betaFeedback

WorksPhysics and Astronomy

The crystal and magnetic structure of cesium superoxide

Abstract

CsO2 is a member of the family of alkali superoxides (formula AO2 with A= Na, K, Rb and Cs) that exhibit magnetic behavior arising from open pp-shell electrons residing on O2- molecules. We use neutron diffraction to solve the crystal and magnetic structures of CsO2, and observe a complex series of structures on cooling from room temperature to 1.6 K. These include an incommensurate modulation along the a-axis of the structure at intermediate temperatures, which then locks into a commensurate modulation that doubles the unit cell compared to the previously supposed orthorhombic unit cell. In both incommensurate and commensurate phases our structural solution involves a staggering of the cesium ion positions along the b-axis, in contrast to studies of other alkali superoxides in which staggered tilts of the O2- dimers relative to the c-axis are seen. Below T ~ 10 K we observe magnetic Bragg reflections arising from an antiferromagnetically ordered structure with a wavevector of k = (0,0,0) (relative to the doubled crystallographic unit cell), with moments that point predominantly along the b-axis with a small component along the a-axis that hints at possible anisotropic exchange coupling (consistent with the crystal structure). Measurements of the magnetic Bragg reflections in an applied magnetic field suggest a spin-flop transition takes place between 2 T and 4 T in which moments likely flop to point along the crystallographic a-axis. Our measurements indicate that CsO2 is an interesting example of magnetic properties being inherently linked to the crystal structure, in that the staggered displacement of the cesium ions activates antisymmetric exchange which then permits the observed spin canting.

Introduction

The complex interplay between spin, orbital and lattice degrees of freedom plays a varied role across many material families, resulting in a plethora of emergent properties. Well known examples include the Jahn-Teller interaction in transition metal oxides [1], magnetoelastic coupling [2], and quantum spin liquid states in strongly spin-orbit coupled materials [3]. This latter example is particularly topical, with the search ongoing for a material showing the bond-directional Ising-like anisotropy necessary for a realisation of the Kitaev model, for instance.

Anionogenic magnetic materials are compounds on which the spin degree of freedom is associated with the partially occupied pp-orbitals of an anionic molecule. Although theoretical predictions exist for other simple anions [4], the most well-known experimental examples are for the cases of oxygen anions. Some such materials are known to be magnetic include solid oxygen [5, 6, 7], alkali sesquioxides, and alkali superoxides [8]. In the latter, the magnetic moment arises due to a partially (half) occupied π∗\pi^* orbital on the O2_2 molecular units. Because the magnetic object is spatially extended, on so-called O2_2 dumbbells, rather than arising from unpaired electrons on a single ion, a natural coupling arises between the magnetism and the crystal structure. Of particular interest is the interplay of magnetism with the orientation of the dumbbells, a possibility which was noted explicitly in the 1980s and referred to as magnetogyration [9]. The alkali sesquioxides have attracted much recent attention because they undergo charge ordering which then leads to the formation of magnetic pyrochlore or dimer lattices [10, 11, 12, 13, 14].

The alkali superoxides AO2_2 (A=NaA = \mathrm{Na}, K, Rb, Cs) have been known for a relatively long time, with studies in the 1970s and 1980s elucidating many of the key physical properties. In particular, the fact that all of them are magnetic and undergo a cascade of structural phase transitions on cooling was recognized at this time [8]. The materials in the family have a common high temperature structure of the NaCl type [15] in which the O2−_2^- dumbbells are able to rotate freely. As they are cooled the rotation of the dumbbells progressively becomes hindered, so they precess about the cc-axis of a tetragonal structure, and then eventually freeze. The details of this freezing are fairly well characterized for KO2_2, which forms a monoclinic structure at low temperature in which the dumbbells are frozen at an angle from the higher temperature tetragonal cc-axis [16, 17]. Neutron diffraction measurements from the 1960s confirm that KO2_2 orders antiferromagnetically [17], however in the data collected at that time only two magnetic Bragg peaks could be discerned above background, making a magnetic structure refinement impossible. On the other hand, the precise details of the structures of RbO2\mathrm{RbO_2} and CsO2\mathrm{CsO_2} were much less well characterized at that time [15, 18]. What was known is that, like KO2\mathrm{KO_2}, both materials take on a tetragonal structure at room temperature, with transitions at ∼150\sim150 K into possible orthorhombic structures. Around the same temperature an incommensurate structure was observed in both materials, with a period around three lattice units along the aa-axis. Neither the orthorhombic structure nor the incommensurate structure were solved in detail in either case, however.

More recent experimental studies of CsO2\mathrm{CsO_2}, RbO2\mathrm{RbO_2} and NaO2\mathrm{NaO_2} have revisited their bulk magnetic properties. In particular, CsO2\mathrm{CsO_2} has been posited to be a pseudo-1-dimensional magnet, with evidence including a broad hump in the magnetic susceptibility reminiscent of what is predicted by a modified Bonner-Fisher model [19]. Magnetic order occurs at ∼10\sim10 K, and DFT calculations of superexchange pathways [20], and NMR [21] and EPR [22] data are consistent with the formation of a Tomonaga-Luttinger liquid phase.

These studies have also revisited the crystal structure, noting the presence of a transition from a tetragonal to an orthorhombic structure below ∼70\sim70 K (a substantially lower temperature than was seen for a similar transition in earlier studies). Notwithstanding, the crystallographic details of the orthorhombic structure were not solved, and nor were the details of the magnetic structure. High field magnetization studies [23] are consistent with CsO2\mathrm{CsO_2} exhibiting reduced dimensionality, however these data are not fully consistent with the models proposed in the other studies, and the authors note that in the absence of detailed knowledge of the magnetic structure all of the data are hard to interpret. Recent studies combining μ\muSR, Raman spectroscopy, and x-ray diffraction on RbO2\mathrm{RbO_2} [24] and NaO2\mathrm{NaO_2} [25] have been interpreted similarly, with progressively lowering symmetry on cooling, including the observation of incommensurate structural Bragg peaks in RbO2\mathrm{RbO_2}. The main difference between the two compounds is that the data on RbO2\mathrm{RbO_2} are more consistent with a 3-dimensional Heisenberg model rather than 1-dimensional magnetism, whereas the data on NaO2\mathrm{NaO_2} suggest that there is low dimensionality and no long range magnetic order.

Application of a relatively modest magnetic field, of ∼2.5\sim2.5 T at 2 K, results in a significant change in the magnetic susceptibility of CsO2\mathrm{CsO_2} [23]. This is interpreted as being a spin flop transition from a low field antiferromagnetic state with easy-axis anisotropy. It is notable that the majority of the NMR measurements interpreted as evidence for the formation of a Tomonaga Luttinger liquid were taken at fields either close to or above the spin flop transition field.

As already discussed, the crystal and magnetic structure of KO2\mathrm{KO_2} is better known, and perhaps as a consequence of this there has been more theoretical interest in this member of the superoxide family. Studies have focused on the coupling between spin, orbital and lattice physics. Several authors have proposed that KO2\mathrm{KO_2} undergoes orbital ordering [26, 27, 28], which then has consequences for the nature of superexchange between O2−\mathrm{O_2^-} molecules via the K ion and hence affects the magnetic order that develops. In particular, anisotropic exchange has been discussed [29] which as well as affecting the propensity to magnetic order may also give rise to unusual magnetic excitations. The calculations rely on a good knowledge of the crystal structure to have predictive power, underscoring the importance of accurate measurements.

Given the ongoing ambiguity concerning the low temperature crystal structure of CsO2\mathrm{CsO_2}, and the ordered magnetic structure, we were motivated to revisit this problem using modern neutron scattering instrumentation and analysis methods. As well as providing a complete picture, such measurements are invaluable for the interpretation of data from other experimental probes as well as informing future theoretical predictions. We found that at low temperature the crystallographic unit cell is doubled along the aa-axis compared to the room temperature tetragonal cell, with a staggered displacement along the bb-axis of the Cs ions but essentially no tilts of the O2−\mathrm{O_2^-} dumbbells. This phase sets in below 80 K. We also observed the incommensurate crystal structure, akin to one seen in early studies, and found that like the low temperature phase it is characterized by an aa-axis modulation of the positions of the Cs ions along the bb-axis. It would seem that this incommensurate structure eventually locks into the aforementioned commensurate one on cooling. The incommensurate phase is visible below 190 K, below which temperature there is a distinct transition from the room temperature tetragonal phase (space group I4/mmmI4/mmm) to another that, aside from the incommensurate reflections, can be indexed using ImmmImmm. In the magnetically ordered phase below 10 K we found that the moments align antiferromagnetically, with the largest component along the bb-axis but a small component along the aa-axis also, indicating possible exchange anisotropy.

Methods

The CsO2\mathrm{CsO_2} powder samples used for this study were prepared using the well-established method of oxidation of distilled Cs metal with dried molecular oxygen [30].

Neutron diffraction experiments were conducted using the E2 [31], E6 and E9 [32] diffractometers at the BER-II reactor at the Helmholtz Zentrum Berlin, and on the WISH [33] and HRPD [34] diffractometers at the ISIS spallation neutron source. Experiments on E2 were performed using a fixed incident neutron wavelength of λ=2.38\lambda= 2.38 Å, selected by a PG monochromator, between temperatures of 1.7 and 100 K in applied magnetic fields between 0 and 6.5 T.

The sample was pressed into cylindrical pellets of diameter 6 mm and height 3 mm, ten of which were placed into a sealed quartz ampoule in which the sample was well fixed in order to avoid movement of the powder grains in an applied magnetic field. This ampoule was placed into a cylindrical vanadium container of diameter 8 mm and height 60 mm. Experiments on E6 were performed with a fixed incident neutron wavelength λ=2.41\lambda= 2.41 Å, also selected by a PG monochromator, for temperatures between 1.6 and 248 K in zero applied magnetic field, using the same sample containment as for E2. One dataset was collected using E9, with a fixed incident neutron wavelength of λ=1.7985\lambda= 1.7985 Å, selected by a Ge monochromator, at a temperature of 4 K, again with the same sample containment. Neutron powder patterns were collected between the diffraction angles 7.8∘7.8^\circ to 83.4∘83.4^\circ (E2), 5.5∘5.5^\circ to 136.4∘136.4^\circ (E6), and 7.5∘7.5^\circ to 141.7∘141.7^\circ (E9), respectively.

AtomSitexxyyzzUisoU_{\mathrm{iso}} (Å2^2)
Cs2a2a0.000.000.000.0323(6)
O4e4e0.000.000.4138(1)0.0404(5)

Table I. Crystal structure parameters for the I4/mmmI4/mmm space group solution, at 290 K measured on the HRPD instrument. The determined lattice parameters were a=4.4589(2)a = 4.4589(2) Å and c=7.3320(3)c = 7.3320(3) Å. The unit cell volume was V=145.796(1)V = 145.796(1) Å3^3. The residuals for the fit were Rp=2.18%R_p = 2.18\% and Rwp=2.88%R_{wp} = 2.88\%.

HRPD and WISH are both time-of-flight (ToF) instruments, with a white beam of neutrons incident on the sample and analysis of the arrival time at the detectors (ToF) used to determine wavelength and hence Bragg reflection d-spacing. Experiments on HRPD were performed at temperatures of 20 K and 80 K, with further measurements taken over fixed time intervals (6 minutes) as the sample was cooled from 290 K. The sample was loaded into an 8 mm diameter vanadium can of height 40 mm for these measurements. All of the data used for subsequent analysis were collected in the highest angle detector bank, with data time-focused to a scattering angle of 2θ=168.33∘2\theta= 168.33^\circ. Experiments on WISH were performed at temperatures in the range 1.6 to 300 K. The sample was loaded into an aluminum can with an annular geometry, with height 40 mm, outer diameter 20 mm and sample thickness 2 mm. The sample was first cooled to 1.6 K, then a measurement of 40 minutes duration was performed at 1.6 K, followed by another of the same duration at 25 K. The sample was then re-cooled to 1.6 K and further measurements were made at fixed temperatures for 5 minutes per temperature, warming back up to 300 K. A sample temperature fluctuation of ≤2%\le2\% was observed for each of the measurements. For crystal structure analysis data time-focused on to the higher angle detector banks, at 2θ=152.9∘2\theta= 152.9^\circ and 2θ=121.7∘2\theta= 121.7^\circ were used. For the magnetic refinements the lower angle detector bank data, at 2θ=58.3∘2\theta= 58.3^\circ was used.

Structural refinements were performed using the FullProf software suite [35, 36] for the E2, E6 and E9 data, and using Jana2006 [37] for the HRPD and WISH structural data. Subsequent magnetic refinements were performed using FullProf for the WISH data. Magnetic symmetry analysis for the magnetic structure refinement was done using the BasIreps tool implemented in FullProf. The incommensurate structure analysis was done using Jana2006 together with the Isodistort software [38, 39].

Results

Crystal structure

Figure 1 shows data collected at 290 K (room temperature) on the HRPD instrument, together with a refinement with the I4/mmmI4/mmm space group and lattice parameters widely reported in the literature. A satisfactory refinement can be achieved for these data collected for a relatively short duration (6 minutes) with correspondingly non-optimized signal to noise.

Neutron powder diffraction data and refinement at 290 K

Figure 1. Data collected at 290 K on HRPD as a function of d-spacing, refined using the I4/mmmI4/mmm space group (see text). Measured data are indicated by black circles, the fit by the red line, and the difference curve (offset by −1200 units) is shown as a blue curve. Vertical ticks indicate the positions of Bragg reflections. The residuals for these fits were Rp=2.18R_p = 2.18 and Rwp=2.88R_{wp} = 2.88.

The details of the crystal structure determined at 290 K are given in table I. Refinements of the data from WISH at the same temperature are consistent with these parameters. On cooling a phase transition is observed at TS1=192(2)T_{\mathrm{S1}} = 192(2) K, that appears to be tetragonal to orthorhombic as illustrated in fig. 2 which shows the lattice parameter as a function of temperature determined from Rietveld refinements using the tetragonal I4/mmmI4/mmm space group above the transition and the orthorhombic ImmmImmm space group below it, from data collected on E6. The latter space group represents the symmetry of the lattice if only macroscopic strain is taken into account [38, 39]. The strain reflects the change of the unit cell metric below the transition and therefore this space group is sufficient to evaluate the thermal evolution of the unit cell parameters. The transition is at a temperature consistent with early structural studies of CsO2_2 [15], but significantly higher than observed in more recent x-ray diffraction experiments [20]. In order to get good fits we found that anisotropic displacement parameters (ADPs) were required. The best fits were obtained with considerable elongation of the ADP ellipsoids of the cesium ions along the bb-axis compared to the other principal axes. Furthermore, the ADP ellipsoids of the oxygen ions were elongated in the abab-plane, albeit with a smaller difference between aa and bb than for the cesium ions, and compressed significantly along the cc-axis. These findings suggest the presence of atomic displacements not accounted for in the ImmmImmm space group.

Close examination of the data collected on WISH shows that at temperatures above ∼72\sim72 K, and at least up to 180 K, weak incommensurate superlattice reflections are visible above background that can be indexed using a propagation vector of kIC=(0.561(2),0,0)\mathbf{k}_{\mathrm{IC}} = (0.561(2), 0, 0) referred to the tetragonal unit cell, i.e. a period of 1.78 lattice units. The appearance of the incommensurate superlattice seems to be concomitant with the putative transition from the tetragonal to orthorhombic structure at TS1T_{\mathrm{S1}}. To probe this further we show in Figure 3 the temperature dependence of the intensity and position of the (1+kx,2,1)(1+k_x, 2, 1) peak (the fact that superlattice peaks such as this one appear at short dd-spacing, and hence large ∣Q∣|\mathbf{Q}|, indicates that they are structural rather than magnetic in origin).

Temperature dependence of incommensurate Bragg reflection intensity and position

Figure 3. Data from WISH showing the disappearance on warming up to 200 K of the incommensurate Bragg reflection at QIC=(1.56,2,1)\mathbf{Q}_{\mathrm{IC}} = (1.56, 2, 1) at d=1.69d = 1.69 Å. The upper panel shows the intensity of the peak, determined by integrating a fitted Gaussian, with black circles. The lower panel shows the fitted position of the peak kxk_x with respect to (1+kx,2,1)(1+k_x, 2, 1) with red squares. In both cases the errorbars are smaller than the point size.

appear at short dd-spacing, and hence large ∣Q∣|\mathbf{Q}|, indicates that they are structural rather than magnetic in origin).

Given that the incommensurate reflections seem to appear at the same temperature as the putative tetragonal to orthorhombic transition, to solve the structural modulation we assumed that the corresponding order parameter is transformed by a single irreducible representation (irrep) of the high temperature tetragonal space group I4/mmmI4/mmm. The irreducible nature of the order parameter was concluded based on the continuous character of the transition evidenced by the temperature dependence of the unit cell parameters (Figure 2). There are four irreps, Σi\Sigma_i (i=1−4i = 1 - 4) associated with the (g,0,0)(g, 0, 0) line of symmetry. The relevant superspace groups were generated by ISODISTORT [38, 39] and compared with the available diffraction data. Some of them could be ruled out due to extinction conditions that predict no intensity where a finite signal was observed (e.g. at d=4.05d = 4.05Å). The remainder were tested using the Jana2006 software. The best agreement (with Rwp=8.83%R_{wp} = 8.83\%) was found for the model with Immm(0,0,g)s00Immm(0, 0, g)s00 symmetry yielded by the Σ4\Sigma_4 irrep and illustrated in Figure 4. The largest displacements are of the cesium ions, which are staggered along the bb-axis following a sinusoidal modulation with an amplitude of 0.0564(9) lattice units. The oxygen molecules also exhibit a sinusoidal displacement along bb, albeit with a smaller amplitude of 0.0090(7) lattice units. Interestingly, this symmetry does not permit the rotation of oxygen molecules widely predicted for CsO2_2, and seen in other alkali superoxides. Rather, the entire molecules’ centres of mass are displaced along bb, with the molecules remaining oriented parallel to the cc-axis. The complete structural solution for the incommensurate phase is given in Table II.

Temperature dependence of lattice parameters

Figure 2. Temperature dependence of lattice parameters refined from E6 data, assuming I4/mmmI4/mmm and ImmmImmm space groups at high and low temperature respectively.

AtomxxyyzzUiso(A˚2)U_{\mathrm{iso}}(\text{\AA}^2)
Cs0.00.00.00.0029(5)
−0.0564(9)-0.0564(9)0.00.0
0.00.00.0
O0.00.40964(12)0.00.0110(3)
−0.0090(7)-0.0090(7)0.00.0
0.00.00.0

Table II. Structural parameters of CsO2_2 obtained from refinement of the neutron diffraction data (T=100T = 100 K) using the Immm(0,0,g)s00Immm(0, 0, g)s00 superspace group with the basis vectors related to the parent tetragonal I4/mmmI4/mmm structure as (0,−1,0,0)(0, -1, 0, 0), (0,0,1,0)(0, 0, 1, 0), (−1,0,0,0)(-1, 0, 0, 0), (0,0,0,1)(0, 0, 0, 1) with the origin at (0,0,34)(0, 0, \frac{3}{4}). Here, AiA_i and BiB_i with i=(x,y,z)i = (x, y, z) are the Fourier coefficients of the first harmonic (n=1n = 1) of the displacive modulation function: ui,j,l(rj,l,kIC)=∑n=0∞Ai,jnsin⁡(2πn[rj,l⋅kIC])+Bi,jncos⁡(2πn[rj,l⋅kIC])u_{i,j,l}(r_{j,l}, k_{\mathrm{IC}}) = \sum_{n=0}^{\infty} A_{i,j}^{n}\sin(2\pi n[r_{j,l}\cdot k_{\mathrm{IC}}]) + B_{i,j}^{n}\cos(2\pi n[r_{j,l}\cdot k_{\mathrm{IC}}]), where rj,lr_{j,l} indicates the position of the jthj^{\mathrm{th}} atom of the average structure in the lthl^{\mathrm{th}} unit cell. The unit cell parameters were a=4.4117(1)a = 4.4117(1) Å, b=7.3438(2)b = 7.3438(2) Å, c=4.3762(1)c = 4.3762(1) Å and kIC=(0,0,0.561(2))k_{\mathrm{IC}} = (0, 0, 0.561(2)), the latter adopted for the setting of the superspace group specified above. The unit cell volume was V=141.782(6)V = 141.782(6) Å3^3. The residuals for the fit were Rp=6.86%R_p = 6.86\% and Rwp=8.83%R_{wp} = 8.83\%.

Representation of the incommensurate crystal structure

Figure 4. Representation of the incommensurate crystal structure determined using the refined WISH data at 100 K. The Cs ions (green spheres) and oxygen dumbbells (red spheres connected by red cylinder to indicate the molecular bond) modulate sinusoidally along the aa-axis, with displacements along the bb-axis (in tetragonal notation), which is cyclically permuted with respect to the notation in table II). Five unit cells along aa are shown in order to illustrate the form of the displacements. This image, and others in the rest of the manuscript illustrating crystal and magnetic structures, were rendered using the VESTA software [40].

On cooling further a significant anomaly in the aa lattice parameter (in the ImmmImmm setting), and a smaller anomaly in the cc parameter, occurs at TS2=71.5(9)T_{S2} = 71.5(9) K.

At the same time the propagation vector of the incommensurate structural modulation changes and locks in to a commensurate value of kC=(12,0,0)k_C = (\frac{1}{2}, 0, 0) (again referred to the ImmmImmm cell), which survives down to the lowest temperatures measured. Data showing the transition locking into QC=(32,2,1)=(1,2,1)+kCQ_C = (\frac{3}{2}, 2, 1) = (1, 2, 1) + k_C are shown in Figure 5.

Data from WISH showing the transition from incommensurate to commensurate crystal structure

Figure 5. Data from WISH showing the transition from incommensurate to commensurate crystal structure. Below ∼80\sim80 K a peak corresponding to QC=(32,2,1)Q_C = (\frac{3}{2}, 2, 1) appears at d=1.71d = 1.71 Å. At higher temperatures a peak corresponding to QIC=(1.57,2,1)Q_{\mathrm{IC}} = (1.57, 2, 1) at d=1.69d = 1.69 Å is visible. Commensurate and incommensurate peak positions are indicated by a black square and a red circle respectively.

In order to understand what is happening crystallographically in this phase we performed careful Rietveld refinements of all of the data collected on the different instruments used. In previous work it was suggested that the orthorhombic space group ImmmImmm provides an appropriate description of the crystal structure. However, that would be inconsistent with the doubled unit cell seen below TS2T_{S2}.

Assuming the most likely scenario, that the low temperature structural transition is a lock-in type, one can deduce the possible symmetries of the crystal structure below TS2T_{S2}. This scenario sets symmetry constraints on the transformational properties of the commensurate order parameter, implying the same active irrep for both incommensurate and commensurate phases. As mentioned above, the modulated orthorhombic structure is associated with Σ4\Sigma_4. This irrep is four-dimensional with the components of the complex order parameter (η1,η1⋆,η2⋆,η2)(\eta_1,\eta_1^\star,\eta_2^\star,\eta_2). Using the matrix operators summarised

in table III, one can verify the existence of a lock-in free-energy invariant, η14+η1⋆4+η24+η2⋆4\eta_1^4 + \eta_1^{\star4} + \eta_2^4 + \eta_2^{\star4}, for the commensurate value of the propagation vector kC=(1/2,0,0)\mathbf{k}_{\mathrm{C}}=(1/2,0,0). This energy term exists only for the single value of the propagation vector and therefore its activation favours the commensurate phase. This symmetry argument further supports the lock-in mechanism for the transition at TS2T_{\mathrm{S2}}.

Symm.mM5+(δ1,δ2)mM_{5}^{+}(\delta_{1},\delta_{2})mΣ3(ξ1,ξ1∗,ξ2∗,ξ2)m\Sigma_{3}(\xi_{1},\xi_{1}^{*},\xi_{2}^{*},\xi_{2})Σ4(η1,η1∗,η2∗,η2)\Sigma_{4}(\eta_{1},\eta_{1}^{*},\eta_{2}^{*},\eta_{2})
{1∣0,0,0}\{1|0,0,0\}(1001)\begin{pmatrix}1&0\\0&1\end{pmatrix}(1000010000100001)\begin{pmatrix}1&0&0&0\\0&1&0&0\\0&0&1&0\\0&0&0&1\end{pmatrix}(1000010000100001)\begin{pmatrix}1&0&0&0\\0&1&0&0\\0&0&1&0\\0&0&0&1\end{pmatrix}
{2z∣0,0,0}\{2_{z}|0,0,0\}(−100−1)\begin{pmatrix}-1&0\\0&-1\end{pmatrix}(0−100−1000000−100−10)\begin{pmatrix}0&-1&0&0\\-1&0&0&0\\0&0&0&-1\\0&0&-1&0\end{pmatrix}(0100100000010010)\begin{pmatrix}0&1&0&0\\1&0&0&0\\0&0&0&1\\0&0&1&0\end{pmatrix}
{2y∣0,0,0}\{2_{y}|0,0,0\}(0110)\begin{pmatrix}0&1\\1&0\end{pmatrix}(0100100000−10000−1)\begin{pmatrix}0&1&0&0\\1&0&0&0\\0&0&-1&0\\0&0&0&-1\end{pmatrix}(0−100−100000−10000−1)\begin{pmatrix}0&-1&0&0\\-1&0&0&0\\0&0&-1&0\\0&0&0&-1\end{pmatrix}
{4z+∣0,0,0}\{4_{z}^{+}|0,0,0\}(0−110)\begin{pmatrix}0&-1\\1&0\end{pmatrix}(001000010−100−1000)\begin{pmatrix}0&0&1&0\\0&0&0&1\\0&-1&0&0\\-1&0&0&0\end{pmatrix}(0010000110000100)\begin{pmatrix}0&0&1&0\\0&0&0&1\\1&0&0&0\\0&1&0&0\end{pmatrix}
{−1∣0,0,0}\{-1|0,0,0\}(1001)\begin{pmatrix}1&0\\0&1\end{pmatrix}(0100100000010010)\begin{pmatrix}0&1&0&0\\1&0&0&0\\0&0&0&1\\0&0&1&0\end{pmatrix}(0100100000010010)\begin{pmatrix}0&1&0&0\\1&0&0&0\\0&0&0&1\\0&0&1&0\end{pmatrix}
{1∣1,0,0}\{1|1,0,0\}(1001)\begin{pmatrix}1&0\\0&1\end{pmatrix}(eπi0000e−πi0000e−πi0000eπi)\begin{pmatrix}e^{\pi i}&0&0&0\\0&e^{-\pi i}&0&0\\0&0&e^{-\pi i}&0\\0&0&0&e^{\pi i}\end{pmatrix}(eπi0000e−πi0000e−πi0000eπi)\begin{pmatrix}e^{\pi i}&0&0&0\\0&e^{-\pi i}&0&0\\0&0&e^{-\pi i}&0\\0&0&0&e^{\pi i}\end{pmatrix}
{1∣0,1,0}\{1|0,1,0\}(1001)\begin{pmatrix}1&0\\0&1\end{pmatrix}(1000010000100001)\begin{pmatrix}1&0&0&0\\0&1&0&0\\0&0&1&0\\0&0&0&1\end{pmatrix}(1000010000100001)\begin{pmatrix}1&0&0&0\\0&1&0&0\\0&0&1&0\\0&0&0&1\end{pmatrix}
{1∣0,0,1}\{1|0,0,1\}(1001)\begin{pmatrix}1&0\\0&1\end{pmatrix}(1000010000100001)\begin{pmatrix}1&0&0&0\\0&1&0&0\\0&0&1&0\\0&0&0&1\end{pmatrix}(1000010000100001)\begin{pmatrix}1&0&0&0\\0&1&0&0\\0&0&1&0\\0&0&0&1\end{pmatrix}
{1∣1/2,1/2,1/2}\{1|1/2,1/2,1/2\}(−100−1)\begin{pmatrix}-1&0\\0&-1\end{pmatrix}(e12πi0000e−12πi0000e−12πi0000e12πi)\begin{pmatrix}e^{\frac{1}{2}\pi i}&0&0&0\\0&e^{-\frac{1}{2}\pi i}&0&0\\0&0&e^{-\frac{1}{2}\pi i}&0\\0&0&0&e^{\frac{1}{2}\pi i}\end{pmatrix}(e12πi0000e−12πi0000e−12πi0000e12πi)\begin{pmatrix}e^{\frac{1}{2}\pi i}&0&0&0\\0&e^{-\frac{1}{2}\pi i}&0&0\\0&0&e^{-\frac{1}{2}\pi i}&0\\0&0&0&e^{\frac{1}{2}\pi i}\end{pmatrix}
TT(−100−1)\begin{pmatrix}-1&0\\0&-1\end{pmatrix}(−10000−10000−10000−1)\begin{pmatrix}-1&0&0&0\\0&-1&0&0\\0&0&-1&0\\0&0&0&-1\end{pmatrix}(1000010000100001)\begin{pmatrix}1&0&0&0\\0&1&0&0\\0&0&1&0\\0&0&0&1\end{pmatrix}

Table 3. Matrices of irreducible representations for generators of I4/mmmI4/mmm space group, associated with the k=(1,1,1)\mathbf{k}=(1,1,1) (MM-point) and the k=(1/2,0,0),(−1/2,0,0),(0,−1/2,0),(0,1/2,0)\mathbf{k}=(1/2,0,0),(-1/2,0,0),(0,-1/2,0),(0,1/2,0) wavevector star (Σ\Sigma-line of symmetry) [41]. Note that here TT is the time reversal operator.

In the commensurate limit of kC=(1/2,0,0)\mathbf{k}_{\mathrm{C}}=(1/2,0,0) there are three isotropy subgroups associated with Σ4\Sigma_{4}. Depending on choice of the global phase of the modulation, one can obtain the PnmaPnma, PmmaPmma and Pmc21Pmc2_{1} subgroups. They were the primary candidates in the refinement of

the low-temperature diffraction data. Independent analysis based on testing of all possible isotropy subgroups of I4/mmmI4/mmm consistent with the doubled unit cell also resulted only into two possibilities PnmaPnma and Pmc21Pmc2_{1}.

We performed refinements of all data collected on HRPD (since it had the highest resolution) at 20 K using PnamPnam and found generally satisfactory fits, with Rp=2.51%R_{p}=2.51\% and Rwp=3.33%R_{wp}=3.33\%. We also investigated refining the data with the Pna21Pna2_{1} space group (No. 33), which is a sub-group of PnamPnam, and found a slightly improved residuals of Rp=2.47%R_{p}=2.47\% and Rwp=3.29%R_{wp}=3.29\%. However, given that the improvement in goodness-of-fit was very small and likely arises due to fewer symmetry constraints on the structure, we take the higher symmetry structure to be the solution. The final refinements using the PnamPnam space group are shown in Figure 6.

Data and refinement plot for $Pnam$ at 20 K

Figure 6. Data and refinement using PnamPnam for data collected on HRPD at 20 K as a function of d-spacing. Black circles indicate the measured data, the red line the refinement, and the blue line indicates the difference between the two (offset by -2000 units). Vertical ticks indicate the positions of Bragg reflections.

The details of the refinements using the PnamPnam space group is given in Table IV, and shown graphically in Figure 7. The most notable feature is that the Cs ions are shifted considerably from their ideal position of 34\frac{3}{4} on the bb-axis, and form a zig-zag pattern in the doubled unit cell along aa. In the PnamPnam structure the oxygen dumbbells are not allowed by symmetry to tilt, whereas they are in principle free to do this in the Pna21Pna2_1 structure that was investigated. In the former structural solution we found that there was a small staggered shift of the centre of mass of the oxygen dumbbells along the bb-axis, albeit of much smaller magnitude than the shift of the Cs ions. For the Pna21Pna2_1 structure (see Table V) this shift of the oxygen dumbbells was found to be of the same nature, i.e. there

AtomSitexxyyzzUiso(A˚2)U_{\mathrm{iso}}(\text{\AA}^2)
Cs4c4c0.1207(4)0.7142(4)0.250.0022(4)
O8d8d0.8747(4)0.2472(4)0.15944(8)0.0082(3)

Table IV. Crystal structure parameters for the PnamPnam space group solution, at 20 K measured on the HRPD instrument. The determined lattice parameters were a=8.7271(1) A˚a = 8.7271(1)\,\text{\AA}, b=4.39758(6) A˚b = 4.39758(6)\,\text{\AA}, and c=7.33860(8) A˚c = 7.33860(8)\,\text{\AA}. The unit cell volume was V=281.640(6) A˚3V = 281.640(6)\,\text{\AA}^3. The residuals for the fit were Rp=2.51%R_p = 2.51\% and Rwp=3.33%R_{wp} = 3.33\%.

Refined crystal structure with $Pnam$ space group

Figure 7. The refined crystal structure with the PnamPnam space group, with oxygen molecules shown in red (spheres connected by a cylinder to indicate the molecular bond) and cesium ions shown in green (spheres only). The unit cell is indicated by the gray box. The view shown is slightly tilted from being parallel to the aa-axis. The periodic displacements of the Cs ions are clear, with a smaller but still visible periodic displacement of the O2−\mathrm{O}_2^- molecules also. The figure shows two unit cells along the aa-axis.

AtomSitexxyyzzUiso(A˚2)U_{\mathrm{iso}}(\text{\AA}^2)
Cs4a4a0.1203(3)0.7816(9)-0.0003(4)0.0043(4)
O14a4a0.1277(4)0.7502(12)0.4091(14)0.0066(3)
O24a4a0.6293(4)0.7575(8)-0.4091(12)0.0091(5)

Table V. Crystal structure parameters for the Pna21Pna2_1 space group solution, at 20 K measured on the HRPD instrument. The determined lattice parameters were a=8.727(3) A˚a = 8.727(3)\,\text{\AA}, b=4.3968(2) A˚b = 4.3968(2)\,\text{\AA}, c=7.3380(2) A˚c = 7.3380(2)\,\text{\AA}. The unit cell volume was V=281.553(3) A˚3V = 281.553(3)\,\text{\AA}^3. The residuals for the fit were Rp=2.47%R_p = 2.47\% and Rwp=3.30%R_{wp} = 3.30\%, which represent a very small improvement compared to the refinement shown in Table IV for the PnamPnam crystal structure.

was essentially no tilt of the oxygen dumbbells even when this would be allowed by symmetry.

Magnetic structure

On cooling below T≈10T \approx10 K additional low QQ / long dd-spacing peaks appear that are consistent with the previously supposed appearance of antiferromagnetic order. Such peaks were consistently visible in the data collected on WISH, E2, E6 and E9. As mentioned in § 2, only the WISH data were used for refining the magnetic structure. Of all the datasets, those from this instrument had the best signal to noise ratio and hence the most magnetic Bragg peaks were discernable, giving the greatest chance of a reliable refinement of the magnetic structure. Subsequent checks of the refinement using the E2, E6 and E9 instruments yielded good fits. An overview of the data collected as a function of temperature below 11 K is shown in Figure 8.

Overview of the data collected in the low angle bank on WISH as a function of temperature between 1.5 K and 11 K

Figure 8. Overview of the data collected in the low angle bank on WISH as a function of temperature between 1.5 K and 11 K. Data have been converted from time-of-flight to dd-spacing. The strong peaks at ∼8.8\sim8.8 Å and ∼5.7\sim5.7 Å correspond to the (1,0,0)(1,0,0) and (1,0,1)(1,0,1) Bragg reflections respectively.

The appearance of strong peaks at (1,0,0)(1, 0, 0) and (1,0,1)(1, 0, 1) indicates the onset of antiferromagnetic order, at 10<TN<1110 < T_{\mathrm{N}} < 11 K. This compares with the peaks in the inverse susceptibility and specific heat, that are taken to give the value of TNT_{\mathrm{N}}, of 9.6 K [42, 43, 44]. The small difference may be due to small differences in thermometry or a slight lag on the true sample temperature since the measurements were collected on warming.

All possible irreducible representations of a magnetic structure with k=(0,0,0)\mathbf{k} = (0, 0, 0) in the PnamPnam space group are given in Table VI. The general expressions of the Fourier coefficients Sk(j)\mathbf{S}_{\mathbf{k}}(j) were obtained from the basis functions calculated from the different representations of the O2−\mathrm{O}_2^{-} units at the Wyckoff position 4c4c of the space group PnamPnam: O1 at (x,y,14)(x, y, \frac{1}{4}), O2 at (−x,−y,34)(-x, -y, \frac{3}{4}), O3 at (12+x,12−y,14)(\frac{1}{2}+x, \frac{1}{2}-y, \frac{1}{4}), and O4 at (12−x,12+y,34)(\frac{1}{2}-x, \frac{1}{2}+y, \frac{3}{4}). Instead of the Wyckoff position 8d8d, where the individual oxygen atoms are located, we used 4c4c which defines the center of gravity of the O2−\mathrm{O}_2^{-} unit.

IrrepSk(1)\mathbf{S}_{\mathbf{k}(1)}Sk(2)\mathbf{S}_{\mathbf{k}(2)}Sk(3)\mathbf{S}_{\mathbf{k}(3)}Sk(4)\mathbf{S}_{\mathbf{k}(4)}
Γ1(4c)\Gamma_1(4c)(0,0,u)(0,0,u)(0,0,u)(0,0,u)(0,0,−u)(0,0,-u)(0,0,−u)(0,0,-u)
Γ2(4c)\Gamma_2(4c)(u,v,0)(u,v,0)(−u,−v,0)(-u,-v,0)(u,−v,0)(u,-v,0)(−u,v,0)(-u,v,0)
Γ3(4c)\Gamma_3(4c)(0,0,u)(0,0,u)(0,0,u)(0,0,u)(0,0,u)(0,0,u)(0,0,u)(0,0,u)
Γ4(4c)\Gamma_4(4c)(u,v,0)(u,v,0)(−u,−v,0)(-u,-v,0)(−u,v,0)(-u,v,0)(u,−v,0)(u,-v,0)
Γ5(4c)\Gamma_5(4c)(u,v,0)(u,v,0)(u,v,0)(u,v,0)(u,−v,0)(u,-v,0)(u,−v,0)(u,-v,0)
Γ6(4c)\Gamma_6(4c)(0,0,u)(0,0,u)(0,0,−u)(0,0,-u)(0,0,−u)(0,0,-u)(0,0,u)(0,0,u)
Γ7(4c)\Gamma_7(4c)(u,v,0)(u,v,0)(u,v,0)(u,v,0)(−u,−v,0)(-u,-v,0)(−u,−v,0)(-u,-v,0)
Γ8(4c)\Gamma_8(4c)(0,0,u)(0,0,u)(0,0,−u)(0,0,-u)(0,0,u)(0,0,u)(0,0,−u)(0,0,-u)

Table VI. Irreducible representations of a k=(0,0,0)\mathbf{k} = (0, 0, 0) magnetic structure in the PnamPnam space group.

We also considered a solution to the magnetic structure using the lower symmetry space group Pna21Pna2_1. The difference between the magnetic structures allowed in the two space groups is that PnamPnam does not permit magnetic order with a general component of the moment in the abab-plane and along the cc-axis, but one or the other. On the other hand, Pna21Pna2_1 does in principle allow magnetic moments to point along a general direction. However, we found that even refining the cc-axis component of the magnetic moment, permitted in Pna21Pna2_1, we found the best fit yielded this component as zero within errorbar. For the following we therefore restrict our discussion just to PnamPnam.

Note that the magnetic form factor for the s=1/2s = 1/2 O2−\mathrm{O}_2^{-} ions was included using the tabulated values obtained from measurements of solid oxygen [45]. Although the size of the oxygen molecules is different (1.29 Å for solid oxygen vs 1.33 Å for CsO2_2) the difference is small enough that we have reasonable confidence that the derived form factor is applicable here.

Before fitting the data we note that the strong intensity at (1,0,0)(1,0,0) and (1,0,1)(1,0,1) implies that the largest component of the magnetic moment will be along the bb-axis, since the magnetic neutron diffraction cross-section is proportional to the component of the magnetic moment perpendicular to the wavevector. This means that we can immediately rule out the Γ1\Gamma_1, Γ3\Gamma_3, Γ6\Gamma_6 and Γ8\Gamma_8 irreps, since they contain moments parallel to the cc-axis only.

We also note that for the Γ7\Gamma_7 irrep the yy-components of the moments are coupled ferromagnetically. Since the bulk susceptibility follows a purely antiferromagnetic trend below TNT_{\mathrm{N}}, for the refinements this component must be fixed to zero, which immediately allows us to rule it out. This then leaves the following possible irreps: Γ2\Gamma_2, Γ4\Gamma_4, and Γ5\Gamma_5, which all have AFM arrangements of spins for the major yy-component of the moment.

The presence of a very weak peak corresponding to (0,0,1)(0,0,1), that appears below TNT_{\mathrm{N}} and hence must be magnetic, also allows us to narrow down our choice of possible irreps. This observation allows us to rule out Γ5\Gamma_5, for which this peak has zero intensity. Comparing Γ2\Gamma_2 and Γ4\Gamma_4, we found substantially better fits (Rwp=6.36%R_{\mathrm{wp}} = 6.36\%) for

@references 大发棋牌 the case of the Γ2\Gamma_{2} irrep, compared to the Γ4\Gamma_{4} irrep (Rwp=7.75%R_{\mathrm{wp}} = 7.75\%). The best fit was found with μy=0.512(16)μB\mu_y = 0.512(16)\mu_{\mathrm{B}} and μx=0.05(2)μB\mu_x = 0.05(2)\mu_{\mathrm{B}}, giving a total moment of 0.514μB0.514\mu_{\mathrm{B}}. As illustrated in the inset of Figure 9, the presence of the (0,0,1)(0, 0, 1) peak is crucial for determining the xx-component of the magnetic moment, as the intensity is rather sensitive to the value of μx\mu_x. The final refinement is shown in Figure 9, together with an inset showing the sensitivity of the (0,0,1)(0, 0, 1) peak’s intensity to μx\mu_x. The final refined magnetic structure is shown in Figure 10. Our result implies the magnetic structure has Pn′a′m′Pn^\prime a^\prime m^\prime symmetry, keeping the setting of the paramagnetic PnamPnam space group.

Data and refinement plot

Figure 9. Data collected in the low angle bank on WISH at 1.5 K (black points), together with the final refinement (solid red line), difference curve (solid blue line). The lower ticks (blue) are for the antiferromagnetic structure and the upper ticks (green) are for the nuclear structure. The inset shows data focused on the region where the (0, 0, 1) peak is to be found. The lines correspond to simulations of the scattering with different values of μx\mu_x, with all other parameters from the refinement fixed. This shows that the intensity of the (0, 0, 1) peak is rather sensitive to the value of μx\mu_x.

Refined magnetic structure of CsO2

Figure 10. The refined magnetic structure of CsO2_2, with black arrows indicating the moment direction on the O2−_2^- dumbbells, the oxygen ions shown as red spheres, connected by a red cylinder to indicate the molecular bond, and the Cs ions are shown as green spheres.

Spin flop transition

Data taken on the E2 diffractometer probed the effect of an applied magnetic field on the two strongest magnetic Bragg peaks, (1,0,0)(1, 0, 0) and (1,0,1)(1, 0, 1), and are shown in Figure 11. Both peaks showed marked decreases in intensity between 2 T and 4 T. The strong (1,0,0)(1, 0, 0) peak intensity reduces in magnitude above background by approximately one third, with a somewhat smaller reduction of around 15% seen for the (1,0,1)(1, 0, 1) peak. In the zero field phase we have established that the moments lie predominantly along the bb-axis, and in a spin flop transition we would expect them to switch to being parallel to the aa or cc directions. Crucially, this would apply only to those (randomly oriented) crystallites in the powder sample whose bb axes lie mostly parallel to the applied magnetic field.

Intensity of magnetic Bragg reflections of CsO$_2$ at 1.7 K in applied magnetic fields

Figure 11. Data collected on E2 at T=1.7T = 1.7 K in applied magnetic fields between 0 T and 6.5 T. A sharp drop in peak intensity of the (1,0,0)(1,0,0) and (1,0,1)(1,0,1) magnetic Bragg reflections is observed for 2<B<42 < B < 4 T, most likely arising from a spin flop transition.

Above the critical field we would therefore expect approximately one third of the crystallites to undergo a spin flop transition. The maximum change in magnetic Bragg peak intensity would therefore be 13\frac{1}{3}, corresponding to the signal from the one third of crystallites involved in the spin flop going to zero. Because the (1,0,0)(1, 0, 0) peak reduces by approximately this amount we can surmise that the spin flop involves spins reorienting from the bb-axis to the aa-axis. If the spins were to flop towards cc, and assuming no change in the size of the ordered moment, then the component of magnetization perpendicular to Q\mathbf{Q} would be unchanged, and hence no change in the intensity of the (1,0,0)(1, 0, 0) Bragg peak would be seen, at variance with our observations. On the other hand, a spin flop towards aa would result in the component of magnetization perpendicular to Q\mathbf{Q} going to zero. We can check this result by looking at the intensity of the (1,0,1)(1, 0, 1) magnetic Bragg peak as well. Spins reorienting from bb to aa would result in a reduction in Bragg peak intensity of ∼40%\sim40\% for the one third of crystallites involved, giving rise to an overall reduction in intensity of ∼13%\sim13\%, which is broadly consistent with what we observe.

Discussion and Conclusion

We have determined the low temperature crystal structure of CsO2_2, finding it to be described by the orthorhombic space group PnamPnam (or equally well by the lower symmetry orthorhombic space group Pna21Pna2_1) in which the aa axis is approximately double the length of the bb axis. Earlier work had indicated a lowering of symmetry on cooling from room temperature, for example via the appearance of extra peaks in the Raman spectrum that could not be accounted for in a high symmetry setting [20] or due to the observation of the weak (12,0,0)(\frac{1}{2},0,0) peaks (indexed in the high temperature tetragonal unit cell) at low temperature in x-ray diffraction measurements [15]. Furthermore, DFT calculations have predicted an enlarged (doubled) unit cell compared to the high temperature tetragonal phase [20]. However, we note neither that the Raman nor the x-ray data were used to determine the detailed crystal structure as we have here. In addition, the DFT calculations were correct in predicting a doubled unit cell, but suggested that the doubling occurs along more than one crystallographic axis, which we did not observe.

Concerning the details of the low temperature crystal structure we determined, we see that the doubling of the unit cell along aa is predominantly due to a zig-zag displacement (‘puckering’) of the Cs ions, with a smaller zig-zag displacement of the centers of mass of the O2−_2^- dumbbells. The staggered displacement of Cs ions was foreseen by DFT [20]. On the other hand, the same DFT also predicted a tilting relative to the cc-axis of the O2−_2^- dumbbells, a scenario with which our data are inconsistent.

Looking at the refined low temperature commensurate structure we can compare it to the incommensurate structure visible for 80≃T≃20080 \simeq T \simeq200 K by looking at the amplitude of the displacement of the Cs ions and O2−_2^- dumbbells respectively. The amplitude of the sinusoidal modulation of Cs ion positions in the incommensurate phase was 0.0564(9) lattice units, whereas the amplitude of the modulation, given by the deviation of the Cs ion position along the bb-axis relative to the ideal coordinate of y=0.75y = 0.75 in the commensurate phase, is 0.0358(1) lattice units. These values are rather comparable. The amplitude of the O2−_2^- dumbbell displacement was 0.0090(7) lattice units in the incommensurate phase, whereas in the commensurate phase it is 0.00284(4) lattice, which is somewhat different.

We see in fig. 2 that cooling below the transition from a tetragonal phase at TS1=192T_{\mathrm{S1}} = 192 K the aa lattice parameter of the average structure shrinks significantly more rapidly down to TS2=71.5T_{\mathrm{S2}} = 71.5 K, reducing by ∼1.2%\sim1.2\% compared to the bb lattice parameter which reduces by just 0.34% and the cc lattice parameter which barely changes (<0.1%< 0.1\% change). At TS2T_{\mathrm{S2}} the aa lattice parameter (when indexed using the ImmmImmm space group) suddenly increases. A possible explanation for this is that the puckering of the Cs ions in the incommensurate structure at comparatively high temperature, when the oxygen molecules’ orientation is fluctuating significantly, results in strain on the lattice due to under-bonding. As the fluctuations decrease on cooling eventually the Cs ions and oxygen molecules can bond, leading to a lock-in to the commensurate structure, a decrease in lattice strain and a corresponding slight increase in aa.

It was already noted that a staggered structure would likely have an impact on the strength of the superexchange interaction between the s=1/2s = 1/2 units, due to differing orbital overlap between the oxygens and the cesium ions [21]. NMR data modelled under the assumption of staggered tilts of the O2−_2^- dumbbells indicated the formation of antiferromagnetic 1d spin chains in CsO2_2. The temperature dependence of the magnetic exchange parameters determined from the NMR was understood in relation to the assumed staggered dumbbell tilts together with a model based changing orbital overlap as a result of librations of the dumbbells. Our findings show that in fact the dumbbells are hardly tilted at all, rather the cesium ion positions are staggered. One might anticipate that the overall effect could be similar, at least as far as superexchange is concerned. It would be interesting to re-model the NMR data based on the crystal structure we have determined.

We note that there is some discussion in the literature of orbital ordering in superoxides [21, 22, 26, 27, 28], which has an impact on the aforementioned spin chain formation. Orbital order is frequently accompanied by structural distortion, as in the cooperative Jahn-Teller effect [46]. In the tetragonal I4/mmmI4/mmm structure of CsO2_2, one can take the 2b(0,0,1/2)2b(0,0,1/2) Wyckoff position with the 4/mmm4/mmm site symmetry as the place where the O2−O_2^- molecules reside. The degenerate π∗\pi^* molecular orbitals of O2−O_2^- form basis of two-dimensional EuE_u site symmetry irrep, which has non-zero subduction frequency in the Σ4\Sigma_4 space group irrep. This implies that local distortions that belong to the site symmetry irrep EuE_u can induce global distortions that belong to the space group irrep Σ4\Sigma_4. In other words, the structural distortions obtained in the present work are compatible with the orbital ordering scenario. The ground state orthorhombic structure is consistent with the orbital pattern proposed by Riyadi et al. [20] although no additional doubling along the bb- and cc-axes suggested by the authors, due to tilting of the oxygen dimers, was detected in our diffraction data. The incommensurate phase implies partial occupancies of the πx∗\pi_x^* and πy∗\pi_y^* orbitals, modulated upon propagation through the crystal (i.e. an orbital density wave). A similar phenomenon takes place in some perovskite manganites containing octahedrally coordinated Mn3+\mathrm{Mn}^{3+} cations with electronic degeneracy. The modulated orbital states in these systems vary from achiral density waves [47] to chiral orbital helices [48].

Considering now the magnetic structure, we have confirmed that, as long anticipated, CsO2\mathrm{CsO}_2 is an antiferromagnet. We note that until now the details of the magnetic structure had remained unknown, with the only proposal being that of ferromagnetism within the abab-plane with an antiferromagnetic stacking between planes [43], which was formed on the basis of analogy with KO2\mathrm{KO}_2 [17] rather than a direct measurement. So, our data clarify this point at last, revealing ferromagnetism in the bcbc-plane and antiferromagnetism along aa. The total magnetic moment determined, 0.514μB0.514\mu_B, is reduced compared to the ideal spin only value, possibly due to fluctuations and/or reduced dimensionality [21].

The presence of a small component of the magnetic moment along the aa axis, as well as the larger component along bb, is significant. These orthogonal spin components are transformed by the same irrep of the paramagnetic PnmaPnma space group (table VI), implying bi-linear coupling between them. This is a typical case of antisymmetric Dzyaloshinskii-Moriya exchange underpinning the coupling at microscopic level [49]. Let us denote the corresponding magnetic order parameters as δ\delta and ξ\xi, respectively. The bi-linear invariant, δξ\delta\xi, in principle also involves single ion terms, however these are not relevant here due to the S=1/2S = 1/2 nature of the interacting spins. The δ\delta and ξ\xi order parameters are transformed by distinct irreps (mM5+mM_5^+ and mΣ3m\Sigma_3) of the tetragonal I4/mmmI4/mmm space group (table III), indicating that they would be decoupled if no structural distortion was present. Decomposition of the experimentally determined ground state crystal structure of CsO2\mathrm{CsO}_2 in respect of symmetrised displacive modes of the tetragonal I4/mmmI4/mmm reveals the presence of Γ1+(k=0)\Gamma_1^+(k = 0), Γ2+(k=0)\Gamma_2^+(k = 0), M5(k=1,1,1)M_5(k = 1,1,1) and Σ4(k=1/2,0,0)\Sigma_4(k = 1/2,0,0) modes. The latter has the largest amplitude (of 0.250.25Å), as expected for the primary order parameter. Further analysis of the allowed free-energy terms reveals that the Σ4(η1,η1∗,η2∗,η2)\Sigma_4(\eta_1,\eta_1^*,\eta_2^*,\eta_2) order parameter forms a trilinear invariant with mM5+(δ1,δ2)mM_5^+(\delta_1,\delta_2) and mΣ3(ξ1,ξ1∗,ξ2∗,ξ2)m\Sigma_3(\xi_1,\xi_1^*,\xi_2^*,\xi_2):

δ1η1ξ1∗+δ1η1∗ξ1−δ1η2ξ2∗−δ1η2∗ξ2−δ2η1ξ1∗−δ2η1∗ξ1−δ2η2ξ2∗−δ2η2∗ξ2\delta_1\eta_1\xi_1^*+\delta_1\eta_1^*\xi_1-\delta_1\eta_2\xi_2^*-\delta_1\eta_2^*\xi_2 -\delta_2\eta_1\xi_1^*-\delta_2\eta_1^*\xi_1-\delta_2\eta_2\xi_2^*-\delta_2\eta_2^*\xi_2

indicating that this structural distortion is responsible for the coupling of the orthogonal magnetic modes in the ground state orthorhombic structure. The term is reduced down to δ1η1ξ1∗+δ1η1∗ξ1−δ2η1ξ1∗−δ2η1∗ξ1\delta_1\eta_1\xi_1^*+\delta_1\eta_1^*\xi_1-\delta_2\eta_1\xi_1^*-\delta_2\eta_1^*\xi_1 for the case of single arm of the Σ\Sigma-star, as observed experimentally. The coupling is fully optimised when the mM5+(δ1,δ2)mM_5^+(\delta_1,\delta_2) order parameter takes the δ1=−δ2\delta_1=-\delta_2 direction (δ1,−δ2)(\delta_1,-\delta_2). This direction represents the symmetry of the experimentally observed secondary antiferromagnetic mode with the components of the spins along the orthorhombic aa-axis.

In conclusion, we have used neutron diffraction to perform a comprehensive study of the crystal and magnetic structure of CsO2\mathrm{CsO}_2. We find that an incommensurate crystal structure appears seemingly simultaneously with a transition from a tetragonal to an orthorhombic structure at 192 K. This incommensurate structure, which is modulated along bb, is composed of displacements of the cesium ions along the bb-axis and of smaller displacements of the O2−O_2^- dimers’ entire centre of mass. On cooling further below 72 K the modulated structure locks into a commensurate wavevector of (12,0,0)(\frac{1}{2},0,0), doubling the unit cell compared to the previously supposed orthorhombic crystal structure. Hints of a structure bearing some similarity to this had previously been suggested from DFT calculations. However, in both calculations and in other superoxides it was found that the O2−O_2^- dimers tilt with respect to the cc-axis, which is different to what is seen here. Magnetic order sets in below ∼10\sim10 K and as previously supposed this order is antiferromagnetic. The spins of the O2−O_2^- dimers modulate along the aa-axis and point predominantly along bb, albeit with a small component along aa. Thus, CsO2\mathrm{CsO}_2 is an interesting example of a structure–properties relationship. The large structural distortion, associated with the staggered cesium and oxygen displacements, activates antisymmetric exchange and couples the orthogonal magnetic modes directly observed in the neutron diffraction experiment. For applied magnetic field 2<B<42 < B < 4 T we see changes in magnetic Bragg peak intensity that are consistent with a spin flop transition in which the magnetic moments reorient to point along the aa-axis.

Note: Just before the completion of this manuscript we became aware of another work, ref. [50], in which neutron powder diffraction was used to examine the magnetic structure of CsO2\mathrm{CsO}_2. The authors also discuss the structural aspects of the order. The main difference to the present work compared to ref. [50] is in how the structural distortions are combined with the magnetic order. Taking the tetragonal I4/mmmI4/mmm space group as the common parent symmetry, the suggested structural distortions and the magnetic order double the aa- and bb-axes, respectively. In the ground state found in our work, both the primary structural distortions and the spin ordering double the same tetragonal axis (the aa-axis in our setting). In addition, the measurements presented in ref. [50] were insensitive to the small aa-axis component of the magnetic moment that we found.

Acknowledgments

We are grateful to Bella Lake and Ryota Ono for insightful discussions. Experiments at the ISIS Neutron and Muon Source were supported by an Xpress beamtime allocation RB1990392 from the Science and Technology Facilities Council. Raw data are available online [51].

References

  1. [1]S. Blundell, Magnetism in Condensed Matter (Oxford University Press, 2001).
  2. [2]E. R. Callen and H. B. Callen, Static Magnetoelastic Coupling in Cubic Crystals, Phys. Rev. 129, 578 (1963).
  3. [3]J. G. Rau, E. K.-H. Lee, and H.-Y. Kee, Spin-Orbit Physics Giving Rise to Novel Phases in Correlated Systems: Iridates and Related Materials, Annual Review of Condensed Matter Physics 7, 195 (2016).DOI
  4. [4]O. Volnianska and P. Boguslawski, Magnetism of solids resulting from spin polarization of p orbitals, Journal of Physics: Condensed Matter 22, 073202 (2010).DOI
  5. [5]P. W. Stephens and C. F. Majkrzak, Magnetic structure and dynamics in the alpha and β phases of solid oxygen, Phys. Rev. B 33, 1 (1986).pubmed.ncbi.nlm.nih.gov/9937875
  6. [6]F. Fernandez-Alonso, F. J. Bermejo, I. Bustinduy, M. A. Adams, and J. W. Taylor, Spin dynamics in liquid and rotationally disordered solid oxygen, Phys. Rev. B 78, 104303 (2008).
  7. [7]S. Klotz, T. Strässle, A. L. Cornelius, J. Philippe, and T. Hansen, Magnetic Ordering in Solid Oxygen up to Room Temperature, Phys. Rev. Lett. 104, 115501 (2010).pubmed.ncbi.nlm.nih.gov/20366482
  8. [8]W. Hesse, M. Jansen, and W. Schnick, RECENT RESULTS IN SOLID STATE CHEMISTRY OF IONIC OZONIDES, HYPEROXIDES, AND PEROXIDES, Prog. Solid St. Chem 19, 47 (1989).
  9. [9]M. A. Boesch, M. E. Lines, and M. Labhart, Phys. Rev. Lett. 45, 140 (1980).
  10. [10]J. Winterlik, G. H. Fecher, C. A. Jenkins, C. Felser, C. Mühle, K. Doll, M. Jansen, L. M. Sandratskii, and J. Kübler, Challenge of Magnetism in Strongly Correlated Open-Shell 2p Systems, Phys. Rev. Lett. 102, 016401 (2009).pubmed.ncbi.nlm.nih.gov/19257217
  11. [11]D. Arčon, K. Andle, M. Klanjšek, A. Sans, C. Mühle, P. Adler, W. Schnelle, M. Jansen, and C. Felser, Influence of O₂ molecular orientation on p-orbital ordering and exchange pathways in Cs₄O₆, Phys. Rev. B 88, 224409 (2013).
  12. [12]P. Adler, P. Jeglič, M. Reehuıs, M. Geiß, P. Merz, T. Knaflıč, M. Komelj, A. Hoser, A. Sans, J. Janek, D. Arčon, M. Jansen, and C. Felser, Every-type charge ordering transition in an open-shell i↑;p↓/i↓-electron compound, Science Advances 4, eaap7581 (2018), https://www.science.org/doi/pdf/10.1126/sciadv.aap7581.
  13. [13]R. H. Colman, H. E. Okur, W. Kockelmann, C. M. Brown, A. Sans, C. Felser, M. Jansen, and K. Prassides, Elusive Valence Transition in Mixed-Valence Sesquioxide Cs₄O₆, Inorg. Chem. 58, 14532 (2019).DOI
  14. [14]T. Knaflič, P. Jeglič, M. Komelj, A. Zorko, P. K. Biswas, A. N. Ponomaryov, S. A. Zvyagin, M. Reehuıs, A. Hoser, M. Geiss, J. Janek, P. Adler, C. Felser, M. Jansen, and D. Arčon, Spin-dimer ground state driven by consecutive charge and orbital ordering transitions in the anionic mixed-valence compound Rb₄O₆, Phys. Rev. B 101, 024419 (2020).
  15. [15]M. Ziegler, M. Rosenfeld, W. Kanzig, and P. Fischer, Strukturuntersuchungen an Alkalihyperoxiden, Helv. Phys. Acta 49, 57 (1976).
  16. [16]F. Halverson, COMMENTS ON POTASSIUM SUPEROXIDE STRUCTURE, J. Phys. Chem. Solids 23, 207 (1962).DOI
  17. [17]H. G. Smith, R. M. Nicklow, L. J. Raubenheimer, and M. K. Wilkinson, Antiferromagnetism in potassium superoxide KO₂, Journal of Applied Physics 37, 1047 (1966).DOI
  18. [18]M. Rosenfeld, M. Ziegler, and W. Kaenzig, Helv. Phys. Acta 51, 298 (1978).
  19. [19]J. C. Bonner and M. E. Fisher, Linear Magnetic Chains with Anisotropic Coupling, Phys. Rev. 135, A460 (1964).DOI
  20. [20]S. Riyadi, B. Zhang, R. A. de Groot, A. Caretta, P. H. M. van Loosdrecht, T. T. M. Palstra, and G. R. Blake, An-tiferromagnetic S = 1/2 Spin Chain Driven by p-Orbital Ordering in CsO₂, Phys. Rev. Lett. 108, 217206 (2012).pubmed.ncbi.nlm.nih.gov/23003298
  21. [21]M. Klanjšek, D. Arčon, A. Sans, P. Adler, M. Jansen, and C. Felser, Phonon-Modulated Magnetic Interactions and Spin Tomonaga-Luttinger Liquid in the p-Orbital Antiferromagnet CsO₂, Phys. Rev. Lett. 115, 057205 (2015).
  22. [22]T. Knaflič, M. Klanjšek, A. Sans, P. Adler, M. Jansen, C. Felser, and D. Arčon, One-dimensional quantum antiferromagnetism in the p-orbital CsO₂ compound revealed by electron paramagnetic resonance, Phys. Rev. B 91, 174419 (2015).
  23. [23]M. Miyajima, F. Astuti, T. Kakuto, A. Matsuo, D. P. Sari, R. Asih, K. Okunishi, T. Nakano, Y. Nozue, K. Kindo, I. Watanabe, and T. Kambe, Magnetism and High-magnetic Field Magnetization in Alkali Superoxide CsO₂, Journal of the Physical Society of Japan 87, 063704 (2018), https://doi.org/10.7566/JPSJ.87.063704.
  24. [24]F. Astuti, M. Miyajima, T. Fukuda, M. Kodani, T. Nakano, T. Kambe, and I. Watanabe, Antionogenic Magnetism Combined with Lattice Symmetry in Alkali-metal Superoxide RbO₂, Journal of the Physical Society of Japan 88, 043701 (2019), https://doi.org/10.7566/JPSJ.88.043701.
  25. [25]M. Miyajima, F. Astuti, T. Fukuda, M. Kodani, S. Iida, S. Asai, A. Matsuo, T. Masuda, K. Kindo, T. Hasegawa, T. C. Kobayashi, T. Nakano, I. Watanabe, and T. Kambe, Spin-gap formation due to spin-Peierls instability in π-orbital-ordered NaO₂, Phys. Rev. B 104, L140402 (2021).DOI
  26. [26]A. K. Nandy, P. Mahadevan, P. Sen, and D. D. Sarma, KO₂: Realization of Orbital Ordering in a p-Orbital Sys- tem, Phys. Rev. Lett. 105, 056403 (2010).
  27. [27]M. Kim, B. H. Kim, H. C. Choi, and B. I. Min, Antiferromagnetic and structural transitions in the superoxide KO₂ from first principles: A 2p-electron system with spin-orbital-lattice coupling, Phys. Rev. B 81, 100409 (2010).DOI
  28. [28]M. Kim and B. I. Min, Temperature-dependent orbital physics in a spin-orbital-lattice-coupled 2p electron Mott system: The case of KO₂, Phys. Rev. B 89, 121106 (2014).DOI
  29. [29]I. V. Solovyev, New Journal of Physics 10, 013035 (2008).
  30. [30]A. Helms and W. Klemm, Die Kristallstrukturen von Rubidium- und Cäsium-Oxyd, Z. anorg. allg. Chem. 242, 33 (1939).DOI
  31. [31]J. U. Hoffmann and M. Reehuis, JLSRF 4, 129 (2018).
  32. [32]D. Többens, N. Stüsser, K. Knorr, H. Mayer, and G. Lampert, Materials Science Forum 378, 288 (2001).
  33. [33]L. C. Chapon, P. Manuel, P. G. Radaelli, C. Benson, L. Perrott, S. Ansell, N. J. Rhodes, D. Raspino, D. Duxbury, E. Spill, and J. Norris, Wish: The New Powder and Single Crystal Magnetic Diffractometer on the Second Target Station, Neutron News 22, 22 (2011).DOI
  34. [34]R. M. Ibberson, Design and performance of the new supermirror guide on HRPD at ISIS, Nuclear Instruments and Methods in Physics Research Section A 458, 454 (2001).
  35. [35]J. Rodríguez-Carvajal, Recent developments of the program FULLPROF, Commission on powder diffraction (IUCr). Newsletter 26, 12 (2001).
  36. [36]J. Rodríguez-Carvajal, Recent advances in magnetic structure determination by neutron powder diffraction, Physica B: Condensed Matter 192, 55 (1993).
  37. [37]V. Petříček, M. Dušek, and L. Palatinus, Crystallographic Computing System JANA2006: General features, Zeitschrift für Kristallographie - Crystalline Materials 229, 345 (2014).
  38. [38]ISODISTORT, ISOTROPY Software Suite, https://iso.byu.edu, accessed: 10/11/2022.
  39. [39]B. J. Campbell, H. T. Stokes, D. E. Tanner, and D. M. Hatch, ISODISPLACE: a web-based tool for exploring structural distortions, Journal of Applied Crystallography 39, 607 (2006).
  40. [40]K. Momma and F. Izumi, VESTA3 for three-dimensional visualization of crystal, volumetric and morphological data, Journal of Applied Crystallography 44, 1272 (2011).DOI
  41. [41]M. I. Aroyo, A. Kirov, C. Capillas, J. M. Perez-Mato, and H. Wondratschek, Acta Cryst. A 62, 115 (2006).
  42. [42]S. Riyadi, S. Giriyapura, R. A. D. Groot, A. Caretta, P. H. V. Loosdrecht, T. T. Palstra, and G. R. Blake, Ferromagnetic order from p-electrons in rubidium oxide, Chemistry of Materials 23, 1578 (2011).DOI
  43. [43]M. Labhart, D. Raoux, Kanzig, and M. A. Bosch, Magnetic order in 2p-electron systems: Electron paramagnetic resonance and antiferromagnetic resonance in the alkali metal hyperoxides KO2, RbO2, and CsO2, Phys. Rev. B 20, 53 (1979).
  44. [44]A. Zumsteg, Magnetische und kalorische Eigenschaften von Alkali-Hyperoxid-Kristallen, Ph.D. thesis, ETH Zürich (1973).
  45. [45]P. W. Stephens, Magnetic form factor of the O₂ molecule, Phys. Rev. B 31, 4491 (1985).pubmed.ncbi.nlm.nih.gov/9936382
  46. [46]J. Kanamori, Crystal Distortion in Magnetic Compounds, Journal of Applied Physics 31, S14 (1960).DOI
  47. [47]N. Perks, R. Johnson, C. Martin, L. Chapon, and R. Radaelli, Associated Equipment.
  48. [48]D. D. Khalyavin, R. D. Johnson, F. Orlandi, P. G. Radaelli, P. Manuel, and A. A. Belik, Emergent helical texture of electric dipoles, Science 369, 680 (2020), https://www.science.org/doi/pdf/10.1126/science.aay7356.
  49. [49]I. Dzyaloshinsky, A thermodynamic theory of “weak” ferromagnetism of antiferromagnetics, Journal of Physics and Chemistry of Solids 4, 241 (1958).
  50. [50]T. Nakano, S. Kontani, M. Hiraishi, K. Mita, M. Miyajima, and T. Kambe, Antiferromagnetic structure of alkali metal superoxide CsO₂ (2023), arXiv:2305.07189 [cond-mat.str-el].DOI
  51. [51]R. A. Ewings, A. S. Gibbs, and A. D. Fortes, The structure of CsO₂ (2020).

Paper details

vlnc:W_gtuzrvgm
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents