Superconductivity in the Chalcogens up to Multimegabar Pressures

Several runs were made with decreasing sample sizes and diamond culet dimensions to successively higher pressures. Measurements are made as a function of temperature. Representative results are shown in Figs. 2-4. No superconductivity was detected up to approximately 90 GPa. At 93 GPa, however, a peak characteristic of the supercondcting transition to the sperconducting state was observed.[15] T_c is identified as the temperature where the signal goes to zero on the high temperature side (e.g., Fig.4), which is the point at which magnetic flux completely enters the sample. [15] Two peaks are clearly seen at $\sim$10-12 K and $\sim$17 K. The second broad peak at lower temperatures arises from the sample outside of the flat culet, where pressure is considerably lower than in the middle of the culet. At pressures over 190 GPa the peak at 17 K becomes visibly split. This splitting is artificial and only reflects the fact that the signal amplitude has increased substantially with respect to the background.

The background signal in our measurements appeared to be ferromagnetic, as its phase is approximately opposite to that of the signal from the sample (diamagnetic) (see Fig. 2). Because the background signal changes smoothly with temperature, we can separate the signal from the backround by the simple procedure illustrated in Fig. 3. We measure amplitude and phase of a sum of signal and background with the lock-in technique. The signal changes very abruptly in the vicinity of the superconducting transition, allowing us to see these changes both in amplitude and phase (Fig. 3). It is straightforward to interpolate the background in the range of the superconducting transition with a smooth polynomial function. The total signal can be represented as the complex variable ${\bf U}=A_{S}e^{i\phi_{S}}$, and the interpolated background as ${\bf B}=A{{}_{B}}e^{i\phi_{B}}$; our signal is then ${\bf S}=A_{S}e^{i\phi_{S}}$=U-B (the difference of two complex variables).

The signal with background subtracted is shown in Fig. 4. As in previous work, [14] we observed the appearance of a T_c signal at $\sim$93 GPa at 10 K. T_c gradually increases and upon entering the $\beta$-Po phase at 165 GPa it jumps to 17.0($\pm$0.5) K. For the next 35 GPa, it still slowly increases, reaching the a weak maximum of 17.3($\pm$0.5) K at 180-200 GPa; it then decreases on further compression, dropping to 15 K at 231 GPa.

Two sets of experiments were performed on Se. In the first experiment, we used 99.999% pure amorphous Se as the starting material with NaCl as pressure transmitting medium. Raman spectra showed that Se crystallized at around 11 GPa.[21] A measurable T_c signal appeared at $\sim$25 GPa at 5.8 K, in agreement with previous studies.[22] Figure 5 shows the dependence of resistance and its modulations by the magnetic field as a function of temperature. In both cases, the onset of the superconducting transition in the sample is clearly seen. After reaching 35 GPa, the pressure was released. The T_c signal was observable down to 18 GPa and decreased with releasing pressure.

In the second set of experiments amorphous Se was pressurized up to 180 GPa and T_c was measured by susceptibility technique. Currently, it is difficult to measure T_c signals of $\sim$30 $\mu$m in diameter samples with this technique if the temperature of superconducting transition is below 4 K due to a developing background signal which is probably related to paramagnetic signal from the gasket (tungsten inset and Ni-Cr-Al alloy gasket material). We were able to detect the superconducting signal in the bcc phase from 140 to 180 GPa. On decompression below 130 GPa, the measurable signal disappeared.

The pressure dependences of T_c in the group VIa elements are shown in Fig. 6, together with the observed transition pressures between the phases. It should be noted that these boundaries were determined by x-ray diffraction at room temperature while the observations of superconductivity were done at temperatures ranging from 2 to 18 K. It may be assumed that the phase boundaries at low temperatures (if any) would be shifted in pressure. The observed structures are illustrated in Fig. 7.

Tellurium becomes superconducting in the monoclinic phase at $\sim$4 GPa. As shown in Fig. 6c, T_c rises linearly with pressure in this phase upon transforming to the bco structure, it levels off and starts to decrease, passing through the field of stability of the $\beta$-Po phase. T_c then jumps by almost factor of three and starts to come down again. This jump in T_c happens at 35 GPa while room temperature phase transition to bcc phase happens at 27 GPa.

Under ambient conditions, Se can be found in trigonal, monoclinic or amorphous forms. Akahama et al. [7] showed that monoclinic and trigonal Se become metallic at 12 and 23 GPa respectively. Under pressure amorphous Se crystallizes in a trigonal structure and becomes metallic at 12 GPa. It was shown [7] that if $\alpha$-monoclinic Se used as the starting material, it transforms at 12 GPa to an unidentified metallic phase which is superconducting at 17 GPa but no pressure dependence of T_c was given. Figure 6b shows our results for Se combined with the results of Ref. [22] The trend in the T_c dependence for Se is remarkably similar to that of Te, although unlike the latter, Se becomes metallic and superconducting in an unidentified phase from 14 to 23 GPa. The structural sequence is identical to that of Te starting from 23 GPa where Se enters the monoclinic phase. Unlike the situation for Te, T_c within this phase changes little with pressure. When Se is compressed further, the abrupt decrease in T_c starts at 33 GPa which can be attributed to the transition to the bco phase, which happens at 27 GPa at room temperature. There are no experimental measurements of T_c for Se from 60 to 150 GPa, but theoretical calculations [24] predict a decrease of T_c in the $\beta$-Po phase, followed by an increase to much higher values around 140 GPa where Se transforms to the bcc structure at 300 K. The values of T_c in Se at 150-170 GPa measured here are very close to those calculated in Ref. 24.

The structures of the semiconducting phases of S below the 90 GPa transition is not known. It becomes metallic and superconducting in the bco phase. Unlike the behavior of Se and Te, T_c in this phase of S increases with pressure and jumps abruptly upon entering the $\beta$-Po-type phase (Fig. 6a). In the $\beta$-Po phase, T_c slowly decreases with further compression. This decrease is consistent with the theoretical calculations of Rudin et al. [12], which predicted a decrease in the electron-phonon coupling parameter $\lambda$ as the sc phase is approached.[12]

It can be clearly seen from the behavior of the T_c for the three elements that the transition temperatures cannot be simply explained by the existence of phase transitions and similar values for electron-phonon coupling alone. On the other hand, the similar trends are broadly consistent with changes towards higher symmetry structures (transformation of layered phases to the close packed structures at the highest pressures) with weakening of the directional covalent bonding, as we now discuss.

Mössbauer studies of Te showed a decrease in the quadrupole splitting with pressure that was ascribed to strengthening interactions between the neighboring chains and weakening of the covalent bonds within the chain.[27] The splitting disappeared when Te was pressurized into the bcc phase where no covalent bonding is present. Thus we assume that the continuous weekening of the covalent bonds with increasing pressure is due to increasing screening by free carriers due to increasing density of states at the Fermi level.

To analyse the possible T_c increase in the lower pressure phases of the chalcogens due to the change in phonon spectra, we estimated T_c in the lower symmetry ”layered” structures (i.e monoclinic, bco) using the Allen-Dynes formula [25] with literature values of $\lambda$ (electron-phonon coupling constant) and $\mu^{*}$ (effective Coulomb repulsion potential). We assumed $\mu^{*}$=0.1 and varied $\lambda$. The values of $<\omega_{log}>$ and $<\omega^{2}>$, the averages of the phonon frequencies weighted to represent the strength of the electron-phonon coupling, were estimated with the phonon density of states shown in Fig. 8. The density of states was assumed to have two cutoff Debye frequencies with values typical for Se and Te corresponding to intralayer bond-stretching and inter-layer vibrations. In our estimates of T_c versus pressure, the lower energy peak which we attributed to interlayer vibrations was allowed to move towards higher energies with pressure increase while the higher energy peak representing intralayer bond-stretching modes was assumed to be pressure-independent. The values of T_c obtained in these calculations are within the range of the experimentally observed T_c for Se and Te in their monoclinic and bco structures. However, T_c is virtually independent of phonon spectra but instead depends on $\lambda$. From experimental data, $\lambda$ increases with pressure, meaning that $N(E_{F})<I^{2}>$ increases faster than $M<\omega^{2}>$ (which does not change substantially in our simplified model), where$\lambda=\frac{N(E_{F})<I^{2}>}{M<\omega^{2}>}$, according to McMillan [26]. Here $N(E_{F})<I^{2}>$ is the Hopfield parameter, which is generally assumed to be inversely proportional to volume in simple s-p metals [26]. Thus, the increase in T_c observed in the monoclinic phase of Te and the unidentified phase of Se where covalent bonding is still significant is most probably governed by the increase in Hopfield parameter with pressure. Sulfur has increasing T_c in its bco phase where covalent bonding is still present.

The bco $\rightarrow$ $\beta$-Po $\rightarrow$ bcc structural sequence may also be possible for oxygen, although the stability of its diatomic state appers to extend to very high pressures ($>$ 100 GPa [28]. It remains to be determined at what pressures the oxygen molecules dissociate (at low temperature), and what crystal structures would form. The additional complication with oxygen is its magnetic properties, which could play important role in determining T_c. If oxygen follows a similar structural trend in its nonmolecular state, a similar T_c dependence might be expected. Recent calculations [29] suggest that could be in 1 TPa range.

We have measured T_c of S up to 231 GPa, a record pressure for both superconductivity and magnetic susceptibility techniques. The value of T_c at the highest pressures (15 K above 200 GPa, decreasing from 17 K at 160 GPa) in accord with the original theoretical predictions for the hypothetical bcc phase [11] and more recent calculations for the $\beta$-Po structure [12]. The superconducting behavior of Se was observed in an unidentified phase from 14 to 20 GPa. The T_c for Se is also in very good agreement with recent first-principles calculations [24] (assuming $\beta$-Po, bcc, fcc structures). The data obtained in the present work, previously published experimental data, and recent theoretical calculations for the chalcogens allow us to conclude that at lower pressures T_c is controlled at least in part by changes in covalent bonding. In higher pressures the bcc phases of these materials, the behavior of T_c resembles normal p metals and decreases with pressure.

We thank A. Liu for sharing results on Se prior to publication and S. Rudin and P. Dera for helpful discussions. This work was supported by the National Science Foundation.