NaGe₆As₆: Insertion of sodium into the layered semiconductor germanium arsenide GeAs

1 Introduction

A large number of ternary pnictides A–M–Pn (A = electropositive metal; M = metal or metalloid from group 13 or 14; Pn = P, As, Sb, Bi) exhibit structures in which A cations are found in conjunction with heteronuclear anionic [M_xPn_y]^n– units [1]. Most of these compounds satisfy the Zintl concept so that there is a definite relationship between the electron count and the crystal structure. In cases where the Pn atoms experience a deficiency of electrons with respect to the 8–N rule, polyanionic networks containing Pn–Pn bonds must necessarily develop. Less commonly, there is a surplus of electrons, which are assumed to reside on the M atoms, leading to formation of polycationic networks containing M–M bonds. Within the A–Ge–As (A = alkali or alkaline–earth metal) systems, the only known phases reported so far are Na₅GeAs₃ [2], K₂GeAs₂ [3], and BaGe₂As₂ [4]. The latter is a polycationic compound with Ge atoms in homoatomic bonding arrangements of infinite cis–trans chains, consistent with the presence of Ge²⁺ species if oxidation states are assigned. A gas-phase molecular species NaGeAs₃ has also been proposed [5].

As part of a broader investigation of other A-Ge-As phases, we report here the preparation and structure determination of the new compound NaGe₆As₆. Remarkably, it is related to the binary arsenide GeAs, a charge-balanced semiconducting compound that has so far been relatively little studied [6–14], through intercalation of Na atoms between layers. We examine the implications of this relationship for the electronic structure.

2 Experimental

3 Results and discussion

NaGe₆As₆ is a new phase in the Na-Ge-As system, in which Na₅GeAs₃ and NaGeAs₃ were the only previously reported representatives. Na₅GeAs₃ is a Zintl phase that contains Na⁺ cations and [Ge₂As₆]¹⁰⁻ anions in the form of discrete edge-sharing tetrahedral units [2]. NaGeAs₃ was prepared in a Knudsen cell and is proposed to consist of gaseous “Zintl molecules” that are aggregates of Na⁺ cations and [GeAs₃]⁻ anions in the form of tetrahedral clusters isoelectronic to As₄ [5].

The structure of NaGe₆As₆ is of a new monoclinic type (space group C2/m) containing [Ge₆As₆] layers alternately stacked with Na atoms along the c direction (Fig. 1a). Each layer is built by condensing chains of edge-sharing octahedra, each centered by Ge–Ge dumbbells, extending along the b direction (Fig. 1b). There are two types of Ge₂As₆ units, each of which resembles ethane in staggered conformation. These units are oriented so that Ge1–Ge1 pairs lie roughly parallel to the layers whereas Ge2–Ge3 pairs lie perpendicular to the layers. The sequence of these pairs (one horizontal Ge1–Ge1 pair followed by two vertical Ge2–Ge3 pairs) as they propagate along the a direction leads to a characteristic motif of two five-membered rings fused to opposite sides of a central six-membered ring. The Na atoms lying between the layers are surrounded by six As atoms in octahedral coordination. Other structures containing ethane-like Ge₂As₆ units have been previously reported in only a few examples: GeAs [12], A₃Ge₂As₄ (A = Ca, Sr) [18], ACdGeAs₂ (A = K, Rb) [19], K₅In₅Ge₅As₁₄ [20], and K₈In₈Ge₅As₁₇ [20] (the latter two exhibit disorder of In and Ge atoms). Similar motifs of fused 5–6–5 rings are found in K₅In₅Ge₅As₁₄ [20] and Ga₃Te₃I [21].

Fig. 1:

Structure of NaGe₆As₆ viewed (a) down the b direction, showing the stacking of layers and (b) down the c direction, showing a layer built up of edge-sharing octahedra centered by Ge–Ge dumbbells. The octahedral coordination geometry is highlighted around one of the Na atoms in (a).

The [Ge₆As₆] layers in NaGe₆As₆ are the same as those found in the binary arsenide GeAs (Fig. 2). As the layers are stacked, they are in perfect registry in NaGe₆As₆ but they are displaced with respect to each other in GeAs. Both structures can also be regarded as being built from rumpled close-packed nets of As atoms stacked in an hcp (AB) arrangement, with Ge–Ge dumbbells centered in the octahedral sites between alternate pairs of layers. This viewpoint draws an analogy to the CdI₂-type structure: when half of all available octahedral sites are filled with Ge–Ge dumbbells within a close-packed array of As atoms, the resulting formula is □_1/2(Ge₂)_1/2(As), or GeAs. When a third of the remaining vacant octahedral sites between the GeAs layers is occupied by Na atoms, the resulting formula is □_2/6(Na)_1/6(Ge₂)_1/2(As), or NaGe₆As₆. This relationship is similar to the intercalation of guest atoms within layered transition-metal dichalcogenides [22].

Fig. 2:

Comparison of the monoclinic structures of (a) GeAs and (b) NaGe₆As₆, both built from similar layers formed by filling octahedral sites with Ge–Ge dumbbells within an hcp stacking of rumpled close-packed As nets.

The possibility that the Ge and As atoms are not ordered, as was assumed in the structure refinement, must be considered given their similar X-ray scattering factors. Evidence to support the ordered model comes from an analysis of the bond valence sums [23], which are 1.3 for the Na atom, 3.5–3.6 for the Ge atoms, and 2.6–3.0 for the As atoms, in agreement with the expected valences. If the assignments of Ge and As atoms are reversed, the bond valence sums would show worse agreement.

The derivation of NaGe₆As₆ by insertion of Na atoms into GeAs poses interesting questions about how the electronic structure is affected. GeAs itself can be considered to be a polycationic Zintl phase, attaining the charge-balanced formulation (Ge³⁺)(As³⁻) consistent with the presence of homoatomic Ge–Ge bonds within the dumbbells. (SiAs [10, 24, 25], GeP [10, 26], and GaTe [27] are the only other known compounds isostructural to GeAs.) On proceeding from GeAs [12] to NaGe₆As₆, there is virtually no change in the Ge–As (2.44(1)–2.47(1) vs. 2.4349(8)–2.4755(9) Å) and Ge–Ge distances (2.43(1)–2.46(1) vs. 2.4376(11)–2.4572(9) Å). It is reasonable to propose that electron transfer will take place from the Na atoms to the rest of the framework to give the formulation Na⁺[Ge₆As₆]⁻. If this increase in electron count is assumed to cause local chemical reduction of the Ge atoms, corresponding to the formulation (Na⁺)(Ge^2.83+)₆(As³⁻)₆, one expects a strengthening of the Ge–Ge bonds, but this is not reflected in the observed distances. Interestingly, although insertion of Na atoms pries the layers further apart (the center-to-center distance between the layers expands from 6.4 to 7.2 Å), the closest As–As separation in the gap between the layers actually decreases slightly from 3.34(1) Å in GeAs to 3.270(1) Å in NaGe₆As₆. These As–As distances are certainly too long to be considered as fully covalent bonds, but it is notable that they are quite a bit shorter than twice the van der Waals radius of As (4.0 Å) [28].

The absence of significant structural changes within the layers implies that the major electronic effect of introducing Na atoms into GeAs is to essentially produce an n-doped semiconductor. This picture is confirmed by comparing the electronic band structures of GeAs and NaGe₆As₆. The density of states (DOS) curve for GeAs clearly reveals valence and conduction bands separated by a gap of 0.58 eV (Fig. 3a), which agrees fortuitously well with an optical band gap of 0.65 eV reported in early experimental measurements [11]. The atomic projections of the DOS show that, as a result of the similar energies of Ge and As atomic orbitals, there is a strong hybridization of these states, which are dominated by 4s character at lowest energy (−14 to −6 eV) and by 4p character at higher energy (−5 to 0 eV). The crystal orbital Hamilton population (COHP) curves indicate that the filled valence bands correspond to Ge–As and Ge–Ge bonding levels. If a rigid band model is applied and the electron count is increased by one, corresponding to NaGe₆As₆, the Fermi level would be raised so that it lies at the bottom of the conduction band. Because antibonding levels now start to be occupied, the expectation is that Ge–As and Ge–Ge bonds will be weakened, or the structure may undergo distortion to avoid this situation. The DOS curve for the actual structure of NaGe₆As₆ (Fig. 3b) is very similar to that of GeAs. The calculated band gap is slightly larger (0.64 eV) and the Fermi level just cuts the bottom of the conduction band. The bonds do weaken, but only very slightly; the integrated COHP values (–ICOHP) change from 2.61 to 2.56 eV/bond for the Ge–As contacts and from 2.74 to 2.66 eV/bond for the Ge–Ge contacts on proceeding from GeAs to NaGe₆As₆. This small destabilization will, of course, be compensated by ionic bonding interactions between the Na atoms and the layers. The situation is reminiscent of Ba₇Ga₄Sb₉, an “electron-rich metallic Zintl phase” in which the apparent violation of charge balance can be reconciled by recognizing the important stabilizing role of the cations [29, 30]. Except for the region around the Na atoms in NaGe₆As₆, the layers are non-interacting and held together only by van der Waals forces, as confirmed by the small and negative –ICOHP values for the interlayer As–As contacts (−0.03 eV/bond in GeAs and −0.10 eV/bond in NaGe₆As₆).

Fig. 3:

Density of states (DOS) and crystal orbital Hamilton population (–COHP) curves for (a) GeAs and (b) NaGe₆As₆.

4 Conclusions

The successful preparation of NaGe₆As₆ suggests that the GeAs structure can tolerate modest changes of its electron count; it may thus be amenable to doping within layers or intercalation between layers by other elements. To establish if the proposed description of intercalation is accurate, it will be important to perform experiments to attempt deintercalation of NaGe₆As₆. Moreover, the physical properties will be expected to be modified; for example, metallic-like electrical conductivity can be predicted for NaGe₆As₆. It will be worthwhile to examine other compounds that are isostructural to GeAs, such as SiAs, or those with related layered structures, such as GaS and GaSe [31–35], to see if they can also accommodate guest atoms between the layers.