You are on page 1of 16

Supplementary Material

Linking Local Environments and Hyperfine Shifts: A Combined Experimental and


Theoretical 31P and 7Li SolidState NMR Study of Paramagnetic Fe(III) Phosphates

by Jongsik Kim, Derek S. Middlemiss, Natasha A. Chernova, Ben Y. X. Zhu, Christian


Masquelier and Clare P. Grey

S1. Derivation of the Forms of the Fermi Contact and ElectronNuclear Dipolar
Hamiltonians

The following derivation follows the approach adopted by Bransden and Joachain.1 We
start by expressing the total Hamiltonian for an electron in an atom as the sum

H = H0 + HMD (S1)

of the normal unperturbed operator H0 including the electronnuclear Coulomb potential and
electron kinetic energy, and
the perturbation HMD (magnetic dipole) due to the interaction of the
electron and nuclear magnetic moments. The vector potential A(r) due to a point magnetic dipole
N at the origin (taken as coincident with the nuclear position) is obtained as


A( r ) = 0 ( N r ) . (S2)
4r 3

The electronic orbital angular momentum [expected to be small in Fe(III)bearing materials] is


neglected hereafter. The
magnetic induction arising due to the nuclear dipole is

1
B = A = 0 N 2 N
4 r
( ) 1r, (S3)

while the spin magnetic moment of the electron is




e = ge BS . (S4)

Combining equations S3 and S4, the Hamiltonian is obtained as


S1
0 2 1 1
HMD =
4
( )(
e N e N
r
)
r
(S5)
1 1 .
= 0 ge g I B N S I 2 S I
( )( )
4 r r


Now we examine the form of the Hamiltonian separately for the cases r=0 and r0. Given that a

hydrogenic wavefunction characterised by angular momentum quantum number l varies as rl at

the nuclear position, it is clear that it is only l=0 (ssymmetry) states that contribute at r=0.

Moreover, given that

1
2 = 4( r ) , (S6)
r

it is clear that the first term in equation S5 vanishes for r0. Further analysis yields


1 S r I r ( )( ) ,
1
( )( r
)
S I = S I 3
r 3 r2
(S7)

for r0, so that




r 2S
0 I 3 S r I r
( )( )
HMD = ge B g I N , (S8)
4 5
r

for r0, representing the dipolar interaction of the magnetic moments associated with the
electron andnuclear spin angular momenta. The generalisation of equation S8 to the form
appropriate for a scalar field of unpaired electron density, and the first order energy shift due to
the perturbation in the uncoupled regime are both straightforward to derive, and are as presented
in the main article. Returning now to the case r=0, we have already noted that the first term in
equation S5 is proportional to (r) at this limit, as shown in equation S6. The second term in
equation S5, for spherically symmetric l=0 states, contributes a further

S2

(S )( I ) 1r = 43 S I (r ) , (S9)

so that the total Hamiltonian for l=0 is obtained as



2 0
HMD = ge B g I N ( r )S I , (S10)
3

representing the Fermi contact interaction of ssymmetry states with the nuclear moment. Again,
the generalisation ofequation S10 to the form appropriate for a scalar field of unpaired electron
density, and the first order energy shift due to the perturbation in the uncoupled regime are both
straightforward to derive, and are as presented in the main article.

S3
S2. Detailed Derivation of the Scaling Factor

We recall that the factor scaling the DFT calculated contact shifts and dipolar tensors into
the paramagnetic regime is obtained as the ratio of the finite temperature paramagnetic and
saturated ferromagnetic magnetizations (M) thus

Mpara
= (S11)
Msat .

The paramagnetic magnetization may be straightforwardly expressed



Mpara = ng J BJB J ( ) (S12)
,

where n is the number density of transition metal (TM) ions; gJ, the Land gvalue; B, the Bohr

magneton; J, the total electronic angular momentum quantum number of the TM ion; and BJ(),
the Brillouin function. Following the argument for =gJBJB0/kB(T)<<1 outlined in the main
article, wherein the Brillouin function is approximated as BJ() (J+1)/3J, the paramagnetic
magnetization is obtained as

nB 0 g 2J 2BJ(J + 1) 2
nB 0eff
Mpara = = (S13)
3k B ( T ) 3k B ( T ) ,

amounting to a restatement of the CurieWeiss law in terms of the magnetization, and in which
the effective moment
. Note that eff and enter as empirical parameters
obtained from fits of the CurieWeiss law to the experimental susceptibility data, so that it is the
final form of equation S13 that is relevant here. Now, the saturated ferromagnetic magnetzation
representative of the DFT calculations is obtained as

Msat = nge BSform (S14)


,

where, in keeping with the absence of orbital angular momentum in the present DFT approach,

ge is the free electron gfactor; and Sform, the formal spin angular momentum quantum number of
the TM ion [here Sform = 5/2 for Fe(III)]. The scaling factor is thus finally obtained as

2
B 0eff
(B 0 , T,,Sform ,eff ) = (S15)
3k B ge BSform ( T ) .


S4
S3. XRay Diffraction

(a)

S5
(b)

Figure S1. X-ray powder diffraction patterns for (a) strengite (black) and DH strengite (blue),
and (b) phosphosiderite (black) and DH phosphosiderite (blue). XRD patterns of the dehydrated
phases were compared with the published data;2-3 red vertical lines denoting XRD patterns of
strengite and phosphosiderite simulated on the basis of the respective structural refinements.

(a)

(b)

S6
(c)

Figure S2. X-ray powder diffraction patterns for (a) LiFeP2O7, (b) monoclinic Li3Fe2(PO4)3, and
(c) rhombohedral Li3Fe2(PO4)3. Comparison with the published data are presented; red vertical
lines denoting XRD patterns simulated on the basis of the respective structural refinements.4-6

S7
S4. Refitting of the Magnetic Susceptibilities of DH Strengite and DH Phosphosiderite

(a)

(b)

Figure S3. Temperature dependence of the magnetic susceptibility of (a) DH phosphosiderite


and (b) DH strengite and corresponding fits to the sum of temperature independent (0) and

S8
Curie-Weiss contributions. For curves (1) the best three-parameter fit is shown with parameters
presented in Table S1. The resulting values of the effective magnetic moments are close to those
reported in Ref. [4] but are unphysical as eff cannot exceed the spin-only value of 5.92 B for
S=5/2 Fe3+ ions. For curves (2) C was fixed at 4.375 emu K/mol corresponding to eff = 5.92 B.
This model fits data very well with parameters presented in Table 1 (two-parameter fit). Curves
(2) and their fittings are shifted down by 10-3 emu/mol for clarity. Insets show temperature
dependences of the reciprocal susceptibility of (a) DH phosphosiderite after subtraction of (1) 0
= -8.310-4 emu/mol and (2) 0= -0.610-4 emu/mol and (b) DH strengite after subtraction of (1)
0 = -1.0410-3 emu/mol and (2) 0= -2.710-4 emu/mol. Straight lines represent least-square fits
to the Curie-Weiss law with C and values summarized in Table S1.

Table S1. Fitted magnetic parameters of DH phosphosiderite and DH strengite. 0 values are
obtained from the non-linear fits (Figure S3); C and values are obtained from linear fits (insets
in Figure S1).

0 (10-4 C (emu K/ Standard deviation


(K) eff (B)
emu/mol) mol) for linear fit
DH Phosphosiderite
3-parameter fit -8.3(4) 5.003(4) -174.89(6) 6.33 0.05
2-parameter fit -0.6(1) 4.375(5) -147.9(1) 5.92 0.07
DH Strengite
3-parameter fit -10.4(5) 4.927(3) -118.9(1) 6.28 0.03
2-parameter fit -2.7(5) 4.375(4) -97.9(1) 5.92 0.04

S9
S5. 31P MAS NMR Spectra of Monoclinic Li3Fe2(PO4)3

Experimental:

31P spin-echo MAS NMR experiments at 11.7 and 4.7 T were performed on Varian Infinity
Plus and CMX-200 spectrometers equipped with 1.3mm and 1.8 mm MAS probes, at Larmor
frequencies of approximately 202 and 81 MHz, respectively. All the 31P NMR spectra were
referenced to 85% wt. H3PO4 solution at 0 ppm. The NMR experiments used a spin-echo pulse
sequence of the form 90x180yacquire, where the evolution period was
synchronized with the spinning speed. A 90 pulse width of 1.15 s and a pulse delay of 0.05 s
were used to acquire the 11.7 T data, while a 90 pulse width of 1.50 s and a pulse delay of 0.05
s were used for the 4.7 T data. 31P MAS NMR spin-echo mapping experiments were performed
at 11.7 T with an irradiation frequency step size of 0.2 MHz, over range of 2 MHz, where the
step size was chosen to be less than 1 (0.22 MHz). A pulse delay of 0.05 s was used.
While the isotropic resonance of a spectrum arising from a single site spectrum can simply be
extracted from the position of the center of gravity of the static/MAS spectrum, the isotropic
resonances in a system with multiple sites is difficult to identify, since the peak positions are
affected by both the spinning frequency and the temperature. The monoclinic form of
Li3Fe2(PO4)3 falls within this troublesome class of materials. The spectra acquired at different
spinning speeds, as shown in Figure S4, illustrate the challenges encountered, given that the
sideband positions move as a consequence of the variation of spinning speed (as for a
diamagnetic compound) while all the peaks (sidebands and isotropic resonance) move (unlike
most diagmagnetic compounds) due to the variations in temperature (and consequently
susceptibility) arising out of frictional heating. Now, the highest intensity resonance in this
spectrum is most likely due to the P2 site, as discussed in the main text, but even with a
relatively modest decrease of ~8 kHz in spinning speed, its shift is observed to vary from 15483
to 15864 ppm; this is consistent with (i) the higher sample temperatures expected under MAS
conditions, and (ii) the increase in temperature with MAS spinning frequency.

S10
Figure S4. The 31P MAS NMR spectra of monoclinic Li3Fe2(PO4)3 at spinning speeds ranging
from 30 to 38 kHz obtained on a 4.7 T magnet. These data were collected without frequency
mapping.

In considering isotropic shifts of order 15000 ppm, we make the reasonable approximation
that the contribution due to the diamagnetic chemical shift is totally negligible and take the
isotropic shift as arising totally out of the hyperfine effects. Given that the hyperfine shift varies
directly with the susceptibility of the sample, which is in turn proportional to a factor 1/(T) in
the normal Curie-Weiss model, we may infer that a change in temperature from Ti to Tf leads to a
linear scaling of the peak positions according to the factor (Ti)/(Tf). Thus, dividing the shift
axis of each spectrum at temperature Tn by a factor Fn=(T0)/(Tn) (referred to hereafter as a
Tscaling of the spectrum) should effectively negate the effects of frictional heating. We
should now be able to identify the isotropic peaks in the normal manner, namely as those peaks
in the Tscaled spectra that do not move under variation of spinning speed. However, in the
event that the sample temperatures at different spinning speeds and values are not readily
available, we may still proceed on the following basis. Identifying the isotropic shift, 0, of one

S11
site in the system, and arbitrarily choosing one spectrum as reference (T0), we obtain the
appropriate Tscaling factor Fn of each spectrum directly as the quotient of the shifts n/0.
Here, for the reasons outlined in the main article, we identify the most intense peak at 15483
ppm in the 38 kHz (4.7 T) spectrum (Figure S4) as the isotropic peak associated with the P2 site.
Adopting the 38 kHz spectrum as reference for T0 and 0, performing the Tscaling discussed
above, and aligning the putative P2 shifts with the 38 kHz value, we obtain the set of spectra
shown in Figure S5.

Figure S5. The Tscaled and aligned 31P MAS spectra at 4.7 T.

It is clear from this that the resolution is still not sufficiently high to unambiguously
distinguish the remaining two isotropic peaks; additional spectra are required for comparison.
Accordingly, we make use of the MAS spectra obtained for the same sample on a 500 MHz
magnet at spinning speeds of 50, 53, 55 and 58 kHz. Since the magnitude of the hyperfine
interactions remains the same in ppm under different static magnetic fields, we may simply
perform further Tscalings and alignments (now adopting the putative P2 shift in the 11.7 T 58
kHz spectrum as the reference), leading to the composite set of spectra shown in Figure S6.
Clearly, the compilation of the two sets of varying field strength data provides welldefined
isotropic resonances of 15281, 14506 and 13674 ppm consistent with a spinning speed of 58
kHz.

S12
Figure S6. Composite set of Tscaled and aligned 31P MAS spectra obtained under static field
strengths of 4.7 T (spinning speeds of 30, 32, 34, 36, and 38 kHz) and 11.7 T (spinning speeds of
50, 53, 55, and 58 kHz). The putative P2 resonance at 15281 ppm in the 11.7 T spectrum at 58
kHz was chosen as the reference. Further isotropic resonances at 14506 and 13674 may then be
clearly identified.

Finally, there remains the problem of extrapolating the shifts obtained for monoclinic
Li3Fe2(PO4)3 above back to effectively static conditions, so as to be consistent with the 31P shifts
obtained for the other Fe(III) phosphates from the spin-echo mapping static experiments. Here,
we make use of the centres of gravity of the static and 58 kHz MAS spectra, lying at 15426.5 and
14530.4 ppm, respectively, to define an effective Tscaling factor of 1.0617 transforming the
latter spectrum to static conditions. Similarly, the 38 kHz MAS spectrum is first Tscaled with
reference to the 58 kHz spectrum, as above, and subsequently to static conditions, resulting in a
net Tscaling factor of 1.0478. An overlay of the two Tscaled MAS and single static spectra is
shown in Figure S7, yielding estimated isotropic shifts under static conditions of 16235, 15401
and 14514 ppm. The latter values may be compared on an equal basis with the 31P hyperfine
shifts presented for the other phosphate phases in the main article.

S13
Figure S7. Overlay of the 38 kHz (spin-echo) and 58 kHz (spinecho mapping) MAS and spin-
echo mapping static 31P spectra, the former two Tscaled so as to be consistent with the latter.

S14
S6. Comparison of the Present and Davis et al. 7Li Shift Assignments in Monoclinic
Li3Fe2(PO4)3

Table S2. Comparison of the bonding environments of Li1, Li2 and Li3 sites in the Patoux et al.5
(used in the present study) and Davis et al.7 structures, and the present and previous7 hyperfine
shift assignments. ni indicates value not included in the calculation of the respective mean or
standard deviation.
Patoux et al. Structure and Present Assignments Davis et al. Structure and Assignments
d(LiFe) (LiOFe) d(LiO) Assigned d(LiFe) (LiOFe) Assigned Reason for Assignment
Shift Shift
3.033 98.564 1.918 3.055 97.83
3.033 97.050 1.985 3.055 98.22
Li1 Li1
2.849 90.322 2.015 2.869 93.91
190 102 Assigned by elimination
2.849 93.758 1.907 2.869 92.45 of the other two sites.
mean 2.941 94.924 1.956 mean 2.962 95.603
0.106 3.665 0.052 0.107 2.864
3.018 93.496 2.08 3.033 102.6
3.018 100.093 1.946 3.033 90.41
Li2 2.824ni 77.297ni 2.456ni Li3 2.872 85.03 Assigned on basis that
2.824 82.829 2.102 90 2.872 77.92 216 this site shows the
2.824 90.621 1.874 2.872 89.51 strongest 90 overlap.
mean 2.921 91.760 2.001 mean 2.936 89.094
0.112 7.153 0.109 0.088 9.023
3.477 124.313 1.896 3.589 111.96
3.694ni 112.125ni 2.367ni 3.540 126.51
Li3 2.714ni 73.203ni 2.435ni Li2 2.742 83.43 Assigned on basis that
this site shows the
2.714 81.438 1.988 40 2.742 87.73 45
weakest
2.714 86.345 1.952 2.742 72.18 LiOFe overlap.
mean 2.968 97.365 1.970 mean 3.071 96.362
0.441 23.466 0.025 0.451 22.241

S15
S7. References

(1) Bransden, B. H.; Joachain, C. J. Physics of Atoms and Molecules; Longman Scientific
and Technical: Harlow, UK, 1983.
(2) Reale, R.; Scrosati, B.; Delacourt, C.; Wurm, C.; Morcrette, M.; Masquelier, C. Chem.
Mater. 2003, 15, 5051.
(3) Song, Y.; Zavalij, P. Y.; Suzuki, M.; Whittingham, M. S. Inorg. Chem. 2002, 41, 5778.
(4) Andersson, A. S.; Kalska, B.; Jonsson, P.; Haggstrom, L.; Nordblad, P.; Tellgren, R.;
Thomas, J. O. J. Mater. Chem. 2000, 10, 2542.
(5) Patoux, S.; Wurm, C.; Morcerette, M.; Rousse, G.; Masquelier, C. J. Power Sources
2003, 119-121, 278.
(6) Rousse, G.; Rodriguez-Carvajal, J.; Wurm, C.; Masquelier, C. Solid State Sci. 2002, 4,
973.
(7) Davis, L. J. M.; Heinmaa, I.; Goward, G. R. Chem. Mater. 2010, 22, 769.

S16

You might also like