In situ studies of a platform for metastable inorganiccrystal growth and materials discoveryDaniel P. Shoemakera,1, Yung-Jin Hub, Duck Young Chunga, Gregory J. Halderc, Peter J. Chupasc, L. Soderholmb,J. F. Mitchella, and Mercouri G. Kanatzidisa,d,2

aMaterials Science Division, bChemical Sciences and Engineering Division, and cX-ray Science Division, Argonne National Laboratory, Argonne, IL 60439;and dDepartment of Chemistry, Northwestern University, Evanston, IL 60208

Edited by Zachary Fisk, University of California, Irvine, CA, and approved June 17, 2014 (received for review April 10, 2014)

Rapid shifts in the energy, technological, and environmental de-mands of materials science call for focused and efficient expan-sion of the library of functional inorganic compounds. To achievethe requisite efficiency, we need a materials discovery and op-timization paradigm that can rapidly reveal all possible com-pounds for a given reaction and composition space. Here weprovide such a paradigm via in situ X-ray diffraction measurementsspanning solid, liquid flux, and recrystallization processes. Weidentify four new ternary sulfides from reactive salt fluxes in amatter of hours, simultaneously revealing routes for ex situ syn-thesis and crystal growth. Changing the flux chemistry, here ac-complished by increasing sulfur content, permits comparison ofthe allowable crystalline building blocks in each reaction space.The speed and structural information inherent to this method ofin situ synthesis provide an experimental complement to compu-tational efforts to predict new compounds and uncover routes totargeted materials by design.

Discovering new materials is a crucial step to address large-scale problems of energy conversion, storage, and trans-

mission and other technological needs whether seeking bulkphases or thin films. Dense inorganic materials are desired fortheir tunable transport, magnetism, optical absorption, and sta-bility, but their existence in general cannot be predicted with thenear certainty of that of metastable organic and organometalliccompounds. Whereas the desire to efficiently locate and as-semble inorganic materials is great, it is hindered by traditionalsolid-state synthetic methods—at high temperatures often only theenergy-minimum thermodynamic product is obtained. To strivetoward an arena where metastable compounds can be discoveredrapidly and made systematically, here we conduct reactions withinliquid fluxes and use in situ monitoring to capture signatures ofnew phases, even when they quickly dissolve in the melt.Convective liquid fluxes (salts, metals, or oxides) can serve as

reaction media that aid diffusion and enable rapid formation ofcompounds at temperatures far below their melting points (1–6).The flux can be nonreactive or reactive; in the latter case the fluxitself becomes incorporated into the product (7, 8). This well-established approach has demonstrated the prolific discovery ofnovel inorganic materials grown out of low-melting fluxes, fromoxides and other chalcogenides (9–12), to pnictides (13, 14), tointermetallics (15), many of which cannot be attained by directcombinations of the elements. Despite the variety of metastablephases formed in these reactions, the classical approach is topredetermine a given set of reaction conditions (e.g., time,temperature, and heating and cooling rates) and wait for com-pletion to isolate and identify the formed compounds. It is notpossible to observe how the reaction system itself has arrived atthe isolated compound, whether the crystalline material formedon heating, on cooling, or on soaking at the given high temper-ature, nor it is possible to know whether any intermediates werepresent and, if so, their influence on product formation. This lackof awareness (“blind synthesis”) hinders our ability to identifythe new materials or to devise successful synthetic processes fordesired and targeted materials. If we are to develop a predictive

understanding of synthesis and to more quickly discover newmaterials, we will greatly benefit from the input of much higherlevels of detail in how syntheses proceed.We show here that in situ synchrotron X-ray diffraction maps

of metastable inorganic compound formation in inorganic fluxesreveal complex real-time phase relationships and permit rapidaccess to new inorganic materials that would be missed usingclassical approaches. Specifically, we have discovered heretoforeunknown phases in systems with simple elemental compositionsof Cu and Sn with molten polysulfide salts K2S3 and K2S5(melting points 302 °C and 206 °C, respectively) as paradigmaticrepresentatives. Complex copper sulfides have been identified aspossible earth-abundant photovoltaics (16) and are a source ofexotic charge-density-wave materials, whereas tin chalcogenidesform the basis for Cu2ZnSnS4 (CZTS) semiconductors (17) andexhibit ion exchange properties useful for heavy metal wastecapture (18). We observe a complex phase space: In all but oneof our reactions we observe additional crystalline phases in situthat are not present by the end of the reaction (as would berecovered ex situ). Both families of ternary compounds canexhibit a variety of coordinations by sulfur, and the range of prop-erties is accordingly large: In just one of our Cu-containing reac-tions we found a previously undiscovered 1D metal (K3Cu4S4)similar to heavy-metal capture materials and observed a differ-ent 2D metal (K3Cu8S6) and a layered semiconductor (KCu3S2).Our approach can be combined with previous work that

probes the formation and stabilization of inorganic materialsfrom liquid media, using structural and spectroscopic data (19,20). Within a given reaction, we can use temperature as a vari-able to probe the relationships among phases with the goal of

Significance

Dense inorganic materials comprise most functional electronic,optical, and magnetic devices. Whereas the discovery of newinorganic materials can increase our technical capabilities anduncover new phenomena, the search is difficult due to theirformation at high temperatures where only the most stable(often known) materials can be isolated postreaction. We finda variety of unexpected and unknown materials nucleating atmoderate temperatures in molten salts. By probing these pro-cesses with in situ diffraction, we are able to identify a largevariety of new phases quickly and pave a path to more effi-cient materials discovery.

Author contributions: D.P.S., D.Y.C., P.J.C., L.S., J.F.M., and M.G.K. designed research; D.P.S.and Y.-J.H. performed research; G.J.H. contributed new analytic tools; D.P.S. analyzeddata; and D.P.S., L.S., J.F.M., and M.G.K. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.1Present address: Materials Science and Engineering Department, University of Illinois atUrbana–Champaign, Urbana, IL 61801.

2To whom correspondence should be addressed. E-mail: m-kanatzidis@northwestern.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1406211111/-/DCSupplemental.

10922–10927 | PNAS | July 29, 2014 | vol. 111 | no. 30 www.pnas.org/cgi/doi/10.1073/pnas.1406211111

Dow

nloa

ded

by g

uest

on

Mar

ch 3

0, 2

020

understanding how the structures may be constructed fromsimilar building units. Additionally, in situ studies of crystalli-zation have been shown to be effective at probing the kinetics ofoxide and sulfide formation, typically from aqueous solutions(21–23).The in situ technique we describe here provides rapid dif-

fraction signatures of all crystalline phases formed duringprocessing, including metastable intermediates. Reactions areperformed in under 2 h (compared with typical flux reactions onthe order of days), but time resolution of the data can be on theorder of seconds. This is a route to efficient exploration ofcomposition–temperature spaces that might be predicted tohouse novel compounds with favorable properties by first-prin-ciples (24) or data-mining calculations (25). Because new struc-ture types cannot be easily predicted by computational theory,the approach described here is a necessary experimental tool innow-budding efforts to understand, validate, and expand thematerials genome.A schematic of the in situ capillary furnace used to investigate

phase formation during flux reactions is shown in Fig. 1A. Thesample tubes, 0.7 mm in diameter, were sealed under vacuum,heated using a resistive coil, and continuously rastered throughthe synchrotron X-ray beam to maintain uninterrupted X-rayexposure of the sample as it melted and flowed within the tube.For each reaction, the accompanying series of raw diffraction

patterns provide a real-time monitor of the reaction progressand, to a first approximation, the number of phase formation anddissolution events. The X-ray diffraction pattern from a combi-nation of Cu and K2S3 before heating is shown in Fig. 1B andfitted with a Rietveld refinement that confirms the contributionsfrom both reagents. The diffraction patterns collected continu-ously while heating and cooling this reaction mixture are shownin Fig. 1C. Wholesale changes in the diffraction patterns indicatethat the reaction pathway proceeded in a series of steps (colorcoded in Fig. 1C). First, signatures of the reagent metal and

polysulfide appeared (blue region), which all resemble the re-fined pattern in Fig. 1B. Upon heating, low-Q peaks appearedin the diffraction data (red region). This real-time information(before any analysis) revealed that more complex ternary K–Cu–Sphases were forming.Continued heating led to the disappearance of all Bragg peaks

(gray region in Fig. 1C). At this point the ternary sulfides dis-solved completely into the molten polysulfide salt flux. Aftercooling, low-Q peaks again signified the presence of ternaryphases (violet region in Fig. 1C). Random variations in peakheights in adjacent patterns provided a morphological clue: Theproduct on cooling formed as large sulfide crystals, leading tosharp diffraction spots rather than powder-averaged rings.Systematic least-squares refinements account for all crystalline

phases present. Phase transformations and amorphous regions(defined by the absence of Bragg peaks) can also be observed inreal time by visual inspection. Representative diffraction pat-terns and their least-squares refinements from each of the fourtemperature regimes in Fig. 1C are shown in Fig. 1 D–F. Data inFig. 1D were taken at 320 °C upon heating and the reactioncontained two ternary phases: KCu3S2 and the previously un-known phase K3Cu4S4.At the time of the experiment, contributions to the diffraction

pattern that do not match known compounds are consideredunknown phases to be solved. Three options to identify newphases are as follows:

i) Deduce structure by chemical analogy to known compounds.This can be accomplished manually or aided by computa-tional tools such as the “Structure Predictor” of the Materi-als Project (26, 27).

ii) Grow larger single crystals of compounds that crystallize outof a melt. In the present case, elemental analysis and single-crystal diffraction were used in this study to identify K4Sn2S6

Fig. 1. Schematic of in situ X-ray diffraction experiment and data analysis. (A) The in situ setup for collecting X-ray diffraction data from melting polysulfidefluxes. The sample tube diameter is 0.7 mm. Two-dimensional patterns are integrated and the 1D data can be fitted using Rietveld refinements to giveindividual contributions to the diffraction patterns in B. Upon heating, diffraction data are collected continuously, giving the set of patterns in C. These dataare divided into regions with distinct phase contributions. A single pattern from each region is shown in D–F, with contributions to each pattern identified.

Shoemaker et al. PNAS | July 29, 2014 | vol. 111 | no. 30 | 10923

CHEM

ISTR

Y

Dow

nloa

ded

by g

uest

on

Mar

ch 3

0, 2

020

from crystals grown ex situ based on information provided bythe in situ experiments.

iii) Solve structure from powder diffraction, using computa-tional packages such as FOX (28).

In the case of the experiment with data shown in Fig. 1D, theunknown phase K3Cu4S4 was identified by option i: comparisonof the isostructural phase Na3Cu4S4.The diffraction pattern at 600 °C in Fig. 1E, taken from the

gray region in Fig. 1B, has no Bragg peaks and represents theamorphous polysulfide melt after dissolution of KCu3S2 andK3Cu4S4. A diffraction pattern from cooling at 50 °C is shown inFig. 1F to contain the 2D metallic compound K3Cu8S6 and therecrystallized flux K2S3. Because K3Cu8S6 grew as large crystals,not a powder, Le Bail full-profile fitting (disregarding peakheights but refining peak profiles and peak locations constrainedby unit cell symmetry) is shown in Fig. 1F to account for all peakcontributions rather than the Rietveld method (29).The last step in diffraction analysis is to refine every powder

pattern sequentially. Using the four diffraction patterns in Fig. 1D–F and their refined phase contributions as anchors, we per-formed automated least-squares refinements to the ∼200 dif-fraction patterns to record the reaction progression as a functionof time and temperature and form a “reaction map.” In thismanner we identified phase fractions and points of crystalliza-tion, melting, and dissolution of all crystalline phases presentduring in situ reactions of Cu and Sn in fluxes of K2S3 and K2S5.

The resulting panoramic reaction maps are shown in Fig. 2 anddiscussed subsequently.

Hidden Compounds upon Heating Cu in K2S3Two in situ reactions of Cu in K2S3 (1:1 and 1:5 metal:flux moleratio) were studied and the time- and temperature-dependentprogression of phases is shown in Fig. 2 A and B. The melting ofthe flux and dissolution of Cu are readily seen in the earlyminutes of heating. The first formed product was the knownphase KCu3S2, a valence-precise semiconductor (30), whichcould be isolated if the reaction were stopped at this point. Soonafter, the previously unknown phase K3Cu4S4 appeared. Thestructures of these phases are closely related and shown inFig. 3: K3Cu4S4 contains 1D chains of CuS trigonal pyramids.In KCu3S2 these chains are linked by tetrahedral CuS4 units toform a 2D network. The shared structural motifs suggesta common precursor may exist in the melt, and the implicationfor directing synthesis in regime can be probed by other in situstudies (19, 20). Na3Cu4S4 accommodates extra charge on Cu–Schains but curiously does not exhibit charge-density-wavemodulations (31). It is not yet known whether K3Cu4S4 behavessimilarly. K3Cu4S4 dissolves into the flux at 410 °C, followed byKCu3S2 at 490 °C.The power of in situ structural characterization to identify

intermediate compounds and events is clear in these systemsbecause both reactions of Cu in K2S3 produced different phases

Fig. 2. Reaction maps obtained from sequential refinements to in situ diffraction data. (A) Reactions of Cu + K2S3 lead to formation of KCu3S2 and the newphase K3Cu4S4 on heating. These dissolve into the melt upon heating, and K3Cu8S6 crystallizes upon cooling. A similar phase progression occurs for Cu + 5K2S3(B), with KCu3S2 also forming on cooling. Reactions of Cu + K2S5 (C) and Cu + 5K2S5 (D) produce the layered phase KCu4S3. Reactions of Sn + K2S3 (E) produceSnS and the new phase K6Sn2S7 on heating. No ternary phases are observed for Sn + 5K2S3 (F). The lowest-melting reaction Sn + K2S5 (G) reveals formation ofthe new phase K5Sn2S8 from a melt upon heating. K4Sn2S6 and K2Sn2S5 crystallize concurrently upon cooling. For Sn + 5K2S5 (H), the progression is similar butK2Sn2S5 is absent.

10924 | www.pnas.org/cgi/doi/10.1073/pnas.1406211111 Shoemaker et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 3

0, 2

020

on heating vs. cooling, indicating metastable compounds. Cool-ing from 600 °C led to nucleation of the known 2D metalK3Cu8S6 (32). This phase was accompanied at 426 °C by KCu3S2when additional flux was used (starting composition Cu +5K2S3), Fig. 2B. The two reactions of Cu in K2S3 produced onepreviously unknown ternary phase in less than 4 h of reactiontime. If this particular reaction procedure of Cu + K2S3 wereperformed ex situ, no evidence of the formation of KCu3S2 andK3Cu4S4 would exist—only K3Cu8S6 would remain. Furthercontrol over these reactions requires a grasp of reaction kinetics,either by diffraction or calorimetry, but our approach can pro-vide clues about the (meta)stability of the competing phases. Forinstance, a separate in situ reaction showed that holding thereaction Cu + K2S3 at 320 °C led to disappearance of K3Cu4S4within minutes whereas KCu3S2 persisted unchanged for 10 huntil the collection was ended. Increasing the flux ratio to Cu +5K2S3 and holding at 320 °C led to dissolution of KCu3S2 backinto the melt. Armed with a structural description of these phasetransitions, simpler calorimetric techniques (differential thermalanalysis) would be a complementary method to probe the energydependence of these reactions and how they can be tuned viatemperature and chemistry.

Distinct Double-Layer Sulfides: Cu in K2S5Quickly establishing phase stability boundaries is a key aspect ofeffective phase discovery. A change in flux chemistry to K2S5(Fig. 2 C and D) led to the gradual formation of a separate 2Dmetallic compound KCu4S3, even before K2S5 had fully melted.The phase KCu4S3 bears little structural relationship to KCu3S2and K3Cu4S4 from reactions in K2S3. It contains corner-sharingCuS4 tetrahedra, reminiscent of a double-layer version of theThCr2Si2-type structure that exhibits superconductivity in alkali–iron–pnictide compounds (33). KCu4S3 is mixed valence andexhibits holes in the S p-band, but modification/doping reactionsto induce a superconducting state have yet to be demonstrated(34). A small amount of digenite Cu2-δS is formed at highertemperatures. Otherwise the reaction mixture is amorphous afterKCu4S3 dissolves at 495 °C.Upon cooling, reactions with metal:flux ratio 1:1 crystallize

KCu4S3 congruently, indicating that this reaction stoichiometry

is a viable medium for growing single crystals of KCu4S3. In-creasing the flux ratio to 1:5 (Fig. 2D) did not change the phaseprogression upon heating, but did suppress crystallization ofKCu4S3 and led to a mixed K–Cu–S glass after cooling at thesame rate of 15 °C/min, which is relatively fast for an inorganicsynthesis.

Implications for Directed Ternary Copper Sulfide SynthesisThis series of four reactions (∼8 h) produced a total of fourternary phases, one of which is new. We also observed that K2S3and K2S5 fluxes produced different sets of phases. In essence,K2S5 contains more S–S bonding (average valence only S0.4−).Note that Cu can be oxidized only to 1+ (35). The only semi-conducting phase (KCu3S2 with nominal S2− valence) was formed inK2S3 fluxes only. Metallic phases K3Cu8S6 and K3Cu4S4 that sharethe same three-coordinate Cu-S motifs (with subtle modifications)were also observed alongside KCu3S2. Using a K2S5 flux reducedthe oxidizing power of the flux, eliminated three-coordinate Cu+,and produced metallic KCu4S3, which has a less-negative nomi-nal S valence (S1.67−) than any other phase seen here. The lack ofoverlap (no other ternary phases existing alongside KCu4S3)implies that there is an intermediate reaction space betweenK2S3 and K2S5 that remains ripe for discovery: It may containcoexistence of these ternary phases or entirely new compounds.These maps may also imply that there is a change in character ofCu–S coordination in the melt between K2S3 and K2S5 thatdivides reactions forming KCu4S3 and those forming three-co-ordinate ternary copper sulfides.

New Dimer Compounds: Sn in K2S3In situ reactions of Sn in K2S3 revealed two new phases in ourfirst 2-h study. These reactions are shown in Fig. 2 E and F andproceeded with the melting of Sn first and then formation ofα-SnS before the flux K2S3 had finished melting at 302 °C. Aftermelting the flux, α-SnS was converted into the previously un-known phases K6Sn2S7 and K4Sn2S6, which are shown in Fig. 3.The structure of K6Sn2S7 was solved based on chemical analogyto the known valence-precise compound Rb6Sn2S7, which con-tains corner-sharing dimers of SnS4 tetrahedra (36).Isolation and structure solution of K4Sn2S6 was straightfor-

ward because rapid in situ reactions of Sn in K2S5 revealed that itcrystallized congruently from the melt. In this case, the reactionmaps provided an immediate “recipe” for growth of clear singlecrystals. K4Sn2S6 contains SnS4 tetrahedra that are edge sharingto form [Sn2S6]

4− units. Both structures are shown in Fig. 3.Continued heating led to disappearance of K6Sn2S7 until onlyK4Sn2S6 remained.The reaction Sn + K2S3 → 1/2 K4Sn2S6 is stoichiometric, so

no liquid flux is necessarily present at 600 °C. However, the chaincompound K2Sn2S5 (structure is shown in Fig. 3; it contains a 3Dnetwork of corner- and edge-sharing SnS5 trigonal bipyramids)formed upon cooling without any decline in the amount ofK4Sn2S6. The presence of SnS and K6Sn2S7 upon heating sug-gests that these phases left behind amorphous K-poor or S-richregions, either of which could lead to formation of K2Sn2S5.Again, kinetic studies may reveal routes to full conversion toK4Sn2S6 in this system.Increasing the flux ratio to Sn + 5K2S3 eliminated the formation

of any ternary compounds in these reactions. Only a small amountof α-SnS was formed on heating, which dissolved into the flux,leaving a K–Sn–S glass after cooling at 15 °C/min, analogous to thecooling behavior of the reaction Cu + 5K2S5 in Fig. 2D.

New Structure Type from a Melt upon Heating: Sn in K2S5Reactions of Sn in K2S5 were the only reactions in this studywhere there was a small induction time after melting of the fluxand Sn (Tm = 232 °C): No crystalline phases were present fora short period upon heating. This “dead time” after 60 min have

Fig. 3. Ternary crystal structures observed from in situ synthesis experi-ments. (Upper) Crystal structures of known phases KCu3S2, K3Cu8S6, andKCu4S3 are shown, along with the sole previously reported Sn-containingstructure K2Sn2S6. (Lower) Structures of four previously unknown phasesthat were discovered: K3Cu4S4, K6Sn2S7, K4Sn2S6, and K5Sn2S8. (Cu, blue; Sn,gray; S, yellow; K, white).

Shoemaker et al. PNAS | July 29, 2014 | vol. 111 | no. 30 | 10925

CHEM

ISTR

Y

Dow

nloa

ded

by g

uest

on

Mar

ch 3

0, 2

020

elapsed is shown in Fig. 2 E and F. This metastable inductionregion has not been observed before in molten salt synthesisand is surprising given the preference of the Cu-containingreactions to form ternary compounds immediately as the fluxmelted. Ex situ or calorimetric measurements alone cannotquickly differentiate between melting events that produce newphases vs. amorphous induction periods—they must be informedby structural methods.The new phase K5Sn2S8 crystallized directly from the amor-

phous induction region upon heating and its structure was solvedfrom high-resolution powder diffraction. Future studies shouldprobe the generality of this formation mechanism. If such be-havior exists in many other chemical syntheses, then it wouldpermit single-crystal growth at very low temperatures and thereforemay access a different cache of phases than those formed fromcooling at high temperatures.The 2D compound K5Sn2S8 is shown in Fig. 3 and contains

infinite zig-zag chains of corner-sharing SnS4 tetrahedra, similarto Na2GeS3, but with extra space between the chains and ac-commodating one S4

2− polysulfide chain, effectively creatinga phase that is K4Sn2S6 + 1/2K2S4 (albeit topologically dissimilarto K4Sn2S6). Continued heating led to formation of K2Sn2S5 andK4Sn2S6. At 430 °C all phases had dissolved, leaving an amor-phous melt that persisted until concurrent crystallization ofK4Sn2S6 and K2Sn2S5 upon cooling past 460 °C at 15 °C/min.Increasing the flux ratio to Sn + 5K2S5 (Fig. 2F) led to a re-action with an induction time before crystallizing K5Sn2S8 andK4Sn2S6. K2Sn2S5 was absent on heating and cooling, so thisreaction medium was used in a large ex situ reaction to prepareoptically clear single crystals of K4Sn2S6 for structural solution.Unlike the ternary copper sulfides, the K–Sn–S phases are

mostly charge-balanced insulators with nominal S2−. The oneexception is the new metastable phase K5Sn2S8. This phase haspolysulfide units incorporated into the structure alongside theSnS4 tetrahedra that persist in all other ternary phases we ob-served. This unique phase forms only after an induction timeon heating, so further kinetic and structural studies may revealnovel strategies to prepare tin sulfides with large open chan-nels or a disproportionation of S5

2− chains in the melt that areincorporated into the crystal.

Implications for Materials Synthesis and DiscoveryIn situ exploratory reactions in inorganic reactive fluxes showthat a variety of new phases are waiting to be found even insystems that are well investigated. The method of collectingdiffraction data in transmission over the full length of a liquid-filled capillary is a powerful tool for materials discovery in itsown right, and these experiments can provide valuable insightwhenever inorganic or metastable materials are being synthe-sized. A single in situ reaction can inform bulk reactions toprepare large crystals, as was done here for K4Sn2S6, especiallywhen the phase is of technical interest or cannot be solved bypowder diffraction patterns alone. The eight reaction spaces weshow in Fig. 2 represent less than 24 h of diffraction time. Ac-cordingly, we present only a limited composition space, becausemany other fluxes (K2Sx) and ratios could be explored. Thismatrix of reactions is large, but could be performed rapidly usingthe methods we describe.Knowledge about the temperatures at which different phases

form (or disappear), when combined with the atomic structuresof the resulting compounds, provides a unique opportunity tobegin to understand the underlying mechanism by which thesematerials are assembled. For example, the new K3Cu4S4 struc-ture is composed of simple Cu4S4 one-dimensional linkages thatare joined together to form the more complex 2D Cu–S networkseen in KCu3S2, raising the possibility that the two phases areclosely linked energetically and hence could be tuned to favorone product over another. Increasing the S content leads

to formation of KCu4S3, using entirely different CuS4 motifs.Experiments using our in situ technique, when combined withtheory, can approach the directed synthesis of targeted materials.Our observation of ternary phases forming upon heating

through an amorphous induction zone in low-melting Sn + K2S5reactions highlights the novel formation mechanisms of meta-stable phases that remain to be uncovered and understood. Be-cause these phases form from solution, studying the nature of themolecular entities that exist in the melt before crystallization willbe a major step toward controlling the formation of inorganiccrystals and designing new materials. A full array of localstructure techniques will be necessary to probe these processes:In situ Raman and X-ray spectroscopy, real-space high-energyX-ray scattering analyses, NMR, and molecular dynamics simu-lations can all provide structural and mechanistic guidance,whereas calorimetric studies on the same timescale as our in situwork can inform the kinetics.The rapid interrogation of temperature–composition space is

a natural complement to computational efforts to acceleratematerials discovery and design, such as the Materials GenomeInitiative. New structure types such as K5Sn2S8 found herecannot be easily predicted by theory and have the potential togreatly expand the library of known materials. Our methodallows unfamiliar diffraction patterns to be quickly comparedwith those of predicted compounds, such as those freelyavailable in the Materials Project (25, 26). This approachshould be integrated into arenas where the discovery of com-plex phases can expand the reservoir of functional materialsthat dictate the state of the art across energy and other ap-plications spaces.

MethodsPrecursor polysulfides K2S3 and K2S5 were prepared by a tube-in-tube vaportransport technique: Balls of metallic K were loaded into a quartz tube, andan alumina crucible filled with a stoichiometric amount of S was suspendedabove. This tube was sealed under vacuum with the K-containing end sub-merged in liquid N2 to prevent vaporization. Care must be taken to preventspilling the suspended S onto the K below. The sealed tubes were heated to300 °C in 12 h, held at that temperature for 4 h, and then air quenched.Ingots of K2S3 and K2S5 powders were ground and sieved to 150 μm andthen ground together with stoichiometric amounts of Cu and Sn inside anN2 glove box. These mixed powders were loaded into 0.7-mm diameterquartz tubes and sealed under vacuum.

In situ X-ray diffraction was performed at beamline 17-BM of the Ad-vanced Photon Source (APS), using X-rays of wavelength λ = 0.605 Å (20 keV),using a beam size ∼0.3 mm2 and a Perkin-Elmer 2D detector and the sampleheating environment developed by Chupas et al. (37). The experimentalconfiguration is shown in Fig. 1A. Diffraction data were collected continu-ously with 30-s collection times, and the sample assembly was continuouslyhorizontally rastered through the beam so that each powder pattern inte-grates intensity over the powder-filled portion of the capillary. All sampleswere heated to 600 °C with a slowdown near the melting temperature ofthe polysulfide salt flux given in SI Text.

High-resolution powder diffraction data were collected on selectedproducts at beamline 11-BM of the APS (λ = 0.414 Å, 40 keV). Rietveldrefinements were performed using the EXPGUI front end to GSAS (38),whereas structure solution from powder diffraction data used FOX (28)and TOPAS (Bruker Corporation). Single-crystal X-ray diffraction was per-formed using a STOE diffractometer, using Mo-Kα radiation. Single crystalsof K4Sn2S6 were suspended in Paratone oil and sealed in 0.3-mm diameterglass capillaries to prevent degradation of the samples during data col-lection. Single-crystal refinements were conducted using SHELX (39). Be-cause the K4Sn2S6 crystals were composites of multiple domains, the spacegroup and approximate atomic positions were extracted using SHELX andthen further refined using powder data. In most cases, atomic displace-ment parameters Uiso given in SI Text are constrained to a single value peratomic species, due to the limited statistics inherent to structure solutionby powder diffraction of multiphase samples and moderate-resolutioninstruments (i.e., 17-BM). Additional structural details and refinementresults are included in SI Text.

10926 | www.pnas.org/cgi/doi/10.1073/pnas.1406211111 Shoemaker et al.

Dow

nloa

ded

by g

uest

on

Mar

ch 3

0, 2

020

ACKNOWLEDGMENTS. Research at Argonne National Laboratory and use ofthe Advanced Photon Source (beamlines 11-BM and 17-BM) is supported by

the US Department of Energy, Office of Basic Energy Sciences, MaterialsScience and Engineering Division, under Contract DE-AC02-06CH11357.

1. Elwell D, Scheel HJA (1975) Crystal Growth from High-temperature Solutions(Academic, London).

2. Kanatzidis MG (1990) Molten alkali-metal polychalcogenides as reagents and solventsfor the synthesis of new chalcogenide materials. Chem Mater 2:353–363.

3. Stein A, Keller SW, Mallouk TE (1993) Turning down the heat: Design and mechanismin solid-state synthesis. Science 259(5101):1558–1564.

4. Manos E, Kanatzidis MG, Ibers JA (2011) The Chemistry of the Actinide and Trans-actinide Elements, eds Morss LR, Edelstein NM, Fuger J (Springer, Dordrecht, TheNetherlands), pp 4005–4077.

5. Liu X, Fechler N, Antonietti M (2013) Salt melt synthesis of ceramics, semiconductorsand carbon nanostructures. Chem Soc Rev 42(21):8237–8265.

6. Fisher IR, Shapiro MC, Analytis JG (2012) Principles of crystal growth of intermetallicand oxide compounds from molten solutions. Philos Mag 92:2401–2435.

7. Scheel HJ (1974) Crystallization of sulfides from alkali polysulfide fluxes. J CrystGrowth 24–25:669–673.

8. Kanatzidis MG, Sutorik AC (1995) The application of polychalcogenide salts to theexploratory synthesis of solid state multinary chalcogenides at intermediate tem-peratures. Prog Inorg Chem 43:151–265.

9. Mugavero SJ III, Gemmill WR, Roof IP, zur Loye H-C (2009) Materials discovery bycrystal growth: Lanthanide metal containing oxides of the platinum group metals (Ru,Os, Ir, Rh, Pd, Pt) from molten alkali metal hydroxides. J Solid State Chem 182:1950–1963.

10. Mahjoor P, Latturner SE (2012) Salt-flux synthesis of complex oxides: Cs0.33MoO3, CsFe(MoO4)2, and the inverse salt-inclusion phase Cs2Mo0.65O0.21Cl5.44. Philos Mag 92:2582–2595.

11. Graf C, Assoud A, Mayasree O, Kleinke H (2009) Solid state polyselenides and poly-tellurides: A large variety of Se-Se and Te-Te interactions.Molecules 14(9):3115–3131.

12. Stoll P, Näther C, Bensch W (2002) Syntheses and crystal structures of a new niobiumpolysulfide K6Nb4S22 and two new dimorphic tantalum polysulfides K6Ta4S22. Z AnorgAllg Chem 628:2489–2494.

13. Ozawa TC, Kauzlarich SM (2003) NaCl/KCl flux single crystal growth and crystalstructure of the new quaternary mixed-metal pnictide: BaCuZn3As3. Inorg Chem42(10):3183–3186.

14. Ni N, et al. (2008) Anisotropic thermodynamic and transport properties of single-crystalline Ba1-xKxFe2As2 (x=0 and 0.45). Phys Rev B 78:014507.

15. Canfield PC, Fisk Z (1992) Growth of single crystals from metallic fluxes. Philos Mag B65:1117–1123.

16. Wadia C, Alivisatos AP, Kammen DM (2009) Materials availability expands theopportunity for large-scale photovoltaics deployment. Environ Sci Technol 43(6):2072–2077.

17. Mitzi DB, Gunawan O, Todorov TK, Wang K, Guha S (2011) The path towards a high-performance solution-processed kesterite solar cell. Sol Energy Mater Sol Cells95:1421–1436.

18. Manos MJ, Kanatzidis MG (2009) Sequestration of heavy metals from water withlayered metal sulfides. Chemistry 15(19):4779–4784.

19. Shoemaker DP, et al. (2012) Understanding fluxes as media for directed synthesis: In

situ local structure of molten potassium polysulfides. J Am Chem Soc 134(22):

9456–9463.20. Hu Y-J, et al. (2013) Understanding the role of aqueous solution speciation and its

application to the directed syntheses of complex oxidic Zr chlorides and sulfates. J Am

Chem Soc 135(38):14240–14248.21. Pienack N, Bensch W (2011) In-situ monitoring of the formation of crystalline solids.

Angew Chem Int Ed Engl 50(9):2014–2034.22. Geselbracht MJ, et al. (2004) Probing molten salt flux reactions using time-resolved in

situ high-temperature powder X-ray diffraction: A new synthesis route to the mixed-

valence NaTi2O4. Chem Mater 16:1153–1159.23. Scarlett NVY, Grey IE, Brand HEA (2013) In situ synchrotron diffraction studies on the

formation kinetics of jarosites. J Synchrotron Radiat 20(Pt 2):366–375.24. Woodley SM, Catlow R (2008) Crystal structure prediction from first principles. Nat

Mater 7(12):937–946.25. Hautier G, Fischer C, Ehrlacher V, Jain A, Ceder G (2011) Data mined ionic sub-

stitutions for the discovery of new compounds. Inorg Chem 50(2):656–663.26. Jain A, et al. (2011) A high-throughput infrastructure for density functional theory

calculations. Comput Mater Sci 50:2295–2310.27. Ong SP, et al. (2011) The Materials Project. Available at http://materialsproject.org/.

Accessed September 14, 2013.28. Favre-Nicolin V, �Cerný R (2002) FOX, ‘free objects for crystallography’: A modular

approach to ab initio structure determination from powder diffraction. J Appl Cryst

35:734–743.29. Le Bail A (2005) Whole powder pattern decomposition methods and applications: A

retrospection. Powder Diffraction 20:316–326.30. Burschka C, Bronger W (1977) KCu3S2, ein neues Thiocuprat. Z Naturforsch B 32:11–14.31. Whangbo MH, Canadell E (1990) Apparent absence of electronic instability in the

one-dimensional metal sodium copper sulfide (Na3Cu4S4). Inorg Chem 29:1395–1397.32. Bair HE, et al. (1987) Charge-density waves in the mixed-valence two-dimensional

metal K3Cu8S6. Phys Rev B Condens Matter 35(4):1932–1938.33. Stewart GR (2011) Superconductivity in iron compounds. Rev Mod Phys 83:1589–1652.34. Brown DB, et al. (1980) Solid-state structure and electronic properties of a mixed-

valence two-dimensional metal, potassium copper sulfide (KCu4S3). Inorg Chem 19:

1945–1950.35. Folmer JC, Jellinek F (1980) The valence of copper in sulphides and selenides: An X-ray

photoelectron spectroscopy study. J Less Common Met 76(1-2):153–162.36. Klepp KO, Fabian F (1999) Neue Chalkogenometallate mit binuklearen Anionen, I:

Darstellung und Kristallstruktur von Rb6Sn2S7. Z Naturforsch B 54:1505–1509.37. Chupas PJ, et al. (2008) A versatile sample-environment cell for non-ambient X-ray

scattering experiments. J Appl Cryst 41:822–824.38. Toby BH (2001) EXPGUI, a graphical user interface for GSAS. J Appl Cryst 34:210–213.39. Sheldrick GM (2008) A short history of SHELX. Acta Crystallogr A 64(Pt 1):112–122.

Shoemaker et al. PNAS | July 29, 2014 | vol. 111 | no. 30 | 10927

CHEM

ISTR

Y

Dow

nloa

ded

by g

uest

on

Mar

ch 3

0, 2

020