3.1. Geometrical Properties and Stability of Perovskite Structures
Table 1 summarizes the calculated lattice constants (
a,
b,
c, α, β, and γ), cell volume (
V) and density (ρ) of
A2AgCrBr
6 (
A = Cs, Rb, K).
Figure 1a shows the relaxed polyhedral model of Cs
2AgCrBr
6, which is analogous with that found for
A2AgCrBr
6 (
A = Rb, K). Because
a =
b =
c and α = β = γ for each of them, each adopts a 3D elpasolite structure with space group Fm-
m.
The lattice constant
a is found to be the largest of 10.68 Å for Cs
2AgCrBr
6, and the smallest of 10.52 Å for K
2AgCrBr
6. The trend in the decrease of
a in the series is consistent with the corresponding decrease in the cell volume
V, with the
V values ranging from 1217.10 Å
3 (Cs
2AgCrBr
6) to 1165.60 Å
3 (K
2AgCrBr
6). These changes are in agreement with the corresponding decrease in ρ, as well as that in the Ag–Cl, Cr–Cl, and
A–Cl bond distances (
Table 2), across the series passing from Cs
2AgCrBr
6 through Rb
2AgCrBr
6 to K
2AgCrBr
6. For comparison, our calculation on Cs
2AgBiCl
6 and Cs
2AgBiBr
6 yielded in lattice constants, density, and volumes that were not only close to experiment (within 2%), but also led to a suggestion that the Cr-substituted compounds are relatively lightweight, thus advocating the reliability of the accuracy of the DFT functional chosen.
The geometrical stability of
A2AgCrBr
6 was examined using the global instability index,
GII, given by:
, where
n is the number of ions and
d is the bond discrepancy factor defined as the deviation of bond valence sum (
BVS) from formal valence [
51,
52].
BVS was calculated using the sum of bond valences (
) around any specific ion given by:
, where
,
lij is a bond length,
l0 is the bond valence parameter empirically determined using experimental room-temperature structure data, and
b is the bond softness parameter. For geometrically stable perovskite structures without steric distortions,
GII equals 0.0 valence unit (v.u.); and for empirically unstable structures,
GII > 0.2 v.u. [
51,
52].
The results of our calculation listed in
Table 1 show that
GII for Cs
2AgCrBr
6, Rb
2AgCrBr
6 and K
2AgCrBr
6 are 0.02, 0.06, and 0.08 v.u., respectively. This means that all the studied systems are associated with very marginal geometrical distortion compared to what might be expected of ideal perovskite structures. The largest instability is observed for K
2AgCrBr
6, which is believed to be due to the small size of the K
+ cation that causes the lattice K
2AgCrBr
6 to contract.
Table 3 lists Shannon’s ionic radii of atoms that were used to understand the geometrical stability of
A2AgCrBr
6. Specifically, we used them for the calculation of the octahedral factor
μ (
μ =
rB/
rX), and Goldschmidt tolerance factor (
). According to numerous previous demonstrations which have appeared in the literature [
48,
49,
50], the combination (
μ and
t) should be suitable to assess the formability of perovskite structures provided
μ and
t values are in the ranges 0.414 <
μ < 0.732 and 0.825 <
t < 1.059, respectively. Our calculation gave a value of 0.45 for
μ for all
A2AgCrBr
6. Similarly, the
t values for these systems were in the range 0.90 <
t < 0.96 (
Table 3). Clearly, the combination (
μ and
t) recognizes the formability of
A2AgCrBr
6 (
A = Cs, Rb, K) as stable perovskites. For comparison, the most studied CH
3NH
3PbI
3 was reported to have a
t of 0.91 (unstable with respect to tilting), whereas NH
4PbI
3 has a
t of 0.76 [
79]. For the latter case, an alternative non-perovskite structure was suggested [
79].
Table 3 also includes the
τ (a newly proposed tolerance factor [
39]) values for
A2AgCrBr
6. These were calculated using the relationship given by
, where
nA is the oxidation state of
A,
ri is the ionic radius of ion
i, and
rA >
rB by definition. The proposal to evaluate the geometrical stability of single and double perovskites (
ABX3 and
A2B′
B″
X6, respectively) using
τ (
τ < 4.18) was emerged after it was being realized that the recommended ranges for
μ and
t do not guarantee the formation of the perovskite structure since they give a high false-positive rate (51%) in the regions of
t (0.825 <
t < 1.059) and
μ (0.414 <
μ < 0.732). In fact,
τ was used to generalize outside of a 1034 training set of experimentally realized single and double perovskites (91% accuracy), and that its application was also useful to identify 23,314 new double perovskites ranked by their probability of being stable as perovskite [
39]. Based on the criterion that
τ < 4.18 for geometrically stable perovskites, we found that
A2AgCrBr
6 (
A = Cs, Rb) are a set of such geometrically stable perovskites (4.04 <
τ < 4.14), and K
2AgCrBr
6 (
τ = 4.22) is a partially unstable structure (a non-perovskite!), where the ionic radius of
B was calculated as the arithmetic mean of the ionic radii of
B′ and
B″. These results are not in exact agreement with that inferred from
GII or the combination of
μ and
t values (see
Table 1 and
Table 3). This may either mean that the value of
τ (
τ < 4.18 [
39]) recommended for identifying unknown perovskite structures may be stringent, causing the rejection of K
2AgCrBr
6 as a stable perovskite, or both the
GII and the combination of
μ and
t mislead the stability features.
To understand the effect of halogen substitution at the X site of
A2AgCrX
6, we have carried out similar calculations only for Cs
2AgCrI
6 and Cs
2AgCrCl
6. The calculated lattice constants, cell volume, and density were 11.48 Å, 1511.9 Å
3, and 5.2 gcm
−3 for Cs
2AgCrI
6, respectively. These were 10.13 Å, 1040.2 Å
3, and 4.08 gcm
−3 for Cs
2AgCrCl
6, respectively. In other words, the halogen substitution does not cause significant strain in the lattice that can lead to the change in the symmetry of the system from Fm-
m. However, the replacement of the Br atoms in Cs
2AgCrBr
6 by the Cl and I anions has indeed resulted in lattice contraction and expansion, respectively, which are expected to affect the electronic, transport, and optical properties of the resulting systems (
vide infra). The presence of lattice contraction and expansion is also evident of metal-halide and alkali-halide bond lengths that are decreasing across the series in this order: Cs
2AgCrCl
6 < Cs
2AgCrBr
6 > Cs
2AgCrI
6. (A comparison detail of the Ag–X, Cr–X, and Cs–X bond distances for Cs
2AgCrX
6 (X = Cl, Br, I) is given in
Table 2).
Moreover, the calculated
μ and
t values were 0.40 and 0.94 for Cs
2AgCrI
6 (
Table 3), respectively. Based on the traditional arguments [
48,
50], one may not disapprove to call Cs
2AgCrI
6 a stable perovskite. The same argument applies to Cs
2AgCrCl
6, as the combination of the
μ and
t values (
Table 3) favors a perovskite structure. This result is consistent with that inferred from the
GII values (0.10 v.u. for Cs
2AgCrCl
6 and 0.00 v.u. for Cs
2AgCrI
6). However, our calculated
τ value of 4.31, far exceeding its recommended value of 4.18, rejects Cs
2AgCrI
6 as a stable perovskite. This is analogously as
τ rejected the K
2AgCrBr
6 system to be called a perovskite. However, this is not the case with Cs
2AgCrCl
6 (
τ = 3.87), in agreement with that reported recently [
39]. Clearly, the aforesaid mismatch between the results associated with different indices suggests that the geometrical properties of a large body of
AB′B″X
6 systems of different B′ and B″compositions, together with X = I and Br and
A = K, need to be analyzed to reach any definitive conclusion on the predictability of
GII- and
τ-based stability features.
3.2. Density of States and Band Structures
Figure 2a–g shows the orbital-projected partial density of states of each atom type, and
Figure 2h shows the atom projected partial density of states for Cs
2AgCrBr
6. From these, it may be said that the dispersion of the HOMO band (VBM) results collectively from contributions of e
g and t
2g states of both Ag (4d) and Cr(3d), and the 4p states of Br. This is also reminiscent of the data provided in
Table 4, in which, the VBM is mainly due to the 4p states of Br (88%), and that the contribution from the 3d (Cr) and 4d (Ag) states is very small.
On the other hand, the LUMO band (CBM) of Cs
2AgCrBr
6 is predominantly of Cr (3d) character. It is substantially composed of the e
g orbital energy states (~72.5%). This, together with the small contributions from the Ag (5s) and Br (4p) states, causes the dispersion of the band. Note that alkali substitution at the
A site has very little effect on the orbital characters of VBM and CBM. However, when the Br atom at the X-site was replaced by the I and Cl atoms, this had a very marginal effect on the orbital character of the CBM, and significant effect on the VBM. In particular, the 3d (Cr) energy states dominate below the Fermi level for Cs
2AgCrCl
6 (46%), which was comparatively larger than those arose from the 3p (35.2%) and 4d (17.0%) states of Cl and Ag, respectively. For Cs
2AgCrI
6, the I(5p) states dominate below the Fermi level, and the 3d (Cr) and 4d (Ag) states contribute marginally to the VBM.
Figure 3 compares the nature of band dispersion of Cs
2AgCrBr
6 with Cs
2AgCrI
6, and
Figure 4 illustrates the orbital- and atom-projected partial density of states for Cs
2AgCrI
6.
From the bandgap (E
g) data listed in
Table 1, it is apparent that the
A2AgCrBr
6 systems are an indirect bandgap material. This is arguably because the VBM is located at the high symmetry L-valley (VBM) and the CBM is located at the Γ-valley (CBM); see
Figure 3a for Cs
2AgCrBr
6. The calculated bandgaps (E
g) of
A2AgCrBr
6 were found between 1.27 and 1.29 eV (
Table 1), thereby suggesting the semiconducting nature of these materials. Small variations in E
g across the alkali series indicate that alkali substitution at the
A-site of
A2AgCrCl
6 does not significantly affect the gap between CBM and VBM, and is consistent with the nature and extent of orbital characters that were involved in the formation of these bands (
Table 4).
By contrast, Cs
2AgCrI
6 has a bandgap of 0.43 eV, and is direct at the Γ-valley (
Figure 3b). Obviously, the replacement of the Br atoms of Cs
2AgCrBr
6 with iodine atoms not only causes the shift of the VBM from the L-valley to appear at the Γ-valley, but also significantly narrows the gap between the CBM and VBM and alters the character of the electronic transition between them. This is not the case with Cs
2AgCrCl
6 since the bandgap for this system is increased giving rise to an E
g value of 1.84 eV (SCAN +
rVV10) and was indirect between the L-valley (VBM) and Γ–valley (CBM). These results lead to a conclusion that halogen substitution from Cl through Br to I at the X-site of Cs
2AgCrX
6 not only triggers the nature of the bandgap from indirect to direct, but also adjusts its magnitude from 1.84 eV to 0.43 eV.
In any case, the charge transport properties of any semiconducting material are strongly dependent on the effective masses of the charge carriers [
80,
81,
82].
Table 5 summarizes the effective masses of conduction electrons and holes for
A2AgCrBr
6, which were calculated by parabolic fitting of the lower conduction band and the upper valence band centered at the Γ- and L-valleys, respectively. Since the HOMO band is formed by several spin-up and spin-down channels (total eight) caused by the Cr
3+ ions, there are spin-up and spin-down holes along the crystallographic directions. The most important of these is the L→Γ direction, which is linked directly with the electronic transition between the VBM and CBM. As can be seen from
Table 5, the effective masses of the spin-up holes (
mh*(
up)) are always lighter than that of the spin-down holes (
me*(
down)), regardless of the nature of the
A-site in
A2AgCrBr
6. This means that (
mh*(
down)) would play an insignificant role in determining the transport phenomena that are usually governed by the band structures at and around the close vicinity of the Γ-point. In contrary, there are only spin-up electrons that are associated with the bottom of the conduction band. They are indeed small (
me*(
up) (values lie between 0.50 and 0.62
m0 for all directions), but are not always heavier than those of
mh*(
up). These results suggest that the studied systems might be suitable as hole (and electron) transporting materials due to their high mobility [
81,
82,
83]. A similar result was found for Cs
2AgCrCl
6 (values not given).
Table 6 collects the
me*(
up) and
mh*(
up) values for Cs
2AgCrI
6. Of the three, two channels associated with the top of the valence band are having significantly smaller
mh*(
up) values along both the crystallographic directions (Γ→L and Γ→X) compared to that of the third one.
me*(
up) is always smaller than
mh*(
up) regardless of the nature of crystallographic directions, which is obviously due to the ‘flatter’ nature of the VBM compared to the parabolic CBM (
Figure 3b).
For comparison, Wang et al. have reported unusually heavier effective masses for holes for the hybrid perovskite series, AEQTBX
4 (B = Pb, Sn; X = Cl, Br, I; AEQT= H
3NC
2H
4C
16H
8S
4C
2H
4NH
32+), with the
mh* values between 0.63
m0 ((AEQT)SnI
4) and 105.21
m0 ((AEQT)PbI
4) [
84]. (AEQT)SnI
4 was observed to exhibit dispersive VBM and CBM, and was accompanied by a moderate fundamental bandgap of 2.06 eV involving a strong direct valence band to conduction band transition. This, together with the relatively light effective masses for electrons and holes (~0.6
m0), and high dielectric constants, has led to a suggestion that this system is suitable for application as a top absorber of the tandem solar cell.
3.3. Optical Properties
The excitonic effect is generally approximated by solving the Bethe–Salpeter equation for the two-body Green’s function, without or with considering local field effect [
85]. Due to the high computational cost and limited computed resources, the frequency-dependent complex dielectric function was calculated within the framework of DFPT without taking into account the electron-hole coupling effect. Studies have shown that neglecting electron-hole coupling can yield reasonable results for semiconductors with small band gaps [
86,
87]. Nevertheless,
Figure 5 shows the plot of the real and imaginary parts of the dielectric function as a function of photon energy for
A2AgCrBr
6. The first peak of the
ε2 curve is located at an energy higher than that of the
ε1 curve, and alkali substitution at the
A-site causes a slight fluctuation in the high frequency dielectric behavior. For instance, these peaks on the
ε1 and
ε2 curves are positioned at energies of 1.45 and 1.63 eV for Cs
2AgCrBr
6, respectively. These are 1.53 and 1.72 eV for Rb
2AgCrBr
6, respectively, and are 1.55 and 1.74 eV for K
2AgCrBr
6, respectively. Similarly, the low frequency limit of the isotropically averaged value of
ε1 (
ω = 0) =
ε∞ (called optical dielectric constant or high-frequency dielectric constant) is found to be 6.1 for Cs
2AgCrBr
6, 5.4 for K
2AgCrBr
6, and 5.8 for K
2AgCrBr
6. This result indicates that the contribution of electrons to the static dielectric constant is appreciable and that the studied systems contain ionic bonds. Zakutayev et al. [
88], as well as others [
84], have previously demonstrated that perovskites with appreciable dielectric constants are defect tolerant. For comparison, MAPbI
3 (X = Cl, Br, I), and other perovskite systems [
73] were reported to have virtually similar optical dielectric constants (viz.
ε∞ = 6.5 for MAPbI
3, 5.2 for MAPbBr
3, 4.2 for MAPbCl
3, and 5.3 for CsPbI
3).
The magnitude of the bandgap is the determinant of the onset of first optical absorption, which can be approximated using the spectra of
ε2. As such, the onset of absorption is approximately lying between 1.15 and 1.40 eV for
A2AgCrBr
6 (
Figure 5), showing an appreciable absorption mostly in the infrared spectral region. Note that the bandgaps of
A2AgCrBr
6 evaluated using SCAN +
rVV10 were lying between 1.27 and 1.29 eV, and the dielectric functions were calculated using PBEsol in conjunction with DFPT, yet there is a close match between bandgap and onset of optical absorption evaluated using the two different theoretical approaches employed.
Figure 6a,b compares the energy dependence of
ε1 and
ε2 for Cs
2AgCrX
6 (X = Cl, Br, I), respectively. From either of the two plots, it is apparent that halogen replacement has a significant effect not only on the magnitude of
ε∞, but also on the onset of optical absorption. For instance, the value of
ε∞ inferred from
Figure 6a is approximately around 4.5, 6.1, and 8.9 for Cs
2AgCrCl
6, Cs
2AgCrBr
6, and Cs
2AgCrI
6, respectively. These indicate that Cs
2AgCrI
6 has a greater ability to screen charged defects compared to Cs
2AgCrBr
6 and Cs
2AgCrCl
6 [
89,
90]. The observed trend in the increase of
ε∞ caused by halogen substitution at the X-site in Cs
2AgCrX
6 is consistent with that reported for the MAPbX
3 (X = Cl, Br, I) series [
73].
The onset of optical absorption corresponding to the
ε2 spectra is positioned around 1.55 eV for Cs
2AgCrCl
6. This is significantly shifted from the near-infrared region to the low energy infrared region of the electromagnetic spectrum caused by the halogen substitution at the X-site, thus showing up at energies around 1.15 eV and 0.45 eV for Cs
2AgCrBr
6 and Cs
2AgCrI
6, respectively. These values are close to their corresponding bandgaps predicted using the SCAN +
rVV10 functional. It is worth noting that we did not calculate the contribution of dielectric constant due to ions to the static dielectric constant, thus it is not possible to provide views on the polarity of the chemical bonds and the softness of the vibrations [
91], as well as the nature and importance of (picosecond) response of lattice vibrations (phonon modes) that explain the ionic and lattice polarizations, and the extent to which the material would control the photovoltaic performance [
74,
89,
90,
92].
In any case, the major peaks in the
ε2 curve of Cs
2AgCrCl
6 are positioned at energies of 2.0 and 2.3 eV, whereas the onset of the optical absorption was located at an energy of 1.55 eV (
Figure 6b). These peaks in the UV/Vis/NIR diffuse reflectance spectrum were experimentally observed at energies of 1.6 and 2.2 eV for hexagonal Cs
2AgCrCl
6, respectively, and were assigned to the d-d transitions of
4A2→
4T2 and
4T2→
4T1, respectively. This is not surprising given that the ground state of the Cr
3+ (t
2g3e
g0) cation in an octahedral O
h symmetry is described by
4A2g(
F), and is accompanied with spin-allowed transitions from the
4A2g(
F) ground state to the excited
4T2 and
4T1 states (that is,
4T2→
4T1 and
4A2→
4T1). Such d–d transitions were reported to occur at energies of 3.06 eV (405 nm) and 2.25 eV (550 nm) in the UV–vis spectrum of isolated Cr
3+ complex of aspartic acid, respectively [
93]. The intense and broad bands observed at energies of 2.93 eV and 2.13 eV for strontium formate dehydrate crystal of Cr
3+ were attributed to the corresponding transitions, respectively [
94]. Also, zinc-tellurite glasses doped with Cr
3+ ion feature two intense broad bands with maxima at 454 nm (2.73 eV) and 650 nm (1.91 eV), which were assigned to the transitions
4A2 →
4T1 and
4A2 →
4T2, respectively [
95]. In Cs
2AgInCl
6:Cr
3+ halide double perovskite, the absorptions peaks at 1.55 eV (800 nm) and 2.19 eV (565 nm) were assigned to the same spin-allowed transitions
4A
2 →
4T
2 and
4T2→
4T1, respectively. Because the VBM is dominated with Br(4p) and I(5p) energy states for Cs
2AgCrBr
6 and Cs
2AgCrI
6, respectively, and the contribution of Cr(3d) orbitals states to the VBM is reasonably small, the two electronic transitions appear in the
ε2 spectra in the region 0.3–2.2 eV were shifted toward the low energy regions (
Figure 6b).
Figure 7 displays the plot of the absorption coefficient, photoconductivity, and complex refractive index for
A2AgCrBr
6. Equations (4)–(7) were used, respectively.
Figure 7b shows the Tauc plot. These plots replicate the oscillator peaks of the dielectric function
ε2 in the region 0.0–3.0 eV (
Figure 6). From
Figure 7a, it is obvious that α for
A2AgCrBr
6, which determines the extent to which light of a particular energy can penetrate a material before it is absorbed [
96], is non-zero (positive) both in the infrared and visible regions. As can be seen, different regions have different absorption coefficients. This may not be unusual given that the absorption coefficients of MAPbI
3 as low as 10
–14 cm
–1 were detected at room temperature for long wavelengths, and were 14 orders of magnitude lower than those observed at shorter wavelengths [
97]. Nevertheless, α increases as the light energy increases and becomes a maximum at the strongest peak occurred at an energy around 1.7 eV for Cs
2AgCrBr
6. The trend in α in the series
A2AgCrBr
6 follows the order: Cs
2AgCrBr
6 (3.3 × 10
5 cm
−1) < Rb
2AgCrBr
6 (3.6 × 10
5 cm
−1) ≈ K
2AgCrBr
6 (3.6 × 10
5 cm
−1). These are somehow larger than those of 0.5 × 10
4 cm
−1 and 1.5 × 10
4 cm
−1 observed at 1.77 eV (700 nm) and 2.25 eV (550 nm) for MAPbI
3, respectively [
98,
99].
The nature of the curves of α is similar to that found for σ at the strongest peak positions in the region 0.0–2.5 eV (see
Figure 7a vs.
Figure 7c). The value of σ at these peaks is 2.64 × 10
9 Sm
−1 for Cs
2AgCrBr
6, 2.76 × 10
9 Sm
−1 for Rb
2AgCrBr
6, and 2.84 × 10
9 Sm
−1 for K
2AgCrBr
6. For comparison, the average value of σ for MAPbI
3 thin films was reported to be 6 × 10
5 Sm
−1 [
100].
The above trend both in α and σ is not exactly similar to that found for the static refractive index
n (ω = 0), with the latter were around 2.5, 2.3, and 2.4 for Cs
2AgCrBr
6, Rb
2AgCrBr
6, and K
2AgCrBr
6, respectively (
Figure 7d). The
n(ω) values increase with respect to the increase of photon energy in the region 0.0–3.0 eV. The maximum of
n at the highest peak in the region 0.0–3.0 eV varies between 3.0 and 3.2 for
A2AgCrBr
6.
Cl and I substitutions at the X-site in Cs
2AgCrX
6 has resulted in a relatively smaller α of 2.4 × 10
5 cm
−1 for Cs
2AgCrI
6 and a relatively larger α of 4.0 × 10
5 cm
−1 for Cs
2AgCrCl
6; all within the range 0.0–2.5 eV (
Figure 8a). Although the peaks in the α spectrum of Cs
2AgCrCl
6 are relatively sharper than those of Cs
2AgCrBr
6 and Cs
2AgCrI
6, the different absorption edges found for the three systems are consistent with the corresponding bandgap properties. Also, the trend in α is concordant with the conductivity spectra shown in
Figure 8b, in which, σ decreases in the series in this order: Cs
2AgCrI
6 < Cs
2AgCrBr
6 < Cs
2AgCrCl
6. On the other hand, the
n (
ω = 0) values were 2.1, 2.5, and 3.0 for Cs
2AgCrCl
6, Cs
2AgCrBr
6, and Cs
2AgCrI
6, respectively (
Figure 9a). The maximum values of
n were approximately 2.9, 3.2, and 3.4 for the corresponding systems, respectively, a feature which is consistent with the trend in the optical dielectric constants (see above). Regardless of the nature of the
A2AgCrX
6 systems examined, the extinction coefficient (imaginary part of the refractive index) was found to be very small (
Figure 7d and
Figure 9b), and is expected for semiconducting materials [
101]. These results lead to a meaning that the studied Cs
2AgCrX
6 (X = Cl, Br, I) perovskites may result in relatively larger reflection at the perovskite/electrode interface. For comparison, the refractive indices of CH
3NH
3PbI
3–xCl
x perovskite thin films were reported approximately to be 2.4 and 2.6 in the visible to near-infrared wavelength region [
101]. He et al. have examined bulk single crystals of CH
3NH
3PbX
3 (CH
3NH
3 = MA, X = Cl, Br, I) and have measured refractive indices that rank in this order: MAPbI
3 > MAPbBr
3 > MAPbCl
3 at the same wavelength [
102]. These results indicate that
n may play an important role in determining the optical response properties of halide perovskites, which may serve as a useful metric in the parameter space for use in optoelectronic device design.