Anal Bioanal Chem (2006) 385: 263271

DOI 10.1007/s00216-006-0374-9

SPECIAL I SSU E PAPER

Nancy J. McMillan . Catherine E. McManus .

Russell S. Harmon . Frank C. De Lucia Jr .
Andrzej W. Miziolek

Laser-induced breakdown spectroscopy analysis of complex

silicate mineralsberyl
Received: 9 November 2005 / Revised: 13 February 2006 / Accepted: 16 February 2006 / Published online: 17 March 2006
# Springer-Verlag 2006

Abstract Beryl (Be3Al2Si6O18) is a chemically complex

and highly compositionally variable gem-forming mineral
found in a variety of geologic settings worldwide. A
methodology and analytical protocol were developed for
the analysis of beryl by laser-induced breakdown spectroscopy (LIBS) that minimizes the coefficient of variance
for multiple analyses of the same specimen. The parameters considered were laser energy/pulse, time delay and
crystallographic orientation. Optimal analytical conditions
are a laser energy/pulse of 102 mJ and a time delay of 2 s.
Beryl compositions measured parallel and perpendicular to
the c axis were identical within analytical error. LIBS
analysis of 96 beryls from 16 countries (Afghanistan,
Brazil, Canada, China, Colombia, India, Ireland, Italy,
Madagascar, Mexico, Mozambique, Namibia, Norway,
Russia, Tanzania and United States), Antarctica, and ten
US states (AZ, CA, CO, CT, ID, ME, NC, NH, NM and
UT) were undertaken to determine whether or not LIBS
analysis can be used to determine the provenance of gem
beryl.

N. J. McMillan (*) . C. E. McManus

Department of Geological Sciences,
New Mexico State University,
Las Cruces, NM 88003, USA
e-mail: nmcmilla@nmsu.edu
Tel.: +1-505-6465000
Fax: +1-505-6461056
R. S. Harmon
Environmental Sciences Division,
US Army Research Office,
Research Triangle Park,
Durham, NC 27709, USA
F. C. De Lucia Jr . A. W. Miziolek
Weapons and Material Research Directorate,
US Army Research Laboratory,
Aberdeen Proving Ground,
Aberdeen, MD 21005, USA

Keywords Laser-induced breakdown spectroscopy .

LIBS . Beryl . Gem

Introduction
LIBS is a minimally destructive, real-time analytical
technique that has the potential to be an important tool
for geochemical analysis both in the laboratory and in the
field. Broadband LIBS captures the full emission spectrum
from 200 to 980 nm with a single laser shot, providing a
complete spectral analysis for a material. One particular
area where LIBS has an intrinsic advantage over other
analytical techniques, such as electron microprobe analysis
or x-ray fluorescence spectrometry, is for spatially-resolved
or in-field geochemical analysis of elements of low atomic
weight (i.e., Li, Be and B). In addition to providing a means
to perform real-time qualitative to quantitative chemical
analysis of geological materials [16], LIBS exhibits
potential as a tool for mineralogical/petrological analysis
[7, 8]. Based on the idea that every material produces a
unique emission spectrum, broadband LIBS analysis has
been applied to material identification and used to
distinguish between materials with similar elemental
signatures using full-spectrum statistical analysis [9, 10].
This paper presents the development of analytical
protocols that support a larger investigation of the
provenance of gem-quality beryl using a minimally
destructive chemical analysis tool. The data set consists
of 20 LIBS analyses for each of 96 beryl specimens of six
different colors from 16 countries, the Antarctic, and ten
different US states (Table 1). This variety of location and
color, which in beryl is generally attributed to chemical
composition, indicates that the database has captured much
of the natural compositional variation in beryl. Initial work
suggested that broadband LIBS could be used to
discriminate between different beryl samples of closely
similar composition. Prior to attempting to relate beryl
composition to location, however, we noted that the data
set has an unusually high variation of backgroundcorrected, internally-normalized peak intensities. We dis-

264

cuss the possible origins of this variation here and suggest

optimal analytical protocols for complex silicate minerals.

Structure, composition and color of beryl

Beryl is a Be- and Al-bearing silicate mineral with a
general chemical formula of Be3Al2Si6O18. Beryl most
commonly crystallizes in pegmatites (coarse-grained veins
and dikes) during the emplacement of granitic magmas into
the continental crust, or as the result of contact metamorphism or hydrothermal alteration. Because the trace
element composition of a beryl-bearing pegmatite should
reflect the chemical characteristics of the crustal domain in
which the deposit developed, it may be possible to
determine the source of a beryl gemstone (geographic
region, local area or possibly even specific mine) on the
basis of its trace element composition.
The molecular structure of beryl (Fig. 1) is dominated by
hexagonal rings comprised of six SiO tetrahedra. These
rings are stacked in a parallel manner to produce hollow
Table 1 Distribution of different colors in the beryl sample suite
Country/State
Afghanistan
Antarctica
Brazil
Canada
China
Colombia
India
Ireland
Italy
Madagascar
Mexico
Mozambique
Namibia
Norway
Russia
Tanzania
USA
Arizona
California
Colorado
Connecticut
Idaho
Maine
North Carolina
New Hampshire
New Mexico
Utah
Unknown locality

Colorless Blue Pink Yellow Green Reda

1

2
1
1

2
1

channels within the mineral structure, which accommodate

the presence of water and a variety of alkali cations with
large ionic radii (Na, K, Cs and Rb) within the crystal
lattice, but outside of the crystal formula unit [11, 12]. The
rings are linked by both tetrahedrally coordinated Be and
octahedrally coordinated Al. Coupled stoichiometric substitutions of ions of different charges lead to highly
complex variations in beryl chemistry. A threefold classification of beryl is based on these substitutions [12]:
(i) Octahedral beryl, in which the main substitutions are
Mg2+, Fe2+, Fe3+, Mn2+, Cr3+, V3+ and Ti4+ for Al3+ in
the octahedral site.
(ii) Tetrahedral beryl, in which the main substitution is Li+
for Be2+ in the Be tetrahedral site.
(iii) Normal beryl, in which both types of substitutions
occur, but to a limited extent.
The SiO2 concentration in beryl is reported to be highly
variable, ranging from 59.5 to 69.5 wt.%, although most
analyses show a smaller range of 63.266.7 wt.% [1115].
Substitutions involving Si4+ include Be2+ or Al3+ in the
silica tetrahedra and Si4+ in the Be tetrahedral site. The
charge imbalance produced by substitution of divalent
cations for Al3+, Li+ for Be2+, and Be2+ or Al3+ for Si4+
requires incorporation of monovalent cations (Na+, K+, Cs+
or Rb+) in the channel site or a tetravalent cation for Al3+
(Fig. 1b).

1
1

2
1
2
1

1
2
4

1
1
2

1
1

8
5

1
1
2
1
2
9

1
3
2

3
2

6
1
6

a
Includes red beryl (bixbite, Utah) and pezzottaite (Madagascar)
The common nomenclature for beryls with different colors is:
colorless=goshenite, blue/blue-green=aquamarine, pale-violet/
pink=morganite, light yellow/gold=heliodor, green=emerald, and
deep red-purple=pezzottaite

Fig. 1ab Molecular structure of beryl, Be3Al2Si6O18. a Structure projected onto (0001), looking down the c axis [11]. b Crosssections of the beryl structure across a ring of silica tetrahedra
[14, 15]

265

Fig. 2 Background-corrected average broadband LIBS spectra for

sample Antarctica-01, a white beryl

Fig. 4 Single-shot background-corrected broadband LIBS spectra

for emerald, the green variety of beryl, for samples from localities in
North Carolina in the USA, Colombia and China

The ARL laboratory broadband LIBS system consists of a

pulsed laser, mirror and focusing lens for laser light
delivery, adjustable sample stage, a fiber optic for plasma
light collection, a broadband spectrometer, and a desktop
computer for system control and data collection. A laser
pulse of 10 ns duration from an actively Q-switched Ultra
Nd-YAG laser (Big Sky, Inc., Bozeman, MT, USA)

operated at 1064 nm is focused by a 50 mm focal length

convex lens to generate a high-temperature microplasma at
the sample surface.
The sample to be analyzed is placed on a sample stage
and the height adjusted manually so that the laser spark is
observed to occur on the sample surface. The laser beam
diameter is 3 mm, with a beam divergence of 1 mr, so that
the laser spot size is about 60 m and the crater generated
on the sample is on the order of 100 m diameter. A bundle
of seven 600 m core diameter optical fibers is oriented to
collect the light emission from the plasma. A lens is placed
before the fiber optic bundle so that the plasma spark is
defocused to the point that spatial effects are eliminated and
the individual fibers collect the same light emission. The
fibers transmit the acquired light to an Ocean Optics, Inc.
(Dunedin, FL, USA) Model 2000 broadband spectrometer,
with a resolution of 0.1 nm, to separate the light from
different atomic, ionic and molecular constituents of the
plasma. The seven optical fibers are connected to seven
separate spectrometers, each of which covers one portion
of the 200980 nm spectral range. The LIBS spectra
collection software and the statistical LIBS library software
were provided by Ocean Optics Inc.
Two sets of experiments (laser energy/pulse and delay
time between the laser pulse and collection of light) were
conducted to determine the optimal analytical conditions
for beryl analysis. Twenty LIBS spectra of a gem-quality
beryl were acquired at a laser energy/pulse of 30, 60, 90,
102, 120 and 140 mJ. The optimal laser energy/pulse,

Fig. 3 Single-shot background-corrected broadband spectra for

aquamarine, the light blue variety of beryl, for samples from
localities in India, Connecticut in the USA, Namibia and
Afghanistan

Fig. 5 Single-shot background-corrected broadband LIBS spectra

for morganite, the pink variety of beryl, for samples from localities
in Madagascar, Italy, Brazil and Connecticut in the USA

In its pure form, beryl is colorless or white in color, but a

variety of trace metals in different oxidation states can
occur in beryl that contribute to a variety of different colors
and gem types [12]. Ultimately, the color exhibited by a
beryl specimen is determined by a complex set of factors
that include not only the amount and valence of an exotic
element and its location within the crystal lattice, but also
such factors as lattice distortion plus the degree and extent
of post-crystallization heating to which the specimen has
been subjected either naturally or artificially [12, 14, 16].
As a result of this highly variable chemistry, there is no
single cation in the beryl structure that can be considered to
be constant in composition from sample to sample.
Although the concentration of oxygen is relatively
consistent in beryl because of its abundance and constant
proportions in the SiO tetrahedra, small but variable
amounts of H2O and OH can be incorporated in the
channel site [1215].

Analytical methodology

266

The operating conditions of a laser energy/pulse of

102 mJ and a delay time for signal capture of 2.0 s were
used to analyze the full beryl sample suite. A single
broadband LIBS spectrum was acquired for the spectral
discrimination experiment described below. For the main
study, each of the 96 beryl samples was placed on the
sample stage and a single-shot broadband LIBS spectrum
was acquired for 20 different points selected at random
across the sample surface, following a cleaning shot at each
point.
Fig. 6 Single-shot background-corrected broadband LIBS spectra
for heliodor, the yellow variety of beryl, from localities in Brazil,
Norway, Namibia and Russia

Results and discussion

102 mJ, was chosen as that which produced the best

precision (here considered to be the minimum relative
standard deviation, or RSD, calculated as 100[/xav],
where is the standard deviation and xav is the arithmetic
average value of the data set) of Si/Be, Al/Be and Si/Al
background-corrected peak intensities. Similarly, twenty
spectra from the same sample were collected at 140 mJ
laser energy/pulse for each delay time (04 s at 0.5 s
intervals); 2.0 s produced the lowest RSDs for the
selected peak ratios.
Table 2 LIBS peaks used in
this experiment, with those used
as internal standards designated
in italics

Spectral discrimination of beryls

Figure 2 presents a typical broadband LIBS spectrum for
common beryl, with the most prominent peaks identified.
In previous work at the ARL laboratory, it was demonstrated that broadband LIBS could be used to discriminate
between different materials of closely similar composition
[8, 1720]. To test this idea, an experiment was conducted
to ascertain how effective LIBS might be at discriminating

Element

Wavelength

Spectrometer

Element

Wavelength

Spectrometer

Ag
Ag
Al
Al
Au
Au
B
Ba
Ba
Be
Be
Be
Ca
Ca
Ca
Cl
Co
Cr
Cr
Cs
Cs
Cu
Cu
F
Fe
Fe
Ga
Ge
K

241.31
328.07
394.36
396.15
226.36
282.23
249.77
455.4
553.55
313.03
234.82
527.08
393.37
422.67
643.9
542.32
350.23
360.53
428.97
460.38
894.56
213.6
324.75
385.1
239.56
373.71
645.59
283.18
766.49

1
2
3
3
1
1
1
3
4
2
1
4
3
3
5
4
2
2
3
3
7
1
2
2
1
2
5
1
6

K
Li
Li
Mg
Mg
Mg
Mn
Mn
Na
O
O
P
P
Pb
Pt
Rb
Rb
Sc
Si
Si
Si
Sn
Sr
Sr
Ti
Ti
V
Y
Zn
Zn

769.9
610.36
670.79
279.55
280.27
285.21
257.61
259.37
309.27
777.54
844.81
214.91
255.32
405.78
340.81
780.02
794.76
361.38
288.16
251.61
634.7
284
407.77
460.73
334.94
336.12
437.9
377.44
213.87
334.5

6
4
5
1
1
1
1
1
2
6
7
1
1
3
2
6
6
2
1
1
5
1
3
3
2
2
3
2
1
2

267

between different types of beryl. A library of single-shot

broadband LIBS spectra was acquired for a diverse suite of
21 beryl samples encompassing a range of colors from 11
different localities. Then, 37 LIBS unknown samples
were randomly selected from this beryl sample suite and a
single-shot broadband LIBS spectrum acquired for each
sample. The statistical best match of the unknown
sample spectrum to those present in the library was
determined by a linear correlation analysis in which peak
intensities for the more than 13,600 spectrometer channels
across the 200980 nm spectral range for each unknown
sample were compared to the 21 broadband LIBS spectra
in the spectral library. The best match assignment was
made to the library specimen that displayed the highest
correlation coefficient with the unknown sample. The
results of this experiment were highly encouraging, with 35
of 37 specimens identified correctly.
Figures 3, 4, 5 and 6 present broadband LIBS spectra of
different types of beryl from a variety locations worldwide.
A distinctive and important feature of each of these figures
is the variability in broadband LIBS spectra observed for
beryls of equivalent type (similar bulk composition). Such
variability is a feature which characterizes all of the data
collected in this study of almost a hundred beryl samples
(Table 1). Variation in LIBS emission line intensity is
expected from shot-to-shot variations resulting from the
extent of coupling of the laser energy to the sample surface
and the resultant temperature of the plasma that is created
on the sample surface. This phenomenon depends on a
large number of factors, some of which are systematic
while others are determined by the specific nature of the
target surface at the sub-mm scale at the point of analysis:
the surface texture, the roughness, and the small-scale
mineralogical or chemical heterogeneity of the sample at
the point on the surface where the plasma was generated.
Internal standard normalization methods
A total of 34 elements, representing the major and trace
element constituents of beryl, were examined in this study.
As noted above, overall spectral peak intensities varied
considerably over the 20 LIBS analyses for each sample. In
order to use the spectra with the most information, spectra
with total intensities less than the average total intensity for
the 20 shots (calculated by adding the intensities of all
channels) were removed from the data set. Appropriate
wavelengths for the elements of interest (Table 2) were
selected after visual inspection for interfering peaks. Peak
intensities were calculated as the maximum intensity
within a five-channel range centered on the peak wavelength. Background intensities were assumed to be the
mode of the intensities for each spectrometer. This
calculation eliminates background differences between
spectrometers but does not correct for nonlinear backgrounds associated with the edges of spectrometers.
Because of these edge effects, peaks near spectrometer
edges were not used in this study.

Two different approaches were used for the internal

normalization of individual peak intensities after background correction but prior to averaging (Table 2). In
Method I, elemental peak intensities were normalized to
the intensity of a major element peak (Be, Al, Si, or O) on
the same spectrometer. Method II normalizes all peak
intensities to the oxygen peak at 844.81 nm. This method is
based on the assumption that the oxygen concentration in
beryl plus that derived from the creation of the plasma in air
is relatively constant.
Both of these internal standard normalization methods
are subject to variability because of the stoichiometric
substitutions inherent in the beryl structure. For instance,
beryls can contain up to 0.5 Li atoms per formula unit
(apfu, based on 18 oxygens), which represents up to 17%
(one-sixth) of the Be apfu. It might be expected that the
Table 3 Average RSDs for multiple analyses of gem-quality beryls
and included beryl specimens, normalized by the spectrometer and
O methods
(nm) Method I
Average
RSD,
gemmy
Al
394.36
Be
234.82
Si
251.61
Ca
393.37
Cr
360.53
Cu
324.75
F 385.1
Fe
239.56
K 769.9
Li
670.79
Mg
279.55
Mn
257.61
Na
309.27
Rb
794.76
Sr
460.73
Ti
334.94
Zn
334.5

Method II
Average
RSD,
included

Average
RSD,
gemmy

Average
RSD,
included

4.9%

4.1%

42.0%

53.9%

17.2%

25.2%

31.9%

43.1%

6.5%

7.5%

32.8%

46.0%

65.8%

56.6%

71.7%

63.2%

46.7%

59.7%

49.5%

64.6%

69.2%

73.7%

68.6%

81.6%

48.5%
62.1%

64.9%
67.3%

50.9%
63.4%

65.4%
65.7%

65.0%
57.3%

73.1%
64.7%

66.2%
44.2%

78.4%
55.3%

47.0%

47.4%

56.8%

56.8%

32.2%

42.7%

41.2%

60.1%

28.5%

39.4%

40.6%

54.4%

22.2%

29.9%

29.1%

36.4%

43.2%

58.9%

50.1%

61.4%

67.5%

75.7%

73.5%

82.2%

43.2%

46.8%

51.8%

52.1%

268
Table 4 Average RSDs for analyses of multiple beryl crystals with different surface textures
(nm)

Method I

Method II

Average RSD, flat and Average RSD, flat

smooth
and rough
Al
394.36
Be
234.82
Si
251.61
Ca
393.37
Cr
360.53
Cu
324.75
F 385.1
Fe
239.56
K 769.9
Li
670.79
Mg
279.55
Mn
257.61
Na
309.27
Rb
794.76
Sr
460.73
Ti
334.94
Zn 334.5

Average RSD,
irregular

Average RSD, flat and Average RSD, flat

smooth
and rough

Average RSD,
irregular

3.5%

4.3%

4.0%

52.1%

46.2%

48.8%

14.5%

17.1%

23.1%

38.0%

30.7%

39.6%

6.6%

6.9%

6.6%

38.7%

31.4%

42.6%

53.3%

53.5%

55.2%

73.5%

68.2%

57.8%

36.3%

37.8%

60.0%

53.8%

42.9%

62.1%

68.3%

55.5%

72.0%

73.9%

63.2%

75.6%

51.1%
46.3%

50.2%
50.2%

61.9%
63.7%

53.7%
72.1%

58.1%
37.5%

59.4%
60.6%

64.4%
40.6%

71.5%
66.3%

69.7%
10.2%

66.6%
53.7%

83.1%
56.4%

71.0%
47.9%

52.1%

31.9%

45.7%

76.1%

40.8%

53.6%

17.9%

25.8%

41.6%

55.7%

47.6%

50.9%

24.0%

25.7%

36.6%

43.4%

38.6%

49.2%

14.2%

21.8%

28.5%

27.3%

32.8%

33.6%

39.0%

42.3%

50.2%

57.7%

51.3%

52.8%

66.5%

58.7%

78.5%

76.9%

71.5%

84.0%

36.5%

35.4%

51.1%

43.5%

44.1%

56.1%

Flat and smooth describes a perfect face on a gem-quality crystal. Flat and rough describes an imperfect crystal face. Irregular
describes a broken surface. Peak intensities normalized by Method I and Method II

Method II approach to normalization would provide a

consistent internal standard and produce more precise

Fig. 7 Variation of O and Be background-corrected peak intensities

for multiple shots of two beryl specimens, illustrating the negative
correlation between O and Be peak intensities observed for some
samples

results because only small amounts of H2O and OH

typically exist in the channel sites of the beryl structure.
The significantly higher RSDs for Method II than for
Method I (Tables 3 and 4) is an unexpected finding,
because O concentrations should vary less in both the beryl
structure and the atmosphere. However, for some specimens, the intensity of the O 844.81 nm peak does not
correlate positively with the intensity of the Be 313.03 nm
peak in multiple spectra of the same sample, as would be
expected if variations in total intensity are related to
analytical rather than mineralogic issues (Fig. 7). We
suspect that this phenomenon is related to the variable
height of the sample surface relative to the laser focal point
that was a feature of the analytical methodology employed.
Because the samples were of highly variable size and
shape, each was placed on an adjustable sample platform
that was subsequently raised by hand to visually bring the
sample surface to the point where the laser spark was

269

Peak intensity variability

Three additional sources of variability for the normalized
peak intensities are considered here:

Fig. 8 Variation in Li and Be background-corrected peak intensities

for multiple shots of Li-rich beryls (high Li concentrations implied
by intense Li 670.79 nm peaks)

1. Analysis of mineral or fluid inclusions within the beryl

structure;
2. Degree of coupling of the laser energy to the mineral
surface, which should be correlated to surface
roughness;
3. Direction of incident laser energy relative to the
crystallographic orientation of the sample.

Inclusions in beryl
created. Although care was taken to focus exactly on the
mineral surface, small differences in the position of the
spark relative to the sample surface probably caused shotto-shot variations in emission line intensity. The negative
correlations for the O and Be peak intensities (Fig. 7) can
be explained by the incorporation of different proportions
of beryl and air as a function of spark height. When the
spark is created slightly above the sample surface, the
analyzed volume should contain more air and less beryl,
resulting in higher O and lower Be intensities.
The applicability of Method I relies on constant
concentrations of Be, Al, Si, and O in all spectra from a
single specimen. If the mineral is chemically zoned
(variations in elemental composition occur on a very fine
spatial scale), the concentrations of the major elements will
vary from spectrum to spectrum. This phenomenon was
evaluated by examining the background-corrected peak
intensities from multiple analyses of the same sample.
Because Li selectively substitutes for Be, one would expect
a negative correlation between these elements, especially in
the Li-rich samples. This relationship is seen in these
samples (Fig. 8), suggesting that mineral zoning is a
contributing factor to the high shot-to-variability observed
in this study.

Fig. 9 Comparison of peak intensity relative standard deviation

(RSD) for gem-type and included-type beryl samples, normalized by
Method I (spectrometer method)

The mineral beryl crystallizes in water-rich environments,

either from hydrous silicate magmas or in metamorphic
rocks in which water and/or CO2 act as a catalysts to
enhance reaction rates. Thus, it is common for very small
bubbles of fluid (fluid inclusions) to be trapped within
the growing crystal structure. Similarly, microcrystals of
coprecipitating minerals can also often be included in
beryls. The degree to which such inclusions are present can
be evaluated visually by the transparency of the specimen,
as inclusions cause the crystal to be cloudy, translucent or
even opaque. The presence of different amounts of
inclusions within each analyzed volume will significantly
affect the precision of the average normalized peak
intensities. Based on visual inspection of a subset of our
samples, 20 samples are gem-quality and 18 are translucent-to-opaque, indicating that they are rich in inclusions.
Average RSDs of multiple analyses for 20 gem-quality

Fig. 10ab Peak intensities, normalized using Method I, for

elements that might be ablated differently when the laser is directed
parallel to or perpendicular to the c crystallographic axis; a Be
versus K and b Si versus Rb

270

specimens and 18 included specimens, normalized by both

approaches, are presented in Table 3 and Fig. 9. The
average RSDs for gem-quality samples are almost
universally lower for the included samples (Table 3,
Fig. 9). Thus, we conclude that mineral and fluid inclusions
do indeed contribute to the high degree of variability in the
data set. However, the gem-quality samples still have RSDs
that are much higher than typically observed [4, 7]. Also,
the precision observed for Method II is much worse,
especially for the major elements Be, Al, and Si, than for
Method I.
Surface roughness
Shot-to-shot variation in LIBS emission line intensity
might be expected to result from differences in the extent
of coupling of the laser energy to a sample surface and
the resultant temperature of the plasma that is created on
the sample surface, particularly where the surface of the
sample is rough or compositionally inhomogeneous at the
sub-mm scale [8].
A subset of 36 of the 96 beryl samples were categorized
on the basis of surface roughness as either (i) flat and very
smooth (gem-quality), (ii) flat with small irregularities, or
(iii) very rough and irregular. The average RSDs for major
and trace element peaks show no regular differences for the
three different surface qualities (Table 4).
Crystallographic orientation during analysis
The beryl structure is quite different when viewed from
different directions (Fig. 1). Parallel to the c crystallographic axis, the structure appears to consist of rings of
Si-O tetrahedra with large monovalent cations in the center.
Perpendicular to the c axis, the structure appears to consist
of layers of SiO, AlO, and BeO coordination
polyhedra. If the LIBS plasma samples a very small
amount of sample, the beryl composition will appear to
change as a function of the direction of incident laser light.
To evaluate the crystallographic control of LIBS analysis, two samples with well-developed crystal faces were
analyzed parallel to and perpendicular to the
c crystallographic axis (Fig. 10). When the laser is focused
parallel to the c axis, one might expect that ions that reside
in channels parallel to the c axis (Na, Rb) could be more
easily ablated than elements that reside in the tetrahedral
sites (Si, Be). However, no differences in normalized peak
intensities were observed related to the direction of
analysis. Because of this, no care was taken during analysis
to orient the samples, except to place the mineral surface to
be analyzed perpendicular to the path of the incident laser
light.

Summary and conclusions

A methodology and analytical protocol were developed for
the LIBS analysis of beryl (Be3Al2Si6O18), a chemically
complex gem mineral, that takes into account laser energy/
pulse, time delay between creation of the laser spark and
signal recording, and crystal orientation. This procedure
minimizes the variance for background-corrected element
emission line intensities. Optimal analytical conditions are
a laser energy/pulse of 102 mJ and a time delay of 2 s.
Beryl compositions measured parallel and perpendicular
to the crystallographic c axis were identical within
analytical error, indicating that the mass ablated by a
single nanosecond laser shot is sufficient to provide a
homogeneous sampling of this mineral of highly asymmetric crystal chemistry. A spectral matching experiment
demonstrated that single-shot broadband LIBS spectra can
provide a basis for discriminating between different
specimens of beryl.
An unexpectedly high variability in emission line
intensities was observed for two different internal normalization approaches based upon 20 LIBS analysis of 96
beryls from 27 different localities worldwide. This result
highlights the difficulties in both analytical procedures and
data manipulation protocols for LIBS analysis of a
complex silicate mineral such as beryl. Three factors
have been demonstrated to contribute to the precision of
our analytical dataset: (i) the presence of mineral or fluid
inclusions, (ii) mineral zoning, and (iii) the position of
plasma volume formation relative to the sample surface.
Because every element in the beryl structure can experience significant stoichiometric substitution, it is difficult, if
not impossible, to find an element to serve as an internal
standard. Even the use of oxygen as an internal standard is
problematic, because small variations in laser focal height
lead to differences in the berylair mix that is analyzed,
enhancing the variability caused by intrinsic mineralogic
factors. Future work will be directed toward ensuring that
the LIBS spark is focused precisely on the sample surface,
and, for both analysis in an Ar atmosphere and doublepulse analysis, toward eliminating or minimizing the effect
of entrained air on the LIBS analysis.
Acknowledgements The authors acknowledge a seed grant to
McMillan from National Science Foundation ADVANCE grant #
0123690 to New Mexico State University. We thank Robert
Whitmore for access to the Palermo Mine, New Hampshire, and
Tori Gomez for logistical and analytical assistance. We are grateful to
Micheal Wise, Smithsonian Institution, George Harlow, American
Museum of National History, and Ed Grew, University of Maine,
who graciously lent us beryl specimens for this work. Thanks also go
to Kevin McNesby, Nancy Flores, and an anonymous reviewer, who
provided valuable comments that improved the paper.

References
1. Grant KJ, Paul GL, ONeill JA (1991) Appl Spectrosc
4:701705
2. Wisbrun R, Schechter I, Niessner R, Schrder H, Kompa K
(1994) Anal Chem 66:29642975

271
3. Bolger JA (2000) Appl Spectrosc 54:181189
4. Rosenwasser S, Asimellis G, Bromley B, Hazlett R, Martin J,
Pearce T, Zigler A (2001) Spectrochim Acta B 56:707714
5. Hilbk-Kortenbruck F, Noll R, Wintjens P, Falk H, Becker C
(2001) Spectrochim Acta 56:933945
6. Capitelli FG, Colao F, Provenzano MR, Fantoni R, Brunetti G,
Senesi N (2002) Geoderma 106:4562
7. Fabre C, Boiron, M-C, Dubessy J, Chabiron A, Charov B,
Crespo TM (2004) Geochim Cosmochim Acta 66:14011407
8. Harmon RS, De Lucia FC, Miziolek AW, McNesby KL,
Walters RA, French PD (2005) Geochem Explor Environ Anal
5:2128
9. Gornushkin IB, Smith BW, Nasajpour N, Winefordner JD
(1999) Anal Chem 71:51575164
10. Anzano JM, Gornushkin IB, Smith BW, Winefordner JD
(2000) Polym Eng Sci 40:24232429
11. Klein C (2002) Mineral science (after James D. Dana). Wiley,
New York

12. Deer WA, Howie RA, Zussman J (1975) Rock forming

minerals, vol 1: Ortho- and ring silicates. Longman, London
13. Aurisicchio C, Fioravanti G, Grubessi O, Zanazzi PF (1988)
Am Mineral 73:826837
14. Sinkankas J (1981) Emerald and other beryls. Chilton,
Radnor, PA
15. Aurisicchio C, Grubessi O, Zecchini P (1994) Can Mineral
32:5568
16. Mukerjee B (1951) Nature 167:602
17. DeLucia FC, Harmon RS, McNesby KL, Winkel RJ, Miziolek
AW (2003) Appl Opt 42:61486152
18. Samuels AC, De Lucia FC Jr, McNesby KL, Miziolek AW
(2003) Appl Opt 42:62056209
19. DeLucia FC Jr, Samuels AC, Harmon RS, Walters RA,
McNesby KL, La Pointe A, Winkel RJ Jr, Miziolek AW
(2005) IEEE Sens J 5:861869
20. Munson CA, De Lucia FC Jr, Piehler T, McNesby KL,
Miziolek AW (2005) Spectrochim Acta B 60:12171224