betaFeedback

WorksPhysics and Astronomy

Melting Points and Formation Free Energies of Carbon Compounds with Sodalite Structure

Abstract

Using first-principles calculations, we investigate the melting temperatures TmT_{\rm m} and formation free energy of carbon compounds with sodalite structures, XXC6 _6, XXC10 _{10}, and XXC12 _{12}, where XX is F, Na, Cl, and so on. These compounds are expected to be phonon-mediated superconductors exhibiting high transition temperatures TcT_{\rm c} of up to about 100 K. We estimate TmT_{\rm m} as a function of pressure PP by using the first-principles molecular dynamics method and show the results as phase diagrams on the PP-TT plane together with the results of TcT_{\rm c}. It indicates that the TmT_{\rm m} of NaC6_{\rm 6}, which has a TcT_{\rm c} up to about 100 K, is about 13001300 K or more at P=30P=30 GPa. Furthermore, the TmT_{\rm m} of FC6_{\rm 6} is about 2200 K even at P=0P=0 GPa, where its TcT_{\rm c} is about 80 K. Similar results are obtained for FC10_{\rm 10} and ClC10_{\rm 10} systems. These results suggest that some compounds can stably exist as high-temperature superconductors even at room temperature and pressure. To examine the feasibility of synthesizing these compounds, we estimate the formation enthalpies and formation free energies. These results suggest that NaC6_6 could be formed under a sufficiently high pressure of about 300 GPa and a high temperature of about 6500 K.

Introduction

Since the high-temperature superconductors of hydrogen compounds with a sodalite structure, such as YH6_6 and lanthanum decahydride (LaH10_{10}),1–7) were predicted to indicate TcT_{\mathrm{c}} over 250 K, much effort has been made to clarify the superconductivity of these compounds and/or to find new hydrogen compounds.1–7) Based on theoretical results,1–5) experiments8–11) succeeded in finding these superconductors whose TcT_{\mathrm{c}} is almost close to the result of first-principles calculations. This high-temperature superconductivity (HTS) is caused by phonon-mediated attraction, and the mechanism of the superconductivity is conventional. Therefore, TcT_{\mathrm{c}} is determined mainly by two parameters: the electron-phonon coupling constant λ\lambda and the characteristic phonon frequency ωlog⁡\omega_{\log}. When both parameters are simultaneously large, HTS can be expected. In fact, many hydrogen compounds with a sodalite structure have a large electron-phonon coupling constant λ\lambda, that is larger than about 2.0.1–6) Typical values of ωlog⁡\omega_{\log} are around 1000 K.

In these hydrogen compounds, high pressure may stabilize these characteristic structures and lead to a high phonon frequency. These materials are only stable over a limited pressure and the phonon frequency decreases with pressure.1–6) Therefore, hydrogen compound HTS can only be achieved with devices using diamond anvil cells at this stage. To realize practical devices, the required pressure should be as low as possible, preferably normal pressure.

A material constructed with carbon atoms, such as diamond, has a high phonon frequency up to about 2000 K at atmospheric pressure. Boron-doped diamond has been studied as a candidate of phonon-mediated HTS, and its TcT_{\mathrm{c}} reaches ∼ 25 K at atmospheric pressure.12–15) Intercalated graphite and alkali-doped C60_{60} compounds are known carbonate superconductors with high TcT_{\mathrm{c}} of up to ∼ 33 K.16–19)

On the basis of the intuitive idea that the sodalite structure plays an important role in HTS, carbon compounds with a sodalite structure have been proposed as candidates for HTS materials.20–25) Based on the density functional theory, materials composed only of carbon atoms are known to be insulators with a large charge gap.26) By combining them with another element, XX, to form a compound, carriers are introduced into the system,22–25) where XX is Li, Na, Cl, or so on. The crystal structures of the compounds are shown in Figure 1, where (a) XC6\mathrm{XC}_{6} and (b) XC10\mathrm{XC}_{10} are shown as representative examples. XC12\mathrm{XC}_{12} has a structure similar to that of XC6\mathrm{XC}_{6} shown in Figure 1(a), except for one XX atom.

Figure 1. (Color online) Structures of the sodalite-type compounds (a) XC6\mathrm{XC}_{6} and (b) XC10\mathrm{XC}_{10}, where large spheres represent XX-atoms and small spheres are carbon. Here, the crystal structure of XC12\mathrm{XC}_{12} corresponds to (a) with one XX atom removed.

By first-principles calculations, it has been shown that some of these compounds exhibit TcT_{\mathrm{c}} of about 100 K even at atmospheric pressure.20,22–25) However, the existence of XC6\mathrm{XC}_{6}, XC10\mathrm{XC}_{10}, and XC12\mathrm{XC}_{12} has not yet been experimentally demonstrated. Furthermore, their stabilities at finite temperatures neither have been clarified. Even if these are stable near T=0T = 0, they will not be practical if the crystals become unstable below room temperature.

In this study, we investigated the stability of carbon compounds with sodalite structure by the first-principles molecular dynamics (FPMD) method. First, we estimated the temperature TuT_{\mathrm{u}} at which the system becomes unstable. Then, we estimated TmT_{\mathrm{m}} using the known relationship between TuT_{\mathrm{u}} and the melting point TmT_{\mathrm{m}}. These results are shown as phase diagrams for compounds on the PP-TT plane, together with the superconducting transition temperature TcT_{\mathrm{c}}.

We also calculated the formation energies of compounds from the ground state energies of the elements. To consider the effect of finite temperature, we obtained the free energies of NaC6_6 and FC6_6 as typical cases. From these results, we estimated the Gibbs free energies and discussed whether the formation of these compounds is possible.

Model and Methods

Calculations were performed using Quantum ESPRESSO (QE), which is an integrated software of open-source computer codes for electronic-structure calculations.27) In the FPMD calculations, we used the N,P,T=const.N,P,T=\mathrm{const.} ensemble system, where the system is a supercell made up of multiple unit cells, as shown in Figure 1. To reduce finite size effects, the direction vector that determines the shape of the supercell is set to change at each time step of the simulation. This variable cell method includes system fluctuations better than the method using fixed direction vectors.

To estimate TmT_{\mathrm{m}}, we used a combination of the FPMD method and the ZZ-method,28) which is used to determine melting points.29–34) Our method uses FPMD simulations to determine the critical temperature TuT_{\mathrm{u}} at which the crystal becomes thermally unstable, and then calculates TmT_{\mathrm{m}} from the relationship Tm≃Tu/1.23T_{\mathrm{m}} \simeq T_{\mathrm{u}}/1.23.28) Details of the ZZ-method are discussed in Appendix § 6.

When the temperature of a system exceeds TuT_{\mathrm{u}}, the motion of atoms in the system changes significantly in a short period of time. Therefore, it is easy to find TuT_{\mathrm{u}} in the simulation. To obtain TuT_{\mathrm{u}}, we calculated the Lindemann parameter defined as Lp=⟨ri2⟩/dL_{\mathrm{p}} = \sqrt{\langle r_i^2\rangle}/d.35–37) Here, rir_i is the displacement of each atom from its equilibrium position at T=0T = 0, and dd is the distance between the nearest-neighbor atoms in the crystal at T=0T = 0. The thermal average is taken over all atoms at each time step in the simulation.

When the value of LpL_{\mathrm{p}} exceeds a certain critical value, the system becomes unstable rapidly.38) Thus, we calculated LpL_{\mathrm{p}} as a function of temperature and found TuT_{\mathrm{u}} from the point where it changes significantly.

At first, we considered the diamond system to test the above method. Figure 2 shows the Lindemann parameter Lp(t)L_{\mathrm{p}}(t) as a function of simulation time tt for a diamond system with N=64N = 64 at P=0P = 0 GPa, where NN is the number of atoms in the system. Here, solid circles and empty circles represent the results for T=6000T = 6000 and 7000 K, respectively. It indicates that the values of Lp(t)L_{\mathrm{p}}(t) at T=6000T = 6000 K are almost constant with tt, and those at T=7000T = 7000 K increase with tt and seem to diverge for t≳170t \gtrsim170 fs. This result suggests that TuT_{\mathrm{u}} is between 6000 and 7000 K.

Figure 2. Lindemann parameter LpL_{\mathrm{p}} as a function of simulation time tt, where solid circles and empty circles represent the results for T=6000T = 6000 and 7000 K, respectively. L‾p\overline{L}_{\mathrm{p}} represents the average of the last four points of LpL_{\mathrm{p}}.

To analyze TuT_{\mathrm{u}} in more detail, we defined L‾p\overline{L}_{\mathrm{p}} as the average value of Lp(t)L_{\mathrm{p}}(t) at the last four points (t=180,190,200t = 180,190,200, and 210 fs), as shown in Figure 2. As shown in the figure, it is expected that the system will reach a certain degree of equilibrium after about 150 fs. Therefore, we used L‾p\overline{L}_{\mathrm{p}} as the approximate equilibrium value of Lp(t)L_{\mathrm{p}}(t).

Figure 3 shows L‾p\overline{L}_{\mathrm{p}} as a function of TT for the systems with N=8N = 8 (solid circles) and 64 (empty circles). The value of L‾p\overline{L}_{\mathrm{p}} increases with temperature and appears to diverge above L‾p∼0.3\overline{L}_{\mathrm{p}} \sim0.3.38) Hereafter, we assumed the temperature corresponding to L‾p=0.3\overline{L}_{\mathrm{p}} = 0.3 to be TuT_{\mathrm{u}}, where the point of L‾p=0.3\overline{L}_{\mathrm{p}} = 0.3 is determined by interpolation. As shown in the figure, we found that Tu≃4500T_{\mathrm{u}} \simeq4500 K for

Figure 3. L‾p\overline{L}_{\mathrm{p}} as a function of TT, where solid circles and empty circles represent the results for the systems with N=8N = 8 and 64, respectively. TuT_{\mathrm{u}} is determined by the condition L‾p=0.3\overline{L}_{\mathrm{p}} = 0.3.

dependence seems not to be so large. However, the size effect of TmT_{\mathrm{m}} is not simple. At a certain pressure, TmT_{\mathrm{m}} is high when the size is large. However, the opposite also appears depending on the pressure. The use of variable supercells is speculated to be the cause of the complex size effects.

Figure 5(b) shows the TmT_{\mathrm{m}} of C10_{10} as a function of pressure PP, where solid circles and solid squares represent the results for N=10N = 10 and 40, respectively. TmT_{\mathrm{m}} is about 3500 K at P=0P = 0 and increases slightly with pressure up to about 100 GPa, but decreases above that. This suggests that C10_{10} would be able to exist stably at room temperature and pressure, although it has not yet been experimentally discovered.

N = 8 and 6800 K for N=64N = 64. This leads to Tm≃3700T_{\mathrm{m}} \simeq3700 K for N=8N = 8 and 5500 K for N=64N = 64. When the system is small, fluctuations of the system are large and TuT_{\mathrm{u}} tends to be low.

Figure 4 shows the TmT_{\mathrm{m}} of diamond as a function of pressure PP along with previous theoretical39,40) and experimental41,42) results. Here, the solid and empty circles represent the results for N=8N = 8 and 64, respectively. In the FPMD calculation, we used mesh numbers k=6k = 6 and 2 in wavenumber space for the systems with N=8N = 8 and 64, respectively. The obtained results are almost in agreement with those of previous works.

The figure also shows that the size effect for TmT_{\mathrm{m}} is greatest when P=0P = 0 and decreases as the pressure increases. This is probably due to the decrease in the fluctuations of the system with increasing pressure. The inset shows the kk-dependence of TmT_{\mathrm{m}} for the systems with N=8N = 8 and 64 at P=0P = 0. This indicates that the kk-dependence is rather small compared with the size effect in both systems.

Results and Discussion

TmT_{\mathrm{m}} as a function of PP for C6_6 and C10_{10}

We applied the above method to pure carbon systems C6_6 and C10_{10} as reference systems for XC6_6 and XC10_{10}. Figure 5(a) shows the TmT_{\mathrm{m}} of C6_6 as a function of pressure PP, where empty circles, empty squares, and solid triangles represent the results for N=12N = 12, 48, and 96, respectively. The TmT_{\mathrm{m}} of C6_6 is at least 2000 K at P=0P = 0 and increases slightly with pressure. Although its TmT_{\mathrm{m}} is lower than that of diamond, it is stable at room temperature and pressure.

The figure indicates that TmT_{\mathrm{m}} for each size is almost in agreement with each other at all pressures, and the size dependence seems not to be so large. However, the size effect of TmT_{\mathrm{m}} is not simple. At a certain pressure, TmT_{\mathrm{m}} is high when the size is large. However, the opposite also appears depending on the pressure. The use of variable supercells is speculated to be the cause of the complex size effects.

Figure 5. (a) TmT_{\mathrm{m}} of C6_6 as a function of PP, where empty circles, empty squares, and solid triangles represent TmT_{\mathrm{m}} for the N=12N = 12, 48, and 96 systems, respectively. (b) TmT_{\mathrm{m}} of C10_{10} as a function of PP, where solid circles and solid squares represent TmT_{\mathrm{m}} for the N=10N = 10 and 40 systems, respectively.

TmT_{\mathrm{m}} and TcT_{\mathrm{c}} as functions of PP for XC6_6 and XC10_{10}

We considered compounds of XC6_6 and XC10_{10}, combining carbon and X-atoms. Mainly, we presented the results for compounds NaC6_6, FC6_6, FC10_{10}, and ClC10_{10} as typical cases.

Figure 6(a) shows the TmT_{\mathrm{m}} of NaC6_6 as a function of pressure PP, where solid circles and solid squares represent the results for N=14N = 14 and 56, respectively. In addition to TmT_{\mathrm{m}}, the superconducting transition temperature TcT_{\mathrm{c}}43) as a function of pressure PP is also shown as empty circles, where TcT_{\mathrm{c}} is multiplied by 10 for ease of viewing. Here, TcT_{\mathrm{c}} is calculated by the tetrahedral method,27) which allows us to avoid the ambiguity in the choice of Gaussian broadening that exists in the interpolation method.

We found that the TmT_{\mathrm{m}} of NaC6_6 is zero and unstable at P=0P = 0. However, TmT_{\mathrm{m}} is about 2,000 K at P=30P = 30 GPa and increases with pressure, reaching about 5,000 K at P=200P = 200 GPa. This result seems to be consistent with those of previous works showing that NaC6_6 is dynamically unstable for P≲30P \lesssim30 GPa.22,23) On the other hand, TcT_{\mathrm{c}} was calculated as 149, 123, and 113 K at P=20P = 20, 30, and 50 GPa, respectively. As the pressure decreases, the melting point decreases and the system becomes unstable. However, TcT_{\mathrm{c}} appears to increase with decreasing pressure for P≲50P \lesssim50 GPa.

Figure 6(b) shows the TmT_{\mathrm{m}} of FC6_6 as a function of pressure PP, where solid circles and solid squares represent the results for N=14N = 14 and 56, respectively. In this case, TmT_{\mathrm{m}} is ∼2000\sim2000 at P=0P = 0 and increases with pressure and exceeds 6000 K at P=200P = 200 GPa. Unlike NaC6_6, FC6_6 is stable even at room temperature and pressure. TcT_{\mathrm{c}} is 83 K at P=0P = 0 and increases slightly with pressure. In a previous study, FC6_6 was shown to be dynamically unstable by an interpolation method.[22] However, this is not the case in the tetrahedron method and we can obtain TcT_{\mathrm{c}} for FC6_6.

Figure 7(a) shows the TmT_{\mathrm{m}} of FC10_{10} as a function of pressure PP, where solid circles and solid squares represent the results for N=11N = 11 and 44, respectively. Similar to Figure 6, TcT_{\mathrm{c}} as a function of pressure PP is also shown as empty circles. Although the size dependence will be somewhat large in this case, TmT_{\mathrm{m}} may be about 2000 K at P=0P = 0. The figure also shows that TmT_{\mathrm{m}} increases with pressure and reaches a maximum at about P=150P = 150 GPa. On the other hand, TcT_{\mathrm{c}} is about 110 K at P=0P = 0 and decreases with increasing pressure. This material is also expected to exhibit superconductivity under normal pressure, similar to FC6_6.

Figure 6. (a) TmT_{\mathrm{m}} and TcT_{\mathrm{c}} of NaC6_6 as functions of PP, where solid circles and solid squares represent TmT_{\mathrm{m}} for the N=14N = 14 and 56 systems, respectively. Here, empty circles represent Tc×10T_{\mathrm{c}} \times10. (b) TmT_{\mathrm{m}} and TcT_{\mathrm{c}} of FC6_6 as functions of PP, where solid circles and solid squares represent TmT_{\mathrm{m}} for the N=14N = 14 and 56 systems, respectively.

Figure 7(b) shows the TmT_{\mathrm{m}} of ClC10_{10} as a function of pressure PP, where solid circles and solid squares represent the results for N=11N = 11 and 44, respectively. Empty circles represent Tc×10T_{\mathrm{c}} \times10. In this system, the behaviors of TmT_{\mathrm{m}} and TcT_{\mathrm{c}} are similar to those of FC10_{10}.

For detailed data on NaC6_6, FC6_6, FC10_{10}, and ClC10_{10}, Table I gives TmT_{\mathrm{m}}, TcT_{\mathrm{c}}, λ\lambda, and ωlog⁡\omega_{\log}, where the numbers in parentheses are pressure in GPa. Here, TmT_{\mathrm{m}} values are for the N=14N = 14 system for XC6_6 and for the N=11N = 11 system for XC10_{10}.

Results for other compounds at P=0P = 0 are shown in Table II. For TmT_{\mathrm{m}}, we present the N=14N = 14 system for XC6_6, the N=11N = 11 system for XC10_{10}, and the N=13N = 13 system for XC12_{12}. The table shows that XC6_6 has a relatively high TcT_{\mathrm{c}}, but TmT_{\mathrm{m}} is not high. On the other hand, XC10_{10} and XC12_{12} have high TmT_{\mathrm{m}}, but TcT_{\mathrm{c}} is low. These results suggest that TcT_{\mathrm{c}} decreases as the proportion of XX atoms in the system decreases. This can be interpreted as a reduction in the number of effective carriers that the XX atom brings to the system.[22] This leads to a decrease in TcT_{\mathrm{c}}. Conversely, an increase in the number of effective carriers reduces the stability of the system and decreases TmT_{\mathrm{m}}.

Figure 7. (a) TmT_{\mathrm{m}} and TcT_{\mathrm{c}} of FC10_{10} as functions of PP, where solid circles and solid squares represent TmT_{\mathrm{m}} for the N=11N = 11 and 44 systems, respectively. Here, empty circles represent Tc×10T_{\mathrm{c}} \times10. (b) TmT_{\mathrm{m}} and TcT_{\mathrm{c}} of ClC10_{10} as functions of PP, where solid circles and solid squares represent TmT_{\mathrm{m}} for the N=11N = 11 and 44 systems, respectively.

NaC6_6 (30)NaC6_6 (200)FC6_6 (0)FC6_6 (200)
PP [GPa]
TmT_{\mathrm{m}} [10310^3K]1.35.12.26.2
TcT_{\mathrm{c}} [K]1241148397
λ\lambda3.811.761.371.45
ωlog⁡\omega_{\log} [K]407744731830
FC10_{10} (0)FC10_{10} (200)ClC10_{10} (0)ClC10_{10} (200)
PP [GPa]
TmT_{\mathrm{m}} [10310^3K]1.54.21.93.8
TcT_{\mathrm{c}} [K]1066011631
λ\lambda1.781.132.070.67
ωlog⁡\omega_{\log} [K]697653662997

Table 1. TmT_{\mathrm{m}}, TcT_{\mathrm{c}}, λ\lambda, and ωlog⁡\omega_{\log} for NaC6_6, FC6_6, FC10_{10}, and ClC10_{10}.

HC6_6LiC6_6BC6_6OC6_6ClC6_6HC10_{10}
TmT_{\mathrm{m}} [10310^3K]0.300.110.170.510.182.2
TcT_{\mathrm{c}} [K]435720527314
λ\lambda0.753.472.371.231.140.53
ωlog⁡\omega_{\log} [K]889209113532825902
OC10_{10}NaC10_{10}BC12_{12}NaC12_{12}FC12_{12}ClC12_{12}
TmT_{\mathrm{m}} [10310^3K]1.51.30.571.23.01.7
TcT_{\mathrm{c}} [K]32136321721
λ\lambda0.950.570.530.730.580.60
ωlog⁡\omega_{\log} [K]463670336765955951

Table 2. TmT_{\mathrm{m}}, TcT_{\mathrm{c}}, λ\lambda, and ωlog⁡\omega_{\log} for other compounds at P=0P = 0.

Formation enthalpies at P=0P = 0 and T=0T = 0

From the perspective of actually synthesizing these compounds, not only TmT_{\mathrm{m}} but also the formation enthalpy may be important. We define the formation enthalpy ΔH\Delta H such as ΔH=HNaC6−(HNa+6Hgra)/7\Delta H = H_{\mathrm{NaC}_6} - (H_{\mathrm{Na}} + 6H_{\mathrm{gra}})/7 for NaC6_6, and similarly for other compounds. Here, HNaC6H_{\mathrm{NaC}_6}, HNaH_{\mathrm{Na}}, and HgraH_{\mathrm{gra}} are the enthalpies at T=0T = 0 per atom of NaC6_6, sodium in a bcc structure, and graphite, respectively. Note that the definition of ΔH\Delta H given here is insufficient to estimate an exact value of the formation enthalpy. To obtain this, it is necessary to compare the energies of all compounds that contain the elements in the target compound in any ratio. Since this is very difficult in practice, we estimated ΔH\Delta H as information that is useful for synthesizing compounds.

Figure 8 shows the ΔH\Delta H of XC6_6, XC10_{10}, and XC12_{12} at P=0P=0, where the horizontal axis is the fraction of carbon contained in compounds. Here, ΔH\Delta H of XC6XC_{6} is represented by empty circles, that of XC10XC_{10} by empty squares, and that of XC12XC_{12} by empty triangles. In addition, there is the ΔH\Delta H of X-atom-doped diamond (denoted as X-diamond), in which one carbon atom in a diamond supercell made up of eight atoms is replaced by an X-atom. These are represented by solid diamonds in the figure. The solid triangle stands for the result of the graphite intercalation compound that has sodium between the layers (Na-GIC).[18] Each point of ΔH\Delta H is accompanied by a value of TmT_{\mathrm{m}} in units of 10310^{3} K. The TmT_{\mathrm{m}} values for XC6XC_{6}, XC10XC_{10}, and XC12XC_{12} correspond to those for the systems with N=14N=14, 11, and 13, respectively.

For reference, the formation enthalpies per atom of graphite, pure diamond, C6C_{6}, C10C_{10}, and C60C_{60} have also been added, where these are represented by solid circles. Here, TmET_{\mathrm{m}}^{\mathrm{E}} stands for the experimental melting temperature, Tm∗T_{\mathrm{m}}^{*} is the melting temperature at the triple point near 12 GPa, and TdET_{\mathrm{d}}^{\mathrm{E}} is the decomposition temperature observed in the experiment.[44] The results for XC6XC_{6} and XC12XC_{12} almost agree with those of the others.[21]

Comparing the difference in the value of ΔH\Delta H with those in other studies,[45] we found that it is about 0.008 eV for diamond and about 0.012 eV for C60C_{60}. For compounds NaCl, B4CB_{4}C, and Be2CBe_{2}C,[45] this difference is about 0.16, 0.005, and 0.11 eV, respectively, although not shown in the figure. For XC6XC_{6} and XC12XC_{12}, the discrepancy between our results and others’ is roughly 0.06 eV.[21] In the case of compounds, the discrepancy of ΔH\Delta H does not seem to be very small.

The figure shows that the ΔH\Delta H of C60C_{60} is about 0.4 eV and close to that of C6C_{6}. C60C_{60} is a material well known as fullerene and can now be synthesized. Therefore, it is expected that the synthesis of C6C_{6} is possible.

Unfortunately, XC6XC_{6} has low TmT_{\mathrm{m}}, except for FC6FC_{6} and OC6OC_{6}, and may be unstable at room temperature. On the other hand, XC10XC_{10} and XC12XC_{12} have higher TmT_{\mathrm{m}} than

Figure 8. Formation enthalpies ΔH\Delta H as a function of the fraction of carbon in compounds with TmT_{\mathrm{m}}, where the unit of TmT_{\mathrm{m}} is 10310^{3} K. Here, X-diamond means that one atom is replaced by an X-atom in the 8-site diamond cell, and Na-GIC represents a graphite intercalation compound with sodium between the layers.

XC6XC_{6}, even though their ΔH\Delta H values are similar to each other. The value of ΔH\Delta H for B-diamond is about 0.25 eV, which is small, where its TcT_{\mathrm{c}} is calculated as 32 K. This result seems to be related to the fact that boron-doped diamond was found to be a superconductor in experiments.[12, 14, 15] On the other hand, ΔH\Delta H values of F-, Cl-, and Na-diamonds are around 1 eV or larger. This might correspond to the fact that superconducting diamonds doped with these elements have not been found.

The Na-GIC (sodium intercalated graphite)[18] considered here has the same crystal structure as the experimentally discovered Li-GIC (lithium intercalated graphite).[45] This type of Na-GIC has not been found experimentally. However, it is an interesting isomer of NaC6C_{6} with the sodalite structure, since finite TcT_{\mathrm{c}} is found theoretically.[18]

Formation enthalpies as a function of PP at T=0T=0

The pressure dependence of formation enthalpies is also interesting. Figure 9 shows ΔH\Delta H as a function of PP for C6C_{6}, C10C_{10}, FC6FC_{6}, NaC6C_{6}, FC10FC_{10}, NaC10C_{10}, and Na-GIC. Here, the ΔH\Delta H values are calculated relative to diamond, except when P=0P=0. This is because the enthalpy of diamond is lower than that of graphite at pressures above about 2 GPa. Furthermore, fluorine and sodium undergo structural phase transitions when the pressure is increased,[46, 47] so the ΔH\Delta H of the compounds must be calculated using the enthalpies corresponding to their structures.

Figure 9. ΔH\Delta H as a function of PP for C6C_{6}, C10C_{10}, FC6FC_{6}, NaC6C_{6}, FC10FC_{10}, NaC10C_{10}, and Na-GIC.

For fluorine, the C2/cC2/c structure is used for P<20P<20 GPa, and the CmcaCmca structure is used otherwise.[46] For sodium, the bcc, fcc, C2/mC2/m, and P63/mmcP6_{3}/mmc structures are used for P<60P<60, P=100P=100, 200, and 300 GPa, respectively.[47]

Figure 9 indicates that ΔH\Delta H increases with PP for C6C_{6}, C10C_{10}, FC10FC_{10}, NaC10C_{10}, and Na-GIC. These substances have relatively low density, so their enthalpy increases under high pressure. In fact, the densities of diamond, C6\mathrm{C}_{6}, and C10\mathrm{C}_{10} are 4.64, 3.97, and 3.79 g/cm3^{3}, respectively, at P=200P = 200 GPa. Furthermore, those of NaC6_{6} and Na-GIC are 4.18 and 3.54 g/cm3^{3}, respectively, at P=100P = 100 GPa. Here, the melting point of Na-GIC is 0 K at P∼110P \sim110 GPa, and it may be thermally unstable at pressures above this.

The figure also indicates that the ΔH\Delta H of NaC6_{6} decreases with increasing PP, while the others generally increase with P=0P = 0 except FC6_{6}. The ΔH\Delta H of FC6_{6} seems to be almost independent of PP. We have confirmed that the ΔH\Delta H values for NaC6_{6} and FC6_{6} are almost the same at P=300P = 300 and 400 GPa. Therefore, it seems that these ΔH\Delta H values will not decrease even if the pressure is increased further.

In contrast to NaC6_{6}, the ΔH\Delta H of NaC10_{10} increases with pressure. This can be interpreted as the crystal structure of NaC10_{10} being more sparse than that of NaC6_{6}. This may also apply to the relationship between FC6_{6} and FC10_{10}.

Formation free energies for NaC6_{6} and FC6_{6}

As shown in Figure 9, the ΔH\Delta H of NaC6_{6} decreases with increasing pressure and becomes about 0.29 eV at P=300P = 300 GPa. In this case, the effects of finite temperature, such as lattice vibrations, may not be negligible near the melting point. To clarify the effects of phonons, we calculated the Helmholtz free energy F(T)F(T) by harmonic approximation. Comparing F(T)F(T) with experimental results,[48] we introduced a phenomenologically corrected free energy F~(T)\widetilde{F}(T). It reproduces the experimental results better than F(T)F(T) and allows us to estimate the free energy of the liquid state. By adding F~(T)\widetilde{F}(T) to the enthalpy at T=0T = 0, we can approximately obtain the Gibbs free energy at finite temperatures. Detailed calculation methods are given in Appendix B.

Figure 10(a) shows F~(T)\widetilde{F}(T) as a function of TT for diamond, NaC6_{6}, and sodium at P=300P = 300 GPa. We also present the formation free energy, ΔF~(T)\Delta\widetilde{F}(T), which is defined as

ΔF~(T)=F~NaC6(T)−F~Na(T)+6F~dia(T)7.\Delta\widetilde{F}(T) = \widetilde{F}_{\mathrm{NaC}_{6}}(T) - \frac{\widetilde{F}_{\mathrm{Na}}(T) + 6\widetilde{F}_{\mathrm{dia}}(T)}{7}.

Here, F~NaC6(T)\widetilde{F}_{\mathrm{NaC}_{6}}(T), F~Na(T)\widetilde{F}_{\mathrm{Na}}(T), and F~dia(T)\widetilde{F}_{\mathrm{dia}}(T) are the F~(T)\widetilde{F}(T) per atom for NaC6_{6}, sodium, and diamond, respectively. These values decrease with increasing temperature owing to entropy. As shown in the figure, ΔF~(T)\Delta\widetilde{F}(T) is about −0.34-0.34 eV at T=6550T = 6550 K, where this temperature corresponds to the TmT_{\mathrm{m}} of NaC6_{6}, and the TmT_{\mathrm{m}} of diamond is about 9300 K under the same pressure. Here, to obtain TmT_{\mathrm{m}}, we used the N=14N = 14 system for NaC6_{6} and the N=8N = 8 system for diamond.

Recalling that ΔH≃0.29\Delta H \simeq0.29 eV, the value of ΔH+ΔF~(T)\Delta H + \Delta\widetilde{F}(T) becomes about −0.05-0.05 eV. Since the value is negative, the formation of NaC6_{6} under this condition can be expected.[49] The result also suggests that high pressure and high temperature may favor the synthesis of NaC6_{6}.

By applying a similar analysis as described above to FC6_{6}, ΔF~(T)\Delta\widetilde{F}(T) is found to be about −0.36-0.36 eV at T=6660T = 6660 K, as shown in Figure 10(b). Here, this temperature corresponds to TmT_{\mathrm{m}} at P=300P = 300 GPa. We assumed that csolid=1.06c_{\mathrm{solid}}=1.06 for fluorine, FC6_{6}, and diamond. For the liquid state of fluorine, cliquidc_{\mathrm{liquid}} was assumed to be 1.2. Since the ΔH\Delta H of FC6_{6} is about 0.43 eV at P=300P = 300 GPa, we obtained ΔH+ΔF~(T)≃0.07\Delta H + \Delta\widetilde{F}(T) \simeq0.07 eV. This value is small but positive. Thus, the synthesis of FC6_{6} may be more difficult than that of NaC6_{6}.

Figure 10. Free energy as a function of temperature for (a) Na, NaC6_{6}, and diamond, and (b) F, FC6_{6}, and diamond at P=300P = 300 GPa. Here, solid circles represent TmT_{\mathrm{m}}. The dotted lines show the formation free energy ΔF~(T)\Delta\widetilde{F}(T), which was determined using a similar definition as for ΔH\Delta H.

In the above discussion, we have not addressed the contribution of electrons at finite temperatures. To consider it, we estimated the free energy of electrons. At P=300P = 300 GPa, NaC6_{6} is metallic and the value of DOS at the Fermi level, DFD_{\mathrm{F}}, is 0.20 states/eV/atom. Using the free electron model, we calculated the contribution of free energy to be ∼−π26(kBT)2DF\sim-\frac{\pi^{2}}{6}(k_{\mathrm{B}}T)^{2}D_{\mathrm{F}} at low temperatures. The value is about −0.11-0.11 eV at T=6550T = 6550 K. If liquid sodium is in the metallic state, the value of the free energy is expected to be of the same order as NaC6_{6}. On the other hand, we found that diamond has an energy gap of 5.6 eV at P=300P = 300 GPa, and the electron contribution to the free energy is negligible. In this case, the ΔF~(T)\Delta\widetilde{F}(T) of NaC6_{6} may decrease by about 0.09 eV.

However, we used F~(T)\widetilde{F}(T), which closely reproduces the experimental results. Since the experimental results should include the effect of electrons at finite temperatures, F~(T)\widetilde{F}(T) should be considered to include the effect of electrons, even if it is not complete. If so, adding the electron contribution to F~(T)\widetilde{F}(T) could result in double counting. Similarly, the effects of volume changes at finite temperatures may already be included. In any case, the contribution of these effects may be small and would not significantly change the results.

Summary

By combining the FPMD method and the Z-method, we investigated the TmT_{\mathrm{m}} of carbon compounds XC6_{6}, XC10X\mathrm{C}_{10}, and XC12X\mathrm{C}_{12} with the sodalite structure. To verify the validity of the method, we applied it to the diamond system and compared the results with those of other theoretical and experimental works. We also discussed the assumptions adopted in the ZZ-method by comparing them with experimental results. In addition to the melting points, we obtained the superconducting transition temperature TcT_{\mathrm{c}} by the optimized tetrahedron method in QE.

These results suggest that FC6\mathrm{FC}_{6}, FC10\mathrm{FC}_{10}, and ClC10\mathrm{ClC}_{10} can exist stably at sufficiently high temperatures (T≳1000T \gtrsim1000 K) even under normal pressure, and the TcT_{\mathrm{c}} values of these compounds reach nearly 100 K. Although the TmT_{\mathrm{m}} of NaC6\mathrm{NaC}_{6} drops rapidly at pressures below 30 GPa, TcT_{\mathrm{c}} is over 120 K for P≳30P \gtrsim30 GPa. We also showed results for several other compounds that have relatively low TcT_{\mathrm{c}} but are stable at ambient pressure. If these compounds can actually be synthesized, it would be possible to observe high-temperature superconductivity mediated by phonons.

To examine the feasibility of these compounds, we calculated the formation enthalpies ΔH\Delta H from the ground state energies of the elements. Unfortunately, all of these values for compounds were positive and above 0.5 eV at P=0P = 0. This suggests that the synthesis of these compounds is difficult. We also examined the pressure dependence of ΔH\Delta H for C6\mathrm{C}_{6}, C10\mathrm{C}_{10}, and several representative compounds. Since C6\mathrm{C}_{6} and C10\mathrm{C}_{10} have lower densities than diamond, ΔH\Delta H increases with pressure. However, the ΔH\Delta H of NaC6\mathrm{NaC}_{6} decreases with increasing pressure, from about 0.94 eV at P=0P = 0 to 0.29 eV at P=300P = 300 GPa.

In this case, the effect of finite temperature may make the Gibbs formation free energy negative. To clarify this, we introduced a phenomenologically corrected formation free energy ΔF~(T)\widetilde{\Delta F}(T). Applying this to NaC6\mathrm{NaC}_{6}, we found the value of ΔF~(T)\widetilde{\Delta F}(T) to be approximately −0.36-0.36 eV at T=6600T = 6600 K and P=300P = 300 GPa. This cancels out the positive ΔH\Delta H, and ΔF~(T)+ΔH\widetilde{\Delta F}(T) + \Delta H, which corresponds to the Gibbs formation free energy, becomes negative.

This result suggests that sufficiently high temperatures and pressures may be effective for the formation of NaC6\mathrm{NaC}_{6}. Applying a similar analysis to FC6\mathrm{FC}_{6}, we found that the Gibbs formation free energy is small but positive under conditions similar to those for NaC6\mathrm{NaC}_{6}. This result indicates that the synthesis of FC6\mathrm{FC}_{6} may be more difficult than that of NaC6\mathrm{NaC}_{6}.

Acknowledgments

This work was supported by JSPS KAKENHI Grant Number JP19K03716.

Simulation Details

All calculations of FPMD and electronic structures were carried out by using the density functional theory of Perdew-Burke-Ernzerhof (PBE) generalized gradient approximation and the projector augmented wave pseudopotential as implemented in QE package.27 The cutoff energies for the wave functions and charge densities were set to the suggested minimum values given in the pseudopotential files.27 For example, for a carbon atom, the two values are 40 and 323 Ry, respectively. Convergence

thresholds were set as 10−410^{-4} a.u. in total energy and 10−310^{-3} a.u. in force.

The k\mathbf{k} of 6×6×66 \times6 \times6 mesh in the first Brillouin zone was used for diamond systems with N=8N = 8, XC6\mathrm{XC}_{6} systems with N=14N = 14, and XC10\mathrm{XC}_{10} systems with N=11N = 11. For other systems with NN less than 30, a 4×4×44 \times4 \times4 mesh was used. For all systems with NN more than 30, a 2×2×22 \times2 \times2 mesh was used. In the calculation of FPMD, we used the ‘vc-md’ option27 in QE, which uses the Beeman algorithm to integrate Newton’s equation. The time step is chosen to be one femtosecond (fs).

A new simulation is started with the equilibrium atomic positions at T=0T = 0, where these atomic positions are obtained by structural optimization under the target pressure. The initial velocity distribution is a Boltzmann distribution determined by the target temperature. The instantaneous temperature TiT_{\mathrm{i}} is calculated at the end of every atomic move, and the temperature of the system is rescaled to the target temperature at each step. The instantaneous pressure PiP_{\mathrm{i}} of the system changes with each step, but is adjusted to the target pressure by modifying the cell volume. The root mean square of the displacement of the atomic positions from their initial positions at t=0t = 0 corresponds to the Lindemann parameter LpL_{\mathrm{p}}. Here, LpL_{\mathrm{p}}, PiP_{\mathrm{i}}, and TiT_{\mathrm{i}} are recorded every 10 steps during the simulation.

Figure A · 1. (a) Instantaneous pressure PiP_{\mathrm{i}} as a function of tt, (b) instantaneous temperature TiT_{\mathrm{i}} as a function of tt, and (c) Lindemann parameter LpL_{\mathrm{p}} as a function of tt, where a straight line represents the average value of LpL_{\mathrm{p}}. The system used is NaC6\mathrm{NaC}_{6} (N=14N = 14), and the target pressure and target temperature are set to 100 GPa and 3500 K, respectively.

Figure A · 1 shows the simulation results for (a) Pi(a)P_{\mathrm{i}}(a), (b) TiT_{\mathrm{i}}, and (c) LpL_{\mathrm{p}} for NaC6\mathrm{NaC}_{6} (N=14N = 14) as a function of tt up to 2000 fs, where the target pressure and target temperature are set to 100 GPa and 3500 K, respectively. In Fig. A · 1(c), the straight line represents the average value of LpL_{\mathrm{p}}. Owing to the small system size, PiP_{\mathrm{i}} and TiT_{\mathrm{i}} show large fluctuations. However, these values are around the target pressure and the target temperature at all times, and no systematic deviations are observed within this calculation. The value of LpL_{\mathrm{p}} also appears to fluctuate around the average value.

Figure A·2. (a) Instantaneous pressure PiP_i as a function of tt, (b) instantaneous temperature TiT_i as a function of tt, and (c) the Lindemann parameter LpL_p as a function of tt, where a straight line represents the average value of LpL_p. Here, the system is FC10_{10} with N=11N = 11. The target pressure and target temperature are set to 0 GPa and 1000 K, respectively.

Figure A·3. (a) LpL_p as a function of tt for NaC6_6 at the target temperature T=5000T = 5000 K (solid circles) and T=4200T = 4200 K (empty circles) with N=14N = 14 at P=100P = 100 GPa. (b) LpL_p as a function of tt at T=4000T = 4000 K (solid circles) and T=3100T = 3100 K (empty circles) for FC10_{10} with N=44N = 44 at P=0P = 0 GPa. (c) L‾p\overline{L}_p as a function of TT for NaC6_6 with N=14N = 14 at P=100P = 100 GPa. (d) L‾p\overline{L}_p as a function of TT for FC10_{10} with N=44N = 44 at P=0P = 0 GPa.

As shown in Figure 6 in the main text, TmT_m and TuT_u were estimated to be approximately 3800 and 4600 K, respectively. The peaks of TiT_i in Fig. A·1(b) appear to reach ∼5000\sim5000 K and exceed the above temperatures. However, the system maintains its crystal structure. It is interesting that very large instantaneous fluctuations do not necessarily lead to melting.

Figure A·2 shows the simulation results for (a) PiP_i, (b) TiT_i, and (c) LpL_p for FC10_{10} (N=11N = 11) as a function of tt up to 500 fs, where the target pressure and target temperature are set to 0 GPa and 1000 K, respectively. In Fig. A·2(c), the straight line represents the average value up to 2000 fs. We confirmed that these values fluctuate around the average value up to 2000 fs.

Similar to the case of NaC6_6, the fluctuations are large. As shown in the figures, it is difficult to distinguish between fluctuations in the equilibrium state and relaxation processes from the initial state at t=0t = 0. Here, we considered the relaxation process by focusing on the behavior of LpL_p in Fig. A·2(c). In our simulation, the value of LpL_p starts from zero at t=0t = 0. The value appears to approach the equilibrium value after ≃100\simeq100 fs. Judging from this behavior, the relaxation time may be considered to be about 100 fs.

If the crystal structure of a system becomes unstable at high temperatures, the value of LpL_p is expected to become very large. Figures A·3(a) and A·3(b) show LpL_p as a function of tt for (a) NaC6_6 (N=14N = 14, target pressure P=100P = 100 GPa) and (b) FC10_{10} (N=44N = 44, target pressure P=0P = 0 GPa). For the case of NaC6_6, LpL_p increases sharply with tt for t≳150t \gtrsim150 K at T=5000T = 5000 K, but it seems to be almost constant at T=4200T = 4200 K. For the case of FC10_{10}, similar to NaC6_6, LpL_p is nearly constant at T=3100T = 3100 K, but seems to increase rapidly with tt at T=4000T = 4000 K. These results suggest that the system becomes unstable

at T∼4600T \sim4600 K for NaC6_6, and at T∼3500T \sim3500 K for FC10_{10}. To calculate TuT_u more systematically, we used L‾p\overline{L}_p, which is defined as the average value of LpL_p at t=180t = 180, 190, 200, and 210 fs. Figures A·3(c) and A·3(d) show L‾p\overline{L}_p as a function of the target temperature TT for (c) NaC6_6 and (d) FC10_{10}. The point where the temperature dependence of L‾p\overline{L}_p changes suddenly may correspond to TuT_u. This behavior of L‾p\overline{L}_p is the same as that observed in the diamond system discussed in the main text. For convenience, we considered that the temperature at which L‾p=0.3\overline{L}_p = 0.3 approximately gives TuT_u. In this simulation, the target temperature is varied in intervals between approximately 1/51/5 and 1/101/10 of TuT_u, and the temperature at which L‾p=0.3\overline{L}_p = 0.3 was found by interpolation.

Although the step width of the target temperature is coarse and the system used is not so large, it may be successful in giving an approximate value of TuT_u. In fact, the values of TmT_m estimated from our TuT_u roughly reproduced the experimental results, as shown in Figure 4 in the main text and Table B·1 in Appendix B.

Figure 4. Melting points of diamond as a function of pressure PP, where empty circles and empty squares represent TmT_{\mathrm{m}} for the N=8N = 8 and 64 systems, respectively. Solid squares, solid circles, and solid triangles represent the results of Liang et al.,42) Correa et al.,40) and Wang et al., respectively.39) The light and broken lines show the result of Bundy et al.,41) where the broken line indicates the phase boundary between graphite and diamond. The inset shows the kk-dependence of TmT_{\mathrm{m}} at P=0P = 0 GPa, where kk is the mesh number in wavenumber space.

Figure A·4 shows the kk-dependence of TmT_m for the systems of NaC6_6 for N=14N = 14 at P=100P = 100 GPa (empty circles), FC10_{10} for N=44N = 44 at P=200P = 200 GPa (solid squares), ClC10_{10} for N=44N = 44 at P=0P = 0 GPa (empty triangles), and FC10_{10} for N=11N = 11 at P=0P = 0 GPa (solid circles). This indicates that the kk-dependence is small in all the calculated systems, similar to the case of the diamond system presented in the main text.

Z-method

The result of the Z-method is derived under the basic assumption Usolid(Tu)=Uliquid(Tm)U_{\mathrm{solid}}(T_u) = U_{\mathrm{liquid}}(T_m), where Usolid(T)U_{\mathrm{solid}}(T) and Uliquid(T)U_{\mathrm{liquid}}(T) are the internal energies of the system in the solid and liquid states, respectively. Moreover, two approximate relations, CV(Tm)≃3kBC_V(T_m) \simeq3k_{\mathrm{B}} and ΔS≃kBln⁡2\Delta S \simeq k_{\mathrm{B}}\ln2, are used,[50] where CV(Tm)C_V(T_{\mathrm{m}}) is the specific heat at a constant volume in the solid state at TmT_{\mathrm{m}}, and ΔS\Delta S is the entropy change between the liquid and solid states.[51] These lead to Tu/Tm=(1+ΔS/CV(Tm))≃1.23T_{\mathrm{u}}/T_{\mathrm{m}}=(1+\Delta S/C_V(T_{\mathrm{m}}))\simeq1.23.[28]

Image not extracted

Figure A·4. kk-dependence of TmT_{\mathrm{m}} for NaC6_6 for N=14N=14 at P=100P=100 GPa (empty circles), FC10_{10} for N=44N=44 at P=200P=200 GPa (solid squares), ClC10_{10} for N=44N=44 at P=0P=0 GPa (empty triangles), and FC10_{10} for N=11N=11 at P=0P=0 GPa (solid circles).

The first assumption is based on the results of MD simulations and seems plausible. To confirm the validity of the second relations, we compared them with experimental results, where the difference between CV(Tu)C_V(T_{\mathrm{u}}) and CP(Tu)C_P(T_{\mathrm{u}}) was neglected. Table B·1 gives these experimental values with melting points TmexT_{\mathrm{m}}^{\mathrm{ex}} for several elements and compounds.[52] For reference, our calculated results of TmT_{\mathrm{m}} were also added.

NaCaAlNaClB4_4CBe2_2C
CP/kBC_P/k_{\mathrm{B}}3.805.203.964.053.793.69
ΔS/kB\Delta S/k_{\mathrm{B}}0.8440.9211.381.580.9171.26
ΔS/CP\Delta S/C_P0.2220.1770.2420.3890.2420.341
TmexT_{\mathrm{m}}^{\mathrm{ex}} [K]3711115933107427432400
TmT_{\mathrm{m}} [10310^3K]0.361.20.940.903.02.7
size (NN)1632484512

Table B·1. Experimental data for CPC_P, ΔS\Delta S, ΔS/CP\Delta S/C_P, and TmexT_{\mathrm{m}}^{\mathrm{ex}} together with calculated TmT_{\mathrm{m}}.

Although the experimental values of CP/kBC_P/k_{\mathrm{B}} and ΔS/kB\Delta S/k_{\mathrm{B}}[52] are larger than the theoretically expected values, the values of 1+ΔS/CP1+\Delta S/C_P are relatively close to 1.23. The above results suggest that the equation Tm=Tu/1.23T_{\mathrm{m}}=T_{\mathrm{u}}/1.23 is a reasonable approximation even for compounds.

To consider the validity of TmT_{\mathrm{m}}, we present some additional results on TmT_{\mathrm{m}} below. It is 431 K for Na with N=54N=54, 1504 K for Ca with N=4N=4, and 585 K for Al with N=32N=32. The results show that the size dependence of TmT_{\mathrm{m}} is not very consistent, and TmT_{\mathrm{m}} for larger systems does not necessarily give better values. Although it may be difficult to precisely estimate TmT_{\mathrm{m}} by this method, it is possible to obtain the approximate TmT_{\mathrm{m}} of complex systems such as compounds with a small amount of calculation.

Estimation of TcT_{\mathrm{c}}

To estimate the transition temperature TcT_{\mathrm{c}} of superconductivity, we calculated the Eliashberg electron phonon spectral function α2F(ω)\alpha^2F(\omega) within the framework of density functional perturbation theory (DFPT)[53] in the QE package, where ω\omega is the phonon frequency. Here, α2F(ω)\alpha^2F(\omega) is defined by

α2F(ω)=1N(εF)∑mn∑qνδ(ω−ωqν)∑k∣gk+q,kqν,mn∣2δ(εk+q,m−εF)δ(εk,n−εF),\alpha^2F(\omega)=\frac{1}{N(\varepsilon_{\mathrm{F}})}\sum_{mn}\sum_{\mathbf{q}\nu}\delta(\omega-\omega_{\mathbf{q}\nu})\sum_{\mathbf{k}}\left|g_{\mathbf{k}+\mathbf{q},\mathbf{k}}^{\mathbf{q}\nu,mn}\right|^2 \delta(\varepsilon_{\mathbf{k}+\mathbf{q},m}-\varepsilon_{\mathrm{F}})\delta(\varepsilon_{\mathbf{k},n}-\varepsilon_{\mathrm{F}}),

where gk+q,kqν,mng_{\mathbf{k}+\mathbf{q},\mathbf{k}}^{\mathbf{q}\nu,mn} is the matrix element of the electron-phonon coupling and N(εF)N(\varepsilon_{\mathrm{F}}) is the density of states at the Fermi energy εF\varepsilon_{\mathrm{F}} per both spins. Within DFPT, the electron-phonon matrix elements are obtained from the first-order derivative of the self-consistent Kohn-Sham potential. The summation over the Fermi surface is performed using the tetrahedron method.[27, 54]

The logarithmic average of phonon frequency ωlog⁡\omega_{\log} is written as

ωlog⁡=exp⁡{2λ∑qkmnlog⁡(ωqν)N(εF)ωqν∣gk+q,kqν,mn∣2δ(εk+q,m−εF)δ(εk,n−εF)}.\omega_{\log}=\exp\left\{\frac{2}{\lambda}\sum_{\mathbf{q}\mathbf{k}mn}\frac{\log(\omega_{\mathbf{q}\nu})}{N(\varepsilon_{\mathrm{F}})\omega_{\mathbf{q}\nu}}\left|g_{\mathbf{k}+\mathbf{q},\mathbf{k}}^{\mathbf{q}\nu,mn}\right|^2 \delta(\varepsilon_{\mathbf{k}+\mathbf{q},m}-\varepsilon_{\mathrm{F}})\delta(\varepsilon_{\mathbf{k},n}-\varepsilon_{\mathrm{F}})\right\}.

The electron-phonon coupling constant λ\lambda is given by

λ=2∫0∞α2F(ω)ω dω.\lambda=2\int_0^\infty\frac{\alpha^2F(\omega)}{\omega}\,\mathrm{d}\omega.

The transition temperature TcT_{\mathrm{c}} of superconductivity was calculated using the Allen and Dynes formulation,[43]

Tc=f1f2ωlog⁡1.2exp⁡(−1.04(1+λ)λ−μ∗(1+0.62λ)),T_{\mathrm{c}}=f_1f_2\frac{\omega_{\log}}{1.2}\exp\left(-\frac{1.04(1+\lambda)}{\lambda-\mu^*(1+0.62\lambda)}\right),

where, the factors f1f_1 and f2f_2 are decided by the λ\lambda, μ∗\mu^*, ωlog⁡\omega_{\log} and mean square frequency ⟨ω2⟩\langle\omega^2\rangle. They are given by

f1=[1+(λ2.46(1+3.8μ∗))3/2]1/3f_1=\left[1+\left(\frac{\lambda}{2.46(1+3.8\mu^*)}\right)^{3/2}\right]^{1/3}
f2=1+(⟨ω2⟩/ωlog⁡−1)λ2λ2+[1.82(1+6.3μ∗)(⟨ω2⟩/ωlog⁡)]2,f_2=1+\frac{\left(\sqrt{\langle\omega^2\rangle}/\omega_{\log}-1\right)\lambda^2}{\lambda^2+\left[1.82(1+6.3\mu^*)\left(\sqrt{\langle\omega^2\rangle}/\omega_{\log}\right)\right]^2},

where ⟨ω2⟩=2λ∫0∞α2F(ω)ω dω\langle\omega^2\rangle=\frac{2}{\lambda}\int_0^\infty\alpha^2F(\omega)\omega\,\mathrm{d}\omega. The screened Coulomb potential parameter μ∗\mu^* is set to 0.1 as a typical value.

In our calculation, kk mesh (qq mesh) is chosen to be 8×8×88\times8\times8 (4×4×44\times4\times4). In the optimization of structures, we used convergence thresholds as 3×10−63\times10^{-6} a.u. in total energy and 3×10−53\times10^{-5} a.u. in force. To verify the validity of our estimation, we calculated TcT_{\mathrm{c}} of YH6_6 at P=166P=166 GPa. We found that Tc≃259T_{\mathrm{c}}\simeq259 K, while the experimental result at the same pressure is 224 K.[10, 11] For LaH10_{10}, TcT_{\mathrm{c}} is estimated to be 212 K at P=220P=220 GPa, but it is 245 K in the experimental results.[8, 9] These results suggest that the above formulation successfully derives TcT_{\mathrm{c}} for phonon high-TcT_{\mathrm{c}} superconductors.

As an example, the result of FC10_{10} at P=0P=0 GPa is shown. In this case, we obtained ωlog⁡≃697\omega_{\log}\simeq697 K, λ≃1.78\lambda\simeq1.78, and Tc≃106T_{\mathrm{c}}\simeq106 K. We confirmed that the results of TcT_{\mathrm{c}} are almost the same even if the meshes are different. For example, if kk and qq meshes were chosen to be $6\times6\times6$ and 3×3×33\times3\times3, respectively, we obtained a result of Tc≃112T_{\mathrm{c}}\simeq112 K. For 12×12×1212\times12\times12 and 6×6×66\times6\times6, we obtained $T_{\mathrm{c}}\simeq104 K.

Using α2F(ω)\alpha^{2}F(\omega), we defined λ(ω)\lambda(\omega) as

λ(ω)=2∫0ωα2F(ω′)ω′dω′.\lambda(\omega)=2\int_{0}^{\omega}\frac{\alpha^{2}F(\omega')}{\omega'}d\omega'.

This allows us to see the contribution of each atom to λ\lambda. Figure C·1 shows λ(ω)\lambda(\omega) as a function of ω\omega (solid line). Figure C·1 also shows the phonon DOS decomposed into the fluorine and carbon atoms, where the dotted line represents the contributions of fluorine atoms, and the dashed line represents the contributions of carbon atoms. It indicates that the phonon spectrum is clearly divided into two parts: components of F and C atoms. Furthermore, the contribution of phonons to λ(ω)\lambda(\omega) is also divided into two parts, and the F atom component seems to be about 1/81/8 of the C atom component. This result indicates that it is mainly the carbon atoms that contribute to the superconductivity of FC10. Similar results have already been obtained for XC6.22)

Figure C·1. Electron-phonon coupling constant λ(ω)\lambda(\omega) and phonon DOS ρ(ω)\rho(\omega) for FC10 at P=0P=0 GPa. Here, the solid line represents λ(ω)\lambda(\omega), the dotted line represents the contributions of fluorine atoms to ρ(ω)\rho(\omega), and the dashed line represents the contributions of carbon atoms to ρ(ω)\rho(\omega).

Free Energy

We considered the Helmholtz free energy F(T)F(T) per atom within the harmonic approximation for phonons. In the harmonic approximation, the volume of the system does not depend on TT. Thus, we neglected the volume change of the system due to temperature hereafter.

Using the free energy formulation for a harmonic oscillator, we obtained

F(T)=∫[ℏω2+kBTln⁡(1−exp⁡(−ℏωkBT))]ρph(ω)dω.F(T)=\int\left[\frac{\hbar\omega}{2}+k_{\mathrm{B}}T\ln\left(1-\exp\left(-\frac{\hbar\omega}{k_{\mathrm{B}}T}\right)\right)\right]\rho_{\mathrm{ph}}(\omega)d\omega.

where ρph(ω)\rho_{\mathrm{ph}}(\omega) is the phonon DOS per atom, which is calculated by QE. The internal energy U(T)U(T) was also obtained by a similar formulation as

U(T)=∫ℏω2[1+1exp⁡(ℏωkBT)−1]ρph(ω)dω.U(T)=\int\frac{\hbar\omega}{2}\left[1+\frac{1}{\exp\left(\frac{\hbar\omega}{k_{\mathrm{B}}T}\right)-1}\right]\rho_{\mathrm{ph}}(\omega)d\omega.

Here, entropy S(T)S(T) can be easily obtained by S(T)=(U(T)−F(T))/TS(T)=(U(T)-F(T))/T.

To compare the experimental results with the above results, we considered the enthalpy of phonons, H(T)H(T). In our approximation, the pVpV term in the enthalpy is independent of temperature. Therefore, H(T)H(T) as a function of TT is equivalent to U(T)U(T), except for a constant. Since enthalpy is typically used in experiments,52) we used H(T)H(T) instead of U(T)U(T) to avoid confusion.

As shown in Fig. D·1(a), we found that the experimental result of the enthalpy Hex(T)H^{\mathrm{ex}}(T) for sodium52) is well fitted by Hex(T)≃csolidH(T)H^{\mathrm{ex}}(T)\simeq c_{\mathrm{solid}}H(T) for T≤TmT\leq T_{\mathrm{m}} and Hex(T)≃cliquidH(T)+ΔgH^{\mathrm{ex}}(T)\simeq c_{\mathrm{liquid}}H(T)+\Delta_{g} for T>TmT>T_{\mathrm{m}}, where csolidc_{\mathrm{solid}} is a constant, 1.06, and cliquidc_{\mathrm{liquid}} is 1.19. Here, Δg\Delta_{g} is the value to reproduce the gap in Hex(T)H^{\mathrm{ex}}(T) at TmT_{\mathrm{m}}, which will be given later. For the case of NaCl,52) we obtained csolid=1.06c_{\mathrm{solid}}=1.06 and cliquid=1.28c_{\mathrm{liquid}}=1.28, as shown in Fig. D·1(b). The csolidc_{\mathrm{solid}} values are the same for Na and NaCl, but the cliquidc_{\mathrm{liquid}} values are slightly different.

As shown in Figs. D·1(c) and D·1(d), we found that similar relationships hold for entropies. That is, Sex(T)≃csolidS(T)S^{\mathrm{ex}}(T)\simeq c_{\mathrm{solid}}S(T) for T<TmT<T_{\mathrm{m}} and Sex(T)≃cliquidS(T)S^{\mathrm{ex}}(T)\simeq c_{\mathrm{liquid}}S(T) for T≥TmT\geq T_{\mathrm{m}}. These relations lead to F~(T)≃csolidF(T)\widetilde{F}(T)\simeq c_{\mathrm{solid}}F(T) for T<TmT<T_{\mathrm{m}} and F~(T)≃cliquidF(T)+Δg\widetilde{F}(T)\simeq c_{\mathrm{liquid}}F(T)+\Delta_{g} for T≥TmT\geq T_{\mathrm{m}}, where F~(T)\widetilde{F}(T) is the phenomenologically corrected free energy that is fitted to closely reproduce the experimental results, and

Δg=(csolid−cliquid)F(Tm).\Delta_{g}=(c_{\mathrm{solid}}-c_{\mathrm{liquid}})F(T_{\mathrm{m}}).

The above phenomenological relations give approximate free energies of the solid and liquid states which closely reproduce the experimental results. The results shown in Fig. D·1 suggest that this approximation can be widely applied to a variety of materials. In fact, we confirmed that this method can also be applied to Al and B4_4C.52) We found that csolid=1.09c_{\mathrm{solid}}=1.09 and cliquid=1.3c_{\mathrm{liquid}}=1.3 in the former case. In the latter case, csolid=1.03c_{\mathrm{solid}}=1.03 and cliquid=1.13c_{\mathrm{liquid}}=1.13.

Figure D·1. Theoretical and experimental results of enthalpy as a function of TT for (a) Na and (b) NaCl, and those of entropy for (c) Na and (d) NaCl, where the experimental results are taken from NIST.52) Here, the experimental values of enthalpy were shifted by a constant to match the theoretical value of H(T)H(T) at T=0T=0.

To test the above analysis, we tried to find the phase boundary between graphite and diamond. When T=0T=0, the enthalpy of diamond is lower than that of graphite and diamond is stable for P≳2P\gtrsim2 GPa.[41, 45] As the temperature increases, the free energy of graphite decreases more quickly than that of diamond, and graphite becomes stable. The boundary is given by ΔF~(T)+ΔH=0\Delta\widetilde{F}(T)+\Delta H=0, where ΔF~(T)\Delta\widetilde{F}(T) and ΔH\Delta H are the formation free energy and the formation enthalpy of graphite, respectively.

Figure D · 2 shows the F~(T)\widetilde{F}(T) of diamond and graphite and the ΔF~(T)\Delta\widetilde{F}(T) of graphite, where csolidc_{\mathrm{solid}} is set to 1.06 for both diamond and graphite. Here, we confirmed that the F~(T)\widetilde{F}(T) of graphite almost reproduces the experimental result.[52] In the figure, the solid lines represent the F~(T)\widetilde{F}(T) of diamond and graphite at P=8P = 8 GPa. ΔF~(T)\Delta\widetilde{F}(T) is shown as a dotted line for P=8P = 8 GPa and as a dashed line for P=15P = 15 GPa. The two lines are close to each other and look nearly identical. The result indicates that the PP-dependence of ΔF~(T)\Delta\widetilde{F}(T) is very small. On the other hand, the PP-dependence of ΔH\Delta H is not so small and dominates the conditions that determine the phase boundary.

The ΔH\Delta H values are −0.129-0.129, −0.0283-0.0283, 0.03180.0318, 0.05140.0514, and 0.12930.1293 for P=0P = 0, 55, 88, 1010, and 1515 GPa, respectively. Using the results above, we found that the boundary temperatures are approximately 00, 750750, 13001300, and 34003400 K for P=6.9P = 6.9, 88, 1010, and 1515 GPa, respectively. This boundary appears to be shifted by about 55 GPa toward higher pressure compared with other theoretical[55] and experimental[41] results. Apart from this, the result seems to agree with those of other works.

The application to the boron-doped diamond system is also interesting. It is synthesized at high pressure (∼10\sim10 GPa) and temperature (∼3000\sim3000 K) and shows superconductivity with Tc∼4T_{\mathrm{c}} \sim4 K.[12] Here, the boron concentration ρb\rho_{\mathrm{b}} is about 3 percent. Figure D · 3 shows the F~(T)\widetilde{F}(T) of diamond, boron-doped diamond, and boron at P=10P = 10 GPa. For boron-doped diamond, we used the N=32(4×4×2)N = 32(4 \times4 \times2) system, in which one atom is replaced by boron.

Figure D · 2. Phenomenological free energies F~(T)\widetilde{F}(T) of diamond and graphite (solid lines) at P=8P = 8 GPa, together with the formation free energies ΔF~(T)\Delta\widetilde{F}(T) of graphite at P=8P = 8 (dotted line) and P=15P = 15 GPa (dashed line).

Figure D · 3. Phenomenological free energies F~(T)\widetilde{F}(T) of diamond, boron, and boron-doped diamond at P=10P = 10 GPa (solid lines). The formation free energy ΔF~(T)\Delta\widetilde{F}(T) of boron-doped diamond is represented by a dotted line.

In a real system, the effect of entropy becomes important[56] because the boron atoms are randomly distributed in the system. We evaluated the additional entropy by treating the boron-doped system as a type of binary alloy.[57] In this case, the entropy is expressed as −kB(ρbln⁡ρb+(1−ρb)ln⁡(1−ρb))-k_{\mathrm{B}}(\rho_{\mathrm{b}} \ln\rho_{\mathrm{b}} + (1 - \rho_{\mathrm{b}}) \ln(1 - \rho_{\mathrm{b}})). For ρb=0.03\rho_{\mathrm{b}} = 0.03, the value is ≃0.12×10−4\simeq0.12 \times10^{-4} eV/K, and we add it to the F~(T)\widetilde{F}(T) of the boron-doped diamond.

In this figure, the dotted line indicates the formation free energy ΔF~(T)\Delta\widetilde{F}(T) of the boron-doped diamond. Here, to obtain ΔF~(T)\Delta\widetilde{F}(T), we assumed that the crystal structure of boron is R3ˉmR\bar{3}m.[45, 58] It decreases with increasing temperature, and ΔF~(T)+ΔH\Delta\widetilde{F}(T) + \Delta H becomes negative at T≳1200T \gtrsim1200 K. Here, ΔH\Delta H is the formation enthalpy of the boron-doped diamond, and its value is about 0.0380.038 eV. This result means that the boron-doped diamond is stable at T≳1200T \gtrsim1200 K.[59] It does not contradict the synthesis conditions of the boron-doped diamond.[12]

As mentioned above, our evaluation of the free energy is expected to be useful in finding conditions for material synthesis. To analyze the free energy of NaC6_6, we used csolid=1.06c_{\mathrm{solid}} = 1.06 and cliquid=1.13c_{\mathrm{liquid}} = 1.13 as the phenomenological parameters for sodium at P=300P = 300 GPa. For simplicity, we also used csolid=1.06c_{\mathrm{solid}} = 1.06 for diamond and NaC6_6. The calculated results are given in the main text.

References

  1. [8]8) A. P. Drozdov, P. P. Kong, V. S. Minkov, S. P. Besedin, M. A. Kuznovikov, S. Mozafari, L. Balicas, F. F. Balakirev, D. E. Graf, V. B. Prakapenka, E. Greenberg, D. A. Knyazev, M. Tkacz, and M. I. Eremets, Nature 569, 528 (2019).
  2. [9]9) M. Somayazulu, M. Ahart, A. K. Mishra, Z. M. Geballe, M. Baldini, Y. Meng, V. V. Struzhkin, and R. J. Hemley, Phys. Rev. Lett. 122, 027001 (2019).
  3. [10]10) I. A. Troyan, D. V. Semenok, A. G. Kvashnin, A. G. Ivanova, V. B. Prakapenka, E. Greenberg, A. G. Gavriliuk, I. S. Lyubutin, V. V. Struzhkin, and A. R. Oganov, Adv. Mater. 33, 2006832 (2021).
  4. [11]11) P. P. Kong, V. S. Minkov, M. A. Kuznovikov, S. P. Besedin, A. P. Drozdov, S. Mozafari, L. Balicas, F. F. Balakirev, V. B. Prakapenka, E. Greenberg, D. A. Knyazev, and M. I. Eremets, Nat. Commun. 12, 5075 (2021).
  5. [12]12) E. A. Ekimov, V. A. Sidorov, E. D. Bauer, N. N. Mel'nik, N. J. Curro, J. D. Thompson, and S. M. Stishov, Nature 428, 542 (2004).
  6. [14]14) A. Kawano, H. Ishikawa, S. Iriyama, R. Okada, T. Yamaguchi, Y. Takano, and H. Kawarada, Phys. Rev. B 82, 085318 (2010).
  7. [15]15) H. Okazaki, T. Wakita, T. Muro, T. Nakamura, Y. Muraoka, T. Yokoya, S. Kurihara, H. Kawarada, T. Oguchi, and Y. Takano, Appl. Phys. Lett. 106, 052601 (2015).
  8. [18]C. M. Hao, X. Li, A.R. Oganov, J. Hou, S. Ding, Y. Ge, L. Wang, X. Dong, H. T. Wang, G. Yang, X. F. Zhou, and Y. Tian, Phys. Rev. B 108, 214507 (2023).
  9. [21]Q. Wei, Q. Zhang, and M. Zhang, Materials 9, 726 (2016).
  10. [22]K. Sano, Y. Masuda, and H. Ito, J. Phys. Soc. Jpn. 91, 083703 (2022).
  11. [27]P. Giannozzi, S. Baroni, N. Bonini, M. Calandra, R. Car, C. Cavazzoni, D. Ceresoli, G. L. Chiarotti, M. Cococcioni, I. Dabo, A. Dal Corso, S. Fabris, G. Fratesi, S. de Gironcoli, R. Gebauer, U. Gerstmann, C. Gougoussis, A. Kokalj, M. Lazzeri, L. Martin-Samos, N. Marzari, F. Mauri, R. Mazzarello, S. Paolini, A. Pasquarello, L. Paulatto, C. Sbraccia, S. Scandolo, G. Scalzuero, A. P. Seitsonen, A. Smogunov, P. Umari, and R. M. Wentzcovitch, J. Phys. Condens. Matter 21, 395502 (2009). https://www.quantum-espresso.org/
  12. [28]A. B. Belonoshko, N. V. Skorodumova, A. Rosengren, and B. Johansson, Phys. Rev. B 73, 012201 (2006).
  13. [41]F. P. Bundy, W. A. Bassett, M. S. Weathers, R. J. Hemley, H. K. Mao, and A. F. Goncharov, Carbon 34, 141 (1996).
  14. [43]P. B. Allen and R. C. Dynes, Phys. Rev. B 12, 905 (1975).
  15. [44]C. S. Sundar, A. Bharathi, Y. Hariharan, J. Janaki, V. S. Sastry, and T. S. Radhakrishnan, Solid State Commun. 84, 823 (1992).
  16. [45]A. Jain, S. P. Ong, G. Hautier, W. Chen, W. D. Richards, S. Dacek, S. Cholia, D. Gunter, D. Skinner, G. Ceder, and K. A. Persson, APL Mater. 1, 011002 (2013). www.materialsproject.org
  17. [46]C. Tandardini, F. N. Jalolov, and A. G. Kvashnin, J. Phys. Chem. C 126, 11358 (2022).
  18. [47]N. E. Christensen and D. L. Novikov, Solid State Commun. 119, 477 (2001).
  19. [48]Y. Huang, M. Widom, and M. C. Gao, Phys. Rev. Mater. 6, 013802 (2022).
  20. [49]To precisely discuss the synthesis conditions, it may be necessary to improve the accuracy of ΔH. Additionally, the presence of other compounds such as NaC may need to be considered.
  21. [50]If the simulations can be performed in sufficient detail, it is possible to determine Tₘ without using these two approximate relations.
  22. [51]S. M. Stishov, Usp. Fiz. Nauk 114, 3 (1974).
  23. [52]National Institute of Standards and Technology Federal Information Processing Standards Publications, 140, 2 (2002). https://janof.nist.gov/
  24. [53]S. Baroni, S. de Gironcoli, A. Dal Corso, and P. Giannozzi, Rev. Mod. Phys. 73, 515 (2001).
  25. [54]M. Kawamura, Y. Ghoda, and S. Tsuneyuki Phys. Rev. B 89, 094515 (2014).
  26. [55]L. M. Ghiringhelli, J. H. Los, E. J. Meijer, A. Fasolino, and D. Frenkel, Phys. Rev. Lett. 94, 145701 (2005).
  27. [56]A. Manzoor, S. Pandey, D. Chakraborty, S. R. Phillpot, and D. S. Aidhy, npj Comput. Mater. 4, 47 (2018).
  28. [57]For a binary alloy, a lattice gas model is often used as a simple approximation. See below for its entropy formula. G. D. With, Liquid-State Physical Chemistry: Fundamentals, Modeling, and Applications (Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim, 2013) Appendix C.
  29. [58]G. Parakhonskiy, N. Dubrovinskaya, L. Dubrovinsky, S. Mondal, and S. v. Smaalen, J. Cryst. Growth 321, 162 (2011).
  30. [59]We confirmed that these results are almost unchanged even if the existence of B₄C is taken into account.

Paper details

vlnc:W_q2xgrn4t
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