Sodium Filling in Superadamantoide Na_1.36(Si_0.86Ga_0.14)₂As_2.98 and the Mixed Valent Arsenidosilicate-Silicide Li_1.5Ga_0.9Si_3.1As₄

Abstract

Na_1.36(Si_0.86Ga_0.14)₂As_2.98 and Li_1.5Ga_0.9Si_3.1As₄ were synthesized by heating mixtures of the elements at 950 °C. The crystal structures were determined by single crystal X-ray diffraction (Na_1.36(Si_0.86Ga_0.14)₂As_2.98: I4₁/a, Z = 100, a = 19.8772(4) Å, c = 37.652(1) Å; Li_1.5Ga_0.9Si_3.1As₄: C2/c, Z = 8, a = 10.8838(6) Å, b = 10.8821(6) Å, c = 13.1591(7) Å). Na_1.36(Si_0.86Ga_0.14)₂As_2.98 crystallizes similar to NaSi₂P₃ with interpenetrating networks of vertex-sharing T4 and T5 supertetrahedra. Gallium substitution at the silicon sites increases the charge of the cluster network, which is compensated for by a 36% higher sodium content. Since in contrast to NaSi₂P₃, all sodium sites are now fully occupied, there is no significant ion mobility, as indicated by ²³Na-NMR. Consequently, the total sodium-ion conductivity of Na_1.36(Si_0.86Ga_0.14)₂As_2.98 amounts to only 1.6(1) × 10⁻⁷ S cm⁻¹ and is therefore three orders of magnitude lower than in NaSi₂P₃. Li_1.5Ga_0.9Si_3.1As₄ crystallizes in a new structure type with layers of edge-sharing (Si_1−xGa_x)As₄ tetrahedra alternating with layers that contain infinite Si_n zigzag chains. Lithium ions reside in channels between the chains, and thus, the structure does not provide three dimensional pathways for ion conduction and the measured total Li-ion conductivity amounts to only 1.3(1) × 10⁻⁷ S cm⁻¹.

1. Introduction

Ion-conducting solids will play a key role as solid electrolytes in future all-solid-state batteries (ASSB) and are therefore the subject of intensive research [1,2,3,4,5,6]. The movement of ions in solids is determined by numerous factors, but the product of mobility and concentration of the ions in question is particularly important. However, these quantities are not independent of each other and are therefore difficult to optimize. During recent years, a series of ion-conducting phosphidosilicates have been published [7,8,9,10,11,12]. Conductivities up to 10⁻³ S cm⁻¹ [8], the easy availability of the constituting elements, and a rich structural chemistry make phosphidosilicates particularly interesting. Among these, we have discovered the compounds Asi₂P₃ (A = Li, Na, K) with interpenetrating networks of supertetrahedra up to T5 [10,11,12]. Alkali metal ions are mobile in spaces between these large units where they experience a significantly lower effective charge than in common phosphidosilicates with highly charged SiP₄⁸⁻ tetrahedra. This is one reason why these compounds are often good ionic conductors (σ > 10⁻⁴ S cm⁻¹); although, they contain considerably fewer cations per unit volume than ion conducting phosphidosilicates with isolated tetrahedra. Nevertheless, the conductivity may be improved by incorporating more alkali metal ions. For this purpose, the negative charge of the supertetrahedra must be slightly increased, which could be achieved by partial substitution of Si⁴⁺ by Al³⁺ or Ga³⁺. Our initial attempts to synthesize Na_1+2x(Si_1−xTr_x)₂P₃ (Tr = Al, Ga) were not yet successful. However, since supertetrahedral clusters likewise occur in the gallium arsenides M₁₅Ga₂₂As₃₂ and M₃Ga₆As₈ (M = Sr, Eu) [13], and also related Si/Ga mixed variants M₄Ga₅SiAs₉ and MgaSiAs₃ (M = Sr, Eu) exist [14], it seemed promising to prepare the homologous arsenides Na_1+2x(Si_1−xGa_x)₂As₃. Here, we report on the synthesis, crystal structure and conductivity measurements of Na_1.36(Si_0.86Ga_0.14)₂As_2.98 as well as about the new compound Li_1.5Ga_0.9Si_3.1As₄ with a previously unknown structure type containing infinite Si_n zigzag-chains.

2. Results and Discussion

The crystal structures of Na_1.36(Si_0.86Ga_0.14)₂As_2.98 and Li_1.5Ga_0.9Si_3.1As₄ were determined from X-ray single crystal data.

Table 1. Crystallographic data of Na_1.36(Si_0.86Ga_0.14)₂As_2.98 and Li_1.5Ga_0.9Si_3.1As₄.

Formula	Na1.36(Si0.86Ga0.14)2As2.98	Li1.5Ga0.9Si3.1As4
formula mass/g mol−1	322.59	460.00
space group	I 41/a (No. 88)	C2/c (No. 15)
a/Å	19.8772(4)	10.8838(6)
b/Å	a	10.8821(6)
c/Å	37.652(1)	13.1591(7)
β/deg	90	101.904(3)
Vcell/Å3	14,876.3(7)	1525.0(1)
Z	100	8
ρX-ray/g cm−1	3.601	4.007
µ/mm−1	18.189	20.887
radiation	Mo-Kα	Mo-Kα
θ-range/°	2.35–30.50	5.36–29.39
reflections measured	157,235	13,182
independent reflections	11,375	13,182
refined parameters	373	98
Rσ	0.0498
Rint	0.0235	0.0894
R1 (F > 2σ(F)/all)	0.0243/0.0300	0.0530/0.0784
wR2 (F2 > 2σ(F2)/all)	0.0543/0.0561	0.1597/0.1766
GooF	1.118	1.034
Δρmax, Δρmin/e Å3	0.14/−0.99	2.63/−1.74
Twin volume fractions		0.499(7)/0.501(7)

shows the final parameters of the refinements, while the fractional atomic positions, equivalent and anisotropic displacement factors are listed in Tables S1–S6 in the Supporting Information. Starting from these crystallographic data, Rietveld refinements of the powder X-ray diffraction data confirmed the crystal structures and purity of the samples (Figure 1). Energy-dispersive X-ray spectra confirm the chemical compositions (Tables S7 and S8).

Na_1.36(Si_0.86Ga_0.14)₂As_2.98 crystallizes in the non-centrosymmetric tetragonal space group I4₁/a (No. 88) and forms a diamond-analogue network of T4+T5 supertetrahedra. The structure is very similar to that of NaSi₂P₃ with each T4 cluster coordinated tetrahedrally by four T5 clusters and vice versa, via one common SiAs₄ tetrahedron (Figure 2).

Two symmetry-related diamond-like networks of clusters interpenetrate each other (Figure 3), as described for NaSi₂P₃ [11] and LiSi₂P₃ [10]. Likewise, the silicon atoms in the centers of the T5 units are missing. Such vacancies also occur in other supertetrahedral compounds like [In₃₄S₅₄]⁶⁻ [15] and HP-B₂S₃ [16]. For NaSi₂P₃, Haffner et al. explained the missing silicon atom with the necessity of charge neutrality inside the cluster [11]. In the present case, the arsenic atom at the center of the T4 clusters is only 60% occupied, which reduces the negative charge in the T4 cluster. The gallium atoms are not evenly distributed over the silicon sites within the clusters; instead, we find positions containing only silicon next to those with different fractions ranging from about 15% to 40% gallium (Figure 2). Without using charge constraints to the partial occupations, the composition is ${\mathrm{N}\mathrm{a}}_{1.360\left(1\right)}^{+}{\mathrm{S}\mathrm{i}}_{1.722\left(3\right)}^{4+}{\mathrm{G}\mathrm{a}}_{0.278\left(3\right)}^{3+}{\mathrm{A}\mathrm{s}}_{2.983\left(1\right)}^{3-}$, which deviates from charge neutrality by only 0.7% (0.13 electrons). Thus, we have increased the sodium content of NaSi₂P₃ by 36% while preserving the crystal structure. This shows that our concept of increasing the alkaline metal concentration by increasing the cluster charge by doping the silicon sites with trivalent atoms works, in principle. The sodium atoms in Na_1.36(Si_0.86Ga_0.14)₂As_2.98 are located at the same positions as in NaSi₂P₃ except for one additional site with large displacements, which has to be described as split position with high uncertainty. Thus, the higher sodium content results from the complete occupation of the sodium positions in contrast to NaSi₂P₃, in which all sodium sites are only partially occupied.

The sodium atoms between the clusters are coordinated in distorted NaAs₆ octahedral or trigonal prismatic polyhedra with Na-As distances from 2.93 to 3.64 Å. One sodium site shows significantly larger Na-As distances up to 4.2 Å without an apparent coordination polyhedron. It is located in the center of the gap between the networks which is not filled with sodium atoms in NaSi₂P₃. Probably this position becomes populated after all other sodium sites are completely filled. We observe extremely elongated thermal displacement ellipsoids at this site (Figure 4) because of the substantial positional disorder. The ²³Na-MAS-NMR spectrum of Na_1.36(Si_0.86Ga_0.14)₂As_2.98 shows one very broad signal with three maxima originating from several overlapping signals (Figure 5). Though we were not able to assign the signals to sodium sites, the fact that NMR distinguishes the sodium atoms at different sites suggests at best low or no mobility of the ions.

Li_1.5Ga_0.9Si_3.1As₄ crystalizes in a monoclinic space group (C2/c, no. 15). The structure contains two alternating layers along the c-axis (Figure 6).

One layer is formed by corner-sharing (Si_1−xGa_x)As₄ tetrahedra with the Si/Ga-As distances close to values in the literature [14]. The second layer contains Si-Si bonded zigzag-chains next to channels filled with lithium atoms. The channels run along [1 1 0] and [−1 −1 0] directions. Silicon forms SiSi₂As₂ tetrahedra with Si-Si distances of 2.345–2.349 Å and an angle of 110.1° between three silicon atoms (Figure 7). Infinite zigzag-chains of silicon atoms are known from the Zintl-phase CaSi with the CrB-type structure and similar Si-Si distances (2.452 Å) [17]. Si₂ pairs occur more frequently, for example, in BaCuSi₂P₃ [18], Ca₃Si₂P₄, Ca₃Si₈P₁₄ [19], and Na₃₁Ba₅Si₅₂P₈₃ [20].

Besides the layer-based structure, there are also T3 supertetrahedral clusters visible, forming an intertwining network with lithium between the clusters (Figure 8). The T3 supertetrahedral units are connected by corner-sharing SiSi₂As₂ tetrahedrons, resulting in an overlap of the clusters along the edges.

The lithium atoms are located in channels between the silicon chains and coordinated in a trigonal prismatic fashion with typical Li-As distances (2.10–2.11 Å). The ⁷Li-MAS-NMR-measurement shows one slightly broadened signal (Figure 9). This signal is presumably a superposition of two signals with similar chemical shifts. The relatively broad signal suggests a weak ion mobility, significantly smaller than in LiSi₂P₃ [10] and more comparable to Ca₂Li₄SiP₄ [21].

Besides the main phase Li_1.5Ga_0.9Si_3.1As₄, the Rietveld refinement (Figure 1) shows additional reflections, which could not be assigned. Table S4 lists the parameters of the Rietveld refinement. Scanning transmission electron microscopy (STEM) visualizes well the different layers in the structure of Li_1.5Ga_0.9Si_3.1As₄ (Figure 10). The brighter spots can be assigned to the (Si_1−xGa_x)As₄ tetrahedra. The Si-Si bonded zigzag-chains perpendicular to the viewing plane are recognizable as very weak spots. These spots alternate with dark sections, which stand for the channels filled with Li atoms. The weak spots of the silicon atoms of the zigzag-chains parallel to the viewing plane are next to the corresponding slightly brighter spots of the arsenic atoms.

Electrochemical impedance spectroscopy (EIS) was used to determine the ionic conductivity of Na_1.36(Si_0.86Ga_0.14)₂As_2.98 and Li_1.5Ga_0.9Si_3.1As₄. All EIS data were fitted using equivalent circuit models containing resistors (R), capacitors (C), and constant phase elements (CPE). Resistors and CPEs were used to model partial ionic and electronic conductivities, and polarization. CPEs were used to account for non-ideal sample behavior [22,23]. Capacitors were used to model stray capacities, which were determined at 25 °C and then kept constant for all other spectra [24]. Figure 11 shows impedance data for Na_1.36(Si_0.86Ga_0.14)₂As_2.98.

The EIS data Na_1.36(Si_0.86Ga_0.14)₂As_2.98 show a depressed semicircle with a spike in the low frequency region. Hence, the data were fitted using an equivalent circuit with a resistor and a CPE in parallel (R1/CPE1) to model a single semicircle and a CPE in series (CPE2) to model polarization at the electrodes. In addition, a capacitor (C1) was used in series to the other circuit elements to account for stray capacity, which was fitted to 4 × 10⁻¹² F. To calculate the capacity of CPE1, the Brug formula was used [23,25]. The capacity of CPE1 at 25 °C is 1.42 × 10⁻¹¹ F, which is of the order of magnitude for both bulk and grain boundary processes [26]. Since an unambiguous assignment is hence not possible, only total conductivities are reported. The total Na⁺ ionic conductivity of Na_1.36(Si_0.86Ga_0.14)₂As_2.98 at 25 °C is 1.6(1) × 10⁻⁷ S cm⁻¹. The inset in Figure 11 shows an Arrhenius plot used for the determination of the activation energy E_a for the Na⁺ transport, yielding 0.339(5) eV. To determine the partial electronic conductivity of Na_1.36(Si_0.86Ga_0.14)₂As_2.98, potentiostatic polarization measurements were carried out. The respective I/t and I/U curves are shown in Figures S1 and S2. Additional information on polarization measurements is also given in the Supporting Information. The partial electronic conductivity determined from polarization measurements of Na_1.36(Si_0.86Ga_0.14)₂As_2.98 is 2.1(5) × 10⁻⁸ S cm⁻¹. The reported partial electronic conductivity can be seen as a maximum, since the system might not have fully equilibrated after 6 h of polarization. To set the partial ionic σ_ion and electronic conductivities σ_eon in relation, the ionic transference number τ_ion = σ_ion/(σ_ion + σ_eon) can be calculated [27]. For Na_1.36(Si_0.86Ga_0.14)₂As_2.98, τ_ion is 0.88. For an ideal ion conductor, τ should be as close to unity as possible; hence, Na_1.36(Si_0.86Ga_0.14)₂As_2.98 can be classified as a mixed ionic–electronic conductor [27].

The EIS data of Li_1.5Ga_0.9Si_3.1As₄ are shown in Figure 12. Again, a depressed semicircle with a barely visible spike in the low frequency region is visible. In contrast to Na_1.36(Si_0.86Ga_0.14)₂As_2.98, the low frequency spike is significantly flatter indicating a relatively large influence of a partial electronic conductivity. To model the EIS data of Li_1.5Ga_0.9Si_3.1As₄, the same equivalent circuit like for Na_1.36(Si_0.86Ga_0.14)₂As_2.98 was used, but in addition another resistor in parallel to the other circuit, element R2 was added in order to account for the apparent partial electronic conductivity [27]. The capacity of the capacitor C1 was again determined at 25 °C and fixed at 6 × 10⁻¹² F for all fits. The Brug capacity of CPE1 is 3.54 × 10⁻¹¹ F at 25 °C; hence again, an unambiguous assignment of the process is not possible [23,25,26]. Calculated solely from R1, the total Li⁺ ion conductivity of Li_1.5Ga_0.9Si_3.1As₄ at 25 °C is 1.3(1) × 10⁻⁷ S cm⁻¹. The E_a for the Li⁺ ion transport in Li_1.5Ga_0.9Si_3.1As₄ determined from an Arrhenius fit is 0.405(5) eV. The corresponding Arrhenius fit is shown in the inset of Figure 12. The partial electronic conductivity determined from EIS σ_eon,EIS using R2 is 3.28(5) × 10⁻⁸ S cm⁻¹ at 25 °C, which is in good agreement with the partial electronic conductivity determined from direct current polarization measurements σ_eon,_DC, which is 1.7(1) × 10⁻⁸ S cm⁻¹ at 25 °C. The respective I/t and I/U curves are shown in Figures S3 and S4. The respective ionic transference numbers τ_ion for Li_1.5Ga_0.9Si_3.1As₄ range from 0.88 to 0.80. Taking the partial equilibration of Na_1.36(Si_0.86Ga_0.14)₂As_2.98 into account, Li_1.5Ga_0.9Si_3.1As₄ is showing a stronger influence of electronic leakage compared to Na_1.36(Si_0.86Ga_0.14)₂As_2.98.

3. Materials and Methods

3.1. Synthesis

The compounds were prepared using solid-state reactions. Samples were handled under inert conditions in an argon-filled glovebox with concentrations of O₂ and H₂O < 0.1 ppm. A stoichiometric ratio of metallic sodium (Alfa Aesar: Haverhill, MA, USA, 99.8%) or lithium (Sigma-Aldrich: Taufkirchen, Germany, 99.99%) were filled with metallic gallium (Alfa Aesar, 99.999%) in an alumina crucible. Silicon powder (Alfa Aesar, 99.999%) and arsenic powder (Alfa Aesar, 99.999%) were grounded and added to the crucible. The crucible was sealed in a silica ampule and the ampoule was sealed in another silica ampoule. The reaction mixture was heated to 950 °C and temperature was maintained for 60 h. After cooling slowly to 750 °C, the furnace was shut off. The sample was grounded and reheated with the same procedure, yielding a dark grey, moisture-sensitive powder.

3.2. Single-Crystal X-ray Diffraction

Single crystals were selected from the polycrystalline sample in an argon-filled glovebox and sealed in glass capillaries (Hilgenberg GmbH, Marsfeld, Germany, 1.0 mm diameter). X-ray intensity data were obtained using a Bruker D8 Quest diffractometer (Bruker AXS Inc.: Madison, WI, USA, Mo-Kα radiation, Göbel mirror optics, Photon-II CPAD detector). For intensity integration, data reduction and adsorption correction the software APEX3 v2016.5-0 ed [28] and SADABS v2012/1 [29] were used. Based on systematically absent reflections, the space group determination was carried out with XPREP v2008/2 [30]. The structure solution and refinement were carried out with OLEX2 v1.5 [31]. The final structure was visualized with the DIAMOND v3.2k software [32].

3.3. Powder X-ray Diffraction

The polycrystalline samples were grounded and filled in glass capillaries to avoid hydrolysis (Hilgenberg GmbH: Marsfeld, Germany, 0.3 mm diameter). The powder patterns were measured at an STOE Stadi-P diffractometer (STOE & Cie GmbH: Darmstadt, Germany, Cu-Kα₁ radiation, Ge(111)-monochromator, Mythen 1k detector (Dectris: Baden, Switzerland)) and the Rietveld refinements of the powder diffraction data were carried out using the single-crystal data as starting parameters using the TOPAS v4.1software [33].

3.4. Solid-State MAS-NMR

The samples were filled into a commercial 2.5 mm zirconium rotor in an argon-filled glovebox. The ²³Na- and ⁷Li-NMR-spectra were measured on a Bruker Avance III 500 (Bruker AXS Inc.: Madison, WI, USA) with a field of 11.74 T under MAS conditions (v_rot = 20 kHz) and Lamor frequencies of v₀(²³Na) = 132.33 Mhz and v₀(⁷Li) = 194.42 MHz. The spectra were indirectly referenced to ¹H in 100% TMS at −0.124 ppm.

3.5. Energy-Dispersive X-ray Spectroscopy

Scanning electron microscopy was carried out with a Carl Zeiss EVO-MA 10 instrument with SE and BSE detector (Carl Zeiss: Oberkochen, Germany), controlled by the SmartSEM v5.07 Beta software [34]. The energy-dispersive X-ray spectroscopy (EDX) measurement was performed with a Bruker Nano EDX detector (Bruker AXS Inc.: Madison, WI, USA), X-Flash detector 410-M) using the QUANTAX 200 v1.9.4.3448 software package [35]. Samples were prepared on carbon pads, oxygen from partial hydrolysis was disregarded.

3.6. Scanning Transmission Electron Microscopy

In an argon-filled glovebox the polycrystalline sample were grounded, and the powder was placed on a copper grid covered with a carbon film (Plano GmbH: Wetzlar, Germany) and mounted on a single-tilt vacuum transfer holder. The sample was transferred to a FEI Titan Themis 300 (Thermo Fischer Scientific: Waltheim, MA, USA) transmission electron microscope equipped with an X-FEG electron source, a Cs DCOR probe corrector, a US1000XP/FT camera system (Gatan: Unterschleissheim, Germany), and a windowless 4-quadrant Super-X energy dispersive X-ray spectroscopy detector. The measurements were performed at an acceleration voltage of 300 kV. For transmission electron microscopy (TEM) images, a 4k × 4k FEI Ceta CMOS camera (Thermo Fischer Scientific: Waltheim, MA, USA) was utilized. Data processing and Fourier filtering were executed with JEMS v3.3.425 (SAED simulations) [36] and Velox v3.0 (STEM images, EDX maps) [37]. After test images and crystal X-ray diffraction scans were acquired, the sample was transferred to a double-tilt holder under air. After the sample was implemented in the transmission electron microscope, the measurement procedure was started.

3.7. Electrochemical Measurements

Electrochemical measurements were carried out under inert conditions in an argon-filled glovebox (O₂ and H₂O < 0.1 ppm). Before measurements, samples were ground in an agate mortar and compacted into pellets with a thickness of about 0.2–0.8 mm and a diameter of 3 mm using uniaxial cold-pressing with a pressure of ~0.75 t. The pellets had relative densities spanning from 74 to 91% for Na_1.36(Si_0.86Ga_0.14)₂As_2.98 and 71–78% for Li_1.5Ga_0.9Si_3.1As₄. The pellets were measured in an RHD Instruments TSC SW closed measuring cell on an RHD Instruments Microcell HC cell stand using an Ivium compactstat.h (24 bit instrument) potentiostat. All measurements were carried out using stainless steel electrodes in a two-electrode, ion-blocking setup. The pressure in the measurement cell was about 700 kPa during measurements. The ionic conductivity σ was determined using EIS. Impedance was measured between 1 MHz and 1 Hz with an excitation voltage of 100 mV. The activation energy was determined from temperature-dependent EIS measurements between 25 and 75 °C in 5 °C steps and an equilibration time of 1 h. The activation energy E_a was extracted by fitting the temperature-dependent data to a linear, modified Arrhenius-type behavior using the equation $\sigma T={\sigma }_0{e}^{(-\frac{E_a}{k_{B}T})}$ with the temperature T, the pre-exponential factor σ₀ and the Boltzmann constant k_B [38]. Data processing and fitting procedures were carried out using the program RelaxIS 3 (RHD Instruments, Darmstadt Germany). The electronic conductivity was determined by potentiostatic polarization measurements. For potentiostatic measurements, voltages of either 0.10, 0.20, 0.30, 0.40 and 0.50 V (Na_1.36(Si_0.86Ga_0.14)₂As_2.98) or 0.05, 0.10, 0.15, 0.20 and 0.25 V (Li_1.5Ga_0.9Si_3.1As₄) were applied for 6 h each with the ohmic drop of the resulting current measured. The resistance was calculated on the basis of Ohm’s law using the current at equilibrium.

4. Conclusions

With Na_1.36(Si_0.86Ga_0.14)₂As_2.98, we show that it is possible to increase the concentration of alkali metal ions in superadamantoide compounds with NaSi₂P₃ type structure. This is achieved when the charge of the supertetrahedra is increased by partial substitution of silicon by gallium. In Na_1.36(Si_0.86Ga_0.14)₂As_2.98, the sodium concentration is increased by 36%, and all sites that are partially occupied in NaSi₂P₃ are now completely filled. This limits the mobility of the sodium ions, and therefore, we measure only low ionic conductivity. However, this approach can potentially lead to a higher ionic conductivity if it is possible to increase the concentration by only 5–10%, so that enough free sites remain for ionic movement. Li_1.5Ga_0.9Si_3.1As₄ is a mixed-valence arsenidosilicate-silicide which contains (Si_1−xGa_x)As₄ tetrahedra with Si⁴⁺/Ga³⁺ and Si_n zigzag-chains with divalent silicon. Lithium ions reside in channels, and ⁷Li-MAS-NMR indicates possible mobility; however, the channels do not provide three-dimensional migration pathways and the ion conduction is likewise low.