betaFeedback

WorksPhysics and Astronomy

Isotope Effects and the Negative Thermal Expansion Phenomena in Ice and Water

Abstract

H2O is a unique substance with exceptional thermal properties arising from the subtle interplay between its electronic, phononic, and structural degrees of freedom. Of particular interest in H2O are the negative thermal expansion (NTE) phenomena, observed in its solid phase (ice) at low temperature, and in its liquid phase (water) near the freezing temperature. Furthermore, ice and water exhibit the abnormal volume isotope effect (VIE), where volume expansions occur when replacing H with its heavier isotope, deuterium (D). In order to capture more conceptual and intuitive understanding of intriguing NTE and VIE phenomena in ice and water, we have explored isotope effects in their NTE and melting properties by employing a type of Born-Oppenheimer-approximation approach and the Lindemann criterion. Our findings demonstrate that unusual isotope effects in these phenomena stem from competition between zero-point-energy phonons, thermal phonons, and the hydrogen bonding in H2O. All these components originate from nuclear quantum mechanical (QM) processes, revealing that QM physics plays a crucial role in the seemingly classical ice/water systems.

Introduction

Thermal expansion is a phenomenon of length, area, or volume change of a material with varying temperature (TT). Most substances exhibit positive thermal expansions (PTEs), implying that they expand upon heating and contract upon cooling. However, H2_2O in its different solid, liquid, and gas phases manifests a fascinating exception to this rule (see the phase diagram in Figure 1(a)) [1]. Namely, contrary to the general rule, upon cooling, liquid water expands below T=4T = 4 °C exhibiting the volume minimum (density maximum) at T=4T = 4 °C, and then its volume increases abruptly by about 9% when it freezes to ice at T=0T = 0 °C (T=273T=273 K), as shown in Figure 1(b). Hence liquid water has the negative thermal expansion (NTE) coefficient between 0 °C and 4 °C (inset of Figure 1(b)), and ice is less dense than liquid water due to its larger volume [2]. In fact, the negative slope (dPdT<0\frac{dP}{dT} < 0) of phase-equilibrium line between ice and water in Figure 1(a) is closely related to the observed NTE at the transition [3, 4, 5, 6]. The density maximum at T=4T = 4 °C in H2_2O is well known to be essential for ecological systems and various environmental processes [2]. Furthermore, H2_2O exhibits a colossal PTE (volume expands by ∼1600\sim1600 times), when boiling into its vapor gas phase.

Image not extracted

Figure 1. (a) Phase diagram of H2_2O [1]. A, B, C, and D correspond to the triple point, freezing point, boiling point, and critical point, respectively. (b) Thermal expansion data near the freezing point [29]. Note the NTE (abrupt volume jump) at the water to ice transition at T=0T = 0 °C (273 K), and the NTE of water between 0 °C (273 K) and 4 °C (277 K) shown in the inset. (c) The thermal expansion data of ice-Ih phases of H2_2O and D2_2O. The NTEs are observed below ∼70\sim70 K for both. D2_2O has the larger volume than H2_2O, manifesting the abnormal VIE. (d) Volume variations of supercooled-water isotopes as a function of TT [2, 30]. (e) Open hexagonal crystal structure of ice-Ih (viewed along the cc axis). (f) Hydrogen-bonding (H:O-bonding) in tetrahedral building block of ice-Ih. Each oxygen forms two-short H-O covalent bonds, and two-long hydrogen bonds (H:O-bonds) with neighboring H ions.

The most common form of ice at the ambient pressure is the ice-Ih phase [2]. As shown in Figure 1(c), ice-Ih exhibits the normal PTE between ∼70\sim70 K and 273 K (0 °C), albeit at a much lower thermal-expansion rate than liquid water. A notable feature in ice-Ih is another NTE phenomenon below ∼70\sim70 K. The NTE behavior in ice-Ih at low TT is known to arise from low-energy vibrational modes of ice-Ih phase, which produce the negative Grüneisen parameters [4, 7, 8, 9, 10, 11]. Ice-Ih exhibits an additional abnormal feature in the volume isotope effect (VIE), where the volume of the ice changes distinctly depending on the isotope of hydrogen [12, 13, 14]. In most materials, replacing lighter isotopes with heavier ones leads to a normal VIE, indicating that the material contracts in volume. However, in ice-Ih, replacing H with its heavier isotope, deuterium (D), leads to the volume expansion, as shown in Figure 1(c). This abnormal VIE was reported to arise from the quantum-mechanical (QM) nature of the hydrogen atoms in the ice lattice [9, 10, 15, 16, 17]. Interestingly, the abnormal VIE is also observed in water near the so-called TMDT_{MD} (TMDT_{MD}: the temperature corresponding to the maximum density), as shown in Figure 1(d).

The anomalous thermal expansion properties of ice and water highlight the complex physics and chemistry of the H2_2O system. Through decades of research, their mechanisms are understood to some extent based on numerical computations of the Grüneisen parameters of phonon modes at low TT [9, 10, 11, 15] and numerical simulations of molecular-dynamics at high TT [18, 19, 20, 21, 22, 23]. Despite extensive studies, however, conceptual understanding of the underlying physics and chemistry of intriguing NTE and VIE phenomena in ice and water is still lacking. So we revisit these issues to capture the comprehensive and intuitive picture from a more fundamental point of view.

In this paper, to identify the physics behind the NTE phenomena in ice and water, we have examined the isotope effects in the NTE properties, such as the volume-minimum temperatures (TvmT_{vm}’s) and the melting temperatures (TmT_m’s) as well as the VIEs. Specifically, we have addressed the issues on (i) the origin of different TvmT_{vm}’s for different isotopes as well as the abnormal VIE observed in ice-Ih at low TT, and (ii) the origin of different TmT_m’s and TMDT_{MD}’s for different isotopes as well as the abnormal VIE observed in ice-Ih and water near the freezing TT. In the following, the approach we have adopted for the present analysis, including a type of Born-Oppenheimer approximation and the Lindemann criterion, is briefly described (§ 2), and then results and discussions of our findings are presented (§ 3), which is followed by concluding remarks (Section IV).

Methodology

Based on the analysis of existing experimental data for ice and water, we have attempted to provide conceptual interpretation of anomalous NTE/VIE and melting properties observed in ice and water. For this purpose, we have newly employed a type of Born-Oppenheimer approximation [24, 25] to explain the former and the Lindemann criterion [26] to explain the latter.

A type of Born-Oppenheimer approximation we have employed is described in the supplement section (ii) [6]. For a system like H2_2O having the fast-moving high-energy (ω\omega) and the slow-moving low-energy (Ω\Omega) phonon modes, which are far-separated in energy, it is possible to separate adiabatically the high-energy and low-energy phonon modes [24, 25]. This procedure is just like a conventional Born-Oppenheimer approximation approach applied for a system with the fast-moving electron and slow-moving ion motions [27]. Then the zero-point energy (12ℏω\frac{1}{2}\hbar\omega) of high-energy phonon mode plays a role of additional potential energy for the low-energy phonon mode. We have examined the role of this zero-point energy in the anomalous NTE/VIE properties of ice and water.

The application of Lindemann criterion to the melting property of ice is rather rare [28]. Here, to explain the unusual isotope effects observed in the melting of ice, we have adopted a simple Lindemann criterion, which claims that a crystal melts when the average amplitude of thermal vibrations (⟨u2⟩1/2\langle u^2\rangle^{1/2}) is larger than some fraction of inter-atomic distance. The derivation of ⟨u2⟩\langle u^2\rangle is provided in the supplement section (vi) [6].

Results and Discussions

Hydrogen (H:O)-bonds in ice and water

The anomalous thermal expansion behaviors in ice-Ih and water are known to arise from their characteristic hydrogen-bonding (H:O-bonding) networks [2]. The H:O-bonding arises from an intramolecular attractive interaction, in which a H atom that is covalently bonded to a highly electronegative O atom is attracted to a lone pair of electrons of an O atom in a neighboring molecule. These intermolecular H:O-bonds, albeit much weaker than intramolecular H-O covalent bonds, play a crucial role in determining the arrangement and dynamics of ice and water.

Ice exists in nearly twenty forms of solid phases, each with a unique arrangement of H2_2O molecules held together by H:O-bonds [2]. Among those, Ice-Ih crystallizes in an open hexagonal lattice, in which oxygen ions form an ordered Wurtzite structure having a low packing ratio, as shown in Figure 1(e) [31]. The H:O-bonding network in ice-Ih is composed of tetrahedral building blocks, where each H2_2O molecule forms four H:O-bonds with its neighboring molecules, as depicted in Figure 1(f). In ice-Ih, light H ions are not ordered due to their large zero-point motions.

The NTE observed at low TT (<70< 70 K) in ice-Ih is related to its open, tetrahedral-network structure formed by H:O-bonds. It has been found that low-energy transverse phonon modes in ice-Ih, which are associated with bending motions of three hydrogen-bonded molecules, bring about the negative Grüneisen parameter and consequently the NTE phenomenon [4, 7, 8, 9, 10, 11, 15]. In fact, such NTE phenomena are manifested in other tetrahedrallyarranged solids having a low packing ratio, such as Si and Ge [4, 7, 32]. However, they do not exhibit anomalous VIEs but just normal VIEs, distinct from ice-Ih [33, 34].

Isotope Effects in the NTE of ice-Ih at low TT: Born-Oppenheimer approximation

In the Born-Oppenheimer approximation, the adiabatic separation of the fast-moving electron wave-function and the slow-moving ion wave-function is possible, and then the obtained volume-dependent eigenvalue of electron (ϵ(V)\epsilon(V)) acts as an additional interionic potential energy for the ion motions [27]. In a similar way, for a system with the high-energy (ω\omega) and the low-energy (Ω\Omega) phonon modes, which are far-separated in energy, one can separate adiabatically the high-energy (described by qq, pp variables with mass mm) and low-energy (described by QQ, PP variables with mass MM) phonon modes [24, 25]. Then, as described in the supplement [6], the zero-point phonon energy of the fast-moving qq mode obtained in the quasi-harmonic approximation plays a role of the additional potential energy for the low-energy phonon modes:

[P22M+U0(Q)+12ℏω(Q)]χ0N(Q)=E0Nχ0N(Q),(1)\left[\frac{P^2}{2M}+U_0(Q)+\frac{1}{2}\hbar\omega(Q)\right]\chi_0^N(Q)=E_0^N\chi_0^N(Q), \tag*{(1)}

where χ0N(Q)\chi_0^N(Q) is the wave function of low-energy phonon (slow QQ) mode, and U0(Q)U_0(Q) is the interionic potential energy. Here kk-dependence of ω\omega is neglected by using the Einstein-phonon approximation.

In the case of hydrogen-bonded systems, the variable QQ corresponds to the intermolecular O−O bond length, and so it can be considered to be a volume (VV) variable [24, 25]. Accordingly, the effective intermolecular potential energy for the low-energy phonon modes Ueff(V)U_{eff}(V) is given by

Ueff(V)=U0(V)+UZP(V),(2)U_{eff}(V)=U_0(V)+U_{ZP}(V), \tag*{(2)}

where U0(V)U_0(V) is the intermolecular potential energy for the static lattice, and UZP(V)(≡12ℏω(V))U_{ZP}(V)(\equiv\frac{1}{2}\hbar\omega(V)) is the QM zero-point energy of the high-energy phonon mode. Consequently, the low-energy vibrational motions become significantly affected by UZP(V)U_{ZP}(V), depending on the volume-dependence of ω(V)\omega(V), as provided in Figure 2 below.

Schematic illustrations of phonon-mode behavior, effective intermolecular potentials, and isotope effects

Figure 2. (a) Typical volume-dependent behaviors of high-energy phonon modes ωs(V)\omega_s(V). The upper one exhibiting the pressure-induced hardening produces the positive Grüneisen parameter and the PTE, while the lower one exhibiting the pressure-induced softening produces the negative Grüneisen parameter and the NTE. (b) Effective intermolecular potential energies UeffU_{\mathrm{eff}}’s (Ueff=U0+UZPU_{\mathrm{eff}} = U_0 + U_{\mathrm{ZP}}) for the PTE-anharmonic (top) and the NTE-anharmonic (bottom) systems, respectively. U0U_0 is assumed to be harmonic (middle). Red dotted lines in the top and bottom illustrate the volume expansion and contraction, respectively, with increasing TT. (c) The intermolecular potential energy, U0(V)U_0(V) (red curve), which is intrinsically PTE-anharmonic, can be approximated by the harmonic potential (blue curve) at low TT. (d) For the NTE case with negative γs\gamma_s, the heavier-mass system of D2_2O ice has the larger volume (red dot) than the lighter-mass system of H2_2O ice (blue dot) (ΔV0≡V0D−V0H>0\Delta V_0 \equiv V_0^{D} - V_0^{H} > 0). Due to the mass-dependent QM zero-point energy UZPU_{\mathrm{ZP}}, D2_2O ice would have larger binding energy than H2_2O ice by ΔUZP\Delta U_{\mathrm{ZP}} (ΔUZP\Delta U_{\mathrm{ZP}} value taken from Refs. [17, 39]).

Now consider a system having the high-energy phonon-mode ωs(V)\omega_s(V), which decreases or increases with volume, as displayed in Figure 2(a). In fact, U0(V)U_0(V) for the static lattice is intrinsically anharmonic in nature. But, at low TT, when the vibrational amplitudes are small, U0(V)U_0(V) can be assumed to be harmonic, i.e., U0(V)≈12K(V−V0)2U_0(V)\approx\frac{1}{2}K(V-V_0)^2, as shown in the middle of Figure 2(b) and Figure 2(c). Then, through the contribution of UZP(V)U_{ZP}(V), Ueff(V)U_{eff}(V) would have either the anharmonic potential of shallower-flattened-right (top) or that of shallower-flattened-left (bottom) with respect to the harmonic potential in the middle. According to the definition of Grüneisen parameter γ\gamma (≡−∂ln⁡(ω)/∂ln⁡V\equiv-\partial\ln(\omega)/\partial\ln V) [6], the former describes the PTE system with positive γs\gamma_s, while the latter describes the NTE system with negative γs\gamma_s. The red arrows in the top and bottom of Figure 2(b) illustrate the positive and negative volume expansions, respectively, with increasing TT.

In the case of ice-Ih, the high-energy phonon mode corresponds to the intramolecular O-H stretching mode with energy of ∼0.4\sim0.4 eV, while the low-energy phonon modes correspond to the intermolecular vibration modes with energy of ≤0.05\leq0.05 eV, which are considered to be responsible for the NTE at low TT [10, 35, 36, 37]. Hence the application of the Born-Oppenheimer approximation would be sufficiently valid in the present case [38].

Note that the high-energy stretching mode in ice-Ih, ωs\omega_s, was observed to possess a pressure-induced phonon softening [35]. This feature is understood in view of the fact that the O-H bond lengths tend to increase with increasing PP. In fact, its Grüneisen parameter was calculated to be negative (γs≈−0.2\gamma_s\approx-0.2) [9, 39]. Then, as described in the bottom of Figure 2(b), the zero-point energy of ωs(V)\omega_s(V) yields an NTE-anharmonic Ueff(V)U_{eff}(V). U0(V)U_0(V), being an electrostatic interaction, is volume independent of isotopic mass, and so U0(V)U_0(V) is same for both H2_2O and D2_2O. Hence, considering the mass-dependent UZPU_{ZP} (≡12ℏωs∼1miso\equiv\frac{1}{2}\hbar\omega_s\sim\frac{1}{\sqrt{m_{iso}}}), D2_2O ice would have a higher binding energy than H2_2O ice by ΔUZP\Delta U_{ZP} (≈50\approx50 meV) [17, 39], as displayed schematically in Figure 2(d).

In addition, because of the mass-dependent UZP(V)U_{ZP}(V), Ueff(V)U_{eff}(V) of D2_2O becomes less anharmonic than that of H2_2O, that is, Ueff(V)U_{eff}(V) of D2_2O is closer to the harmonic potential than that of H2_2O [40]. For simplicity, let us assume that there is just a single low-energy phonon mode, corresponding to an intermolecular vibration in ice-Ih, in the presence of Ueff(V)U_{eff}(V). Then, as shown in Figure 2(d), H2_2O having a stronger contribution from the anharmonic NTE potential would have smaller equilibrium volume (blue dot) than D2_2O (red dot) having a weaker contribution from the anharmonic NTE potential. Thus the asymmetric nature of Ueff(V)U_{eff}(V) of an NTE anharmonic type explains why ice-Ih exhibits the abnormal VIE, namely, D2_2O ice has the larger equilibrium volume than H2_2O ice (V0D>V0HV_0^D>V_0^H at T=0T=0).

The volume at the minimum of Ueff(V)U_{eff}(V), VmV_m, can be obtained quantitatively from the volume derivative of Ueff(V)U_{eff}(V). Considering the expansion of 12ℏωs(V)\frac{1}{2}\hbar\omega_s(V) up to the first order, Ueff(V)=K2(V−V0)2+12ℏωs(V0)+ℏ2(V−V0)∂ωs∂V∣V0U_{eff}(V)=\frac{K}{2}(V-V_0)^2+\frac{1}{2}\hbar\omega_s(V_0)+\frac{\hbar}{2}(V-V_0)\left.\frac{\partial\omega_s}{\partial V}\right|_{V_0}. This yields Vm=V0−ℏ2K∂ωs∂V∣VmV_m=V_0-\frac{\hbar}{2K}\left.\frac{\partial\omega_s}{\partial V}\right|_{V_m} from ∂Ueff∂V∣Vm=0\left.\frac{\partial U_{eff}}{\partial V}\right|_{V_m}=0. Therefore, the isotope volume difference between the D2_2O and H2_2O ice, ΔVm\Delta V_m (≡VmD−VmH\equiv V_m^D-V_m^H), is approximately given by

ΔVm≈ℏ2K(∂ωsH∂V−∂ωsD∂V)≈ℏ2K∂ωsH∂V(1−mHmD),(3)\Delta V_m\approx\frac{\hbar}{2K}\left(\frac{\partial\omega_s^H}{\partial V}-\frac{\partial\omega_s^D}{\partial V}\right)\approx\frac{\hbar}{2K}\frac{\partial\omega_s^H}{\partial V}\left(1-\sqrt{\frac{m_H}{m_D}}\right), \tag*{(3)}

which is positive because ∂ωsH∂V>0\frac{\partial\omega_s^H}{\partial V}>0, implying the abnormal VIE.

Indeed, the observed NTE for both H2_2O and D2_2O be low ∼70\sim70 K in Figure 1(c) is explained well by the NTE-anharmonic UeffU_{\mathrm{eff}} shown at the bottom of Figure 2(b). But, as TT increases above ∼70\sim70 K, the PTE is observed for both H2_2O and D2_2O, as if Ueff(V)U_{\mathrm{eff}}(V) changes from the NTE-anharmonic to the PTE-anharmonic type. This feature can be understood based on the change of U0(V)U_0(V) from the harmonic to the anharmonic form with increasing TT, as shown in Figure 2(c). Remember that, in Figure 2(d), the anharmonic U0(V)U_0(V) was assumed to be harmonic at low TT [41]. But, with increasing TT, the intrinsically PTE-anharmonic nature of U0(V)U_0(V) starts to be more and more effective, and so Ueff(V)U_{\mathrm{eff}}(V) of ice-Ih eventually becomes PTE-anharmonic, leading to volume expansions for both H2_2O and D2_2O. Namely, with increasing TT, the phonon modes with positive Grüneisen parameters become populated, and thereby reduce the NTE. Interestingly, even at finite TT above ∼70\sim70 K, the abnormal VIE will be maintained because Ueff(V)U_{\mathrm{eff}}(V) of H2_2O is still shallower-fattened-left than that of D2_2O due to a mass-dependent UZPU_{\mathrm{ZP}} contribution, and thereby the volume of H2_2O is still smaller than that of D2_2O.

As shown in Figure 1(c), there is also an isotope effect on the volume-minimum temperatures (TvmT_{\mathrm{vm}}), albeit not so prominent. In fact, TvmT_{\mathrm{vm}} in H2_2O ice is only slightly higher (∼1\sim1 K) than that in D2_2O ice [12, 14]. TvmT_{\mathrm{vm}} would be determined by the crossing point at which Ueff(V)U_{\mathrm{eff}}(V) changes from an NTE-anharmonic type to a PTE-anharmonic type. As described above, the NTE-anharmonic feature will persist up to higher TT in H2_2O than in D2_2O, and so TvmT_{\mathrm{vm}} in H2_2O is higher than TvmT_{\mathrm{vm}} in D2_2O (∼72\sim72 K for H2_2O vs. ∼71\sim71 K for D2_2O) [12].

Isotope Effects in the NTE of water near the freezing (melting) TT

As TT increases, the thermal vibrational energy of the H2_2O molecules in ice-Ih increases, which weakens the H:O-bonds and causes the molecules to fluctuate around their equilibrium positions. Then H:O-bonds become weaker and weaker so as to be broken, and eventually the volume collapse occurs at the melting temperature Tm=0 ∘CT_m = 0\,^{\circ}\mathrm{C}. This feature of ice-Ih is different from those of most solids, for which the extra molecular motion in the liquid phase requires more space and therefore melting is accompanied by volume expansion.

As shown in Figure 1(b), H2_2O water has the minimum volume at TMD=4 ∘CT_{\mathrm{MD}} = 4\,^{\circ}\mathrm{C}. This volume minimum is brought about by the balance between two competing effects with increasing TT: the volume contraction driven by melting vs. the normal volume expansion. As mentioned above, the crystalline H:O-bonding network is to be broken at Tm=0 ∘CT_m = 0\,^{\circ}\mathrm{C}. In the liquid phase, however, substantial amounts of H:O-bonds are still maintained in cluster forms of hexagonal ice-Ih phase keeping large empty space [42]. Upon heating, the number of these remaining H:O-bonds decreases slowly, resulting in the volume contraction. Meanwhile, due to the increased kinetic vibrational energy of unbonded H2_2O molecules, water expands upon heating like most substances. The crossing of these two effects is supposed to occur at TMDT_{MD}.

As introduced in Figure 1(d), the isotope effects are also manifested in water near TMDT_{MD}. As provided in Table I, D2_2O water has significantly higher TMDT_{MD} (∼11.2 ∘C\sim11.2\,^{\circ}\mathrm{C}) and the higher freezing (melting) temperature TmT_m (∼3.7 ∘C\sim3.7\,^{\circ}\mathrm{C}) than H2_2O. Heavier T2_2O (T: tritium) water has even higher TMDT_{MD} and TmT_m [20, 30]. Thus D2_2O and T2_2O waters exhibit wider temperature ranges of the NTE phenomena (7.5 and 8.9 ∘C^{\circ}\mathrm{C}) than H2_2O water (4 ∘C^{\circ}\mathrm{C}),

TmT_m [K][°C]TMDT_{MD} [K][°C](TMD−Tm)(T_{MD}-T_m) [K]
H2_2O273.150277.133.983.98
D2_2O276.833.68284.3411.197.51
T2_2O277.644.49286.5513.408.91

Table I. Isotope effects on the freezing (melting) temperature (TmT_m) and the maximum-density temperature (TMDT_{MD}) [20]. Here H, D, and T represent hydrogen, deuterium, and tritium, respectively.

Such differences are expected to occur due to the competition between the isotopic mass-dependent H:O-bonding (EHBE_{HB}) and the kinetic molecular-vibration (EkE_k) energies of H2_2O and D2_2O. EkE_k increases upon heating, and melting occurs when EkE_k overcomes the intermolecular energy holding the crystal structure together that corresponds to the cohesive energy (EcohE_{coh}) [43]. When H atoms are replaced with D atoms, the molecules would have stronger intermolecular D:O-bondings due to the reduced zero-point QM fluctuations of heavier D ions. On the other hand, larger intramolecular quantum delocalization of H ions in H2_2O brings about the enhanced dipole moment of the water molecule, producing a stronger H:O-bonding network in H2_2O than in D2_2O. Hence the energy difference in H:O and D:O-bonding energies will be much reduced due to two opposite effects of intermolecular and intramolecular nuclear QM fluctuations [44]. Nevertheless, the D:O-bonding network is expected to be still more robust than the H:O-bonding network [20, 45], and so EcohE_{coh} of D2_2O is larger than that of H2_2O by their difference in EHBE_{HB} (ΔEHB\Delta E_{HB}). EcohE_{coh} of H2_2O ice was reported to be ∼600\sim600 meV experimentally [46, 47], and EcohE_{coh} of D2_2O was estimated to be larger than EcohE_{coh} of H2_2O by ΔEHB\Delta E_{HB} of ∼15\sim15 meV [39]. Meanwhile, EkE_k including the zero-point motions in H2_2O ice was reported to be as much as ∼350\sim350 meV at low TT, whereas EkE_k in D2_2O is supposed to be smaller than EkE_k in H2_2O by ΔEk∼90\Delta E_k \sim90 meV due to its heavier mass [23]. Both EkE_k’s increase slowly with increasing TT.

In Figure 3(a) are plotted EkE_k’s and −Ecoh-E_{coh}’s of H2_2O and D2_2O ices as a function of TT. It is shown that the larger EkE_k in H2_2O ice will break the weaker H:O-bonds in H2_2O ice at a lower TT than in D2_2O ice, resulting in the lower

Image not extracted

Figure 3. (a) Schematic plot of TT-dependent variations of the kinetic molecular vibration energies (EkE_k’s) and the (−1×-1 \times) cohesive energies (−Ecoh-E_{coh}’s) of H2_2O (in blue) and D2_2O (in red) ices. The variation of EkE_k is somewhat exaggerated. Melting occurs when EkE_k overcomes EcohE_{coh}. It is shown that TmT_m of D2_2O is higher than that of H2_2O. ΔEk\Delta E_k and ΔEHB\Delta E_{HB} values are taken from Ref. [23] and Ref. [39], respectively. (b) Thermal-average amplitudes of ionic displacements ⟨u2⟩1/2\langle u^2\rangle^{1/2} of H2_2O (blue) and D2_2O (red) ices. Melting occurs when ⟨u2⟩1/2≥fa\langle u^2\rangle^{1/2} \ge fa (a: intermolecular distance, ff: fractional number (≤0.2\le0.2)).

TmT_m for H2_2O ice than for D2_2O ice (TmH<TmDT_m^H < T_m^D). Also, as a consequence, the lower TMDT_{MD} is naturally expected for H2_2O water than for D2_2O water.

This feature can also be analyzed by using the Lindemann criterion [26]. According to this criterion, a crystal melts when the average amplitude of thermal vibrations of atoms (⟨u2⟩1/2\langle u^2\rangle^{1/2}) is larger than some fraction of interatomic distances (fafa), i.e.,

⟨u2⟩1/2≥fa,(4)\langle u^2\rangle^{1/2} \ge fa, \tag*{(4)}

where uu is the atomic displacement, ff is some fractional number called the Lindemann parameter (f≤0.2f \le0.2 in general), and aa is the interatomic distance. As described in the supplement [6], ⟨ui2⟩\langle u_i^2\rangle for ii-th atom is given by ∑k,λℏ2Miωqλcoth⁡βℏωqλ2\sum_{\mathbf{k},\lambda} \frac{\hbar}{2M_i\omega_{\mathbf{q}\lambda}}\coth\frac{\beta\hbar\omega_{\mathbf{q}\lambda}}{2} (β=1kBT\beta= \frac{1}{k_BT}), indicating that all the phonon modes contribute to ⟨ui2⟩\langle u_i^2\rangle. Note that ⟨ui2⟩\langle u_i^2\rangle has isotopic-mass dependence through MiM_i and ωqλ\omega_{\mathbf{q}\lambda}, and, at T=0T = 0 K, ⟨ui2⟩=ℏ2Miωqλ\langle u_i^2\rangle= \frac{\hbar}{2M_i\omega_{\mathbf{q}\lambda}}. To examine the isotope effect on TmT_m, let us consider just one specific phonon mode (qλq\lambda) for simplicity, ⟨ui2⟩=ℏ2Miωqλcoth⁡βℏωqλ2\langle u_i^2\rangle= \frac{\hbar}{2M_i\omega_{q\lambda}}\coth\frac{\beta\hbar\omega_{q\lambda}}{2}. Accordingly, TmT_m can be obtained from

⟨ui2⟩1/2=[ℏ2Miωqλcoth⁡βmℏωqλ2]1/2=fai,(5)\langle u_i^2\rangle^{1/2} = \left[\frac{\hbar}{2M_i\omega_{q\lambda}}\coth\frac{\beta_m\hbar\omega_{q\lambda}}{2}\right]^{1/2} = fa_i, \tag*{(5)}

as schematically plotted in Figure 3(b). Due to a heavier mass of D2_2O than that of H2_2O, ⟨ui2⟩1/2\langle u_i^2\rangle^{1/2} of D2_2O is to be smaller than that of H2_2O. On the other hand, since the volume of D2_2O is larger than that of H2_2O, faif a_i of D2_2O is to be larger than that of H2_2O. TmT_m is obtained from the crossing of these two graphs, and so TmDT_m^D of D2_2O is to be higher than TmHT_m^H of H2_2O.

Let us try to estimate TmT_m for a specific phonon mode ωqλ\omega_{q\lambda}. The intermolecular H:O-bond stretching vibration mode of H2_2O ice is expected to be the most relevant mode to the melting of ice-Ih [10, 11, 17]. So let us select ωqλH=215 cm−1\omega^{H}_{q\lambda}=215\ \mathrm{cm}^{-1} (≈27\approx27 meV) for H2_2O, which is close to a center of phonon density of states of H:O-bond stretching vibration modes. For D2_2O, let us select ωqλD=204 cm−1\omega^{D}_{q\lambda}=204\ \mathrm{cm}^{-1} (≈25\approx25 meV), which is determined by counting the molecular-mass difference between D2_2O and H2_2O (ωqλD=ωqλHMH2O/MD2O\omega^{D}_{q\lambda}=\omega^{H}_{q\lambda}\sqrt{M_{H_2O}/M_{D_2O}}). Then, in the case of H2_2O, TmHT^{H}_{m} is obtained to be 274.6 K, for f=0.2f=0.2, mH2O=18m_{H_2O}=18 g/mole, and aH=2.76a^{H}=2.76 Å [48] in Eq. (5). Here aHa^{H} corresponds to the intermolecular O−O distance. Similarly, in the case of D2_2O, TmDT^{D}_{m} is obtained to be 280.4 K, for f=0.2f=0.2, MD2O=20M_{D_2O}=20 g/mole, and aD=2.80a^{D}=2.80 Å [48] in Eq. (5). Therefore, the higher TmDT^{D}_{m} of D2_2O than TmHT^{H}_{m} of H2_2O is elucidated well, and the intermolecular bond-stretching vibration modes are suggested to play an important role in the melting of ice-Ih.

As displayed in Figure 1(d), water also exhibits the abnormal VIE, namely, larger volume for D2_2O water than for H2_2O water. As mentioned above, near TMDT_{MD}, substantial amounts of H:O-bonds in water remain in the ice-like clusters, which are embedded in and in equilibrium with the unbonded water molecules [42]. This implies that the nuclear QM effects, the hallmark in ice-Ih phase, are retained even in liquid phases above the melting temperature. Then one can understand the abnormal VIE in water, by employing again the UZPUZP-induced NTE-anharmonic intermolecular potential energy, similarly as discussed in Figure 2(d) for ice-Ih. That is, UeffU_{eff}’s of water are quite similar in shape to those of ice-Ih, so that H2_2O having a larger contribution from the anharmonic NTE potential would have a smaller equilibrium volume than D2_2O having a smaller contribution from the anharmonic NTE potential.

Further, due to existing nuclear QM effects in liquid phases, D2_2O and H2_2O waters would have different H:O bonding energies, EHBWE^{W}_{HB}’s. Figure 4 summarizes the energy configurations and thermal expansion behaviors of D2_2O and H2_2O waters and ice-Ih’s. Figure 4(a), constructed based on the data in the literature, provides the following informations: (i) D2_2O has larger volume than H2_2O for both ice-Ih and water, (ii) waters are located at a higher energy (by ΔE∼60\Delta E\sim60 meV) [19, 42] and at a smaller volume than those of ice-Ih, (iii) D2_2O and H2_2O have different H:O bonding energies, ΔEHBH∼15\Delta E^{H}_{HB}\sim15 meV for ice-Ih [39] and ΔEHBW∼10\Delta E^{W}_{HB}\sim10 meV for water [49].

In Figure 4(b), TT-dependent thermal expansion behaviors expected from the energy configurations given in Figure 4(a) are provided. It is shown that D2_2O possesses the larger volume for both ice-Ih and water phases and the higher TmT_m than for H2_2O.

Energy configurations and thermal expansion behaviors of H2O and D2O waters and ice-Ih

Figure 4. (a) Energy configurations of H2_2O and D2_2O waters and ice-Ih’s. The nuclear QM effects are still retained in the liquid phase near TmT_m, and so D2_2O water has the larger H:O-bonding energy EHBWE^{W}_{HB} and the larger volume than H2_2O water, as in ice-Ih phase. The value of ΔE\Delta E is taken from Refs. [19, 42], and that of EHBHE^{H}_{HB} is taken from Ref. [49]. (b) Thermal expansion behaviors of H2_2O and D2_2O ice-Ih’s and waters expected from the energy configurations of (a).

Concluding Remarks

Prominent isotope effects observed in ice and water indicate that the nuclear QM nature of light H and D ions plays an important role in their thermal physics. Indeed, by employing a type of Born-Oppenheimer approximation for ice-Ih, which has very high-energy phonon modes arising from intramolecular H-O and D-O bond stretching vibrations, we are able to capture a conceptual and intuitive picture of their intriguing NTE and VIE phenomena in the same framework. We demonstrate that the abnormal VIE phenomena in ice and water are attributed to the NTE-anharmonic UeffU_{eff} driven by the QM zero-point energy of the high-energy intramolecular stretching phonon modes with negative γ\gamma. Furthermore, it is found that isotope effects on TvmT_{vm} and TmT_m of ice, as well as TMDT_{MD} of water, originate from the enhanced nuclear QM fluctuations in H2_2O compared to D2_2O.

These findings reveal that the interplay between zero-point energy phonons, thermal phonons, and the H:O-bondings, all of which are of QM origin, governs the unique NTE, VIE, and melting behaviors of ice and water. A comprehensive and conceptual understanding of these mechanisms would provide valuable insights into not only the NTE but also other hitherto-mysterious thermal properties observed in ice and water, including specific-heat, viscosity, compressibility, and sound velocity.

Acknowledgments

Helpful discussions with Prof. Kwang-Sei Lee and Prof. Byeong June Min are greatly appreciated. JSK acknowledges the support by the National Research Foundation (NRF) of Korea (Grant No. RS-2023-00275779).

Negative Thermal Expansion (NTE) when dPdT<0\frac{dP}{dT}<0

The negative slope (dPdT<0)\left(\frac{dP}{dT}<0\right) of phase-equilibrium line at the ice-water transition in Figure 1(a) of the main text suggests the NTE behavior across the ice-water transition upon heating. It is because the thermal-expansion coefficient αV\alpha_V is given by

αV=1V(∂V∂T)P=1B(∂P∂T)V,(6)\alpha_V=\frac{1}{V}\left(\frac{\partial V}{\partial T}\right)_P=\frac{1}{B}\left(\frac{\partial P}{\partial T}\right)_V, \tag*{(6)}

where BB is the bulk-modulus, B=−V(∂P∂V)TB=-V\left(\frac{\partial P}{\partial V}\right)_T, which is always positive [27]. Here we used the thermodynamic relation, (∂V∂T)P=−(∂P/∂T)V(∂P/∂V)T\left(\frac{\partial V}{\partial T}\right)_P=-\frac{\left(\partial P/\partial T\right)_V}{\left(\partial P/\partial V\right)_T}. Accordingly, αV\alpha_V can be negative, provided dPdT<0\frac{dP}{dT}<0 in the pressure-temperature (PTPT) phase diagram.

Adiabatic separation of the high- and low-energy phonon modes

Let us consider a system having the high-energy and low-energy phonon modes, which are far-separated in energy. Then, as done in the Born-Oppenheimer approximation for the fast electron and slow ion motions [27], one can separate adiabatically the high-energy (fast qq) and low-energy (slow QQ) phonon modes.

Consider the total Hamiltonian for qq (with mass mm) and QQ (with mass MM) ion motions:

H(q,Q)Ψn(q,Q)=En(q,Q)Ψn(q,Q).(7)H(q,Q)\Psi_n(q,Q)=E_n(q,Q)\Psi_n(q,Q). \tag*{(7)}

where

H(q,Q)=p22m+P22M+U(q,Q).(8)H(q,Q)=\frac{p^2}{2m}+\frac{P^2}{2M}+U(q,Q). \tag*{(8)}

Here U(q,Q)U(q,Q) corresponds to the static inter-ionic potential obtained by using the Born-Oppenheimer approximation. In the harmonic approximation for U(q,Q)U(q,Q),

H(q,Q)=p22m+12mω(Q)2(q−q0)2+P22M+12MΩ2(Q−Q0)2.(9)\begin{aligned} H(q,Q)&=\frac{p^2}{2m}+\frac{1}{2}m\omega(Q)^2(q-q_0)^2\\ &\quad+\frac{P^2}{2M}+\frac{1}{2}M\Omega^2(Q-Q_0)^2. \tag*{(9)} \end{aligned}

Now consider the adiabatic separation of fast (qq) and slow (QQ) ionic motions,

Ψn(q,Q)=χn(Q)ϕn(q,Q).(10)\Psi_n(q,Q)=\chi_n(Q)\phi_n(q,Q). \tag*{(10)}

Here ϕn(q,Q)\phi_n(q,Q) is the wave function of high-energy phonon (fast qq) mode, satisfying

[p22m+12mω(Q)2(q−q0)2]ϕn(q,Q)=(n+12)ℏω(Q)ϕn(q,Q).(11)\left[\frac{p^2}{2m}+\frac{1}{2}m\omega(Q)^2(q-q_0)^2\right]\phi_n(q,Q)=\left(n+\frac{1}{2}\right)\hbar\omega(Q)\phi_n(q,Q). \tag*{(11)}

Here QQ is considered just as a given parameter. Substituting Eq. (10) and (S6) into Eq. (7), and using the adiabatic approximation [24, 25], then we have the following equation for the low-energy phonon mode,

[P22M+12MΩ2(Q−Q0)2+12ℏω(Q)]χ0N(Q)=E0Nχ0N(Q).(12)\left[\frac{P^2}{2M}+\frac{1}{2}M\Omega^2(Q-Q_0)^2+\frac{1}{2}\hbar\omega(Q)\right]\chi_0^N(Q)=E_0^N\chi_0^N(Q). \tag*{(12)}

In Eq. (12), we considered the n=0n=0 ground-state eigenvalue of qq phonon mode, and so the effective potential for QQ phonons is given by

Ueff(Q)=12MΩ2(Q−Q0)2+12ℏω(Q).(13)U_{\mathrm{eff}}(Q)=\frac{1}{2}M\Omega^2(Q-Q_0)^2+\frac{1}{2}\hbar\omega(Q). \tag*{(13)}

Therefore, in this adiabatic approximation, the zero-point phonon energy of fast motion of qq mode (UZP≡12ℏω(Q)U_{\mathrm{ZP}}\equiv\frac{1}{2}\hbar\omega(Q)) plays a role of additional potential energy for the slow motion of QQ modes. In the case of H2_2O ice and water, the intramolecular O-H stretching mode corresponds to the fast qq mode, while intermolecular vibrations correspond to the slow QQ mode.

Thermal expansion coefficient of phonon origin

In the quasi-harmonic approximation, pressure exerted by phonons can be obtained from the volume-dependent free energy of phonons, Fph(V)F_{\mathrm{ph}}(V), by using P=−(∂Fph(V)∂V)TP=-\left(\frac{\partial F_{\mathrm{ph}}(V)}{\partial V}\right)_T [27 S4, 9]:

Fph(V)=∑k12ℏωk(V)+kBT∑kln⁡[1−exp⁡(−ℏωk(V)kBT)].(14)F_{\mathrm{ph}}(V)=\sum_k\frac{1}{2}\hbar\omega_k(V)+k_{\mathrm{B}}T\sum_k\ln\left[1-\exp\left(-\frac{\hbar\omega_k(V)}{k_{\mathrm{B}}T}\right)\right]. \tag*{(14)}

The first term corresponds to the energy of the quantum mechanical zero-point vibrations of the normal modes. The second term corresponds to the thermal phonon energy at T≠0T\ne0. Then

P=−∂∂V(∑k12ℏωk(V))+∑k(−∂∂Vℏωk(V))nk.(15)P=-\frac{\partial}{\partial V}\left(\sum_k\frac{1}{2}\hbar\omega_k(V)\right)+\sum_k\left(-\frac{\partial}{\partial V}\hbar\omega_k(V)\right)n_k. \tag*{(15)}

where nk=1eβℏωk−1n_k=\frac{1}{e^{\beta\hbar\omega_k}-1} is the boson distribution function [27]. Substituting Eq. (15) into Eq. (6), the thermal-expansion coefficient αV\alpha_V of phonon origin is given by

αV=1B∑k(−∂∂Vℏωk)∂nk∂T.(16)\alpha_V=\frac{1}{B}\sum_k\left(-\frac{\partial}{\partial V}\hbar\omega_k\right)\frac{\partial n_k}{\partial T}. \tag*{(16)}

Now let us define the kk-dependent Grüneisen parameter γk\gamma_k and the kk-dependent specific heat cv(k)c_v(k) contributed by a phonon mode with the energy ℏω(k)\hbar\omega(k) (total specific heat cv=∑kcv(k)c_v=\sum_k c_v(k)), as follows:

γk=−Vωk∂ωk∂V,(17)\gamma_k=-\frac{V}{\omega_k}\frac{\partial\omega_k}{\partial V}, \tag*{(17)}
cv(k)=1Vℏωk∂nk∂T.(18)c_v(k)=\frac{1}{V}\hbar\omega_k\frac{\partial n_k}{\partial T}. \tag*{(18)}

Let us also define the overall Grüneisen parameter γ\gamma corresponding to the weighted average of γk\gamma_k,

γ≡∑kγkcv(k)∑kcv(k).(19)\gamma\equiv\frac{\sum_k \gamma_k c_v(k)}{\sum_k c_v(k)}. \tag*{(19)}

Then αV\alpha_V can be expressed as

αV=1V(∂V∂T)P=γBcv.(20)\alpha_V = \frac{1}{V}\left(\frac{\partial V}{\partial T}\right)_P = \frac{\gamma}{B}c_v. \tag*{(20)}

Thus, whether a system is the NTE or PTE is determined by the sign of γ\gamma.

Volume change and the Grüneisen parameter

From Eq. (20),

dVV=γBcvdT,(21)\frac{dV}{V} = \frac{\gamma}{B}c_vdT, \tag*{(21)}
∫V0VdVV=ln⁡VV0=1B∫γcvdT,(22)\int_{V_0}^{V}\frac{dV}{V} = \ln\frac{V}{V_0} = \frac{1}{B}\int\gamma c_vdT, \tag*{(22)}

and using Eq. (19),

∫γcvdT=∫[∑kγkcv(k)]dT=∫[∑kγk1V0ℏωk∂nk∂T]dT.(23)\begin{aligned} \int\gamma c_vdT = \int\left[\sum_k \gamma_k c_v(k)\right]dT \\ &= \int\left[\sum_k \gamma_k\frac{1}{V_0}\hbar\omega_k\frac{\partial n_k}{\partial T}\right]dT. \tag*{(23)} \end{aligned}

Note that, for V=V0+ΔVV=V_0+\Delta V,

ln⁡VV0=ln⁡(1+ΔVV0)≈ΔVV0.(24)\ln\frac{V}{V_0} = \ln\left(1+\frac{\Delta V}{V_0}\right) \approx\frac{\Delta V}{V_0}. \tag*{(24)}

Accordingly, from Eqs. (22), (23), and (24),

ΔVV0=1B∑kγk1V0ℏωknk.(25)\frac{\Delta V}{V_0} = \frac{1}{B}\sum_k \gamma_k\frac{1}{V_0}\hbar\omega_k n_k. \tag*{(25)}

Phonon frequency and the Grüneisen parameter

On the other hand, from Eq. (17),

γkln⁡VV0=−ln⁡ωk(V)ωk(V0),(26)\gamma_k\ln\frac{V}{V_0} = -\ln\frac{\omega_k(V)}{\omega_k(V_0)}, \tag*{(26)}
ωk(V)=ωk(V0)(VV0)−γk.(27)\omega_k(V) = \omega_k(V_0)\left(\frac{V}{V_0}\right)^{-\gamma_k}. \tag*{(27)}

Eq. (27) indicates that, for the NTE with negative γk\gamma_k, phonon frequency increases with volume (pressure-induced softening), while, for the PTE with positive γk\gamma_k, phonon frequency decreases with volume (pressure-induced hardening) (see Ref. [9]).

Average amplitude of thermally vibrating atoms: ⟨u2⟩1/2\langle u^2\rangle^{1/2}

For a single harmonic oscillator in one dimension, the atomic displacement uiu_i can be described by the quantized creation and annihilation operators, aa and a†a^\dagger, as follows:

ui=ℏ2Miω(a+a†).(28)u_i = \sqrt{\frac{\hbar}{2M_i\omega}}(a+a^\dagger). \tag*{(28)}

Then

⟨ui2⟩=ℏ2Miω⟨(a+a†)(a+a†)⟩=ℏ2Miω⟨aa†+a†a⟩=ℏ2Miω(⟨2a†a⟩+1)=ℏMiω(n+1/2).(29)\begin{aligned} \langle u_i^2\rangle&= \frac{\hbar}{2M_i\omega}\langle(a+a^\dagger)(a+a^\dagger)\rangle\\ &= \frac{\hbar}{2M_i\omega}\langle aa^\dagger+a^\dagger a\rangle\\ &= \frac{\hbar}{2M_i\omega}(\langle2a^\dagger a\rangle+1) \\ &= \frac{\hbar}{M_i\omega}(n+1/2). \tag*{(29)} \end{aligned}

where nn (≡⟨a†a⟩\equiv\langle a^\dagger a\rangle) corresponds to the boson distribution function n=1eβℏω−1n=\frac{1}{e^{\beta\hbar\omega}-1}. At T=0T=0, n=0n=0, and so ⟨u2⟩=ℏ2Miω\langle u^2\rangle=\frac{\hbar}{2M_i\omega}. At finite TT, ⟨u2⟩\langle u^2\rangle increases with TT as ⟨u2⟩=ℏ2Miωcoth⁡βℏω2\langle u^2\rangle=\frac{\hbar}{2M_i\omega}\coth\frac{\beta\hbar\omega}{2}. In three dimensions, factor 3 needs to be multiplied in the above expressions [S6]. Note that ⟨u2⟩\langle u^2\rangle is isotopic-mass dependent through MiM_i and ω\omega. But as T→∞T\to\infty, ⟨u2⟩\langle u^2\rangle becomes kBTMiω2\frac{k_BT}{M_i\omega^2}, and so it does not have isotopic-mass dependence because of Miω2=KM_i\omega^2=K being constant.

For a real solid, the atomic displacement uiu_i can be described, as follows [S7]:

ui=i∑k,λ(ℏ2NMiωkλ)1/2ϵkλ(akλ+a−kλ†)eik⋅Ri,(30)u_i=i\sum_{k,\lambda}\left(\frac{\hbar}{2NM_i\omega_{k\lambda}}\right)^{1/2}\epsilon_{k\lambda}(a_{k\lambda}+a^\dagger_{-k\lambda})e^{ik\cdot R_i}, \tag*{(30)}

where ωkλ\omega_{k\lambda} and ϵkλ\epsilon_{k\lambda} are frequency and polarization vector of (k,λ)(k,\lambda) phonon, respectively. Then ⟨ui2⟩\langle u_i^2\rangle is given by

⟨ui2⟩=∑k,λℏMiωkλ(nkλ+1/2)=∑k,λℏ2Miωkλcoth⁡βℏωkλ2.(31)\begin{aligned} \langle u_i^2\rangle=\sum_{k,\lambda}\frac{\hbar}{M_i\omega_{k\lambda}}(n_{k\lambda}+1/2) \\ &=\sum_{k,\lambda}\frac{\hbar}{2M_i\omega_{k\lambda}}\coth\frac{\beta\hbar\omega_{k\lambda}}{2}. \tag*{(31)} \end{aligned}

which is quite similar to Eq. (29) (here, nk,λ=1eβℏωkλ−1n_{k,\lambda}=\frac{1}{e^{\beta\hbar\omega_{k\lambda}}-1}). Therefore, even in the case of real solid, ⟨ui2⟩\langle u_i^2\rangle for a specific phonon mode ωkλ\omega_{k\lambda} of Einstein type can be described by a simple form of the single harmonic oscillator ℏ2Miωkλcoth⁡βℏωkλ2\frac{\hbar}{2M_i\omega_{k\lambda}}\coth\frac{\beta\hbar\omega_{k\lambda}}{2}. (1968).

[25] Y. Marechal, The Hydrogen Bond and the Water Molecule, Amsterdam, Elsevier (2007).

[S4] D. J. Kim, T. Kizaki, N. Miyai, J. Iwai, N. Hino, S. Fukumoto, J. Korean Phys. Soc. 28, 203 (1995).

[9] B. Pamuk, J. M. Soler, R. Ramirez, C. P. Herrero, P.W. Stephens, P. B. Allen, and M.-V. Fernandez-Serra, Phys. Rev. Lett. 108, 193003 (2012)

[S6] C. Kittel, Introduction to Solid State Physics, seventh edition, John Wiley & Sons, Inc. (1996).

[S7] G. D. Mahan, Many Particle Physics, Plenum, New York (1990).

References

  1. [1]Introductory Chemistry by LibreTexts project (CK-12), website (see Ch. 13.20), (https://chem.libretexts.org/Bookshelves/Introductory_Chemistry/Introductory_Chemistry_(CK-12)).
  2. [2]Water Structure and Science, website by M. Chaplin, (https://water.lsbu.ac.uk/water/water-structure-science.html), and references therein.
  3. [3]Z.-K. Liu, Y. Wang, and S.-L. Shang, Scr. Mater. 65, 664 (2011).
  4. [4]Z.-K. Liu, Y. Wang, and S.-L. Shang, Sci. Rep. 4, 7043 (2014).
  5. [5]J. Kim, J.-S. Kang, C.-J. Kang, and B. I. Min, Phys. Rev. B 107, 115157 (2023).
  6. [6]See the Supplementary Material for the derivations of (i) The NTE for the negative slope of dP/dT < 0, (ii) Adiabatic separation of the high- and low-energy phonon modes, (iii) Thermal expansion coefficient of phonon origin, (iv) The relation between the volume change and the Grüneisen parameter, (v) The relation between the phonon frequency and the Grüneisen parameter, and (vi) Average amplitude of thermally vibrating atoms: ⟨u²⟩¹ᐟ².
  7. [7]J. S. O. Evans, J. Chem. Soc., Dalton Trans., 3317 (1999).
  8. [8]Th. Strässle, A. M. Saitta, S. Klotz, and M. Braden, Phys. Rev. Lett. 93, 225901 (2004).
  9. [9]B. Pamuk, J. M. Soler, R. Ramirez, C. P. Herrero, P.W. Stephens, P. B. Allen, and M.-V. Fernandez-Serra, Phys. Rev. Lett. 108, 193003 (2012).
  10. [10]K. Umemoto, E. Sugimura, S. de Gironcoli, Y. Nakajima, K. Hirose, Y. Ohishi, and R. M. Wentzcovitch, Phys. Rev. Lett. 115, 173005 (2015).
  11. [11]M. K. Gupta, R. Mittal, Baltej Singh, S. K. Mishra, D. T. Adroja, A. D. Fortes, and S. L. Chaplot, Phys. Rev. B 98, 104301 (2018)
  12. [12]K. Röttger, A. Endriss, J. Ihringer, S. Doyle, and W. F. Kuhs, Acta Cryst. 850, 644 (1994).
  13. [13]A. D. Fortes, Acta Cryst. B 74, 196 (2018).
  14. [14]D. T. W. Buckingham, J. J. Neumeier, S. H. Masunaga, and Y.-K. Yu, Phys. Rev. Lett. 121, 185505 (2018).
  15. [15]M. A. Salim, S. Y. Willow, and S. Hirata, J. Chem. Phys. 144, 204503 (2016).
  16. [16]M. Cherubini, L. Monacelli, and F. Mauri, J. Chem. Phys. 155, 184502 (2021).
  17. [17]S. Rasti, E. Ö. Jonsson, H. Jonsson, and J. Meyer, J. Phys. Chem. Lett. 13, 11831 (2022).
  18. [18]J. A. Morrone and R. Car, Phys. Rev. Lett 101, 017801 (2008).
  19. [19]R. Ramirez and C. P. Herrero, J. Chem. Phys. 133, 144511 (2010).
  20. [20]C. McBride, J. L. Aragones, E. G. Noya, and C. Vega, Phys. Chem. Chem. Phys. 4, 15199 (2012).
  21. [21]J. Xu, C. Zhang, L. Zhang, M. Chen, B. Santra, and X. Wu, Phys. Rev. B 102, 214113 (2020).
  22. [22]A. E. Altereb, G. E. Lopez, and N. Giombattista, Phys. Chem. Chem. Phys. 234, 6914 (2021).
  23. [23]G. Romanelli, C. Andreani, A. Bocedi, R. Senesi, J. Chem. Phys. 160, 234503 (2024).
  24. [24]Y. Marechal and A. Witkowski, J. Chem. Phys. 48, 3697 (1968).
  25. [25]Y. Marechal, The Hydrogen Bond and the Water Molecule, Amsterdam, Elsevier (2007).DOI
  26. [26]F. A. Lindemann, Phys. Z. 11, 609 (1910).
  27. [27]N. W. Ashcroft and N. D. Mermin, Solid state physics, New York: Saunders College Publishing (1976).
  28. [28]G. de With, Chem. Rev. 123, 13713 (2023).
  29. [29]website The Engineering ToolBox (https://www.engineeringtoolbox.com/).
  30. [30]The melting temperatures (Tₘ’s) for H₂O, D₂O, and T₂O are 0 °C, 3.68 °C, and 4.49 °C, respectively. The volume-minimum points corresponding to the density maximum points (Tᴹᴰ’s) are 4.0 °C, 11.19 °C, and 13.40 °C for H₂O, D₂O, and T₂O, respectively. Molecular volumes at Tᴹᴰ’s are 29.9 ų, 30.1 ų, and 31.1 ų, respectively.
  31. [31]E. D. Isaacs, A. Shukla, P. M. Platzman, D. R. Hamann, B. Barbiellini, and C. A. Tulk, Phys. Rev. Lett. 82, 600 (1999).
  32. [32]T. Soma, J. Phys. Soc. Jpn. 42, 1491 (1977).
  33. [33]J. C. Noya, C. P. Herrero, and R. Ramírez, Phys. Rev. B 56, 237 (1997).
  34. [34]C.P. Herrero, Solid State Comm. 110, 243 (1999).
  35. [35]Y. Yoshimura, S. T. Stewart, M. Somayazulu, H. K. Mao, and R. J. Hemley, J. Phys. Chem. B 115, 3756 (2011).
  36. [36]C. Q. Sun, X. Zhang, X. Fu, W. Zheng, J.-I. Kuo, Y. Zhou, and J. Zhou, J. Phys. Chem. Lett. 4, 3238 (2013).
  37. [37]X. Zhang, Y. Huang, P. Sun, X. Liu, Z. Ma, Y. Zhou, J. Zhou, W. Zheng, and C. Q. Sun, Sci. Rep. 5, 13655 (2015).
  38. [38]The other vibrational modes located inbetween the high and low energy phonon modes, which correspond to the H₂O libration and H-O:H bending modes, are irrelevant to the NTE phenomenon in ice-Ih, and so are ignored in the present Born-Oppenheimer analysis.
  39. [39]B. Pamuk, P. B. Allen, and M.-V. Fernandez-Serra, Phys. Rev. B 92, 134105 (2015).
  40. [40]It is because the slope of ωₛ(V) with respect to volume, ∂ωₛ/∂V, is stiffer for H₂O than for D₂O, i.e., (∂ωₛᴴ/∂V − ∂ωₛᴰ/∂V) ≈ ∂ωₛᴴ/∂V (1 − √mᴴ/mᴰ) > 0.
  41. [41]If the intrinsic PTE-anharmonicity of U₀(V) at low T is dominant over the NTE-anharmonicity of the QM zero-point energy (½ℏωₛ(V)) of high-energy phonon mode, the NTE phenomenon will not emerge. Other ice phases under high pressure are likely to be in this situation, which is why no NTE phenomena are observed at low T.
  42. [42]G. Nemethy and H. A. Scheraga, J. Chem. Phys. 36, 3382 (1962).
  43. [43]F.A. Deeney and J.P. O’Leary, Phys. Lett. A 372, 1551 (2008).
  44. [44]S. Habershon, T. E. Markland, D. E. Manolopoulos, J. Chem. Phys. 131, 024501 (2009).
  45. [45]M. Ceriotti, W. Fang, P. G. Kusalik, R. H. McKenzie, A. Michaelides, M. A. Morales, and T. E. Markland, Chem. Rev. 116, 7529 (2016).
  46. [46]E. Whalley, Trans. Faraday Soc. 53, 1578 (1957).
  47. [47]E. Whalley, J. Chem. Phys. 81, 1 (1984).
  48. [48]G. Giubertoni, M. Bonn, and S. Woutersen, J. Phys. Chem. B 127, 8086 (2023).
  49. [49]I. F. Hakem, A. Boussaid, H. Benchouk-Taleb, M. R. Bockstaller, J. Chem. Phys. 127, 224106 (2007).

Paper details

vlnc:W_phxzxfjg
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