Understanding fluxes as media for directed synthesis: In situ local structure of molten potassium polysulfides

Article abstract of DOI:10.1021/ja303047e

Rational exploratory synthesis of new materials requires routes to discover novel phases and systematic methods to tailor their structures and properties. Synthetic reactions in molten fluxes have proven to be an excellent route to new inorganic materials because they promote diffusion and can serve as an additional reactant, but little is known about the mechanisms of compound formation, crystal precipitation, or behavior of fluxes themselves at conditions relevant to synthesis. In this study we examine the properties of a salt flux system that has proven extremely fertile for growth of new materials: the potassium polysulfides spanning K₂S₃ and K ₂S₅, which melt between 302 and 206 °C. We present in situ Raman spectroscopy of melts between K₂S₃ and K ₂S₅ and find strong coupling between n in K ₂S_n and the molten local structure, implying that the S_n^2- chains in the crystalline state are mirrored in the melt. In any reactive flux system, K₂S_n included, a signature of changing species in the melt implies that their evolution during a reaction can be characterized and eventually controlled for selective formation of compounds. We use in situ X-ray total scattering to obtain the pair distribution function of molten K₂S₅ and model the length of Sn^2- chains in the melt using reverse Monte Carlo simulations. Combining in situ Raman and total scattering provides a path to understanding the behavior of reactive media and should be broadly applied for more informed, targeted synthesis of compounds in a wide variety of inorganic fluxes.

Full text of DOI:10.1021/ja303047e

Article
pubs.acs.org/JACS
Understanding Fluxes as Media for Directed Synthesis: In Situ Local
Structure of Molten Potassium Polysulfides
Daniel P. Shoemaker,† Duck Young Chung,† J. F. Mitchell,† Travis H. Bray,‡ L. Soderholm,‡
Peter J. Chupas,¶ and Mercouri G. Kanatzidis*,§,†
‡
¶
†Materials Science Division, Chemical Sciences and Engineering Division, and X-ray Science Division, Argonne National
Laboratory, Argonne, Illinois 60439, United States
§Department of Chemistry, Northwestern University, Evanston, Illinois 60208, United States
ABSTRACT: Rational exploratory synthesis of new materials
requires routes to discover novel phases and systematic
methods to tailor their structures and properties. Synthetic
reactions in molten fluxes have proven to be an excellent route
to new inorganic materials because they promote diffusion and
can serve as an additional reactant, but little is known about
the mechanisms of compound formation, crystal precipitation,
or behavior of fluxes themselves at conditions relevant to
synthesis. In this study we examine the properties of a salt flux
system that has proven extremely fertile for growth of new
materials: the potassium polysulfides spanning K2S3 and K2S5, which melt between 302 and 206 °C. We present in situ Raman
spectroscopy of melts between K2S3 and K2S5 and find strong coupling between n in K2Sn and the molten local structure,
2−
implying that the Sn chains in the crystalline state are mirrored in the melt. In any reactive flux system, K2Sn included, a
signature of changing species in the melt implies that their evolution during a reaction can be characterized and eventually
controlled for selective formation of compounds. We use in situ X-ray total scattering to obtain the pair distribution function of
2−
molten K2S5 and model the length of Sn chains in the melt using reverse Monte Carlo simulations. Combining in situ Raman
and total scattering provides a path to understanding the behavior of reactive media and should be broadly applied for more
informed, targeted synthesis of compounds in a wide variety of inorganic fluxes.
of the reaction itself. For that reason we focus on fluxes of the
molten alkali polychalcogenides (A2Qn, where A is an alkali
metal and Q a chalcogenide) which exhibit low melting
temperatures and Q−Q bonding. Interactions between the
solvent and the solute allow the formation of phases that are
unattainable by direct combination of their constituents.4−11
Reactions in these fluxes have produced promising new
materials for use in photovoltaics,12−14 thermoelectrics,15−18
nonlinear optics,13,19−21 and heavy metal and nuclear waste
remediation.22,23 Alkali species in the flux can dismantle a
preformed structure, leading to tunable properties by a process
known as dimensional reduction.24,25 It is also possible for
structural motifs from the flux itself to be incorporated in the
final product.26,27 But how do these compounds form, and from
what?
INTRODUCTION
■
The concept of inorganic materials synthesis by design is
predicated on the ability to control the formation of metastable
phases. This is a particular challenge in the creation of energy-
relevant inorganic materials, where slow diffusion rates are
often countered by high-temperature reactions which produce
only the thermodynamically stable products. This is epitomized
in the case of oxides, which are commonly refractory due to
strong ionic bonding.
A more desirable approach would share the advantages of
organic synthesis: low-temperature reactions and gentler
chemistry allow the selective assembly of tailored molecules.
In solid-state chemistry the closest approximation is reactive
flux synthesis, where a low-melting element (In, Ga, Sn, S, etc.)
or salt (oxide, other chalcogenide, halide) permits facile
diffusion and may take part in the reaction itself. Especially
when salts of the less electronegative chalcogenides (S, Se, or
Te) are used, the combination of gentle cation−anion bonding
and low melting temperatures permits the formation of many
novel or metastable phases.1−3 Despite the versatility of these
reactions and the vast array of products discovered, little is
known about the reaction mechanisms, species present in the
flux, or the behavior of the flux itself at high temperatures.
The ability to selectively nucleate complex phases is aided
when a flux not only maintains diffusion but is an integral part
Only one study by Durichen and Bensch has attempted to
̈
probe reaction mechanisms in a polyanionic flux.28 They found
that Nb in K2Sn fluxes at 350 °C gave four products for varying
n, but only ex situ measurements were performed. There is no
consensus on suitable methods to characterize these systems at
conditions relevant to crystal growth. Attempts to model the
stability of polyanionic species have been limited to the gas
Received: March 29, 2012
Published: May 14, 2012
© 2012 American Chemical Society
9456
dx.doi.org/10.1021/ja303047e | J. Am. Chem. Soc. 2012, 134, 9456−9463

Journal of the American Chemical Society
Article
phase,29,30 but these isolated molecules do not share the
structure or stability of their counterparts in liquid salts.31
In order to understand, and eventually control, directed
syntheses of inorganic materials in molten fluxes, it is first
necessary to characterize the reaction medium. Once details are
known about chemical species contained in the flux, behavior of
these species can be monitored in situ to understand
mechanisms of crystal growth and compound formation.
In this study we examined the K2Sn flux system due to its
demonstrated utility for low-temperature reactions with many
metals.12,28,33−35 The K−S phase diagram in Figure 1 reveals
ionic salts K2Sn, where 1 ≤ n ≤ 6.32 The shaded region indicates
compositions where a low-melting flux can be used as a
medium for crystal growth.
Figure 2. Unit cells of K2S3 (Cmc21, 7.31 × 9.91 × 7.47 Å) and K2S5
2−
(P212121, 6.49 × 6.60 × 17.41 Å) containing polysulfide chains of S3
and S5 respectively, shown here connected by bonds.
2−
distribution function (PDF). The PDF provides atom−atom
distances in real space, regardless of the disappearance of long-
range symmetry in the melt.45 We employed large-box reverse
Monte Carlo (RMC) simulation of the PDF, which is crucial
for investigating highly disordered materials.46−48 A chain-
building algorithm was developed to probe the formation of
2−
Sn polyanions. At present this technique cannot independ-
2−
ently resolve Sn chain lengths in molten K2S5, but it should
have utility for fluxes with smaller or more rigid polyanions.
EXPERIMENTAL PROCEDURE
■
Potassium polysulfides were prepared using a crucible-in-tube
technique. This technique avoids direct contact between K and S,
which can react violently, and is a safe, one-step alternative to synthesis
in liquid ammonia.32 Elemental K (Aldrich, 99.5%) was placed in the
bottom of an 18-mm-diameter quartz tube. Elemental S (Cerac,
99.999%) was placed in a 2-mL cylindrical Al2O3 crucible resting on
indentations approximately 2 cm above the K. The tube was flame-
sealed under vacuum while the K-containing end was submerged in
liquid nitrogen to prevent reaction during the sealing process. Tubes
were heated upright in a box furnace with a 12-h ramp to 300 °C, at
which point the S had vaporized and reacted with K to form a melt
(brown for K2S3, dark red for K2S5). The tube was removed from the
furnace while hot, then cracked open in a N2-filled glovebox for
collection of the product.
Figure 1. K−S phase diagram containing the low-melting compounds
K2S3 (Tm = 302 °C), K2S4 (Tm = 154 °C), which is glassy, K2S5 (Tm =
206 °C), and K2S6 (Tm = 183 °C). Adapted from Sangster;32 dotted
region is shown as an inset in the bottom pane. The shaded region
represents conditions where the low-melting flux may enable
controlled growth of new materials.
2−
The fluxes K2S3 and K2S5 crystallize to form chains of Sn
High-resolution synchrotron X-ray powder diffraction patterns were
collected using the mail-in configuration of beamline 11-BM at the
Advanced Photon Source (APS), running at 60 keV (λ = 0.414 Å),
with samples ground under N2, sieved to 90 μm, and sealed under N2
in 0.5-mm-diameter glass capillaries. Rietveld refinements were
conducted using the EXPGUI frontend for GSAS.49,50
X-ray total scattering was performed at beamline 11-ID-B at the
APS, using the sample heater described by Chupas,51 with samples in
0.7-mm-diameter quartz capillaries sealed under vacuum. Data were
collected at 90 keV, and a maximum momentum transfer was chosen
at Qmax = 20 Å−1 for Fourier transformation of the total scattering
function S(Q) to the PDF using the PDFgetX2 software.52 Least-
squares refinements were performed using PDFgui.53
Reverse Monte Carlo simulations were run using the RMCProfile
package54 running on single 2 GHz cores. Supercells of K2S5 were
created using 9×9×3 unit cells of the P212121 average structure, for a
total of 6804 atoms in a box of dimensions 58 × 59 × 52 Å. Hard-
sphere nearest-neighbor cutoffs were placed at 1.88, 2.70, and 3.30 Å
for S−S, K−S, and K−K, respectively, but no bunching was observed
at these distances. Atomic visualization presented here is output using
VESTA.55
polyanions. Unit cells of these two compounds are shown in
Figure 2. The distribution of these chains in the melt and their
response to the addition of reagents should be important
factors in the growth of new phases.
2−
Large distributions of Sn chain lengths were predicted in
molten Na2Sn to explain the electromotive force in Na−S
batteries, which is nearly linear versus n, but these were not
substantiated by any structural or spectroscopic techniques.36,37
In fact, neither conductivity nor density of Na2Sn varies
smoothly over the same range.38,39
We utilized in situ high-temperature Raman spectroscopy to
2−
probe Sn chain formation in molten K2Sn. Previous Raman
studies have only examined solid polysulfides40,41 and solutions
of Sn2− in liquid ammonia.40,42−44 Our Raman analysis does not
2−
support Na−S battery models which propose molten Sn to
exhibit a wide range of chain lengths in equilibrium. Instead we
find that the melt structure is highly dependent on
stoichiometry.
We investigated the real-space structure of K2S5 melts using
in situ high-temperature X-ray total scattering to obtain the pair
Raman spectroscopy was performed using a Renishaw inVia Raman
microscope, utilizing a 532-nm laser. High-temperature in situ spectra
9457
dx.doi.org/10.1021/ja303047e | J. Am. Chem. Soc. 2012, 134, 9456−9463

Journal of the American Chemical Society
Article
were recorded using a Linkam TMS600 heating stage. Samples were
sealed under vacuum in 0.5-mm-diameter glass capillaries and placed
flat on the heating stage with a heating rate of 20 °C/min. Preliminary
experiments were conducted to ensure that the laser power was
modest enough (5% power, 20 s exposures) to avoid significant local
heating of the sample. No changes in Raman spectra were observed
during 20 min spent at 400 °C.
RESULTS AND DISCUSSION
■
Structural Refinements. Phase purity of K2S3 and K2S5
synthesized using the crucible-in-tube technique was confirmed
using Rietveld refinements to powder X-ray diffraction data,
shown in Figure 3. The unit cells of K2S3 and K2S5 are shown in
Figure 4. Room-temperature Raman spectra of as-synthesized K2S3
(top), K2S5 (bottom) and mixtures of the two samples at varying
molar ratios, before in situ reaction. Dashed fits are weighted averages
of K2S3 and K2S5. Samples are labeled A−E to emphasize continuity
across all Raman data presented here.
< n < 6 at 300 °C.36 In this case, where a wide distribution of
2−
Sn species would be present in the melt, the Raman spectra
should be smoothly varying from molten K2S3 to K2S5, with all
modes present and in gradually shifting intensity.
If, however, liquid polysulfides of exact K2Sn stoichiometry
have a strong preference to form chains of only length n, the
Raman spectrum of each melt with integer n should be distinct,
just as they are for each of the solid phases.
To test the two models (large n distribution versus precise
chains of length n), powders of pure K2S3 and K2S5 were
ground and mixed at K2S3/K2S5 atomic ratios of 75−25, 50−50
(corresponding to an average compositions of K2S4), and 25−
75. The Raman spectra of the powder mixtures at room
temperature are shown in Figure 4. In all Raman spectra
presented here, each sample is designated A−E (in order of
increasing S content) to emphasize that they correspond to the
same samples at different stages of in situ reaction. Only modes
from K2S3 and K2S5 are seen at room temperature. Dashed lines
in Figure 4 are fits to linear combinations of K2S3 and K2S5 (cAA
+ cEE), where ci is the weighting and ∑ci = 1. Upon heating,
these mixtures melt and react.
Figure 3. Rietveld refinements to (a) K2S3 and (b) K2S5 synchrotron
powder diffraction data from APS beamline 11-BM, confirming the
phase purity of samples synthesized by a crucible-in-tube method.
Figure 2. Each consists of S32− or S52− chains, separated by lone
K+ ions. The phase diagram of Sangster32 (Figure 1) and our
high-temperature X-ray diffraction confirm that both these
phases melt congruently. This implies that, for compounds
K2Sn, where n ≤ 6, the Sn2− chains persist at least to the point of
melting.
In Situ Raman Spectroscopy. Raman spectroscopy is an
excellent tool for the study of molten salts because in situ
studies are accessible, and the work of Janz et al. has shown that
the measured room-temperature spectra of solid K2Sn (1 ≤ n ≤
6) are distinct.40 Lengthening of the Sn2− chains should lead to
salient differences between the chains in the melt.30 Our room-
temperature Raman spectra for K2S3 and K2S5 are shown in
Figure 4. These samples were confirmed to be phase-pure by
Rietveld refinement in Figure 3, and they agree with the
published spectra. Peaks in the 400−500 cm−1 range
At 400 °C, the range of compositions between K2S3 and K2S5
are all molten, and fits to the two models were conducted. In
2−
the case of a model with a broad distribution of Sn lengths,
the spectra of the intermediate 25−75, 50−50, and 75−25
compositions should be combinations of modes that are
present in the K2S3 and K2S5 end members (which would be
themselves mixtures of many Sn2− chain lengths). An attempt to
fit the experimental spectra using this model is shown in Figure
5, where least-squares fits to the data were made using a linear
combination of the experimental K2S3 and K2S5 spectra (cAA +
cEE). Spectrum C at 50−50 composition in particular is very
correspond to S−S stretching modes of the Sn chains.40
2−
When K−S and Na−S batteries were under heavy
investigation, attempts were made to use thermodynamic
knowledge of the systems to estimate Sn2− chain lengths in the
melt and thereby explain the electromotive force across Na−S
cells.56−58 Cleaver and Sime’s investigation of Na2Sn proposed a
broad distribution of Sn2− chain lengths, smoothly varying for 2
2−
poorly fit, and the model of a wide distribution of Sn chain
lengths is not supported by Raman spectroscopy.
Fits to the molten K2Sn Raman spectra using a model where
2−
Sn
chain lengths are tightly bound by the overall
9458
dx.doi.org/10.1021/ja303047e | J. Am. Chem. Soc. 2012, 134, 9456−9463

Journal of the American Chemical Society
Article
(cAA + cCC) and (cCC + cEE) and the agreements in Figure 6
are excellent.
2−
These findings imply that Sn chain lengths in molten K2Sn
are coupled to the composition and tuned by temperature.
Since this system is a medium for low-temperature crystal
growth, the considerations for directed synthesis are profound.
The variety of possible species in the melt, and factors
influencing their interconversion, may explain why many
metastable compounds can be stabilized from polychalcogenide
fluxes. The temperature dependence of these relations should
vary from Tm, where chain distribution is probably its tightest,
to high T, where polyanions must eventually break down.
Further Raman spectroscopy should be able to contribute to
this area, although signal markedly decreases with temperature.
After cooling to room temperature, the K2S3−K2S5 mixtures
display the Raman spectra in Figure 7, with spectra of the
Figure 5. Raman spectra of K2S3−K2S5 mixtures at T = 400 °C (solid
lines), fit using linear combinations (dashed lines) of the K2S3 and
K2S5 spectra. These fits of the type (cAA + cEE) do not reproduce the
spectra for the mixed compositions, unlike before these samples were
2−
subjected to heating (Figure 4), implying that the distribution of Sn
is not smoothly varying across the series.
stoichiometry (a liquid line compound model) are shown in
Figure 6. In this case, the 50−50 composition C must be
treated as a distinct K2S4 species itself (after all, the glass K2S4
appears as a congruently melting phase in the K−S phase
diagram in Figure 1). The 25−75 and 75−25 compositions B
and D should represent linear combinations of the K2S3−K2S4
and K2S4−K2S5 spectra, respectively. These fits can be written
Figure 7. Raman spectra of the polysulfides after cooling, showing
formation of K2S4, and mixtures with K2S3 and K2S5.
stoichiometric n = 3, 4, and 5 compounds all matching those
reported by Janz.40 The intermediate compositions B and D
correspond to mixtures of K2S3−K2S4 and K2S4−K2S5. In the
solid state, these are line compounds as prescribed by the phase
diagram in Figure 1. In the melt, this composition-structure
2−
relationship extends to the tight distributions of Sn chains.
Assignment of the exact polysulfide species responsible for
2−
modes in the molten Raman spectra is nontrivial because Sn
in the flux are very different from the isolated, gas-phase
molecules amenable to electronic structure calculations.31
However, local coordination can be probed by the PDF,
where a direct measurement of the real-space structure of melts
is possible.
In Situ X-ray Total Scattering and Local Structure. X-
ray total scattering data were collected up to 400 °C in order to
probe the local structure of molten K2S5. PDF techniques
provide an excellent complement to Raman since the
information is purely structural, not spectroscopic, while still
describing local interactions.59−61 The PDF of pure K2S5 at
room temperature in Figure 8 was fit using a least-squares
refinement to the P212121 average structure. This least-squares
Figure 6. Raman spectra (same data as Figure 5) of K2S3−K2S5
mixtures at T = 400 °C (solid lines), fit using linear combinations
(dashed lines) of the K2S3, K2S4, and K2S5 spectra. The data can be
reproduced by assuming that K2S4 has a distinct molten structure, as it
does in the solid phase.
9459
dx.doi.org/10.1021/ja303047e | J. Am. Chem. Soc. 2012, 134, 9456−9463

Journal of the American Chemical Society
Article
2−
agreement also precludes the existence of glassy K2S4 or K2S6
impurities.
contain enough information to probe S5 chains in molten
K2S5 is a difficult task since (i) the distance between the ends of
the chain is 5.16 Å at room temperature, and the PDF contains
many correlations of r < 5.16 Å (including interatomic
distances to neighboring chains which are not of interest)
that effectively comprise a high amount of noise and (ii) the
2−
Sn chains may exhibit rotations around S−S bonds and
stretch at each position, leading to a broad distribution of
possible distances between the ends of chains in the melt.
Since the K2S5 melt is devoid of long-range symmetry and
may contain an infinite variety of chain conformations, small-
box modeling such as the least-squares refinement of the K2S5
2−
unit cell in Figure 8 cannot faithfully model the Sn chain
lengths in the melt. Instead, we turned to RMC modeling of the
PDF, which is invaluable for the modeling of disordered
structures from total scattering data.46,48
In order to test the sensitivity of the molten K2S5 PDF data
2−
to Sn chain length, we prepared two disparate models: one
2−
where S are required to maintain S5 chains at all times
Figure 8. Room-temperature PDF least-squares refinement of K2S5,
showing a pure phase. The peak at r = 2 Å corresponds to S−S
bonding.
(strongly constrained, as our Raman suggests), and one where
randomly distributed S are driven by the data to segregate
2−
freely into Sn
.
2−
As K2S5 is heated past melting, loss of long-range structure
causes details in the PDF at higher r to disappear. This
progression is shown in Figure 9. The S−S nearest-neighbor
In the first model, the K2S5 unit cell (where all S are in S5
chains) was tiled to create a 9×9×3 supercell, and constraints
were applied during RMC simulation that require S5 to
2−
always remain intact and precisely of length n = 5. This was
accomplished by creating “distance window” constraints where
S chain-end atoms are allowed one and only one neighbor
closer than 2.31 Å, and chain-middle S atoms are restricted to
precisely two neighbors within 1.88−2.31 Å. The PDF of the
starting configuration (containing very sharp peaks since no
experimental broadening is applied) is shown in the top of
Figure 10a. After fitting this highly restrained model to the data
using RMC, the experimental PDF was reproduced very well, as
Figure 9. K2S5 PDF in the vicinity of melting, showing a loss of long-
range structure from 215 to 224 °C. The linear combination (LC,
dashed) of the 215 and 224 °C data reproduces the data from 219 °C,
showing that K2S5 is congruently melting. Difference curve of the 219
°C data minus LC is shown below. Data curves offset by 1 for clarity.
peak at r = 2.07 Å remains sharp, but farther correlations
outside r > 4 Å weaken. The PDFs above and below the
melting temperature can be averaged (best fit within 1% of even
weighting) to precisely reproduce the intermediate PDF at 219
°C in Figure 9. Only K2S5 is present, so we confirm that it is
congruently melting. For K2S5, Tm = 204 °C; this difference can
be attributed to the extra quartz tube encasing the sample and
overheating due to the 10 °C/min heating rate.
Figure 10. RMC fits to 224 °C PDF data starting from (a) the average
P212121 K2S5 structure (the undisturbed crystalline starting model
gives an overly sharp PDF) and (b) a random glass, as described in the
text. After RMC, both models fit the data equally well, despite S5
being constrained in (a) and unconstrained in (b).
In the context of our results from Raman spectroscopy, we
2−
wish to see whether S5 chains in the crystalline solid persist
2−
into the melt, and whether the PDF is a viable way of probing
distinct Sn2− species. Determining whether our measured PDFs
9460
dx.doi.org/10.1021/ja303047e | J. Am. Chem. Soc. 2012, 134, 9456−9463

Journal of the American Chemical Society
Article
shown in Figure 10a. This agreement implies that our Raman
findings can be supported by PDF data. However this rigid S5
model is not necessarily a unique solution−the unconstrained
model must be considered.
Figure 11c. After only 5 min of a 10-h simulation, the chain
2−
lengths have begun to fluctuate around the final average values.
2−
There is no trend toward formation of S5 in particular.
2−
Instead, most S are contained in S2 dimers. This trend of
The unconstrained model was created using a supercell of
the same dimensions, and populated by placing K and S atoms
completely at random, with the only constraint being the hard-
shell nearest-neighbor cutoffs common to all simulations
(specified in the Experimental Procedure). The result is a
glassy, albeit chemically unreasonable, supercell. The PDF of
the random model is shown in the top of Figure 10b. There are
no features in the starting model that resemble the S−S or K−S
correlations of r < 4 Å. This model was used as a starting point
for RMC simulation and the experimental PDF was
reproduced, as shown in the middle of Figure 10b. The S−S
peak at r = 2.07 Å is fit by RMC, so the increase in S−S
“locking-in” a disordered arrangement has been seen before in
RMC studies of amorphous materials,62 and techniques exist to
impose similar local environments on equivalent atoms during
the simulation.63 It remains to be seen whether polyanionic
fluxes might obey such constraints.
Because the S52−-constrained model and the random glass
model fit the data equally well but do not converge on a similar
answer, we can say that our experimental PDF requires
additional, independent information to definitively specify the
2−
length (or distribution of lengths) of Sn in the melt. In our
study, Raman spectroscopy fills this role. Nevertheless we
propose that this method of extracting chain lengths from RMC
modeling of total scattering data may prove valuable for
identifying other molten chemical species, especially if the
cluster of interest is small. Early total scattering studies found
signatures of Zintl polyanions in the A-Sn, and A-Te
systems64−66 (where A is an alkali metal), both of which are
viable fluxes for crystal growth.15,67,68 Constraints on the
number of distinct anion coordinations might be gleaned from
X-ray absorption spectroscopy or nuclear magnetic resonance,
although the latter is not straightforward for sulfides.69
Implications for Future Experiments. Equipment and
methods are available for in situ crystal growth experiments in
molten fluxes. High-temperature Bragg diffraction indicates flux
melting, crystal formation, and conversion between metastable
phases. If cations fully dissolve, additional probes of local
structure are necessary to monitor changes in the flux and
identify the solute complex. For these purposes PDF and
Raman analysis are complementary. Best for rigid species and
nanoscale precipitates, the PDF gives quantitative coordination
information and describes the real-space local structure and
extent of crystalline ordering.
2−
correlations should be commensurate with formation of Sn
chains in this melt. The question is whether these uncon-
2−
strained fits have converged to the same distribution of Sn
2−
chain lengths as the highly constrained S5 model.
2−
Statistical analysis of the Sn chain lengths (taken to be any
string of S atoms connected by <2.31 Å) in the glassy model
2−
after RMC fitting gives the histogram of Sn chains in Figure
11a. The finished model contains a large number of lone S
Raman spectroscopy is sensitive to solute complexation, and
we show here that changes in the flux itself are evident in the
vibrational spectra. If structural motifs from the flux
(polyanions) are incorporated into the product, it may be
possible to monitor their depletion from the melt. The changes
in spectra of molten K2S3 to K2S5 imply that Raman may serve
as a monitor of the A/Q ratio, which defines the basicity of the
flux and has been seen to define the crystalline product in many
systems.5,28
When reagents are introduced to any reactive flux, new
bonds appear and corresponding modes in the flux spectra
should be evident. A special case occurs if the reagents are
solvated before the formation of a long-range ordered phase. In
that critical regime it may be possible to guide the complexes
into new or metastable structures. We hope to examine these
scenarios in the near future.
2−
Figure 11. (a) Sn chain length histograms for the unconstrained
glassy model before and after RMC fitting to data. The differences
between the starting and ending histograms are given in (b). Fractions
of lone and dimerized sulfur atoms decrease while longer chains grow.
Results are averages of five independent simulations. (c) Distribution
of chain lengths versus fitting time, showing oscillation around their
final values after only 5 min, implying that these simulations will never
2−
converge to a model where long S5 chains are favored.
CONCLUSIONS
■
2−
atoms and S2 dimers, which is not thermodynamically
We have investigated potassium polysulfides at conditions
relevant to crystal growth in order to gain an understanding of
their behavior and to determine viable methods of probing the
reasonable.31 For comparison the Sn2− histogram of the starting
random model before RMC is shown.
The difference in the histogram of Sn chain lengths in the
2−
2−
stability and local structure of their constituent Sn species.
supercell after RMC fitting is shown in Figure 11b. Indeed, the
number of lone sulfur atoms and dimer S22− decreased while all
chains of n ≥ 3 became more prevalent. The approach to an
equilibrium distribution of chain lengths can be seen in the plot
K2S3 and K2S5 were prepared by a crucible-in-tube vapor
transport method. Raman spectroscopy was performed in situ at
400 °C for samples in the compositional range between K2S3
and K2S5. It was necessary to treat the K2S4 composition as
possessing molten species that are separate from those in K2S3
2−
of Sn chain lengths as a function of RMC fitting time in
9461
dx.doi.org/10.1021/ja303047e | J. Am. Chem. Soc. 2012, 134, 9456−9463

Journal of the American Chemical Society
Article
and K2S5. This implies that Sn2− species present in the melt are
strongly coupled for compositions n. This dependence of local
structure on stoichiometry may influence mechanisms of crystal
growth in these media. The temperature range over which this
relationship holds should be investigated. At 400 °C we do not
find a broad distribution of molten polysulfide chain lengths
(15) Chung, D.; Hogan, T.; Brazis, P.; Rocci-Lane, M.; Kannewurf,
C.; Bastea, M.; Uher, C.; Kanatzidis, M. G. Science 2000, 287, 1024−
1027.
(16) Kyratsi, T.; Chung, D.; Ireland, J. R.; Kannewurf, C. R.;
Kanatzidis, M. G. Chem. Mater. 2003, 15, 3035−3040.
(17) Sankar, C. R.; Bangarigadu-Sanasy, S.; Assoud, A.; Kleinke, H. J.
Mater. Chem. 2010, 20, 7485−7490.
(18) Sankar, C. R.; Bangarigadu-Sanasy, S.; Kleinke, H. J. Electron.
Mater. 2012, DOI: 10.1007/s11664-011-1846-z.
2−
where Sn are in equilibrium with Sn−12−, Sn+12−, etc., as was
originally proposed to explain the electromotive force of Na2Sn
batteries.36
(19) Lekse, J. W.; Moreau, M. A.; McNerny, K. L.; Yeon, J.;
Halasyamani, P. S.; Aitken, J. A. Inorg. Chem. 2009, 48, 7516−7518.
(20) Mei, D.; Lin, Z.; Bai, L.; Yao, J.; Fu, P.; Wu, Y. J. Solid State
Chem. 2010, 183, 1640−1644.
Analysis of the PDF obtained from X-ray total scattering
indicated that the chain distributions in K2S5 are not readily
visible. Reverse Monte Carlo fits to these data revealed that the
PDF of K2S5 requires additional information to distinguish the
species present. However, the chain-building techniques
presented here may be useful for determining the more rigid
species in melts or amorphous materials.
We have discussed the viability of PDF analysis and Raman
spectroscopy for future studies probing crystal growth in
molten fluxes, with the ultimate goal of understanding and
controlling synthetic reaction mechanisms.
(21) Mei, D.; Yin, W.; Feng, K.; Lin, Z.; Bai, L.; Yao, J.; Wu, Y. Inorg.
Chem. 2011, 51, 1035−1040.
(22) Manos, M. J.; Chrissafis, K.; Kanatzidis, M. G. J. Am. Chem. Soc.
2006, 128, 8875−8883.
(23) Manos, M.; Kanatzidis, M. Chem.Eur. J. 2009, 15, 4779−
4784.
(24) Axtell, E. A., III; Liao, J.; Pikramenou, Z.; Kanatzidis, M. G.
Chem.Eur. J. 1996, 2, 656−666.
(25) Androulakis, J.; Peter, S. C.; Li, H.; Malliakas, C. D.; Peters, J.
A.; Liu, Z.; Wessels, B. W.; Song, J.; Jin, H.; Freeman, A. J.; Kanatzidis,
M. G. Adv. Mater. 2011, 23, 4163−4167.
AUTHOR INFORMATION
Corresponding Author
m-kanatzidis@northwestern.edu
■
(26) Axtell, E. A., III; Park, Y.; Chondroudis, K.; Kanatzidis, M. G. J.
Am. Chem. Soc. 1998, 120, 124−136.
(27) Nguyen, S. L.; Jang, J. I.; Ketterson, J. B.; Kanatzidis, M. G.
Inorg. Chem. 2010, 49, 9098−9100.
Notes
The authors declare no competing financial interest.
(28) Durichen, P.; Bensch, W. Z. Naturforsch. B, Chem. Sci 2002, 57,
̈
1382−1386.
ACKNOWLEDGMENTS
■
(29) Berghof, V.; Sommerfeld, T.; Cederbaum, L. S. J. Phys. Chem. A
1998, 102, 5100−5105.
The authors thank Troy Lutes and the APS Detector Pool for
providing the heating stage for the Raman microscope. We also
thank Suntharalingam Skanthakumar and Richard E. Wilson
with assistance during data collection, and Kevin Beyer and
Melanie Francisco for experimental assistance and helpful
discussion. Argonne National Laboratory, a U.S. Department of
Energy Office of Science Laboratory, is operated by UChicago
Argonne, LLC, under contract No. DE-AC02-06CH211357.
(30) Rouquette, C.; Digne, M.; Renaudot, L.; Grandjean, J.;
Ballaguet, J. Energy Fuels 2009, 23, 4404−4412.
(31) Steudel, R. Top. Curr. Chem. 2003, 231, 127−152.
(32) Sangster, J.; Pelton, A. D. J. Phase Equilib. 1997, 18, 82−88.
(33) Chung, D.; Iordanidis, L.; Choi, K. S.; Kanatzidis, M. G. Bull.
Kor. Chem. Soc. 1998, 19, 1283−1293.
(34) Evenson; Dorhout, P. K. Inorg. Chem. 2001, 40, 2409−2414.
(35) Selby, H. D.; Chan, B. C.; Hess, R. F.; Abney, K. D.; Dorhout, P.
K. Inorg. Chem. 2005, 44, 6463−6469.
REFERENCES
■
(36) Cleaver, B.; Sime, S. Electrochim. Acta 1983, 28, 703−708.
(37) McKubre, M. C. H.; Tanzella, F. L.; Smedley, S. I. J. Electrochem.
Soc. 1989, 136, 303−305.
(1) (a) Kanatzidis, M. G. Chem. Mater. 1990, 2, 353−363. (b) Park,
Y. B.; Kanatzidis, M. G. Angew. Chem., Int. Ed. Engl. 1990, 29, 914−
915.
(38) Cleaver, B.; Davies, A.; Hames, M. Electrochim. Acta 1973, 18,
719−726.
(2) Kanatzidis, M. G.; Sutorik, A. C. Prog. Inorg. Chem. 1995, 43,
151−265.
(39) Cleaver, B.; Davies, A. Electrochim. Acta 1973, 18, 727−731.
(40) Janz, G. J.; Coutts, J. W.; Downey, J. R.; Roduner, E. Inorg.
Chem. 1976, 15, 1755−1759.
(3) Kanatzidis, M. G. Curr. Opin. Solid State Mater. Sci. 1997, 2, 139−
149.
(4) Stoll, P.; Nather, C.; Bensch, W. Z. Anorg. Allg. Chem. 2002, 628,
̈
(41) Janz, G. J.; Downey, J. R.; Roduner, E.; Wasilczyk, G. J.; Coutts,
J. W.; Eluard, A. Inorg. Chem. 1976, 15, 1759−1763.
(42) El Jaroudi, O.; Picquenard, E.; Demortier, A.; Lelieur, J.; Corset,
J. Inorg. Chem. 1999, 38, 2394−2401.
(43) El Jaroudi, O.; Picquenard, E.; Demortier, A.; Lelieur, J.; Corset,
J. Inorg. Chem. 2000, 39, 2593−2603.
(44) Pinon, V.; Lelieur, J. P. Inorg. Chem. 1991, 30, 2260−2264.
(45) Billinge, S. J. L.; Kanatzidis, M. G. Chem. Commun. 2004, 749−
760.
(46) Keen, D. A.; McGreevy, R. L. Nature 1990, 344, 423−425.
(47) McGreevy, R. L. Nuovo Cim. D 1990, 12, 685−701.
(48) McGreevy, R. L. Int. J. Mod. Phys. B 1993, 7, 2965−2980.
(49) Larson, A.; Von Dreele, R. Los Alamos National Lab. Rep. 2000,
86, 748.
(50) Toby, B. H. J. Appl. Crystallogr. 2001, 34, 210−213.
(51) Chupas, P. J.; Chapman, K. W.; Kurtz, C.; Hanson, J. C.; Lee, P.
L.; Grey, C. P. J. Appl. Crystallogr. 2008, 41, 822−824.
(52) Qiu, X.; Thompson, J. W.; Billinge, S. J. L. J. Appl. Crystallogr.
2004, 37, 678−678.
2489−2494.
(5) Palchik, O.; Marking, G. M.; Kanatzidis, M. G. Inorg. Chem. 2005,
44, 4151−4153.
(6) Deng, B.; Chan, G. H.; Huang, F. Q.; Gray, D. L.; Ellis, D. E.;
Van Duyne, R. P.; Ibers, J. A. J. Solid State Chem. 2007, 180, 759−764.
(7) Wu, Y.; Bensch, W. Inorg. Chem. 2009, 48, 2729−2731.
(8) Graf, C.; Assoud, A.; Mayasree, O.; Kleinke, H. Molecules 2009,
14, 3115−3131.
(9) Wu, Y.; Bensch, W. Z. Naturforsch. A 2010, 65b, 1219−1228.
(10) Chung, I.; Biswas, K.; Song, J.; Androulakis, J.; Chondroudis, K.;
Paraskevopoulos, K. M.; Freeman, A. J.; Kanatzidis, M. G. Angew.
Chem., Int. Ed. 2011, 50, 8834−8838.
(11) Wu, Y.; Bensch, W. J. Alloys Compd. 2011, 509, 4452−4456.
(12) Liao, J. H.; Kanatzidis, M. G. Chem. Mater. 1993, 5, 1561−1569.
(13) Bera, T. K.; Song, J.; Freeman, A.; Jang, J.; Ketterson, J.;
Kanatzidis, M. Angew. Chem. 2008, 120, 7946−7950.
(14) Bera, T. K.; Jang, J. I.; Song, J.; Malliakas, C. D.; Freeman, A. J.;
Ketterson, J. B.; Kanatzidis, M. G. J. Am. Chem. Soc. 2010, 132, 3484−
3495.
9462
dx.doi.org/10.1021/ja303047e | J. Am. Chem. Soc. 2012, 134, 9456−9463

Journal of the American Chemical Society
Article
(53) Farrow, C. L.; Juhas, P.; Liu, J. W.; Bryndin, D.; Bozin, E. S.;
Bloch, J.; Proffen, T.; Billinge, S. J. L. J. Phys. Cond. Mat. 2007, 19,
335219.
(54) Tucker, M. G.; Keen, D. A.; Dove, M. T.; Goodwin, A. L.; Hui,
Q. J. Phys. Cond. Mat. 2007, 19, 335218.
(55) Momma, K.; Izumi, F. J. Appl. Crystallogr. 2008, 41, 653−658.
(56) Crosbie, G. M. J. Electrochem. Soc. 1982, 129, 2707−2711.
(57) Lindberg, D.; Backman, R.; Hupa, M.; Chartrand, P. J. Chem.
Thermodyn. 2006, 38, 900−915.
(58) Morachevskii, A. G.; Maiorova, E. A. Russ. J. Appl. Chem. 2002,
75, 1580−1584.
(59) Goodwin, A. L.; Tucker, M. G.; Cope, E. R.; Dove, M. T.; Keen,
D. A. Phys. Rev. B 2005, 72, 214304.
(60) Shoemaker, D. P.; Li, J.; Seshadri, R. J. Am. Chem. Soc. 2009,
131, 11450−11457.
(61) Shoemaker, D. P.; Seshadri, R.; Hector, A. L.; Llobet, A.;
Proffen, T.; Fennie, C. J. Phys. Rev. B 2010, 81, 144113.
(62) Gereben, O.; Pusztai, L. Phys. Rev. B 1994, 50, 14136−14143.
(63) Cliffe, M. J.; Dove, M. T.; Drabold, D. A.; Goodwin, A. L. Phys.
Rev. Lett. 2010, 104, 125501.
(64) Fortner, J.; Saboungi, M.; Enderby, J. E. Phys. Rev. Lett. 1992,
69, 1415−1418.
(65) Reijers, H. T. J.; Saboungi, M.; Price, D. L.; van der Lugt, W.
Phys. Rev. B 1990, 41, 5661−5666.
(66) Saboungi, M.; Fortner, J.; Richardson, J.; Petric, A.; Doyle, M.;
Enderby, J. J. Non-Cryst. Solids 1993, 156−158, 356−361.
(67) Hahn, S. E.; Lee, Y.; Ni, N.; Canfield, P. C.; Goldman, A. I.;
McQueeney, R. J.; Harmon, B. N.; Alatas, A.; Leu, B. M.; Alp, E. E.;
Chung, D. Y.; Todorov, I. S.; Kanatzidis, M. G. Phys. Rev. B 2009, 79,
220511.
(68) Johnsen, S.; Peter, S. C.; Nguyen, S. L.; Song, J.; Jin, H.;
Freeman, A. J.; Kanatzidis, M. G. Chem. Mater. 2011, 23, 4375−4383.
(69) Jakobsen, H. J.; Bildsøe, H.; Skibsted, J.; Brorson, M.; Srinivasan,
B. R.; Nather, C.; Bensch, W. Phys. Chem. Chem. Phys. 2009, 11,
̈
6981−6986.
9463
dx.doi.org/10.1021/ja303047e | J. Am. Chem. Soc. 2012, 134, 9456−9463