Phase transition of fluid H2 inducing ortho-para conversion
Raman spectra from pure rotation modes, S0 (J) (J = 0 to 2), and vibration-rotation modes, Q1 (J) (J = 0 to 3), of H2, measured at each pressure, are shown in Fig. 1, where J is the rotational quantum number. The S0 (3) peak was also measured, however the intensity was very weak, so that was omitted from the analysis. Every S0 (J) peak showed a decrease in peak height and linewidth expansion with an increase in pressure up to 2.05 GPa. In case of Q1 (J) modes, only the Q1 (0) peak disappeared under pressure above 0.68 GPa; further pressurization caused increase of overlap between the Q1 (J) (J = 1 to 3) peaks, which looks like one peak under pressure above 1.48 GPa.
Line shapes of the S0 (J) peaks were analyzed using Lorentz functions; pressure dependence of Raman shift is shown in Fig. 2a. All S0 (J) mode show monotonous changes in Raman shifts except a sharp risie at around 0.5 GPa as shown in an enlarged view of Fig. 2b. To estimate the transition pressure Pc, two tangent lines were drawn on the Raman shift-pressure plots of S0 (J). The value of Pc is estimated from the position of intersection point between the two lines, which is obtained to be in average 0.56 GPa for the S0 (J) peaks. The S0 (J) peaks are the Stokes spectral lines caused by the rotational transition, J → J + 2, respectively, as shown in Fig. 2c; the selection rule for the rotational and vibrational quantum numbers has ∆ J = +2 and ∆ν= 0 in this case5. The magnitude of Raman shift for each S0 (J) mode corresponds to the transition energy. Correlation of the energy levels of J could be changed under pressure above 0.56 GPa. Pressure dependences of full widths at half maximums, FWHMs, of S0 (J) peaks also show anomalies at around 0.56 GPa, as marked with an arrow in Fig. 2d. In addition to the anomaly observed in the pressure dependence of Raman shift, sharp widening of the linewidth could be caused by pressure-induced phase-transition of H2. The pressure changes in relative intensities for the S0 (J) modes normalized using that for S0 (1) were shown in Fig. 2e. The intensity for the S0 (0) mode is almost constant under pressure up to 1 GPa but slightly increases with increase the pressure beyond 1 GPa; it means the increase of transition probability, J = 0 to J = 2, in the para-H2. In contrast that for the S0 (2) mode is almost constant up to 2.04 GPa.
The Q1 (J) peaks overlapping in the narrow frequency range, were decomposed using multiple Lorentz functions as shown in Fig. 3a. The vibration-rotation spectrum consists of the Stokes Q1 (J) lines, which are generated by the vibrational transition, n = 0 → n = 1, and the rotational transition, J → J, respectively5, as shown in Fig. 3b. The line-shapes for Q1 (J) modes drastically change in the pressure range between 0.68 and 1.29 GPa. The Q1 (0) band disappears at the pressure above 0.68 GPa; the Q1 (2) band shows the increase of peak intensity. Although the Q1 (1) band shows a decrease of peak height and an increase of FWHM with an increase in pressure, the Q1 (3) peak intensity looks constant even though the pressure is increasing.
Figure 3c and d show the pressure dependences of Raman shifts and FWHMs for Q1 (J) modes of H2, respectively, which show the rapid increasing at pressure around 0.5 GPa as same as those for S0 (J) modes. The Raman shifts of Q1 (J) modes except Q1 (0) one showed monotonous changes with increasing the pressure above 0.5 GPa. The intensity for Q1 (0) mode was broaden and disappeared under pressure above 0.68 GPa. The value of Pc on average for Q1 (J) modes except Q1 (0) one was estimated at 0.56 GPa, which is the same value for S0 (J) modes; the data for Q1 (0) mode were not enough to obtain the value Pc. At pressures above 0.68 GPa, the FWHMs of Q1 (1) and Q1 (3) modes increased with increase of pressure. On the other hand, FWHM for Q1 (2) mode showed constant in the pressure dependence above 1.0 GPa. The pressure changes in relative intensities for the Q1 (J) modes were normalized where using those Q1 (1) as shown in Fig. 3e. The intensity for Q1 (0) mode decreased with an increase in pressure and disappeared at the pressure above Pc. The Q1 (2) mode showed drastic enhancement of the intensity under pressure above Pc; the intensity for Q1 (2) mode overtakes that for Q1 (1) under pressure higher than about 1 GPa. It means that the vibrational-excited state of J = 2 increased with pressure in comparison with that of J = 1, suggesting progression of the pressure-induced ortho-para conversion at room temperature. The ortho-para conversion initially revealed at low temperature20, has recently been found in catalytic reactions23 or in high-pressure condition at low temperature24.
The following interpretation can be considered for the pressure dependence of peak intensities for Q1 (J) modes. Para-H2 has an isotropic orientation of nuclear spins in case of J = 0. When H2 molecules get close to each other through compression, the inter-molecular interaction becomes strong; the molecular rotation is suppressed and causes anisotropy of the orientation. For this reason, the state of J = 0 is mixed with that of J = 2 by pressurization; the increase of intensity of Q1 (2) occurs in parallel with the decrease of Q1 (0). If the state of J = 1 for ortho-H2 are occupied in all directions with equal probability, H2 molecules do not orient in any special direction. Pressurization generally suppresses the rotational motion, resulting in anisotropy for ortho-H2, as well as the case of para-H2. Since the state of J = 3 is mixed into that of J = 1, depressing of the Q1 (1) peak intensity should be accompanied with enhancement of the Q1 (3) peak intensity. Contrary to such assumptions, in fact, the intensity of Q1 (3) did not increase significantly though diminishing that of Q1 (1); only Q1 (2) mode showed drastic increases of intensity. Therefore, it is speculated that the pressurization beyond 0.56 GPa accelerates ortho-para conversion of H2 even at room temperature.
Fluid H2 causing anomaly of graphite structure under megapascal-class pressure
We also carried out the Raman measurements of H2 pressurized with graphite, which were compared with those of pure H2 in order to confirm the reproducibility of transition which was found in pure H2 and to explore the possibility of intercalation of H2 into graphite. In addition to the spectra for S0 (J) (J = 0 to 2) and Q1 (J) (J = 0 to 3) modes of H2, G-band from graphite are also shown in Fig. 4. The signal to noise (SN) ratio of G band became low at pressure around 0.15 GPa as shown in Fig. 4a, which was improved by further pressurization. It is considered that the SN ratio temporarily deteriorated at low pressure due to the change in the reflectance at the interface between diamond and compressed H2. All the S0 (J) modes except S0 (2) modes, and Q1 (J) modes showed the minimums in the intensities at the pressure around 0.47 GPa as shown in Fig. 4b and c, respectively. In order to check the reproducibility of such strange behaviors, the Raman spectra of 0.47 GPa were measured again after an interval of several hours. All peak positions obtained in the second observation shifted slightly in comparison with those of the first observation. The intensities particularly for S0 (J) and Q1 (J) modes drastically increased in the second observation of 0.47 GPa. In this way pressurization at around 0.5 GPa caused the instability of vibrations. On the other hand, under the pressure more than 0.6 GPa, the S0 (2) mode which was once difficult to observe under pressure above 0.33 GPa, was revived again; after that their linewidths and intensities increased when increasing the pressure. The intensity of G-band also grew at a pressure higher than 0.68 GPa. Overlapping of the linewidths for Q1 (J) modes proceeded with an increase in pressure, while the intensity of Q1 (0) mode flattened and expanded with increasing pressure, which was difficult to analyze under pressure above 0.68 GPa. This means that such a series of drastic variations in the spectral intensities is caused by a major change in the reflectivity at the interface between the upper diamond anvil and H2. It can be considered that an alternation in the reflective index of H2 was brought by something like a phase transition or change of state in H2.
The line-shapes of G-band, the S0 (J) and Q1 (J) peaks, respectively, were analyzed using Lorentz functions; the pressure dependence of each Raman shift is shown in Fig. 5. The values of Raman shift of G-band which were observed at around 0.5 GPa, caused fluctuation of wave numbers in the range between 1583 and 1586 cm-1 as shown in Fig. 5a. The magnitude of FWHM for the G-band also showed fluctuation at around 0.5 GPa, after that expanded with increasing pressure as shown in Fig. 5b except for the range from 0.9 to 1.3 GPa. Such a significant increase of FWHM of the G-band under pressure beyond around 0.5 GPa suggests a shortening of phonon-relaxation time for graphite due to scattering by the densified H225,26 which may be accommodated in the interlayer space of graphite. Because the molecular size of H2 is smaller than the interlayer distance or the diagonal length of benzene ring composing the honeycomb lattice, H2 may be accommodated and densified in the nano space of graphite. In short, it means that the densified H2 changes the phonon-lattice relaxation mechanism of graphite. We suppose that the change observed at around 0.5 GPa is one of the proofs of intercalation of H2 into graphite.
The pressure dependences of Raman shift and FWHM for the S0 (J) and Q1(J) modes of H2, which were obtained in graphite-H2 mixture, are expressed in Fig. 5c to f. The S0 (J) and Q1 (J) modes also caused flip-flop of the Raman frequencies and FWHMs at the pressure where the fluctuations were observed in the G-band. Although the variation in the pressure dependence observed in graphite-H2 mixture is larger than that for pure H2, both this flapped variation of data in common. We were also concerned that only the S0 (0) mode in graphite-H2 causes large dispersion of the Raman shift under pressure above 1.6 GPa, which was not emerged in pure H2. As a result of the peak decomposition, the pressure changes in Q1 (J) peak intensities for H2 in graphite-H2 mixture were almost same as those of pure H2 in Fig. 3a. Although the magnitudes of linewidths for Q1 (1) and Q1 (3) modes increased with an increasing of the pressure, the that of Q1 (2) showed constant not only pure H2 but also in the graphite-H2 mixture.
We suppose that such a series of abnormal behaviors following the flip-flops are probably caused by a phase transition in H2: a liquid phase is one of the candidates for the post-phases at room temperature, which has been previously unreported. We need further verification through alternate investigation and experiment. The values of Pc estimated using the pressure changes in Raman shift for the S0 (J) and Q1 (J) modes of pure H2 and graphite-H2 mixture, are summarized in Table 1. The values of Pc estimated using the results for the S0 (J) and Q1 (J) modes obtained in graphite-H2 were calculated to be 0.46 and 0.48 GPa in average, respectively; the difference is slightly 0.02 GPa, which is within the margin of error. In any case, the magnitude of Pc on H2 obtained in graphite-H2 mixture is smaller than that in pure H2. It means that graphite plays a role of reducing the magnitude of Pc of H2. Therefore, it is speculated that graphite suppresses the vibration of H2 through accommodation of H2 in the interlayer space of graphite or adsorbed on the surface even at room temperature.
According to the high-resolution transmission-electron microscope (TEM) image as shown in Fig. 6a, it was confirmed that the inner part of each grain before pressurization showed a well-stacked graphite-structure with almost all the same spacing between graphene layers. All the XRD experiments gave Debye-Scherrer rings, which have sufficient quality to do the structural analysis as shown in Fig. 6b. The uniformed diffraction ring evidenced the appearance of an unknown peak, which is shown in Fig. 6b, was also confirmed at each pressure, though it was never observed in the pristine sample. The 2D XRD-images from which some contaminations were removed, were integrated along the rings with 0.02 degrees of 2q step. The diffraction pattern of graphite obtained at atmospheric pressure, showed the 002, 100, 101, and 004 reflections from hexagonal symmetry with a space group P63/mmc as shown in Fig. 6c. Pressurization of the graphite-H2 mixture caused the unknown reflection at around 16 degrees, which also remained in the sample released from the pressure. Furthermore, the second unknown peak appeared at around 15 degrees only at the pressure of 0.46 GPa. We confirmed these peaks do not come from the metal gasket. No peaks were observed at angles lower than 2q =10 degrees. The lattice parameters, a and c, were obtained by fitting of the peak intensities using Gaussian functions. The values of lattice parameters of 1 atm were determined to be a0 = 2.4556 ± 0.0004 and c0 = 6.7053 ± 0.0002 Å, respectively. The pressure dependence of lattice parameters is shown in Fig. 6d. The c-axis length monotonously decreases when intensifying the pressure above 0.16 GPa.
On the other hand, the a-axis length compressed with H2 continues to be elongated to 0.6 GPa with some degrees of fluctuations. The maximum of a-axis length is longer than the a0-axis length by 0.27%, which was almost same as that of MWCNTs. Under pressure above 0.6 GPa the a-axis length decreases with an increase of pressure, which does not become shorter than the a0-axis length even if the pressure is beyond 2 GPa. We suppose that the change in the a-axis length indicating the maximum is generated by H2 intercalated into the interlayer space of graphite or the nano-space of the honeycomb lattice. The compressibility of H2 which changes before and after the phase transition, should be reflected on the change in the a-axis length. The unknown peak appeared next to the 100 reflection under pressure above 0.11 GPa; the d-value was obtained at d ≈ 2.5, which is about 2% larger than the a-axis length. The values of lattice parameters of the sample released from the pressure were basically restored to the initial state. On the other hand, the unknown peak observed at around d ≈ 2.5, remained after releasing the pressure from the sample. Appearance of the unknown extra peaks by pressurization suggests that H2 is accommodated in the interlayer space of graphite. We speculate that the alternating stacking rule of graphite was partially broken by filling H2 into the inter layer space.