRmay provide further insights
into whether the ZFC cusp observed in DMS should be associated with a SG
transition or whether their behavior is rather more typical of mictomagnets,
where the cusp results from cluster freezing. As mentioned above, only few
measurements of M R in DMS were reported so far, and the data disagreed.
Oseroff (1982) and Oseroff and Gandra (1985) fitted the time dependence of
MR for Cdl-,Mn,Te and Cdl-,Mn,Se with a log(t) law,
MR(t) = M R ( t o ) k s l o g t / t o ,
(19)
114
0.7 6 -
cdO.60Mn0.40Te
0)
3 0.74-
$. 0.72-
z
oc
!-
&
0.70-
0.68
where to was the initial time and S the slope of the logarithmic decay. Figure
31 gives S vs. H in Cdo.soMn0.40Tefor the thermo-remanent magnetization
(TRM) and the isothermal remanent magnetization (IRM). A similar
behavior was observed in metallic and insulating SG (see Guy, 1978, Ferre
et al., 1980, and Oseroff et al., 1982).
Kinzel(l978) and Dasgupta et al. (1979) performed Monte Carlo calculations of a SG system consisting of an Ising model with random interactions
3.
115
T( K)
FIG.29. TRM vs. Tfor Cdl-xMnxTe, obtained by FC to T = 1.8 K in 100 Oe and measured
after the field was turned off for 3 minutes. The arrows indicate the maximum of ZFC
susceptibility for the three concentrations measured. [After Oseroff and Gandra (1985).]
(20)
116
Cd0.60Mn0,40Te
TRM
+ IRM
T=6K
Cd0.60 Mn0.40Te
0.3
"0
10
20
30
H (kOe)
40
50
FIG. 31. Variation of the slope S for the log(t) decay of TRM (0)and IRM (0)for
Cdo.soMno.aoTeas a function of magnetic field. The slope was obtained from the change of M R
between 1 0 0 s and 1ooOs. [After Oseroff and Gandra (1985).)
3.
117
T = 6K
.I
'
0
10
20
30
40
50
H (kOe)
FIG. 32. Field dependence of the exponent a of the power law M ( t ) = M(t,)t-" for
Cdo.mMno.4oTemeasured at T = 6 K, (0)IRM data and (0)TRM data. [After Oseroff and
Gandra (1985).]
118
3.
119
Above x k 0.17 with T 2 1.3 K the EPR, susceptibility and specific heat
data show several features that can be associated with spin glass transition.
All DMS show a cusp for the ZFC susceptibility measured at low field. The
temperature dependence of the maximum with magnetic field is similar to
other systems exhibiting a spin glass behavior. The specific heat shows a
linear temperature dependence for T < Tf and no anomaly around 3.For
x < 0.17, the low temperature magnetization results appear to indicate a
similar behavior to those observed in the high concentration region.
However, at this point there is not enough data to assure that DMS show a
spin-glass behavior in the whole range of concentrations in which these
materials exist. Conspicuous in this field is our lack of understanding of the
phenomena related to remanent magnetization in the spin glass phase. Also,
the question of the nature of the antiferromagnetic ordering in very concentrated samples (x 5: 0.60) requires still further investigation (for discussion
of this problem, see the contribution by Giebultowicz and Holden in this
volume).
It is to be hoped that from the current intense activity a deeper and more
thorough understanding of magnetic properties of DMS and of mechanisms
underlying these will shortly emerge.
References
Abragam, A. (1950). Phys. Rev. 19, 534.
Abragam, A., and Bleaney, B. (1970). Electron Paramagnetic Resonance of Transition Ions,
Oxford University Press.
Aggarwal, R. L., Jasperson, S. N., Shapira, Y., Foner, S., Sakakibara, T., Goto, T., Miura,
N., Dwight, K., and Wold, A. (1984). In Proceedings of the 17th InternationalConference
on the Physicsof Semiconductors(J. D. Chadi and W. A. Harrison, eds.) Springer-Verlag,
p. 1419.
Aggarwal, R. L., Jasperson, S. N., Becla, P., and Galazka, R. R. (1985). Phys. Rev. B32,5132.
Aharony, A. (1978). J. Phys. C11, L457.
Amarasekara, C. D., Galazka, R. R., Yang, Y. Q., and Keesom, P. H. (1983). Phys. Rev. B27,
2868.
Anderson, P. W. (1979). In 111-CondensedMatter, Les Houches, 1978 (R. Balian, R. Maynard,
and G. Toulouse, eds.) North-Holland, Amsterdam 1979.
Andrianov, D. G., Gimelfarb, F. A., Kushnir, P. I., Lopatinskii, I. E., Pashkovskii, M. V.,
Savelev, A. S., and Fistul, V. I. (1976). Sov. Phys. Sernicond. 10, 66. (Fiz. Tekh.
Poluprov. 10, 111).
Andrianov, D. G., Pavlov, N. M., Savelev, A. S., Fistul, V. I., andTsiskarishvili, G. P. (1980).
Sov. Phys. Semicond. 14, 711. (Fiz. Tekh. Poluprov. 14, 1202).
Aresti, A., Manca, P., and Spiga, A. (1979). Chem. Phys. Lett. 63, 139.
Ayadi, M., Nordland, P., Ferre, J., Mauger, A., and Triboulet, R. (1986). J. Magn. Magn.
Muter. 54- 57, 1223.
Bastard, G., and Lewiner, C. (1979a). Phys. Rev. B20,4256.
Bastard, G . , and Lewiner, c. (1979b). J. Phys. Lett. 40, L177.
Bastard, G., and Lewiner, C. (1980). J. Phys. C13, 1469.
120
Bhagat, S. M., Spano, M. L., and Lloyd, J. N. (1981). Solid State Commun.38, 261.
Bloembergen, N., and Rowland, T. J. (1955). Phys. Rev. 97, 1679.
Brandt, N. B., Moshchalkov, V. V., Skrbek, L., Taldenkov, A. N., and Chudinov, M. (1981).
JETP Letr. 35, 401. (Pisma v Zh. Eksper. Teor. Fiz. 35, 326).
Brandt, N. B., Tsidilkovskii, I. M., Moshchalkov, V. V., Ponikarov, B. B., and Skrbek, 1.
(1982). Sov. J. Low. Temp. Phys. 6. (Fiz. Nizkikh Temper 6).
Brandt, N. B., Moshchalkov, V. V., Orlov, A. O., Skrbek, L., Tsidilkovskii, I. M., and
Chudinov, S. M. (1983). Sov. Phys. JETP57, 614. (Zh. Eksper. Teor. Fiz. 84, 1050).
Brumage, W. H., Yarger, C. R., and Lin, C. C. (1964). Phys. Rev. 133, A765.
Candea, R. M., Hudgens, S. J., and Kastner, M. (1978). Phys. Rev. Bl8, 2733.
Cannella, V., and Mydosh, J. A. (1972). Phys. Rev. B6, 4220.
Chamberlin, R. V., Hardiman, M., and Orbach, R. (1981). J. A p p l Phys. 52, 1771.
Ching, W. Y., and Huber, D. L. (1984). Phys. Rev. B30, 179.
Dasgupta, C., Ma, S., and Hu, C. (1979). Phys. Rev. 820, 3837.
Davydov, A. B., Noskova, L. M., Ponikarov, B. B., and Ugodnikova, L. A. (1980). Sov. Phys.
Semicond. 14, 869. (Fiz. Tekh. Poiuprov. 14, 1461).
Deigen, M. F., Zevin, V. Ya., Maevskii, V. M., Potykevich, I. V., and Shanina, B. D. (1967).
Sov. Phys. Solid State 9, 773. (Fiz. Tvevdogo Tela 9, 983).
DeJonge, W. J. M., Otto, M., Denissen, C. J . M., Blom, F. A . P., van der Steen, C., and
Kopinga, K. (1983). J. Magn. Magn. Mater. 31-34, 1373.
De Seze, L. (1977). J. Phys. C10, L353.
Domb, C., and Dalton, N. W. (1966). Proc. Phys. Soc (London) 89, 859.
Dorain, P. B. (1958). Phys. Rev. 112, 1058.
Dormann, E., and Jaccarino, V. (1974). Phys. Lett. A48, 81.
Escorne, M., and Mauger, A. (1982). Phys. Rev. B25, 4674.
Escorne, M., Mauger, A , , Triboulet, R., and Tholence, J . L. (1981). Physica 107B, 309.
Estle, T. L., and Holton, W. C. (1966). Phys. Rev. 150, 159.
Falkowski, K. (1967). Acta Phys. Polon. 32, 831.
Ferre, J . , Pommier, J., Renard, J. P., and Knorr, K. (1980). J. Phys. C 13, 3697.
Fidone, I . , and Stevens, K. W. H. (1959). Proc. Phys. SOC. (London) A73, 116.
Fischer, K. H. (1976). Physica (Utrecht) 86-88, 813.
Frisch, H. L., Sonenblick, E., Vyssotsky, V. A,, and Hammersley, J. M. (1961). Phys. Rev.
124, 1021.
Furdyna, J. K. (1982). J. Appl. Phys. 53, 7637.
Gabriel, J. R., Johnston, D. F., and Powell, M. D. J. (1961). Proc. Roy. SOC. (London) A264,
503.
Gaj, J . A., Planel, R., and Fishman, G. (1979). Solid Stare Commun.29, 435.
Galazka, R. R., Nagata, S., and Keesom, P. H. (1980). Phys. Rev. B22, 3344.
Giebultowicz, T., Minor, W., Kepa, H., Ginter, J., and Galazka, R. R. (1982). Magn. Magn.
30, 215.
Giebultowicz, T. M., Rhyne, J. J., Ching, W. Y., and Huber, D. L. (1985). To be published.
Ginter, J., Kossut, and Swierkowski, L. (1979). Phys. Stat. Sol. (b) 96, 735.
Goldfarb, R. B., and Patton, C. E. (1981). Phys. Rev. B24, 1360.
Gray, E. M. (1980), J. Magn. Magn. Mater 15-18, 177.
Grochulski, T., Leibler, K., Sienkiewicz, A., and Galazka, R. R. (1979). Phys. Stat. Sol. (b)
91, K73.
Guy, C. N. (1978). J . Phys. F8, 1309.
Hall, T. P. P., Hayes, W., Williams, F. I. B. (1961). Proc. Phys. Soc (London) A78, 883.
Heidrich, H . (1981). Phys. St. Sol. (a) 67, 163.
Heiman, D., Shapira, Y., and Foner, S. (1983). Solid Stare Commun. 45, 899.
3.
121
Heiman, D., Shapira, Y., Foner, S., Khazai, B., Kershaw, R., Dwight, K., and Wold, A. (1984).
Phys. Rev. B29, 5634.
Huber, D. L. (1972). Phys. Rev. B6, 3180.
Ishikawa, Y. (1966). J. Phys. SOC. Japan 21, 1473.
Ivanov-Omskii, V. I., Kolomiets, B. T., Ogorodnikov, V. K., Rud, Yu. V., and Tsmots, V.
M. (1972a). Phys. Stat. Sol. (a) 13, 61.
Ivanov-Omskii, V. I . , Kolomiets, B. T., Ogorodnikov, V. K., Smekalova, K. P., and Tsmots,
V. M. (1972b). Phys. Stat. Sol. (a) 14, 51.
Jaczynski, M., Kossut, J., and Galazka, R. R. (1978). Phys. Stat. Sol. (b) 88, 73.
Keesom, P. H. (1985). Private communication.
Keesom, P. H., (1986). Phys. Rev. B33, 6512.
Khattak, G. K., Amarasekara, C. D., Nagata, S., Galazka, R. R., and Keesom, P. H. (1981).
Phys. Rev. B23, 3553.
Kikuchi, C., and Azarbayejani, G. H. (1962). J. Phys. SOC. Japan 17, Suppl. B-I, 453.
Kinzel, W. (1979). Phys. Rev. B19, 4595.
Kittel, C., and Abraham, E. (1953). Phys. Rev. 90, 238.
Koh, A. K., Miller, D. J., and Grainger, C. T. (1984). Phys. Rev. B29, 4904.
Kondo, J . (1965). Prog. Theor. Phys. 33, 575.
Kreitman, M. M., and Barnett, D. L. (1965). J. Chem. Phys. 43, 364.
Kreitman, M. M., Milford, F. J., Kenan, R. P., and Daunt, J. G. (1966). Phys. Rev. 144,367.
Kremer, R. E., and Furdyna, J. K. (1983). J. Magn. Magn. 40, 185.
Kremer, R. E., and Furdyna, J. K. (1985a). Phys. Rev. B31, 1 .
Kremer, R. E., and Furdyna, J . K. (1985b). Phys. Rev. B32, 5591.
Lambe, J., and Kikuchi, C. (1960). Phys. Rev. 119.
Larson, B. E., Hass, K. C., and Ehrenreich, H. (1985). Solid State Commun. 56, 347.
Larson, B. E., Hass, K. C., and Aggarwal, R. L. (1986), Phys. Rev. 833, 1789.
Leibler, K., Giriat, W., Checinski, K., and Wilamowski, 2. (1970a). Proc. XVICong. Ampere,
Bucuresti 1970, p. 993.
Leibler, K., Giriat, W., Checinski, K., and Wilamowski, 2. (1970b). Proc. XVICong. Ampere,
Bucuresti 1970, p. 987.
Leibler, K., Giriat, W., Wilamowski, Z., and Iwanowski, R. (1971). Phys. Stat. Sol (b)47,405.
Leibler, K., Giriat, W., Checinski, K., Wilamowski, 2.. and Iwanowski, R. (1973). Phys. Stat.
Sol. (b) 55, 447.
Leibler, K., Biernacki, S., Sienkiewicz, A., and Gaiazka, R. R. (1976). Proc. XIX Congresse
Ampere, Heidelberg 1976.
Leibler, K., Sienkiewicz, A., Checinski, K., Galazka, R., and Pajaczkowska, A. (1977). Proc.
of the International Conf. on Phys. of Narrow Gap Semiconductors (Rauluszkiewicz, J.,
Kaczmarek, E., and Gorska, M., eds.) P W N Publishers, Warsaw 1978, p. 199.
Lewicki, A., Spalek, J., Furdyna, J. K., and Galazka, R. R. (1985). ProceedingsInternat. Conf.
Magnet., San Francisco. ( J . Magn. Magn. Muter. 54-57, 1221 (1986).)
Lewiner, C., Gaj, J. A., and Bastard, G . (1980). J. Physique C5, 289.
Liu, L. (1982). Phys. Rev. B26, 975 and 6300.
Manoogian, A., Chan, B. W., Brun del Re, R., Donofrio, T., and Woolley, J. C. (1982). J.
Appl. Phys. 53, 8934.
Matamura, 0. (1959). J. Phys. SOC.Japan 14, 108.
Matarrese, L. M., and Kikuchi, C. (1956). J. Phys. Chem. Solids 1, 117.
McAlister, S. P., Furdyna, J. K., and Giriat, W. (1984). Phys. Rev. B29, 1310.
Mulder, C. A. M., van Duyneveldt, A. J., van der Linden, H. W. M., Verbeek, B. H., van
Dongen, J. C. M., Nieuwenhuys, G. J., and Mydosh, J. A. (1981a) Phys. Lett. 83A, 74.
Mulder, C. A. M., van Duyneveldt, A. J., and Mydosh, J. A. (1981b). Phys. Rev. B23, 1384.
122
Mullin, D. P., Galazka, R. R., and Furdyna, J. K. (1981). Phys. Rev. 824, 355.
Mycielski, A., Rigaux, C., Menant, M., Dietl, T., and Otto, M. (1984). Solid State Commun.
50, 257.
Mydosh, J. A. (1981). Lecture Notes in Physics 149, Springer-Verlag.
Nagata, S.,Keesom, P. H., and Harrison, H. R. (1979). Phys. Rev. B19, 1633.
Nagata, S.,Galazka, R. R., Mullin, D. P., Akbarzadeh, H., Khattak, G . D., Furdyna, J. K.,
and Keesom, P. H. (1980). Phys. Rev. B22, 3331.
Nagata, S., Galazka, R. R., Khattak, G. D., Amarasekara, C. D., Furdyna, J. K., and Keesom,
P. H.(1981). Physica 107B, 311.
Nawrocki, M., Planel, R., Mollot, F., and Kozielski, M. J. (1984). Phys. Stat. Sol. (b) 123,99.
Novak, M. A., Oseroff, S., and Symko, 0. G. (1981). Physica 107B, 313.
Novak, M. A., Symko, 0. G., Zheng, D. J., and Oseroff, S.; (1984). Physica 126B,469.
Novak, M. A., Symko, 0. G., Zheng, D. J., and Oseroff, S. (1985). J. Appl. Phys. 57, 3418.
Okada, 0.(1980). J. Phys. Soc. Japan 48, 391.
Oseroff, S. B. (1982). Phys. Rev. 25, 6584.
Oseroff, S. B. (1984). Unpublished results.
Oseroff, S. B., and Acker, F. (1981). Solid State Commun. 37, 19.
Oseroff, S. B., and Gandra, F. (1985). J. Appl. Phys. 57, 3421.
Oseroff, S. B., Calvo, R., and Giriat, W. (1979). J. Appl. Phys. 50, 7738.
Oseroff, S. B., Calvo, R., Giriat, W., and Fisk, Z. (198Oa). Solid Stafe Commun. 35, 539.
Oseroff, S. B., Calvo, R., Fisk, Z., and Acker, F. (1980b). Phys. Lett. 80A, 311.
Oseroff, S.,Mesa, M., Tovar, M., and Arce, R. (1982). J. Appl. Phys. 53, 2208.
Oseroff, S., Mesa, M., Tovar, M., and Arce, R. (1983). Phys. Rev. B27, 566.
Otto, M., D i d , T., Mycielski, A., Dobrowolska, M., and Dobrowolski, W. (1980). Proc. of
the X Conf. On Physics of Semicond. Compounds, Jaszowiec, Poland 1980, p. 225.
Pajaczkowska, A. (1978). Prog. Cryst. Growth and Charact. 1, 289.
Pajaczkowska, A., and Pauthenet, R. (1979). J. Magn. Magn. Muter. 10, 84.
Price, M. H. L., and Stevens, K. W. H. (1950). Proc. Phys. SOC.A63, 36.
Savage, H., Rhyne, J. J., Holm, R., Cullen, J. R., Carroll, C . E., and Wohlfarth, E. P. (1973).
Phys. St. Sol (b) 58, 685.
Sayad, H. A., Bhagat, S. M. (1985). Phys. Rev. B31, 591.
Schneider, J., Sircar, S. R., and Rauber, A. (1963). Z . Naturf. 18a, 980.
Schultz, S.,Gullikson, E. M., Fredkin, D. R., andTovar, M. (1980). Phys. Rev. Left.45, 1508.
Shapira, Y. (1985). Private communication.
Shapira, Y . , and Foner, S. (1985). Unpublished.
Shapira, Y . , Foner, S., Ridgley, D. H., Dwight, K., and Wold, A. (1984). Phys. Rev. B30,4021.
Seehra, M. (1972). Phys. Rev. 86, 3186.
Simpson, A. W. (1970). Phys. Stat. Sol. (b) 40, 207.
Singh. S., Singh, P. (1980). J. Phys. Chem. Solids. 41, 135.
Sondermann, U. (1976). J. Magn. Magn. Muter. 2, 216.
Sondermann, U. (1979). J. Mugn. Magn. Mafer. 13, 113.
Sondermann, U. (1980). J. Magn. Magn. Muter. 21, 228.
Sondermann, U.,and Vogt, E. (1977a). J. Mugn. Magn. Muter. 6,223.
Sondermann, U., and Vogt, E. (1977b). Physicu 86-88B,419.
Spalek, J., Lewicki, A., Tarnawski, Z., Furdyna, J. K., Galazka, R. R., and Obuszko, 2.
(1986). Phys. Rev. B33, 3407.
Swarup, P. (1959). Can. J. Phys. 37, 848.
Tholence, J. L., and Tournier, R. (1984). J. Physique 35, C4-229.
Title, R. S. (1963a). Phys. Rev. 130, 17.
Title, R. S. (1963b). Phys. Rev. 131, 2503.
3.
123
CHAPTER 4
and
T. M. Holden
ATOMIC ENERGY OF CANADA LIMITED,
CHALK R N E R NUCLEAR LABORATORIES,
CHALK RN ER, ONTARIO, CANADA
INTRODUCTION.. . . . . . . . . . . . . . . . . . .
1 . Motivafion and Introductory Remarks . . . . . . . . .
2 . Theory of Neutron Scattering in Systems with Short-Range
Magnetic Order . . . . . . . . . . . . . . . . . .
3 . Theoretical Models of Spin Correlation Phenomena and
Spin Dynamics in DMS Systems. . . . . . . . . . . .
11. POWDER
DIFFRACTION
EXPERIMENTS
ON Znl-,Mn,Te.
. . . .
4. Elernentsof the Experimental Technique. . . . . . . . .
5 . Results of Experiments . . . . . . . . . . . . . . .
111. NEUTRONSCATTERING
STUDIES
OF Cdl -,Mn,Te.
. . . . . .
6 . Samples and Elements of Experimental Technique. . . . .
7 . Dvfraction Studies of Cdl -,Mn,Te Single Crystals. . . . .
8 . Inelastic Neutron Scattering in Cd0. 35M no.ssT e . . . . . .
IV. CONCLUDING
REMARKS. . . . . . . . . . . . . . . . .
REFERENCES
. . . . . . . . . . . . . . . . . . . . .
I.
125
125
128
135
140
140
141
155
155
157
173
177
179
I. Introduction
1 . MOTIVATION
AND INTRODUCTORYREMARKS
125
Copyright 0 1988 by Academic Press, Inc.
All rights of reproduction in any form reserved.
ISBN 0-12-752125-9
126
127
antiferromagnet is the face centered cubic (FCC) array of spins, each spin
interacting antiferromagnetically only with its nearest neighbors. The
A:'_,Mn,BV' DMS are very close realizations of such a system. The consequences of the frustration of the spin lattice are far reaching in the context
of magnetic properties of these materials. For example, as a consequence of
frustration, the ground state of the system is highly degenerate, there is no
long range order in three dimensions, and the mechanism of spin excitations
becomes very complex. Frustration may also result in a spin-glass behavior
of the system. Frustrated FCC systems were subjects of many recent
theoretical analyses and modelling studies (see, e.g., Villain, 1978; Grest and
Gabl, 1979; de Seze, 1979; Alexander and Pincus, 1980; Ching and Huber,
1981, 1982a,b, 1984; Fernandez et al., 1983), where various aspects of this
general problem are discussed. Since the cubic members of the DMS family,
Cdl-,Mn,Te and Znl-,Mn,Te, resemble closely the prototype FCC
frustrated antiferromagnets considered in these theoretical works, the results
obtained in both elastic and inelastic neutron scattering experiments on these
crystals can therefore be directly compared with the predictions of the models,
providing valuable feedback for further development of the theories.
In this chapter, we give a review of neutron scattering studies performed
on Cdl-,Mn,Te and Znl-,Mn,Te. Since some readers of this book may be
unfamiliar with neutron scattering methods, certain basic elements of
neutron scattering theory are given in Sec. 2, together with a brief explanation of the data analysis procedures, in order to establish a common
language. We then follow, in Sec. 3, with a short description of the models
of diluted antiferromagnets used in the interpretation of the experimental
results.
Cdl-,Mn,Te can be obtained in single crystal form for Mn concentrations
0 s x 4 0.7. On the other hand, Znl-,Mn,Te samples used in the experimental studies described in this chapter were available only as polycrystals.
Thus, different experimental techniques and data processing procedures had
to be applied in these two cases. Although the materials are quite similar as
far as their magnetic properties are concerned, we chose to discuss them in
separate sections, hoping that this will contribute to the clarity of the
presentation. Part I1 of this chapter is devoted to ZnI-,Mn,Te, where the
experimental findings obtained by neutron scattering on powder samples are
presented. The analysis of the short-range antiferromagnetic order observed
in this material yielded direct information about the spin-spin correlation
coefficients between magnetic neighbors as a function of their separation,
concentration of Mn ions, and the temperature. Then, in Part 111, we discuss
neutron diffraction experiments on single crystals of Cdl-,Mn,Te. In
addition to the information similar to that presented in Part 11, these
experiments led to the observation of the anisotropy of the magnetic
128
correlation length and permitted the study of the dynamics of the spin system
by inelastic scattering methods.
2. THEORYOF NEUTRONSCATTERING
IN SYSTEMS
WITH
SHORTRANGEMAGNETIC
ORDER
a. Basic Definitions
The number of neutrons counted in a detector in an experiment to measure
both the energy and the angular distributions of neutrons scattered by a
sample is proportional to the double differential scattering cross section,
dza/dQdc'. The incident (scattered) neutrons are characterized by wave
vector k(k') and energy
The solid angle into which the neutrons are
scattered is designated by Q. The momentum transfer, hQ, and energy
transfer, h o , to the sample are given by
&(&I).
hQ = hk - hk',
and
h o = E - E'.
(4)
' j+m
s ( Q , 0)= 277
(F&(O)FQ(t))=
v'a'ua
-m
dt e-'"''(F&O)&(t))
pyPa((v',
( Y ' I F Q I Y , O!)12E?iwa'mf.
(5)
(6)
The states of the sample are specified by a and the energy transfer to the
sample is the difference of the initial and final states, h a , , , = Em<- Em.
The correlation function is averaged over the initial states of the sample
4.
129
FQ(O)=
c (bt
In Eq. (7,b, is the nuclear scattering length of atom I , S?'is the projection
of the electronic spin perpendicular to the scattering vector Q, s is the neutron
spin and f ( Q ) is the magnetic form factor. Rt denotes the position of site I
in the lattice. The neutron gyromagnetic ratio and classical electron radius
are denoted by y and r e , respectively.
In the case of a pure magnetic material, such as a type-I11 antiferromagnet
which we will discuss later, the scattering function at a particular wave vector
Q has the form of a delta function at the characteristic frequency of the
elementary excitation (a spin wave in this case), W Q , where
S(Q,W )
OQ).
(9)
S(Q)
-m
S(Q, 0)d o
Uii(W~(0)).
(10)
is equal to the static structure factor in the limit of very high incident energies
m. For this reason, it is important in diffraction measurements of S(Q)
to employ high incident neutron energies.
For an alloy of the type Al-,M,B, where there is a random substitution
of magnetic atoms, M, in place of the non-magnetic atoms, A, the differential scattering cross section is given by
E
-+
130
1I [VI(FM.I
- FA,/) + FA,/ + FB,I].
(13)
The independent quantity r,q is unity if an M atom occupies the lth lattice site
and zero otherwise. The individual scattering amplitudes are:
F M ,=~ [- bM
FA,I= -bAe
+ 2yref(Q)eiQ'R's sf],
*
iQ.Ri
(14)
(15)
FB,I= -bBe i Q . U b + r )
(16)
The average in Eq. (12) has to be taken over the statistical distribution of
magnetic atoms, the electron spin direction, the neutron spin direction
(assuming the beam is unpolarized) and, for a powder sample, over all
orientations of the crystal lattice with respect to the incident beam. The
nuclear scattering length b~ is taken to be identical on each M atom, and it
is assumed for the present that the nuclear incoherent scattering is absent,
but it is an easy matter to include it at a later stage.
On carrying out the first two averages, with the assumption that V I and
Si] are statistically independent, the resulting expression for the cross
section in a single-crystal is
with
G(Q) = ~ S ( +
S 1)
C ( ~ -dS-j)eiQ'R[.
1#0
(18)
Here h = (h, k,I) labels the reciprocal lattice sites, and Kh are corresponding
reciprocal lattice vectors.
is the volume of the primitive unit cell, x is the
manganese concentration and Fh is the unit cell structure factor,
Fh = x b -t
~ (1 - X)bA + bgeiKh.'.
(19)
The scattering is made up of three parts: the nuclear Bragg scattering from
the site-averaged scattering length, the diffuse nuclear component coming
from the site-to-site fluctuation in occupancy on the A-sites, and the
magnetic diffuse scattering corresponding to the correlations between spins
131
c PeXP[i&
r
'
r],
(20)
where p = f1 for the two antiparallel spin directions, and the sum extends
over all spin sites in the magnetic unit cell. Consider now the unit cell of the
type-111 antiferromagnetic structure, depicted in Fig. lb. One can readily
check that in this case, the magnetic superstructure reflections occur at h'
being odd integer, and for k' and I' being even or odd integers. Here the
symbols h', k',I' refer to the tetragonal unit cell (2a x a x a). Usually, both
magnetic and nuclear reflections are described in terms of a common
notation involving the cubic indices. In the latter notation, the magnetic
reflection indices are: h is an odd half-integer, while k and I are a combination
of odd and even integers.
c. Powder Diffraction
On carrying out the powder average, the differential cross section is found
to be
where 26' is the scattering angle, Oh is the Bragg angle and Mh is the multiplicity of the h-th reflection. In reality, the Debye-Scherrer peaks have a
finite width due to experimental resolution, so for comparison with experiment, the delta-function may be replaced by an appropriate Gaussian. The
function of interest, that measures the correlations between spins, is
1#0
with
At =
S(S
+ 1 ) ( S o - SI),
132
and
The spherical Bessel functions j o , j 2 arise from averaging over the spin
directions and from the axial symmetry of the spin tensor with respect to the
vectors RI.The z-axis for each value of I is the vector joining the positions
of the zeroth and I sites. If the absolute cross section is measured, it is possible
to determine both S and the spin correlation functions. For the range of Q
values studied in powder diffraction experiments, jz(QZ?l) is approximately
equal to -jo(QR1), so that in effect only A I - Br can be determined by fitting
Eq. (21) to experimental results of the accuracy normally achieved. Similar
expressions for the magnetic correlations in amorphous materials have been
used by Nagele et al. (1978).
It is convenient to reinterpret the parameters ( A - BI)in Eq. (22) as the
sum over all sites with the same R I , i.e., as the correlation within a given
coordination shell. The quantity A I - BI is usually treated as a fitting
parameter for each shell of neighbors and measures the size and range of the
correlations. It is instructive and straightforward to derive A / for each shell
of neighbors on the basis of what appears to be the ideal spin structure
observed in DMS, the type-I11 antiferromagnet. The parameter BI depends
on the spin direction in the ideal structure. In the absence of information
about the actual spin direction we often cannot calculate B I , but the ideal
spin-spin correlation coefficients A I, including multiplicity, can be derived
(see, e.g., Table I1 discussed later in this chapter).
The additional contribution to the total cross section from the incoherent
nuclear scattering from each component is given by
d&nc
--
dC2
- "OF
4n
+ xu$ + (1
- x)ajy].
The contribution to the diffuse intensity from thermal diffuse scattering (i.e.,
the sum of the phonon scattering) may be written in the incoherent approximation as
4.
133
The static susceptibility per formula unit, allowing for spin correlations, is
therefore given by
d. PhenomenologicaI Approach
134
(SO * S r ) S R = D(RI)(SO * s I ) P ,
(29)
Equation (31) shows that the diffuse scattering intensity is described by the
Fourier transform of the correlation damping function D(R). This fact
makes it possible to calculate the shapes of the diffuse scattering maxima for
various types of correlation damping by carrying out relatively easy Fourier
inversions. For instance, in the case of D(R) in the form
R- exp(- KR)
(324
which are often employed in the theory of critical scattering, one obtains the
following lineshapes, respectively:
1
44) = q2
1 + 7
K
(Lorentzian lineshape)
(334
4.
135
or
Additionally, it can be concluded on the basis of Eq. (31) that any symmetric
damping function D(R) = D(-R) leads to a distribution of the scattered
intensity that is symmetric with respect to a given reciprocal lattice point.
So far, we have assumed that D(R)is isotropic. In general, however, the
rate of correlation damping may very well be different for various crystallographic directions. This effect is clearly observed in Cdl -,Mn,Te. Shortrange spin correlations that depend on direction can be simply modeled by
replacing the isotropic exponent K Rin Eq. (32) by an anisotropic form, e.g.,
[ d x 2 + ~,?y' + K$?]'",
where K , , K ~ and
, K* are correlation lengths for the
three principal axes. By applying an uncomplicated transformation of
variables, one can then easily obtain the following lineshapes:
or
I
136
where the first term describes isotropic exchange between Mn ions, labeled
byj, on sublattice k (1 or 2). The second term introduces an anisotropy field
that gives a gap at q = 0 in the excitation spectrum. For non-cubic magnetic
structures, this can arise from dipole-dipole interactions or from a
Dzialoshinski-Moriya interaction that is allowed (Keffer, 1962) in the type111 structure. The positive value of the exchange interaction constant in Eq.
(35) signifies an antiferromagnetic coupling.
The stability of magnetic structures of the face-centered lattices of spins
was discussed by Smart (1966) in terms of molecular field theory. For
collinear spin arrangements, there are four domains of stability: ferromagnetism, and three types of antiferromagnetism, labelled I, 11, and 111,
that are determined by the sign and relative magnitude of the first and second
neighbor exchange parameters J I , J 2 . The phase diagram for face-centered
cubic lattices and the type-111 magnetic structure are shown in Fig. 1.
Antiferromagnetism of type-I (AF-I) is stable for J I antiferromagnetic and
J2 ferromagnetic, whereas AF-111 is stable for both J I and J2 antiferromagnetic. Since the second neighbors exchange appears to be small in the
compounds with which we are concerned, the AF-I1 phase is not encountered. As we shall see, the observed correlations in DMS strongly resemble
those of AF-111, indicating that J2 is antiferromagnetic. In fact, the analysis
of single crystal experiments, discussed later, provides a quantitative
measurement of 5 2 .
Cubic P-MnS is one of the few materials that exhibits AF-I11 in the pure
state. Hastings et al. (1981) have shown experimentally that the magnetic
phase transition is of first order. Theoretical arguments given by Hastings
et al. suggest that the transition will always be first order in this structure.
The stability of various magnetic structures was also investigated by Ter
Haar and Lines (1962), who showed that the spin-wave frequency along the
tetragonal axis in reciprocal space would be zero in the absence of J 2 ,
implying that the ground state is unstable with respect to spin fluctuations.
In spin-wave theory for a type-I11 antiferromagnet, an antiferromagnetic J2
raises the spin-wave frequency along the tetragonal axis, whereas a ferromagnetic J2 gives imaginary frequencies. Spin-wave theory thus gives the
same stability criteria as molecular field theory.
The spin-wave dispersion relation for FCC AF structures takes the form
(ha)'
A+A
B'B,
(36)
with
A
(37)
2SJ(12q).
(38)
4.
137
FIG.1. (a) Phase diagram for magnetic order in the ( J I , J z )plane for a face-centered cubic
lattice. JI and JZ are first and second neighbor exchange constants, and a positive sign denotes
an antiferromagnetic exchange interaction. (b) Unit cell for antiferromagnetic order of type-I11
on an fcc lattice. Open and closed circles denote the opposing spin sublattices.
The functions J(kkq)are the Fourier transforms of the real space exchange
interactions defined by
and have to be worked out for the appropriate spin structure. For the AF-111
structure, with the y-axis, as the tetragonal axis we have
138
(40)
with
There are three antiferromagnetic domains corresponding to the tetragonal
axis being along x,y , or z directions in a macroscopic specimen. Thus the
observed neutron scattering from spin-wave excitations at (1, ),0) which
corresponds to q = 0 for one domain would have contributions from the
other two domains.
With this background in the structure and dynamics of FCC antiferromagnets, we proceed to discuss the properties of the disordered FCC lattice,
as described by Ching and Huber (1982a,b).
6 . Ching-Huber Modeis
By means of numerical simulation methods, Ching and Huber (1982a,b)
have investigated the dynamics of FCC lattices of spins, interacting only via
nearest neighbors isotropic Heisenberg exchange, for both the fully occupied
lattice and the lattice diluted with non-magnetic ions. For the fully occupied
lattice, (001) sheets of spins order antifferomagnetically under the influence
of JI. However, the spin direction varies randomly from layer to layer, so
that there is no three dimensional magnetic order. Each spin does experience
the same local molecular field, -4SJ1, from its neighbors in the same plane.
Presumably, because of the two-dimensional order, the excitations within
this structure are well defined in energy and wave vector, and follow a
dispersion relation that is-practically identical with that of a type-I11 AF. The
numerical calculation also gives a finite density of magnon modes at zero
energy. The presence of these modes indicates that the ground state of the
disordered AF lattice is unstable with respect to spin fluctuations.
In the diluted FCC array, there is a distribution of local molecular fields
about the mean value, and the latter shifts to lower energies according to the
approximate form for x > 0.5,
HM= -4SJ
+ 2SJ(1 - x).
(43)
4.
139
The calculations of the dynamics show that the excitations fall in energy
as the dilution increases, with an accompanying increase in lifetime. The
average spin-wave energy is nearly equal in magnitude to the frequency of
the type-I11 AF structure after taking the cubic average and scaling linearly
with concentration.
Ching and Huber (1984) have also calculated S(Q) numerically for diluted
FCC antiferromagnets with J2 = 0.1 J1 . They find that S(Q) is a maximum
at wave vectors like (1, i,0), which is an ordering wave vector of the type-I11
A F structure, and rapidly decreases with wave vector offset from (1, &, 0).
Close to the characteristic wave vector, t, they found that S(Q) could be
closely represented by the following analytical form
140
4. ELEMENTS
OF
THE
EXPERIMENTAL
TECHNIQUE
4.
141
information about sample homogeneity from the widths of the Bragg peaks
of Znl-xMn,Te. The observed counts must be corrected for multiple
scattering and self-shielding. Following Sears (1975a, b), the observed counts
C are related to the true cross section by
where Kis the instrumental normalization factor, H(k0, k') is the absorption
factor (= 10%) which is independent of angle for our experimental arrangement, and m is the multiple-scattering correction which is =1.6% for
Znl-xMnxTe samples.
With improved accuracy of the coherent and incoherent scattering cross
sections (Koester, 1977) due to the individual components of the alloy, the
analysis of the Bragg peak intensities permits the calculation of the concentration of Mn atoms in the samples as well as an effective Debye-Waller
factor and the normalization factor to put the observed diffuse scattering on
an absolute scale, The relative intensities of the three families of peaks in the
zinc-blende structure (h + k + 1 = 4n, 4n k 2, and 4n -t 1, where n is an
integer) are very sensitive to concentration, since the scattering lengths of zinc
and manganese have opposite signs, The accuracy in measurement of
concentration is in fact about +- 1 at.9'0. The effective Debye-Waller factor
permits an estimate to be made of the thermal diffuse scattering via Eq. (26).
The agreement between intensity calibrations at different temperatures was
k 5 % . The normalization factor was also measured with a cylindrical
vanadium sample of known diameter and irradiated length, correcting the
observed counts for self-shielding and multiple scattering (8.7%). Agreement
between the internal Bragg scattering normalization and with vanadium was
within 5%.
5. RESULTSOF EXPERIMENTS
The diffraction pattern for Zno.435Mno.565Te taken with neutrons of
wavelength 1.4064 A at 4.2 K (Holden et al., 1982) for scattering angles
between 5 and 75" is shown in Fig. 2. We see the expected pattern of Bragg
peaks for the zinc-blende structure together with A1 powder peaks from the
sample can, superposed on a diffuse background. The principal feature of
the diffuse scattering is a broad peak centered on 14.5", which may be
indexed as (1, i,0) in the zinc-blende structure. This peak is clearly much
wider than the experimental resolution in reciprocal space given by the width
of the Bragg peaks. Two other diffuse peaks are identifiable. The peak
near 23.3", which is partly overlaid by the adjacent (111) nuclear peak of
the zinc-blende structure, may be indexed as (1, *, 0) and the peak near
142
50001
i
I
v,
I-
4000
3000
0
IT
'r
w
'
2000
x = 0.565
T=42K
(Ill)'
A = I3863 6
(220)
II
.1
10
15
20
SCATTERING
25
30
35
40
ANGLE 2 8
30.5" may be indexed as (1, *, 2). Similar diffraction patterns were recorded
for Zno.6~4Mno.376Te, Zn0.406Mno.s94Te (Dolling el d., 1982),
and Zno.32Mno.68Te. For the x = 0.68 sample, there was evidence in the
diffraction pattern for small amounts of other phases, such as MnTez and
hexagonal MnTe.
The manganese concentration, Debye-Waller factor, lattice spacing and
intensity scaling factors, obtained by fitting to the integrated powder
diffraction intensities and angles are collected in Table I. The table also
contains estimates, derived from the (220) and (442) nuclear reflections of
Znl-,Mn,Te, of the homogeneity of the samples based on the widths of these
Bragg peaks. Excess width, above the experimental resolution, was assumed
to originate from a spread of lattice parameters in the sample. The homogeneity of the three samples with lowest c is excellent by this criterion.
However, there appear to be concentration fluctuations in the x = 0.68
sample, which may be connected with the appearance of impurity phases.
The temperature dependence of the diffuse scattering between 5 and 33"
for the concentrations x = 0.376,0.594, and 0.68 is shown in Figs. 3, 4, and
5 . In each case the dashed-dot lines give the nuclear diffuse scattering,
TABLE I
MANGANESE
CONCENTRATION,
TEMPERATURE,
LATTICEPARAMETERS,
DEBYE-WALLER
FACTOR,AND BRAGGPEAKDATAINDICATING
HOMOGENEITY
OF THE Zn,-,Mn,Te
SAMPLES
STUDIED.
37.6 k 0.3
59.4 f 0.2
59.3 k 0.3
68.0 & 0.7
56.7 f 0.05
56.3 k 0.6
56.5 k 0.6
77.6
240
4
300
4
300
17
6.176
6.236
6.230
6.263
6.218
6.234
6.218
0.0051 t O.OOO4
0.0131 t 0.0006
0.0047 f 0.0008
0.0156 t 0.0021
0.0035 f 0.0017
0.0162 t 0.0026
0.0051 0.0021
0.47 f 0.02
0.90
0.04
0.06 rt 0.04
0.0 t 0.15
0.45 k 0.02
0.53 k 0.02
0.92 f 0.04
0.95 f 0.04
0.04 0.05
0.097 f 0.20
0.03 t 0.06
0.048 f 0.050
_+
...
I600
1200 -
X =0.376
4.2 K
M=3
.
-. . '--0.5
000 400 -
- 1.0
/
-
--- - ----
- - - - - - - - - - - - - - - - - - - - - - - - - - - -_ -
&
0
u
3
I
ln
1200
*... ...
000-
0
V
400 -
1200 -
800 400-
77 K
I-
z
2
. .
..
- . , - ... _ :.
---- I
5
-----
I
10
_ . . . a
- .. - .
I-
r0.5 u
w
e.
--
-.
ul
ul
ul
- 1.0
300 K
....
.*
- . - - -... .
*
0.5
- - - - - - - - - - - - - - - - - --- - - - - - -_ _
I
15
20
I
25
I
30
-2
z
0
- - - - - - - - - - - - - - - - - - - - - - - - - -___
-- --------1
* -
-1.0
7;
50
cl
..800 -
.?
+-\
X *0.!594
M=I
400
- 2.0
4.2 K
-- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - I
0
In
.-
(200)
I
I
-2.0 7.
78 K
I-
(111)
.*
400-
.*
1.0
-- - - _- - - - - - - - _ _ - _ - - -- - - _ - -- - - - - - - - - - - - - - - - - - - 0.
I
I
400 -
.. - - .
~*
-.. .
-- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - 0-
I
10
-2.0
240 K
------v
I
15
I
20
1
25
I
30
-1.0
V
W
In
In
g
a
u
:
2000
z
3
1000
1000
78.0K
.... - _T1.0
. . - * - -
. - .'-<. .. . .\ - . ..
- 2.0
.-.. .
_ _ _ _ _ _ _ _ - - ------------ -------_
(2005 -
.-. .. j I . 0
- - - - --
I-
v)
4.
147
including both incoherent and thermal diffuse scattering, that accounts for
the results at high angles, where the magnetic scattering is negligible. As the
temperature is raised the diffuse peak near 14.5" weakens, becomes broader
and shifts to lower angles. The peak indicates the existence of magnetic
correlations between manganese spins that are strongest at low temperatures.
The correlations persist to high temperatures even for the x = 0.376 sample,
although they are not very long-ranged, since the 300 K data are ill-described
by a MnZ+form factor alone. The development of the spectrum as a function
of concentration and temperature can be quantitatively described by the
growth of correlations of a site with many shells of neighbors.
There does not appear to be a tendency for either clustering or superlattice
formation on the (Mn, Zn) sublattice. There are no sharp features persisting
to room temperature in this system, and in each case the spectrum evolves
smoothly to a high temperature state with short-ranged correlations. Little
chemical short-range ordering is expected, since MnZ+and Znz+ ions have
similar ionic sizes and the transition metal lattice is interpenetrated by the Te
lattice. Thermally driven short-range order would involve interchanging
transition metal and tellurium ions and this process is strongly inhibited by
the chemical bonding.
The question of long-range magnetic order, that is, a phase transition to
an ordered antiferromagnetic state, can be definitely settled for x = 0.594.
The width of the main magnetic diffuse peak certainly does not approach that
of a Bragg peak at the lowest temperatures. For x = 0.68, this conclusion is
not so clear cut. It may be possible, though unlikely, that a small fraction
of the sample shows long-range order, since the top 30% of the main peak
has the FWHM of a Bragg peak. This portion could be associated with the
concentration spread noted earlier for this sample. It is clear, however, that
the bulk of the sample does not show long-range order. The analysis of the
data supports this view.
The maximum cross section observed for all the samples is shown as a
function of temperature in Fig. 6. In every case there is a gradual increase
in the cross section as the temperature is decreased, with saturation below
20 K. By contrast, the static susceptibility of these alloys shows the cusp
behavior of spin-glasses, and the position of the cusp, Tg (McAlister et al.,
1984), is also shown in Fig. 6. There is no anomaly in the neutron scattering
at T g .
It is well-known that the cusp in the susceptibility, that is a characteristic
of spin-glass behavior, becomes rounded in an applied magnetic field. The
neutron cross section was therefore measured in a magnetic field above or
below Tg . The x = 0.594 sample was mounted in a superconducting magnet
capable of providing fields up to 4T in a direction normal to the horizontal
scattering plane. No magnetic field effects were observed either on field
148
T. M. GIEBULTOWICZ A N D T . M. HOLDEN
c
.-
n
Y
0.68
0.594
o x - 0.565
x x = 0.376
0
X=
o x=
TEMPERATURE
(K)
FIG.6. Maximum magnetic cross section obtained from the magnetic counts observed in the
vicinity of the ( 1 , i,0) diffuse peak for the four samples studied, as a function of temperature.
In each case the intensity increases smoothly with decreasing temperatures, and saturates below
10 K . Arrows indicate spin-glass temperatures for these samples. [After Holden et al. (1982).]
41
tl +I
?N
0 0
909
2
m
0909???df
0 0 0 0 0 0 0
tl ti +I +I +I +I +I
?99???09
e * m r 4 r 4 - 0
l 1 + 1
r 4 - - - - m - - - -
+I $1
tI +I +I +I t i +I +I +I
09dN
+I +I +I
0 0 0
?W?
W N O
tl
l
tl tl tl
-!?Z
+I +I
tl
09b.N"
+I +I +I +I
?P-1'4
9
m
W
--=
9 9 9 9 9 9 9 9 9 9
0 0 0 0 0 0 0 0 0 0
7%"
0 0 0
+I +I +I
099Y
m m o
???
- 0 0
+I +I
?'4?
0 0 0
1?09
r4-0
- m m
-00
+I +I +I t
I
0 0 0 0
232
W ? ? N
909N1
r4"O
+I +I +I tl
0 0 0 0
t ? N N
0 0 0 0
- 0 - 0
+I +I +I +I +I +I +I
0 0 0 0 0 0 0
F w N 0 0 0 0
??f????
?dP-d:Yb.W
0 0 0 0 0 0 0
v,
150
m,
4. NEUTRON
151
DIFFRACTION STUDIES
three shell parameters are in good agreement with those previously obtained.
The dashed curves in Figs. 3, 4, and 5 represent the best fit to this model.
The analysis of single crystal measurements of Cdl-xMnxTe, discussed later
in this chapter, suggests that the best description of the decrease of spin
correlations with distance is provided by an anisotropic Lorentzian function.
It may be shown that the functional form of the wave vector dependence of
the intensity in a powder experiment is altered if an anisotropic Lorentzian
function is used instead of an isotropic Lorentzian. This may well be the
reason why the isotropic Lorentzian gave an unsatisfactory description of the
powder data.
TABLE I11
MANGANESE
CONCENTRATION,
TEMPERATURE,
SPINCORRELATIONS
AND INVERSE CORRELATION
FROM FITTING
Znl-,Mn,TE NEUTRON
DIFFRACTION
DATATO EQ. (22). x
LENGTHS
OBTAINED
INDICATESTHE QUALITY
OF THE FITS.
X
(at. '70)
37.6
56.5
59.4
(K)
4.2
4.2
15
25
35
45
55
77
300
4.2
25
50
68.0
78
240
4.2
40
78
170
300
Ai - B I
A2 - B2
-5.OfO.l
-3.3 *O.l
-3.4*0.1
-3.4 f 0.1
- 3.2 0.1
-3.2 f 0.1
- 2.8 i0.1
-2.5 f 0.1
- 1.8 i 0.1
- 2.6 f 0.1
-2.8 f 0.1
-2.6 f 0.1
- 1.6 f 0.1
-0.30 i 0.1
- 2.8 f 0.2
-0.4 f 0.1
A3 - B3
KlaNN
2.0 0.2
1.7 +- 0.1
1.8 i 0.1
1.6 f 0.1
1.5 0.1
1.3 0.1
1.0 f 0.1
2.1 f 0.2
-0.3 C 0.1
1.6 i 0.1
1.5 f 0.1
1.1 i 0.1
0.7 f 0.1
0.2 f 0.1
1.4 f 0.3
3.4 f 0.1
4.6 f 0.2
4.5 i 0.2
4.1 f 0.2
3.6 f 0.2
2.9 f 0.2
2.6 f 0.2
1.05 f 0.04
0.3 f 0.2
3.6 i 0.2
3.3 0.2
2.0 f 0.1
1.8 f 0.2
0.6 f 0.1
5.2 0.2
0.62 f 0.02
0.48 f 0.01
0.49 f 0.01
0.54 i 0.01
0.64 f 0.02
0.74 f 0.02
0.86 rt 0.03
0.9
1.7 f 0.2
0.53 i 0.01
0.56 f 0.01
0.86 i 0.02
1.23 f 0.04
7f9
0.38 f 0.01
1.2
1.3
1.1
1.2
0.1 f 0.1
0.1 f 0.1
1.9 f 0.2
1.4 f 0.1
0.7 f 0.1
*
*
1.5 5 0.1
0.9 i 0.1
3.6 f 0.2
0.68 i 0.02
1.15 f 0.04
1.8 f 0.2
0.3
*2
1.1
1.o
1.1
0.7
1.O
1.5
1.3
0.8
0.7
3.4
2.5
1.9
1.4
0.9
152
0.4 -o
0.2-
0.o
0
10
20
30
40
50
60
TEMPERATURE ( K )
70
00
90
FIG. 7. The inverse correlation range ( K I U N N ) , deduced by fitting the data to a model with
exponential fall-off of the ideal correlations with the 4th-336th shell. The solid line is a guide
to the eye. The numbers labelling various symbols denote the Mn mole fractions.
concentration shows that the inverse correlation length can be broken into
a part determined by geometry (and, in our case, exchange interactions) and
a part governed by temperature alone, as follows:
Kill"
= (K/UNN)G
+ (K/UNN)T,
4.
153
t,
0
X
ox
I
0.1
0.2
0.3
I
0.6
CONCENTRATION X
I
0.4
0.5
I
0.7
I
0.8
I
0.9
1.0
FIG.8. Comparison of the zero temperature inverse correlation lengths squared, calculated
for Znl-,Mn,Te with JZ = 0.1 J I (Ching and Huber, 1983), with the experimental result for
powder sample. There are expected to be two inverse correlation lengths, corresponding to
coupling between planes and within planes, and the experimental result on a powder is an
average over both.
154
0.9 -
0.8 -
0.70.6
0 0.68
0.594
0 0.565
0.5(K,Q")f
0.4-
0.3-
0.20.1
TEMPERATURE ( K
FIG. 9. The temperature dependent portion (Kl(INN)T of the inverse correlation length for
x = 0.68, 0.594, 0.565 which lies on a common curve. The solid curve is the result for a
Heisenberg linear chain with J I = 11 K, and is not in close agreement with experiment.
Similar results have been obtained for ferromagnetic CrFe (Burke et al.,
crystals (Cowley et al., 1980).
However, in both these cases the experiments were carried out near the
percolation concentration, where the clusters are ramified, i.e., major
clusters are joined by linear pathways of adjacent sites occupied by
manganese ions. The temperature dependence of the inverse correlation
length for Znl -,Mn,F2 was successfully modelled as a Heisenberg linear
chain. Here we are far from the percolation threshold, and the one-dimensional linear chain model is not very plausible. In any case, the slope of the
variation of KlaNN with temperature is steeper (Fig. 9) than the variation of
KIaNN calculated for a Heisenberg linear chain with classical spins (Fisher,
1964), making use of J I = 11 K deduced from the high temperature static
susceptibility (McAlister et al., 1984).
1983) and antiferromagnetic Zn1-,Mn,F2
4.
155
6. SAMPLES
AND ELEMENTS
OF EXPERIMENTAL.
TECHNIQUE
Because of the extremely high neutron absorption cross section of
cadmium (ou = 2650 barns for I = 1.08 A, see Bacon, 1975), in practice
only single crystal diffraction studies are possible in the case of Cdl -*Mn,Te,
the powder signals being too weak. Fortunately, this material can be obtained
in single crystal form in the whole zinc-blende composition range ( x s 0.70).
The earliest diffraction data for this alloy were obtained using very thin
crystal plates (-0.1 cm) for x = 0.60-0.70. However, the increase of
absorption with increased Cd content did not permit extending these studies
to lower x values. Fortunately, samples prepared from the low-absorbing
Il4Cd isotope created new opportunities for neutron scattering studies, and
three such samples (x = 0.65, 0.44, and 0.35) have been investigated in
various diffraction experiments. The largest of these specimens, a
Cd0.3&ln0.6sTe crystal with a volume of 1.5 cm3, also made it possible to
study inelastic magnetic scattering in a DMS system for the first time.
In the above experiments, the crystal axis [OOl] was set vertical, because
this orientation permitted observation of the largest number of type-I11
antiferromagnetic superstructure reflections (i.e., the type of ordering seen
in Cdl-xMnxTe). The inelastic studies and most of the diffraction measurements were carried out with crystal spectrometers (see Brockhouse, 1961) in
either the triple-axis mode of operation, or the diffractometer mode, after
the analyzing crystal was removed. With a diffractometer, a given
momentum transfer vector Q may be selected by setting both the angle that
the crystal makes with the incident beam and the angle of scattering. The
magnetic component of the observed intensity provides in this case a direct
measure of the magnetic structure factor S(Q) (see Eqs. (10) and (1 1)). Using
the triple axis configuration, one can observe inelastic scattering for a chosen
momentum transfer QO and energy transfer 0 0 . The intensity of magnetic
156
y,
4.
157
158
FIG.10. The contour map of the scattered neutron intensity for a sector of the (001) plane
of the reciprocal space for Cdo.56Mno.wTeat (a) 4.2 K, (b) 78 K, and (c) 295 K. The Bragg nuclear
reflections are shown by black dots, and the crossed circles show the ideal positions of the typeI11 superstructure reflections. The plots were obtained by multiple scanning parallel to the [OlOl
axis within the trapezoidal area marked as ABCD in (a); the upper part of the plot is the
mirror image with respect to the line BC. Plot (b) shows that magnetic maxima begin to form
at 78 K. As can be seen in (c), no maxima occur at 295 K. The shape of the equal intensity lines,
however, suggest that very weak residual magnetic correlations persist even at this temperature.
[After Giebultowicz et al. (1986b).]
159
160
2.8
2.9
3.0
3.1
3.2
WAVEVECTOR COMPONENT Q y
FIG.11. Diffuse magnetic scattering in 114Cdo.,sMno.~sTe
single crystal at 6 K. The plots show
the results of radial (or 0 - 28) scans through points in the reciprocal space with coordinates
(4,1, O), and ($, 3 , O ) . The shaded contour is a nuclear Bragg reflection shown for comparison.
[After Giebultowicz et al. (1982).]
o (3/2,I,<)scan
(A 1
( 312, I
,Oo0t
+ <,O)scan
A (3/2+<,I,O)scan
2000 -
o ( 3/2,l,( )scan
(8)
(312, I + <,O)scan
Q - VECTOR
COMPONENT,
FIG. 12. Scans through reflection points (i,1,O) along three cubic axes for Cdl-,MnxTe single crystals, with (a) x = 0.65 and (b)
0.44 at 4.2 K. The ellipses in the figures show the resolution function width (FWHM) within the experimental plane (with the scale on
the x and y axes the same as on the horizontal axes in the main plots). The vertical resolution width (i.e., along the z axis) was in each
case slightly larger than the longer ellipse axis. The shaded contours show the (200) nuclear reflections for comparison. Note that the
horizontal scales are different in both parts of the figure. [After Giebultowicz ef al. (1986a).]
162
4.
163
Scans carried out over such points show convincingly that there is no
detectable type-I component in Cdl-,Mn,Te.
164
165
166
tJ
w
K
TEMPERATURE ( K 1
FIG. 13. Temperature variation of the correlation length if! = KC in Cdo.&In0.65Te,
deduced from the half-width (FWHM) of the magnetic diffuse (1, +,0) maximum, and of the
integrated intensity of this maximum. [After Steigenberger el a/. (1986).]
4.
167
168
+,
4.
169
As noticed in Sec. 7e, the absolute values of K and the details of the
lineshapes obtained by the method of Ching and Huber show some discrepancies from the observed data. These differences may be caused by
finite size effects in the simulations: because of the finite dimensions of
the supercell used (i.e., 12a x 12a x 12a), the model data reflect primarily
the character of the correlations between relatively close spin pairs. On the
other hand, in the case of the measured results, the contribution of more
distant neighbors appears to be significant. For example, as can be estimated
from Eq. (32a), for Ornstein-Zernicke damping the pairs with RI IK - ~give
rise to only -26% of the peak intensity. For Cdo.3sMno.asTe, however,
K ; ~= 9a. Secondly, a description in terms of K independent of the distance
between the neighbors is certainly an approximation. The fact that the results
obtained for Znl-,Mn,Te (see Sec. 5) are in a better agreement with the
model simulations seems to support these arguments. A comparison with the
theory has been made in that case on the basis of correlation coefficients for
the few first shells of neighbors (determined directly from powder measurements), that better corresponds to the situation described by the model.
In view of these shortcomings of the model, the accuracy of the determination of JZ may be questioned. However, test calculations carried out for
systems of various sizes suggest that the ratio K ~ / K I isI decidedly less
sensitive to finite size effects than the absolute value of K (Ching and Huber,
1985).
In simulations carried out by Ching and Huber (1983) for the model
systems with 0.40 Ix I0.80, the short-range character of the spin ordering
persisted up to x = 0.80. The variation of K in this range of x does not show
any anomaly, suggesting the existence of two different phases, in agreement
with the conclusions drawn from diffraction data for Cdl-,Mn,Te.
Similarly, Monte Carlo simulations do not indicate a phase boundary as
proposed by Galazka et al. (1980). Studies done by the latter method for
Cdl -,MnxTe (using a supercell of 4000spins, with JZ = 0.1J1) in the range
0.70 I x I 1 showed only a short-range type-I11 ordering for concentrations
lower than x = 0.85 (Giebultowicz, 1986). Above this concentration, the
systems exhibited a long-range (non-collinear) order at low temperatures.
Since in this case the scale of long-range order is necessarily the scale of the
supercell, what the model suggests is that a phase boundary may occur even
at x higher than 0.85.
g . Variation of Magnetic Scattering with Temperature
170
4.
171
I
I
1
Cd1-x MnXTe (312 10 )reflection
8 0 X 50.70
0 0 X = 0.65
o X=0.63
T (K)
FIG. 14. The variation of the integrated intensity of the magnetic diffuse (1, +,0) maximum
(related to the intensity of the (200) nuclear Bragg reflection) for Cdl-,MnxTe single crystals
with x = 0.60, 0.63, 0.65, and 0.70. [After Giebultowicz ef al. (1981).]
172
T. M. GIEBULTOWICZ A N D T. M. HOLDEN
4.
173
8. INELASTICNEUTRON
SCATTERING
STUDIES
OF
Cd0.3sMn0.65Te
174
800 In
0
0
0
(D
I-
600-
400 -
I.
0
I
1
I
2
I
3
I
5
I
6
FIG. 15. Inelastic neutron scattering spectra for Cdo.~jMn0.6jTeat T = 1.8 K for two
different wavevectors. The lines are guides for the eye. [After Giebultowicz el at. (1984).]
figure, the excitation energy increases away from the superstructure point
Q = (2n/a)(l, *, 0) in both directions, but the growth is faster for the inplane direction [ 1001. Such behavior, reflecting the features of spin-wave
dispersion relations in type-Ill structure (see Sec. 3a, Eqs. (36)-(42)),
confirms that the excitations in the spin system in Cdo.3sMno.ssTe have the
character of propagating magnon modes.
Unusually large linewidths of the inelastic spectra observed in
Cdo.35Mno.ssTe indicate at the same time that the magnon excitations are
strongly damped. The experimental data cannot therefore be interpreted
quantitatively in terms of Eqs. (36)-(42), which apply to the case of
undamped spin-waves. For the purposes of quantitative analysis one can,
however, utilize the model of spin dynamics in diluted FCC antiferromagnets
developed by Chiang and Huber (see Sec. 3b), which enables one to calculate
the dynamic structure factor S(Q, a),
Eqs. (5)-(6). As we note in Sec. 6, the
theoretical dynamic structure factor can be directly compared with the
inelastic neutron scattering data obtained by the variable incident energy
method. Problems such as excitation damping and contributions of different
domain types (see the comment in Sec. 3a) are inherently solved by this
modelling technique.
4.
175
112.0 1
( I,I
Wavevector
FIG. 16. Results of model calculation (a) and measurements (b) of the inelastic magnetic
scattering in Cd0.35Mno.65Te for various momentum transfers. The circles and the squares
indicate, respectively, the calculated and measured positions of the spectral maximum, and the
bars in both plots show the spectrum width at half maximum (not the standard deviation!). The
energy of the modes observed in Raman experiments (see text) is shown in (b) by the arrow.
[After Giebultowicz et al. (1986a).]
The initial step in the modelling method of Ching and Huber is a numerical
simulation of the ground state spin configuration, which is not known
apriori. It must be checked subsequently that the results agree with the actual
spin arrangement in the material. In the case of Cdl-,Mn,Te, a comparison
with diffraction data indeed confirms that the numerical procedure
reproduces satisfactorily the short-range antiferromagnetic order seen in this
compound (see Sec. 7f). Results of the calculation of S(Q, o)for a model
of Cd0.3sMn0.65Te(i.e., for x = 0.65 and JZ = 0.1J1)displayed in Fig. 16a
show that the theory describes the measured dynamic characteristics quite
well. The fit of the model values to the measured data indicates that JI (the
only free parameter in the calculations) in Cdl-,Mn,Te is (14 f 1) K.
The results presented in this section have not been corrected for twinning
effects in the sample (see Sec. 6 ) . In general, the scattering vector Q for
a twinned grain is different from the remainder of the crystal, so that the
scattering contributions from such grains may lead to some distortion of
the measured spectra. It is primarily for this reason that the simplified
176
method of data analysis (concerning only the widths and positions of the
maxima, but not the exact profiles) has been applied. In order to carry out
a detailed analysis of the inelastic lineshapes, in a more recent experiment
(Giebultowicz et al., 1986c), the scans and calculations were done for the Qpoints lying along the [120] and [ l i O ] directions in the (001) plane. As
explained in Sec. 6, the scattering vectors for the two twinned components
are equivalent in these two special cases, making corrections for the twinning
unnecessary. Examples of fits of the calculated S(Q, w ) curves to the
measured data (including corrections for the instrumental effects and for the
background) are displayed in Fig. 17. As clearly illustrated by these plots,
only for Q-vectors in the vicinity of the superstructure point ( 2 n / a ) ( l , 0)
is there a certain discrepancy between the measured and the calculated
4,
FIG. 17. Comparison of the experimental and theoretical inelastic magnetic neutron
scattering lineshapes for Cdo.&n0,65Te for several Q-vectors. The circles show the intensities
measured using a variable incident energy spectrometer operation mode vs. a fixed monitor
count rate. The solid lines show the fits of the theoretical lineshapes (convoluted with the
instrumental resolution) to these data. The values of the nearest-neighbor exchange constant
JI (the only adjustable parameter) obtained from each fit are displayed on the plots.
[Giebultowicz e l al. (1986c).]
4.
177
lineshapes. At all other investigated points the model of Ching and Huber
provides an excellent description of the experimental data. The values of the
exchange constant J1obtained from these fits are very close in each case. The
average result for all runs, JI = (13.8 f 0.2) K, is in very good agreement
with the value J1 = (13.8 k 0.3) K determined by Lewicki et al. (1986) (see
also Spalek et al., 1986), from the high-temperature magnetic susceptibility,
and close to the value J1 = (12.6 & 0.6) K reported by Larson et al. (1986)
for very dilute systems ( x < 0.05). The lack of quantitative agreement for
Q = (27c/a)(l, 3,O)is probably due to some shortcomings of the model, such
as the size of the model supercell being too small.
It can be concluded that the excitation mode observed in neutron diffraction is probably the same as that seen in Raman measurements. The
measured energies are in fair agreement in both cases. The point in Q-space,
Q = (2n/a)(l, *,O), at which the mode is observed in the neutron case is
consistent with the interpretation of the Raman data in terms of a onemagnon process at the zone center proposed by Ramdas (1982). Thus the
neutron scattering data do not confirm the interpretation in terms of a twomagnon Raman process (Grynberg and Picquart, 1981), primarily by
showing that there are no excitations in Cd0.35Mn0.6sTe with energies
sufficiently low to be consistent with such a description.
178
4.
179
the two alternative explanations of the origin of the magnon mode observed
in Raman scattering studies in Cdl-,Mn,Te is more likely. All these successes
of the Ching-Huber model tempted its authors to apply their approach to
the case of the heat capacity calculations (Ching and Huber, 1984), again
with good results.
To summarize the results of the neutron scattering investigations in
Cdl-,Mn,Te and Znl-,Mn,Te, let us repeat that both materials exhibit the
same short-range antiferromagnetic ordering of type-111, at least for
0.35 C x < 0.7. The temperature behavior of the magnetic scattering
observed in all investigated samples shows features typical of spin-glasses,
thus providing evidence against the spin-glass antiferromagnet phase
boundary in the vicinity of x = 0.6. Unfortunately, the existing neutron
scattering data do not give us much insight into the mechanism of the spinglass transition itself. This is because of the fact that the methods employed
so far were sensitive to the spin-spin correlations in space, while it is the
temporal behavior of the correlations that is of importance in the context of
spin-glasses. Methods involving high energy resolution, such as the time-offlight method and spin-echo spectrometry (see Mezei, 1982), are very
promising in this context. Experiments of this kind should be undertaken on
DMS in order to extend our understanding of the nature of the various
clustering and spin-glass effects that take place in these materials.
References
Aggarwal, R. L., Jasperson, S. N., Becla, P., andGalazka, R. R. (1985). Phys. Rev. 832,5132.
Alexander, S . , and Pincus, P . (1980). J. Phys. A 13, 263.
Als-Nielsen, J. (1976). In Phase Transitions and Critical Phenomena, 5A (C. Domb and
S. M. Green, eds.), Academic Press, London.
Anderson, P. W. (1053). Phys. Rev. 79, 705.
Bacon, G. E. (1975). Neutron Diffraction. Clarendon Press, Oxford.
Balzarotti, A., Motta, T., Kisiel, A., Zimnal-Starnawska, M., Czyiyk, M. T., and Podgorny,
M . (1975). Phys. Rev. B 31, 7526.
Binder, K., Kinzel, W., and Stauffer, D. (1979). Z . Phys. B 36, 161.
Binder, K., and Schroder, K. (1976). Phys. Rev. B 14, 2142.
Brockhouse, B. N. (1961). In Inelastic Scattering of Neutrons from Solids and Liquids,
Vol. 1 (IAEA, Vienna, 1961). p. 147.
Burke, S. K., Cywinski, R., Davis, J. R., and Rainford, B. D. (1983). J. Phys. F: Met. Phys.
13, 451.
Brockhouse, B. N., Becka, L. N., Rao, K. R., Woods, A. D.(1963). In Second Symposium
on Inelastic Scattering of Neutrons from Solids and Liquids, Vol. 11 (IAEA, Vienna,
1963), p. 23.
Ching, W. Y., and Huber, D. L. (1981). J. Appl. Phys. 52, 1715.
Ching, W. Y., and Huber, D. L. (1982a). Phys. Rev. B 25, 5761.
Ching, W. Y., and Huber, D. L. (1982b). Phys. Rev. B 26, 6164.
Ching, W. Y., and Huber, D. L. (1983). Unpublished.
Ching, W. Y., and Huber, D. L. (1984). Phys. Rev. B 30, 179.
180
4.
181
CHAPTER 5
I. INTRODUCTION
. . . . . . . . . . . . . . . . . . . .
INTERACTION BETWEEN
MOBILES OR p CARRIERS AND
11. EXCHANGE
LOCALIZED
3d OR 4f ELECTRONS
IN SEMICONDUCTORS.
....
1. Interaction Hamiltonian . . . . . . . . . . . . . .
2 . Exchange Constants . . . . . . . . . . . . . . . .
111. BANDSTRUCTURE
OF NARROW-GAPDILUTED
MAGNETICSEMICONDUCTORS IN QUANTIZING MAGNETIC
FIELDS
. . . . . . .
IV. TRANSPORT
MEASUREMENTS
IN THE QUANTUM
REGIME:
BAND STRUCTURE MODEL. . .
3. Hgl-,MnxTe and Hgl -, M n, Se . . . . . . . . .
4. (Cdl-,Mn,),As2
. . . . . . . . . . . . . . .
5. Pbl-xMn,Te and Pbl -, M n, S.
. . . . . . . . .
V. TWO-DIMENSIONAL
ELECTRON
GASIN Hgl-,Mn,Te AND
Hgl-,,Cd,MnYTe
. . . . . . . . . . . . . . .
6. Experiments . . . . . . . . . . . . . . . .
7. Energy Levels of 2d Electron Gas. . . . . . . .
8. Future Possibilities . . . . . . . . . . . . .
REFERENCES
. . . . . . . . . . . . . . . . . .
CONFIRMATION OF THE
. . .
. . .
. . .
. .
. . .
. . .
. . .
183
185
185
186
190
204
204
214
216
211
217
219
224
225
I. Introduction
In this chapter, we deal with narrow-gap diluted magnetic semiconductors
184
J . KOSSUT
5.
185
1. INTERACTION
HAMILTONIAN
The exchange interaction between mobile, s-like band electrons and
localized electrons from partially filled 3d (or 4f) shells was first considered
in transition (or rare-earth) metals. Early ideas of Zener (1951a-c) and
Vonsovskii and Turov (1953) led to the development of the so-called s-d (or
s-f) model. The case of rare-earth metals was particularly stimulating in this
development, since magnetic properties of these materials could hardly be
explained in terms of direct exchange interaction involving the overlap of 4f
wave functions localized on neighboring rare-earth atoms. The localization
of these wave functions is so strong (thus their overlap so small) that the
estimated values of the direct exchange constants are far too small to
account, e.g., for magnetic transition temperatures observed. Therefore
indirect coupling between the localized magnetic moments of rare-earth ions
via conduction electrons was invoked in order to explain the observed
magnetic behavior. Several authors dealt with the problem of the form of this
coupling (see, e.g., Vonsovskii and Turov, 1953; Kasuya, 1955; Abrahams,
1955; Yosida, 1957; Mitchell, 1957). It was found that the exchange interaction between the conduction electrons and 3d or 4f electrons (i.e., those
responsible for the formation of localized magnetic moments) can be
represented by a simple Heisenberg-like formula
186
J. KOSSUT
The second assumption made when deriving Eq. (1) concerns the validity
of the description of the localized moments in terms of the spin eigenvalues.
Such description often requires the use of a total angular momentum J,
instead of the spin only. It was shown by Liu (1961) that even in such a
case a simple form involving the scalar product can be retained if S, in
Eq. (1) is replaced by J, and the coupling coefficient J(r - R,) is slightly
redefined:
All expressions quoted so far are isotropic. It can be shown that this is only
an approximation, valid if spatial extensions of 3d (4f) wave functions are
small compared to the de Brolie wavelength of mobile s- or p-electrons. Only
then the mobile carriers do not feel an anisotropic character of 3d (40
functions. This condition is sometimes not easy to meet in metals. Therefore,
a generalization of Eq. (1) was found by Kaplan and Lyons (1963) that is
explicitly anisotropic. In semiconductors, however, the mobile carriers which
usually play a role, are characterized by much smaller values of momenta k
than in metals; that is, they have longer de Brolie wavelengths. This fact
justifies the use of the isotropic expression (1) in our further considerations
of diluted magnetic semiconductors.
2. EXCHANGE
CONSTANTS
We shall now turn our attention to the nonlocal potential J(r - Rn)
appearing in Eq. (1). All papers quoted in the preceding section defined
rather the Fourier transform of J(r - Rn) than this function itself. This
transform can be written as
J(k, k)
= -2
.r
q&(r)+;(r)V(r
- rr)~,lk(r)+d(r)d3rrd3r
(4)
5.
187
where an(k) are coefficients of the combination. The sum in Eq. (5) can often
be restricted to only a few terms. This is particularly true in the case of
narrow-gap semiconductors. Then the coefficients an(k) can be found exactly
together with their explicit dependence on the wave vector k. Inserting wave
functions ( 5 ) into (4), we obtain
(6)
As we can see a part of the k, k dependence is already taken out from the
integral. The remaining k-dependence is of lesser importance. As we
188
J . KOSSUT
(7)
where ur6 = S&/S(r) and tirs = Qo/X(r)while SZO stands for the volume
of a unit cell. The elements as ( S I J I S ) , etc., commonly used in the literature,
represent symbolically the integration indicated by Eq. (7) with appropriate
5.
189
190
J . KOSSUT
between the models. One is, then, still in a situation in which the relevant
exchange constants have to be treated as phenomenological quantities, with
values treated as adjustable parameters of the given model of a diluted
magnetic semiconductor.
The majority of papers dealing with narrow-gap diluted magnetic semiconductors is devoted to alloys based on 11-VI mercury compounds, e.g.,
Hgl-,Mn,Te or Hgl-,Mn,Se. They were also the first studied experimentally
in an extensive manner. Therefore, we describe here the band structure of
these materials. We shall discuss briefly the band structure of tetragonal
(Cdl-,Mn,)3Asz and DMS based on lead chalcogenides (or rather only the
modifications of the general model associated with their different crystallographic symmetries), when we describe in the following sections the
experimental results obtained for these materials.
Early investigations (Morrissy, 1973; Delves, 1963and 1966)indicated that
semiconducting properties of Hgl -,Mn,Te cannot be entirely understood in
terms of existing theoretical models that could be successfully applied to such
semiconducting mixed crystals as Hgl -,Cd,Te. The unexplained features of
Hgl -,Mn,Te were a non-monotonic temperature behavior of the Shubnikovde Haas oscillation amplitude (Morrissy, 1973) and the value of the spinsplitting of electrons and its temperature dependence observed in magnetooptical study of Bastard et al. (1978) (see also Rigaux, this volume). On the
other hand, when no strong magnetic field was applied, the band structure
of Hgl-,MnxTe appeared to be qualitatively the same as in Hgl-,Cd,Te, as
indicated in the study of Stankiewicz et al. (1975). As in other narrow-gap
materials with the zinc blende lattice symmetry, the minimum of the conduction band and the maximum of the valence band occur at the central point
r of the Brillouin zone. There are three bands with similar energies (see
Fig. 1): a fourfold degenerate band of TSsymmetry, and doubly degenerate
r6 and r7 symmetry bands. The sequence of these bands varies with the crystal
composition. For low values of Mn mole fractionx, we deal with the so-called
inverted band ordering (or symmetry-induced zero-gap configuration)
depicted in Fig. la. For increasing values of x, r6 valence band shifts
upwards, and at x = 0.065 (at 4.2 K), it coincides with the position of the TS
band (Fig. 1b). Above this semimetal-semiconductor transition point, r6
band starts to form the conduction band (Fig. lc) while the TSband becomes
the light and heavy hole band. The energy gap Eg defined as the energy
difference between TSand r6 levels is then positive, and we are dealing with
5.
191
x
A
FIG.1 . The energy level scheme in the vicinity of the conduction and valence band edges in
narrow-gap semiconducting mixed crystals with zinc blende structure: (a) inverted (zero-gap)
band structure; (b) semimetal-semiconductor transition point; (c) open-gap structure.
the standard semiconducting band ordering. The variation of the energy gap
with the crystal composition in Hgl-,Mn,Te is shown in Fig. 2. The r7
symmetry band (or spin-orbit split-off band) remains a deeper valence
band.
It is well known from the work of Kane (1957) that the proper description
of a narrow-gap semiconductor requires that at least the three bands r6, r7
and l-8 , because of their energetic proximity, are considered at the same level
of accuracy in a k p treatment of the band structure. The remaining more
remote bands, often referred to as higher bands, may be treated in an
approximative way. Their influence on the shape of the conduction and
valence bands is usually evaluated with accuracy to terms proportional to k2,
where k is the electronic wave vector. Only in the case of HgSe, and of mixed
crystals similar in composition to HgSe, does one have to include in the set
of closely spaced (quasi-degenerate) bands also the nearest higher band of
r15 (no spin notation) symmetry, as found by Mycielski et al. (1982). This
fact is connected with the relation EE = 2/3A, where A is the spin-orbit
splitting (or the distance between TS and r7 bands), that is fulfilled in HgSe.
However, a very steep slope of the Eg vs. x relationship in Hgl-xMn,Se
(Dobrowolska et al., 1981) limits this anomalous region to x s 0.0005.
Also, the modifications due to such enlargement of the quasi-degenerate
set of bands concern mainly the r6 band (light holes) and therefore are
possibly important in an analysis of the interband magnetoabsorption
192
J. KOSSUT
500
400
300
-> 2 0 0
*
X
0 1 00
0.05
0.10
0.15
0.20
0
un
-100
-200
ml
54
?
I
-3 00
X
experiments. In this chapter, which is focused mainly on the magnetotransport phenomena in n-type DMS, we shall neglect this complication in
further considerations.
There exist in the literature on narrow-gap semiconductors the simplified
models of their band structure in a presence of a strong magnetic field. In
particular, the so-called quasi-Ge model of Luttinger (1 956) includes into the
quasi-degenerate set of states only the Tg band. The three-band model
(e.g., Kacman and Zawadzki, 1971) although considering in detail all
closely positioned r6, l-7 and TSbands, neglects totally the influence of the
higher bands. The advantage of these models lies in the fact that they can
be solved analytically. However, their applicability is only of limited range.
193
194
J . KOSSUT
c = -h2,
mo
s=-
eH
hc '
Eg + 2CsF(N + 9)
+ Cs(N + 1) + (F + +)ck:
iP&a
5.
195
D, =
i
- -Pk,
43
- C a y 3 k,a
- 3 N y 3 kza+
The operators a and a+ in Eqs. (14)-(16) are the harmonic oscillator lowering
and raising operators, respectively, defined as
196
J . KOSSUT
and
N = aa.
Fn
(LyLz)-2exp[i(kyy
+ k,z)]
197
where
DL =
198
J . KOSSUT
DL =
where ST = SJ?f iSj and, as in Part 11, the only nonvanishing and
independent exchange constants are defined as
Here QOis the unit cell volume, and NO = V / Q , is the number of cation
sublattice sites.
The matrix (22) is of rather complex form. The problem is further
complicated by the fact that after addition of (22) to (1 l), its solutions can
no longer be sought as a simple vector-column of harmonic oscillator
eigenfunctions. The matrix (22) can be, however, considerably simplified
when an approximation analogous to the mean field approximation from the
physics of magnetism is invoked. Namely, the wave functions of conduction
and valence band electrons are spread over the whole volume of the crystal.
Therefore, one may say that these electrons interact with all magnetic
moments present in a sample, thus feeling in effect the average value of
the localized moments. One may thus say that the band electrons are subject
to a certain average or resultant field of the localized moments. It is therefore
permissible to carry out an averaging procedure over all possible states of Mn
moments already at the level of the Hamiltonian matrix. For a paramagnetic
system only the S, component of the localized spin does not vanish after the
averaging procedure. This very major approximation was confirmed
a posteriori by Gaj et af. (1979) who determined experimentally a direct
proportionality between spin splittings and magnetization that is proportional to (Sz>x(note that C,(S~>/NO
= x(S,>).
After this averaging procedure the entire off-diagonal blocks DL in Eq. (22)
vanish and one is left with two nearly diagonal 4 x 4 blocks DL and Di,. The
matrix resulting from the addition of Eq. (1 1) and Eq. (22), with DL blocks
5.
199
200
J. KOSSUT
-3
-3
-2
-2
-I
-I
201
real energy gap opens. Therefore, it may happen that the band gap in DMS
opens only at elevated temperatures.
The magnitude of the band overlap is a result of an interplay between this
exchange interaction-induced shift and the movement of the lowest conduction band that occurs also in the upward direction on the energetic scale. For
very small values of Eg(close to the semimetal-semiconductor transition), the
conduction band levels (due to the smallness of their effective mass) move
up faster than the top of the valence band, and the overlap may not occur
even at very low temperature. Also, for sufficiently high magnetic fields the
shift of the top of the valence band reverses its sign and the level beings to
move downwards, as it does in standard narrow-gap semiconductors. Then
the energy gap induced by the field starts to open.
This strange behavior of the top of the valence band may be qualitatively
understood when one bears in mind how ( S , ) depends on the temperature
and magnetic field. Let us assume for a moment that we are dealing with a
perfect paramagnet where the magnetic moments do not interact with each
other. Then ( S , ) varies as 1/T, so that at higher temperatures the additional
terms proportional to (S,) may become negligible compared to normal
terms brought by Eq. (11). The magnetic field tends to align the magnetic
moments along its direction. When all the moments are already aligned a
saturation is reached and ( S , ) = - 512. The normal terms in Eq. ( l l ) , on
the other hand, are directly proportional to the field. A value of the magnetic
field may thus be reached when the contribution of these normal terms wins
over the exchange induced contribution. For the top of the valence band such
a return to normalcy occurs if the following condition is met:
The above general remarks remain true even if the interaction between the
localized moments becomes appreciable and ( S , ) cannot be described by a
simple Brillouin function applicable to non-interacting spin systems.
Shape of the Heavy Hole Landau Levels
202
J. KOSSUT
40
->
-E
30
20
10
T=5 K
B(T)
(a)
FIG.4. The energies of the minima of the conduction subbands ( k =~0) in Hg,-,Mn,Te
(x = 0.1). calculated at (a) T = 5 K, (b) T = 15 K and (c) without the contribution of the
01
5.
203
40
3c
2c
10
T = 15 K
2
B(T)
(c)
FIG.4 (continued)
204
J . KOSSUT
AND
Hgl-,MnxSe
5.
205
Further in this section we shall try t o describe the most striking differences
between the quantum transport phenomena observed in diluted magnetic and
nonmagnetic semiconductors, and we shall discuss how the anomalies
occurring in DMS can be understood.
a. Temperature Dependence on the Shubnikov-de Haas Peak Positions
The magnetic field positions of the maxima (or minima) observed in the
Shubnikov-de Haas (SdH) effect are mainly determined by the crossing of
the Fermi level with consecutive Landau levels. Their temperature dependenCe is, in usual materials, only very slight. It was, therefore, quite surprising
to observe a very strong temperature shift of the maxima of the SdH effect
in Hg,-,Mn,Te (Jaczynski et al., 1978). The shift was particularly pronounced in the case of the spin split peaks. The small temperature variation
of the effective mass of conduction electrons could not possibly account for
the observed shifts.
These shifts are quite naturally explained in terms of the band structure
model incorporating the s-d exchange terms presented in Section I11 of this
chapter. They are due to a strong temperature dependence of the terms
proportional to ( S , ) , that govern the spin splittings of the Landau levels, and
also, but to a much lesser degree modify the positions of the unsplit Landau
levels. Similar strong temperature dependence was also observed in later
Shubnikov-de Haas effect studies both in Hgl-,Mn,Te by Byszewski et af.
(1979) and Hgl-,Mn,Se by Takeyama and Galazka (1979), Byszewski et af.
(1980) and Lyapilin et al. (1983).
It should be mentioned that in the case of the very last oscillation, i.e., that
occurring at the highest magnetic field (often labeled 0 - ) , the Shubnikov-de
Haas peak exhibits a relatively strong dependence on the temperature also
in nonmagnetic semiconductors. This has its source in the partial lifting of
a strong degeneracy of the electron gas. However, in diluted magnetic
semiconductors, as noted in Hgl-,MnxSe by Lyapilin et al. (1983), the 0peak observed in intense magnetic fields may shift towards smaller fields
when the temperature is increased. This is exactly the opposite direction of
the temperature shift of the 0- peak to that observed in nonmagnetic
semiconductors. Hence, the usual explanation may again not be applied to
the case of diluted magnetic semiconductors. On the other hand, both the
direction of the 0- shift and its magnitude can be readily explained by the
variation of the Landau level (n = 0, set b) with temperature due to the
exchange terms proportional to (&).
Probably the most striking feature in the SdH behavior of DMS was the
evolution the temperature of a well resolved doublet structure of Shubnikovde Haas peaks that remained unsplit at lower temperatures. This can be
clearly seen in Fig. 5 in the case of the oscillation maxima associated with
206
J . KOSSUT
160
-2 1 2 0
E
w
80
40
12
15
B(T1
10
21
24
27
18
21
24
27
(a)
160
z120
80
40
12
15
B(T)
(b)
FIG.5 . The Landau levels at k~ = 0 in Hg,-,Mn,Te (x = 0.02) calculated at (a) T = 4 K and
(b) T = 36 K using the value of /3 = 0.6 eV and a = - 0.4 eV. Open circles show the positions
of the maxima of the Shubnikov-de Haas oscillations observed by Byszewski et al. (1979).
Arrows indicate the peak which splits at elevated temperatures. Note that at the lower
temperature the energy gap opens only above 8T.
5.
207
208
J. KOSSUT
FIG. 6. Resistivity in Hgl-,Mn,Te (x = 0.009) versus temperature showing thermooscillation maxima. The arrows indicate the theoretically expected positions of the maxima.
[After Dobrowolska et al. (1979).]
5.
209
15
-s
-d 1 0
0
Y
+
3
.-n
E
d
6
TIK)
10
-hPOp-
r=l
f i X r exp(- arTD/H)
cos(mr) cos[ 2 n r e - y ) f
GsinhX,
:]
(29)
where a = 2n2m*kdeA, TD is the Dingle temperature describing phenomenologically the broadening of the Landau levels, X , = arT/H, v is the
spin splitting factor defined as
v
1 m*
-g*2 mo
with m* and mo standing for, respectively, the effective mass and the free
electron mass. The oscillatory character of the magnetoresistance is given by
the last cosine factor in Eq. (29) with the fundamental period PSH.For
spherical Ferrni surfaces, this can be related to the electron concentration N
by
Often only the first few terms of the expansion (29) are of importance,
because of the exponential damping factor involving TD. The temperature
210
J. KOSSUT
g* = - i ;
m*
* l , *3.
When theg-factors given by the band structure calculation for Hgl -,Mn,Te
were calculated, it turned out that, indeed, it is possible to meet condition
(30) in the temperature range where the Shubnikov-de Haas effect was
studied. A superposition of the factors Xl/sinh XI and cos(nv) gave a good
description of the amplitude vs. temperature curve (see the solid line in Fig.
7). The condition (30) suggests, further, that more than one zero of the
amplitude can be expected as a function of the temperature. In fact, two zeros
were detected in Hgl-,Mn,Se by Lyapilin et al. (1983). The temperatures at
which both amplitude zeros occurred could be well accounted for by the
simple explanation presented above.
Let us note that condition (30) corresponds to the situation where the
Landau spin-split sublevels are equally spaced in energy. This is an easy
situation for the collision broadening of the levels to smear out the quantum
oscillations, i.e., to produce a zero of the amplitude. So, one expects an
occurrence of a clear zero in relatively impure samples with rather high values
of the Dingle temperature TD. Such values enable one to neglect all higher
terms in the harmonic series, Eq. (29). For purer samples, when more terms
in the harmonic expansion are important, only a minimum of the amplitude
vs. temperature relationship can be expected (Kossut, 1978).
The analysis of the positions of the zeros of the amplitude as a function
of temperature can serve as a rather precise method for determining the gfactor, as well as the exchange constants a and fl. This method was actually
employed in the case of (Cdl-,Mn,)sAst mixed crystals (see further in this
part).
Recently, Reifenberger and Schwarzkopf (1983) have suggested a different
explanation for the beating of the amplitude of the Shubnikov-de Haas
effect observed by them in Hgl-,MnxSe with small x and also in HgSe at
weak magnetic fields. They suggest that the beating effect, which can be
interpreted as a zero of the amplitude at a certain magnetic field, is due to
5.28
5.
211
the magnetic breakdown between the closed-orbits in the spin-split conduction band. The splitting results from the lack of inversion symmetry
characteristic for the zinc blende lattice structure. Actually, the fact that the
beating effect is also found by Reifenberger and Schwarzkopf (1983) in HgSe
leads them to reject the explanation in terms of the cos(nv) factor described
above. In order to account for the temperature shift of the beat fields
observed in Hgl-,Mn,Se samples, Reifenberger and Schwarzkopf (1983)
conclude that substitution of Mn for Hg atoms makes the linear k terms in
the k * p Hamiltonian also sensitive to the temperature in this material.
4eH
1'2
2
where we assumed that A = OD.
The overlap corresponds to a positive AE, while for an open gap the value
of AE is negative. The value of the band overlap is a resultant of several
factors. First there is an upward movement of the lowest conduction band
level a,(O) with the magnetic field, determined mainly by the effective mass
(or, more precisely, by the cyclotron frequency) characterizing the conduction band. The variation of this level with the magnetic field is modified
by exchange interaction to a lesser degree by the exchange interaction than
that of the uppermost valence band Landau level bv(- 1). The latter level,
whose energy in nonmagnetic semiconductor decreases with increasing
magnetic field, may be shifted in DMS to considerably higher energies if the
212
J. KOSSUT
Since ( S , ) has a tendency to saturate in the magnetic field while the right
hand side of Eq. (32) is linear in the field, the above condition can only be
met in relatively small field region, if at all. Of course, the lower the
temperature, the easier it is to satisfy the inequality (32). Because of antiferromagnetic interactions between the localized magnetic moments in
Hgl- .Mn,Te and Hgl- ,MnxSe, the value of x(S,) at a given field and
temperature is a decreasing function of x for samples with the Mn mole
fraction greater than x = 0.05. At the same time, the rate at which the a,(O)
level moves up in energy with the field increases with x in zero-gap materials,
because the effective mass at the bottom of the conduction band becomes
smaller. Therefore, the existence of the overlap is limited to rather low Mn
contents region. It is estimated that in both HgI- .Mn,Se and Hgl- .MnxTe,
the highest value of x for which the overlap may be observable is about
0.04-0.05.
The detailed knowledge of the magnetization (i.e., of x ( S , ) ) is of course,
quite crucial for this estimate. The use of values of x(S,) taken from more
or less substantiated theoretical models of magnetization can lead to an
overestimation of the extent of the overlap.
The existence of the band overlap may lead to a substantial redistribution
of carriers between the two bands in question. This, in turn, is reflected as
an anomaly of various transport coefficients in very pure samples. In fact,
the analysis of the conductivity tensor components by Sandauer and
Byszewski (1982) gave the first direct evidence for the band overlap. It was
found that the warping asymmetry of the heavy hole band, proportional to
y 2 - ~ 3 , is also observable as the anisotropy of the band overlap in
Hgl- .Mn,Te (Sandauer et al., 1983).
5.
213
also caused a total alignment of the localized Mn spins, i.e., the region of
saturation ( S t ) = - $ was reached. The sample was then illuminated by a
microwave radiation with energy chosen in such a way that it correponded
to the condition of electron paramagnetic resonance of the Mn-subsystem.
The resonant absorption by the Mn ions and the change of the polarization
of the localized moments associated with it were visible as an increase of the
resistivity of the sample. The possibility of other mechanisms (e.g.,
bolometric effect, influence of the optical excitation of the carriers between
the Landau levels, etc.) contributing to the observed increase was ruled out
by the authors. On the other hand, when the resistivity change connected with
the sudden variation of the spin dependent scattering at the resonance
condition is calculated (Wittlin et al., 1980b), the correct magnitude of the
observed effect is obtained.
214
J. KOSSUT
5.
215
two such constants PI and fill which, in general, may have different values.
This fact may contribute additionally to the anisotropy of the g-factor of
(Cdl- xMnx)3Asz.
Much less experimental information is available on this diluted magnetic
semiconductor system than in the case of Hgl- *Mn,Te. The Shubnikov-de
Haas effect was investigated in this material by Neve et af. (1981). Because
of relatively high electron concentration in these crystals (of the order of
5 x 10" ~ m - ~it) was
, impossible to determine the variation with x of such
parameters as the energy gap and the momentum matrix elements P I ,PII
(again, because of the tetragonal distortion, the model contains two
independent momentum matrix elements).
However, it was noted that the pattern of the oscillations was quite distinct
from that in pure Cd3Asz crystals. In particular, the oscillations exhibited a
strong beating effect, with the position of the nodes sensitive to temperature.
When the amplitude at a fixed field was plotted versus the temperature, it
was noted that two zeros were observed (see Fig. 8) that, when interpreted
as being due to the cos(nvr) term (see subsection IV.3c of this chapter),
rendered the values of the exchange constants PI = = 4.9eV and
a = -3.4eV. These values are remarkably greater than those in diluted
magnetic semiconductors with the zinc blende lattice structure.
x=O.Ol
I x.0.02
8-1.53'1
B=1.90T
10
T (K)
216
J. KOSSUT
The anisotropy of the conduction band was also studied in (Cdl- .Mnx)3As2
by Blom et al. (1983) who studied the Shubnikov-de Haas effect in monocrystalline samples. The angular dependence of the effective mass was found
to be qualitatively similar to that in Cd3As2. The anistropy of the g-factor
was found to be smaller than predicted by a simple theory of Neve et al.
(1982), and even less pronounced than in Cd3Asz. This fact led Blom et al.
(1983) to conclude that there must be a strong dependence of the energy gap
Eg and the crystal field parameter 6 on the crystal composition given by the
Mn mole fraction x.
5. Pbl- .Mn,Te
AND
Pbl- .Mn,S
5.
217
6. EXPERIMENTS
In the first series of experiments by Grabecki et al. (1984a), the MIS
structures were prepared by placing a thin sheet of Mylar as an insulator on
the Hgl- .Mn,Te surface. The p-type samples were used with the Mn mole
fraction x = 0.1, that corresponded to an open energy gap of the order of
about 100 meV (c.f., Fig. 2). The relatively high value of x was chosen to limit
the effects of tunneling of electrons from the surface inversion layer to the
valence band in the bulk of the crystal.
It was found that the conductivity at low temperatures is dominated by a
flow of electrons within a thin quasi two-dimensional inversion layer
produced by an attracting surface potential. This potential restricts the
electronic motion in the direction perpendicular to the surface. The motion
218
J. KOSSUT
5.
219
grain boundary. The mobility of electrons in the layer at the boundary was
quite high (of the order of 104cm2/Vs), enabling observation of the
Shubnikov-de Haas oscillations in magnetic fields as low as 0.2T. Two or
three periods of the oscillations were observed, indicating again that several
electric subbands were occupied by 2d electrons.
The electron concentrations and effective masses at the Fermi level were
determined for each of the subbands and are listed in Table I. They are compatible with results of a calculation based on the model to be described in
the following section. The results of this calculation are also listed in Table I .
No spin splittings of the Landau levels were observed in the Shubnikov-de
Haas oscillations studied in the samples of Hgl- .Mn,Te ( x = 0.1). This fact
is due to the smallness of electronic g-factor, which at low temperatures, is
greatly reduced by the exchange interaction in this open-gap material. This
arises because the large and negative contribution to the g-factor due to the
spin-orbit interaction is nearly compensated by the exchange induced terms
proportional to the magnetization. Parenthetically, this fact enabled a
precise determination of the effective masses from the temperature study of
the oscillation amplitude.
To avoid the compensation of the g-factor, Grabecki et al. (1985) prepared
(x = 0.23, y = 0.02)
also samples from quaternary Hgl- ,-,Cd,Mn,Te
containing the grain boundaries. The energy gap in these samples
(=200 meV) is open mainly by the presence of the Cd atoms. The effects of
the exchange interaction due to the presence of the small number of Mn
atoms are here more pronounced than in Hgl- .Mn,Te (x = 0. I), because the
reduction of the magnetization in the latter material is greater. This results
from a stronger antiferromagnetic coupling between localized Mn moments
in the less diluted case. In fact, the Shubnikov-de Haas oscillations of the
2d electrons in Hgl- ,-,Cd,Mn,Te samples show well-resolved spin splittings
in the available range of magnetic fields. This enabled the first determination
of the g-factor of the 2d electrons in a diluted magnetic semiconductor. Not
surprisingly, the g-factor observed showed a strong temperature dependence
(see Fig. 9). It is also well accounted for by the model of the subband structure
(see next section). The effective mass at the Fermi level and the population
of the lowest electric subband (only one period of oscillation was observed)
in Hgl- x-yCdxMnyTeare given in Table I.
7. ENERGY
LEVELS
OF 2d ELECTRON
GAS
TABLE I
EXPERIMENTAL
AND THEORETICAL
VALUES OF 2d ELECTRON
CONCENTRATION, n: AND THE EFFECTIVE
MASSESAT THE FERMILEVEL( E F ) mf IN THE r-TH
ELECTRIC
SUBBAND
IN SAMPLES
CONTAINING
GRAINBOUNDARY.
E , IS THE CALCULATED
VALUEOF THE BOTTOM(n: = 0) OF THE r-m SUBBAND.
THE
FITTEDVALUE OF THE POTENTIAL SLOPE uo IS GIVEN. THESPATIAL EXTENSION
OF THE r-TH STATE, W,, WAS ESTIMATED
FROM Wr = E r / U o
Hgo.s9Mn0.11Te
110
0
1
2
H g 0 . 9 0 M n o .l o T e
85
5.3
1
0
Hg,-.-,Cd,Mn,Te
(X =
0.23, y = 0.02)
205
15.4 f 2
6.0 + 0.5
1.8 0.2
5.0 f 1
3.3 + 0.5
2.2 i 0.2
12.18
5.98
1.87
3.60
3.01
2.40
57.2
111.8
151.3
163
319
432
2.6
0.2
5.29
2.49
52.7
192
1.3 f 0.1
2.0 f 0.1
1.34
1.85
102.1
371
3.2
2.6 f 0.4
3.2
2.4
44.2
161
+ 0.3
f 0.1
172
7.0 lo4
122
5.5 lo4
80.9
5.5 lo4
5.
221
T(K)
FIG.9. Theg-factor of 2d electrons at agrain boundary in Hgo Kdo.z,Mno.o2Teas a function
of the temperature. The symbols with error bars represent experimental data, based on the
splitting of the magnetoresistance maximum at approximately H = 4.5T.Solid lines show the
results of the calculation based on the model described in the text. [After Grabecki ef a/.(1985).]
version of the model, neglecting the presence of the spin-orbit split-off band
r7,aiming at the analytic form of solutions. Actual calculations, the results
of which are given in Table I, were done without making this approximation.
However, the influence of the higher bands was neglected in both versions.
We start with the Hamiltonian in the matrix form given by Eqs. (13)-(16)
with y1 = y2 = y3 = F = K = 0.
The presence of the confining potential at the surface or at the grain
boundary requires an addition of an appropriate potential U to the diagonal
terms of these matrices. When the z-axis is chosen perpendicular to the
surface (or to the grain boundary), the confining potential depends only on
z. In the present model, very simple forms of this potential were assumed.
In the case of the surface layer, the potential was taken to be a triangular well
with infinite height on the insulator side; in the case of the grain boundary,
it was assumed in the form of a symmetric triangular potential well. The slope
of the potential is the only fitting parameter of the model. The other band
structure parameters can be taken from electronic studies of bulk crystals.
This simple form of the confining potential is probably the most severe
approximation of the model. We assume also that the magnetic field is
parallel to the z-axis. Finally, the exchange terms involving the constants a!
and must be added to the Hamiltonian. The matrices obtained in this way
represent two sets of differential equations for the envelope functions. These
222
3. KOSSUT
where
2 P2
A=-3 A2'
at
= [E f
a!'
- &f
(34)
- P'
U(z)] Eg +
&+
7+ E f
w].
(35)
In Eqs. (34) and ( 3 9 , Pdenotes the interband momentum matrix element and
a' = + X a ! ( S Z ) ,
(36)
8' = +x/3<S,>,
(37)
while E denotes the eigenvalue of energy that we are looking for. Plus and
minus indices in Eqs. (33) and (35) refer, respectively, to solutions for
predominately spin-up and spin-down states in the conduction band. The
parameter cf is defined by
with
+ $),
r = 0, 1 , 2 , ...
(40)
where za = 0 and q5 = $ for the case of the surface layer potential, while
zo = - zo and $ = for the layer at the grain boundary. Here zo denotes the
classical turn point of the motion in a confining potential well, i.e., it is
5.
223
defined by
PZ(Z0) =
The smallest root of Eq. (41) corresponds to solutions for the conduction
band
20 =
f ayl - Ef
uo
U(z) = Uolzl
(44)
for the grain boundary potential. The integration indicated by Eq. (40) can
be carried out analytically, giving the following equations
where
a*
=E 7
a - ~ f ,
b* = Eg +
and
G=
+ ~f F B
-,
3
Equation (45) has now to be solved numerically for the energies of the n-th
Landau level with spin (+) or (-) in the r-th electric subband.
A similar calculation can be carried out when no external magnetic field
is applied. The result is formally very similar to that obtained for InSb, with
the quantity ~f redefined (cf., Zawadzki 1983):
224
J. KOSSUT
The g-factors calculated in this way give a fairly good description of the
values observed in Hgl- x-yCdxMnyTe.At lower temperatures, they turn out
to be positive due to the large contribution of the exchange interaction terms.
As T increases (and ( S , ) decreases), the value of the g-factor drops,
becoming nearly zero at the highest temperatures at which the oscillations
were studied. Let us mention also that the values of the g-factor for the 2d
gas of electrons at the grain boundary are slightly different from those
expected for bulk electrons. The above results indicate that the influence of
the localized magnetic moments also manifest itself in the properties of 2d
electrons in diluted magnetic semiconductors.
8 . FUTURE
POSSIBILITIES
5.
225
References
Abrahams, E. (1955). Phys. Rev. 98, 387.
Adler, M. S., Hewes, C. R., and Senturia, S. D. (1973). Phys. Rev. B7, 5186.
Anderson, P. W. (1961). Phys. Rev. 124, 41.
Ando, T., Fowler, A. B., and Stern, F. (1982). Rev. Mod. Phys. 54, 437.
Bastard, G., Rigaux, C., Guldner, Y . , Mycielski, J., and Mycielski, A. (1978). J. dePhysique
39, 87.
Bhattacharjee, A. K., Fishman, G., and Coqblin, B. (1983). I n Physics of Semiconductors
(Proc. 16th Internat. Conf. on the Physics of Semiconductors, Montpellier, 1982,
M. Averous, ed.) Part I, p. 449, North Holland, Amsterdam.
Blom, F. (1980). I n Narrow Gap Semiconductors Physics and Applications (W. Zawadzki,
ed., Proceedings International School, Nimes, 1978, Lecture Notes in Physics, vol. 133,
p. 191, Springer-Verlag).
Blom, F. A. P., Neve, J. J., and Nouwens, P. A. M. (1983). In Physics of Semiconductors
(M. Averous, ed., Proc. 16th Internat. Conf. Montpellier, 1982) Part I, p. 470, North
Holland, Physica 117-118 B + C).
Bodnar, J. (1978). In Physics of Narrow Gap Semiconductors (J. Rauluszkiewicz,
M. Gorska, E. Kaczmarek, eds., Proceedings of International Conference, Warsaw, 1977)
p. 311, P.W.N. Publishers.
Byszewski, P., Szlenk, K., Kossut, J., and Galazka, R. R. (1979). Phys. Stat. Sol. (b) 95, 359.
Byszewski, P., Cieplak, M. Z., and Mongird-Gorska, A. (1980). J. Phys. C13, 5383.
Delves, R. T. (1963). J. Phys. Chem. Solids 24, 885.
Delves, R. T. (1966). Proc. Phys. SOC.87, 809.
Dobrowolska, M., Dobrowolski, W., Galazka, R. R., and Kossut, J. (1979). Solid State
Cornmun. 30, 25.
Dobrowolska, M., and Dobrowolski, W. (1981). J. Phys. C14, 5689.
Dobrowolska, M . , Dobrowolski, W., Galazka, R. R., and Mycielski, A. (1981). Phys. Star. Sol.
(b) 105, 477.
Elsinger, G. (1983). Thesis, Montanuniversitat Leoben, unpublished.
Friedel, J. (1958). Nuovo Cimento 52, 287.
Gaj, J. A., Planel, R., and Fishman, G. (1979). Solid State Commun. 29, 435.
Galazka, R. R., and Kossut, J. (1980). I n Narrow Gap Semiconductors Physics and
Applications (W. Zawadzki, ed., Proc. Internat. School, Nimes, 1979) Lecture Notes in
Physics, vol. 133, p. 245, Springer-Verlag.
Galazka, R. R., and Kossut, J. (1983). In Landolt-Borstein New Series, Group 111
(0.
Madelung, M. Schulz, and H. Weiss, eds.) vol. 17b, p. 302, Springer-Verlag.
Grabecki, G., Dietl, T., Kossut, J., and Zawadzki, W. (1984a). SurJ. Sci. 142, 588. (Proc. 5th
Internat. Conf. on Electron Properties of Two-Dimensional Systems, Oxford, 1983.)
Grabecki, G., Dietl, T., Sobkowicz, P., Kossut, J., and Zawadzki, W. (1984b). Appl. Phys.
Lett. 45, 1214.
Grabecki, G., Dietl, T., Sobkowicz, P., Kossut, J., and Zawadzki, W. (1985). Acta Phys.
Polon. A67, 297.
Hass, K. C., and Ehrenreich, H. (1983). J . Vac. Sci. Techno/. 1, 1678.
Irkhin, Yu. P. (1966). Zh. Eksp. Teor. Fiz. 50, 379. (Sov. Phys. JETP23, 253.)
Izyumov, Yu. A., and Noskova, L. M. (1962). Fiz. Tverdogo Tela 4, 217. (Sov. Phys. Solid
State 4, 153.)
Jaczynski, M., Kossut, J., and Galazka, R. R. (1978). Phys. Stat. Sol. (b) 88, 73.
Kacman, P., and Zawadzki, W. (1971). Phys. Stat. Sol. (b) 47, 629.
Kane, E. 0. (1957). J. Phys. Chem. Solids 1, 249.
226
J . KOSSUT
5.
227
Stankiewicz, J., Giriat, W., and Bien, M. V. (1975). Phys. Stat. Sol. (b) 68, 485.
Story, T., Lewicki, A., and Szczerbakow, A. (1985). Acta Phys. Polon. A67, 317.
Takeyama, S.,and Galazka, R. R. (1979). Phys. Stat. Sol. (b) 96, 413.
Toth, G., and Leloup, Y. Y. (1970). Phys. Rev. B1, 4573.
Uchida, S., Landwehr, G., and Bangert, E. (1983). Solid State Commun. 45, 9869.
Vonsovskii, S. V., and Svirskii, M. S. (1964). Zh. Eksp. Teor. Fiz. 47, 1354. (Sov. Phys. JETP
20, 914.)
Vonsovskii, S. V., and Turov, E. A. (1953). Zh. Eksp. Teor. Fiz. 24, 419.
Vul, B. M., and Zavaritskaya, E. T. (1979). Zh. Eksp. Teor. Fiz. 76, 1089. (Sov. Phys. JETP
49, 551.)
Wallace, P. R. (1979). Phys. Stat. Sol. (b) 92, 49.
Watson, R. E., and Freeman, A. J. (1966). Phys. Rev. 152, 566.
Watson, R. E., and Freeman, A. J. (1969). Phys. Rev. 178, 725.
Wittlin, A., Knap, W., Wilamowski, Z., and Grynberg, M. (1980a). Solid State Commun. 36,
233.
Wittlin, A., Grynberg, M., Knap, W., Kossut, J., and Wilamowski, Z. (1980b). J. Phys. SOC.
Japan 49 (suppl. A), 635. (Proc. 15th Internat. Conf. Phys. Semicond. Kyoto, 1980.)
Yosida, K. (1957). Phys. Rev, 106, 893.
Zawadzki, W. (1983). J. Phys. C16, 229.
Zener, C. (1951a). Phys. Rev. 81, 440.
Zener, C. (1951b). Phys. Rev. 82, 403.
Zener, C. (1951~).Phys. Rev. 83, 299.
CHAPTER 6
INTRODUCTION. . . . . . . . . . . . . . . . . . . .
THEORETICAL
DESCRIPTION
OF ELECTRONIC
STATESIN NGDMS
1 . Band Structure . . . . . . . . . . . . . . . . . .
2. Landau Levels . . . . . . . . . . . . . . . . . .
3. Magnetooptical Transitions . . . . . . . . . . . . .
111. INTERBANDMAGNETOOPTICS.
. . . . . . . . . . . . . .
4. Zero Gap Hg,-,Mn,Te. . . . . . . . . . . . . . .
5. Narrow Gap H g l - , M n x T e . . . . . . . . . . . . . .
6 . Other Semimagnetic Compounds. . . . . . . . . . .
IV. INTRAEIANDMAGNETOOPTICS
. . . . . . . . . . . . . .
7. Far IR Magnetospectroscopy in Zero Gap DMS. . . . .
8 . Harmonics of the Cyclotron and Combined Resonances in
Narrow Gap Semiconductors . . . . . . . . . . . .
V. RESULTS
OF MAGNETOOPTICAL
STUDIES.. . . . . . . . . .
9. Band Parameters . . . . . . . . . . . . . . . . .
10. Exchange Parameters . . . . . . . . . . . . . . .
VI. CONCLUDING
REMARKS. . . . . . . . . . . . . . . . .
REFERENCES
. . . . . . . . . . . . . . . . . . . . .
I.
11.
229
23 1
232
233
237
240
24 1
248
252
257
257
264
267
267
269
27 1
272
I. Introduction
Diluted Magnetic Semiconductors (DMS) are characterized by the
presence of magnetic ions of a transition element, that enter substitutionally
into the semiconductor lattice. The interest in this class of magnetic semiconductors arises from the possibility of controlling the dilution of magnetic
ions and the well known band structure of the host semiconductor (typically
a 11-VI compound). Thus DMS offer an unique opportunity of gradually
incorporating the exchange phenomena, characteristic of magnetic semiconductors, into the electronic properties of classical semiconductors.
Narrow-gap DMS are ternary random alloys obtained by substituting
magnetic ions (Mn, Fez+) for Hgz+ in mercury chalcogenides (HgTe,
HgSe). Among them, the system Hg1-xMnxTe has been studied particularly
229
Copynght 8 1988 by Academic Press. Inc.
All rights o f reproduction in any form reserved.
ISBN 0-12-752125-9
230
C. RIGAUX
6.
231
6 .I
232
C . RIGAUX
1. BAND STRUCTURE
In II-VI semiconductors, the band edges of relevance are located at the
Brillouin zone center (r point), where the states transform according to the
representation of the Td crystal symmetry group. The six p-like states are
decoupled by the spin-orbit interaction into a quadruplet Ts (J = 4) and a
doublet r7 (J =
the latter forming a lower-lying valence band, with the
energy A = Er, - Er, determined by the spin-orbit splitting. In HgTe, the
s-like band, transforming as r6 (J = *), lies below the Ts quadruplet, leading
to a symmetry-induced zero-gap semiconducting configuration; the Ts levels
then form the degenerate valence and conduction band edges, separated from
the r6 valence band by the interaction gap EO = Er6 - Ers = -0.3eV at
4.2 K (Fig. 1).
CdTe and (hypothetical) cubic MnTe are wide gap semiconductors. The
Tsquadruplet forms two degenerate valence bands and the conduction band
edge belongs to r6 representation. The fundamental energy gap (1.6 eV in
CdTe) is estimated to be 3.2 eV in cubic MnTe (Gaj et al., 1979).
The band structure of ternary alloys Hg, -,Cd,Te and Hg, -,Mn,Te evolves
continuously from the zero gap configuration to an open-gap configuration,
as shown in Fig. 1. The interaction gap EO , defined as a difference between
energies of the r.5and rs states, increases with the alloy composition. These
systems undergo a zero gap to semiconductor transition that occurs, at 4.2 K,
for x = 0.075 in Hgl-,Mn,Te and x = 0.165 in Hgl-,Cd,Te. In the absence
i),
6.
233
positive gop
$i
\ I
4-
0.14
2. LANDAULEVELS
In the presence of an applied magnetic field, the electronic states in DMS
are obtained by including the exchange interaction as an additive term X,,
in the proper effective mass Hamiltonian XO. For zinc blende narrow gap
semiconductors, the most accurate effective mass Hamiltonian XO is
provided by the eight-band Pidgeon-Brown model (1966). In this model the
k p interactions between closely spaced r6, r7, Ts bands are treated exactly
and the interactions between these states, and the more remote higher bands
are included up t o order k2. The magnetic Hamiltonian XO, in the absence
of exchange, consists of an 8 x 8 matrix D,that reduces to two 4 x 4 matrices
Da and Db ,at the r point when inversion asymmetry and warping in Tg bands
are neglected. The matrix D has the form:
in the basis of the band edge Bloch functions ur(J, MJ) associated with
r
6
234
(f
= 1,2),
C . RIGAUX
Ts (f
= 3,
..., 6 ) , r7(I
i.;(;,-;)
rU2(;-;,)
= 7,8) levels:
= -&(Xi
iY)t
+ Zl)
= ijS1)
where Si is a localized spin at the site Ri (Si = 5);0 is the spin of the mobile
electron, and J(r - Ri) is an exchange integral rapidly varying over a unit
cell. Since the mobile electron has a spatially extended wave function, it
interacts simultaneously with a larger number of Mn ions. This allows one
to replace the Mn spin operators by their thermal average ( S ) . In the absence
of magnetic field ( S ) vanishes. When the magnetic field H is applied along
the z direction, ( S ) has only a non-vanishing component (S,) .
6.
235
Thus, by using the VCA and the molecular field approximation, the
exchange perturbation becomes:
-NOxgMnpB(Sz),
(4)
where No is the number of unit cells per unit volume, x is the Mn molar
fraction and p~ is the Bohr magneton.
For 3d5 electrons of Mn, ( S , ) = -Sfnorm;where S = 5 and fnom denotes
the normalized magnetization. For non-interacting magnetic moments,
2S+1
coth
2S+1
~
1
x
-coth-.
2
2
where
/3Aa/P
0
-3AdP
Mb=[
0
0
0
-2Afi
(7)
0
0
A
0
0
-3A
2Afi
0
The off diagonal r7 - TSterms in Eq. (7) may be neglected in the case of large
spin-orbit splittings. Here M, and Mb involve the s-d and p-d exchange
integrals (Y = (SIJ(r)lS),B = (XIJ(r)IX) = (YIJ(r)lY)= ( Z I J ( r ) l Z ) ,that
are the only non-vanishing matrix elements of J(r) involving S , X , X Z
236
C . RIGAUX
(DO- Eaf))lva = 0 ,
(Db
Ebf))Wb =
0,
(8)
wa(n) = a i nf n u l
vb(n)
+ a 3 n f n - i ~ + a 5 n f n + i ~ 5+ a 7 n f n + i ~ 7 ,
+ h n f n - 1 ~ 4 + b6nfn+lU6 + b8nfn-lu8.
(9)
b2nfn~2
where m?6 and g?6 are the band-edge effective mass and gyromagnetic factor
6.
237
g *r , -2
z.,:*>*
I------
The last term in Eq. (11) represents the s-d exchange contribution. The
exchange effect leads to a modification of the effective g-factor, that can be
expressed in the form:
For the l-8 bands, the energies of a and b ladders are given in the parabolic
limit by the analytical expressions of the Luttinger model (Luttinger, 1959),
in which the parameter K is replaced by R = K - 2A/hw0. In the classical
limit (i.e., large n), the l-8 band-edge effective masses are
238
C. RIGAUX
Here EOis the amplitude and E the unit vector in direction of the electric field
of the radiation, and ll is the operator - ih V + e A / c , where A is the vector
potential due to the external magnetic field (H 11 d ) .
The probability for a direct transition of an electron from an initial state
IZ) to a final state IF) is proportional to the square of the modulus of the
matrix element nt = (FIE* IIlZ). The absorption coefficient is given by:
where I/ is the crystal volume, q is the refractive index, and g(E1) and (EF)
denote the occupation factors for initial and final states, respectively.
We consider direct transitions between Landau levels described within the
modified Pidgeon-Brown model by wave functions, Eq. (9), of the form:
#I(m) =
C Ajffrn;, ujr ;
j'
yF(n) =
Cj p j f n j u j .
As the envelope function f(r) and the vector potential A(r) are slowly varying
over the unit cell, after taking into account the orthogonality of the bandedge Bloch functions the matrix element connecting initial and final states is:
The dominant contribution to the transition results from the first term. The
absorption coefficient is finally obtained by summing Eq. (17) over initial
( m , ky ,k,) and final ( m , ky , k,) states:
ha
(19)
6.
239
TABLE I
SELECTION
RULESFOR INTERBAND MAGNETOOPTICAL
[m. n DENOTELANDAUQUANTUM
TRANSITIONS
a , b CORRESPOND
TO A GIVENSET OF
NUMBERS;
SOLUTIONS,
EQ.(9)]
Polarization
Selection rules
E-
E+
o(m)
b(m)
b(n) n
a(n) n
=
=
m
m
+1
-
-+
-+
--t
-+
240
C . RIGAUX
(Weiler, 1981) or the simplified three band model (Kacman and Zawadzki,
1971) are required to describe the Landau levels away from the r point and
the electric dipole spin resonance absorption (McCombe, 1969).
The inversion asymmetry of the zinc blende lattice leads to additional
that have been neglected in the above
terms in the magnetic Hamiltonian KO,
formulation as these terms are usually very small. The inversion asymmetry
effects lead to induce the electric dipole spin resonance and the combined
resonance for any polarization. Of the two possible processes for the
excitation of intraband spin-flip resonances, the s-p interband coupling was,
until recently, thought to be the dominant mechanism in HgCdTe compounds (McCombe et al., 1970,1971). However, more recently, it was clearly
demonstrated in the case of InSb by Chen et al. (1985) that the lack of
inversion symmetry in the zinc blende lattice is indeed the dominant
mechanism which allows the transitions involving spin-flip resonances by
electric dipole excitation.
-+
6.
241
4. ZEROGAPHgl-,MnxTe
+ 1);
&+(oro+):br6(n + 1) 4 brs(n);
E
(1 H: br6(n +
1)
(20)
ar8(n).
The line identification is indicated in the figures. Only for very diluted alloys
(x = 0.0014), the relative ordering of the transitions is identical to that
observed in the spectrum of HgTe (Guldner et al., 1973) or Hgl-,Cd,Te
alloys (Guldner et al., 1977; Rigaux, 1980). For x L 0.004, a change of
regular transition sequence appears, resulting in an inversion of the relative
positions of the circular polarization transitions ar6(n) ar,(n + 1)(0-) and
br,(n) 4 brs(n - l)(a+)(Figs. 2-4). Comparison between Hgl-,MnxTe and
Hgl -,Cd,Te alloys of identical interaction gaps shows that this modification
originates from anomalous spin effects in the semimagnetic compounds,
whereas the energy separation between consecutive transitions for a given
polarization (i.e., the period of oscillations of a given spectrum) remains
+
H( kG 1
FIG. 2. Energies of r6 Ts transitions vs. magnetic field for Hgl-,Mn,Te of Mn content
x = 0.004 at 4.2 K . Symbols are experimental. Solid lines represent theoretical fits using the
modified Pidgeon-Brown model. The quantum number n identifying the transitions according
to Eq. (20) is indicated for each line. [After Bastard et 01. (1978).]
+
242
C . RIGAUX
10
20
30
50
40
H(kG)
FIG. 3. Energies of rs
--f
l-8
x = 0.004 at 2 K. Symbols are experimental. Solid lines represent theoretical fits using the
modified Pidgeon-Brown model. The quantum number n identifying the transitions according
to Eq. (20) is indicated for each line. [After Bastard et al. (1978).]
360
I,.
'
3 20
2
v
280
HW)
FIG. 4. Energies of r,5 + Tg transitions vs. magnetic field for Hgl-,Mn,Te of Mn content
x = 0.008 at 2 K . Symbols are experimental. Solid lines represent theoretical fits using the
modified Pidgeon-Brown model. The quantum number n identifying the transitions according
to Eq. (20) is indicated for each line. [After Bastard ef 01. (1978).]
identical for both compounds. These features are illustrated in Fig. 5 by plots
of the period ha and the spin splitting
= Eb(l) - Ea(l)of the n = 1
conduction Landau level, directly measured form the energy difference at a
given field, between the r7+ and E 11 H transitions, c.f., Eq. (20). The quantity
ha, representing in the parabolic limit the sum of the cyclotron energies in
r' and Ta conduction bands, has comparable values for HgMnTe and
HgCdTe, demonstrating the similarity of the band parameters in both
families. The strong enhancement of the spin splitting in the semimagnetic
6.
I
243
H=20kG
20
2
E
UJ
10
-300
-250
-200
e, ( meV 1
- 150
-100
FIG.5. Spin and cyclotron splittings, Sc(l) and An; respectively, as a function of EO in
HgCdTe (open symbols) and HgMnTe (black symbols) at 4.2 and 2 K, for H = 20 kG. [After
Bastard et al. (1978).]
compounds provides direct evidence of the modification of electron spin sublevels induced by the exchange interaction.
Another manifestation of exchange interaction in zero gap HgMnTe is the
0 spectrum
observation of a new and intense magnetoabsorption line in the '
of alloys for x 2 0.004 (Fig. 6). This line, quantitatively interpreted as being
0 polarization, provides direct
due to br6(0) br,(-l) allowed for '
evidence for the emergence of the br8(- 1) heavy hole level above the Fermi
energy.
One of the most remarkable properties of r6 * Ts magnetooptical spectra
is their strong temperature dependence in striking contrast to HgCdTe alloys.
On decreasing the temperature, an energy shift is observed for the circular
'
, negative for 0-),
but E 1) H transitions are
polarization lines (positive for o
unaffected. The temperature behavior is illustrated in Fig. 6 by the shift of
the resonance field of the inter-valence transition br6(0) br,(- 1). The
observed temperature dependence of l-6 Ts transitions results in an
enhancement of the conduction electron spin splitting between 4.2 and 2 K
(Fig. 5). This effect originates entirely from the exchange contribution, which
induces an (&)-dependent term in the gyromagnetic factor (11.2, Eq. (16)).
A remarkable peculiarity of the spectra is the opposite curvature exhibited
by the u+ and o- transitions in the energy diagrams E ( H ) (Figs. 3, 4).
This bowing effect, particularly pronounced at 2K, gives rise to an
unusual crossing of the lines associated with br,(n) * br,(n - I
)(o') and
ar,(n) ar,(n + l)(a-)transitions.
All these manifestations of the exchange interaction are particularly
pronounced at very low temperature (4.2 and 2 K). In the temperature range
-+
-+
244
C. RIGAUX
0 10
20
30
1
LO
50
H (kG)
10-65 K, the results of Dobrowolska and Dobrowolski (1981) show that the
magnetooptical transitions conserve the same sequence as in the spectrum of
HgTe, and the inter-valence transition is not observed. The main effect of
the exchange interaction consists in the temperature dependence of the spin
splitting Sc(n), which is seen to decrease with increasing temperature.
b. Theoretical Interpretation
Interband magnetooptical spectra were successfully interpreted within the
framework of the Pidgeon-Brown model generalized to include the s-d and
p-d exchange contributions (Part 11, section 2 ) . A numerical fitting
procedure of the r6 + Ts transition energies to the experimental lines was
achieved at a fixed magnetic field. The variables of the minimization
procedure were restricted to Ep(or P = ( - ih /m )(Sl p x l X ))and the exchange
parameters. The spin-orbit splitting and the Luttinger parameters were fixed
to the values obtained for HgCdTe alloys of low composition: A = 1 eV;
y1 = 3; p = 0; K = -1.65 (Guldner et al., 1977). The interaction gap EO was
determined for each composition from the extrapolated zero field energy of
the r6 r8 transitions and then adjusted to obtain the best theoretical fits.
For dilute HgMnTe compounds ( x < 0.015), EPwas found between 18.3
and 18.8 eV ( T = 4.2 and 2 K), in excellent agreement with the results
+
6.
245
OL
0
'
I
WG)
50
246
C . RIGAUX
0.035
1
-I
0.0251.
i-"
0
-300
, &@",j
-250
-200
FIG.8. Band edge effective masses and gyromagnetic factors for the conduction band in zerogap HgCdTe (crosses) and HgMnTe alloys. Open circles are for T = 4.2 K, and black circles
for T = 2 K. [After Bastard (1978).]
6.
241
20
2
E
10
20
10
0
0
10
20
30 40
H (tG)
10
20
30 40
248
C. RIGAUX
5 . NARROW
GAP Hgl-,Mn,Te
a. TS
r
6
Magnetoabsorption
bs(n
l),
TI 2
-1.
(21)
For the 0- polarization, transitions from heavy and light valence ladders
are expected to be strong, the dominant series corresponding to:
-1.
(22)
6.
249
.i
T=2K
27
li
Y
230
210
10
20
30
40
50
H(kG)
(4
2w
10
20
30
40
50
H(kG)
(b)
FIG. 10. Energies of Ts + Ts transitions vs. magnetic field for Hgl-,MnxTe of x = 0.128
(EO = 219 meV), at 2 K, (a) E I( H polarization, and (b) 0 - polarization. Points are experimental.
Solid lines represent theoretical fits obtained within the modified Pidgeon-Brown model. The
quantum number n identifying the transitions according to Eqs. (21) and (22) is indicated for
each line. The line marked Imp corresponds to the acceptor -+ br,(O) transition. [After
Bastard er a/. (1981).]
250
C. RIGAUX
H(kG)
Eq. (10)). This explains the observed temperature and magnetic field
dependence of bra(- 1) -+ br,(O)(a-)and ars(- 1) br,(O)(E (1 H) transitions
(Fig. 10).
Comparison between theoretical and experimental transition energies vs.
magnetic field is shown in Fig. 10 for x = 0.128 (EO = 219 meV). The band
parameters eo ,E p , and the exchange parameters r, A(H, 7')were determined
at 4.2 and 2 K in the range of composition 0.12 5 x 5 0.17. The results are
presented in Part V. The magnetization parameter A ( H ) varies almost
linearly with the magnetic field (Fig. 7) and is weakly temperature dependent
in this range of composition.
Hgl-,Mn,Te semiconductors are usually p-type. The relatively high
density of acceptors (Nu = 5 x 10'6cm-3) present in the investigated
compounds makes possible the observation of acceptor-to-conduction
Landau level transitions. An intense absorption line satellite of the interband
br,(O) br,(l) transition of weaker intensity (labelled Imp in Fig. lob),
extrapolates at zero field to about 5 meV below the energy gap. This line may
result from the transition between an acceptor state (bound to br,(O))and the
conduction band Landau level br,( 1).
+
-+
6.
b. Effect of Exchange on
r
6
251
Landau Levels
The exchange interaction affects the Landau level spectrum of the narrow
(positive) gap HgMnTe in a striking fashion. The magnetic field dependence
of the conduction Landau level edges, for a semiconductor of EO = 219 MeV
shown in Fig. 12, illustrates the new important features originating from the
exchange phenomena.
In the conduction band, the relative positions of the levels ar,(n) and
br,(n), corresponding, respectively, to nt and nl electron spin states, are
inverted with respect to the usual ordering in zinc blende semiconductors, so
that in the present case E&) > Eb(n). This effect is a direct consequence of
the s-d exchange interactions, which shift the spin sublevels upward for a(n)
and downward for b(n). Thus, in semimagnetic alloys with positive energy
gap, the gyromagnetic factor of conduction electron is positive. This is in
complete contrast to zero gap HgMnTe and to non-magnetic narrow gap
semiconductors, such as InSb or HgCdTe, where the electron gyromagnetic
factors are always negative.
The origin of the sign reversal of the g-factor occurring in these materials
can be understood from the analytical expression (14) for the effective
gyromagnetic factor in the parabolic limit. The first term gc* resulting from
the k * p contribution is negative, whereas the second term, representing the
s-d exchange contribution, is positive (a < 0). Thus, with increasing
magnitude of the exchange interactions, the effective gyromagnetic factor
301
2
w
20[1
.--+a
FIG. 12. Magnetic field dependence of Landau levels of the conduction band at k, = 0 for
HgMnTe with Mn content x = 0.128 calculated at T = 2 K . The labels a, b correspond to two
sets of solutions of Eq. (8); the Landau quantum numbers are given at each curve. [After Bastard
et al. (1981).]
252
C . RIGAUX
g& will go through zero and subsequently reverse sign when the exchange
-+
-+
6.
20I
'
I
40
'
I
60
253
'
HlkGI -
H/kGl--
FIG.14. Magnetic field dependence of the electron g* factor calculated for the n = 1 Landau
level for HgSe (dashed lines) and Hg0.9~Mno.ollSe(solid lines). [After Dobrolowska et al.
(1981).]
254
C . RIGAUX
b. HgFeTe
New DMS containing substitutional Fe2+ions were recently synthetized (for
x I 0.02) at Purdue University. The first magnetooptical study carried out
by low temperature interband magnetoabsorption in zero gap Hg, -,Fe,Te
alloys have emphasized the semimagnetic character of these compounds
(Serre et al., 1981, 1982). The r6 l-8 magnetooptical spectra obtained for
the rn and ll polarizations show (similarly as in HgMnTe) exchange-enhanced
spin splittings, as well as the occurrence of the inter-valence transition
br,(O) -+ h 8 ( - 1 ) in the rn+ spectrum of alloys with ~ E O I > 260meV. In
complete c6ntrast to HgMnTe, however, the magnetooptical spectra are
independent of temperature between 4.2 and 2 K. The r6 r8 magnetooptical spectra of very dilute Hgl -*FexTe alloys were quantitatively
interpreted in the framework of the modified Pidgeon-Brown model and the
exchange parameters r = -0.3 and A ( H ) were determined. The negative
values of r and A imply that a < 0 and fl > 0, similarly as in the case of
HgMnTe, leading to exchange-enhanced gyromagnetic factors in both the r6
and the Ts conduction bands. The exchange effects in Hgl-,Fe,Te are,
however, much weaker than in Hgl-xMn,Te alloys of comparable interaction gap, as indicated by the values of A ( H ) obtained for both alloy
systems: e.g., for EO = -230 meV, at H = 2T and 2 K, A is about twice as
large in HgMnTe as it is in HgFeTe.
The most striking peculiarities of HgFeTe compounds are the linear field
dependence of A ( H ) and its invariance with temperature. The results
obtained for several values of the Fe content also show a linear dependence
of the magnetization on composition, as illustrated in Fig. 15 by plotting
- A / H vs. the relative interaction gap energies. This implies that in dilute
HgFeTe alloys, ( S , ) is independent of composition, in sharp contrast with
-+
6.
--
255
1-
%r
E
c
0
u-
0.3
0 1
0
10
20
30
H(kG)
40
50
60
FIG. 16. Normalized magnetization &,, vs. magnetic field determined at 2K for
Hg,-,Mn,Te alloys of several Mn concentrations. (A: x=0.0014; B: x=0.004; C: x = 0.008;
D x=O.Ol; E x = 0.015. Dots are experimental data obtained form interband magnetooptics.
[After Bastard ef a/. (1978).] Continuous lines represent theoretical fits obtained by using the
generalized Brillouin function given in Eq. (6) with Treplaced by T + r0. The value of the fitting
parameter %(x) for each composition is indicated in the figure. [After Bastard and Lewiner
(1980).]
256
C. RIGAUX
drastically contrasts with the results obtained in dilute Hg, -*Mn,Te alloys,
where magnetization manifests a strong temperature dependence.
The behavior observed on Hgl-,FexTe alloys originates from paramagnetism of Fe2+ions in the zinc blende lattice (Serre et al., 1982). The d
shell of the Fe2+ ions has 6 electrons, and in vacuum the ground state
corresponds to S = 2 and L = 2. The cubic crystal field lifts the resulting
25-fold degeneracy into a 10-fold-degenerate level and a 15-fold-degenerate
excited state separated by A = 0.3 eV from the ground state (Mahoney et al.,
1970). The spin-orbit coupling AL .S (with A = 12 meV) further lifts the
degeneracy of the ground state into 5 equidistant levels with an energy
separation equal to 6A2/A, the degeneracy of the levels being 1, 3, 2, 3,
and 1, respectively. Finally, the Zeeman termpB(L, + gFeSz)Hisintroduced.
The resulting spectrum of the d level of an Fe2+ ion is shown in Fig. 17.
The ground state is a singlet, that is, there is no permanent moment at zero
magnetic field. Consequently, there is no linear Zeeman shift. The quadratic
Zeeman shift
Eo(H) = EO - 4-PB2 A H 2
3
A
physically corresponds to a field-induced magnetic moment, that varies
linearly with magnetic field and arises from the virtual transitions between
the ground state and the excited states. The magnetic susceptibility
x = 83 p2~ ( A / A 2
is) therefore independent of magnetic field and of temperature, as long as the contribution of the first excited triplet can be neglected
1
1,
/ /
IL
'I
6.
257
(i.e., kT Q 6A2/A). Thus at low temperatures, in contrast with the CurieWeiss paramagnetism of Mn2+ions, the Fe2+magnetism is of the Van Vleck
type and arises from the lack of a permanent moment at zero field.
IV. Intraband Magnetooptics
7. FARIR MAGNETOSPECTROSCOPY
IN ZEROGAPDMS
In non-magnetic zero gap semiconductors, FIR spectroscopy has been
successfully applied in carrying out detailed studies of the Fg valence and
conduction bands. Magnetotransmission experiments performed in the
submillimeter and FIR region on unoriented n-type samples of HgTe
(Tuchendler et al., 1973) and Hgl-,CdxTe alloys (Kim and Narita, 1976;
Guldner et al., 1977) with low electron densities ( h a > EF) led to the
observation of intraband electronic and interband Ti' Ti transitions.
Measurements performed on oriented p-type samples with magnetic field
applied along the principal crystallographic directions were especially useful
for determining the Fg band anisotropy in HgTe (Pastor et al., 1978). HgSe
is characterized by large electron densities (n, = 10'' ~ m - leading
~ )
to an
opacity of the crystal at long wavelengths, due to plasma cut-off. FIR
measurements carried out on HgSe at frequencies larger than the plasma
frequency have provided observation of intraband electronic transitions
(plasma-shifted cyclotron resonance for E I H, combined resonance for
E I( H, and electric dipole spin resonance for both E 11 H and E I H). The
interband Ti' -+ Fitransitions requiring photon energies larger than the Fermi
energy (typically 14 meV) could also be excited in the investigated spectral
region (7.61-12.85 meV) (Pastor et al., 1980, 1981).
In zero gap DMS, FIR magnetooptics provides one of the best experimental methods for studying the behavior of Tg Landau levels in the presence
of magnetic exchange. Pastor et al. (1979) observed at 4.2K a strong
influence of the exchange effects on the positions of the magnetotransmission minima associated with intraband and interband Ti --* r; transitions in
dilute Hgl -,Mn,Te compounds with low electron densities. Similar features
were reported by Kim et al. (1982) in quaternary Hgl-,-,Cd,Mn,Te
compounds for several Mn compositions. In zero gap Hgl-xMn,Se, FIR
magnetooptical experiments were particularly difficult due to the high
electron concentration (typically 10'' cmP3), leading to a plasma edge cutoff
at wavelengths about 100pm. Using samples of exceptionally low electron
,
by appropriate annealing,
concentration (5 x 1OI6~ m - ~ ) prepared
Witowski et al. (1982) performed a detailed study of intraband spin-flip
resonances in zero-gap Hgo.97Mn0.03Se.The strong influence of exchange on
the conduction spin sublevels is experimentally manifested by a dramatic
+
258
C . RIGAUX
TS
rb
30
io
50
FIG. 18. Magnetotransmission spectra at frw = 3.68 meV for Hgl-,Mn,Te alloys of low x
observed in the E 1) H polarization. Experimental lines are identified as follows according to the
notation of the Luttinger model: CB: combined resonance ~ ( 1 +
) &2(2);SR: spin resonance
&](I).+ &$(I); E: d ( 2 ) EZ(~); 8:$(2)
&](I);A : d ( 2 ) E I ( ~ ) .[After Pastor ef a/. (1979).1
-+
-+
6 . MAGNETOOPTICS IN
1.o
0.5
259
15
2.0
x (/)
FIG.19. Comparison of transmission minima observed at I = 118.6pm in several oriented
Hgl-,MnxTe samples (H 11 (1 11)) with the predictions of the theory. (a:transitions observed for
E 11 H; 0: transitions observed for E I H; 0 : transitions observed for both E (1 H and E 1 H.)
[After Pastor et d.(1979).1
field varies linearly with the photon energy for the three different laser
wavelengths, the non-parabolicity effects in r s bands are expected to be
negligible. The experimental data were thus analyzed on the basis of the
Luttinger model, with the effect of anisotropy of the Ts bands ( y 2 # y 3 )
included. In this model, the warping contribution introduces a mixing of the
Ts Landau levels at kH = 0, and the k p Hamiltonian D thus becomes an
infinite matrix.
The exchange interaction between mobile carriers and Mn2+ ions was
introduced into the model by using an additional Heisenberg-type
Hamiltonian Xex,(Eq. (2)). In the representation of the matrix D,the
260
C. RICAUX
exchange interaction X,, has the form of an infinite matrix that depends only
on the exchange parameter A = - & Z V ~ f l x f ~The
~~~
effective
.
Hamiltonian
D + X,, was solved numerically, and the calculated transition energies were
fitted to the experimental data, treating NODas the only fitting parameter.
In the analysis, the normalized magnetization was taken from interband
magnetooptical data, and the Luttinger parameters were estimated from
those obtained for HgTe (Pastor et al., 1978), taking into account the
composition dependence of the energy gap. Comparison of the experimental
resonances for different compositions with the prediction of the Luttinger
model is shown in Fig. 19, obtained from theoretical fits of the data that
yielded the value of Nofl as 650 meV.
(b) Hgl-,-,Cd,Mn,Te
The quaternary system Hgl -,,Cd,Mn,Te
offers the interesting and
important possibility of independently varying the energy gap and the
magnetic ion concentration in DMS. The first magnetooptical study of
quaternary alloys of low Mn content ( y < 0.01) was reported by Kim et al.
(1982). FIR magnetotransmission measurements were carried out at 4.2 and
1.6 K, in both Faraday and Voigt geometries, on n-type samples of low
electron concentration (8 x 1014~ m - ~Energies
).
of the resonance lines vs.
magnetic field are shown in Fig. 20 for alloys with interaction gaps EO = - 26
and - 56 meV, respectively. The FIR spectra consist of several Ts" TsC
transitions, as well as the spin, combined, and cyclotron resonances. In the
low field region, the crossing of the cyclotron and spin resonances lines
directly demonstrates crossing of a(1) by b(0) conduction levels. The band
edge effective masses and gyromagnetic factors of conduction electrons are
estimated from these experiments for several compositions. The g factors are
found to be twice as large in the quaternary semimagnetic alloys as those in
Hgl -,Cd,Te alloys ofidentical interaction gap, whereas the effective masses
remain comparable in both alloys.
-+
FIG. 20. Energies of transmission minima vs. magnetic field for quaternary Hg,-,-,Cd,Mn,Te
alloys of interaction gaps
= -26meV (a) and EO = -56meV (b). CR, CO, SR, and 11.2.3 denote cyclotron resonance, combined resonance, spin
resonance, and interband transitions, respectively. [After Kim et al. (1982).]
EO
262
C . RIGAUX
FARADAY
I
I.o
0.5
I.5
50-
119pm
6.
263
field shifting by about 50% in the same temperature range. This feature is
evident in Fig. 24, which shows the temperature variation of the spin-flip
resonance at 119pm. The strong temperature shift of the spin-flip transitions
represents a direct influence of the (&)-dependent term on the energy
scheme of the T8 spin sublevels due to the p-d exchange interaction.
The experimental data were quantitatively interpreted in terms of the
modified Pidgeon-Brown model. A generalized Brillouin function for an
effective temperature
T(see, Gaj et al., 1979) was used to describe the
thermal average of the Mn spins. The parameter To, as well as the fundamental band structure parameters EO and P , were treated as adjustable
parameters in the numerical fitting procedure of the transition energies vs.
magnetic field. The values of the remaining band parameters were taken from
those determined earlier for HgSe (Pastor et al., 1981; Dobrowolska et al.,
1980a) and the values of the exchange integrals a,p were taken from the
quantum transport results on dilute HgMnSe alloys (Byszewski et al., 1980).
The best theoretical fit of the transition energies vs. magnetic field to the
experimental line positions at 4.2K is shown in Fig. 23. Two groups of
transitions are distinguished: at higher energies are combined resonances,
corresponding to a@) b(n + l), and at lower energies occur the EDSR
transition a(n) b(n). The quantum numbers n, n, ascribed to each
transition, are indicated in the figure. Since the Fermi level lies at about
16 meV above the r point, several spin-flip transitions are possible in the
investigated spectral region. From the fitting procedure, Witowski et a/.
determined(at4.2K)~o= -83.1 meV,P= 5.58 10-8eVcm,and%= 7.23K
for x = 0.032. The temperature dependence of the resonance positions for
several wavelengths was subsequently quantitatively interpreted by fitting
%(T) as the only fitting parameter. Comparison between the calculated and
experimental line positions for I = 119pm is shown in Fig. 24 indicating
excellent agreement. From theoretical fits performed at different wavelengths, the temperature dependence of the parameter %( T ) was estimated.
The results show that the parameter To( T ) increases from 6 K to about 11 K
between 2 and 6.5 K.
Conduction band Landau levels at the r-point, obtained from the analysis
of spin-flip transitions in Hgo.97Mno.o~Se
at T = 1.6 and 6 K, are depicted
in Fig. 25. The b(n)levels are the most strongly affected by the temperature,
their energies increasing when the temperature is lowered, similarly as in zero
gap HgMnTe alloys. The behavior of the T8 Landau levels explains the
important shift of the EDSR at I = 119pm (a(0) b(0);Fig. 22), as well as
the transformation of the transition 4 2 ) b(2) into a(1) b(1) observed at
L = 163pm when the temperature is raised from 1.6 to 6 K. As one can see
in Fig. 23, the spin splitting b(n) - a(n) exceeds the cyclotron splitting
a(n + 1) - a(n) in the investigated temperature range.
-+
264
C. RIGAUX
O.8.5
I.o
1.5
MAGNETIC FELD
20
2.5
3bO
(TI
FIG.23. Transition energies vs. magnetic field for EDSR and combined resonance at 4.2 K
observed in Hgl-,MnxSe (x = 0.032). Symbols are experimental. Solid lines represent the best
fits to the experimental data. Unprimed and primed integers refer to EDSR and combined
resonance transitions, respectively. [After Witowski et at. (1982).]
6.
265
2.2
2.0
2
0
-I-
1.8-
1.6-
1.4-
z 1.2-
wLL
9
I-
1 I .o
0.8:
I
266
C . RIGAUX
MAGNETIC FIELD ( T I
b(n) (or to excited donor states bound to these levels). The sensitivity of the
combined resonance harmonic transitions b(0) a(n) to the exchange
parameter A has made it possible to determine the magnetic field variation
A ( H ) up to 18 Tesla. Satisfactory agreement is also found with the interband
magnetoabsorption data obtained at lower fields for HgMnTe of similar
composition.
Low temperature magnetoabsorption and stimulated spin-flip Raman
scattering experiments in n-type Hgo.g9Mno.l1Tewere also reported by Geyer
and Fan (1980). Weak magnetotransmission structures observed at photon
energies smaller than the energy gap were attributed to cyclotron resonance
harmonic transitions coupled with LO phonon emission. A line observed in
scattering measurements under applied magnetic field was found to be due
to spin-flip scattering, and the magnitude of the Land6 factor of conduction
electron was estimated as [gel = 45. The observed spin-flip scattering was
determined to be stimulated on the basis of its intensity dependence on the
+
6.
261
laser excitation intensity and the calculated photon occupation number of the
scattered radiation.
V. Results of Magnetooptics
9. BANDPARAMETERS
*O
0
X
2100
oOl
x
Hg,,
Mn, TI .
10
15
(%I
FIG.26. Interaction gap vs. composition in Hg,-,MnlTe alloys for T = 4.2 K. Black circles
are from Bastard et a1 (1978, 1981), open circles from Dobrowolska et at. (1981), crosses from
Jaczynski et at. (1978), and triangles from McKnight et at. (1978).
268
C. RIGAUX
-200
-070
005
--"'OO
FIG. 27. Energy gap vs. composition in Hgl-,Mn,Se for T = 10 K . Open circles are from
interband magnetooptical data (Dobrowolska et al., 1981 ) ; full circles are from quantum
transport (Takeyama and Galazka, 1979).
- I
-I
-300 -200 -100
100
200
300C&meV
FIG.28. The momentum matrix element P v s . the interaction gap for Hgl-,Mn,Te alloys for
T = 4.2 K. 0 are from Bastard et al. (1978, 1981); 0 are from Dobrowolska el al. (1981);
3 are from Jaczynski er al. (1978).
and C
The Kane matrix element P = - i h / m ( S IpzlZ> was determined from theoretical fits of interband and intraband experiments for several compositions.
The results are presented in Figs. 28 and 29 together with those obtained from
Shubnikov-de Hass measurements. In both Mn-based systems, the interband momentum matrix element P shows a systematic decrease with
x , unlike the behavior observed for Hgl-,Cd,Te
(Guldner et al.,
1978).
6.
269
0
h
5
2
80
90
470,
..
++4I
10. EXCHANGE
PARAMETERS
TABLE I1
EXCHANGE
CONSTANTS
IN NARROWGAP DMS. (IBMO: INTERBAND
MACNETOOPTICS;
FIRMO: FARINFRARED
MAGNETOOPTICS.)
THE CONSTANTS
MODEL(PB) OR THE LUTTINGER
MODELOF THE BANDSTRUCTURE
WEREDERIVEDUSINGTHE PIDCEON-BROWN
T (K)
NOD(ev)
NOCU
(eV)
0.0014
0,004-0.02
0.12-0.17
0.02; 0.03; 0.04
0.007; 0.01 1
x < 0.02
2-4.2
2-4.2
2-4.2
4.2
10-65
4.2
-0.5
-0.5
-0.8
1.5
- 0.75
1.4
0.6
0.65
- 0.7
Hg,-,Mn,Se
10-40
1.6-30
1.6-30
1.6-30
4.2
4.2
0.7
1.54
1.35
0.92
0.9
0.9
- 0.4
- 0.9
- 0.96
-0.32
- 0.35
-0.35
Hgl-,Fe,Te
5 0.016
2-4.2
Compound
Hgl-,Mn,Te
Composition x
- 0.3
- 0.4
Experiments
IBMO (PB) Bastard et al. (1978)
IBMO (PB) Bastard et a / . (1978)
IBMO (PB) Bastard et al. (1981)
Quantum transport (PB) Jaczynski et a / . (1978)
lBMO (PB) Dobrowolska & Dobrowolski (1981)
FIRMO (Luttinger) Pastor et al. (1981)
IBMO (PB) Dobrowolska et a / . (1981)
Quantum transport (PB) Takeyama & Galazka (1979)
Quantum transport (PB) Takeyama & Galazka (1979)
Quantum transport (PB) Takeyama & Galazka (1979)
Quantum transport (PB) Byszewski et al. (1980)
Quantum transport (PB) Byszewski et al. (1980)
IBMO (PB) Serre et al. (1982)
6.
271
272
C. RIGAUX
2o
"0
FIG. 30. Magnetic phase diagram of Hgl-xMnxTe, P and S denoting the paramagnetic and
spin-glass regions, respectively. The origin of the data points is as follows: crosses: ac susceptibility (Mycielski et al., 1984); open circles: Faraday rotation experiments (Mycielski et al.,
1984); open triangles: dc susceptibility (Nagata et al., 1980); full triangles: Brandt et al. (1983);
stars: Faraday rotation data for Hgo.s~Mn0.~5Cdo.zoTe
and Hgo.~oMno.&do.~oTe(Mycielski et
al., 1984).
References
Andrianov, D. G., Gimelfarb, F. A., Kushnir, P. I., Lopatinskii, I . E., Pashkovskii, M. V.,
Savelev, A. S . , and Fistul, V. I . (1976). Sov. Phys. Semicond. 10,66. (Fit. Tekh. Pafuprov.
10, 111.)
Bastard, G., Rigaux, C., and Mycielski, A. (1977). Phys. Stat. Sol. (b) 79, 585.
Bastard, G . , Rigaux, C., Guldner, Y., Mycielski, J., and Mycielski, A. (1978). J. Physique
39, 87.
Bastard, G. (1978a). These de Doctorat d'Etat, Paris (unpublished).
Bastard, G. (1978b). (14th Int. Conf. Phys. Semicond.) Institute ofphysics Con$ Series 43
(Wilson, London), 231.
Bastard, G . , and Lewiner, C. (1979). Phys. Rev. B 20, 4256.
Bastard, G., Gaj, J. A,, Planel, R., and Rigaux, C. (1980). J. Physique 41, C5, 247.
Bastard, G., Rigaux, C., Guldner, Y . , Mycielski, A., Furdyna, J. K., and Mullin, D. P. (1981).
Phys. Rev. B 24, 1961.
Brandt, N. B., and Moshchalkov, V. V. (1984). Adv. Phys. 33, 193.
Brandt, N . B., Moshchalkov, V. V . , Orlov, A. O . , Skrbek, L., Tsidil'kovskii, I. M., and
Chudinov, S. M. (1983). Sov. Phys. JETP57, 614. (Zh. Eksp. Teor. Fiz. 84, 1050.)
6.
273
Byszewski, P., Cieplak, M. Z., and Mongird-Gorska, A. (1980). J. Phys. C 13, 5383.
Chen, Y .-F., Dobrowolska, M., Furdyna, J . K., and Rodriguez, S. (1985).Phys. Rev. B32,890.
Davydov, A. B., Noskova, L. M., Ponikarov, B. B., andugodnikova, L. A. (1980). Sov. Phys.
Semicond. 14, 869. (Fiz. Tekh. Poluprov. 14, 1461.)
De Seze, L. (1970). J. Physique 10, L353.
Dietl, T. (1981). Solid State Science 24, 344.
Dobrowolska, M., Dobrowolski, W., and Mycielski, A. (1980a). SolidState Commun. 34,441.
Dobrowolska, M., Dobrowolski, W., Otto, M., Dietl, T., and Galazka, R. R. (1980b). (15th
Int. Conf. Phys. Semicond.) J. Phys. SOC.Japan 49, Suppl. A 477.
Dobrowolska, M., and Dobrowolski, W. (1981). J. Phys. C 14, 5689.
Dobrowolska, M., Dobrowolski, W., Galazka, R. R., and Mycielski, A. (1981). Phys. Stat. Sol.
(b) 105, 477.
Furdyna, J. K. (1982). J. Vac. Sci. Technol. 21, 220.
Gaj, J. A., Byszewski, P., Cieplak, M. Z., Fishman, G., Galazka, R. R., Ginter, J., Nawrocki,
M., Nguyen The Koi, Planel, R., Ranvaud, R., Twardowski, A. (1978). (14th Int. Conf.
Phys. Semicond.) Institute of Physics Conf. Series 43 (Wilson, London), p. 1113.
Gaj, J . A., Planel, R., and Fishman, G. (1979). Solid State Commun. 29, 435.
Gaj, J. A. (1980). (15th Int. Conf. Phys. Semicond.) J. Phys. SOC. Japan 49, Supple. A, 747.
Series 43 (Wilson,
Galazka, R. R. (1978). (14th Int. Conf. Semicond.) Instituteo~PhysicsCon~.
London), 133.
Galazka, R. R., and Kossut, J. (1980). In Narrow Gap Semiconductors: Physics and
Applications, Lecture Notes in Physics 133, p. 245. Springer, Berlin.
Galazka, R. R., Nagata, S., and Keesom, P. H. (1980). Phys. Rev. B 22, 3344.
Galazka, R. R. (1982). In Physics of Narrow Gap Semiconductors (4th Int. Conf.), Lecture
Notes in Physics 152, p. 294. Springer, Berlin.
Geyer, F. F., and Fan, H. Y. (1980). IEEE J. of Quantum Electron. QE16, 1365.
Grynberg, M., Martinez, G., and Brunel, L. C. (1982). Solid State Cornmun. 43, 153.
Guldner, Y., Rigaux, C., Grynberg, M., and Mycielski, A. (1973). Phys. Rev. B 8, 3875.
Guldner, Y., Rigaux, C., Mycielski, A., and Couder, Y. (1977). Phys. Stut. Sol. (b) 81,615 and
Phys. Stut. Sol. (b) 82, 149.
Holm, R. T., and Furdyna, J. K. (1977). Phys. Rev. B 15, 844.
Kacman, P., and Zawadzki, W. (1971). Phys. Stat. Sol. (b) 47, 629.
Jaczynski, M., Kossut, J., and Galazka, R. R. (1978). Phys. Stat. Sol. (b) 88, 73.
Kane, E. 0. (1957). J. Phys. Chem. Sol. 1, 249.
Khattak, G. D., Amarasekara, C. D., Nagata, S., Galazka, R. R., and Keesom, P. H. (1981).
Phys. Rev. B 23, 3553.
Kim, R. S., and Narita, S. (1976). Phys. Stat. Sol. (b) 73, 741.
Kim, R. S., Mita, Y., Takeyama, S., and Narita, S. (1982). I n Physics of Narrow Gap
Semiconductors (4th Int. Conf.), Lecture Notes in Physics 152, p. 316. Springer, Berlin.
Kierzek-Pecold, E., Szymanska, W., and Galazka, R. R. (1984). Solid State Commun. 50,658.
Kossut, J. (1976). Phys. Stat. Sot (b) 78, 537.
Leibler, K., Sienkiewicz, A., Checinski, K., Galazka, R. R., and Pajaczkowska, A. (1978). I n
Physics of Narrow Gap Semiconductors (3rd Int. Conf.), p. 199. Polish Scientific
Publishers, Warsaw.
Luttinger, J. M. (1953). Phys. Rev. 102, 1030.
Mahoney, J., Lin, C., Brumage, W., and Dorman, F. (1970). J. Chem. Phys. 53, 4286.
McCombe, D. B. (1969). Phys. Rev. 181, 1206.
McCombe, D. B., Wagner, R. J., and Prinz, G. A. (1970). Phys. Rev. Lett. 25, 87.
McCombe, D. B., and Wagner, R. J. (1971). Phys. Rev. B 4 , 1285.
McKnight, S. W., Amirtharaj, P. M., and Perkowitz, S. (1978). Solid State Commun.25,357.
274
C . RIGAUX
CHAPTER 7
I . INTRODUCTION:POSSIBILITIES
OPENEDBY STUDIES
OF LARGE
GAPDMS . . . . . . . . . . . . . . . . . . . . .
1 . Optical Access to Magnetic Properties. . . . . . . . .
2 . Studies of Excitons in Magnetic Field . . . . . . . . .
3 . Precise Investigation of Selection Rules . . . . . . . .
TECHNIQUES
. . . . . . . . . . . . . .
I1 . EXPERIMENTAL
4 . Faraday Rotation . . . . . . . . . . . . . . . . .
5 . Reflectivity . . . . . . . . . . . . . . . . . . .
6 . Transmission Measurements . . . . . . . . . . . .
I . Photoluminescence . . . . . . . . . . . . . . . .
8 . Other Techniques. . . . . . . . . . . . . . . . .
111. EXCHANGE
SPLITTING
OF THE FREEEXCITON
GROUND
STATE .
9 . Splitting Valuesand Selection Rules . . . . . . . . . . .
10. Relation of Splittings to Magnetization of the Ion System
1 1 . Transitions from the Spin-Orbit Split-Off Component of
the Valence Band . . . . . . . . . . . . . . . . .
IV . THEORETICALDESCRIPTIONOF BASIC MAGNETOOPTICAL
PROPERTIES
OF DMS . . . . . . . . . . . . . . . . .
12. Hamiltonian of the Problem and Typical Approximations
13 . Band EIectron States Split by Exchange . . . . . . . .
14. Zeeman Splitting of the Free Exciton Ground State
. .
15 . Faraday Rotation . . . . . . . . . . . . . . . . .
16. Analysis of the Mean Field Approximation in Terms of the
Magnetic Polaron . . . . . . . . . . . . . . . . .
V . DETAILED
MAGNETOABSORPTION
STUDIES
OF EXCITON
STATES
. .
17. Experiment Results . . . . . . . . . . . . . . . .
18 . TheoreticalAnalysis . . . . . . . . . . . . . . . .
VI . EXCITON
LUMINESCENCE
. . . . . . . . . . . . . . . .
VII . TRANSITIONS
AWAYFROM
THE BRILLOUIN
ZONECENTER. . .
19. Experimental Data . . . . . . . . . . . . . . . .
20 . Attempts of Theoretical Description . . . . . . . . .
REFERENCES . . . . . . . . . . . . . . . . . . . .
275
216
216
211
211
211
211
219
219
282
282
283
283
286
281
288
288
290
293
296
291
300
300
303
303
305
305
306
308
276
J . A. GAJ
OPTICAL
ACCESSTO MAGNETIC
PROPERTIES
FIG. 1 Faraday rotation of Cdl-xMn,Te versus magnetic field with microwave excitation for
two wavelengths (1 and 2). Note a drop in rotation at 3.5 kOe (resonance field). [After Komarov
et al. (1977).]
7.
MAGNETOOPTICAL PROPERTIES
277
2. STUDIES
OF EXCITONS
IN MAGNETIC
FIELD
Due to high values of the Zeeman splittings in semimagnetic semiconductors, absorption lines observed in different polarizations are sufficiently
separated to resolve and study weak lines due to forbidden transitions (they
are, typically, lo2times weaker than the allowed transitions). This possibility
has been used by Nguyen The Khoi et al. (1983), who explained the observed
asymmetrical splitting of the absorption edge of Cdl-,Mn,Te in a magnetic
field by a small contribution from forbidden transitions. This contribution
has been measured as a function of magnetic field. The origin of the deviation
from strict selection rules remains still an open question.
11. Experimental Techniques
4. FARADAY
ROTATION
278
J. A. GAJ
bothering to turn the polarizers, to plot the intensity of the transmitted light
as a function of magnetic field. Subsequent minima of the resulting curve
correspond to multiples of of the rotation angle. Furthermore, varying the
wavelength at a fixed magnetic field one obtains also an oscillating curve,
permitting a quick determination of the whole Faraday rotation spectrum
from a single recorder trace (Fig. 3). In some experiments, however, it is
desirable to measure small rotation angles, of the order of minutes of arc.
This is the case in the spin-glass phase magnetooptical studies, where use of
a small magnetic field is of a great importance. Small rotations occur also
in reflection of polarized light from a crystal placed in a magnetic field-the
so-called magneto-optical Kerr effect.
In small angle measurements various modulation techniques prove useful.
An example is shown in Fig. 4. An oscillating plate consisting of two pieces
of polarizing foil with mutually perpendicular polarization directions is
placed in the optical system in such a way that the outgoing light passes
alternatively through two pieces of the foil. If the incident light is polarized
along a direction at n / 4 to polarization directions of both pieces of foil and
for zero Faraday rotation, no modulation will arise. For small deviations of
the outgoing light polarization from the equilibrium direction, a modulation
Md
sc
FIG. 2. Experimental setup for Faraday rotation measurements in semimagnetic semiconductors. The elements of the setup are designated as follows: Sc-source, P-polarizers,
C-superconducting coil, S-sample, Md-modulator, M-monochromator, D-detector,
E-electronics.
0.77
078 0.79
2 (pm)
7.
MAGNETOOPTICAL PROPERTIES
219
the plate marked by arrows. The incident light polarization is shown by the bottom arrow.
280
J. A. GAJ
I64
I63
I62
I61
PHOTON ENERGY ( e V )
FIG.6. Reflectivity spectra of Cdl-,MnxTe at 1 . 1 Tin the Faraday configuration (a' and upolarizations) and in the Voigt configuration (n polarization). [After Gaj el of. (1979).]
7.
MAGNETOOPTICAL PROPERTIES
281
c+
B=5T
1.61
1.62
FIG.8. Sample etching arrangement. The optical system is used to interrupt etching when the
sample becomes sufficiently thin to transmit light (Sc-source, L-lens, E-etching solution,
B-bar
of a magnetic stirrer, S-sample, P-plastic
screens, M-microscope of low
magnification).
282
J. A. GAJ
M1
sc
7.
MAGNETOOPTICAL PROPERTIES
283
9. SPLITTING
VALUESAND SELECTION
RULES
Zeeman effect on the free exciton ground (1s) state has been studied in a
number of semimagnetic semiconductors. Results vary depending on crystal
structure of the semiconductor.
a. Cubic Structure
Magnetoreflectivity and magnetotransmission have been studied in a
number of cubic (zinc blende) large gap DMS, such as Cdl-,Mn,Te
(Komarov et al., 1977; Twardowski et al., 1979; Gaj et al., 1978, 1979;
Rebman et al., 1983), Znl-,Mn,Te (Komarov et al., 1978; Twardowski etaf.,
1984) or Znl-,MnxSe (Komarov et af., 1980b; Twardowski et al., 1983). The
exciton ground state in those alloys splits in six components: four (denoted
A through D) visible in the Q polarization (electric field of light perpendicular
to the external magnetic field) and two (E and F) visible in the 72 polarization
(electric field of light parallel to the external magnetic field). In the Faraday
configuration (light propagation along the external magnetic field), the four
a-components exhibit a' (A and B) or a- (C and D) circular polarizations
(see also Fig. 16).
The whole splitting pattern is approximately symmetric relative to the zero
field position. Components B and C are significantly weaker than A and D.
The lowest-lying component A is much sharper (especially at high fields) than
the remaining ones.
The features described above are visible in Figs. 6 and 7 for reflectivity and
transmission, respectively, of Cdl -,Mn,Te.
For a given semiconductor in a given magnetic field and temperature, the
Zeeman pattern can be characterized by splitting values of two pairs of acomponents: strong (A and D) and weak ones (B and C). The splitting of the
two n components, E and F, is not independent. It equals one half of the
difference between splittings A-D and B-C. The ratio of the two latter
splittings has been found to be independent of magnetic field or temperature
for a given semiconductor.
The polarization selection rules are quite strong. However, absorption
measurements allow one to observe very small deviations from perfect
selection rules. Nguyen The Khoi el al. (1983) have observed a small (of the
order of 1%) admixture of Q- polarization in some a+-polarized Zeeman
components in Cdl-,Mn,Te.
Figure 10shows the positions of the Zeeman components of the free exciton
ground state in a sample of Cdl-,Mn,Te as a function of magnetic field.
Splittings increase linearly at low fields and tend to saturate at higher field
values. That behavior can be examined in detail in Fig. 11, where the splitting
284
J. A. GAJ
of the strong D components A and D is plotted vesus magnetic field for various
compositions. In crystals of low Mn mole fraction x , the splitting increases
with x and the saturation is clearly visible, while at higher x values the curves
show only a slight deviation from linearity. At higher temperatures the values
of the splittings becomes smaller and the tendency to saturate is weaker.
FIG. 10. Energy of the Zeernan components of the exciton versus magnetic field for
CdI-,Mn,Te in the Faraday configuration. Continuous lines are drawn to guide the eye. [After
Twardowski (1981b).]
FIG. 11. Splitting of the two strong u components of the exciton Zeernan structure of
Cdl-,Mn,Te versus magnetic field at 1.4 K. Manganese concentration values are given in
percentage. [After Gaj e t a / . (1979a).]
7.
285
MAGNETOOPTICAL PROPERTIES
b. Hexagonal Structure
Magnetoreflectivity and magnetoabsorption measurements have also been
reported on hexagonal (wurtzite) semimagnetic semiconductors, such as
Cdl-,Mn,Se (Arciszewska and Nawrocki, 1982; Aggarwal et al., 1983;
Komarov et al., 1980c) or Cdl-,Mn,S (Gubarev, 1981). The results show a
behavior generally more complicated than in the cubic case. A relatively
simple situation can be obtained by applying the magnetic field along the
hexagonal axis of the crystal. The resulting splitting pattern and selection
rules are then analogous to those in cubic crystals', except for a displacement
of the two strong o components relative to the remaining ones due to the
crystal axial anisotropy (see Fig. 12). The variation of the effect with
composition and temperature are also like in cubic semimagnetic semiconductors.
For an arbitrary direction of the magnetic field, the results are more
complicated.
-2
2.31
MAGNETIC FIELD f T )
FIG. 12. Energy of the Zeeman components of the exciton versus magnetic field for
Cdl-,Mn,Se in the Faraday configuration. Magnetic field is parallel to the crystal hexagonal
axis. [After Arciszewska and Nawrocki (1984).]
CdS seems to be an exception. The strong u+ component, being narrower (as in other alloys)
is placed above its u- counterpart (Gubarev, 1981).
286
J. A. GAJ
10. RELATIONSHIP
OF THE SPLITTINGS
TO THE MAGNETIZATION
OF THE IONSYSTEM
AX( S z ) .
x 0.5
-0
0
+
2
5
0 10
A 20
30
I
-10'~
-10-
-10-
FIG. 13. Splitting of the two strong u components of the exciton Zeeman structure of
Cdl-,Mn,Te versus magnetization, expressed in terms of the mean value of the Mn spin (per
cation). The straight line corresponds to a proportionality with a slope of 1.10 eV. Manganese
concentration is indicated in percentage. [After Gaj et a/. (1979b).]
7.
MAGNETOOPTICAL PROPERTIES
11. TRANSITIONS
FROM THE SPIN-ORBIT SPLIT-OFF
OF THE VALENCEBAND
COMPONENT
An exciton state can also be created with a hole from the lower component
of the valence band, split off by the spin-orbit interaction. Such a state is
more or less difficult to observe depending on the value of the spin-orbit
splitting A. It is quite easy to see in CdS (A = 0.065 eV), fairly easy in CdSe
(A = 0.45 eV) whereas in CdTe (A = 0.95 eV) it takes some effort to observe
it. In an alloy, disorder introduces an additional broadening, making its
observation more difficult.
In CdTe an exciton absorption line associated with transitions from the r7
valence band has been observed by Twardowski et al. (1980). Substituting
2% of Cd by Mn broadens the structure enough to make the absorption peak
disappear (only a threshold remains, as seen in Fig. 14). Any magnetooptical
measurements on such an ill-defined structure must be inaccurate. Within
2.550
2.600
[eV]
ENERGY
FIG. 14. Transmission versus photon energy for CdTe ( B = 0) and Cdo.9~Mn~.o~Te
(B = 0
and 4 T ) . Straight lines were drawn to determine transition energies (marked by arrows). [After
Twardowski ef al. (1980).]
288
J.
A.
GAJ
12. HAMILTONIAN
OF THE PROBLEM
AND TYPICAL
APPROXIMATIONS
We shall look for solutions of the eigenproblem of a band electron in a
given external magnetic field and a given state of the magnetic ion system.
Such an approach can be justified if the band electron wave function is
sufficiently delocalized to make an influence of the electron on each of the
magnetic ions negligible.
The Hamiltonian for a band electron will have the form,
Ho
+ HB + HC + Hex + Heh
(1)
H,
P2
=-
2m
+ V(r)
The term
e
HB = - - p ' A
mc
+ g/fBb'B
(3)
Hex =
-c J(r
n
- R,)s~
-G
(4)
7.
MAGNETOOPTICAL PROPERTIES
289
He, =
-Ci
J(r - R i ) x a Z ( S Z ) .
(5)
It must be pointed out that, within the framework of the effective mass
theory, the above Hamiltonian has the form identical to the spin part of HB
(Zeeman energy of the electron spin). Therefore the introduction of the
exchange term He, does not in principle create any new formal difficulties
(within the approximation outlined above) compared to the non-magnetic
case.
Let us make now another essential approximation, neglecting the term HB.
This is justified in wide gap DMS when the temperature is not excessively
high, i.e., when the magnetic field effects due to carrier-ion exchange are
strong compared to nonmagnetic effects described by H B . Now the
Landau quantization is absent from the problem, and we may speak of the
electron wave vector k in the presence of a magnetic field.
In what follows we present a k * p calculation of the energy bands, using
the following further approximations:
(1) The k * p term is included as a second order perturbation (Luttinger
model). In the absence of exchange, this leads to parabolic bands;
(2) Warping of the valance band is neglected, as well as effects due to lack
of inversion symmetry of the crystal;
(3) Spin-orbit splitting of the valence band is assumed to be large
compared to the ion-carrier exchange energy.
290
J. A . GAJ
13. BANDELECTRON
STATES
SPLIT BY
EXCHANGE
Let us consider a cubic zinc blende crystal. Under the above assumptions,
we can treat separately the conduction band and each component of the
valence band. Following Bir and Pikus (1972), we shall use as basis vectors
for the conduction (r6)band:
I),
+)C
= st
(Wv3/2IHexlWu3/2)
where
3
O
0
0
O
B
O
0 - B
0
O
O
0
-3B
7.
MAGNETOOPTICAL PROPERTIES
291
Here No denotes the number of unit cells per unit volume, and a = (slJls)
and p = ( X l J l X ) are exchange integrals for the conduction and valence
bands, respectively.
Within the assumed approximations the total Hamiltonian will now
consist of on!y two terms,
H = Ho + H e x .
(13)
In the k * p calculation the term H , will be represented by
Let us point out that at k = 0, where the k p term vanishes, the presence
of exchange produces two-fold or four-fold and equal splittings of bands,
with the splitting values proportional to x( S z ) and therefore (if diamagnetic
effects are neglected) to the magnetization of the crystal.
The solutions of the problem follow immediately for the conduction, as
well as the lower valence bands. In those cases the entire Hamiltonian
matrices (Hexand k * p terms) are diagonal, where the energies
Ec = E,"
h2k2
++- 3A
2mc
and
- [(X + iY)l
d3
-i
vui/2(k) = - [(X - i Y ) t
\r5
+ Zt]eIb
- Zl]eih
(18)
292
J. A. GAJ
(A4+ f N ) k 2 - SN(k:
+ k,?)
(--
h2k2 + 3 h2 - Gh2) ( k :
E2,3 = -2mlh 4 2mrh
+ k,?) f B.
\--I
We thus obtain four bands whose constant energy surfaces are ellipsoids:
cucumber-(E1,4) or disc-like (E2,3).
The diagonalization of the k p Hamiltonian alone produces the well
known spherical and parabolic light and heavy hole bands:
h2k2
Ehh(k) = E," + Mk2 = E," - ---,
2mhh
The wavefunctions corresponding to Eq. (23) and (24) can be easily obtained
by a suitable rotation of the basis functions. Note that for k along the z-axis
the Hamiltonian given by Eq. (19) is diagonal, considerably simplifying the
calculations (see, e.g., Kane, 1957).
7.
MAGNETOOPTICAL PROPERTIES
293
t
0
- 10
-20
3
+lC;
Z("= 100)-
FIG. 1s. Energy band structure of Cdl-,Mn,Te in dimensionless energy ( E ) and wave vector
(K)coordinates at two values of the angle between the wave vector and the magnetization: 90"
(KJ and 10". IAfter Gaj e t a / . (1978).]
B) = Eih(k) f B d 4
Ehh(k,
3 COS2 8
B) = E l h ( k ) f 3 8 C O S 6
(25)
(26)
where B is the angle between the wave vector k and the magnetization
direction.
A numerical diagonalization of the Hamiltonian matrix in a general case
produces a highly nonparabolic and anisotropic band structure. This is
shown in Fig. 15 for two directions of the wave vector: at angles 90" and 10"
to the magnetization direction.
14. ZEEMANSPLITTING
OF FREEEXCITON
GROUNDSTATE
The problem of exciton states in the presence of an exchange field has been
treated, e.g., by Cieplak (1980), who followed an approach developed by
Altarelli and Lipari (1973). We shall not enter here into the details of the
complete exciton calculation, but will simply note that in the effective mass
approximation the Coulomb potential of the electron-hole interaction HC
does not mix the basis functions, enabling us to think of an exciton with a
hole from one of the exchange-split valence subbands and an electron from
one of the conduction subbands, if the electron-hole exchange is to be
neglected. This allows us to describe excitonic magnetooptical splittings in
terms of interband transitions. Remembering that direct exciton creation
294
J . A. GAJ
cond. G
band
valence
bands
G
r,
7.
295
MAGNETOOPTICAL. PROPERTIES
TABLE I
CALCULATED
ENERGIES
AND RELATIVE
INTENSITIES
OF THE ZEEMAN
COMPONENTS
OF
THE FREEEXCITON
IN A ZINCBLENDE
SEMIMAGNETIC
SEMICONDUCTOR
States
Valence
Conduction
Energy
(Relative to
zero field
position)
3B
B
B
-B
-B
-3B
-B
-B
B
B
Polarization
- 3A
U+
+ 3A
-
U+
3A
+ 3A
- 3.4
+ 3A
+ 3A
- 3A
+ 3A
- 3A
U
U
U+
n
0
Relative
intensity
Symbol
3
1
2
2
1
3
2
1
I
2
A
B
E
F
-'
A = --x NO(Y(S')
B = sxNoP(S'>
Attention: (S') is normally negative!
TABLE I1
EXCHANGE
CONSTANTS
IN VARIOUS
LARGEGAPDMS DETERMINED
FROM MAGNETOREFLECTNITY
OR MAGNETOABSORPTION
COMBINED WITH MAGNETIZATION
MEASUREMENTS
~~
Crystal
structure
Naa
NaP
Compound
[evl
[evl
Cdl-,Mn,Te
zinc blende
0.22
-0.88
Znl-xMnxTe
zinc blende
0.19
- 1.09
Znl-xMnxSe
zinc blende
0.29
- 1.4
Cdl-,MgSe
wurtzite
0.23
Cdl-,Mn,S
wurtzite
Reference
1.27
5.6"
Gubarev (1981)
'A positive value was obtained as a consequence of defining the Heisenberg Hamiltonian
without a minus sign [see Eq. (4)]. Both the sign and the unexpectedly high absolute value must
be regarded with caution (see also the footnote in Section 9).
296
J . A. GAJ
exchange constant 3/ for all the valence band functions (see Table 11). In
principle, two different values px= ( X I J I X ) and pz = ( Z l J l Z ) should be
allowed, since the X and Z functions are no longer equivalent.
For an arbitrary direction of the magnetic field, the exchange splittings are
no longer proportional to the magnetization, since the two perturbations,
crystal field and exchange field, are now represented by non-commuting
Hamiltonians of comparable magnitude. The situation at k = 0 resembles
the eigenproblem in a cubic semiconductor k # 0, which has rather complicated non-linear solutions, as mentioned above.
In general, magnetooptical studies of hexagonal DMS also permit one to
find (by extrapolation) parameters of the band structure of the host nonmagnetic semiconductor. The assumption that px = pz can be discarded if
transitions from the spin-orbit split-off component of the valence band are
also studied.
15. FARADAY
ROTATION
Using the results of the previous sections, one can now calculate Faraday
rotation spectra. This has been done using the excitonic absorption model
developed by Elliott (1957). The model predicts a series of discrete hydrogenic lines corresponding to creation of consecutive exciton bound states of
energies En = Eg - R*/nZ (R*-exciton binding energy) and of oscillator
strength varying as l / n 3 .
Above the energy gap, absorption is a continuous function of the photon
energy, with zero slope at Eg,and slowly rising above that value to approach
a square-root curve at high energies ( A o - Eg S= R*). The value of the
absorption coefficient at ho = Eg is equal to the value obtained by an
infinitesimal broadening of the discrete lines at A o E g . In reality a
broadening must always be expected, so that only a limited number of states
of the hydrogenic series can be distinguished (two states are observable in
Fig. 7). There is no particular difficulty in calculating the Faraday rotation
spectra using the full Elliott model. However, a very simplified approach,
consisting in the calculation of two first nonvanishing moments of the
circular dichroism, gives quite an adequate description of the experimental
rotation spectra near the band gap (Gaj, 1981). In the calculation, performed
in a linear (low field) approximation, the line broadening, as well as any
dichroism above the energy gap, has been neglected. The resulting solution
for the rotation angle 8 is:
7.
MAGNETOOPTICAL PROPERTIES
297
c
0
c
0
035
0.20
log ( E l l eV1
FIG. 17. Faraday rotation spectrum of Cdl-,Mn,Te fitted to Eq. (4): x = 0.05,T = 77 K,
= 0.12, Eg = 1.66eV.
KO
where
KO
d(i 4)
Eo = Eg - 2R*[
n=l
4
n
n = n
~
i 41
n
n=1
= Eg - 1.14R* (28)
298
J . A. GAJ
=
I
N.32
gpB/ kT=0
..........
-----
ll
0-
(b)
FIG. 18. Exciton absorption spectra calculated by Gaj and Golnik (1985): (a) zero field, no
broadening; (b) broadening included.
7.
MAGNETOOPTICAL PROPERTIES
299
FIG. 19. Splittingof the two strong acomponents of the excitonic Zeeman structure measured
in Cd,-,Mn,Te (a), compared with a calculation for N = 32 Mn ions interacting with the exciton
(b). [After Gaj and Golnik (1985).]
300
J. A. G k l
The model discussed above includes a localization of the exciton. The nature
of this localization remains an open question. One of the possible explanations is to view it as due to alloy composition fluctuations.
V. Detailed Magnetoabsorption Studies of Exciton States
17. EXPERIMENTAL
RESULTS
1,700
-2
Y
>
a
W
z
W
1,650
7.
MAGNETOOPTICAL PROPERTIES
301
Let us limit our discussion to the case of strong exchange effects (x 2 0.02
at liquid He temperatures). It turns out that, within each set, the positions
of the lines related to the principal (i = 1) line depend only on the magnetic
field, and not on magnetization.
These two types of behavior, governed by magnetization (ground states of
sets) or by magnetic field (relative position within sets) can be seen in Figs.
21 and 22. In Fig. 21, where magnetization is varied by changing the
temperature at a given magnetic field, constituent lines of the sets A and B
follow the variation of the corresponding ground states A1 and BI. The
magnetization dependent behavior is eliminated in Fig. 22, where the relative
positions within sets A and B are plotted as a function of magnetic field. The
data from various compositions (and thus various magnetization values)
coincide well in the plot.
Such a distinction is not possible in CdTe or in very diluted crystals
( x I0.01), where the exchange energy is comparable to, or smaller than, the
cyclotron energy.
302
J . A. GAJ
I
5'10
*
1Oo/o
B(TJ
(b)
FIG.22. Line positions within sets Ai (a) and Bi (b) versus magnetic field. Continuous curves
correspond to a calculation described in Section 18. [After Twardowski (1981a).]
7.
MAGNETOOPTICAL PROPERTIES
303
18. THEORETICAL
ANALYSIS
The magnetotransmission data reviewed above support the idea that the
exchange and Coulomb terms may be considered independently, as brought
up in Section 14. In order to include excited excitonic states, we can no longer
continue neglecting the orbital part of H B , that must be now added to the
Coulomb terms HC .
We are thus left with the problem of a simple exciton created from a pair
of non-degenerate bands. It should be pointed out that due to a difference
in form between the valence subbands of mj = +* and those of mj = +*,
two types of solutions may be expected in this case, consistently with the
experiment.
A calculation along these lines has been performed by Twardowski and
Ginter (1982), who used an improved version of the adiabatic method. Comparison of their theoretical results with experimental data is shown in Fig.
22, where continuous lines represent results of the calculation. The
agreement is quite good, except for a certain difference in slope for line A1 .
Line A2, interpreted as originating from a 3d-2 exciton state, is not included
in the calculation.
VI. Exciton Luminescence
304
J. A. GAJ
-.-
H = I IT
H = 5.6T
I .oo
0 00
ENERGY ( e V )
k.,
\,'"\-..
."-a,\
I
XMn=
5%
X%
*@.
-2 1650 >
c3
LL
a
'.
I605
W
Z
W
I630
x-x-x-+x_x
0
.
0
'
0
'
0-0-0
7.
MAGNETOOPTICAL PROPERTIES
305
The only magnetooptical results off the Brillouin zone center in a large
gap semimagnetic semiconductor are those obtained at the L point in
Cdl-xMnxTe. Those transitions give rise to reflectivity structures at energies
denoted commonly by E I = 3.5 eV and El + A1 = 4 eV. Figure 25 shows a
relevant fragment of the reflectivity spectrum of Cdl -xMnxTe after ZimnalStarnawska et af. (1984). The first magnetoreflectivity measurements
reported by Dudziak et af. (1982) have already shown that the Zeeman
splitting of the El structure is much smaller than the splittings at the r point.
In fact it was not possible to detect any Zeeman structure at the L point; there
was only a small difference in the structure position when measured with the
(T'
or the (T- polarizations indicating that the splitting (if any) was much
smaller than the structure width. To resolve it, a modulation technique has
been applied. Figure 26 shows an experimental setup with an elasto-optical
modulator used in a later work (Ginter et af., 1983, see also Coquillat et af.,
1986). Results obtained on Cdl-,Mn,Te samples can be summarized as
follows:
(1) At low (pumped liquid He) temperatures, the structures El and
El
306
J. A. GAJ
00
0 05
0 12
0 2
03
04
......- .............
I
2.0
05
.........................
I
I
I
4.0
-,.-
06
"'1
0.7
b.
6.0
8.0
Energy ( e V )
FIG 25. Reflectivity curves of Cdl-,MnxTe at room temperature. El and EI + At peaks are
marked by arrows. [After Zimnal-Starnawska er af. (1984).]
There is clearly a need for systematic and precise magnetooptical measurements of these transitions in other materials and for a better theoretical
model.
20. ATTEMPTS
OF THEORETICAL
DESCRIPTION
It must be said that up to the present moment there exist no satisfactory
explanation of the observed strong reduction of the splitting at the L point.
Certain features of the observed effects have been explained by Ginter et al.
(1983) who used a simple tight binding model to find a relationship between
the exchange splittings and the selection rules at the L point, and those at the
point. The energy bands in the model are developed from the atomic 5s
7.
MAGNETOOPTICAL. PROPERTIES
307
PM
308
J. A. GAI
(3) r he selection rules are much less strict at the L-point than they are at
the center of the Brillouin zone.
Acknowledgments
It is the authors pleasure to thank Dr. J. Ginter for many valuable
discussions, as well as Ms. M. Arciszewska, Drs. M. Nawrocki, and
A. Twardowski for permission to reproduce their unpublished results.
References
Aggarwal, R. L., Jasperson, S. N., Stankiewicz, J., Shapka, Y., Foner, S., Khazai, B., and
Wold, A. (1983). Phys. Rev. B28, 6907.
Altarelli, M., and Lipari, N. 0. (1973). Phys. Rev. B7, 3798.
Arciszewska, M., and Nawrocki, M. (1982). In Physics of Semiconducting Compounds (1 1th
Conf.), p. 225,. Pol. Acad. Sci., Warsaw, Poland.
Arciszewska, M., and Nawrocki, M. (1984). Unpublished.
Bastard, G., Rigaux, C., Guldner, Y., Mycielski, J., and Mycielski, A. (1978). J. Physique 39,
87.
Bir, G. L., and Pikus, G. E. (1972). Simmetria i deformatsionnye effekty vpoluprovodnikakh,
Nauka, Moscow, Sections 23 anll 24 (Symmetry and Strain-induced Effects in Semiconductors, Wiley, 1974).
Cieplak, M. Z. (1980). Phys. Stat. Sol. (b) 97.
Coquillat, D., Lascaray, J. P., Dejardins-Deruelle, M. C., Gaj, J. A., andTriboulet, R. (1986).
Solid State Commun. 59, 25.
Dudziak, E., Brzezinski, J., and Jedral, L. (1982). In Physics of SemiconductingCompounds
(11 Conf.), p. 166. Pol. Acad. Sci., Warsaw, Poland.
Elliott, R. I. (1957). Phys. Rev. 108, 1384.
Gaj, J. A., Ginter, J., and Galazka, R. R. (1978). Phys. Stat. Sol. (b) 89, 655.
Gaj, J. A., Byszewski, P., Cieplak, M. Z., Fishman, G., Galazka, R. R., Ginter, J., Nawrocki,
M., NguyenTheKhoi, Planel, R., Ranvaud, R., and Twardowski, A. (1979a). In Physics
of Semiconductors (14th Int. Conf.), p. 1113, Inst. Phys. Conf. No. 43.
Gaj, J. A., Planel, R., and Fishman, G. (1979b). Solid State Commun. 29, 435.
Gaj, J . A. (1981). In Exchange Interaction of Paramagnetic Ions With Band Electrons in
Cdl-,MnxTe, Editions of University of Warsaw (in Polish).
Gaj, J . A., and Golnik, A. (1985). Acta Phys. Polon. A67, 307.
Ginter, J . , Gaj, J. A., and Le Si Dang (1983). Solid State Commun. 48, 849.
Golnik, A., Ginter, J., and Gaj, J. A. (1983). J. Phys. C16, 6073.
Gubarev, S. I. (1981). Zh. Eksp. Teor. Fiz. 80, 1174. (Sov. Phys. JETP53, 601.)
Kane,. E. 0. (1957). J . Phys. Chem. Sol. 1 , 249.
Kett, H., Gebhardt, W., Krey, U., and Furdyna, J. K. (1981). J. Magn. Magn. Muter. 25,215.
Kierzek-Pecold, E., Szymanska, W., and Galazka, R. R. (1984). Solid State Commun. 50,658.
Komarov, A. V., Ryabchenko, S. M., Terletskii, 0. V., Zheru, I. I., and Ivanchuk, R. D.
(1977). Zh. Eksp. Teor. Fit. 73, 608. (Sov. Phys. JETP46, 318.)
Komarov, A. V., Ryabchenko, S. M., and Vitrikhovskii, N. I. (1978), Pismz v Zh. Eksp. Teor.
Fiz. 27. (JETP Letter 27, 413.)
Komarov, A. V., Ryabchenko, S. M., Terletskii, 0. V., Ivanchuk, R. D., and Savitskii, A. V.
(1980a). Fiz. Tekh. Popuprovod. 14, 17. (Sov. Phys. Semicond. 14, 9.)
7.
MAGNETOOPTICAL PROPERTIES
309
Komarov, A. V., Ryabchenko, S. M., and Terletskii, 0. V. (1980b). Phys. Stat. Sol. (b) 102,
603.
Komarov, A. V., Ryabchenko, S. M., Semenov, Y. G., Shanina, B. D., and Vitrikhovskii, N.
1. (1980~).Zh. Eksp. Teor. Fiz. 79, 1554. (Sov. Phys. JETP52, 783.)
Kossut, J. (1976). Ph.D. Thesis, Institute of Physics, Polish Academy of Sciences, Warsaw.
Lindstrorn, M., Kuivalainen, P., Heleskivi, J., and Galazka, R. R. (1983). Physica 117B and
118B, 479.
Malyavkin, A. V. (1983). Phys. Stat. Sol. (b) 115, 353.
Nguyen The Khoi, Ginter, J., and Twardowski, A. (1983). Phys. State Sol. (b) 117, 67.
Ortenberg, M. von (1984). Solid State Commun. 52, 111.
Oseroff, S. B. (1982). Phys. Rev. B25, 6584.
Planel, R., Gaj, J., and Benoit a la Guillaume, C. (1980). J. Physique 41, C5-39.
Rebman, G., Rigaux, C., Bastard, G., Menant, M., Triboulet, R., and Giriat, W. (1983).
Physica 117B and 118B, 452.
Ryabchenko, S. M., Terletskii, 0. V., Mizetskaya, I. B., and Oleinik, G. S. (1981). Fiz. Tekh.
Poluprovodn. 15, 2314. (Sov. Phys. Semicond. 15, 1345.)
Twardowski, A. (1981a). Ph.D. Thesis, University of Warsaw.
Twardowski, A. (1981b). Unpublished.
Twardowski, A,, Nawrocki, M., and Ginter, J. (1979). Phys. Stat. Sol. (b) 96, 497.
Twardowski, A., Rokka, E., and Gaj, J. A. (1980). Solid State Commun. 36, 927.
Twardowski, A., Dietl, T., and Demianiuk, M. (1983). Solid State Commun. 48, 845.
Twardowski, A., Swiderski, P., Ortenberg, M. von, and Pauthenet, R. (1984). Solid State
Commun. 50, 509.
Twardowski, A., and Ginter, J. (1982). Phys. Stat. Sol. (b) 110, 47.
Zimnal-Starnawska, M., Podgorny, M., K i d , A., Giriat, W., Demianiuk, M., and Zmija, J.
(1984). J. Phys. C17, 615.
CHAPTER 8
. .
I. INTRODUCTION
. . . . . . , . . . . . . . . . .
1. Shallow and Resonant Acceptors in DMS in the Absence of
Magnetic Field . . . . . . .
. . . . . . . . , .
2. Effect of Exchange Interactionon Band Structure. . . .
11. THEORY
OF SHALLOW
ACCEPTORS
IN THE ABSENCE OF MAGNETIC
FIELD . .
. .. . . . . ... .. . .. . . ...
3. Acceptor Hamiltonian . . . . . . . , . . . . . . .
4. Variational Procedure
. . . . . . . . . . . . .
111. ACCEPTORS
IN WIDE-GAP
DMS IN A MAGNETIC
FIELD.. . . .
5. Exchange Splitting. . . . . . . . .
,
. . . .
6. ExperimentalResults. . . . .
. . . . . .
IV. ACCEPTORS
IN NARROWGAP DMS IN A MAGNETIC
FIELD. . . .
7. Acceptor Hamiltonian in Ultra-Quantum Limit. . . .
8. Acceptor States in Ultra-QuantumLimit; Boil-Off Effect. .
9. Arbitrary Magnetic Fields. . . . . . . . . . . . . .
10. Experimental Binding Energies.
.,.... . ., .
IN p-TYPENARROW-GAP
DMS
V. IMPURITYMAGNETOCONDUCTIVITY
1 1 . Hopping Magnetoconductivityand Giant NegativeMagnetoResistance in Hgl-,Mn,Te . . . .
. , .
. .
12. Nonmetal-to-Metal Transition . . . . . . . . . . , .
VI. RESONANT
ACCEPTORS
IN ZERO-GAP
DMS . . . . . . . .
13. MagnetoopticalData. . . . . . . . . . . . . . . .
14. MagnetotransportData. . . . . . . . . . .
. . .
REFERENCES
. . . . . . . . . . . . . . . . . . . . .
. .
. . .
. . .
.
.
. . .
.
311
312
313
3 16
316
318
322
323
324
325
325
321
329
330
333
333
331
338
339
341
342
I. Introduction
Probably the best known diluted magnetic semiconductors (DMS) that can
be made p-type are Cdl-xMnxTe and Hgl-,MnxTe (see the review articles:
Galazka, 1979; Galazka and Kossut, 1979; Gaj, 1980; Bastard et al., 1980;
Mycielski, 1981; Dietl, 1981; Galazka, 1982; Furdyna, 1982a,b; Grynberg,
i ( 1930- 1986)
311
Copyright 0 1988 by Academic Press. Inc.
All rights of reproduction in any form IeSeNed.
ISBN 0-12-752125-9
312
J . MYCIELSKI
1983; Mycielski, 1983 and 1984). Both are A:'_,Mn,BV' systems of the zincblende cubic structure, and both have wide conduction and valence bands.
However, Cd, -,Mn,Te is a wide-gap semiconductor, while Hgl -,MnxTe is
a narrow-gap semiconductor (for x > 0.075), or a zero-gap semiconductor
(for x c 0.075). For specificity, in this chapter we will only consider
acceptors in the above materials.
We limit our considerations to shallow .acceptors, and ignore more
complex systems like excitons bound to acceptors (for excitons see: Gaj, this
volume). Taking into account the effect of the exchange interaction of a hole
with magnetic ions on the acceptor states, we will neglect the feedback: ordering of spins of magnetic ions by their interaction with the hole, i.e. the bound
magnetic polaron effect (for magnetic polarons see: Wolff, this volume).
1. SHALLOW AND RESONANTACCEPTORS
IN DMS IN THE
ABSENCE
OF
MAGNETIC
FIELD
In the absence of magnetic field the spins of magnetic ions are not ordered.
Thus there is no net exchangeinteraction of these ions with a hole. The acceptor
states in DMS are then similar to those in non-magnetic semiconductors.
The theoretical ionization energy EA of a shallow acceptor in CdTe is
87.4meV (Baldereschi and Lipari, 1973), and should be the same in
Cdl-,Mn,Te, as the parameters of the valence band (Twardowski and
Ginter, 1982) and the static dielectric constant (Gebicki and Nazarewicz,
1980) were found to be independent of composition in this material. In a
photoluminescence experiment values of 57.8, 58.8, 108, and 147 meV were
found for EA in CdTe and were interpreted as corresponding to the substitutional impurities Li, Na, Ag, and Cu, respectively (Molva et al., 1982).
In Cdl-,MnxTe the absorption due to acceptor photoionization was
measured and analyzed by Wojtal et a/. (1979). The photoionization
spectrum was fitted using a formula derived by the quantum defect method
(Bebb, 1969). This gave a rather high quantum defect for the ground state,
and ionization energies of 165 and 170meV for x of 0.05 and 0.10,
respectively. Thus, the acceptor is not very shallow.
In the narrow-gap Hgl -,Mn,Te, the theoretical ionization energy of a
shallow acceptor increases withx, and is about 9 meV for x = 0.15 (Mycielski
and Mycielski, 1980). From transport measurements the value 9.4 meV was
obtained for x = 0.15 (Wojtowicz and Mycielski, 1984), while the photoconductivity experiment revealed-for the same value of x-a deeper
acceptor, with the ionization energy about 30 meV (Wrobel et al., 1984). A
second acceptor level-with an ionization energy about 25 meV-was also
reported in Hgl-,Mn,Te with a wider gap (Anderson et al., 1983). Acceptors
are believed to be due to mercury vacancies or foreign impurities.
8.
313
a
4
I+,
-+> = i 2 - 9 x - i Y ) l ) ,
(2)
(3)
(4)
314
J. MYCIELSKI
where s and S; are the spin operators (in units of h ) of the band electron and
of the i-th magnetic ion, respectively, and the summation is over all lattice
sites occupied by magnetic ions. The exchange constant J is a short-range
function of distance.
Using the mean-field and virtual-crystal approximations we can write Eq.
(5) in the form (Bastard et a/., 1978 and Jaczynski et al., 1978)
which has the periodicity of the crystal lattice, as the summation is now over
all metal lattice sites. Here x is, as before, the molar fraction of the magnetic
component, and ( S ) a v is the spin of the magnetic ion, averaged thermally
and over all magnetic ions. Thus, ( S ) a v determines the macroscopic
magnetization of DMS. In the absence of an external magnetic field
(S), = 0. In the presence of a magnetic field ( S ) a v increases with decreasing
temperature. In the following, we take the external magnetic field parallel
to the z-direction. Thus, (S)avhas a non-vanishing z-component (Sz)av that
tends to -5 for strong fields and low temperatures.
The exchange Hamiltonian, Eq. (6), written in the basis given by Eqs.
(1)-(4) has the form
3Bexch
0
0
Hexch
with
Bexch
-Bexch
Bexch
= ;XNoP(Sz)av
0
-3Bexch
(7)
(8)
(9)
where NOis the number of unit cells in unit volume. It is worth noting that
&xch may be modified to include the Luttinger parameter K of the valence
band (Luttinger, 1956).
It follows from Eq. (7) that the effect of exchange interaction results in
splitting the I-" level into equally spaced singlets. This splitting dependsthrough (S,)av-On
the external magnetic field and the temperature, and
can approach values as large as 100 meV.
= (ZIJIZ),
a. Wide-Gap DMS
In a wide-gap DMS like Cdl-xMnxTe, because of the high effective masses
of holes, the exchange splitting is much higher than the Landau and Zeeman
splittings. It is, therefore, justified to perform the k p calculation using
Eq.(7) while neglecting the direct effect of the external magnetic field on the
8. SHALLOW ACCEPTORS
315
spin splitting and the orbital motion of the valence electrons (Gaj eta/., 1978
and Gaj, this volume). One then obtains a dramatic dependence of the
splitting of the valence bands on both the direction and the absolute value
of the electron wave-vector k. Of particular interest for the following are
dispersion relations close to the r point, i.e., for
where mh is the light-hole mass. Neglecting warping, the four non-degenerate valence bands are:
316
J. MYCIELSKI
where coo is the free-electron cyclotron frequency and ~ 1 3 / 2is the cyclotron
frequency corresponding to the transverse effective mass of the bands
E,3/z(k) given by Eq. (1 1). The wave function of bu(-1) is a ground-state
Landau envelope multiplied by @4. For the usual values of parameters (i.e.,
p > 0), bu(-l) is the highest valence band level in both the open- and zerogap DMS and is well-separated from other valence levels. It is the uppermost
heavy-hole Landau subband. From Eq. (13) and from the form of the wave
function, one can see that bu(-l) is the first Landau level originating from
E-3/z(k) valence band. In spite of the fact that it corresponds to a heavy-hole
Landau subband (in the z-direction), it is well-separated from other valence
levels due to a low transverse effective mass of the E-3/2(k) valence band,
and to the high value of &ch. It should be stressed, however, that other
Landau levels cannot be calculated in a narrow- or zero-gap DMS from the
simple expresssion, Eq. (1 1).
Because of the -33Bexch term in Eq. (13), &,(-I)
increases with the
magnetic field (even for x = lop3) up to the saturation of
(at a few
T ) and only then starts to decrease slowly. In a zero-gap DMS at moderate
fields, it may overlap in energy with the lowest conduction-band levels.
11. Theory of Shallow Acceptors in the Absence of a
Magnetic Field
We will present here the theory of shallow acceptors in a diamond or zincblende-type open-gap semiconductor in the absence of an external magnetic
field, following the variational method of Schechter (1962). Being less
accurate than other methods (see Baldereschi and Lipari, 1973), it is, however, more analytic. We present here Schechters theory in a simplified and
more explicit (while equivalent) version due to Mycielski and Rigaux (1983).
3, ACCEPTOR
HAMILTONIAN
v/; =
C mj4j;,
j=
1
i = 1, ..., 4,
(14)
v/i
are
8.
317
where
Di2
Di3
D = ( Dt2
D22
D?3
DT3
-Diz
Dii
D22
-D:2
-A
D22
0
-A - iB
0
+ )B
0
0
-ti(3B2
=3
+B
-A
+ )B
0
D12
-A
+ C2)'/'
- (31/2/2)B
-443~' +
0
-A -B
-ti(3B2 + C2)'/2
-$(3B2 + C2)'/2
- i ( 3 B 2 + C2)'/2
0
0
0
- )i(3B2 + C2)"2
(31/2/2)B
")
(21)
Here A , B, and Care material constants (see Dresselhaus et al., 1955). Signs
of the elements of D12 and D13 involving C 2 correspond to the case of the
negative value of the constant N defined by Dresselhaus et al. (1955). If
N > 0, the signs should be opposite. The physical meaning of A , B, and C
can be seen from the dispersion law of electrons in the valence band:
E(k)
+ k,k: + k,"kx')]"2.
(22)
A and B are real but C may be imaginary (i.e., C 2may be negative). In order
318
J. MYCIELSKI
to have E(k) real and negative for all k f 0, the following conditions must
be satisfied:
C 2 L -3B2,
(23)
(24)
+ IBI),
(25)
= h2/2(IAI - 1B1).
(26)
m/i, = A2/2(IAI
mhh
4. VARIATIONAL
PROCEDURE
We assume now that the wave functions of the acceptor ground state are
of the form (see Schechter, 1962 and Kaczmarek, 1966):
where CI, c2, c3, rl , and r2 are real parameters, and rl and r2 are positive.
The functions defined by Eq. (27)-(30) fulfill the orthonormalization
condition, Eq. (15), if
2 3
mlr
1 + 9 n c X + (:)ncfr? = 1.
(31)
8. SHALLOW
319
As one can see from Eqs. (27)-(30), the envelope functions are composed
from an s-type part (that proportional to CI) and two d-type parts (those
proportional to cz and c3). The latter parts do mix different @ j in a given v / i .
In order to determine the parameters CI, CZ, c3, r l , and r2, and the
ionization energy EA, we have to maximize expression (16). From Eqs.
(16)-(21)and Eq. (27) we get-after some algebra-E[yl] as a quadratic
form of dimensionless quantities a1 , at, and a3, with coefficients depending
on dimensionless PI and pz , where
ez
pz = -
EJAI
The quantities d , a$,and a: are the norms of the s-type and the two d-type
parts of the envelope, respectively (c.f., condition (31)).
Dividing E [ ~ / by
I ] d + a: + a: and maximizing the ratio with respect to
a2/a1 and 0 3 / 0 1 , we obtain
r;l
are given by
320
J. MYCIELSKI
The sign of adal is that of B , and a3/al is positive. If the constant Ndefined
by Dresselhaus et al. (1955) is positive, nothing is changed in our results
except for adal being negative.
It follows from Eq. (23) and Eq. (24) that
(43)
O<q<$.
In the absence of warping (i.e., for C = 0) we have from Eqs. (25), (26),
and (42)
where
=
(45)
mhh/mlh
(46)
- I)<$ - q)11-'.
(47)
8.
FIG.1.
321
The maximum value of the functionfdefined by Eq. (40)and the corresponding values
of the variables p1 and p2, for different values of the parameters q. [After Mycielski and Rigaux
(1983).]
17
FIG.2. The function F(p1, pz; q)-giving the acceptor radii [Eqs. (38), (39), and (401-as a
function of the material parameter q defined by Eq. (42). The values of p1 and p2 were chosen
so as to maximize the acceptor energy, Eq. (37). [After Mycielski and Rigaux (1983).]
Inserting the optimal values of p1 and pz in Eq. (41), we obtain the function
F(q) shown in Fig. 2.
Substituting f = f(q) and F = F(q) in Eqs. (37)-(39), we obtain the
acceptor binding energy EA and the norms of the two d-type parts of the
envelope. To calculate EA for CdTe-and therefore also for Cdl-,Mn,Te
with moderatex(see Gebicki and Nazarewicz, 1980; Twardowski and Ginter,
322
J . MYCIELSKI
1982)-we take the parameters values used by Baldereschi and Lipari (1973):
& = 9.7,
] A1 = 5.29h2/2mo,
IBI = 3.78h2/2mo,
ICI
= 5.45h2/2rno,
where mo is the free-electron mass. From Eq. (42), Fig. 1, and Eq. (37), we
then obtain the acceptor ionization energy,
EA = 79 meV.
= 15,
( A (= (136.3,41.1, 20.8)h2/2mo,
c = 0.
This yields
8. SHALLOW
323
5. EXCHANGE
SPLITTING
We will write now the exchange Hamiltonian (7) in the basis of the acceptor
ground state functions, Eq. (14) and Eqs. (27)-(30). Using also Eq. (33) and
Eq. (34), we obtain after some algebra,
01Bexch
Hexch
where
0
0
-02Bexch
0
0 2 Bexch
0
0
(48)
- 01Bexch
0 1
= 3 - 2at - +a$,
(49)
02
= 1 - 2d.
(50)
It follows from Eq. (48) that the functions (14) and (27)-(30) are eigenfunctions also in the presence of exchange. However, the fourfold degenerate
ground state of the acceptor is split by exchange. Comparing Eq. (7) with Eqs.
(48)-(50), one can see that the acceptor splitting is smaller than the splitting
of the free-carrier Tg state, and that the split levels are not equidistant. This
quenching of the acceptor splitting results from the fact that the acceptor
wave function is composed of different functions of the Tg level (with d-type
envelopes).
Substituting Eqs. (38) and (39) into Eqs. (49) and (50), we obtain 01 and
02 as the functions of the material parameters
and (C/B)2:
01=30 2 = 1 -
12
+ (8/3)(C/B)2
F(V)
5 + (C/B)
9
4
F(r;l).
(C/B)
These functions are shown in the form of a map in Fig. 3. The region of points
representing all possible materials is limited by Eq. (47).
Comparing again Eq. (7) with Eq. (48), we can see that the acceptor
ionization energy in the presence of exchange is reduced, and equals
EAexch =
EA - (3
- 01)IBexchI.
(53)
Using the material parameters given in Section 4, we get from Fig. 3 for
Cdl-*Mn,Te (for moderate x )
01
= 2.25,
02
= 0.83,
324
J . MYCIELSKI
I .o
0.5
I5
2.0
2.5
7
FIG.3. Acceptor splitting as a function of the material parameters. Bold lines: the boundaries
defined by Eq. (47) of the region of points representing all possible materials, C = 0 line corresponding to the absence of warping. Thin continuous lines correspond to the various values of
6
1 and broken lines to the various values of 02 as marked. Open circle is for Cdl-,Mn,Te, and
the cross is for Hgo.sfMno.l5Te. [After Mycielski and Rigaux (1983).]
= 2.26,
02
0.75.
8. SHALLOW
325
In Section 2, we have discussed the highest valence subband in a narrowgap DMS in the presence of an external magnetic field: that is, a heavy-hole
(in z-direction) band originating from the level b u ( - l ) ,of energy given by
Eq. (13). It is separated from lower subbands roughly by h 0 1 3 / 2 or 21B,,,1,l.
If
2IBexch1, h013/2 B EA,
(54)
the acceptor wave functions may be constructed from the wave functions of
this single subband (ultra-quantum limit), at least for the acceptor states
of energies above the bv(-l) level (Mycielski and Mycielski, 1980).
In the symmetric gauge A = )Bz(- y , x , 0 ) (where Bz is the magnetic
induction), an orthonormal electron wave function of the subband bu(-1)
has the form (see, e.g., Gawron and Mycielski, 1984):
ymkZ =
e x p ( i k Z z ) ~ , ( r ~ 4 ,
LT
Xm(r) = [ ( 2 1 ~ ) / ~ L ]I m
- ( !)-I2 exp(imr$)(p/21/2L)1m1
exp( -p2/4L2),
(55)
(56)
-00;
(57)
326
J. MYCIELSKI
eL
Y = EY,
(59)
(61)
where the normalized envelope function Tms(z) and the acceptor state energy
fulfill the following effective one-dimensional wave equation :
Ems
<
8.
327
we use the notation of Abramowitz and Stegun, 1965). It is worth noting that
6 is the ratio of the magnetic length to the heavy-hole Bohr radius.
For m = 0, one obtains
(64)
(65)
Thus, Vm(1tI) decreases with increasing ImJand tends to 0 for ImJ 00.
The selection rules for optical transitions between the functions defined by
Eq. (61) were discussed by Gawron and Mycielski (1984).
-+
8. ACCEPTOR
STATESIN
THE
THE BOIL-OFFEFFECT
ULTRA-QUANTUM
LIMIT;
328
J . MYCIELSKI
(67)
d B 1,
a. States with m
0.
Substituting Eq. (65) into Eq. (62), we obtain an equation of the Airy type.
Its eigenfunctions (vanishing for both
03 and (
- 00, and continuous
with the first derivative at = 0) are of the form
<-+ +
<
Tos(()
= (sgn
Ai(25/661/31(l+ xs)Cs,
(68)
= - ~ ~ ) 1 / 2 6 (+
1
21/6n-1/2d-1/3 Xs) .
(69)
Here Ai(x) is the Airy function, C, is a normalization factor, and xs are zeros
of the Airy function (odd s) or of its derivative (even s): xs 3 - 1.02, - 2.34,
-3.25, -4.09, -4.82, -5.52, ... for s = 0, 1,2, 3,4, 5, .... The envelope
function To, is symmetric for even s and antisymmetric for odd s. It follows
from Eq. (69) that the ground state is the symmetric state s = 0.
Equation (69) can also be written in the form
whereEAh = e4mhh/2h2&2
is the shallow acceptor binding energy (for BZ = 0)
calculated for the heavy-hole (i.e., longitudinal) effective mass. Because of
Eq. (67), the acceptor binding energy in the ultra-quantum limit is much
lower than EAh (which is not much higher than acceptor ionization energy
EA when the energy gap is not too low),
EA(&)
Eoo - &,,(-I)
Q EAh.
(7 1)
8.
329
330
J. MYCIELSKI
numerical results show the development of four levels originating from the
four-fold degenerate acceptor ground state.
10. EXPERIMENTAL
BINDINGENERGIES
One can expect that the strange behavior of the acceptor in narrow-gap
DMS in a magnetic field should affect substantially the low-temperature
magnetoresistance of p-type samples. In fact, a strong drop of the resistivity
of Hgl -,Mn,Te in magnetic field was observed at low temperatures by Delves
(1966) and Morrissy (1971). In the extrinsic range a very strong negative dc
magnetoresistance was observed in several p-type open-gap Hgl -,Mn,Te
samples with a net acceptor concentration of the order of 106cm-3 by
Mycielski and Mycielski (1980) for 0.12 c x c 0.17 and Wojtowicz and
Mycielski (1983) for 0.081 Ix I0.11. Anderson et at. (1983) observed a
strong negative ac magnetoresistance and a decrease of the ac Hall coefficient
of p-Hgl-,Mn,Te with 0.12 < x c 0.20 in the extrinsic range.
In the extrinsic p-type conductivity range the negative magnetoresistance
and the decrease of the Hall coefficient are due to a decrease of the acceptor
ionization energy with increasing magnetic field (the boil-off effect). Such
a decrease was demonstrated by the dc resistivity p vs. inverse temperature
and the ac Hall-coefficient vs. inverse temperature plots (Anderson et al.,
1983). It should be stressed, however, that the slope of Inp vs. l / T a t a given
magnetic field does not determine quantitatively the value of the acceptor
ionization energy at this field, since the ionization energy itself depends on
(at non-zero magnetic field). This was taken into
the temperature via (S,),
account recently by Wojtowicz and Mycielski (1984). They have measured
the transverse magnetoresistance and the Hall effect of p-type Hgl -,Mn,Te
in the composition range 0.085 Ix 5 0.15 and for the acceptor concentrations from 6 x 1015cm-3 to 3 x 10 ~ m - in~ the
, magnetic field B, up to
7Tand in a broad range of temperatures. In analyzing the data they assumed
that the acceptor ionization energy EA depends on Bz and T only through
(S,),
, thus neglecting the direct effect of the magnetic field on EA. Let us
denote by B,(Mz, T ) the magnetic induction at which (S,),
(i.e., also the
magnetization M,(B,, T)of the sample) has a given value at temperature T:
Mz[Bz(Mz
n,TI = Mz
(74)
8. SHALLOW
331
'
x.0.15
10.
No I
0.5
I .o
I .5
M, [emu41
FIG.4. Acceptor ionization energy in Hgl-xMnxTe as a function of magnetization. For each
composition x, the magnetic induction corresponding to a given value of magnetization at 15 K
is also given. Full circles are the experimental results of Wojtowicz and Mycielski (1984). The
acceptor concentration of the samples 1, 3, and 5 are 6 x lof5,4 x
and 9 x 1016cm-3,
respectively. Open circles correspond to Eq. (37), broken line to Eq. (53), and dotted line to
Eq. (70).
332
J . MYCIELSKI
For sample 1 , the experimental data in Fig. 4 are compared with the
dependence of EA on M , given by Eq. (53), with GI calculated at the end of
Section 5 and Bexchobtained from Eq. (8) with NO@= 0.65 eV (see Kossut,
this volume, and Rigaux, this volume). From Eqs. (70) and (26), and the
material parameters given at the end of Section 4, we have calculated EA in
the ultra-quantum limit. These values of EAare shown in Fig. 4 as a function
of magnetic induction B, for sample 1 . The agreement of both low- and highfield theory with experiment is reasonable, in spite of the fact that the
applicability of Eq. (53) to Hgl-,Mn,Te is questionable (see Section 5 ) and
the first part of the ultra-quantum limit condition (54) is not fulfilled: i.e.,
at a few Tesla the quantity 2/&xchl is slightly lower than EA at zero field. The
second part of Eq. (54) and the condition (67) for the validity of Eq. (70) are
fulfilled: for several Tesla Aco13/2 is about twice as high as EA in zero field,
and 6 is about 5 . It should also be noted that the method of determining EA
based on Eq. (75) is not well-justified for strong magnetic fields, which
influence EA not only through Mz but also directly (i.e., through orbital
magnetic quantization).
It is interesting that the decrease of EAwith increasing magnetization (i.e.,
with decreasing temperature at a given magnetic field) may be so strong for
samples with low x and high acceptor concentration that the concentration
of free holes will increase in the extrinsic range with decreasing temperature.
This effect was observed by Wojtowicz and Mycielski (1984) in sample 5 ,
both in the conductivity and the Hall constant measurements.
Some information on the acceptor states in narrow-gap Hgl -,Mn,Te were
obtained from the magnetotransmission measurements of p-type samples at
liquid helium temperatures (Rigaux et a/., 1980 and Bastard et a/., 1981). A
line identified as the transition from an acceptor level to the b,(l) Landau
state in the conduction band was observed. The acceptor level was found to
decrease with respect to the top of the valence band at B, = 0 as the magnetic
field increased (see Fig. 5 ) . The transmission measurement in the far infrared
(at B, = 0) has shown an absorption line at about 5 meV, which may
correspond to the transition from the valence band to this acceptor level
(Rigaux eC al., 1980). At 4T, this level is about 12meV below the bu(-l)
Landau level. Thus, it is not the shallow acceptor ground state, but probably
one of the excited states.
The measurements of impurity photoconductivity of p-type Hgl -,Mn,Te
with x = 0.15 at the temperature 4.2K and in magnetic field up to 8.3T
revealed a deeper acceptor, with a zero-field ionization energy of about
30 meV (Wrobel et al., 1984). The ionization energy decreases strongly in
magnetic field, and at 4Tis only about 17 meV. At higher fields the ionization
energy starts to increase with field. The strong decrease of the ionization
energy in magnetic field, much stronger than for shallow acceptors, may be
8.
333
20
40
B,(kG)
FIG. 5 . Magnetic field dependence of the acceptor level observed in Hgl-,MnxTe with
x = 0.128 (corresponding to an energy gap of 219 rneV). Zero energy is taken to be at the top
of the valence band at B, = 0. [After Bastard et al. (1981).]
11. HOPPINGMAGNETOCONDUCTIVITY
AND THE GIANT
NEGATIVE
MAGNETORESISTANCE
IN Hg, -,Mn,Te
It can be expected that the increase of the transverse radius of the acceptor
wave function in magnetic field should lead to a strong increase of the p-type
hopping conductivity in narrow-gap DMS and, furthermore, that the
conductivity should become anisotropic due to the magnetic-field-induced
anisotropy of the overlap integral of neighboring acceptors. This was
observed and interpreted for narrow-gap Hgl-,Mn,Te (0.12 < x < 0.17) by
Mycielski and Mycielski (1980). The samples were p-type and partially
compensated, with net acceptor concentration of the order of 10l6~ m - The
~ .
liquid-helium-temperature dc resistivity at 7 T was typically two orders of
magnitude lower than for zero magnetic field, and the transverse resistivity
334
J. MYCIELSKI
in magnetic field was lower than the longitudinal. It was also observed that
the activation energy for dc hopping conductivity decreases strongly with
magnetic field. Anderson et al. (1983) observed a more than one order of
magnitude decrease of the ac hopping resistivity of p-Hgl -,Mn,Te
(x = 0.19) at liquid helium temperatures in a magnetic field of 2.6T, and a
drop of the ac hopping activation energy.
The giant negative dc hopping magnetoresitance of Hgl-,MnXTe was
measured and analyzed in more detail by Wojtowicz and Mycielski (1983).
The samples used were narrow-gap (0.081 Ix I0.11) partially compensated p-type, with acceptor concentration of the order 106cm-3. The
transverse and longitudinal magnetoresistance was measured at temperatures
down to 1.4 K and magnetic fields up to 7T. The transverse resistivity vs.
reciprocal temperature for one of the samples is shown in Fig. 6 for different
magnetic fields. The hopping conductivity dominates below about 7 K. At
1.4 K, as B, increases to 7T, the resistivity drops by nearly seven orders of
magnitude! Both transverse and longitudinal resistivities of another sample
are shown as functions of magnetic field in Fig. 7, for different temperatures.
One can observe the anisotropy of the resistivity in magnetic field (the
transverse resistivity being smaller) and, ultimately an increase of resistivity
in very strong magnetic fields. The latter effect is expected when the true
ultra-quantum limit is finally reached, as the acceptor radius decreases slowly
in this limit with increasing magnetic field (see discussion at the end of
Section 8a).
In the hopping regime, the resistivity may be expressed in the form (Miller
and Abrahams, 1960)
P = ~3 e x p ( - ~ d k ~ T ) ,
(76)
8.
335
10
loc
10
lo4
-t6
10:
lo2
10
Ioo
lo-;
FIG.6 . Transverse resistivity as a function of temperature and magnetic field for a sample
of p-type Hgl-,MnxTe withx = 0.11. The hopping resistivity region corresponds approximately
to T - l > 0.1 (above the knee of the data).
(42), Fig. 1 , and the material parameters at the end of Section 4). On the other
hand, the transverse radius of the acceptor wave function in the ultraquantum limit, i.e., 2A, is 98 A at 6.8T
The decrease of e3 in the magnetic field may be caused by the approach
of the nonmetal-metal transition (see Section 12) or may correspond to the
destroying action of the magnetic field on the bound magnetic polaron and
on fluctuations of the magnetization (see Wolff, this volume). The latter
336
J . MYCIELSKI
field and
'wicz and
8.
337
TABLE I
TRANSVERSE
HOPPINGRESISTIVITY
PARAMETERS FROM
THE SAMPLE
OF p-Hg,-,MnxTe IN FIG. 6 (x = 0.1 1) AS
FUNCTIONS
OF MAGNETIZATION.
M Z
Iemukl
0
0.5
1 .o
1.5
2.2
BAM,, 4.2 K)
IT1
0
1.1
2.4
4.0
6.8
P3
E3
[Qcm]
[meVl
175
112
32
7.8
1.8
1.45
1.1
0.7
0.4
0.2
1984, for DMS). If the radius of the wave function increases, 63 will decrease
as the fluctuations are averaged over a bigger volume. Solving Eq. (62) [with
Eq. (64)] numerically, computing the overlap integral, and using percolation
theory, Gawron (1986b) has calculated recently the (positive) magnetoresistance in the ultra-quantum limit, as well as the anistropy of the
resistivity. The anisotropy is of the order of 2 at 4T, and drops to less than
unity at about 10T.
12. NONMETAL-METAL
TRANSITION
The increase of the overlap of the neighboring acceptor wave functions
with increasing magnetic field may induce the Mott nonmetal-to-metal
transition. This transition corresponds, in principle, to the vanishing of the
energy gap between the band of localized acceptor ground states and the
empty band of mostly delocalized positive acceptor states (i.e., acceptors
with two holes; see Mott, 1974). Before the transition occurs, two conductivity mechanisms are possible: excitation of holes from the localized
ground states to the positive acceptor band (more precisely, to the states
above the mobility edge of this band), and the hopping between the localized
ground states themselves. The conductivity activation energy of the former
mechanism is denoted by c t , and of the latter by 6 3 . As 8 3 < 8 2 , the hopping
conductivity dominates at lower temperatures. For different values of 6 2 the
conductivity extrapolated to T = 00 is roughly the same and is given by
amin
=
0.026e2N,3/tr,
(77)
i.e., the minimum metallic conductivity, NJ3 being the critical acceptor
concentration at which the Mott transition occurs, i.e., at which the
conductivity activation energy vanishes.
338
J. MYCIELSKI
8. SHALLOW
339
10'
loi
10'
loa
10-
0.1
0.2
0.3 0.4
T-' [ K-' ]
0.5
0.6
DATA
13. MAGNETOOPTICAL
Several magnetoabsorption measurements were performed on zero-gap
Hgl-,Mn,Te (Bastard el al., 1978, 1979; Dobrowolska et al., 1980;
340
J. MYCIELSKI
10
20
30
40
Bz (kG)-+
FIG. 9. Submillimeter absorption lines vs. magnetic field for zero-gap Hg,-,Mn,Te with
x = 0.0014, at T = 1.7 K in the Faraday configuration. The data represent two transitions
between the r8-valence and the rs-COndUCtiOnLandau levels, and the transition from a revalence
Landau level to an (impurity) acceptor level. [After Bastard ef af. (1979).]
8.
341
2
I
t
W
-I
-2
10
20
B, (kG)+
FIG. 10. Submillimeter absorption line vs. magnetic field for zero-gap Hgl-,Mn,Te with
x = 0.008, observed at T = 1.7 K in the Faraday configuration. The absorption corresponds
to a transition from an acceptor level to the bu(-l) valence Landau level. [After Bastard et al.
( 1979).]
occupied acceptor state to the bv(- 1) valence Landau level was observed
(Fig. 10). From an extrapolation of the transition energy to zero field (i.e.,
to negative energies), the zero-field acceptor energy level was estimated to be
2.2 meV. Therefore, the level was identified as the A1 (mercury vacancy)
level, observed also in non-magnetic zero-gap Hgl-xCdxTe (Guldner et al.,
1977). The existence of a transition from this level to the b4-I) level
indicates that the acceptor wave function involves a contribution from the
function (P4 = 15, -$) of the Tg basis. The energy of A1 was found to be
nearly independent of magnetic field.
14. MAGNETOTRANSPORT
DATA
In a partially compensated p-type zero-gap semiconductor with a low
conduction band density of states (i.e., a low conduction band effective
mass), the Fermi level is usually pinned to the resonant acceptor level (in the
absence of magnetic field). Measuring the low field Hall coefficient and the
period of the Shubnikov-de Haas oscillations, and knowing the conduction
effective mass, one can determine the position of the Fermi level, i.e., the
resonant acceptor energy (Sawicki et al., 1983; Sawicki and Dietl, 1983).
Energies of about 4 meV and 7 meV were reported for x 0.06 (i.e., for the
342
J. MYCIELSKI
r
6 band only slightly below the TSbands), and it was observed that the energy
increases with increasing magnetic field.
Conductivity in the impurity band of resonant acceptors was observed by
Davydov et af. (1980), Ponikarov et af. (1981), and Davydov et af. (1981).
The conductivity increases strongly in the field range from 1 to 2T, in
particular in the transverse direction. This effect was interpreted as the
decrease of spin-disorder scattering in the impurity band. However, it may
also be interpreted by the increase of the mean radius of the acceptor wave
function, as discussed earlier in this chapter.
Transformation of resonant acceptor states in zero-gapp-type Hgl -,Mn,Te
into non-resonant states was investigated by magnetotransport measurements in high magnetic fields and/or under high pressures. Temperatures
down to 30 mK and hydrostatic pressures up to 1.5 GPa were used (Sawicki
et al., 1983). For x 4 0.06, the r6 band is situated only slightly below the TS
bands. A high hydrostatic pressure can then reverse the relative band
positions thus opening the gap of the material and transforming the resonant
states into the non-resonant acceptors. The magnetic field may open a gap
between the TSconduction and the TSvalence band (possibly after initially
overlapping them by an upward shift of the bu(-1) level; see Section 2), and
also make the acceptor states non-resonant. The latter effect was observed
by Byszewski et af. (1979), Ponikarov et af. (1981), Davydov et af. (1981),
Sawicki et af. (1983), and Sawicki and Dietl (1983). The field intensity
necessary to open the gap decreases from about 10T for low (but non-zero)
x to about 1T for x = 0.065.
After opening the gap, hopping conductivity involving non-resonant
acceptors was observed (Ponikarov et af., 1981; Davydov et al., 1981). It
decreases sharply with increasing (strong) magnetic field, due to shrinking
of the acceptor wave function. The transverse conductivity is then lower than
the longitudinal conductivity. In a sample with x = 0.065 (where the r6band
is only slightly below the TSbands) and with a net acceptor concentration of
2 x 106cm-3m, the acceptors were found to be in a metallic phase after
opening the gap by the magnetic field of about 1T (Sawicki and Dietl, 1983).
For magnetic fields higher than 3T, they show a positive magnetoresistance.
References
Abramowitz, M., and Stegun, I. A . (1965). Eds. of Handbook of Mathematical Functions.
Dover, New York.
Akimoto, O., and Hasegawa, H. (1967). J. Phys. SOC. Japan 22, 181.
Anderson, J. R., Johnson, W. B., and Stone, D. R. (1983). J. Vac. Sci. Technol. A l , 1761.
Baidereschi, A., and Lipari, N. 0. (1973). Phys. Rev. B8, 2697.
Bastard, G. (1979). In Physics of Semiconductors, 1978 (B. L. H. Wilson, ed.), Conf. Ser.
43, p. 231. The Institute of Physics, Bristol.
8.
343
Bastard, G., Rigaux, C., Guldner, Y., Mycielski, J., and Mycielski, A. (1978). J. Physique 39,
87.
Bastard, G . , Rigaux, C., Couder, Y., Thome, H., and Mycielski, A. (1979). Phys. Stat. Sol.
(b) 94, 205.
Bastard, B., Gaj, J. A., Planel, R., and Rigaux, C. (1980). J. Physique C5, 247.
Bastard, G., Rigaux, C., Guldner, Y., Mycielski, A., Furdyna, J. K., and Mullin, D. P. (1981).
Phys. Rev. B24, 1961.
Bebb, H. B. (1969). Phys. Rev. 185, 1 1 16.
Beneslavskii, S. D., and Entralgo, E. (1975). Zh. Eksp. Teor, Fiz. 68,2271. (Sov. Phys.: JETP
41, 1135.)
Beneslavskii, S. D., Brandt, N. B., Golyamina, E. M., Chudinov, S. M., and Yakovlev, G. D.
(1974). Zh. Eksp. Teor. Fiz. Pisma 19, 256. (Sov. Phys. JETP Letters 19, 154.)
Byszewski, P., Szlenk, K., Kossut, J., and Galazka, R. R. (1979). Phys. Star. Sol. (b) 95, 359.
Davydov, A. B., Ponikarov, B. B., andTsidilkovskii, I. M. (1980).Phys. Stat. Sol. (b) 101,127.
Davydov, A. B., Ponikarov, B. B., and Tsidilkovskii. I. M. (1981). Fiz. Tekh. Poluprov. 15,
881. (Sov. Phys. Semicond. 15, 504.)
Delves, R. T. (1966). Proc. Phys. Soc. 87, 809.
Dietl, T. (1981).In Physicsin High Magnetic Fields (N. Miura, ed.), p. 344. Springer, Berlin.
Dietl, T., Antoszewski, J., and Swierkowski, L. (1983). Physica B117 and 118, 491.
Dobrowolska, M., and Dobrowolski, W. (1981). J. Phys. C: Solid St. Phys. 41, 5689.
Dobrowolska, M., Dobrowolski, W., Otto, M., Dietl, T., and Galazka, R. R. (1980). J. Phys.
Soc. Japan 49, Suppl. A, p. 815.
Dobrowolski, W., Ortenberg, M. von, Sandauer, A. M., Galazka, R. R., Mycielski, A., and
Pauthenet, R. (1982). In Physics of Narrow Gap Semiconductors (E. Gornik,
H. Heinrich, and L. Palmetshofer, eds.), p. 302. Springer, Berlin.
Dresselhaus, G., Kip, A. F., and Kittel, C. (1955). Phys. Rev. 98, 368.
Elliot, R. J., and Loudon, R. (1960). J. Phys. Chem. Solids 15, 196.
Furdyna, J. K. (1982a). J. Vac. Sci. Technol. 21, 220.
Furdyna, J. K. (1982b). J. Appl. Phys. 53, 7637.
Furdyna, J. K. (1985). Solid State Commun. 53, 1097.
Gaj, J. A. (1980). J. Phys. Soc. Japan 49, Suppl. A, p. 797.
Gaj, J. A., Ginter, J., and Galazka, R. R. (1978). Phys. Stat. Sol (b) 89, 655.
Galazka, R. R. (1979). In Physics of Semiconductors, 1978 (B. L. H. Wilson, ed.), Conf.
Ser. 43, p. 133. The Institute of Physics, Bristol.
Galazka, R. R. (1982). In Physics of Narrow Gap Semiconductors (E. Gornik, H. Heinrich,
and L. Palmetshofer, eds.), p. 294. Springer, Berlin.
Galazka, R. R., and Kossut, J. (1979). In Proceedings of the International School of NarrowGap Semiconductors, Nimes, Lecture Notes in Physics, 133, p. 245. Springer-Verlag.
Gawron, T. R. (1986a). J. Phys. C19, 21.
Gawron, T. R. (1986b). J. Phys. C19, 29.
Gawron, T. R., and Mycielski, J. (1984). Phys. Stat. Sol. (b) 125, 341.
Gawron, T. R., and Trylski, J. (1982). In Physics of Narrow Gap Semiconductors
(E. Gornik, H. Heinrich, and L. Palmetshofer, eds.), p. 312. Springer, Berlin.
Gelmont, B. L., Gadzhiev, A. R., Shklovskii, B. J., Shlimak, J. S., and Efros, A. L. (1974).
Fiz. Tekh. Poluprov. 8, 2377. (Sov. Phys. Semicond. 8, 1549.)
Gebicki, W., and Nazarewicz, W. (1980). Phys. Stat. Sol. (a) 61, K135.
Gortel, 2. W., Szymanski, J., and Swierkowski, L. (1979). In Physics of Semiconductors,
1978 (B. L. H. Wilson, ed.), Conf. Ser. 43, p. 261. The Institute of Physics, Bristol.
Grynberg, M. (1983). Physica 117B and 118B, 461.
Guldner, Y., Rigaux, C., Mycielski, A., and Couder, Y. (1977). Phys. Stat. Sol. (b) 81, 615.
344
J. MYCIELSKI
Gupta, 0. P., Joos, B., and Wallace, P. R. (1982). Phys. Rev. B25, 1101.
Jaczynski, M., Kossut, J., and Galazka, R. R. (1978). Phys. Stat. Sol. (b) 88, 73.
Jaroszynski, J., Dietl, T., Sawicki, M., and Janik, E. (1983). Physica 117B and 118B,473.
Kaczmarek, E. (1966). Acta Phys. Polonica 30, 267.
Kossut, J . (1976). Phys. Stat. Sol. (b) 78, 537.
Lipari, N. O . , and Altarelli, M. (1980). Solid State Commun. 33, 47.
Luttinger, J . M. (1956). Phys. Rev. 102, 1030.
Luttinger, J. M., and Kohn, W. (1955). Phys. Rev. 97, 869.
McKnight, S. W., Amirtharaj, P. M., and Perkowitz, S. (1978). Solid State Commun. 25, 357.
Miller, A., and Abrahams, E. (1960). Phys. Rev. 120,745.
Molva, E., Chamonal, J . P., and Pautrat, J. L. (1982). Phys. Stat. Sol. (b) 109, 635.
Morrissy, J. H. (1971). Ph.D. Thesis, Oxford University.
Mott, N. F. (1968). Phil. Mag. 17, 1259.
Mott, N. F. (1974). Metal-Insulator Transitions. Taylor and Francis, London.
Mycielski, A., and Mycielski, J. (1980). J. Phys. SOC.Japan 49, Suppl. A, p. 807.
Mycielski, J . (1981). In Recent Developments in Condensed Matter Physics (J. T. Devreese,
ed.), Vol. 1, p. 725. Plenum Press, New York.
Mycielski, J. (1983). In Application of High Magnetic Fields in Semiconductor Physics (G.
Landwehr, ed.), p. 431. Springer, Berlin.
Mycielski, J. (1984). Progr. Cryst. Growth Character 10, 101.
Mycielski, J., and Rigaux, C. (1983). J. Physique 44, 1041.
Mycielski, J., Witowski, A., Wittlin, A., and Grynberg, M. (1984). In Proceedings 17th
Intern. Conf. on Physics of Semiconductors, San Francisco (D. J. Chadi and W. A.
Harrison, eds.). Springer, Berlin, 1415.
Ortenberg, M. von (1973). J. Phys. Chem. Solids 34, 397.
Pidgeon, C. R., and Brown, R. N. (1966). Phys. Rev. 146, 575.
Pollak, M. (1980). Phil. Mag. B42, 781.
Ponikarov, B. B., Tsidilkovskii, I. M., and Shelushinina, N. G. (1981). Fiz. Tekh. Poluprov.
15, 296. (Sov. Phys. Semicond. 15, 170.)
Rigaux, C., Bastard, G., Guldner, Y., Rebmann, G., Mycielski, A., Furdyna, J. K., and Mullin,
D. P . (1980). J. Phys. SOC. Japan 49, Suppl. A, p. 81 1.
Sawicki, M., and Dietl, T. (1983). In Physics of Semiconducting Compounds (R. R. Galazka
and Rauluszkiewicz, eds.), Polish Academy of Sciences, Institute of Physics, Proc. Conf.
in Physics, Vol. 6, p. 400. Ossolineum, Wroclaw.
Sawicki, M., Dietl, T., Plesiewicz, W., Sekowski, P., Sniadower, L., Baj, M., and Dmowski,
L. (1983). I n Application of High Magnetic Fields in Semiconductor Physics
(G. Landwehr, ed.), p. 382. Springer, Berlin.
Schechter, D. (1962). J. Phys. Chem. Solids 23, 237.
Shlimak, I. S., Efros, A. L., and Yantchev, I. Ya. (1977). Fiz. Tekh. Poluprov. 11,257. (Sov.
Phys. Semicond. 11, 149.)
Trzeciakowski, W. (1982). J. Phys. C.: Solid St. Phys. 15, 1199.
Twardowski, A., and Ginter, J. (1982). Phys. Stat. Sol. (b) 110, 47.
Wojtal, R.,Golnik, A., and Gaj, J. A. (1979). Phys. Stat. Sol. (b) 92, 241.
Wojtowicz, T., and Mycielski, A. (1983). Physica 117B and 118B. 476.
Wojtowicz, T., and Mycielski, A. (1984). Acta Phys. Polonica A67, 363.
Wrobel, J., Kuchar, F., and Meisels, R. (1984). Acta Phys. Polonica A67, 369.
CHAPTER 9
I.
11.
111.
IV.
V.
VI.
INTRODUCTION . . . . . . . . . . . . . . . . . . . .
INELASTIC LIGHTSCATTERING
PRINCIPLES.
. . . . . . . . .
EXPERIMENTAL
TECHNIQUES.
. . . . . . . . . . . . . .
RAMAN SCATTERING
BY VIBRATIONAL
EXCITATIONS.
. . . . .
1 . Phonons in Perfect Crystals . . . . . . . . . . . . .
2 . Phonons in Mixed Crystals. . . . . . . . . . . . . .
3. Experimental Results and Discussion. . . . . . . . . .
MAGNETIC
EXCITATIONS
. . . . . . . . . . . . . . . .
4. General Considerations . . . . . . . . . . . . . . .
5 . Paramagnetic Phase . . . . . . . . . . . . . . . .
6 . Magnetically Ordered Phase . . . . . . . . . . . . .
7. Spin-Flip Raman Scattering . . . . . . . . . . . . .
VIBRATIONAL,ELECTRONIC
AND MAGNETICEXCITATIONS
IN
SUPERLATTICES
. . . . . . . . . . . . . . . . . . . .
REFERENCES.. . . . . . . . . . . . . . . . . . . .
345
346
350
353
353
351
365
313
313
316
389
396
406
410
I. Introduction
In this chapter the focus is on Raman scattering studies on the localized and
collective excitations in Mn-based 11-VI diluted magnetic semiconductors
(DMS), the excitations being vibrational, electronic or magnetic in character.
As has been extensively discussed elsewhere in this volume, the DMS alloys
possess a unique combination of semiconducting and magnetic properties.
In common with their non-magnetic counterparts (e.g. Hgl-,Cd,Te), they
exhibit the well-known composition-dependent vibrational spectra (Barker
and Severs, 1975), electronic band structure (Zanio, 1978) and transport
properties (e.g. carrier mobilities). In addition, in a Mn-based 11-VI alloy like
Cdt-,Mn,Te, new properties arise which can be traced to the substitutional
Mn2+ions randomly replacing the group I1 element (Cd, Zn, or Hg). In these
tetrahedrally co-ordinated semiconducting alloys with the zinc blende or the
wurtzite structure, the ground state of Mn2+ is
(ignoring small crystal
345
Copyright 0 I988 by Academic Press. Inc.
All rights 01 reproduction in any form reserved.
ISBN 0-12-752125.9
346
field effects); these ions thus have an effective magnetic moment of 5.92 Bohr
magnetons @B). The exchange interaction between the Mn2+ ions and the
band electrons produces large spin splittings at the conduction band
minimum and valence band maximum, in turn resulting in the large Zeeman
splitting of the free exciton (Gaj et al., 1978a) and the giant Faraday rotation
(Gaj et al., 1978b). When the Mn2+ concentration (x) exceeds a certain
critical value, long range magnetic ordering sets in below a certain temperature (Galazka et al., 1980)-a consequence of the anti-ferromagnetic interaction between neighboring Mn2+ions. Raman scattering has proved to be
very fruitful in discovering and delineating vibrational modes, localized and
collective; electronic transitions within the Zeeman multiplet of the ground
state of Mn2+ split by an external magnetic field; the spin-flip of electrons
bound to donors in large, effective mass orbits; and the collective magnetic
excitations in the low temperature, magnetically ordered phases. In this
chapter, the theoretical and experimental background on Raman scattering
are first presented in general terms. Each specific topic with its relevant
theory is then developed along with the presentation of experimental results
and their interpretation.
11. Inelastic Light Scattering: Principles
9.
347
RAMAN SCATTERING
Here
5 (Eyv
(v
1d 1 v
-
) ( v 1 d 1 yo)
E., - AWL
1 1
1 I v
( v d yo) ( v d
E,,l
- E,
+h
))
o ~
(3)
the states Iv) are the energy eigenvectors with eigenvalues E, and d, the
electric dipole moment operator of the system. The quantities n~ and ns are
the photon populations in the incident and scattered beams, respectively. The
summation in Eq. (3) extends over all states Iv) of the scatterer.
Often, it is convenient to describe the motion of the scattering system as
consisting of two or more types of excitations coupled by very small interactions. Typical examples are molecules with weakly coupled electronic,
vibrational and rotational motions, and crystals where one can, in a first
approximation, disregard the electron-phonon interaction. In such cases, the
states Iv) described above are the stationary states of the system taking due
account of the coupling between all modes of motion. To fix the ideas, we
think of crystals in which the electron-phonon interaction Hepis sufficiently
348
A . K . RAMDAS A N D S. RODRIGUEZ
where the EoYare the energies associated with the 4,. Clearly each Eo, is the
sum of electronic and phonon energies. The sum omits the v' = v.
An equivalent procedure is to take He, and the electron-photon interaction as perturbations. Then, the scattering cross section is expressed as a
sum reflecting the use of third order perturbation theory as in Loudon (1963,
1964). For example, a Raman process associated with emission of an optical
phonon can be viewed as the virtual creation of an electron-hole pair (step
l), emission by the electron or by the hole of an optical phonon (step 2, Hep)
followed by electron-hole recombination with emission of a scattered photon
(step 3). This process is illustrated in Fig. 1. The steps involved in this
description are implicit in the nature of the states Iv) in Eq. (4) that explicitly
take the electron-phonon interaction into account.
Equation (3) for CY,~,, contains matrix elements between the exact states
I VO), I v), and I v' ) as well as the energy eigenvalues E, . These are related to
6,0, 4v,
4,~and EoYby standard perturbation theory. It is sometimes useful
to view C Y as~ being
~
~modulated
~
by the internal motions. The amplitude of
the polarization giving rise to a Raman line at OL - wi is of the form
9.
349
RAMAN SCATTERING
k'
'
FIG. 1 . Schematic diagram representing inelastic light scattering in crystals with the creation
of a phonon. The dashed lines represent photons, the solid lines with arrows directed towards
the right (left) correspond to electrons (holes) and the wavy line indicates a phonon.
the interaction between the radiation field and the scattering system changes
by factors eikr'" and e-jks'" upon translation. If q and qo are the wave
vectors of the states I v ) and I VO), respectively, the amplitude of the resulting
transition probability is a lattice sum, each term being proportional to
ei(-q
+ kr- ks +qo).
II
where G is a vector of the reciprocal lattice. For visible light, /kLl and (ksl
are negligible compared to [GIexcept for G = 0; therefore, this conservation
law reduces to
kL
ks
+ q - 90.
(5)
19) = 21k~Isin-,
2
350
where 8 is the scattering angle. Thus in first order Raman scattering, only
states whose wave vectors are near the center of the fundamental Brillouin
zone can be excited (or de-excited).
A large number of investigations accessible to Raman scattering involve
collective or localized excitations with frequencies in the infrared. For this
reason an excitation observed with the absorption or emission of a photon
is labeled infrared active whereas that seen in a Raman spectrum, Raman
active. Raman and absorption spectroscopy are complementary tools in the
study of the internal motions of centro-symmetric molecules and crystals.
(See, for example, Bhagavantam and Venkatarayudu, 1969.)
In addition to wave vector conservation, Raman transitions are governed
by selection rules appropriate to the second rank polarizability tensor a. In
this context symmetry arguments can be used to deduce selection rules
including details of the polarization features for different scattering
geometries. A transition from a state vo belonging to the irreducible
representation ry0
of the space group of the crystal to a state v belonging to
r, occurs only if rY* x ,?I x rvocontains the totally symmetric representation. Here r, is the representation generated by the components of a. We
also note that, for Raman shifts associated with visible exciting radiation,
since the wave vector of the excitation responsible is near the center of the
Brillouin zone, one is justified in treating these zone-center excitations
using the point-group of the crystal.
111. Experimental Techniques
9.
351
RAMAN SCATTERING
In Table I we list the various ion lasers currently available and the wavelengths of the visible radiations they provide. Monochromatic lines in the
range 1-12 pm are generated with Nd:YAG, CO, and C02 lasers. Frequency
doubling with non-linear crystals is a procedure often exploited to obtain
exciting lines in a desired range. Tunable dye lasers with suitable dyes,
pumped with a Ar+ laser, a Kr laser, a frequency doubled 1.06 pm line of
the Nd:YAG laser, or a nitrogen laser allow a complete coverage from the
blue to the red. Though CW operation is by far the standard in Raman
spectroscopy, pulsed lasers are exploited in time resolved as well as in
stimulated Raman studies.
TABLE I
LASERAND VISIBLE OUTPUT WAVELENGTHS USED
BRILLOUIN
SPECTROSCOPY
Laser
Krypton ion
He-Ne
Dye laser
Argon ion
Output wavelength
IN
RAMANAND
(A)
Since Raman signals are typically orders of magnitude weaker than the
parasitically scattered, unshifted laser radiation, spectrometers have to be
able to reject it efficiently, thus allowing Raman lines to stand out. Double
and triple grating monochromators are especially designed to address this
problem; with currently available spectrometers it is possible to observe
Raman lines with shifts as small as a few wave numbers, and even Brillouin
components under favorable experimental conditions. Holographic gratings
are particularly effective in rejecting efficiently the parasitic radiation.
Brillouin components and sharp Raman lines can be experimentally studied
by isolating a small spectral range with an optical filter and performing
further spectral analysis with a scanning Fabry-Perot interferometer.
Multipassing of a Fabry-Perot has proved a significant measure to enhance
its contrast, and tandem operation allows an advantageous increase of the
free spectral range.
Scattered radiation after its spectral analysis is typically detected with
thermoelectrically cooled photomultipliers followed by photon counting
electronics. The state-of-the-art data acquisition and data processing systems
incorporate microcomputers. These can be organized to multi-scan a given
352
spectral range and the data can be displayed on a multichannel analyzer, X-Y
or chart recorder or a monitoring screen and plotted on a plotter after the
data are suitably processed.
Figure 2 shows the schematic diagram of the experimental set up used in
the Raman scattering experiments in the authors laboratory. Variable
temperature cryostats operating from room temperature down to 1.8 K using
liquid helium as a coolant; uniaxial stress attachments and hydrostatic stress
apparatus (e.g. the diamond anvil cell); variable temperature cryostats
incorporating a superconducting magnet-such equipment currently are in
an advanced and sophisticated state of development and are commercially
available or readily fabricated.
We refer the reader to Chapter 2 of Hayes and Loudon (1978) for a very
readable account of Raman and Brillouin spectrometers. For a comprehensive description of Raman spectrometers, we recommend the article by
Hathaway (1971). A complete description of a piezo-electrically scanned
tandem Fabry-Perot interferometer appears in Sandercock (1982).
RATEMETER
MICROCOMPUTER
SYSTEM
PLOTTER
FIG.2. Schematic diagram of the experimental setup used in Raman scattering experiments.
PR stands for polarization rotator, A for analyzer, DP for Dove prism, and HWP for half-wave
plate. L1, L2, and L3 are lenses. [After Peterson (1984).]
9.
RAMAN SCATTERING
353
(NMu)-12eaexp[i(q n - at)],
(8)
354
C [Caar(q)- o26,,!1e,,
where
Caa,(q)= C(MolMol~)-2Ca,~(n
- n)exp[-iq
n
(9)
- (n - nf)l.
(10)
02601aI
= 0.
(1 1)
For each wave vector q, Eq. (1 1) gives 3f solutions for the frequency o(q)
yielding 3f branches of the dispersion curves of the vibrational spectrum of
the crystal, The values of q are restricted to the reduced Brillouin zone to
avoid redundancy. For crystals containing f atoms per primitive cell, there
are 3 acoustic branches and 3(f - 1) optical branches. For acoustic modes,
o is linear in q near q = 0 and vanishes at the center of the Brillouin
zone. However, optical branches have a non-vanishing frequency at the
zone center. Except for degenerate polar optical phonons in crystals free
from improper symmetry operations (Pine and Dresselhaus, 1969, 1971;
Grimsditch et a/., 1977; Imaino el al., 1980), the frequency w(q) is of
the form
9.
355
RAMAN SCATTERING
CHARACTER
TABLEFOR
&
8C3
3Cz
AI
A2
E
FI
F2
1
-1
2
3
3
0
0
2
-1
-1
60d
6s4
1
-1
-1
-1
1
-1
Td
Basis functions
x 2 +Y 2 + Z 2
222 - x2 - Y2, A ( X 2 - Y2)
Rx,Ry,Rz
x, Y , Z ; Y Z , Z X , X Y
X , Y, Z are the components of a polar vector field along the cubic axis x, y, z . R,,
Ry , and R , are the components of a pseudo-vector along the cubic axes.
356
TABLE 111
CHARACTER TABLEFOR
A1
1
1
1
1
Az
B1
Bz
El
EZ
2
2
1
1
-1
-1
-2
2
1
1
1
1
-1
-1
1
-1
-1
1
1
-1
-1
z; xz
R,
-1
1
-1
0
0
-1
c
6
v
0
0
+ Y2; z2
Y(Y2 - 3 x 7
X ( X 2 - 3Y2)
X , K Z X , ZV; R , , Ry
xz-Y Z , - 2 X Y
~~
~~
X , Y, and Z are the components of a polar vector field with respect to a coordinate
system x , y , z in which 1is along c6.. and the xz plane is a Ud R,, R,, and R, are the
components of a pseudo-vector along the coordinate axes.
The wurtzite structure, having four atoms per primitive cell, has optical
phonons with symmetries A, ,El , B1, Ez ,there being two of each of the last
two irreducible representations. Of these A1 and E1 are both Raman and
infrared active, E2 is only Raman active, whereas B1 is active in neither.
In order to explain the polarization features of the Raman spectra of
phonons we determine the form of the Raman scattering tensor a, assumed
to be symmetric away from resonance, by reducing the representation it
generates into its irreducible components. This reduction is r, = A1 E FZ
for Td and r, = 2A1 + El + EZ for C6".
Next we use a theorem in group theory which we quote: the inner product
of two vectors (or quantum mechanical states if we are dealing with a
dynamical system) is zero unless the vectors belong to the same row of
identical, unitary irreducible representations. As an example, we deduce the
form of a for the E modes in the zinc-blende structure. Clearly, for the E
mode belonging to the row 22' - X' - Y z ,
+ +
aXx
= aZx= av = 0.
The quantity 2aZz- axx- or,, does not vanish. If we denote it by 66 we find
ayy = - b, azz = 2b. We find the components of a for the second row
by noting
aXx=
f i ( ~ ~ x-x ~ l y y )=
while
ffxx
ffyy
6b
= ay, = ffzx =
ffp
= 0.
These equations yield cyZz = 0 and aXx= -ayy = b f i . The forms of a for the
transitions from the totally symmetric ground state A1 to the
9.
357
RAMAN SCATTERING
TABLE IV
RAMANTENSORS
FOR
AI
a 0 0
10 a 0 1
O O a
1-i
fib
E
0
- b0
Td
F2
0 0 0
2b
: l , 2 Z 2 - X 2 - Y 2 1 0O 0d O
d / , X
I:
FI
0
:I,R.
--c0
O O d
-c
O d O
-fib
d O O
Note that the tensors associated with A l , E, and FZare symmetric whereas those with FIare
antisymmetric.
TABLE V
RAMAN TENSORS
FOR c6"
Symmetric
AI
EI
Antisymmetric
E2
A2
EI
0 0
0 0
10 b
0
- b , Y,Ry
01
allowed final states in first order Raman scattering are displayed in Tables
IV and V for Td and c 6 u , respectively. The symmetry classification of the
lines in an experimental spectrum can be arrived at on the basis of polarization studies for a variety of scattering geometries.
2. PHONONS
IN MIXEDCRYSTALS
A number of physically interesting aspects of crystal physics have their
origin in the presence of imperfections, e.g., foreign atoms, lattice defects,
.... Alloys of two or more compounds forming a homogeneous crystalline
358
phase also exhibit novel physical phenomena. These systems lack strict
translational symmetry. The lattice dynamics of such crystals is the subject
of this sub-section. We illustrate our discussion of mixed crystals with the
results on alloys of two compound semiconductors like CdTe and MnTe in
atomic proportions of ( 1 - x)/x. The resulting compound is designated by
Cdl-,Mn,Te. It has been found that for x 5 0.7, it has the zinc-blende
structure with Mn atoms replacing at random sites Cd atoms (See Chapter 2).
We investigate first the small vibrations of a crystal containing imperfections located at the atomic sites of the perfect crystal. We follow procedures
and nomenclature similar to those in Sec. IV. 1. To second order in the small
displacements from equilibrium, una,the potential energy is of the form
(14)
MAAiina
=-
na
Cik$(n,n)unsa,,
(15)
where Mi;) is the mass associated with the a-th degree of freedom in the cell
n. Periodic solutions of Eq. (15) of the form
(0) - i d
una(t) = unae
exist if
na
IMAL)U2Saa1Snnl- Ci3(n,n)l = 0.
(18)
(19)
9.
RAMAN SCATTERING
359
As seen in Sec. IV. 1, in a perfect crystal, the size of the eigenvalue problem
can be reduced, invoking the translational symmetry, to one involving a
Hermitian matrix in 3f dimensions. In fact, in such a case Mi:) = M,
and Ci;j(n, n') = CPPl(n- n'). For each of the N wave vectors q in a
quasi-continuous distribution within the fundamental Brillouin zone of the
crystal, we must solve the eigenvalue problem in Eq. (9) in 3f dimensions.
The eigenvalues are designated by wl(q) and the eigenvectors by ePs(q),
(s = 1,2, ..., 3f) which can be chosen to satisfy the orthogonality condition
The general solution for the displacement of the atoms in the lattice is of
the form
with
QJq, t ) =
- iws(q)Qs(q, t ) .
(22)
allow
The orthogonality of the vectors e5(q) and of the functions N-1/2eiq*"
us to solve for the coordinates QS(q,t ) in terms of u,,(t). We obtain
360
A . K . RAMDAS A N D S. RODRIGUEZ
+ G(co)A)M~/~u(O)
= 0.
(25)
+ G(o)Al
0.
(26)
A=
@ ~ - 1
1)02
- ~ - 1 / 2 ( @ ) - c)M-/~.
(29)
We consider now the case of a foreign atom of mass M , less than M , the
mass of the atom in a crystal that it replaces. If we assume further that the
force constants remain the same, A is a matrix having zeros everywhere
except at three positions along the diagonal in which the value of the matrix
element is
M - M
o2c 0.
A =
M
The eigenfrequencies of the imperfect crystal are obtained equating a 3 x 3
determinant to zero, namely
where the indices CY and aare those corresponding to the atom replaced by
the foreign atom of mass M .
On the basis of the above theory the presence of a Mn atom replacing Cd
in CdTe will result in a high-frequency local mode. On the other hand, Cd
in the hypothetical MnTe of zinc-blende structure can have only a gap mode.
The variation of the local mode of Mn in CdTe as a function of x can be
calculated, at least for small x , by considering the mutual interactions of the
local modes centered around the different Mn ions. This problem has been
9.
361
RAMAN SCATTERING
d2u
MI= -M'w;u
dt2
+ zeEp,
where Ep is the local electric field at the position of a Mn ion and ze is the
effective charge of the Mn ion upon displacement. We have assumed here
that the local mode has an inertia equal to that of the free Mn atom ; in any
case, (z/M') can be viewed as a phenomenological constant. The local
electric field Ep with the system having uniform polarization P is
Ep = E
4n
+ -P,
3
(33)
E = -4nP.
(35)
362
A . K . RAMDAS A N D S. RODRIGUEZ
8n
Ep=--P
3
(37)
Thus, from Eqs. (32) (34) (36), and (37), we find that for transverse (TO)
waves
For
respectively.
The theory outlined above has the advantage of simplicity and provides
a satisfactory account of the two-mode behavior of Cdl-,Mn,Te. It neglects
the polarization due to ionic displacements of the Cd and Te atoms and all
electronic polarizabilities except for the use of the index of refraction n. We
outline below the theory proposed by Genzel et al. (1974), that also considers
the displacements (for q = 0) of all three types of atoms and the corresponding restoring forces.
The model is based on the following equations of motion for an A B 1 - G
mixed crystal of zinc-blende symmetry, taking into account nearest-neighbor
and second-neighbor interactions:
UB)
- Xfc(UA - UC)
UA)
(42)
(43)
(44)
9.
363
RAMAN SCATTERING
and
ffA
+ UC
vo
Ej.
(45)
In these equations, m ,u, and a are the masses, displacements, and electronic
polarizabilities of the ions A, B, and C, respectively. The volume of the
primitive cell, vo ,will vary with composition and is given by vo = a3/4, where
a is the lattice parameter of the zinc-blende crystals. The subscripts b and c
on the force constants f and the Szigeti-effective charges, e, refer to the
compounds AB and AC, respectively. Equations (42)-(45) are those of
Genzel et al. (1974), with additional terms involving the second-neighbor
force constant between the B and C ions, fs. Based on the experimental
evidence that the extrapolation of the frequencies of the phonon modes in
Cdl-,Mn,Te and Znl-xMnxTe to x = 1 yields the same values, the force
constants between the Mn and Te ions, f c , must be the same in both alloy
systems as x + 1. Hence, it is necessary to let the force constants exhibit a
dependence on the lattice parameter. This dependence is approximated by a
linear function
P/(
(47)
364
A . K . RAMDAS A N D S . RODRIGUEZ
Here fib and p c denote the reduced masses of ions A and B, and of ions A
and C, respectively; v b and vc are the corresponding volumes of the primitive
cells. The static and high frequency dielectric constants of the crystals AB
and AC are given by &Ob ,&mb ,E ~ and
~ E, ~ and
~ the
, corresponding TO phonon
In addition to the force
frequencies at zero wave vector are o m and
constants, the polarizabilities and effective charges should also exhibit an xdependence, since the nearest-neighbor distances change with x. To a first
approximation, these effects will be neglected and the force constants are
assumed to have a linear dependence on the lattice parameter. The eigenfrequencies of Eqs. (42)-(44) have the form
where
and
= VOt
In the limits of x
CYA
- (1
x)aB
XaC.
(58)
9.
365
RAMAN SCATTERING
Fe
= [-&-(I
arc
+ Fs
a, - ab
+4 3 7 4
1/2
(59)
The microscopic parameters of Eqs. (54)-(58) are determined from the BornHuang relationships of Eqs. (48)-(53) and the boundary conditions of Eqs.
(59) and (60).
3. EXPERIMENTAL
RESULTS
AND DISCUSSION
Figure 3 shows the room temperature Raman spectrum of Cdo.6MnoaTe
(Venugopalan et al., 1982). The scattering configuration is y(zz)x where
x,y , and z are along [loo], [Oll], and [Oil], respectively; it allows the
observation of the Raman active phonon features having A I , E, and FZ
symmetries. The first order Raman spectrum of pure CdTe consists of a pair
of LO-TO lines of F2 symmetry occurring at 140 and 171 cm-, respectively
(Mooradian and Wright, 1968; Selders et al., 1973). We note that the alloy
exhibits a rather intense and quasicontinuous spectrum below 130 cm-, in
addition to the two pairs of relatively sharp lines, one at 143 and 158 cm-
and the other at 189 and 203 cm-. This clearly shows that the mixed crystal
has a richer and more complex spectrum as compared to that of CdTe.
30
50
I00
150
200
RAMAN SHIFT (cm-I)
FIG. 3. Raman spectrum of Cdo.6Mno.aTeat 295 K, excited with 6764 A Kr laser line. x,
y, and z are along [loo], [Oll], and [Oil], respectively. The polarization geometry and the
allowed phonon symmetries are also indicated. [After Venugopalan et at. (1982).]
366
A. K. RAMDAS A N D S. RODRIGUEZ
00
02
04
06
08
10
08
06
04
02
00
X
FIG.4. The frequencies of the Cdl-,Mn,Te and Znl-,MnxTe zone-center optical phonons
at T = 80 K. The curves were generated using the MREI model described in the text. [After
Peterson er a/. (1986).]
9.
RAMAN SCATTERING
367
368
30
90
I50
210
270
-+
9.
RAMAN SCATTERING
369
ZnTe :Mn, becomes the TO vibrational mode of MnTe. The two remaining
phonons, TO2 and L02, merge to become the gap mode of Zn in the zincblende MnTe (MnTe: Zn). As can be seen in Fig. 4, the extrapolation
of the Cdl-xMn,Te and Znl-,Mn,Te phonons to x = 1 results in the same
values for the frequencies of the TO and LO modes for the hypothetical zincblende MnTe. This effect has also been observed for other physical characteristics of these alloys. The lattice parameters (Furdyna et al., 1983) and the
energy band gaps (Lee, Ramdas, and Aggarwal, 1987; Brun del Re et al.,
1983) of Cdl-xMn,Te and Znl-xMnxTe extrapolate to the same values for
x = 1.
The curves in Fig. 4 were determined from the modified random element
isodisplacement (MREI) model. The fundamental assumptions of the
random element isodisplacement model are that in the long wavelength limit
(q 0), the anion and cation of like species vibrate with the same phase and
amplitude, and that the force each ion experiences is provided by a statistical
average of the interaction with its neighbors. This MREI model is a
modification of that developed by Genzel et al. (1974) which emphasizes the
use of the local field and is completely defined by the macroscopic parameters
of the pure end members. The modified theory is described in the theoretical
discussion given above and embodied in Eqs. (42)-(60). The additional
constraint that the frequencies of the LO and TO modes for Cdl-,Mn,Te be
equal to the corresponding modes of Znl-,MnxTe when x = 1 allows one to
incorporate second-neighbor force constants and a linear dependence of the
force constants on the lattice parameter into the model without resorting to
microscopic fitting parameters. The only fitting parameters necessary for this
model are the frequencies of the TO and LO modes of MnTe and of the gap
modes MnTe :Cd and MnTe :Zn. The frequencies of the vibrational modes
in Cdl-,Mn,Te and Znl-,Mn,Te as a function of x (Fig. 4) were determined
from the macroscopic parameters of Table VI. The resulting force constants
are also given in the table. The magnitude of the second-neighbor force
constants is about one third that of the nearest-neighbor constants. As
illustrated in Fig. 4, the curves generated from this MREI model follow the
experimental results quite well, except for the TO1 modes for which there is
a significant discrepancy in the curvature of the theory and the experimental
results. However, considering the simplifying assumptions of this model, it
is gratifying to note that the theory provides an adequate description of the
phonon frequencies in CdI-xMn,Te and ZnI-,MnxTe over the entire composition range. The theoretical model used by Venugopalan et al. (1982) for
interpreting the composition dependence of the optical phonon frequencies
in Cdl-xMnxTe is a simplification of the model presented here and, though
strictly valid only for small x, gave an adequate description of the distinct
two mode behavior in Cdl-,Mn,Te.
310
TABLE VI
PARAMETERS
IN THE MREI MODEL
Cdl -,Mn,Te
(experimental)"
Znl-,Mn,Te
(experimental)"
"fitting parameters"
wTo(MnTe) = 185 cm-'
wLo(MnTe) = 216 cm-'
wI(MnTe :Cd) = 147 cm-'
or(MnTe: Zn)= 171 cm-'
Fcd-Mn = 0.58
Fz"-M"= 0.34
T = 80K.
0.003
0.05
0.13
0.19
0.37
0.60
0.70
53
54
73
72
50
69
84
82
82
109
109
108
108
107
101
99
(IN
cm- ')
119
119
119
120
120
120
120
150
150
148
146
143
140
138
9.
371
RAMAN SCATTERING
x = 0.05, is shown in Fig. 5(b). The low frequency Raman spectrum for
x = 0.05 is similar to that for x = 0.003, except that the 2TA(L) mode has
372
151
50
too
I50
200
phonons and a less intense peak at 110 cm- due to longitudinal acoustic
(LA) phonons. Furthermore, their comparison of the frequency distributions
of CdTe and InSb shows that as compared to InSb, the optic and LA modes
of CdTe show a clear softening, but the TA modes lie closely within the
same range of frequencies for both compounds. We therefore conclude that
the peak in Fig. 7(a) at 42 cm- originates due to disorder-activated, firstorder Raman scattering by TA phonons and that it reflects the corresponding
TA phonon density of states in the mixed crystal.
9.
RAMAN SCATTERING
373
The Raman spectra of Figs. 5 and 6 demonstrate that the low frequency
disorder-induced features become more intense in Znl -,Mn,Te as the
disorder in the lattice increases. The lack of strict translational symmetry in
the alloy allows the observation of Raman features associated with nonzero
wave vector excitations. However, since the masses of Zn and Mn are more
closely matched, the deviation from strict translational symmetry is not as
severe as in Cdl-xMnxTe, for which there is a large difference in the masses
of Cd and Mn. This conclusion is based on the observation that the onephonon density of states features dominate the low frequency Raman spectra
in Cdl-,Mn,Te, in contrast to the situation that prevails for Znl-,Mn,Te.
V. Magnetic Excitations
4. GENERAL
CONSIDERATIONS
x(M) * EL.
(61)
374
A. K . RAMDAS A N D S. RODRIGUEZ
...
(62)
(63)
(64)
(66)
9.
375
RAMAN SCATTERING
where
S2
-7H
(67)
(68)
~ ( W L=F
S2)tI.
(69)
This shows that in this geometry there is a Stokes line with polarization
(6+,
2) and an anti-Stokes line with (6-,2).
In a similar way, if the incident wave propagates at right angles to H but
is polarized along H,
EL = E02 exp[- i ~ ~ f ]
(70)
and
P(') = ~ G M Esin
o 8[(f - ij)exp[-
- (f + ij)exp[-
~(OL
~(WL
-
Q)t]
+ S2)t]].
(71)
H =
--(Y
i<j
(72)
Here yl(Ri) is the electronic wavefunction normalized over the primitive cell
and evaluated at R ; ; aNo is the s-d exchange integral, NObeing the number
of primitive cells per unit volume.
376
5. PARAMAGNETIC
PHASE
We now consider Raman transitions between Zeeman sublevels of the
individual Mn2+ ions in an external magnetic field, the sample being in its
paramagnetic phase. In this phase the exchange interaction between Mn2+
ions is smaller than the thermal energy kg T and the ions can be considered
as being independent of one another. The 6S5/2 ground state of the Mn2+ion
has a total spin S = 5 / 2 , orbital angular momentum L = 0 and total angular
momentum J = 5 / 2 . In this subsection we will discuss Cdl-,Mn,Te as an
illustrative example. The cubic crystalline field (site symmetry 5)splits the
six-fold degenerate ground state into a Tg quadruplet state at + a, and a r7
doublet at -2a, where 3a is the crystal field splitting. From electron
paramagnetic resonance (EPR) experiments, Lambe and Kikuchi (1960)
obtained 3a = 0.0084 cm-' for Mn2+in CdTe. This crystal field splitting is
too small to be observed with the resolution of a standard Raman spectrometer and we treat the ground state of MnZ+in Cdl-,Mn,Te as an atomic
'S5/2 level. The application of an external magnetic field, H, results in the
removal of the six-fold degeneracy of the ground state, the energy levels being
E(ms) = gpBHms. Here ms, the projection of S along H, has the values
- 512, - 312, ..., + 5 / 2 . These energy levels form the Zeeman multiplet of
the ground state of Mn2+.
In the paramagnetic phase, Raman scattering associated with spin-flip
transitions between adjacent sublevels of this multiplet has been observed by
Petrou et al. (1983). The results in Cdl-,Mn,Te are shown in Fig. 8 for
x = 0.40. As can be seen, a strong Stokedanti-Stokes pair is observed with
a Raman shift of UPM = 5.62 f 0.02cm-' at room temperature and
H = 60 kG. Taking H and incident light parallel to 2 and denoting light
polarization with positive and negative helicity b+ and b- , respectively, the
Stokes line is observed in the (I?,, 2) configuration, whereas the anti-Stokes
line is seen in (6-, 2). When the incident light propagates at right angles to
H(2), the Stokes component appears in the polarization (i?,b-),while the
anti-Stokes is observed in (2,&+). Within experimental error the frequency
shift is linear in H . With the energy separation between adjacent sublevels
of the Zeeman multiplet given by AE = gPBH = ~ U P M ,it is found that
g = 2.01 f 0.02. The Raman line at UPM in Cdl-,Mn,Te has been observed
for a variety of compositions ranging from x = 0.01 to x = 0.70.
Following the arguments given by Fleury and Loudon (1968) one can
consider, as a possible mechanism for the UPM Raman line, a two step
process having as the intermediate state one of the excited states of the Mn2+
ion (L = 1, S = 5/2). Figure 9 shows such mechanisms for the Stokes and
antistokes components of the UPM line. For the Stokes component an
incident photon of energy tzu~
and polarization 6, induces a virtual electric
9.
40 r
377
RAMAN SCATTERING
- F,z,
A
__
u- , z
T.300 K
H.60 kG
10
5
0
5
RAMAN SHIFT ( c m ' )
10
FIG.8. Stokes (S) and anti-Stokes (AS) Raman lines at WPM resulting from Ams = 1 spinflip transitions within the Zeeman multiplet of MnZ+in Cdl-,Mn,Te, x = 0.40: -(a+, 8;
(a_,2). The wavelength of the exciting laser line A L = 6764 A; the applied magnetic field
H = 60 kG; the temperature T = 300K; x , y , and z are along [OOI], [ITI], and [IIO],
respectively. [After Petrou et al. (1983).]
_._.
378
A. K . RAMDAS A N D S . RODRIGUEZ
L=0
s= 512
mJ=mS J = 512
mJ+I
I '
FIG.9. Raman mechanism for the WPM line involving the internal transitions of the Mn2+
ion. The arrows indicate virtual electric-dipole transitions. The energy level scheme is not to scale
and the energy difference between the excited and ground states EOis much greater than ~ W P M .
[After Petrou et al. (1983).]
Si* s
= S?)$(Z)
+Si(+)$(-)
+ +&W$(+).
(73)
here S f * ) and s(*) are the spin raising and lowering operators for a Mn2+ion
and band electron, respectively, and S?) and s(') are the corresponding
projections of spin along 2. The second term of Eq. (73) raises the spin of
a Mn2+ion while simultaneously lowering the spin of a band electron, i.e.,
(mS)Mn2+lmJ)e
ImS
l)MnZ+ImJ- 1 ) e .
(74)
379
9. RAMAN SCATTERING
In a similar fashion, the third term lowers the spin of an ion while raising
the spin of a band electron, i.e.,
ImS)Mn2+(mJ)e4ImS.- 1)Mn2+(mJ-b 1)e.
(75)
Hence, these terms can induce simultaneous spin-flips of the band electrons
on the one hand and the Mn2+ ions on the other, corresponding to
Arns(Mn2') = f 1 and AmJ(e)= =F 1.
In Fig. 10 we show the above mechanism for both the Stokes and the antiStokes component and the two right-angle geometries considered above. In
the presence of a magnetic field the Ts valence band splits into four subbands
with m~ = - 3/2, - 1/2, + 1/2, and + 3/2, and the l-6 conduction band splits
into ~ T Z J= + 1/2 and - 1/2 subbands. The possible processes for the Stokes
2) configuration are shown in Fig. 10a. In
component appearing in the (6+,
Stokes
anti -Stokes
FIG.10. Raman mechanism for the WPM line involving the band electrons: CB and VB refer
to the conduction and valence bands, respectively, which are labeled by the electronic quantum
number r n .~The single arrows indicate virtual electric-dipole transitions, while the double
arrows refer to transitions induced by the electron-Mn*+ exchange interaction. The allowed
(c), while those for the antipolarizations for the Stokes component are (8,.2): (a) and (i,&-):
Stokes are (&-, .?): (b) and (2, &+): (d). [After Petrou et al. (1983).]
380
9.
RAMAN SCATTERING
Stokes
381
(&+,&-I
+ 1/2
-3/2
-112
+ 1/2
+3/2
(a 1
(C)
- 3/2
- 1/2
+ 1/2
+3/2
(d)
FIG. 1 1 . Raman mechanism for the Ams = i=2 spin-flip transition, i.e., the 2mPM line,
involving the band electrons. The Stokes line is allowed in (6+,
6-)polarization, whereas the
anti-Stokes appears in (6-,
6+).[After Petrou et nl. (1983).]
laser as the exciting radiation. The forward scattering geometry was used in
order to obtain the (a,, 6-) and (6-, 6+)configurations. As can be seen in
Fig. 12, the Stokes and the anti-Stokes lines appear with the expected
polarizations. Under the resonant conditions of this experiment the OPM line
is extremely intense (- lo6 counts/sec); although this line is forbidden for
these polarizations, it is not surprising that the leakage observed is quite
strong.
It is known that electrons and holes in polar crystals interact strongly with
zone center longitudinal optical (LO) phonons through the Frohlich interaction (see Hayes and Loudon, 1978). An LO phonon can be created or
annihilated as a result of such an interaction. Referring to the mechanism
responsible for the OPM line, shown in Fig. 10, one can visualize a fourth step
382
A . K . RAMDAS A N D S. RODRIGUEZ
RAMAN SHlFTkm-')
FIG.12. Stokes and anti-Stokes components of 2WPM in Cdl-,Mn,Te, x = 0.10, recorded for
the (a+,&) and (a_,8+)polarizations. The sample temperature T = 120 K, H = 60 kG, and
A L = 7525 A; x , y , and z are along [I lo], [TlO], and (0011, respectively. The WPM line appearing
in the spectra is due to leakage. [After Petrou ef al. (1983).]
in which the excited electron or hole interacts with the lattice and creates or
annihilates an LO phonon. Such a mechanism would result in a scattered
photon with a Raman shift of WLO f W P M . The net result for the Stokes
process with a shift of WLO + WPM is that an LO phonon is created and a
Mn2+ion is excited to the next sublevel of the Zeeman multiplet. A Stokes
shift of WLO - WPM corresponds to the creation of an LO phonon and the
de-excitation of a Mn2+ion to the next lower sublevel of the multiplet. The
W L O + OPM Stokes Raman line is expected to appear in the (&+ ,2) or the
( f , b - ) configurations, whereas the WLO - WPM Stokes line is allowed for
(8- , 2) or (2,8+).
In addition to the above, the creation or annihilation of two LO phonons
in a similar process is also possible. An electron or a hole can create or
annihilate two LO phonons in a single step. Another possibility is that an
electron as well as a hole each separately create an LO phonon. In either case,
the scattered photon can have the following Raman shifts: 2 ~ ~ f0WPM
, ,
WLO, + WLO, f WPM or 2 ~ ~ f0W,P M , where LO1 refers to the CdTe-like
and LO2 to the MnTe-like zone center longitudinal optical phonons
(Venugopalan et al., 1982).As before the Stokes lines involving the excitation
of a Mn2+ion are expected in the (a+,2) or (2, 8-) polarization geometries,
while those involving a de-excitation of a Mn2+ ion appear in (8-,2) or
(2, 8+).
9.
RAMAN SCATTERING
383
The new lines described above should occur only under conditions of bandgap resonance. Under such conditions we have indeed observed the new
Raman lines with shifts of WLO f WPM in Cdl-,MnxTe for a variety of
compositions. The Raman spectra in the region of the longitudinal and
transverse optical (TO) vibrational modes are shown in Fig. 13 for
Cdl-,MnxTe with x = 0.10. The 7525 A Kr+ laser line was used to excite the
spectra. The zero magnetic field LO and TO phonon spectrum is shown in
Fig. 13b. Here the CdTe-like TO and LO and the MnTe-like LO modes are
quite distinct, while the MnTe-like TO appears as a shoulder to the LO. The
corresponding Raman spectra, recorded in the presence of a magnetic field
of 60 kG and in the (2,6+)
and (2,8-) configurations, are presented in Fig.
13a and 13c, respectively. The additional Raman lines with Stokes shifts
WLO f up^ are clearly present with the proper polarization characteristics.
The sample temperature was 120K; at this temperature, the sublevel
occupation probability ratio of adjacent levels in the Zeeman multiplet of
Mn2+is exp(- h ~ p ~ / Tk )g= 0.94. Thus the intensities of the WLO + WPM
and the oL0- WPM lines are expected to be approximately equal. The
WLO f WPM lines can be observed only when the exciting photon energy is
I
I
24
H=GOkG
61d
L A
~~
"130
170
210
Ix)
170
21
384
A . K . RAMDAS A N D S. RODRIGUEZ
-k
+ WPM
- WL)(Wo - U L )
+ 5 additional terms
l2
. (76)
+ B,
hub = Ex - 3A + 3B
hwa
Ex - 3A
(77)
+xNo/3(S,Mn>.
(78)
9.
385
RAMAN SCATTERING
Cd,-,Mn Te
X=0.05
PHOTON ENERGY ( e V
FIG.14. The photoluminescence spectra of Cdc.ssMno.o5Teat T = 5 K with H = 0 for (a)and
(b), H = 60 kG for (c). I L = 7525 A for (a) and (c), A L = 6764 A for (b), and laser power
PL = 25 mW for all cases. Raman features are denoted by 'R'. In (c) the features between 1.55
and 1.6eV are also displayed on a scale reduced by 10. [After Peterson et al. (1985a).]
the 3d states of Mn2+ and conduction (valence) electrons and <S,"n)is the
average value of the component of Mn2+ spin along H given by
(5/2)&/2(g,uBH/kB T ) in the paramagnetic phase. Thus the energies h&,b
are characterized by an effective g factor that is a non-linear function of
(H/T).
In Fig. 14(a) and (b), we show the photoluminescence spectrum of
Cdl-xMnxTe,x = 0.05, at 5 Kin the absence of a magnetic field. The feature
labeled X is attributed to free exciton recombination (Plane1 et al., 1980) and
has an energy of Ex = 1.665 eV, an increase of 70 meV from the value in
CdTe (Zanio, 1978, p. 100). The energy of this exciton varies linearly with
Mn2+ concentration (Lee and Ramdas, 1984) in Cdl-,Mn,Te and the value
386
measured here is in good agreement with that expected. The feature labeled
AoX is attributed to the exciton bound to a neutral acceptor (Plane1 et al.,
1980) with a binding energy of 9meV. The feature at 1.608eV, when
corrected for the change in the energy gap, corresponds to the 1.54 eV feature
in CdTe which may result from free electron to acceptor transitions (Zanio,
1978, p. 151); the features at 1 3 3 7 and 1.566 eV are the LO phonon replicas
of this transition. Similarly, the series of luminescence peaks beginning at
1 . 5 eV are the LO phonon replicas of the transitions associated with a
vacancy-donor complex (V&D) (Zanio, 1978, p. 139), where the donors in
this case are presumed to be interstitial copper atoms (Zanio, 1978, p. 144)
introduced during the crystal growth. As the magnetic field is increased the
free exciton peak increases in intensity and shifts to lower energy sweeping
across the AoX feature (Plane1 ef al., 1980). At sufficiently high magnetic
fields a splitting in the free exciton feature is observed (Ryabchenko et al.,
1981). Two of these are present in Fig. 14(c); the low energy component
labeled X + appears in the 6+polarization, while the other, X , , is polarized
along 2. The feature associated with the free electron to acceptor transition
and its LO phonon replicas shift by 14 meV in a magnetic field of 60 kG.
On the basis of the luminescence spectra, appropriate choices of laser
wavelength, sample temperature and magnetic field can be made to achieve
conditions of in resonance or out resonance which can selectively
enhance specific features in the Raman spectrum. In the experiments
reported here, the OLO - WPM line, i.e., the Raman shift with the creation
of an LO phonon and the de-excitation of the MnZ+by Ams = - 1 involves
a virtual transition at an energy tiw, for the incident radiation polarized along
i and another at hob for scattered radiation having 6+polarization. With
the 7525 A ( ~ O L = 1.648 eV) laser line, the in resonance condition is nearly
fulfilled; with A and B in Eq. (78) for x = 0.05, T = 5 I<, and H = 60 kG,
we estimate the energies of X , and X , , the components which move to lower
energies with increasing H , to be 1,.636eV and 1.603 eV, respectively (here
we use Noa = 220 meV and NoP = - 880 meV given by Gaj et al. (1979)).
Thus, it is clear that the in resonance condition is satisfied even more closely
whereas the out resonance for OLO - OPM can now be realized. The results
in Fig. 15 show the resonantly enhanced lines at O P M , 2 0 P M , 3 W P M , and
4UpM. We discuss some of the underlying physical considerations later. Here
we emphasize the dramatic enhancement in the intensity of the OLO, - OPM
line as illustrated in Figs. 16 and 17. The spectra were recorded with the
incident polarization along 2, unanalyzed scattered radiation and ks I/ H. The
relatively broad feature at
220 cm- is the X + luminescence feature
attributed to the exciton component at Ex - 3A + 3B. In addition to the
LO1, LO1 f PM, LOz, and LO2 f PM lines, that have been reported
before (Petrou et al., 1983), two additional features with Raman shifts of
9.
387
RAMAN SCATTERING
I
Cd - M n xTe
OO:
10
20
388
A. K . RAMDAS A N D S . RODRIGUEZ
Cd - Mn Te
X =
0.05
5K
60 k G
L O , t PM
[rLOl
0'
140
180
t 2PM
220
We now discuss the origin of the Raman lines in Fig. 15 with shifts of 3 0 p M
and 4 0 p M , corresponding to Ams = 3 and 4, respectively. Mutiple spin-flip
Raman scattering from electrons bound to donors (Oka and Cardona,
1981a, b) has been reported in the literature and explained invoking either an
exchange coupling among donor spins (Economou et al., 1972) or multiple
scattering (Wolff quoted by Geschwind and Romestain, 1984). The multiple
spin-flip features in DMS can be accounted for in terms of excitations within
neighboring pairs of Mn2+ions coupled antiferromagnetically and assuming
an anisotropic exchange interaction between the ground state multiplet of
one and an excited state of the other. In fact, a pair of neighboring Mn2+,
say 1 and 2, in their 'S5/2 states in a magnetic field fill 2 can be described
by the Hamiltonian
H = gpBH& -
J(s2 - y ) ,
(79)
9.
389
RAMAN SCATTERING
LO1 - PM
Cd,-,Mn, Te
X = 0.05
5K
35 k G
J
LO,-2PM
40
180
2 20
6 . MAGNETICALLY
ORDERED
PHASE
As mentioned earlier in this Section, Cdl -*Mn,Te exhibits a magnetically
ordered low temperature phase for x > 0.17. The transition from the
paramagnetic to the magnetically ordered phase is accompanied by the
390
9.
01
391
RAMAN SCATTERING
50
100
I50
200
M(r, t ) =
c M4 exp(iq
4
r - iw,t).
(80)
In first order Raman scattering, only the long wavelength magnons can be
excited. In an antiferromagnetic system these excitations can be described by
classifying the spins into those which, in the state of equilibrium, point in
one direction and those pointing in an anti-parallel direction. This classification gives rise to magnetizations MI and M2 where MI is the magnetic
moment per unit volume of the spins of the first class and Mz that of the
second. In equilibrium MI + M2 = 0. Now the agents responsible for the
preferential orientation of a spin of type 1 are those of type 2 and conversely,
the former are in a molecular field of the form - AM2 + HY where the first
term, called the exchange field, is isotropic and the second points in a
preferred crystallographic direction and i s , thus, called the anisotropy field.
392
3
3.6 -
3.2 -
- 2.8-
0
v)
al
$
=
2.4-
40.0 K-
mg 2.0-
34.0 K -
>-
5z
1.6-
32.0 K-
1.2-
30.0 K
20.0 K
0.8 0.4
01
5.0 K
\L-44
-p-----,--~5.0
I
K I
+ Ha")
dMz
- - - yMz x (H - AM1
dt
+ Ha")
--
and
9.
393
RAMAN SCATTERING
-M ;
d
dt
d
-M?
dt
iy(H -t HA-k H E ) M 7
~
M$
(83)
r i y ( H - HA - HE)M+ f iyHEMt
(84)
where M t and M? are MI, f iM1, and MzX f iMzy, respectively. Equations
(83) and (84) have non-trivial solutions of frequency u& given by
a&
fyH + I y l ( H i
+ 2HAHE).
(85)
M S = (hgpBs/a3)BS(r),
(86)
where pug is the Bohr magneton, a the lattice constant, and Bs(y)the Brillouin
function with
+1)~.
(88)
394
A . K . RAMDAS A N D S. RODRIGUEZ
I
Cd ,+ MnxTe
A x.0.7
TEMPERATURE (K)
FIG.21. Temperature dependence of the magnon peak frequency in the magnetically ordered
phases of Cdl-,Mn,Te. Triangles: x = 0.7, antiferromagnetic phase; circles: x = 0.4, spinglass phase. The solid curve passing through the points for x = 0.7 is calculated for TN = 40 K
solving Eq. (87) for MS as a function of T 5 TN and assuming WM is proportional to MS and
equal to 12.5 cm- at T = 0 K. [After Venugopalan el a/. (1982).]
9.
1
AS
(a)
(&.a
H=GOkG
(C)
($.
395
RAMAN SCATTERING
I
M-
AS
2)
H=GOkG
M-
24
12
0
12
24
RAMAN SHIFT (cm-')
FIG. 22. Effect of theommagnetic field on the magnon feature of Cd,-,Mn,Te, x = 0.70, at
T = 5 K with A L = 5682 A. x , y , andz are along [110], [TlO], and [Ool], respectively: (a) (6-, t),
H = 60 kG; (b) (6+,
2). H = 0; (c) (a+,2). H = 60 kG. Owing to imperfect polarizationresults,
2) configuration.
leakage of the fairly strong featureM+ appears as a small shoulder in the (6-,
[After Petrou et a/. (1983).]
initially broadens and then moves towards higher Raman shifts as a consequence of the increased importance of the exchange interaction. An
increase in Raman shift is observed at temperatures well above TN 40 K.
As the temperature is lowered through and below TN, the line becomes the
magnon component observed in the (a+,2) configuration.
396
A. K . RAMDAS A N D S. RODRIGUEZ
I
I
OOK
300K
I
xi6
6K
xi6
56c4
~ 6 c i
10
10
10
20
1 -
10
20
FIG.23. Evolution of the Raman line at WPM of Cdl-xMn,Te, x = 0.70, into the magnon
feature as the temperature is lowered from room tempefature to below the Ntel temperature.
The spectra were recorded with H = 60 kG, I L = 6764 A in the (a,, 2) polarization; x , y , and
z are along [IlO],[TlO], and [@I], respectively. [After Petrou et a/. (1983).]
Finally, the conclusion that the magnetic feature observed in the magnetically ordered phases is a one magnon excitation was initially deduced from
its polarization characteristics and temperature behavior (Venugopalan
et al., 1982); this is supported by the results of Ching and Huber (1982a,
1982b). The fact that this feature shows a splitting in the presence of a
magnetic field and that the OPM line of the paramagnetic phase, clearly
associated with a single ion excitation, evolves smoothly into the higher
energy component of the magnon provides a strong confirmation of this
interpretation. In the same spirit, one might expect a two magnon feature
associated with the 2 0 p M line; however, given the intensity of the 2 0 p line
~
compared to that of the OPM line, the intensity of such a feature would
preclude its observation. A two magnon feature similar to that seen in MnF2
by Fleury and Loudon (1968) would have symmetric polarization characteristics; such a feature has also not been observed in Cdl -,Mn,Te. In the light
of these experimental results, the two magnon interpretation advanced by
Grynberg and Picquart (1981) to explain the feature at OM is clearly excluded.
7. SPIN-FLIPRAMANSCATTERING
Since the first observations of electron spin-flip Raman scattering in the
narrow band gap semiconductor InSb by Slusher et al. (1967) followed by
9.
RAMAN SCATTERING
397
that in the wide band gap semiconductor CdS by Thomas and Hopfield
(1968), there has been a continuing interest in this magneto-optical phenomenon. An important contribution to this field is the first demonstration
of the spin-flip Raman laser by Patel and Shaw (1970). The success of the
InSb spin-flip Raman laser as a practical source of magnetic field tuned,
coherent radiation in the 5 to 15 pm spectral region (Patel and Shaw, 1970;
Mooradian et al., 1970) can be traced to the large effective g-factor of the
electron, 1g*I = 50. In contrast, the wider band gap II-VI semiconductors
CdS, CdSe, and CdTe have rather small spin splittings (Thomas and
Hopfield, 1968; Walker et al., 1972) characterized by 1g*I 5 2, since their
conduction and valence band spin-orbit interactions are small in comparison
to their energy band gaps.
When detectable, spin-flip Raman scattering provides a practical means
of probing the electronic structure of semiconductors, as dramatically
illustrated in DMS. The large Raman shifts associated with spin-flip
scattering from electrons in DMS were first observed in the narrow gap
Hgl-,Mn,Te by Geyer and Fan (1980). The first evidence of a finite spinsplitting of the electronic level in the absence of a magnetic field was reported
by Nawrocki et al. (1980, 1981) in the wide gap diluted magnetic semiconductor Cdl-,Mn,Se. The effects of Mn concentration and the antiferromagnetic coupling among the Mn*+ ions on the spin-flip Raman shifts
were first observed (Peterson et al., 1982) in a study of Cdl-,Mn,Te. These
studies established the nature of Raman scattering associated with the spinflip transitions of electrons bound to donors in DMS. The large Raman shifts
depend not only on the applied magnetic field, but also on temperature
and manganese concentrations. The far-infrared absorption spectra
(Dobrowolska et al., 1982) of Cdl-,Mn,Se also provided evidence for these
spin-flip transitions. Following these initial reports, there have been several
investigations (Alov et al., 1981, 1983; Heiman et al., 1983a,b, 1984a,b;
Douglas et al., 1984; Peterson et al., 1985b) of spin-flip Raman scattering
in DMS. The extensive results currently available allow a detailed comparison with the theory of Diet1 and Spalek (1982, 1983).
The Raman spectra of Cdl-,Mn,Te(Ga), x = 0.03, are shown in Fig. 24
for the (6+,2) and (6-,2) polarization configurations with T = 40 K and
H = 60 kG. The two Stokes features labeled PM and SF are present only in
(a+,i),while the corresponding anti-Stokes features appear only in the
(6-, 2) configuration. The observed width of the PM line is instrument
limited, while that of the SF feature is 23 cm-'. The PM feature is associated
with the spin-flip transitions.
The SF feature of Fig. 24 is attributed to spin-flip Raman scattering from
electrons bound to gallium donors. It has the same polarization characteristics as those of the PM line appearing (Peterson et al., 1982)in the (a+,2)
398
20
I0
-10
-20
-3
FIG. 24. Raman spectra of Cdl-,Mn,Te(Ga), x = 0.03, showing the Ams = f 1 transitions
within the Zeeman multiplet of MnZf(PM) and the spin-flip of electrons bound to Ga donors
(SF). kcps = 10 countslsec. [After Peterson et a/. (1985b).]
or (&a_)polarizations for Stokes scattering and in (8-, 2) or (2, a+)for antiStokes. As illustrated in Fig. 25, the peak Raman shift of the spin-flip
feature, 6,exhibits a strong .dependence on both temperature and magnetic
field. The primary source of the spin splitting of the electronic level is the
exchange coupling with the Mn2+ ions (first term of Eq. (72)) with the
Zeeman effect (second term of Eq. (72)) making a relatively small contribution. Hence, the Raman shift should be approximately proportional to the
magnetization of the Mn2+ ion system, which amplifies the effect of the
magnetic field on the electron. As can be seen in Fig. 25, a finite Raman shift
is observed for zero magnetic field. This effect is attributed by Diet1 and
Spalek (1983) to the bound magnetic polaron (BMP). The electron
localized on a donor in a diluted magnetic crystal polarizes the magnetic ions
within its orbit, creating a spin cloud that exhibits a net magnetic moment.
9.
I- Cd,-,Mn,Te
(Ga 1
X = 0.03
'E
90 -
z
rn
+ + +
I-
LL
399
RAMAN SCATTERING
0
0
60-
0
A
LL
+
o
A
20
:40Ki
40
60
MAGNETIC FIELD ( k G )
FIG.25. Magnetic field and temperature dependence of the Raman shift associated with the
spin-flip of electrons bound to donors in Cdl-,Mn,Te(Ga), x = 0.03. [After Peterson el uf.
(1985b).]
An additional effect on the binding energy of the electron bound to the donor
originates from thermodynamic fluctuations of the magnetization and the
resulting spin alignment of the magnetic ions around the donor.
According to Eq. (72), the spin splitting of the donor energy levels in DMS
arises from the combined effect of the magnetization of the Mn2+ions and
the external field H. Due to the strong s-d coupling, the effect due to the
magnetization dominates. The Raman shift associated with spin-flip
scattering from the donor states has the form
where MOis the macroscopic magnetization. The magnetization is proportional to the thermal average of the Mn2+spin projection along H multiplied
400
A . K . RAMDAS A N D S. RODRIGUEZ
X C Y N O ( S+~ g*pBH.
~)
(90)
Here R is the concentration of Mn2+ ions that contribute to the magnetization. For small x, the crystal is paramagnetic and the thermal average of the
Mn" spins is
160 C d
,-, MnT,e
'E
120
0
LL
+s
80
z
a
0 O
Ox:o.05
--
+x=o
A A
x=o.3
A
A
A
A
+O
+O
oX=0.20
A
__
X = 0.03
__
h
+X=O.Q- Cd,-,Mn,Te
T=1.8K
n
+n
I-
V,
+ +
T = 1.8K
40
20
40
20
40
60
MAGNETIC FIELD ( k G )
FIG.26. Magnetic field and composition dependence of the peak spin-flip Raman shift in
the Cdl-,Mn,Te samples at T = 1.8 K . [After Peterson et a/. (1985b).]
9.
RAMAN SCATTERING
401
but less pronounced. For x = 0.10, the deviation from the paramagnetic
behavior is quite evident. For H = 60 kG, the Raman shift for x = 0.10 is
only four times that for x = 0.01. As x exceeds 0.10, the Raman shifts for
a given field actually decrease; note that the shifts for the x = 0.20 sample
lie below those for the x = 0.10 sample. And the Raman shifts for the
Cdl-,Mn,Te(Ga), x = 0.30, sample are significantly smaller than those for
the x = 0.10 and x = 0.20 samples. These trends have their origin in the
decrease of the mean magnetic field due to the increasing antiferromagnetic
pairing of MnZ+neighbors.
In order to discuss the experimental results on the bound magnetic
polaron, it is useful to review the theory for it as developed by Dietl and
Spalek (1982, 1983). The electron is assumed to be bound to a shallow donor
interacting through an s-d coupling with a paramagnetic subsystem of
localized magnetic moments. Only the large polaron case is considered,
allowing the continuous-medium, effective-mass, and molecular-field
approximations. This model assumes that the donor electron interacts with
a large cloud of spins behaving classically.
Neglecting field induced anisotropy, Dietl and Spalek derive an effective
BMP Hamiltonian given by
where ED(^) is the binding energy of the electron arising from the Coulomb
potential, A is the magnitude of the spin-splitting with A, parallel to the
effective local magnetic field, defining the direction of the spin quantization
of the electron,
is the characteristic BMP energy for an s-type wave function with an effective
Bohr radius a. Here Mo is the magnetic field induced magnetization and x
is the magnetic susceptibility. At a finite temperature, a range of A beyond
the minimum of H ( A )is accessible to the system. The probability of a thermal
fluctuation of the magnetization giving rise to a specific A is
I : [: :
P(A) = Cexp --
(95)
402
A. K . RAMDAS A N D S. RODRIGUEZ
where
where Cis a constant related to the scattering cross section, and f refers to
the Stokes and anti-Stokes components of the line. The peak position of the
Raman line, 6, satisfies
'F
("4"'
- dAocoth - & p k ~ T
) - 4&pk~T=
0,
(100)
9.
RAMAN SCATTERING
403
96 na3No
a
0
a
?
?;I
*wg
u
&
l3
3
15
10
15
10
15
MAGNETIC FIELD ( k G )
FIG. 27. Magnetic field and temperature dependence of the peak spin-flip Raman shift in the Cdl-,Mn,Te(Ga), x = 0.05, 0.10, and 0.20,
samples at low field. [After Peterson et al. (1985b).]
a
N
9.
RAMAN SCATTERING
405
406
A. K. RAMDAS A N D S . RODRIGUEZ
9.
501
-2b
407
RAMAN SCATTERING
-10
10
20
FIG. 29. Stokes (S) and anti-Stokes (AS) components of the folded longitudinal acoustic
x = 0.5 and y = 0.11. [After
branch of the superlattice Cdl-,Mn,Te/Cdl,Mn,Te,
Venugopalan et al. (1984).]
408
A . K . RAMDAS A N D S. RODRIGUEZ
A.
SL
1500
325 - 4
H=60KG
v)
CL
>
1000
v)
z
W
500
200
400
600
9.
409
RAMAN SCATTERING
(Venugopalan et al., 1982). Their respective positions are 167 (LOI) and 199
presumably due to the flatness of the dispersion curves of these modes. The
additional groups of peaks seen here, centered at 370, 530,and 700 cm-',
originate from strong resonant scattering by overtones and combinations of
the two fundamental LO phonons. The clear observation of second and third
overtones in a very thin (1.6,um) sample, where the effective scattering
volume is necessarily small, attests to the resonant enhancement. It is
probable that the frequencies of all these features are influenced by any
strains present within the superlattice. We further note that although
transverse optic modes are also allowed in the experimental geometry, their
intensity was too weak to be detected. With 7525A excitation-that lies
below the energy gap of the superlattice-none of the features, including LO
modes, could be seen. Based on these facts and that the excitation employed
for the spectrum shown in Fig. 30 is above the energy gap, the selective
enhancement seen for the LO modes and their overtones suggests an underlying Frohlich mechanism. The sloping background beyond 640 cm-' is due
to the onset of the photoluminescence peak. Under an applied magnetic field,
the position of the luminescence peak shifts towards lower photon energy
with an effective g-factor of 100 at 5 K; these data, demonstrate that the
very large electronic g-factors first reported by Gaj et al. (1978a) for
Cdl-,Mn,Te alloys are also realized in DMS superlattices.
We finally discuss the question of magnetic excitations within a DMS
superlattice. As noted earlier, Galazka et af., (1980) have shown that
Cdl -,Mn,Te alloys possess a spin-glass phase at low temperatures for the
composition range 0.17 Ix 5 0.60. In single crystals with x = 0.5 and 0.4,
Venugopalan et af. (1982) detected, in the absence of an external magnetic
field, a distinct magnon Raman line characteristic of the spin-glass phase.
In the superlattice, there appears to be no evidence for such a zero-field
excitation. Instead, with an applied field of 60 kilogauss, one observes an
intense peak originating from the Raman-paramagnetic resonance (OPM)
between the Zeeman sublevels of the ground state of MnZ+ions (Petrou
et af.,1983). In addition, the first and second overtones of OPM are also easily
detectable. Venugopalan et af. (1984) attribute this to the possibility that even
at 5 K the 59 A thick layers (with x = 0.5) of the superlattice may lack the
spin-glass ordering, although it does occur for the same composition of the
bulk alloy. If so, this would also signify that the dimensional constraints
created in a superlattice can exert a decisive influence on magnetic phase
transitions.
In summary, this study illustrates that Raman scattering can serve
as a sensitive, diagnostic probe of diverse structural aspects of DMS
superlattices.
410
Acknowledgments
The support from the National Science Foundation during the preparation of the chapter is
gratefully acknowledged (Grant No. DMR-84-03325).
References
Alov, D. D., Gubarev, S. I., Timofeev, V. B., andshepel, B. N. (1981). PismaZh. Eksp. Teor.
Fiz. 34, 76; JETP Lett. 34, 71.
Alov, D. L., Gubarev, S. I., and Timofeev, V. B. (1983). Zh Eksp. Teor. Fiz. 84, 1806; Sov.
PhyS.-JETP 57, 1052.
Barker, A. S . , and Loudon, R. (1972). Rev. Mod. Phys. 44, 18.
Barker, A. S., and Severs, A. J. (1975). Rev. Mod. Phys. 47, Suppl. 2.
Bhagavantam, S., and Venkatarayudu, T. (1969). Theory of Groups and Its Application to
Physical Problems. Academic Press, New York.
Bicknell, R. N., Giles-Taylor, N. C., Blanks, D. K., Yanka, R. W., Buckland, E. L., and
Schetzina, J. F. (1985). J. Vac. Sci. Technol. B3, 709.
Born, M., and Huang, K. (1968). Dynamical Theory of Crystal Lattices. Oxford University
Press. London.
Brun del Re, R., Donofrio, T., Avon, J., Majid, J., and Woolley, J. C. (1983). I1 Nuovo
Cimento 2, 1911.
Cardona, M. (1975). Light Scattering in Solids, Vol. 1. Springer, Berlin.
Cardona, M., and Giintherodt, G. (1982a, b). Light Scattering in Solids, Vols. 2,3.Springer,
Berlin.
Cardona, M., and Giintherodt, G. (1984). Light Scattering in Solids, Vol. 4.Springer, Berlin.
Ching, W. Y., and Huber, D. L. (1982a). Phys. Rev. B25, 5761.
Ching, W. Y., and Huber, D. L. (1982b). Phys. Rev. B25, 6164.
Colvard, C., Merlin, R., Klein, M. V., and Gossard, A. C. (1980). Phys. Rev. Lett. 45,298.
Colvard, C., Cant, T. A., Klein, M. V., Merlin, R., Fischer, R., Morkoc, H., and Gossard,
A. C. (1985). Phys. Rev. B31, 2080.
Datta, S., Furdyna, J. K., and Gunshor, R. L. (1985). SuperlatticesandMicrostructures 1,327.
Dietl, T., and Spalek, J. (1982). Phys. Rev. Lett. 48, 355.
Dietl, T.,and Spalek, J. (1983). Phys. Rev. B28, 1548.
Dingle, R. (1975). Advances in Solid State Physics (H. J. Queisser, ed.), Vol. 15, p. 21.
Pergamon Vieweg, Braunschweig.
Dobrowolska, M., Drew, H. D., Furdyna, J. K., Ichiguchi, T., Witowski, A., and Wolff, P.
A. (1982). Phys. Rev. Lett. 49, 845.
Douglas, K., Nakashima, S., and Scott, J. F. (1984). Phys. Rev. B29, 5602.
Economou, E. N., Ruvalds, J., and Ngai, K. L. (1972). Phys. Rev. Lett. 29, 110.
Fleury, P. A., and Loudon, R. (1968). Phys. Rev. 166, 514.
Frohlich, H. (1958). Theory of Dielectrics, 2nd Edition. Oxford University Press, London.
Furdyna, J. K., Giriat, W., Mitchell, D. F., and Sproule, G. I. (1983). J. Solid State Chem.
46,349.
Gaj, J. A., Ginter, J., and Galazka, R. R. (1978a). Phys. Stat. Solidi (b) 89, 655.
Gaj. J. A., Galazka, R. R., and Nawrocki, M. (1978b). Solid State Commun. 25, 193.
Gaj, J. A., Planel, R., and Fishman, G. (1979). Solid State Commun. 29, 435.
Galazka, R. R., Nagata, S., and Keesom, P. H. (1980). Phys. Rev. B22, 3344.
Genzel, L., Martin, T. P., and Perry, C. H. (1974). Phys. Status Solidi B62, 83.
Geschwind, S., and Romestain, R. (1984). Light Scattering in Solids IV (M. Cardona and
G. Giintherodt, eds.). Springer, Berlin.
9.
RAMAN SCATTERING
411
412
A . K . RAMDAS A N D S . RODRIGUEZ
Petrou, A., Peterson, D. L., Venugopalan, S., Galazka, R. R., Ramdas, A. K., and Rodriguez,
S. (1983). Phys. Rev. B27, 3471.
Pine, A. S., and Dresselhaus, G. (1969). Phys. Rev. 188, 1489.
Pine, A. S., and Dresselhaus, G. (1971). Phys. Rev. B4, 356.
Planel, R., Gaj, J., and Benoit a la Guillaume (1980). J. Phys. (Paris) Colloq. 41, C5-39.
Ploog, K., and Dohler, G. H. (1983). Advan. Phys. 32, 285.
Rodriguez, S, and Rarndas, A. K. (1985). Inelastic Light Scattering in Crystals, Highlights
of Condensed Matter Theory, Proceedings of the International School of Physics Enrico
Ferrni (Course LXXXIX) (F. Bassani, F. Furni, and M. P. Tosi, eds.), pp. 369-420. North
Holland, Amsterdam.
Rowe, J . M., Nicklow, R. M., Price, D. L., and Zanio, K . (1974). Phys. Rev. B10, 671.
Ryabchenko, S. M., Terletskii, 0. V., Nizetskaya, I. B., and Oleinik, G. S. (1981). Fiz. Tekh.
Poluprovodn. 15, 2314; Sov. Phys. Semicond. 15, 1345.
Rytov, S. M. (1956). Akust. Zh. 2, 71; Sov. Phys. Acoust. 2 , 68.
Sandercock, J. R. (1982). Light Scattering in Solids (M. Cardona and G. Giintherodt, eds.),
Vol. 3, p. 173.
Sapriel, J., Michel, J. C., Toledano, J. C., Vacher, R., Kervarec, J., and Regreny, A. (1983).
Phys. Rev. B28, 2007.
Selders, M., Chen, E. Y., and Chang, R. K. (1973). Solid State Commun. 12, 1057.
Sennett, C. T., Bosomworth, D. R., Hayes, W., and Spray, A. R. L. (1969). J. Phys. C2,1137.
Slusher, R. E., Patel, C. K . N., and Fleury, P. A. (1967). Phys. Rev. Lett. 18, 77.
Thomas, D. G., and Hopfield, J. J. (1968). Phys. Rev. 175, 1021.
Twardowski, A., Nawrocki, M., and Ginter, J. (1979). Phys. Stat. Solidi. (b) 96, 497.
Venugopalan, S., Petrou, A., Galazka, R. R., Ramdas, A. K., and Rodriguez, S. (1982). Phys.
Rev. B25, 2681.
Venugopalan, S., Kolodziejski, L. A., Gunshor, R. L., and Ramdas, A. K. (1984). Appl. Phys.
Lett. 45, 974.
Walker, T. W., Litton, C. W., Reynolds, D. C., Collins, T. C., Wallace, W. A., Gorrell, J .
H., and Jungling, K. C. (1972). Proceedings of the XI International Conference on the
Physics of Semiconductors, Warsaw, 1972, p. 376. Elsevier, New York.
Weinstein, B. A. (1976). Proceedings of the XI11 International Conference on the Physics of
Semiconductors, Rome, 1976 (F. A. Fumi, ed.), p. 326. Tipografia Marves, Rome.
Zanio, K. (1978). Semiconductors and Semimetals (R. K. Willardson and A. C. Beer, eds.),
Vol. 13. Academic Press, New York.
CHAPTER 10
I. INTRODUCTION . . . . . . . . . . . . .
POLARONS
. .
11. FREEVERSUSBOUNDMAGNETIC
111. THEBMP MODEL. . . . . . . . . . . .
IV.
V.
VI .
VII.
VIII.
IX.
X.
. . . . . .
. . . . . .
. . . . . .
413
415
417
420
420
423
424
425
42.5
428
430
430
433
436
436
438
440
44s
449
45 1
45 1
45 1
452
1. Introduction
In semiconductors containing magnetic ions, there is generally a sizable
exchange interaction between carrier spins and those of the ions. Exchange
causes novel spin-dependent phenomena in such materials including giant
413
CopYnEht 0 1988 by Academic Press, Inc.
All rights of reproduction in any form reserved.
ISBN 0-12-752125-9
414
P. A. WOLFF
10. THEORY
415
The simplest possible magnetic polaron-the free magnetic polaronconsists of a carrier trapped, via the exchange interaction, in a magnetic
potential well created by locally aligning the spins of magnetic ions. The
stability of such complexes was investigated by Kasuya et al. (1970). Their
analysis shows that the polaron energy is determined by a balance of two
terms of opposite sign: a positive kinetic energy due to carrier localization,
and a negative energy due to the exchange well. Free polarons are stable in
certain regions of the polaron parameter space, unbound in others. The
existence of free magnetic polarons in a given material thus depends, in
detail, on the parameters that characterize it. Calculations suggest that they
are generally not stable in dilute (x < 0.05) semimagnetic alloys.
The energetics of BMP are quite different from those of free magnetic
416
P . A. WOLFF
where (r,s) are the coordinate and spin of the trapped carrier, and ( R j , Sj)
those of a magnetic ion at site R j . The free energy is calculated in the
molecular field approximation, assuming a variational form for the carrier
wave function. For the semimagnetic case, the free energy becomes:
where a is the radius of the carrier wave function, x the susceptibility of the
magnetic ion system, and NO = a,' is the unit cell density of the crystal. This
expression is valid until the spin correlation function, ( s - S j ) , begins to
saturate, i.e., approaches its maximum value, of $ for the Mn2+ ion case.
Equation (2) becomes negative for sufficiently small values of a. If saturation
has not occurred prior to this crossover point, the polaron is bound and will
continue to increase its binding energy by decreasing its radius until saturation occurs. Though approximate-since it relies on mean field theory-this
argument is qualitatively correct.
10.
417
The simplest bound magnetic polaron is the donor-BMP in a semimagnetic semiconductor of moderate (say x c 0.1) magnetic ion concentration.
The single electron in such a complex moves in a nondegenerate conduction
band, and has an extended orbit that is well-described by the effective mass
approximation. Moreover, for small x, the magnetic ion susceptibility in such
crystals as Cdl-,MnxS and Cdl-,Mn,Se is nearly Curie-like in the temperature range of interest. Both of these facts simplify the analysis. Thus,
the donor case will be used to develop the theory; these ideas will later be
extended, in a less rigorous way, to acceptor-BMP. Donor-BMP have been
extensively studied via SFRS in CdMnS (Alov et al., 198 1 , 1983; Douglas et
al., 1984; Nawrocki et al., 1984; Heiman et al., 1983a), CdMnSe (Nawrocki
et al., 1980; Heiman et al., 1983b; Plane1et al., 1984), and ZnMnSe (Douglas
et al., 1984; Heiman et al., 1984), which are n-type as grown; there has also
been one report (Peterson et al. , 1982) of SFRS by donor-bound electrons
in normally p-type CdMnTe. These experiments were an important stimulus
to the theory, by providing the first clear evidence for spin fluctuation effects
in BMP. The donor-BMP in semimagnetics is not, however, sufficiently
strongly coupled to exhibit the full range of BMP behavior; even at 2 K there
is only moderate spin correlation in these complexes. Complete correlation,
i.e., spin saturation, is illustrated by the more complicated, but strongly
interacting, acceptor-BMP case.
The model Hamiltonian of the donor-BMP is:
where m* is the effective mass, EO the dielectric constant, and a the exchange
constant (Gaj, 1980) for the conduction band. In applying Eq. (3) to wurtzite
crystals, such as CdMnS or CdMnSe, we will ignore the small ( = 5 % )
anisotropy of m* and E O . A more serious approximation is the neglect of
direct Mn2+-Mn2+ spin interaction in such materials. Experimentally,
Cdl -,Mn,X crystals with small x-values have nearly Curie-like magnetic
susceptibilities (Kreitman et al., 1966; Nagata et al., 1980; Galazka et al.,
1980) in the temperature range 4-40 K, but with magnitudes corresponding
to effective Mn2+concentrations (2)somewhat smaller than x . For example,
for x = 0.05 one finds R = 0.03. This behavior, which has caused some
confusion in the literature, has recently been shown (Shapira et al., 1984;
Aggarwal et al., 1984) to be the result of antiferromagnetically-coupled Mn2+
spin clusters (singlets, doublets, triplets, etc.). Isolated Mn2+spins, with no
nearest neighbor Mn2+ ions, give a full Curie contribution to the susceptibility; MnZ+pairs (doublets) are antiferromagnetically coupled and make no
418
P . A. WOLFF
_1 - S(S + 1) R((oIN0)2(rn*)2
A-
6n
A4No
(5)
10. THEORY
419
and
where
and
where = l/kB T .
The resulting Schrodinger equation for +(r)has the same form as Eq. (6).
However, (s * S j ) is now defined as the ensemble average:
420
P. A. WOLFF
Eqs. (6) and (12) determine 4(r) self consistently. Their solution is usually
calculated-or estimated-variationally by assumin a simple form for W ) ,
such as the hydrogenic wave function +(r) = e-r'u/ na ,calculating G from
Eq. (1 l), and minimizing it with respect to a.
$-J
1. ZEROMAGNETIC
FIELDCASE
= -a
C [ ( ~ * S j ) l 4 ( R j )=I ~- C]
j
[Kj(s.Sj)]E
-(S*r),
(14)
The expression for the coupling constant, Kj, involves the exchange interaction and the square of the BMP wave function. Both factors are smaller in
donor-BMP than in acceptor-BMP; when combined they imply that polaron
interactions in donors are an order of magnitude weaker than in acceptors.
The quantity l- is proportional to the effective field experienced by the
electron in the BMP. It determines the Zeeman splitting of the complex. r
is a vector operator whose components do not commute with one another;
that fact prevents an exact evaluation of the trace in Eq. (1 1). On the other
hand, if r were a classical variable, the trace over spin variables could be
calculated via the standard identity:
10.
421
where
55/2(x)
= i [ e5/2x
e 3 / 2 ~ + e l / 2 ~+ e-1/2X
e-3/2~
+ e-52x].
(19)
Finally, the A-integral is calculated by moving the contour back to the real
422
P. A. WOLFF
It is interesting to note that the first factor in this expression is the partition
function for an Ising model of the BMP-with Hamiltonian
XIsing
C[~j(szsjz)l.
j
where
wd'=
(;:)
- C(K,?).
j
Equation (23) is valid well beyond the fluctuation range ( y WO),but fails
before y reaches its maximum value, $X(CYNO).
Substitution of Eq. (23) into
Eq. (18) yields the following formula for the partition function in the
Gaussian approximation:
(25)
Equation (25) is similar to that of DS, and will later be used to make a
connection between the DS and HWW theories.
10.
423
2 . FINITEFIELDCASE
Eqs. (18)-(25) can easily be extended to the finite magnetic field case.
When fields are present, the Hamiltonian takes the form:
Xspin
=-
(26)
(27)
where
b*
pBg*Bo
and
bMn
PBgMnBo.
(28)
where
The effective field (beff)experienced by the carrier is the sum of the external
field (b*)and the internal, exchange field defined by:
424
P. A. WOLFF
At low temperatures, the exchange field is often much larger than the applied
field. In CdMnSe, for example, aNo = 0.26 eV; if x = 0.05 and T = 4 K, the
dimensionless factor in brackets in Eq. (31) then has the value 100. Field
enhancement factors of that size are not unusual in semimagnetics. Such
materials act as amplifiers of external magnetic fields, at least insofar as their
effect on carrier spin degrees of freedom is concerned. It should be clearly
understood, however, that the exchange interaction does not affect carrier
orbital motion.
3. SOLUBLE
POLARON
MODEL
Several authors (Planel, 1982; Golnik et al., 1983; Ryabchenko and
Semenov, 1983) have discussed an exactly soluble model of BMP. Though
unrealistic, it provides a valuable test of the classical approximation. The
model assumes that a donor-bound electron interacts with N Mn2+ ions
through an exchange interaction of constant strength:
N
x = -K C
j= 1
(s.Sj)
K
2
(32)
where S = Cj,,(S,). This Hamiltonian would follow from Eq. (13) if NMn2+
ions were distributed on a spherical shell surrounding the impurity.
The operators J = ( S + s), J z , and S 2 commute with the model
Hamiltonian. Thus, it has eigenvalues
E(J, J z , S ) = - - J(J
K2 [
+ 1) - S(S + 1) - -431 ,
(33)
that are independent of J z . For a given S , the allowed values of the total
angular momentum are J = ( S f i).The corresponding energies are:
KS
E ( + , J , , S ) = --,
2
In these two groups of levels the electron spin is aligned parallel and antiparallel to the internal field, respectively. SFRS causes transitions between
them, with S conserved.
The partition function for the model is:
=2
(35)
10.
425
where D ( S ) is the number of Mn2+spin states, of fixed S,, with total spin
S . Yanase and Kasuya (1968) describe a technique for calculatingD (S);when
1 Q S Q 5/2N, it can be shown that:
3/2
~ ( 2 s l)e-PKS/2(S+l) 6S*/3SN
1e1.
+
(37)
This summation over S is similar in form to the integral over y in Eq. (25).
We now compare the classical partition function for the model to the exact
result [Eq. (35)]. If S is assumed to be a classical variable, the trace over the
carrier spin in the partition function can be evaluated with Eq. (17); one finds
Z = Tr[eK(S'S)]
I
:
b(S)(ZS + 1)cosh b K Y ] j .
The function of S under the summation sign in this formula differs from that
appearing in Eq. (35) by terms of order S - ' . Since the main contribution to
the sums comes from terms with S N"' , or larger, we conclude that the
classical approximation to the partition function is accurate to order N-"2.
426
P. A. WOLFF
u = - pWo2
-[
4
1.
12 + (pwo)2
4 + (pwo)2
a-
m.m
--.)
(r2)is calculated from Eq. (25) with the aid of the classical approximation
r. One finds
for
10.
427
4
=:[1+
(4+/32WgZ) - 4+pZwgZ
4pW0
>I
(43)
and
On the other hand, in the BMP regime the energy is quite sharply defined
since
428
P . A. WOLFF
The second factor in Eq. (44) is the BMP contribution to the magnetization.
At low temperatures, the BMP moment is much larger than that of the
carrier (geffP g*). The extra moment is provided by Mn2+ spins in the
vicinity of the BMP, whose alignment with the magnetic field is enhanced by
the exchange interaction. In the low field limit, the BMP has susceptibility:
This expression can be shown to be identical to Eq. (4.8) of DS, via the
replacements discussed after Eq. (40). Note that XBMP is a smooth function
10.
429
This result follows directly from mean field theory (MFT), which assumes
that each type of spin aligns in the external field plus the average exchange
field created by the other type. The MFT equations for the BMP are:
and
<Sjz>
SP[~M~
+ Kj(sz)I,
(49)
(Sj,) gives
the MFT
When ~ W6O1, the second term in the argument of the hyperbolic tangent
is small compared to the first. If it is neglected, Eqs. (48) and (49) are
equivalent to Eq. (47).
The approximations s (Sj) = (s Sj) = (s) Sj ,used in deriving Eqs. (48)
and (49), eliminate the spin-spin correlations responsible for BMP formation. Thus, Eq. (50) cannot give correct results for ~ W O
F )l . In fact, it
predicts a divergence in K B M f~ o r b WO= 2 and finite (sz) in zero applied field
if ~ W >O2. Neither effect is physical. To correctly describe the BMP
susceptibility in the BMP regime ~ Ws Ol), we must return to Eq. (44).
When ~ W sO1, Eq. (44) implies:
where
S(q) =
I:
ctnh(q) - -
430
P. A, WOLFF
6. SPINCORRELATION
Finally, we consider the spin correlation function, (s * Sj), for the BMP.
Note that (s * Sj) = - aX/aK,. For such an operator-the derivative of the
Hamiltonian with respect to a c-number parameter-one can use the cyclic
property of the trace to prove that:
5bz(X)/55/2(X).
(54)
This formula is based on Eq. (22) which includes the effects of spin
saturation. When PKj a 1, the second term, in Eq. (53) is small and
(s S,) + $ 8 5 / 2 -, $ as anticipated. In the fluctation regime, where
PWO Q 1, one finds a small correlation: ( s - Sj) = 105/48(pKj) 4 1. This
statement seems, at first glance, to contradict our earlier assertion that there
can be considerable alignment of the carrier spin (with the fluctuating
internal field) in the PWo e 1 regime. Actually, both statements are true. To
reconcile them, remember that the fluctuating moment of the BMP involves
of the total number of Mn2+ spins within
only a small fraction, 0 (1/N1),
it. Thus, even full alignment of the carrier spin does not represent, on
average, a high degree of spin correlation. For pWo Q 1, one expects
( s - Sj) = l/N/.
VI. The Dietl/Spalek Formalism
10. THEORY
431
432
P. A. WOLFF
!
!
d3A
-( [ 5 5 / 2 ( - iAKc)]ceiXYc)
(27CI3
where
P(y)= (const.)!
d3M(rl)
P(y)d3y.
(57)
d3M(rN)[cosh(PTY)
p[M(r)l4A - A[M(r)llBM(r),
where
and
10.
433
The maximum is relatively sharp at low temperatures (PWO>> I), where BMP
formation occurs. In that limit, @ym) % 1 and the factor 2 on the right hand
side of Eq. (63) become unimportant; the formula is then equivalent to the
mean field theory (MFT) result [Eq. (SO)] in zero magnetic field. At higher
temperatures, on the other hand, the full equation is required to avoid the
spurious phase transition predicted by MFT. In this regime, the distribution
of y-values is broad and dominated by fluctuations.
To summarize, MFT predicts the low temperature spin alignment of BMP,
and the high temperature response to weak fields. It is totally inadequate in
describing fluctuations or the gradual transition from the fluctuationdominated to the collective regime.
VII. The Spin-Spin Correlation Function
434
P. A. WOLFF
Here (with A = 1)
s(t) = ei"rs(0)e-ixt,
(66)
and 2 is given by Eq. (18) or Eq. (27). The unit vector (Y is determined by
the geometry of the experiment and by the laser frequency. For our purposes,
a can be viewed as a constant-involving neither the electron nor the Mn2+
spin variables.
In the semiclassical approximation
d3y6(y -
r - b*)eflz(Sj'bMd
This identity can now be used to evaluate the electron spin trace in Eq. (67).
A tedious, but straightforward, calculation yields the result:
10. THEORY OF
435
Finally, the trace over (Sj) is evaluated after substituting Eq. (69) into Eq.
(67). In the Gaussian approximation, one finds (Heiman et al., 1983)
where
[1:
The integral in Eq. (70) is similar to that determining the partition function
in the Gaussian approximation [Eq. (29)]. This fact substantitates the DS
hypothesis that the integrand of the partition function determines the spin
flip spectrum, with the two terms in the cosh(By/2) factor corresponding to
Stokes and anti-Stokes scattering, respectively. This statement will become
clearer in Part VIII, where Eqs. (69) and (70) are compared with SFRS
experiments. We will also see there that the factor T [ t ; y , p l in Eq. (69)
predicts striking polarization effects in SFRS. Their observation provides a
stringent test of the theory.
The Fourier transform of Eq. (70) determines the spin flip spectrum:
We are primarily interested in the second and third terms of this formula,
that describe Stokes and anti-Stokes SFRS. The first, zero-frequency term
can be shown (Romestain et al., 1975) to cause Faraday rotation.
Equation (71) contains no damping. The electron spin precesses freely,
without spin relaxation, in an effective field (beff)that is the sum of the
external field and an internal, exchange field of the MnZ+ions [Eqs. (30)
and (31)]. Damping is eliminated by the classical approximation which
ignores the (s'S7) terms in the exchange interaction. In practice, Eq. (71)
reproduces the measured SFRS spectra almost perfectly. The theory
succeeds because the Mn2+spins relax slowly at low temperatures; thus, the
electron spin has ample time to align with the slowly fluctuating effective
field.
436
P. A. WOLFF
10.
431
438
P . A. WOLFF
8. OPTICAL
EVIDENCE
FOR BMP
IN
SEMIMAGNETICS
Optical experiments have given the most convincing evidence for the
existence of BMP in semimagnetic semiconductors. Two techniques are used:
spin flip Raman scattering and luminescence. When observable, SFRS gives
detailed information concerning the BMP whose interpretation is relatively
straightforward since the center under study involves a single carrier plus
associated Mn2+ spin cloud. Luminescence lines that exhibit BMP effects
are, in contrast, the result of recombination by complexes, such as excitons
bound to neutral impurities (a three body problem) or donor-acceptor pairs
(whose energy of emission depends on the separation of donor and acceptor).
Neither spectrum is easy to interpret. Unfortunately, to date SFRS has only
been seen in donor-BMP where polaron effects are weak. The observation
of SFRS by more strongly coupled acceptor-BMP may be precluded, in cubic
crystals, by valence band degeneracy. Here, nature conspires against us since
the only wide gap semimagnetic that naturally occurs p-type is cubic
p-CdMnTe; (CdMn)S and (CdMn)Se, with simpler, wurtzite valence band
structures, are invariably n-type.
The luminescence spectra of lightly-alloyed Cdl-,Mn,S, Cdl-,Mn,Se,
and Cdl-,Mn,Te crystals with x I0.05 contain features that are similar,
though less sharp, to those observed in the parent binary compoundsincluding free exciton lines, bound exciton lines, and donor-acceptor pair
(DAP) lines. The bound exciton lines are generally those of excitons bound
to neutral acceptors (AOX) or neutral donors (D'X). These identifications
are made by comparing alloy luminescence spectra to that of the binary. In
Cdl-,Mn,Te, for example, the spectra contain a prominent line that continuously evolves (Plane1 et al., 1980) into the known bound exciton (A'X)
line of pure CdTe as x -,0. Thus, in the dilute alloys, at least, one can be
sure that this feature is caused by A'Xrecombination. Its position determines
the binding energy of the exciton to the neutral acceptor. Measurements of
this energy were performed by Golnik et al. (1980). They found, in contrast
to the CdTe case, that in the semimagnetic alloys, the A'X binding energy
is a rapidly varying function of temperature; at low temperatures, it exceeds
that in CdTe by a sizable factor. The data are shown in Fig. 1. Large
temperature variations of impurity energies are not ordinarily observed in
semiconductors. Golnik et al. attribute them to BMP formation around the
bound exciton complex, and give a qualitative interpretation of the results.
This important experiment was one of the first to give convincing evidence
for BMP in semimagnetics. Unfortunately, the complexity of the three body
problem has so far prevented the development of a quantitative theory of the
AOX-BMP.
There have also been studies (Huber et al., 1983) of D'X luminescence in
10.
C31-, Mn,Te
0 I
439
=0.05
z0.2
zo.3
aa
a
4
30
10
I0
30
T(K)
50
FIG. 1 . Variation of AoX binding energy with temperature in (Cd,Mn)Te alloys. Solid line:
theory of GGNPB; dashed line gives AoX binding in pure CdTe. [After Golnik et al. (1980).]
(CdMn)Se. Here, again, one faces a three body problem in interpreting the
data. Experimentally, (Cdl-,Mn,)Se samples with x = 0.05 have a DoX
binding energy larger than that of CdSe, and nearly temperature independent. The increased binding is attributed to carrier spin alignment in the
exchange field caused by Mn2+ spin fluctuations. On the other hand, in
crystals with x = 0.10 the DoX spectrum abruptly shifts and broadens as
temperature is lowered from T = 10 K to 2 K, an unexpected result ascribed
to polaron formation.
The complexity of the AoX and DoX centers has forced investigators to
seek luminescence techniques for studying isolated donor- or acceptorBMP. That goal can be achieved via DAP luminescence, though at a
considerable experimental price. In DAP emission the energy of radiation
from a given donor-acceptor pair varies, because of the residual Coulomb
interaction between their charges, with the distance R, between them. The
complete DAP line is the superposition of radiation from all pairs. Usually
it is too broad to give information concerning BMP formation. This
convolution can be unfolded, experimentally, by studying DAP spectra as
a function of delay time after pulsed excitation. The effective distance at
440
P. A. WOLFF
which pairs recombine is related to the delay time (t) by the approximate
expression:
e2
R ( t ) = -ln(t/to),
EO ED
(72)
where ED is the donor binding energy and to = 10--lO-o sec. Nhung and
Plane1 (1983) have performed a remarkable series of measurements of this
type in CdMnTe, inferred acceptor-BMP binding energies from them, and
interpreted the results via the DS theory. We will discuss their work in some
detail in Part VIII, Section 4.
SFRS is the other important optical method for studying BMP. In the
(CdMn) chalcogenides, it was first studied by Nawrocki et al. (1980). Their
observation of a finite, zero-field spin splitting was unambiguous evidence
for BMP effects. It motivated recent theoretical work on BMP and
stimulated further experiments. We will see below that the SFRS data clearly
demonstrate the role of fluctuations in BMP.
9.2 ANALYSIS
OF SPINFLIP RAMAN
SCATTERING
EXPERIMENTS
10.
441
These formulas interpolate smoothly into one another, and the latter can be
shown to give a good approximation to the spectrum over the whole field
range below saturation (0 < b~~ < k T ) . Equation (74) is essentially the
formula used by DS [their Eq. (3.25)] in analyzing.spin flip spectra.
FIG.2. Variation of spin-flip energy with magnetic field in (Cd,Mn)Se. [After Heiman et al.
(1 983 b).]
442
P. A. WOLFF
I
>
T=l 9 K
3 4
69
I28
x
A
I
4
+
o
8
18 0
283
I0
B (kG)
FIG. 3. Low field spin-flip energies in (Cd,Mn)Se. [After Heiman et ul. (1983b).]
The theoretical curves in Figs. 2 and 3 were calculated from Eq. (74). They
illustrate the magnetic field variation of the Stokes peak of that spectrum.
Figure 4 compares the complete spectrum at zero field [Eq. (73)] with
experiment. The agreement is excellent in both cases. DS achieve comparable
fits to the data of Nawrocki et al. (1980). The fits determine two parameters:
and the temperature TAFused to correct the Mn2+ susceptibility for next
nearest neighbor interactions (Part V, Section 2). For Cdo.~oMno.loSe,the
parameters are WO= 0.66 f 0.07 meV and T A =~ 2.3 K. The value of WO
Calculated from Eq. (75) is WO= 0.56 f 0.06meV and is in reasonable
agreement with experiment. At the lowest temperature ( T = 2 K) achieved
in these measurements, ~ W =O4. The system is then just beginning to enter
the polaron regime. At higher temperatures fluctuations control its behavior.
In particular, the zero field splitting is temperature independent as implied
by the discussion after Eq. (42). Temperature independent energy shifts are
characteristic of the fluctuation regime.
10. THEORY OF
443
C d , _ , Mnx S e
'
x = o 10
STOKES SHIFT ( m e V )
FIG. 4. Zero field SFRS spectrum of (Cd,Mn)Se. Points are experimental; solid line the
theory. [After Heiman el al. (1983b).]
"
x4
)at12
11
ly*a*(2
i y - ( a * x a**)
Y2
--
where 6' is the angle between the vector y and the &axis. The polarization
444
P . A. WOLFF
ratio is:
Here the average is over the angular weighting function in Eq. (76). For
example:
ex@pdp=
sinh(x)
2( 4
X
- (1
+ cos2#) =
ex@(l+ p 2 )d p = X
Hence
Alov et al. (1983) have derived a formula of precisely this form, but with
argument x = /3MeffBoin a more intuitive way, and shown that it fits the field
dependence of their polarization data. The BMP moments (Meff)
they infer
vary, with SFRS frequency shift, from 2 5 , to
~ ~5 6 ~ This
~ . frequency
variation agrees with that predicted by Eqs. (78)-(80), and thus fully accounts
for the polarization data. The experiment also implies that the SFRS line is
inhomogeneously broadened with different frequency shifts corresponding
to BMP with different effective moments and internal fields.
It is illuminating to briefly review the arguments of Alov et al. (1983) and
Plane1 et al. (1984) leading to Eq. (80). Both groups emphasize that the
donor-bound electron experiences an effective field that, because of Mn2+
spin fluctuations, need not be parallel to the applied field. As a consequence,
the electron spin wave functions take the form:
XT = cos(8/2))t) - sin(B/2)Jl),
XJ = sin(8/2)lt)
+ cos(8/2)11),
(81)
where the spinors It), 1.1) are referred to the c-axis. These states are then
used, in conjunction with B-valence band edge wave functions
10.
445
to calculate spin flip matrix elements. One can easily show that the US,+1/2
channels generate negatively or positively polarized SFRS, with amplitudes
cos2(O/2) or sin2(O/2). Equation (80) follows if it is assumed that the angular
probability distribution has the classical form, e-BMeffBo.
These polarization experiments clearly indicate that BMP have large
( 2 5 - 5 0 ~ ~moments
)
associated with them, whose alignment by a field
follows the classical law-but with a different effective moment at each point
in the SFRS spectrum. Thus, in the fluctuation regime, where the spectrum
is relatively broad [Eq. (43)], it is not correct to speak of the BMP moment
as if it were a rigid, time invariant entity. Whenp WO4 1, each BMP can have
a wide range of moments and gradually fluctuates, in time, over many values.
The situation is different in the collective regime. When ~ W%-O1, the
distribution of effective fields is sharply peaked about the value
The BMP then has a well-defined moment, whose average alignment is truly
described by the classical Langevin formula [Eq. (51)].
a.
10. ANALYSIS
OF ACCEPTOR-BMP
EXPERIMENTS
446
P. A. WOLFF
Since then two groups (Wolff and Warnock, 1984 and Nhung et al., 1984)
have developed formalisms which do not require assumptions (l), (3), and
I0 010
A
400
10
20
T(K)
30
5 '10
2 010
0 '10
40
FIG.5. Acceptor-BMP energies vs. temperature in (Cd,Mn)Te alloys. Dashed line: theory of
Nhung and Planel; solid line: theory of Warnock and Wolff. [After Nhung and Planel (1983).]
10. THEORY OF
441
(4). The analysis parallels that of Part IV. For the valence band one
postulates, in the single band hydrogenic approximation, a spin Hamiltonian
of the form:
x = - J Cj [(s
Ci [ K j ( s
~ j ) ~ $ ( ~ j z=
) l~ ]
~ j ) ] .
(83)
where the hole spin s = and $(r) = e - ' / " / G is the acceptor wave
function. The radius a is determined variationally. The spin partition
function, in the classical approximation, is then:
2 = Tr
{s.sjl
-1
- Tr
S[y -
S[y -
C (KjSj)]eacs")d3y
C (KjSj)][&20Y
+ e1/20Y + e-'/2@Y+ e - 3 / 2 8 Y ] d 3 y
{SJI
= ( 6 ) N j d '(2n)3
dd'yeiX.ynIss/z(-UKj)][e3/20y
11.
(84)
This expression is the analogue, for a spin$ carrier, of Eq. (18). The integrals
can again be evaluated as outlined in Part IV, Section 1. The result is:
a formula quite similar to Eq. (22). Wolff and Warnock (1984) calculated
BMP energies from this expression by making the continuum approximation, though in principle the discrete sum could have been evaluated. Their
results are indicated in Fig. 5 along with earlier calculations of NP. The only
adjustable parameter (a = 13 A ) in this fit was determined from the acceptor
binding energy measured by N P in CdTe. Other quantities @NO,X) are
known from independent measurements. Both of the calculations illustrated
in Fig. 5 imply that Mn2+ spins within the acceptor orbit are fully aligned,
with the hole spin, at low temperatures. The net moment of the BMP is then
. radius of the acceptor wave function could, in principle,
about 20 p ~ The
vary with temperature via the changing exchange interaction. However, this
effect is found to be small.
The very recent calculations of Nhung et al. (1985) are compared with their
experiments in Fig. 6. These results were obtained by numerically integrating
an expression for the internal energy derived from Eq. (85):
448
P. A. WOLFF
u=--a(ln z)
aP
= -
S ~ i ~ y y ~ ( y ) [ + e +~ /+e/2r
~ y - )e-1/28r
- )e-3/28r]
jd3y~(y)[e3/2P
+~e 1 / 2 6 ~+ e-l/Wr + e - 3 / 2 8 ~ ~
(86)
10
20
30
40
Temperature ( K )
FIG. 6. Comparison of theory and experiment for acceptor-BMP energies. The solid curve
is the exact, numerical calculation; the triangles are experimental points for Cd0.9sMno.osTe.The
dashed curve corresponds to a truncated Gaussian approximation. [After Nhung et al. (1985).]
10.
449
450
P. A. WOLFF
2mh:a2
35 R(JNol2 = o.
48 (8na3No)kT
Here we have used Eq. (40),in the collective limit, to evaluate the exchange
energy where the wave function has been assumed hydrogenic. Equation (87)
for AF remains valid until Mn2+ spin saturation occurs. That condition is
not precisely defined since Mn2+spins at different positions in the BMP orbit
saturate at different temperatures. Roughly speaking, however, we may
consider saturation to occur when PKj > 1 for a Mn2+ ion at the center of
the BMP. To avoid saturation we require:
Note that this criterion, for the stability of free polarons, is quite sensitive
to temperature and x-value.
Electrons have small masses and small exchange constants (JNo= 0.25 eV).
For such carriers, Eq. (89) cannot be satisfied at any temperature above
T = 1 K. Thus, free electrons do not form polarons in the known
semimagnetics.
Holes (or excitons) have larger exchange constants and larger masses; both
factors favor polaron formation. With (JNo) = 1 eV, m* = 1, and
T = 10 K, Eq. (89) is satisfied for R > 0.05. This critical value of R agrees
surprisingly well, considering the crudeness of the approximations, made in
deriving Eq. (89), with the R determined by Golnik et al. from the L1 -,L2
crossover of the luminescence spectra. We conclude, therefore, that there is
considerable circumstantial evidence for free magnetic polarons, and that the
evidence is not inconsistent with the criterion of Kasuya et al. for their
stability.
10.
451
X. Conclusion
11. CURRENT
STATUS
OF BMP THEORY
The theory outlined in Parts IV and V predicts the equilibrium properties
Spalek picture of their behavior, whose key feature is the gradual transition
from a high temperature, fluctuation-dominated regime to a low temperature, collective one. The fluctuation regime is exemplified by donor-BMP
and has been thoroughly studied via SFRS in CdMnS and CdMnSe. On the
other hand, only BMP involving holes have exchange interactions strong
enough to produce the complete range of possible BMP effects, culminating
in spin saturation at low temperatures. Unfortunately, the centers containing
holes that are easy to study experimentally (AOX,D o X ) involve three
carriers-two holes and one electron or vice versa. To date, the complexity
of the resulting theoretical problem has prevented quantitative comparisons
of theory with experiment for such complexes. A theoretical advance in this
direction could provide several interesting, new tests of the BMP model.
An exception to these statements is the elegant, but complicated, time
resolved DAP luminescence technique of Nhung and Plane1 (1983). Their
measurements determine binding energy versus temperature for the simple
acceptor-BMP in Cdl -xMnxTe,and clearly demonstrate BMP formation at
low temperatures. Though theory, including the effects of spin saturation,
gives energies in surprisingly good agreement with their measurements, there
are still some uncertainties in the analysis.
The DS and HWW theories of BMP use a classical approximation to
calculate Mn2+spin statistics. Carrier spins are assumed to follow the slowly
varying internal field of the Mn2+ions. This approach accurately predicts the
equilibrium properties of BMP, but is inadequate for treatment of Mn2+or
carrier spin relaxation processes.
12. DIRECTIONS
FOR FUTURE
WORK
452
P. A. WOLFF
excitation, and propose that this delay is the BMP formation time. Optical
pumping experiments (Warnock et af., 1984) show that free carrier spin
relaxation rates are much faster than those of polarons in CdMnSe. These
preliminary, but tantalizing, results suggest a thorough study of BMP spin
dynamics; the problem seems naturally suited to recently-developed picosecond optical spectroscopy techniques. On the theoretical side, a fully
quantum mechanical theory of BMP is required. The work of Yanase and
Kasuya (1968) suggests that such a theory may not be inordinately difficult,
since they have rigorously solved a quantum mechanical BMP problem with
spatially varying exchange constants. Hopefully, a complete quantum
mechanical theory will clarify the roles of spin relaxation and spin diffusion
in the BMP formation process.
The luminescence studies of Golnik et af. (1983) demonstrate a dramatic
change in the spectrum of Cdl-,MnxTe as x increases beyond x = 0.07. The
properties of the new L2 feature they observed are consistent with those
expected from recombination of a magnetically self-trapped exciton. That
interpretation is supported by optical pumping experiments (Warnock et al.,
1984). However, further measurements to confirm this hypothesis , and
better theoretical estimates of the free polaron energy, are needed. In
particular, for crystals with larger x-values, competition between the strong
carrier-Mn2+ exchange interaction, which tends to align Mn2+spins parallel
to that of the carrier, and the direct Mn2+-Mn2+interaction which favors
antiferromagnetic or spin-glass alignment, will play an important role in
determining the polaron binding energy.
Cdl-xMnxTecrystals with x 5; 0.2 have a spin-glass phase at low temperatures. Golnik et al. (1983) see substantial structure, in their measurements of
the energy of the L2 feature versus temperature, at temperatures near the
spin-glass transition in samples withx = 0.2,0.3, and 0.4.These remarkable
observations imply that a localized exciton, whose center-of-mass wave
function may be no larger than 1OA in radius, can sense the Mn2+spin-glass
phase. It would be interesting to exploit this opportunity, by using the strong
(exchange fields approaching 100T), localized magnetic probe that nature
here provides us, as a tool to study the spin-glass phase in semimagnetics.
References
Aggarwal, R . L., Jasperson, S., Shapira, Y . , Foner, S., Sakikabara, S., Goto, T., Miura, N.,
Dwight, K., and Wold, A. (1985). Proc. XVII Intl. Conf. Phys. of Semiconductors, Sun
Francisco, I984 (J. D. Chadi and W. A. Harrison, eds.), p. 1419. Springer.
Ahn, K., and Shafer, M. (1970). J. Appl. Phys. 41, 1260.
Alov, D., Gubarev, S., Timofeev, V., and Shepel, B. (1981). JETP Letfers 34, 71. (Pisma v
Zh. Eksp. Teor. Fiz. 34, 76.)
10.
453
Alov, D., Gubarev, S., and Timofeev, V. (1983). Sov. Phys. JETP57, 1052. (Zh. Eksp. Teor.
Fiz. 84, 1806.)
Busch, G., Junod, P., and Wachter, P. (1964). Phys. Letters 12, 11.
Busch, G., and Wachter, P. (1966a). Phys. Condensed Matter 5, 232.
Busch, G., Magyar, B.. and Wachter, P. (1966b). Phys. Letters 23, 438.
Dietl, T., and Spalek, J. (1982). Phys. Rev. Letters 48, 355.
Dietl, T., and Spalek, J. (1983). Phys. Rev. B28, 1548.
Douglas, K., Nakashirna, S., and Scott, J. (1984). Phys. Rev. B29, 5602.
Gaj, J. (1980). Proc. X V Intl. Conf. Phys. of Semiconductors, Kyoto; J. Phys. SOC. Japan,
Suppl. A49, 797.
Galazka, R., Nagata, S., and Keesom, P. (1980). Phys. Rev. B22, 3344.
Golnik, A., Gaj, J., Nawrocki, M., Planel, R., and Benoit a la Guillaume, C. (1980). Proc.
X V Intl. Conf. Phys. of Semiconductors, Kyoto; J. Phys. SOC. Japan, Suppl. A49, 819.
Golnik, A,, Ginter, J., and Gaj, J. (1983). J. Phys. C16, 6073.
Gubarev, S. (1981). Sov. Phys. JETP53, 601. (Zh. Eksp. Teor. Fiz. 80, 1174.)
Harbeke, G., and Pinch, H. (1966). Phys. Rev. Letters 17, 1090.
Harris, J., and Nurrnikko, A. (1983) Phys. Rev. Letters 51, 1472.
Heikes, R., and Chen, C. (1964). Physics 1, 159.
Heiman, D., Shapira, Y., and Foner, S. (1983a). Solid State Commun. 45, 899.
Heiman, D., Wolff, P., and Warnock, J. (1983b) Phys. Rev. B27, 4848.
Heiman, D., Shapira, Y., and Foner, S. (1984). Solid State Commun. 51, 603.
Holtzberg, F., McGuire, T., Methfessel, S., and Suits, J. (1964). Phys. Rev. Letters 13, 18.
Holtzberg, F., McGuire, T., Methfessel, S., and Suits, J. (1965). Proc. Intl. ConJ Magnetism,
Nottingham (The Institute of Physics and the Physical Society, London, 1965).
Huber, C., Nurmikko, A., Gal, M., and Wold, A. (1983). Solid State Commun. 46, 41.
Kasuya, T., and Yanase, A. (1968). Rev. Mod. Phys. 40, 684.
Kasuya, T., Yanase, A., and Takeda, T. (1970). Solid State Commun. 8, 1543.
Kreitman, M., Milford, F., Kenan, R., and Daunt. J. (1966) Phys. Rev. 144, 367.
Kiibler, J., and Vigren, D. (1975). Phys. Rev. B11, 4440.
Kuivalainen, R., Sinkkonen, J., Kaski, K., and Stubb, T . (1979). Phys. Stat. Solidi(b)94, 181.
Landau, L., and Lifshitz, E. (1977). Statistical Physics (Pergamon Press, New York),
Chap. 12.
Leroux-Hugon, P. (1972). Phys. Rev. Letters 29, 939.
Leroux-Hugon, P . (1976). J. Magn. Magn. Muter. 3, 165.
Mauger, A. (1983). Phys. Rev. B27, 2308.
Nagaev, E., and Origin, A. (1974). Phys. Stot. Solidi (b)65, 457.
Nagaev. E . (1983). Physics of Magnetic Semiconductors (MIR Publishers, Moscow).
Nagata, S., Galazka, R., Mullin, D., Akbarzadeh, H., Khattak, G., Furdyna, J., and Keesom,
P. (1980). Phys. Rev. B22, 3331.
Nawrocki, M., Planel, R., Fishman, G., and Galazka, R. (1980). Proc. X V I n t l . Conf. Phys.
of Semiconductors, Kyoto; J. Phys. SOC.Japan Suppl. A49, 823.
Nawrocki, M., Planel, R., Molot, F., and Kozielski, M. (1984). Phys. Stat. Solidi (b)123, 99.
Nhung, T., and Planel, R. (1983). Proc. XVI Intl. Conf. of Semiconductors, Montpellier;
Physica 117B-l18B, 488.
Nhung, T., Planel, R., Benoit A la Guillaume, C., and Battacharjee, A. (1985). Phys. Rev. B31,
2388.
Oliver, M., Kafalas, J., Dirnmock, J., and Reed, T. (1970). Phys. Rev. Letters 24, 1064.
Oliver, M., Dimmock, J., McWhorter, A., and Reed, T. (1972). Phys. Rev. B5, 1078.
Peterson, D., Petrou, A., Datta, M., Ramdas, A., and Rodriguez, S. (1983). Solid State
Commun. 43, 667.
454
P. A. WOLFF
Planel, R., Gaj, J., and Benoit a la Guillaume, C. (1980). Proc. ist Intl. Meeting on Magnetic
Semiconductors, Montpellier; J. Phys. CON.41, C5-39.
Planel, R. (1982). Proc. XVI Intl. Conf. on Appl. of High Magnetic Fields in Semiconductor
Physics, Grenoble; Lecture Notes in Physics 111, p. 309. Springer, Berlin.
Planel, R., Nhung, T., Fishman, G., and Nawrocki, M. (1984). J. de Physique 45, 1071.
Romestain, R., Geschwind, S., and Devlin, G. (1975). Phys. Rev. Letters 35, 803.
Ryabchenko, M., and Semenov, Yu. (1983). Sov.Phys. JETP57, 825. (Zh. Eksp. Teor. Fiz.
84, 1419.)
Schoenes, J., and Wachter, P . (1974a). Phys. Rev. B9, 3097.
Schoenes, J., Wachter, P., and Rhys. F. (1947b). Solid State Commun. 15, 1891.
Shapira, Y., Foner, S., Ridgley, D., Dwight, K., and Wold, A. (1984) Phys. Rev. B30,4021.
Spalek, J. (1980). J. Mag. Magn. Muter. 15-18, 1289.
Torrance, J., Shafer, M., and McGuire, T . (1972). Phys. Rev. Letters 29, 1168.
Tu, K., Ahn, K., and Suits, J. (1972). IEEE Trans. on Magnetic 8 , 651.
von Molnar, S., and Methfessel, S. (1967). J. Appl. Phys. 38, 959.
Warnock, J., Heiman, D., Wolff, P., Kershaw, R., Ridgley, D., Dwight, K., Wold, A., and
Galazka, R. (1985). Proc. XVII Intl. Conf. Phys. of Semiconductors, Sun Francisco ( J .
D. Chadi and W. A. Harrison, eds.), p. 1407. Springer.
Wolff, P., Ramos, J., and Yuen, S. (1977). Theory of Light Scattering in Condensed Matter,
edited by Bernard Bendow, Joseph L. Birman, and Vladimir M. Agranovich (Plenum, New
York).
Wolff, P., and Warnock, J. (1984). J. Appl. Phys. 55, 2300.
Yanase, A. (1972). Internat. J. Magnet. 2, 99.
Yanase, A., and Kasuya, T. (1968). J. Phys. SOC.Japan 25, 1025.
Index
A
455
11
456
INDEX
457
INDEX
dependence on composition
Cdl-,Mn,Te, 167
Zn,-,Mn,Te, 150-155
temperature dependence
Cdl-,MnrTe, 170
Znl -xMnr.Te, 150- 155
Covalent radii of elements, 6, 11
Crystal structure, A:L,MnxBV1 alloys, 3-4,
7-14, 79
D
Debye temperature, 107
Diamagnetic susceptibility, A"BV'
compounds, 91, 98
Dingle temperature, 209, see ufso
Shubnikov-de Haas effect
Donor-acceptor pair luminescence, 439-440
E
Electron paramagnetic resonance, in DMS
alloys, 80-90, see also Raman
scattering, magnetic excitations
in Cdl-,Mn,Se, 84-85, 89
in Cdl-,Mn,Te, 82-86, 88-90
g-factor, of Mn++ ions, 82
internal field, 82, 89
linewidth, effect of d-d exchange
interaction, 82-90
in quaternary DMS alloys, 90
Energy gap
bowing effects, 38-39
Cdl-,Mn,S, 47
sp-d exchange contribution, 39, 47-49
Znl-xMn,Se, 47, 49
Znl-,MnxTe, 48
concentration dependence, 41, 42
Cdl-,Mn,S, 46, 47
Cdl-xMnxSe, 44-45
Cdl-,Mn,Te, 40, 44
Hgl-,Mn,Se, 268
Hgl-,Mn,Te, 192, 267
Znl-xMnxSe, 47-49
Znl-,Mn,Te, 48
MnB"' compounds, 43
pressure dependence, 51-53
Cdl-,Mn,Te, 52
Znt-,MnxSe, 56, 68-69
Znl-xMn,Te, 55
temperature dependence, 50-51
Cdl-,Mn,S, 55
Cdt-,Mn,Se, 55
Cdl-xMn,Te, 52
Znl -xMnxSe, 56
Znl-,Mn,Te, 55
Epitaxy, A:'_,Mn,Bv' alloys, 28
EXAFS, in ternary DMS alloys, 11-13
Exchange integral, nearest-neighbor d-d
interaction
Cdl-,Mn,S, 100
Cdl-,MnxSe, 100
Cdl-,Mn,Te, 100, 168, 177-178
Hgl-,Mn,S, 100
Hgl-,MnxSe, 100
Hgl-,Mn,Te, 100
Znl.,Mn,S, 100
Znl-,Mn,Se, 100
Znl_,Mn,Te, 100, 154
Exchange integral, sp-d interaction,
188-198, 296, 375
(Cdl-,Mnx)3As2, 215
Cdl-,Mn,S, 295
Cdl-,Mn,Se, 295
Cdl-,Mn,Te, 294-295
definition, 188, 198, 235, 291, 296, 314
Hgl-,Mn,Se, 213-214, 270
Hgl-,Mn,Te, 213, 270
Pb1-,MnxBV' alloys, 217
physical interpretation, 188-189
Znl-,MnxSe, 295
Znl-,Mn,Te, 295
Exchange interaction, d-d, see also
Antiferromagnetic order; Exchange
integral, nearest-neighbor d-d
interaction; Raman scattering,
magnetic excitations; Spin-glass
transition
Bloembergen-Rowland mechanism, 80,
105, 107, 271
458
INDEX
Hamiltonian, 86
for specific cluster forms, 78
interaction beyond nearest neighbor, 77,
103, 118
superexchange, 13, 80, 84
Exchange interaction, ion-carrier, see
Exchange interaction, sp-d
Exchange interaction, sp-d, see also Spin
splitting; Bound magnetic polaron
Hamiltonian, 185-186, 197, 234-235,
288-289, 313-314, 375, 378, 416-417,
420
basis vectors, 193, 196, 234, 290, 313
diagonalization, 194-198, 235-236,
290-291, 314
eigenvalues, 197-198, 235-236,
290-293, 314-315
Pidgeon-Brown model, modification of,
197-199, 233-237
valence band, effect on, 292-293,
313-318
Exchange splitting, see Exchange
interaction, sp-d; Spin splitting
Excitons, in magnetic field
in Cdl-,MnxS, 285
in Cdl-,Mn,Se, 285
in Cdl-,Mn,Te, 281, 283-284, 294, 300
complexes, in Cdl-,Mn,Te, 303-305
excited states, 300-303
luminescence, in Cdl-,Mn,Te, 303-305
magnetic polaron, effect of, 297-299
selection rules, magnetooptical
transitions, 283, 294-295
from spin-orbit split-off valence band,
287
spin splitting, ground state, 277, 279,
283, 286, 293-296
in Znl-,Mn,Se, 283
in Znl-xMnxTe, 283
Extended x-ray absorption fine structure,
11-13
F
Faraday rotation
Cdl-,MnxTe, 276, 297
experimental technique, 277-279
Hgl-,Mn,Te, 271
G
g-factor, effective, 202, 237, 428
Hgl-,Mn,Se, 253
Hgl-,MnxTe, 246
2-dimensional electron gas, 221, 224
H
Hgl-,-,Cd,Fe,Se,
29-30
Hg~-,-,Cd,Mn,Te
magnetooptical studies, 260-261
spin-glass transition, 27 1
thin film, 28
2-dimensional electron gas, 217, 219, 221
Hgl-,FexSe, 29-30
Hgl-,Fe,Te
energy gap, 254
magnetooptics, 254-257, 270
Van Vleck paramagnetism, 256-257
Hgl-,Mn,S
crystal growth, 27
crystal structure, 3, 7-10, 79
exchange integral, d-d interaction, 100
lattice parameter, 8-10
Hgl-,Mn,Se
band parameters, 253, 263, 267-269
combined resonance, 264
conduction-valence band overlap,
exchange-induced, 21 1-212
crystal growth, 26-27
crystal structure, 3, 7-10, 79
energy gap, 268
exchange integral, d-d interaction, 100
exchange integral, sp-d interaction,
213-214, 270
g-factor, 253
lattice parameter, 8-10
magnetic breakdown, 210-21 1
magnetic phase diagram, 104
momentum matrix element, 269
Shubnikov-de Haas effect, 204-21 1
spin splitting of Landau levels, 205
459
INDEX
Hgl-xMnxTe
acceptors, in magnetic field, 330-333
band parameters, 246, 267-269
band structure, near r-point, 191-192,
233, 267
combined resonance, 258-264
conduction-valence band overlap,
exchange-induced, 199-200, 21 1-212
crystal growth, 25-26
crystal structure, 3, 7-10, 79
cyclotron resonance, 264
effective mass, 246
energy gap, 192, 267
exchange integral, d-d interaction, 100
exchange integral, sp-d interaction, 213,
270
Faraday rotation, 271
g-factor, 246
hopping conduction, 333-338
lattice parameter, 8-10
magnetic phase diagram, 104, 106, 272
magnetic susceptibility, 92
momentum matrix element, 268
negative magnetoresistance, 333-338
nonmetal-metal transition, 338
resonant acceptor states, 339-342
shallow acceptors, 322
Shubikov-de Haas effect, 204-209
spin-glass transition, 95, 106, 272
spin splitting of Landau levels, 199-206
stimulated spin-flip Raman scattering,
266
2-dimensional electron gas, 217-219
Hopping conduction, in magnetic field
in Hgl-,Mn,Te, p-type, 333-339
magnetoresistance, giant negative,
334-339
anisotropy of, 336-337
I
Impurity band conduction, see Hopping
conduction
K
Kerr effect, magnetooptical, 278
L
Landau levels, see Band structure
Lattice parameters
A"BV' compounds, 4-6
A'l',Mn,BV' alloys, 7-10
Lattice vibrations. see Phonons
M
Magnetic breakdown, 210-21 1
Magnetic excitations, see also Spin
dynamics
magnons, in Cdl-,Mn,Te, 173-177,
390-396
neutron scattering, inelastic, 173-177
Raman scattering, 373-375, 390-396
Magnetic phase diagram, see specific DMS
alloys
Magnetic polaron, see Bound magnetic
polaron; Free magnetic polaron
Magnetic susceptibility, see also Spin-glass
transition
Cdl-,Mn,Se, 92
Cdl-,Mn,Te, 91
Curie-Weiss form, 91, 97
Hgl-,Mn,Te, 92
irreversible effects, 93-94, 96
of specific cluster configurations, 76-77
Magnetization, 198, 235, 314, 400-401, 428
Brillouin function, 235
effective Mn concentration, 101-102,
403, 418
modified, 101
equations of motion, for magnetic
excitations, 373-375
in Hgl-,Fe,Te, 254-255
magnetic polaron correction, 297-299
mean field approximation, 198, 289, 314,
400
remanent, 112-117
spin splitting, relation to, 198, 235, 286
Magnetooptical transitions, see a h
Excitons, in magnetic field
narrow-gap DMS, 237-240
inter-band, 241-250
intra-band, 257-266
selection rules, 239
wide-gap DMS
away from zone center, 305-308
selection rules, 283, 294-295
460
INDEX
Magnetoresistance, negative, in
Hgl-,Mn,Te, 333-337, see also
Hopping conduction, in magnetic
field
Magnons, see also Magnetic excitations;
Raman scattering; Spin dynamics
neutron scattering, CdI-,Mn,Te, 173-177
Raman scattering, Cdl-,Mn,Te, 390-396
Manganese
covalent radius, 6, 11
distribution in A:L,Mn*BV1 lattice, 76-77,
101-102, 111, 147, 157
frustrated antiferromagnetic lattice, 79
percolation, nearest-neighbor, 77-78,
93, 103, 110
probability, specific cluster forms,
76-77
effective concentration, for
magnetization, 101-102, 400, 418
similarity to group-I1 elements, 4
Manganese d-states
charge transfer states, 70
hybridization, p-d, 70
intra-ion transitions, 58-68
pressure shifts
Cdl-xMnxTe, 61
Znl-,Mn,Se, 65, 68, 69
Znl-*Mn,Te, 63
Racah parameters, 58-60
Znl-,Mn,Se, 60
Znl-*Mn,Te, 60
splitting and degeneracy, 57
Tanabe-Sugano diagram, 58
Mean field approximation, 198, 289, 314
magnetic polaron correction, 297-299
Mn++-Mn++interaction, see Exchange
interaction, d-d
MnS
bond length, 11
lattice parameter, tetrahedrally bonded
phases, 8-10
synthesis, 17
MnSe
bond length, 11
lattice parameter, tetrahedrally bonded
phases, 8-10
synthesis, 17
thin films, zinc blende, 28
MnTe
bond length, 11
N
Neutron scattering, magnetic
Cdl-,Mn,Te
diffuse scattering, 155-172
inelastic scattering, 155, 173-177
in short-range-ordered systems, 128-135
theory, for magnetic short range order
powder, 131-135
single crystal, 130, 135-140
Znl-xMn,Te, diffuse scattering, 140-155
Neutron scattering, nuclear
Bragg scattering, 130
Cdl-,MnxTe, 157
Znl-xMnxTe, 141-143
diffuse, 130, 157
Nonmetal-metal transition, exchangeinduced, 337-338, see also Hopping
conduction, in magnetic field
0
Optical absorption, in zero magnetic field
Varshni equation, 50
Mn d-electron transitions, 59-69
Overlap of bands, exchange-induced,
199-201, 2 11, 246-247
P
Pbl-,Mn,BV1 alloys, 30
exchange integrals, sp-d interactions, 217
Shubnikov-de Haas effect, 216
Percolation, see Manganese, distribution in
A{~,Mn,B"' lattice
Phonons, see also Raman scattering
in Cd,-,Mn,Te, 365-373
in Znl-,MnxTe, 366-373
Preparation
of ATTBV'
compounds, 15-17
of Mn chalcogenides, 17
Purification of elements, 14-15
INDEX
Q
Quantum oscillations, see Shubnikov-de
Haas effect
R
Raman scattering
cross section, 347
experimental technique, 350-352
magnetic excitations
magnons, in Cdl-,Mn,Te, 390-396
Raman antiferromagnetic resonance,
393
Raman electron paramagnetic
resonance, 376-389
theory, 373-376
phonons
mixed crystals, 357-365
perfect crystals, 353-357
selection rules, 348-350
spin-flip, 396-405, 414, 417, 440-445
Cdl-,Mn,S, 417
Cdl-,Mn,Se, 403, 405, 414, 417,
440-445
Cdl-,Mn,Te, 397-401, 404, 414
Hgl-,Mn,Te, 266
Znl-,Mn,Se, 417
tensors
zinc-blende structure, 356-357
wurtzite structure, 356-357
theory, 346-350
Rare-earth-based DMS alloys, 30
Remanent magnetization, 93-94, 112-1 17
Resonant states, see Acceptors
461
462
INDEX
T
Tetrahedral bond, 4, 6
Tetrahedral radius, 6, 11
Twinning
Cdl-,MnxTe, 156, 165, 175
zinc-blende crystals, 156
Two-dimensional electron gas, in DMS
alloys, 217-224
V
Valence band, in DMS alloys, see also
Acceptors
effect of sp-d exchange
narrow-gap DMS, 315-316
wide-gap DMS, 314-315
anisotropy, exchange-induced, 292-293,
315, 318-320, 327-329
Van Vleck paramagnetism, in A!'-,FexBV'
alloys, 257