Sergei A.
Novikov
,
James
Casey
,
Hope A.
Long
and
Vladislav V.
Klepov
*
Department of Chemistry, University of Georgia, Athens, Georgia 30602, USA. E-mail: klepov@uga.edu
First published on 11th June 2025
A group of otherwise covalent ternary copper sulfides and selenides demonstrate metallic p-type conductivity and Pauli paramagnetism, which distinguishes them from semiconducting counterparts built up from the same elements. This difference originates from the electron deficiency delocalized over structural units of copper chalcogenides. Considering the significant technological importance of chalcogenide materials, we surveyed ternary copper sulfides and selenides with alkali and alkaline earth cations to provide insights into their crystal structures and transport properties. We found that the compositions and structures of many chalcogenides can be rationalized based on two 2D nets with honeycomb and square lattice net topologies, respectively. We have also shown how simple electron-orbital counts can be applied to metallic chalcogenides with formally charge unbalanced compositions and demonstrated that these compounds are indeed electron deficient. In agreement with experimental data and DFT calculations, such phases demonstrate metallic p-type conductivity.
Numerous natural and synthetic copper chalcogenides with metallic conductivity and Pauli paramagnetism have been known for decades.1–4 Unlike graphite, materials such as covellite CuS, klockmannite CuSe, or umangite Cu3Se2, are p-type conductors. While the physical properties of many chalcogenides have been thoroughly studied, the theoretical basis of their physical behavior has been developed afterwards. The common trait of these phases is a deviation from the oxidation state formalism. It became widely acknowledged that the oxidation state of Cu in these peculiar phases is +1, while mixed +1/+2 states were ruled out,5–8 resulting in the deficit of the formal negative charge in binary (CuS, CuSe, Cu3Se2, CuS2, CuSe2) and ternary (NaCu4S3, NaCu4Se3, NaCu4S4, NaCu4Se4) phases. This distinguishes copper from other metals, demonstrating truly mixed oxidation states in chalcogenides.9,10 DFT study of covellite confirmed the presence of holes in the valence band and the absence of a bandgap in CuS (more precisely Cu6S6, as the structure is built on stacking of CuS and Cu2S2 layers and has S–S bonds), and thus, the observed p-type metallic conductivity aligned perfectly with the experiments.11 The holes originate from a mismatch between the number of molecular orbitals and the number of available valence electrons to fill them. The same study showed a significant level of hybridization between copper and sulfur valence orbitals near the Fermi level, indicating the delocalization of holes over Cu3S3 building blocks in the covellite structure. The delocalization of holes over copper chalcogenide slabs (hereinafter, slab is defined as a layer made by connecting the simplest layers) suggests slightly higher positive charges on both copper and chalcogenide atoms (in other words, more positive on copper, less negative on sulfur), which must not be confused with mixed valence on both copper and chalcogen atoms.12 The same phenomenon was confirmed in the recently reported paramagnetic NaCu4S3 phase13 and a series of related electron deficient phases.14–16
The unique electronic structure of covalent metals made them attractive for applications as a new class of catalyst for CO2 conversion into value-added products.17 Apart from catalysis, metal chalcogenides were studied as prospective materials for a wide range of vital applications, including rechargeable batteries,18–27 thermoelectric power generation,28–35 heterogeneous photo- and electrocatalysis,36–39 photovoltaics,40,41 solar cells with unmatched performance,42–46 radionuclides removal,47 neuromorphic engineering,48,49 cancer treatment,50,51 and others.52,53 Although most chalcogenide research focuses on the semiconductors, recent advances in understanding the electronic behavior of materials with an intermediate type of bonding challenged the traditional systematics of covalent/metallic/ionic bonding mechanisms.54–56 This intermediate type of bonding, which has also been termed “metavalent”, is responsible for unconventional properties of materials such as GeTe, SnTe, and PbTe,54–56 which have promising applications in thermoelectrics. Although these materials started attracting more interest, a detailed understanding of their electronic structures and crystal design principles is still missing.
The goal of this highlight is to illustrate electron deficiency in metallic covalent compounds by using the structural and electronic features of ternary copper sulfides and selenides with alkali and alkaline metals as an example. These materials are divided into two major groups: formally charge balanced and formally charge unbalanced (i.e. electron deficient). A common and most apparent indication of a potential chalcogenide covalent metal is the difficulty of assigning oxidation states in the composition using the common oxidation states formalism, which provides a starting point for their systematic analysis. Following a traditional crystallographic approach, we start with the local coordination of copper and chalcogenide atoms and consider building blocks in the structures to show how they can be arranged in more complex structures. While many hypothetical building unit stackings are possible, few of them have been realized in real materials, leaving this field awaiting the synthesis of many new compositions.
Besides polychalcogenides, other sources of alkali cations, including carbonates, thiocyanates, and hydroxide–halide mixtures, can be employed to create fluxes. To avoid pressure buildup during the synthesis, reactions with carbonates are performed in open crucibles under an inert atmosphere.14,66–71 Mixed alkali hydroxides-sodium iodide flux was successfully employed for the synthesis of charge unbalanced Na3Cu4Se4 phase isostructural with K3Cu4Se4.72 Such flux, however, is highly reactive and cannot be used for synthesis in a glass ampule or alumina crucible. Therefore, preparations of Na3Cu4Se4 and some other phases were done in a glassy carbon boat under nitrogen flow in a tube furnace.72 Thiocyanate fluxes were used to prepare BaCu2S2 and CsCu5S3 phases in two-step processes.73,74 Potassium thiocyanate, used for the synthesis of Ba phase, acted more like a solvent, as the resulting structure does not contain potassium. On the contrary, the synthesis of CsCu5S3 was assisted by cesium thiocyanate flux.
Although, in most cases reported so far, elemental starting materials are used as primary reagents, one can employ boron–chalcogen mixtures (BCM) to reduce oxides and form chalcogenide phases. This method is generally utilized for the synthesis of copper chalcogenides with elements that have high oxygen affinity and, thus, tend to oxidize in air. For example, NaCuUS3 phase can be formed by reacting U3O8, Cu, Na2CO3, B, and S.75 In this reaction, B binds oxygen in the system and S reacts with the metals, providing both metal chalcogenides to form the final phase and the polychalcogenide flux, NaSn, to grow its single crystals. Similarly, the CsCu4Q3 phase has been prepared using BCM.76 Unlike the previous report on the synthesis of CsCu4S3 phase, where the reaction was performed under an Ar atmosphere,70 the BCM reaction was performed in a sealed silica tube. Alternatively, the CsCu4Se3 phase can be obtained by using a hydrothermal route, which is relatively less common for chalcogenides.77 This phase formed in the reaction between ≈250 mg of a mixture of K2Se4, Cu, and CsCl in a sealed quartz ampoule with 0.5 ml of DI water at 120 and 170 °C.77 These examples show interchangeability between the synthetic techniques. Additionally, new phases can be obtained by solid state postsynthetic modifications of the known phases. For example, a rare case of utilizing ternary sulfide as a starting material was the preparation of the CsCu5S3 phase via solid state reaction of the CsCu4S3 phase with extra copper.78 Overall, rich variability of synthetic techniques offers a convenient playground for the synthesis of new copper chalcogenide phases with desired compositions and properties.
Naturally, the formation of solid solutions between isostructural sulfides and selenides can be assumed. As a recent report on this matter demonstrated, the mismatches of chalcogenide sizes and covalency result in the pronounced structural ordering in mixed chalcogenides.79 This ordering was rationalized based on crystallographic data and hard–soft acid–base principles, which allows for the avoidance of costly DFT calculations. The results were confirmed experimentally, and the new KCuZrTe2S ordered mixed anionic phase was prepared.
Composition | Orbital count | Electron count | Electron-orbital mismatch | Type | sql phases | hcb phases |
---|---|---|---|---|---|---|
A(CuQ)n, n = 1 | ||||||
A+(CuQ)− | 18/2 | 17 + 1 | 0 | Balanced | LiCuSe (ref. 82), NaCuSe (ref. 83) | KCuSe (ref. 83) |
n = 2 | ||||||
A+(Cu2Q2)− | 36/2 | 34 + 1 | 1 | Unbalanced | KCu2Se2 (ref. 84) | — |
A2+(Cu2Q2)2− | 36/2 | 34 + 2 | 0 | Balanced | t-BaCu2Q2 (ref. 73) | CaCu2S2 (ref. 85) |
A+(Cu3Q3)− | 54/2 | 51 + 1 | 2 | Unbalanced | Likely unstable | |
A2+(Cu3Q3)2− | 54/2 | 51 + 2 | 1 | Unbalanced | — | — |
n = 4 | ||||||
A+(Cu4Q4)− | 72/2 | 68 + 1 | 3 | Unbalanced | Likely unstable | |
A2+(Cu4Q4)2− | 72/2 | 68 + 2 | 2 | Unbalanced | Likely unstable | |
A+(Cu4Q2(Q2))− | 70/2 | 68 + 1 | 1 | Unbalanced | — | NaCu4Q4 (ref. 16 and 63) |
A+(Cu4Q3)− | 64/2 | 62 + 1 | 1 | Unbalanced | ACu4Q3 (ref. 60, 71, 76 and 86) | NaCu4Q3 (ref. 13 and 15) |
n = 6 | ||||||
A+(Cu6Q4)− | 92/2 | 90 + 1 | 1 | Unbalanced | — | NaCu6Se4 (ref. 61) |
![]() | ||
Fig. 2 (a) hcb copper selenide net with trigonal planar Cu coordination in KCuSe phase; (b) sql copper selenide net with tetrahedral Cu coordination in LiCuSe structure. |
Basic single CuQ nets can be found in charge balanced A+CuSe phases (A+ = Li–K). The tetragonal structures of LiCuSe and NaCuSe phases are built of sql nets, whereas the hexagonal KCuSe is comprised of hcb nets (Fig. 2, Table 1). By doubling the basic structural units, one can derive the structures with [Cu2Q2]n− stoichiometries. Depending on the counter cation (either alkali or alkaline earth metal), these compounds can represent either charge balanced (such as BaCu2Q2 series) or charge unbalanced compositions (for example, orthorhombic KCu2Se2 phase). Nevertheless, both examples feature isolated sql nets separated by the cations. Multiplying of the layers can also occur by connecting atoms between them and forming a slab. For example, the structure of charge balanced trigonal CaCu2S2 phase is built by connecting some nodes of two hcb nets to build Cu2S22− slabs (Fig. 3a–c). Interactions between two flat hcb layers change the local environment of both Cu and S atoms, leading to their distorted tetrahedral coordination environment. Ca atoms in this structure also have octahedral sulfur coordination and occupy the interlayer space. Two neighboring sql nets with common nodes form a Cu4S3− slab in KCu4S3 phase (Fig. 3d–f).
![]() | ||
Fig. 3 (a–c) Double hcb and (d–e) double sql slabs in the structures of charge-balanced CaCuS2 and charge-unbalanced KCu4S3 phases, respectively. |
The electron deficiency of copper chalcogenide phases can be easily calculated from the chemical composition and the knowledge about Q–Q bonds in the structure (Table 1). For a simple NaCuSe case, the total number of valence orbitals is given as a sum of five 3d orbitals from the Cu atoms, one 4s and three 4p orbitals from the Se atoms, resulting in a total of nine. In these orbital count calculations, we assume that Cu 4s and 4p orbitals have significantly higher energy and have only a marginal (yet, very important in some cases) contribution to the top of the valence band. For the electrons, each Cu atom provides 11 electrons, and each S atom offers six electrons, resulting in 17 electrons per (CuS) unit. For A+CuQ phases (A = Li–K), an additional electron is coming from the alkali metals, charge balancing the layers with one electron to result in nine orbitals with 18 electrons. Doubling the (CuS) units yields 18 orbitals with only 34 electrons, requiring an alkaline earth metal to charge balance it, as in BaCu2Q2. On the other hand, if only one alkali cation is balancing the layers in an A+Cu2Q2 composition, such as KCu2Se2, the layers are one electron short of the charge balanced state, resulting in a one-electron-deficient structure. Further increase in the number of stacking layers increases the electron deficiency. The copper chalcogenide phases reported to date exhibit a maximum of one electron deficiency per formula unit (or 0.5 electron per single CuS unit in NaCu2S2), allowing one to surmise that the structures with two or more electron deficiency sites per formula unit are unstable. Therefore, ternaries with A+Cu3Q3 stoichiometry have not been reported to date, with (CuQ)3− 2D units being likely unstable due to high electron deficiency, 2/3 electrons per CuS unit.
It is easy to see that the higher homologues of the (CuS)n family should also be unstable. There are, however, ways to alleviate the electron deficiency in compositions with high copper content, by either forming dichalcogenide bonds or sharing atoms between layer units. A good example of the former is NaCu4Q4 compounds, which formulas can also be written as Na+[Cu4Q2(Q2)]−. Forming a Q–Q bond effectively eliminates one orbital from the orbital count by shifting Q–Q antibonding state significantly above the Fermi level,13 resulting in 9 × 4 − 1 = 35 orbitals with 17 × 4 + 1 = 69 electrons. Such units are present in the NaCu4Q4 phases, which contain a slab of four hcb layers with a dichalcogenide bond connecting two inner ones. This dichalcogenide bond binds the layers into a double layer with a kagome dual net topology (Fig. 4, Table 1). The former approach to reducing orbital count achieves the stabilization of the structure by sharing atoms from different layers, typically a chalcogen atom. One common example is A+(Cu4Q3)− and A2+(Cu4Q3)2− compositions. In A+(Cu4Q3)−, 4 × 5 + 3 × 4 = 32 orbitals host 11 × 4 + 3 × 6 + 1 = 63 electrons, having a deficiency of one hole per formula unit. In the ACu4Q3 (A = K–Cs) series of charge unbalanced tetragonal phases, Cu4Q3− slabs are formed by two sql nets with shared Q atoms. The slabs are separated by alkali cations in the structures. When a smaller Na+ cation occupies the A site, the resulting structure contains vertex-sharing hcb layers combined into Cu4S3− slabs. Similar to the CaCu2S2 phase, the copper coordination number in the Cu4S3− slabs is increased to four and the hcb layers are visibly corrugated. The way two inner hcb layers of a Cu4S3− slab are connected also matches the topology of kagome dual net. Finally, the highest homologue reported so far has a NaCu6Se4 composition. In NaCu6Se4, two CuSe hcb nets are divided not by one but by two Cu2Se double hcb, forming slabs with a Cu6Se4 composition. The formula unit contains 46 orbitals and 91 electrons, showing the same one hole per formula unit electron deficiency. To the best of our knowledge, no ternary phase built on more single and double hcb layers has been reported; however, a slab made of two single and three double layers can be found in the structure of Cu9S5 binary. Overall, shared atoms and dichalcogenide bonds are an efficient way to stabilize higher homologues of copper chalcogenide ternaries, enabling the formation of phases with lower hole concentration per Cu atom. Since higher homologues require using small A:
Cu molar ratios, their synthesis may be challenging in the presence of commonly used polychalcogenide fluxes. Alternative synthetic routes using fluxes or solvents that dissolve the starting materials but do not incorporate into the final products are therefore required to stabilize those phases.
![]() | ||
Fig. 4 Cu4S4 slab in the NaCu4S4 structure made from four hcb layers. Two central layers are connected in kagome dual net motif (kgd). |
The remaining ternary copper sulfide and selenide structures are not based on hcb and sql nets (Table 2), and their dimensionality varies from 0D to 3D. For example, unlike the A+(Cu4Q3)− series, the charge balanced Na2Cu4S3 phase, as well as both the orthorhombic primitive (oP) and base centered (oC) modifications of the BaCu4S3 phase, feature complex 2D (Na, oC Ba) and 3D (oP Ba) structures. Another example of isomerism among copper chalcogenide units appears in A3+Cu4S4 phases (A+ = Na+ or K+) containing 1D Cu4S4 units with decreased dimensionality due to a high content of alkali cations. Besides oP BaCu4S3, anionic frameworks can be found in the structures of o-BaCu2Q2, ACu5Q3 (A = Na, Cs), and ACu7S4 (A = K, Rb) series with alkali and alkaline earth cations occupying channels of frameworks. 2D anionic slabs and layers are quite common among chalcogenides with various stoichiometries. Usually, alkali and alkaline cations separate 2D anionic units, however, in the recently reported K3BaCu7S6 phase Ba2+ cations are inserted in Cu7S6 double layers, while K+ cations separate such layers.72 Interstitial sites in copper sulfides and selenides layers are usually vacant. However, in mixed anion ACu4.2TeS2 series of compounds isostructural to KCu4S3, there are additional copper atoms partially occupying the voids of the CuTeS2 layers.91 The volume of the Voronoi polyhedron92 for the interstitial copper atom in the KCu4.2TeS2 structure is about 3.6 Å3 larger than the void in the KCu4S3 phases. This discrepancy volume is due to the larger size of the Te atom, which allows for the accommodation of additional Cu atoms. High content of cations per formula unit results in chalcogenides dimensionality reduction, while the presence of polychalcogenide anions facilitates the formation of 1D anionic chains: for example, CuQ4 and CuS6 chains contain Q42− and S62− anions, respectively (Table 2). Finally, the charge balanced Na4Cu2S3 phase appears to be the only compound with discrete Cu2S34− 0D anions. As can be seen, the complexity of anionic units exceeds structures solely built on sql and hcb motifs in many chalcogenides. Nevertheless, the same principles of orbital and electron count can be applied to them (Table 2). Similar to phases built on CuQ fragments, the electron deficiency in phases without such fragments does not exceed one electron per formula unit.
Composition | Orbitals count | Electron count | Electron-orbital count mismatch | Type | Dimensionality of CuxQy unit |
---|---|---|---|---|---|
Cu![]() ![]() |
|||||
CsCu((S6)2−) (ref. 93) | 58/2 − 10/2 | 47 + 1 | 0 | Balanced | 1D |
Cu![]() ![]() |
|||||
A+Cu(Q42−) (ref. 77, 94–96) | 42/2 − 6/2 | 35 + 1 | 0 | Balanced | 1D |
Cu![]() ![]() |
|||||
Ba4Cu8Se2(Se22−)4(Se34−) (ref. 97) | 184/2 − 10/2 | 166 + 8 | 0 | Balanced | 2D |
Na4Cu2S3 (ref. 64) | 44/2 | 40 + 4 | 0 | Balanced | 0D |
Cu![]() ![]() |
|||||
o-BaCu2Q2 (ref. 73) | 36/2 | 34 + 2 | 0 | Balanced | 3D |
KCuS (ref. 65) | 18/2 | 17 +1 | 0 | Balanced | 1D |
A3Cu4S4, A = Na, K (ref. 59 and 67) | 72/2 | 68 + 3 | 1 | Unbalanced | 1D |
Cu![]() ![]() |
|||||
K3BaCu7S6 (ref. 72) | 118/2 | 113 + 5 | 0 | Balanced | 2D |
Cu![]() ![]() |
|||||
Na7Cu12S10 (ref. 58) | 200/2 | 192 + 7 | 1 | Unbalanced | 1D |
Cu![]() ![]() |
|||||
Cs2Cu5Se4 (ref. 69) | 82/2 | 79 + 2 | 1 | Unbalanced | 2D |
Cu![]() ![]() |
|||||
BaCu4S3 (ref. 98), Na2Cu4S3 (ref. 99) | 64/2 | 62 + 2 | 0 | Balanced | 3D (Ba), 2D (Na) |
A3Cu8Q6, A = K–Cs (ref. 62, 66 and 68) | 128/2 | 124 + 3 | 1 | Unbalanced | 2D |
ACu3Q2, A = K–Cs (ref. 70 and 100–102) | 46/2 | 45 + 1 | 0 | Balanced | 2D |
Cu![]() ![]() |
|||||
ACu5Q3, A = Na, Cs (ref. 78 and 87–90) | 74/2 | 73 + 1 | 0 | Balanced | 3D and 2D |
Cu![]() ![]() |
|||||
ACu7S4, A = K, Rb (ref. 103) | 102/2 | 101 + 1 | 0 | Balanced | 3D |
![]() | ||
Fig. 5 Electronic band structure and DOS for (a and b) CuS, (c) NaCu4S3, (d) NaCu4S4, and (e) NaCu4Se4 phases. Adapted with permission from ref. 11, 13 and 16. Copyright: American Chemical Society. |
DFT calculations show crucial common features of all four phases under consideration: first, the upper bands in the band structure cross the Fermi level in the proximity of the Γ point. This accounts for the metallic properties of the chalcogenides and agrees with the experimental data on electrical conductivity and magnetism. Second, the dispersion in the upper bands significantly depends on the direction. Bands contributed by Cu 3dx2–y2, 3dxy, and S 3px, 3py orbitals are highly dispersed in directions parallel to CuxSy slabs. This agrees with the structural data, as orbitals forming these bands participate in the copper–chalcogen bonding within the CuxSy slabs. On the contrary, bands with predominant contributions from Cu 3dxz and 3dyz orbitals are weakly dispersed in all directions, as the corresponding orbitals do not participate in the bonding. Notable differences in the band structures of CuS,9 NaCu4S3,11 NaCu4S4,11 and NaCu4Se4 (ref. 14) phases are observed in the direction perpendicular to the slabs. Cu 3dz2 and Q npz orbitals are oriented in this direction. In NaCu4S3, the overlap of these orbitals results in the bands crossing the Fermi level and their dispersion in the c direction in the crystal structure. In CuS, NaCu4S4 and NaCu4Se4 Q–Q bonds form, and the overlap of npz orbitals significantly surpasses the overlap of 3dz2 and npz orbitals. The resulting antibonding band has much higher energy and lies entirely above the Fermi level. DFT calculations confirm strong covalency of the Cu–Q bonds and the metallic nature of the electron deficient chalcogenides. It is important to note that this view agrees well with the XPS results, which only show that Cu atoms in a selected structure have the same oxidation state, which is different from the +2 oxidation state. Having a slight hole delocalized over Cu 3d orbitals only marginally increases the formal charge on Cu atoms.
This journal is © The Royal Society of Chemistry 2025 |