Synthesis of the new metastable skutterudite compound NiSb3 from modulated elemental reactants

Article abstract of DOI:10.1021/ic011131j

A new metastable binary compound with the skutterudite crystal structure has been synthesized from modulated elemental reactants, through an amorphous intermediate, using a novel low-temperature synthesis technique. The amorphous reaction intermediate undergoes nucleation at 87 °C, an extremely low temperature for solid-state reactions. When heated above 350 °C, the metastable phase NiSb₃ disproportionates into the thermodynamically stable phases NiSb₂ and Sb. Also, if the sum of the individual elemental layer thicknesses is greater than 30 A, a mixture of different phases forms. Simulation of the high-angle powder X-ray diffraction spectrum confirms that NiSb₃ is isostructural with CoSb₃.

Full text of DOI:10.1021/ic011131j

Inorg. Chem. 2002, 41, 4127−4130
Synthesis of the New Metastable Skutterudite Compound NiSb3 from
Modulated Elemental Reactants
Joshua R. Williams and David C. Johnson*
Department of Chemistry and Materials Science Institute, UniVersity of Oregon,
Eugene, Oregon 97403
Received November 1, 2001
A new metastable binary compound with the skutterudite crystal structure has been synthesized from modulated
elemental reactants, through an amorphous intermediate, using a novel low-temperature synthesis technique. The
amorphous reaction intermediate undergoes nucleation at 87 °C, an extremely low temperature for solid-state
reactions. When heated above 350 °C, the metastable phase NiSb3 disproportionates into the thermodynamically
stable phases NiSb2 and Sb. Also, if the sum of the individual elemental layer thicknesses is greater than 30 Å,
a mixture of different phases forms. Simulation of the high-angle powder X-ray diffraction spectrum confirms that
NiSb3 is isostructural with CoSb3.
Introduction
orders of magnitude smaller than found in bulk reactions,
the stacked elemental layers can be annealed at low tem-
perature (usually below 200 °C) to complete diffusion of
the elemental layers. For most systems, an initial annealing
temperature can be chosen which will allow complete
diffusion of the elemental layers without providing enough
energy to overcome the activation energy barrier for nucle-
ation. This produces an amorphous intermediate that, on
further annealing, will undergo nucleation of a crystalline
phase. Due to the low-temperature reaction conditions, long-
range diffusion processes are very slow. Therefore, the
composition of the amorphous phase determines which phase
will nucleate, and the compound that is easiest to nucleate
at the given composition will undergo nucleation, whether
it is the most thermodynamically stable phase or not. In this
way, we are able to “trap” kinetically stable phases.3
The ability to prepare new metastable compounds raises
the question of how to predict the existence of new
metastable phases. There are two main directions that one
could take. The first would be to search for compounds with
new and novel structures.4 A second approach is to start with
known compounds and structure types, and to search for
variations of those structures that could potentially be
synthesized. In this paper we used this second strategy. An
examination of the cobalt-antimony phase diagram, as
shown in Figure 1, reveals the existence of the phase CoSb3,
Due to the extremely slow diffusion rates typical of solid-
state reactions, long annealing times and high temperatures
are generally required to drive reactions to completion. These
synthesis conditions, however, are not conducive to the
formation of compounds that are only metastable, or kineti-
cally stable, particularly if they are thermodynamically
unstable with respect to disproportionation. Unfortunately
there are not many alternative solid-state synthesis methods
available to prepare compounds that will only be accessible
at low temperature.1 Indeed, even the preparation of ther-
modynamically stable compounds that incongruently melt
at low temperatures requires knowledge of the phase diagram
for successful synthesis.2
In the past few years, we have developed a novel synthesis
method that allows the synthesis of solid-state inorganic
materials at low temperatures, usually lower than 500 °C.
This method, called the modulated elemental reactant
method, involves using a high-vacuum physical vapor
deposition system to sequentially deposit ultrathin elemental
layers onto a substrate. The thickness of these elemental
layers, usually on the order of angstroms, determines the
important diffusion distances in the subsequent interdiffusion
of these layers. Since these diffusion distances are many
* Author to whom correspondence should be addressed. E-mail: davej@
oregon.uoregon.edu.
(1) Stein, A.; Keller, S. W.; Mallouk, T. E. Science 1993, 259, 1558-
1564.
(3) Johnson, D. C. Curr. Opin. Solid State Mater. Sci. 1998, 3, 159-
167.
(4) Schon, J. C.; Jansen, M. Angew. Chem. 1996, 35, 1286-1304.
(2) Corbett, J. D. In Synthesis of solid-state materials; Corbett, J. D., Ed.;
Clarendon Press: Oxford, 1987; pp 1-38.
10.1021/ic011131j CCC: $22.00 © 2002 American Chemical Society
Published on Web 07/17/2002
Inorganic Chemistry, Vol. 41, No. 16, 2002 4127

Williams and Johnson
Figure 3. The Ni-Sb phase diagram.8
Figure 1. The Co-Sb phase diagram.8
Experimental Section
Layer deposition was performed in a custom-built high-vacuum
deposition chamber with multiple sources and targets, which is
described elsewhere.7 A personal computer and custom automation
software are used to control the deposition. Nickel was evaporated
from a Thermionics electron-beam gun at a rate of ∼0.5 Å/s.
Antimony was evaporated from a custom-built Knudsen cell at a
rate of ∼0.8 Å/s. Deposition rates and thicknesses were monitored
using Leybold Inficon XTC/2 quartz crystal thickness monitors.
The films were deposited onto polished silicon wafer pieces (for
low-angle X-ray diffraction), miscut quartz pieces (for high-angle
X-ray diffraction), and silicon wafers coated with poly(methyl
methacrylate) (PMMA). The coated wafers were then soaked in
acetone to dissolve the PMMA and allow the multilayer films to
lift off of the wafer surface. The pieces of film were then collected
by vacuum filtration using Teflon filter papers.
Figure 2. The skutterudite crystal structure where the small-diameter balls
indicate the positions of cobalt atoms and the larger diameter circles
represent the positions of antimony atoms. Projection is along the (100)
axis.
Sample composition was determined using electron probe
microanalysis (EPMA) using a Cameca S-50 instrument, a 10 keV
accelerating voltage, a 10 nA beam current, and a 1 µm spot size.
High-angle diffraction data on the powder after differential scanning
calorimetry was obtained on a Scintag SDS-2000 θ-2θ diffrac-
tometer using a piece of miscut quartz as the sample holder. High-
angle diffraction data on the film-coated miscut quartz pieces, as
well as all low-angle diffraction data, was obtained using a Philips
X’Pert five-circle diffractometer. All X-ray experiments were
performed using Cu KR radiation. A TA Instruments DSC 2920
modulated DSC was used to collect all DSC data. The heat flow
in and out of the sample is monitored as the sample temperature is
elevated at a specific rate to determine at what temperature phase
changes and nucleation events occur. Approximately 0.500 mg of
sample is placed and sealed inside an Al sample pan. The samples
were heated under flowing nitrogen at a rate of 2 °C/min up to
either 200 or 250 °C.
which has the skutterudite structure type shown in Figure 2.
This structure type is currently of great interest to the
thermoelectric community due to the demonstrated potential
of “filling” the void in the structure with small heavy atoms,
resulting in a reduction of the thermal conductivity of the
material.5,6 Upon comparing the nickel-antimony phase
diagram shown in Figure 3 with the cobalt-antimony phase
diagram in Figure 1, one observes that they are very similar
above 40% antimony. They both contain 1:1 and 1:2
compounds with identical structures stable in similar com-
position ranges. The main difference in this region is the
existence of the incongruently melting CoSb3 and the absence
of an isostructural compound in the Ni-Sb phase diagram.
Considering the similarities in these phase diagrams, we
decided to search for the existence of an analogous meta-
stable skutterudite phase. With this goal in mind, several
modulated elemental reactants were synthesized, varying the
relative layer thickness of nickel and antimony to adjust the
overall composition of the sample as well as the repeat layer
thickness (the sum of both elemental layer thicknesses).
Results
A list of the samples prepared in this study is shown in
Table 1. The intended repeat thickness for each sample (the
sum of the individual elemental layer thicknesses) is shown
(7) Fister, L.; Li, X. M.; Novet, T.; McConnell, J.; Johnson, D. C. J. Vac.
Sci. Technol., A 1993, 11, 3014-3019.
(8) Massalski, T. B.; Okamoto, H.; Subramanian, P. R.; Kacprzak, L. In
Binary Alloy Phase Diagrams, 2nd ed.; Massalski, T. B., Okamoto,
H., Subramanian, P. R., Kacprzak, L., Eds.; ASM International:
Materials Park, OH, 1990; Vol. 3, pp 2664-2665.
(5) Sales, B. C.; Mandrus, D.; Williams, R. K. Science 1996, 272, 1325-
1328.
(6) Nolas, G. S.; Slack, G. A.; Morelli, D. T.; Tritt, T. M.; Ehrlich, A. C.
J. Appl. Phys. 1996, 79, 4002-4008.
4128 Inorganic Chemistry, Vol. 41, No. 16, 2002

Synthesis of NiSb3 from Modulated Elemental Reactants
Table 1. Summary of Samples Prepared Showing Layer Thickness (Å), Composition, and Phase Formation
intended thickness (Å)
atomic %
bilayer thickness (Å)
intended actual
thickness %
Sb
phase(s)
formed
sample
Ni
Sb
A% Ni
A% Sb
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
0.5
0.75
1
1.25
0.4
1.5
0.75
0.8
1.6
2
0.4
0.5
0.35
0.4
0.35
15
15
15
15
15
30
15
30
60
75
15
15
15
15
15
0.97
0.95
0.94
0.92
0.97
0.95
0.95
0.97
0.97
0.97
0.97
0.97
0.98
0.97
0.98
17.3
24.8
32.2
34.3
15.1
30.2
25.6
22.6
17.9
17.1
24.9
27.7
20.3
18
82.3
75.2
67.7
65.7
84.9
69.7
74.4
77.4
82.1
82.9
75.1
72.3
79.7
82
15.5
15.8
16
16.25
15.4
31.5
15.8
30.8
61.6
77
15.4
15.5
15.4
15.4
15.4
15.1
15.6
15.0
15.2
14.6
28.0
15.2
30.2
56.9
62.9
14.7
15.5
13.8
13.9
14.0
NiSb3
NiSb3
NiSb2, Sb
NiSb2, Sb
NiSb3
NiSb2, Sb
NiSb3
NiSb3, Sb
NiSb3, Sb
NiSb2, Sb
NiSb3
NiSb2, Sb
NiSb3
NiSb3
NiSb3
20.3
79.6
Figure 6. Intended composition (from intended layer thicknesses) vs
measured composition determined using microprobe analysis.
Figure 4. A graph of the intended bilayer thickness and the observed
bilayer thickness determined using low-angle X-ray diffraction. The line
drawn is the expected linear relationship between these thicknesses.
antimony thickness (i.e., thickness of Sb/thickness of Sb +
thickness of Ni) versus atomic percent antimony determined
from microprobe analysis. In general there is good agreement
between the intended and measured composition. The largest
source of error results from small positional changes of the
deposition sources as they are reinstalled over the time
required for this study.
We began this study by preparing several samples, with
compositions both antimony rich and antimony poor relative
to the ideal 1:3 Ni:Sb ratio of our proposed compound, to
determine the effect of composition on phase formation. To
minimize interference from interdiffusion effects, this initial
set of samples was prepared with repeat layer thickness below
20 Å. The absence of Bragg diffraction maxima in the low-
angle diffraction spectra for most of these samples suggests
that either some diffusion has occurred on deposition or the
layering is very nonuniform. We observed that the phase
formed depends very strongly on the overall composition of
the sample. If the Sb:Ni ratio is even slightly below 3:1, the
thermodynamically stable phase NiSb2 forms, as well as Sb.
If the composition is more Sb rich, NiSb3 forms. We
observed no instances of samples composed of thin repeating
bilayers where the phase formation deviated from this
composition dependence. This strong dependence on the
overall sample composition suggests that the samples do in
fact become homogeneous before nucleation occurs.
The formation of both NiSb2 and NiSb3 is clearly visible
in a DSC trace of the as deposited samples. On heating
samples with an Sb:Ni ratio below 3:1, a sharp exotherm is
observed at 92 °C. Diffraction scans before this exotherm
show no high-angle diffraction peaks, while after this
Figure 5. Low-angle X-ray diffraction spectrum illustrating both a Bragg
diffraction maximum resulting from the repeat thickness and the front/back
film reflections.
as well as the intended thickness of each elemental layer.
Figure 4 shows the agreement of the intended thickness
(determined from the deposition parameters) to the measured
layer thicknesses determined directly from low-angle X-ray
diffraction data. Layer thickness is determined using Bragg
diffraction maxima in the low-angle X-ray diffraction, a
sample of which is shown in Figure 5, or by finding the
total film thickness from the front/back reflections in the low-
angle X-ray diffraction spectra, and dividing by the number
of repeats deposited. In some cases neither front/back
reflections nor Bragg peaks were observed presumably as a
result of sample roughness. While the agreement between
the intended thickness and the actual thickness of the layers
in the samples is within an angstrom, the small changes in
intended nickel layer thickness are less than the standard
deviation in Figure 4. In Figure 6 we graph the percent
Inorganic Chemistry, Vol. 41, No. 16, 2002 4129

Williams and Johnson
To determine the stability limits of this compound, we
heated a sample in the DSC under flowing nitrogen. NiSb3
appears to be only kinetically stable, as it disproportionates
into the thermodynamically stable phases NiSb2 and Sb upon
annealing above 250 °C.
Since bulk samples do not react to form NiSb3, there must
exist a critical repeat layer thickness below which the samples
interdiffuse without nucleating NiSb2 at the reacting interface.
To determine this critical thickness, samples 8-10 were
prepared. These samples show a distinct change in behavior
as a function of bilayer thickness. Samples with bilayer
thickness below 56.9 Å and compositions antimony rich of
NiSb3 nucleated NiSb3. Samples with layer thicknesses above
56.9 Å nucleated a mixture of different phases, including
Sb with highly preferred orientation. This suggests that, in
these thicker samples, nucleation is occurring at the interfaces
between elemental layers before the layers can completely
interdiffuse. Thus the critical bilayer thickness must lie
between these two values. It should be noted that we have
observed no relationship between heating rate in the DSC
and critical bilayer thickness for the range of heating rates
typically used in DSC experiments (1-10 °C/min).
Figure 7. A differential scanning calorimetry (DSC) trace showing a clear
exothermic nucleation event for the formation of NiSb3.
Figure 8. High-angle X-ray diffraction of a sample forming NiSb3 before
and after annealing. A powder diffraction simulation assuming the skut-
terudite structure is superimposed over the postannealing scan. Additional
peaks correspond to metallic Sb.
Summary
We have demonstrated the ability to synthesize a new
metastable binary skutterudite compound using the modu-
lated elemental reactant method. The abrupt composition
dependence of phase formation as well as low-angle dif-
fraction data suggests that the sample becomes amorphous
and homogeneous on initial annealing, and nucleates homo-
geneously upon further annealing. The new compound
appears to be only kinetically stable, as it disproportionates
into the thermodynamically stable phases NiSb2 and Sb upon
annealing above 250 °C. We have also demonstrated the
importance of diffusion distance in this synthesis method
by showing that the repeat layer thickness must be below
30 Å in order for the metastable phase to form.
exotherm the expected diffraction maxima from the known
compound NiSb2 are observed. If the ratio is at or above
3:1, however, the targeted metastable skutterudite phase
NiSb3 forms. A representative sample of the DSC traces
observed from samples that form NiSb3 is shown in Figure
7. A clear exothermic peak is observed at 87 °C. High-angle
diffraction of this sample before and after the annealing in
the DSC is shown in Figure 8. The diffraction data clearly
demonstrates that the exothermic peak observed in the DSC
trace results from the nucleation and growth of the targeted
NiSb3 phase. The high-angle X-ray diffraction data is
modeled very well by replacing cobalt with nickel in the
published structure of CoSb3 and adjusting the lattice
parameter to 9.13 Å, also shown in Figure 8. Rietveld
refinement of the structure proved difficult due to a complex
preferred orientation, which is unusual for a cubic system.
Acknowledgment. The authors are grateful for the
support of this work by ONR Contract N00014-98-1-0447
and NSF Grant DMF-9813726 at the University of Oregon.
IC011131J
4130 Inorganic Chemistry, Vol. 41, No. 16, 2002