Localized Spin Dimers and Structural Distortions in the Hexagonal Perovskite Ba3CaMo2O9.

Extended solid-state materials based on the hexagonal perovskite framework are typified by close competition between localized magnetic interactions and quasi-molecular electronic states. Here, we report the structural and magnetic properties of the new six-layer hexagonal perovskite Ba3CaMo2O9. Neutron diffraction experiments, combined with magnetic susceptibility measurements, show that the Mo2O9 dimers retain localized character down to 5 K and adopt nonmagnetic spin-singlet ground states. This is in contrast to the recently reported Ba3SrMo2O9 analogue, in which the Mo2O9 dimers spontaneously separate into a mixture of localized and quasi-molecular ground states. Structural distortions in both Ba3CaMo2O9 and Ba3SrMo2O9 have been studied with the aid of distortion mode analyses to elucidate the coupling between the crystal lattice and electronic interactions in 6H Mo5+ hexagonal perovskites.

Introduction

Transition metal oxides (TMOs) display a wealth of exotic physical phenomena due to the intrinsic coupling between their lattice and electronic degrees of freedom. In recent years, extended solid-state TMOs containing closely spaced transition metal atoms have attracted widespread interest for their ability to form clustered quasi-molecular (QM) states.[1] Typically, these “orbital molecules” manifest as a result of orbital ordering and direct bonding between the metal atoms.[2] A pertinent example is found in the naturally occurring mineral magnetite (Fe3O4), in which subtle contractions in the Fe2+–Fe3+ distances accompany the formation of three-site Fe3+–Fe2+–Fe3+ bonded trimerons.[3,4] In general, such metal–metal bonded clusters arise when the geometrical connectivity between neighboring atoms imposes considerable overlap between the metal d orbitals; in particular, materials with edge-sharing or face-sharing polyhedra are promising in this regard,[5] with other notable examples including TMOs such as VO2,[6,7] Li2RuO3,[8,9] and Y2Mo2O7.[10] Orbital molecules are often formed in tandem with metal–insulator transitions and are characterized by spin-gapped magnetic excitations. As such, orbital molecules exhibit fundamentally interesting physical properties that may hold future applications in spintronic or orbitronic technologies.

In recent years, orbital molecules have been proposed to emerge in hexagonal perovskites of the form Ba3B′Ru2O9 (B′ = Na;[11] Y, In, and Lu;[12] and Ce[13]). These materials crystallize in a six-layer hexagonal (6H) structure consisting of corner-sharing B′O6 octahedra and face-sharing M2O9 bioctahedral dimers (Figure a). The M2O9 dimers are characterized by short intradimer M–M distances that favor electronic interactions between the metal atoms. When M is a 4d/5d transition metal (such as Ru or Ir), the spatially diffuse t2g orbitals can overlap, and with sufficiently weak Hund’s coupling, direct M–M bonds can then form between the metal atoms.[14,15] However, some Ba3B′M2O9 compositions retain localized magnetic moments on the metal atoms so that, in this case, the M2O9 dimers are better described as unbonded cluster magnets (rather than orbital molecules).[16−19] This is due to the strong competition between Hund’s coupling and M–M bonding in a face-sharing polyhedral geometry.[20,21] Such geometries are idiosyncratic motifs of hexagonal perovskites, so these materials present a particularly promising phase space in which to discover and study new orbitally molecular species.

Figure 1

Crystal structures of Ba3CaMo2O9 at (a) 290 K (space group: P63/mmc) and (b) 5 K (space group: Pnma). Projections of the P63/mmc and Pnma structures down the long axes are provided in panels (c) and (d), respectively.

We have recently reported an unprecedented mechanism of electronic phase separation (EPS) in the 6H-perovskite Ba3SrMo2O9.[22] A single hexagonal P63/m phase is observed at room temperature (P63/m-Ba3SrMo2O9), but upon cooling below 230 K, a new and electronically distinct monoclinic P21/m phase emerges (P21/m-Ba3SrMo2O9) and coexists with the primary phase down to 1.6 K. P63/m-Ba3SrMo2O9 features localized Mo2O9 spin dimers, while P21/m-Ba3SrMo2O9 contains a 50:50 mixture of Mo–Mo bonded Mo2O9 orbital molecules and localized spin dimers. The observed phase separation appears to have an electronic origin related to competition between Mo–O–Mo superexchange and direct Mo–Mo bonding. This contrasts established electronic materials such as colossal magnetoresistant manganite perovskites, where EPS appears to depend on the presence of some chemical disorder.[23,24] 6H-Ba3B′Mo2O9 compositions hence comprise a promising new system in which to study how unusual electronic states emerge from competing orbital and magnetic interactions.

Here, we report the synthesis and characterization of the new 6H-perovskite Ba3CaMo2O9. Our neutron scattering experiments reveal that Ba3CaMo2O9 undergoes a structural phase transition near 200 K, but, unlike Ba3SrMo2O9, no phase separation is observed down to 5 K. Magnetic susceptibility measurements, combined with our neutron scattering experiments, show that the Mo2O9 dimers adopt a nonmagnetic ground state but do not appear to have QM character. The diffraction data of both Ba3CaMo2O9 and Ba3SrMo2O9 have been further probed by distortion mode analysis to explore the structural features that dictate orbital molecule formation in this structure type.

Experimental Section

Polycrystalline samples of Ba3CaMo2O9 were synthesized via a standard solid-state reaction. Stoichiometric quantities of BaCO3 (99.999%, Aldrich), CaCO3 (Purity, Aldrich), and MoO3 (99.5 +%, Aldrich) were ground and pressed into 13 mm pellets before being heated in air to 900 °C for 10 h. The pellets were then reground and repressed before heating under 5% H2/N2 at 1150 °C for 20 h, with one intermittent regrinding after 10 h.

X-ray diffraction (XRD) measurements were recorded on a Panalytical EMPYREAN diffractometer equipped with a Johansson monochromator (Cu Kα1 radiation, λ = 1.5406 Å). Scans were recorded for 1 h in the range of 10–100° with a 0.013° step size. High-resolution neutron powder diffraction (NPD) data were collected on a D2B diffractometer at the Institut Laue-Langevin (ILL) in Grenoble (λ = 1.59432 Å). The sample was placed in an 8 mm vanadium can and measured on heating at selected temperature intervals between 5 and 290 K. Scans were recorded for 2 h at each temperature. High-intensity NPD data were also collected on the D20 diffractometer at the ILL (λ = 2.40 Å) for 1.5 h at 5, 150, and 290 K to check for long-range magnetic ordering.

Rietveld refinements[25] were performed using the GSAS/EXPGUI software.[26,27] The background was modeled using a Chebyschev polynomial with 11 background coefficients, while peak profiles were modeled using a pseudo-Voigt function. Distortion mode analyses were performed using the ISODISTORT[28,29] and AMPLIMODES[30] programs.

DC magnetization measurements were performed on a Quantum Design SQUID magnetometer between 2 and 400 K. Data were collected after zero-field-cooling (ZFC) the sample in an applied field of 50 Oe. Heat capacity measurements were recorded on a Quantum Design PPMS from 2 to 300 K in 5 K intervals. DC resistivity measurements were also performed on a PPMS using the four-probe method on a polycrystalline bar between 4 and 300 K.

Results

Rietveld refinement from high-resolution XRD and NPD data shows that Ba3CaMo2O9 crystallizes in the hexagonal P63/mmc space group at room temperature (RT). The fit to the P63/mmc model at 290 K from NPD data collected on D2B is shown in Figure S1, while selected crystallographic data are reported in Table S1. An excellent fit to the P63/mmc structural model (Figure a) is observed, and there is no sign of any impurity phases. The P63/mmc model is the typical crystal structure adopted by 6H-Ba3B′M2O9 compositions. Ba3CaMo2O9 then differs from the B′ = Sr analogue, which instead crystallizes in the P63/m space group at 290 K due to in-phase tilting of the SrO6 octahedra about the hexagonal c axis.[22] The atomic site occupancies could be refined to within ±1% of their nominal values and were thus fixed at 1 for the remainder of the refinements. There is no evidence of chemical disorder between the Ca and Mo sites; additionally, the large difference in ionic radii between Ca2+ and Mo5+ (1.00 Å vs 0.61 Å, respectively[31]) makes this highly unlikely. Selected bond distances and angles at 290 K are reported in Table S2. Bond valence sum (BVS) calculations gave a value of BVS(Mo) = 5.11(1), which is in good agreement with the nominal +5 oxidation state.

Variable-temperature NPD measurements recorded on D2B confirmed that Ba3CaMo2O9 retains the P63/mmc structure down to 230 K. Below 200 K, a structural phase transition occurs as evidenced by the appearance of new peaks in the diffraction pattern. It was not possible to obtain a good fit to the previously reported structural models for 6H-perovskites, such as Cmcm for Ba3CoRu2O9[16] or C2/c for Ba3CaIr2O9.[32] Attempts to perform Rietveld fits with all direct subgroups of P63/mmc were also unsuccessful. However, very good fits were obtained to two structural models described by the Pnma and P21/m space groups (Rwp = 4.34 and 4.28% for Pnma and P21/m, respectively, at 5 K). We found no evidence of any intermediate phase transition that might favor one of these two options, and close examination of the diffraction patterns found that the P21/m model offered no tangible improvement to the peak positions, intensities, or profiles. Ultimately then, Pnma was selected as the correct space group assignment.

The Rietveld fit and crystallographic data for the Pnma structural model at 5 K are provided in Figure and Table S3. Selected bond lengths and angles for the Pnma phase at this temperature are reported in Table S4. The Pnma structure (Figure b) accounts for tilting of the CaO6 octahedra along the orthorhombic [001] and [101] directions but does not generate any additional Ca or Mo crystallographic sites. This is in contrast to P21/m-Ba3SrMo2O9, in which there are two crystallographically inequivalent Sr and Mo sites. The Pnma structure is retained down to 5 K.

Figure 2

Rietveld fit to the Pnma structural model from NPD data collected on D2B at 5 K for Ba3CaMo2O9.

No magnetic diffraction peaks were detected on the high-intensity D20 diffractometer down to 5 K. Ba3CaMo2O9 was too resistive to measure from our DC resistivity measurements, showing that it is an insulator below 300 K. Heat capacity measurements confirmed that the structural transition occurs at 220 K (Figure S2). Variable-temperature magnetic susceptibility (χ(T)) measurements (Figure ) corroborated the lack of long-range magnetic ordering down to 4 K so that the structural transition at 220 K is not tied to magnetic ordering of the Mo5+ moments. At low temperatures, a paramagnetic Curie tail is observed, while above 44 K, χ(T) weakly increases as a function of temperature. A small Curie constant of 3.55(2) × 10–3 emu K Oe–1 mol–1 was extracted from an initial Curie–Weiss fit of the data below 44 K (shown in Figure a); we attribute this to minor magnetic impurities or structural defects in the sample. A subtraction of the Curie–Weiss contribution from χ(T) shows that the residual magnetic contribution tends to zero with cooling. This response is reminiscent of spin-gapped materials such as La4Ru6O19[33] or SrCu2O3[34] and shows that Mo2O9 dimers in Ba3CaMo2O9 have a spin-gapped nonmagnetic ground state.

Figure 3

(a) Variable-temperature magnetic susceptibility (χ(T)) measurements of Ba3CaMo2O9. The initial Curie–Weiss (C-W) fit is shown by the solid black line, and its subtraction from χ(T) is shown in purple. (b) Fit to the spin-gap function listed in the main text, with additional temperature-independent (χ0) and C-W terms.

Prior to the fit, a diamagnetic correction was performed. To fit the full susceptibility response χ(T), we used a modified spin gap expression of the form χSG = a(T) exp(−Δ/kBT)[35] (Figure b). This expression accounts for spin-gapped Mo2O9 dimers in which there is a gap between a nonmagnetic singlet state and a triplet excited state. Similar functions have been used to model spin gaps in previous 6H-perovskites such as Ba3BiRu2O9[36,37] and Ba3BiIr2O9.[38] In Ba3CaMo2O9, the spin gap expression—including additional temperature-independent and C-W terms—gives an excellent fit to χ(T) up to 320 K. From the fit, the following parameters were extracted: χ0 = 3.571(1) × 10–4 emu Oe–1 mol–1, Cimp = 3.35(1) × 10–3 emu K Oe–1 mol–1, θ = −0.23(1) K, a = 6.54(1) × 10–3 emu Oe–1 mol–1 K–0.5, and Δ/kB = 233(1) K.

Discussion

The structural and physical properties of Ba3CaMo2O9 are surprisingly different from Ba3SrMo2O9. Most notably, no phase separation is observed in Ba3CaMo2O9 down to 5 K as demonstrated by our high-resolution diffraction measurements. The Mo2O9 dimers are described well by a simple model of spin-gapped dimers, and there is no pronounced asymmetry in the χ(T) curve that would suggest that a more complex mixture of dimers forms (as previously observed for Ba3SrMo2O9). The properties of the spin dimers do not appear to be affected by the structural transition near 200 K, and there is also no marked lattice contraction (Figure S3). Therefore, the spin dimers in Ba3CaMo2O9 are not strongly coupled to the crystal lattice.

Figure compares the Mo–Mo distances in Ba3SrMo2O9 and Ba3CaMo2O9. Close inspection of the Mo–Mo distances in the Ca analogue (Figure S4) reveals that no Mo–Mo bond forms upon cooling. The Mo–Mo separation in Ba3CaMo2O9 is 2.523(3) Å at 290 K and remains largely similar with cooling so that the Mo–Mo separation is 2.527(3) Å at 5 K. This is in contrast to the QM Mo2O9 dimers in P21/m-Ba3SrMo2O9, where the Mo–Mo distances contract by ∼4% below 230 K due to the formation of a Mo–Mo bond. The Mo t2g orbitals in Ba3CaMo2O9 are hence too far apart to hybridize and form a quasi-molecular electronic state. Competition between Mo–Mo bonding and Mo–O–Mo superexchange is therefore less pronounced in Ba3CaMo2O9 than in Ba3SrMo2O9.

Figure 4

Comparison of the Mo–Mo distances in Ba3B′Mo2O9 (B′ = Ca and Sr22). “AFM” and “QM” refer to the antiferromagnetic spin dimers and quasi-molecular Mo2O9 clusters, respectively, in Ba3SrMo2O9.

Due to the differing structural and magnetic properties of Ba3CaMo2O9 and Ba3SrMo2O9, we explored sources of structural coupling in these materials that may influence whether localized spin dimers are obtained over quasi-molecular clusters. We performed distortion mode analyses using the ISODISTORT[28,29] and AMPLIMODES[30] programs to examine the behavior of structural degrees of freedom in both compositions. Distortion modes comprise symmetry-adapted linear combinations of irreducible representations that relate a high-symmetry crystal structure to a lower-symmetry counterpart. In particular, distortion modes describe symmetry-related sets of collective atomic displacements that represent distinct structural degrees of freedom (for example, octahedral rotations or ferroelectric polarizations). Each mode is quantified by an amplitude that describes the magnitude of the deviation from the high-symmetry structure and denotes its contribution to the total distortion observed. Hence, the advantage of this approach is that it can identify and characterize distinct structural degrees of freedom that produce a structural distortion. For Ba3CaMo2O9, the amplitudes of each distortion mode were derived and normalized with respect to the parent P63/mmc unit cell. Three distortion modes are active in the low-temperature Pnma structure, corresponding to the irreducible representations Γ1+, Γ5+, and Μ2+. Their amplitudes are shown in Figure a, while the effects of the Γ5+ and Μ2+ modes on the crystal structure are depicted in Figure b and Figure c, respectively. The Γ1+ mode is a minor strain mode resulting from the overall compression of the unit cell along c with cooling. The Γ1+ mode shows no abrupt variation across the structural transition and is not associated with any octahedral tilting distortion. The Γ5+ mode consists of an in-phase tilting distortion of the CaO6 octahedra about the orthorhombic [001] direction, while the Μ2+ mode consists of an out-of-phase tilt of the CaO6 octahedra about the [101] direction. Overall, these tilting modes serve to “buckle” the Mo2O9 dimers; despite this disruption to the local MoO6 octahedral environment below the structural transition, the lack of any accompanying change in the magnetic susceptibility (Figure b) shows that the magnetic properties of the Mo2O9 dimers are largely insensitive to the structural distortion. As this distortion does not appear to be tied to any specific electronic instability, it is instead driven by purely structural fluctuations. This substantiates the lack of Mo–Mo bond formation at the transition and ultimately reflects the localized character of the Mo2O9 spin dimers. This is a notable contrast to Ba3SrMo2O9, where distinct changes in the magnetic susceptibility can be identified at the phase separation temperature[22] so that the magnetic and electronic properties of the dimers are more sensitive to the distortion of the local MoO6 environment.

Figure 5

Distortion mode analysis of the structural transition to the Pnma phase in Ba3CaMo2O9. (a) Variable-temperature mode amplitudes, normalized with respect to the parent P63/mmc structure; error bars are too small to be visible. (b) Γ5+ distortion mode, corresponding to an in-phase rotation of the CaO6 octahedra about the orthorhombic [001] direction. (c) Μ2+ distortion mode, corresponding to an anti-phase rotation of the CaO6 octahedra about the orthorhombic [101] direction. The arrows in panels (b, c) highlight the rotations of the CaO6 octahedra and Mo2O9 dimers.

In Ba3CaMo2O9, the Μ2+ mode forms the dominant contribution to the observed distortion. Analysis of the group–subgroup relations between the P63/mmc and Pnma space groups confirms that the Μ2+ mode forms the primary order parameter of the transition as it is the only mode capable of producing the final Pnma symmetry. Pnma is not a direct subgroup of P63/mmc, so the transition is necessarily first order. Despite this, the unit cell volume shows no clear contraction upon cooling below the phase transition temperature (Figure S5), which would appear to be inconsistent with a first-order transition. However, Figure shows that the Μ2+ mode amplitudes (X) can be fit to a critical equation of the following form:where Xu and X0 are the amplitudes at the transition temperature (Tu) and 0 K, respectively, W is a fitting parameter (W ≈ 2), and tu is the reduced temperature defined as tu = (Tu – T)/Tu. This empirical expression describes the critical variation of order-parameter-like structural quantities below a phase transition.[39]Equation is particularly successful in describing the variations of such quantities at temperatures well below the transition temperature (T ≪ Tu); hence, it can be used to model the behavior of distortion modes across extended temperature ranges. Equation has been used previously to model the behavior of structural quantities that vary as order parameters in materials featuring quasi-molecular clusters such as Fe3O4[39] and GaV2O4.[40] Here, a transition temperature of Tu = 225 K was extracted from the fit (Figure ), which is in excellent agreement with the temperature determined from our heat capacity measurements. The parameters X0 and Xu can also be compared to characterize the behavior of the transition, that is, whether the structural distortion is largely frozen below Tu (Xu/X0 ≈ 1), or whether it is quasi-continuous in character (Xu/X0 ≈ 0).[39] Frozen transitions feature largely temperature-invariant structural changes below Tu, whereas quasi-continuous transitions display more temperature-dependent behavior. In the case of Ba3CaMo2O9, the ratio Xu/X0 ≈ 0.45 tends toward the quasi-continuous limit over the frozen limit. This explains the lack of any notable discontinuity in unit cell volume upon cooling below the structural transition. It appears then that thermal effects largely dictate the structural behavior of Ba3CaMo2O9, and the transition does not appear to be tied to any specific electronic instability. This substantiates the lack of Mo–Mo bond formation at the transition and ultimately reflects the localized character of the Mo2O9 spin dimers.

Figure 6

Temperature dependence of the Μ2+ distortion mode in Ba3CaMo2O9. The solid cyan line depicts the fit to the order-parameter-like expression given in the main text.

We performed a comparative distortion mode analysis on Ba3SrMo2O9 based on our previously reported NPD data.[22] The distortion mode amplitudes for both phases of Ba3SrMo2O9 are shown in Figure . Here, the distortion modes have been labeled with respect to the parent P63/m symmetry as the P63/mmc phase has not been reported for Ba3SrMo2O9. For the P63/m phase, the low-temperature distortion is described by a single Γ1+ mode. Predominantly, the Γ1+ mode corresponds to in-phase tilting of the SrO6 octahedra and Mo2O9 dimers about the hexagonal c axis; in 6H-perovskites, this distortion has been canonically attributed to the size mismatch between the B′ and M sites.[41] The amplitude of the Γ1+ mode in P63/m-Ba3SrMo2O9 shows a discontinuous increase below 210 K (Figure a). Interestingly, this discontinuity overlaps with a previously identified lattice anomaly in this phase, where there is an abrupt contraction of the lattice along c below 210 K without any accompanying structural transition.[22] We previously showed that the phase fractions of the P63/m and P21/m phases in Ba3SrMo2O9 were sensitive to this lattice anomaly: the weight fraction of the P21/m phase attains a maximum of 47.6% at 210 K, but this fraction gradually decreases below this temperature so that the lattice anomaly promotes the P63/m phase instead. The distortion stemming from the Γ1+ mode is then associated with the anomalous lattice contraction in this phase. We also note that the amplitude of this mode correlates with the Mo–O(1)–Mo bond angles (Figure S6), where the Mo–O(1)–Mo bond angles tend toward the ideal angle for magnetic superexchange interactions. This structural correlation suggests that the lattice anomaly serves to promote intradimer superexchange pathways in P63/m-Ba3SrMo2O9 so that this is the mechanism by which the P63/m phase becomes favored over the P21/m phase upon further cooling.

Figure 7

Distortion mode analyses of the (a) P63/m and (b) P21/m phases of Ba3SrMo2O9. Mode amplitudes were derived and normalized with respect to their respective parent unit cells. Where not apparent, error bars are smaller than the data points. Data for the P21/m phase at 230 K have been omitted as the small phase fraction at this temperature did not allow for a reliable quantification of the mode amplitudes.

For P21/m-Ba3SrMo2O9, there are three distortion modes active in the low-temperature regime. The Γ1+ and Γ3,5+ modes predominantly correspond to in-phase tilting distortions of the SrO6 octahedra about the pseudohexagonal c and a axes, respectively, while the M1+ distortion mode corresponds to an out-of-phase tilt of the SrO6 octahedra about the pseudohexagonal b axis. The structural distortions in Ba3SrMo2O9 and its Ca analogue are hence largely similar, with the notable exception of the SrO6 tilting associated with the Γ1+ mode in Ba3SrMo2O9. The M1+ mode comprises the dominant contribution to the structural distortion in P21/m-Ba3SrMo2O9 at lower temperatures (Figure b) so that it is the primary order parameter for the transition. Although the Γ1+ mode also has a significant amplitude in this phase, its amplitude shows no apparently critical variation with cooling. This suggests that this mode behaves as a secondary order parameter so that it is coupled to the primary M1+ distortion. In contrast, the M1+ mode shows an overall increase upon cooling (Figure S7). The temperature variation of the M1+ amplitude can be fit with the critical expression given in eq , as shown in Figure S8. A transition temperature of Tu = 241 K was extracted from the fit, which is in agreement with the experimental temperature previously identified from neutron diffraction and heat capacity measurements. The transition is also characterized by the ratio Xu/X0 = 0.80, suggesting that Ba3SrMo2O9 tends toward the frozen limit over the quasi-continuous limit.

Closer examination of the individual atomic displacements for Ba3SrMo2O9 reveals a subtle distinction to its B′ = Ca counterpart. The M1+ mode in P21/m-Ba3SrMo2O9 changes the intradimer Mo–Mo distances according to the [Mo:f]A(a) displacement (Figure ). The two crystallographically distinct Mo2O9 dimers in this phase feature an antiferrodistortive displacement of the Mo atoms along c so that they displace toward each other within the QM clusters and away from each other within the AFM spin dimers. In contrast, the displacement of the Mo atoms in Ba3CaMo2O9 is limited by symmetry to the Μ2+ [Mo:f]E(a) distortion, which corresponds to an antiferrodistortive displacement within the pseudohexagonal basal plane; as the intradimer Mo–Mo distances lie parallel to the crystallographic long axis, there is hence no contraction of the intradimer Mo–Mo distances in Ba3CaMo2O9. Mo–Mo bonding is therefore strongly coupled to the crystal lattice in P21/m-Ba3SrMo2O9, and the near-frozen character of the structural transition in this phase reflects the crystallization of the QM dimer network. In contrast, there appears to be no such behavior in Ba3CaMo2O9 so that orbital molecules do not form in this composition and the electrons remain entangled in a localized spin state. It is likely that the more covalent Ca–O bond draws electron density away from the Mo2O9 dimers so that Mo–Mo bonding is less favored than in the B′ = Sr counterpart.

Figure 8

Asymmetric M1+ [Mo:f]A(a) displacement in Ba3SrMo2O9. Quasi-molecular dimers are shown in pink, while the localized spin dimers are shown in blue.

Summary

Ba3CaMo2O9 is a novel 6H-perovskite featuring localized Mo2O9 spin dimers below 320 K. A structural transition is observed near 200 K, but the electronic and magnetic properties of the Mo2O9 dimers appear to be largely insensitive to the resulting distortion so that the transition does not produce Mo–Mo bonding. Our distortion mode analyses suggest that tilting of the B′O6 octahedra along c in 6H-perovskites is correlated with the intradimer superexchange pathways. Size mismatch between the B′ and M sites could hence constitute an appropriate chemical parameter to tune intradimer exchange interactions and discover new ground states in new or existing 6H-perovskites. This approach could also be extended to other materials featuring orbital molecules to probe the sensitive interplay between bonding interactions and magnetic exchange.