J. Phys. Chem. 1993,97, 63396341
6339
Water II is a "Strong"Liquid C. A. Angell Department of Chemistry, Arizona State University, Tempe, Arizona 85287- 1604 Received: March 8, 1993; In Final Form: May 3, 1993
Rationalization of anomalous devitrification relationships bctween hyperquenched vitreous water (and related ASW forms) and vitrified aqueous solutions requires the assumption that the viscosity-temperature relationship of water near but above its glass transition temperature is that of a "strong" liquid, like Si02. This would support Speedy's claim that water I1 (ASW above Tg)is a separate and distinct state of liquid water. The conclusion is supported by analysis of low-temperature data on water-swollen polymers (hydrogels).
One of the remarkable developmentsin the rich phenomenology of condensed H20 has been the recent discovery that amorphous solid water produced by such diverse routes as vapor deposition,' liquid hyperquen~hing,~.~ and most recently low-temperature pressure-induced amorphization4 ends up (after appropriate normal pressure annealing procedures) in the same physical ~ t a t e . ~This , ~ phase evidently undergoes a very weak glass transition at 136 KS4 on warming and then persists as a supercooled liquid until 150-165 K (depending on method of preparation) before it devitrifies to ice I,. Speedy9recently offered a persuasivethermodynamicargument that this material, in the range 136-150 K, is a distinct phase of liquid water which at 1 atm pressure cannot be connected to normal water by any reversible thermodynamic path. In consequence he designated this phase "water 11". The purpose of this Letter is to argue that water I1 is a uniquely interesting liquid in that it is the only molecular example of a "strong" liquid in the "strong/fragile" classification scheme.1° Furthermore, it must be formed during fast cooling by what could be described as a "fragile liquid-to-strong liquid" transition. That it is a strong liquid is the only way we can see of understanding the otherwise extraordinary observation that, despite the exotic routes required for its preparation, ASW is more stable against crystallization (in terms of the temperature range above T, traversed before crystallization occurs) than are a range of LiCl HzO solutions that are easily made by simple liquid nitrogen cooling procedures.lOJ1Tomake thecase, wemust first examinewhat isknown about other hyperquenched systems. Much is known about glasses formed by hyperquenching from the study of metallic glasses, most of which can only be formed in thismanner.l2-l4 Atypicalquenchingrateusedin theformation of metallic glasses is lo7 K/s. This is comparable with that estimated for the cooling rate of the poorer-conducting HzO droplets vitrified by the aerosol quenching procedure utilized by Mayer and colleag~es.~.~ The majority of the metallic glasses have been thought to crystallize so readily on reheating that the glass transition is never reached. In cases of exceptional stability, such as A u - S i 4 1 , ~Pd-Si-Cu,14 ~ Pd-Ni-P,15 and Fe40Ni40pl&lo16 the entire glass transition can be observed, and viscosity measurements have then been made above T,in the range 109P.13J4J7 At 1OloPthe (relatively slow) crystallizationkinetics are usually sufficient to prevent any further measurements. On the basis of the above information, we would expect that water above its T, would be limited to a temperature range in which its viscosity is greater than 10'0 P. With this in mind, we reexamine the meaning of the observation reported some time ago'* that ASW (water 11), despite a lower glass transition temperature than LiCl-HZO solutions prepared by normal liquid nitrogen cooling procedures, survives to a higher temperature before crystallization terminates its existence. The data are presented in Figure 1.
+
0022-3654/93/2097-6339$04.00/0
I
T = 160K
III
\
C
T/K
Figure 1. Comparison of DSC warm-up behaviors of vapor-deposited ASW and a LiCl.llH20 (8.33 mol 46 LICl) aqueous solution (from ref 17). Warm-up behavior of hyperquenchedliquid water is essentially the same as that of the vapor deposit. Arrow marks position of feeble glass transition detected by Johari and colleague^.^^ vlscoslty
n 8
7 8
0
Si02
t
5 4
3 2: 1-
01: 237 -4-1---.00
, 0 2
,
,
.
04
, 06
.
,
.
08
I1 0
Tg/T
Figure 2. T,-scaled Arrhenius plots ofviecosity data for water and aqueous
LiCl solutions and other selected liquids. The arrows show temperatures of crystallization on reheating of glasses (see Footnote 1) and the Mered viscosity behavior for ASW above its glass transition temperature (water 11). The viscosity behavior of supercooled water below -35 OC according to the power law fits (T,= 225 K) of availabledata is indicated by dashed line. The dashed curve shows a plot of eq 1 for the case of D = 30, with which the findings from the water-in-hydrogel studies an be compared.
To assess the approximate viscosity of LiCl-HZO solutions at their temperature of crystallization, we present a selection of extensive but mostly unpublished data on LiCl H z 0 solutions19 on a scaled Arrhenius plot, using the calorimetric glass transition temperatures as scaling parameter; see Figure 2. Included in the plot are data covering a wider viscosity range for some concentrated solutions reported by Moynihan et al.zo Also included in the plot are a selection of data for other liquids of differing fragility1° so that the position of the LiCl-HZO solutions in the overall strong-to-fragile pattern can be seen.
+
0 1993 American Chemical Society
Letters
6340 The Journal of Physical Chemistry, Vol. 97, No. 24, 1993 An arrow at T,/T = 0.91 indicates the temperature 152 K at which the LiCl-llH20 (8.33% LiCl) solution crystallizes on reheating (Le., where it devitrifies). From the viscosity data of the solutions, which all seem of comparable fragility, it would appear that, at this temperature, the viscosity of the LiCl.1 lHzO solution depicted in Figure 1 is about lo6 P, which is typical of aqueous solutions at the ice-saturated limit of their glass-forming ranges for normal cooling procedures. For water I1 to crystallize at t = 1OloP a s expected, at the observed temperature of 160 K, Tg/T = 0.85, it isclear that itsviscosity-temperature relationship must be quite different from that of the LiCl solutions. The viscosity-temperature relation which permits viscosity to be 1Olo P at 160K (T,/ T = 0.85) has been added to Figure 2 and marked “water II?”. It can be seen that the required behavior places water I1 near the extreme of strong liquid behavior.21 Strong liquid behavior for the viscosity of water near T, would be consistentlo with the very weak thermal manifestation of the glass transition which is observed.”* Indeed, both thermal and viscosimetric features seem very reasonable in view of the interpretation given the strong/fragile liquid pattern (Figure 2) in terms of extent of intermediate range order and resistance to thermally induced structural degradation.l0 ASW, after all, has been shownZ2to have an almost perfect random tetrahedral network with a structure that is even more open than that of vitreous SiO2. Nevertheless, it is provocative in view of the known behavior of normal and supercooled water which occupies a fragile/intermediate position in Figure 2 (based on the arguable use of a 136 K scaling temperature, see below). Evidently, there would be some sort of structural transition occurring in fastcooling water-a transition which may involve passage through unstable states, such that it is intrinsically irrever~ible.~It is tempting to relate this to the LDA-HDA transition in computersimulated water, both ST2 and TIP4P. recently described by Poole et al.,233” though these latter studies would indicate that the transition should be continuous if crystallization is avoided. To find support for our conclusions, we turn to observations of Johari and colleague^^^,^^ on water in hydrogels. These latter are polymeric materials in which the presence of strongly hydrophilic groups on the chain drives the polymer to absorb large weight fractions (up to 42 wt %) of water. Hofer et al.25 showed, in the same manner as they used to reveal the glass transition in ASW (water 11),5,6that in the most water-rich hydrogels a similarly feeble glass transition occurs. Furthermore, they obtained structural relaxation data both by the variable scan speed DSC technique27and by standard dielectric relaxation spectroscopy, from which we can estimate the fragility m defined below and thence the strength parameter D of the familiar VogelTammann-Fulcher (VTF) relaxation equation, 7 = A exp[DTo/(T-To)] (1) With this information we can compare hydrogel water with the other systems in Figure 2. To obtain m and D from the activation energy found from the DSC we can use the relationship~~~,~~
m = AHh/2.3Q3RT, (2) (where A H h is the apparent Arrhenius activation energy for enthalpy relaxation) and
D = 589/(m-17) (3) To obtain D from the cited Arrhenius law preexponent (fo = ( 2 ~ 7 0 )= - ~1.06 X loz1s-l) of the dielectric relaxation plot,26 we use the additional relation28 m = log T,/log
T~
(4)
This yields, for the enthalpy relaxation, m = 39 and D = 26, while for the more accurate and reliable dielectric relaxation we obtain m = 23.8 and D = 86. The latter in particular is a
remarkably large value for the strength parameter and implies that H-bonded water near 136 K is far stronger than any other molecular liquid and indeed is nearly as strong as silica, Si02; see Figure 2. Since the water structure in the hydrogel must be at least partially perturbed by its interaction with the polymer support, such a value can be taken as strongly supportive of the earlier conclusion on the high strength of the pure water 11. According to a theoretical treatment by Vilgi~,~O a large D parameter is to be interpreted as a consequence of a very small mean square fluctuation in the coordination number for the particles making up the liquid and thus would be consistent with the existence of a well-formed tetrahedral network. It is also consistent with expectations from the simple “bond lattice” or “independentbond”mode131*32for thecasewherethereisnochange in degeneracy on bond rupture-the condition which also minimizes the heat capacity increase. (A modified version of this model has recently been applied to ASW by J~hari.’~)Thus, it would seem that water near its T, indeed has a very different character from ordinary water and is deserving of Speedy’s designation, water 11. However, if water I1 is, as Speedy argues, completely distinct from water I and has no point of thermodynamic contact with it (i.e., their normal pressure free energy surfaces never touch), then our representation of normal water in Figure 2 could be criticized. It could be argued that, as a wholly distinct phase which would vitrify by viscosity divergence at 228 K if crystallization did not intervene, water I should be represented by a quite separate curve scaled by a T8of 228 K. We have not done this because of the appeal of the continuous fragileto-strong liquid transition suggested by Figure 2. If however we do plot water I scaled separately, then we observe that this phase of water would also be unique among molecular liquids, lying at the opposite extreme of behavior from water 11. Most of the 12 orders of magnitude change in viscosity between observed values to -35 OC and the value of T, would then occur in a very few kelvin near T8; Le., the liquid would behave like a classic spin glass and exhibit a pseudo-first-order transition from fluid to glass. We note that something akin to this happens during the cooling of many hydrated denatured protein molecules as they renature, by a cooperativeH-bonding process, to give a “frozen” tertiary structure involving a significant fraction of the system’s configurationaldegrees of freedom;34the analogy may be worth developing.
References and Notes (1) Burton, E. F.; Oliver, W. F. Proc. R. Soc. London 1935, A153,166. (2) (a) Brueggeller, P.; Mayer, E. Nature 1980,2888,569; 1982,298, 715. (b) Mayer, E. J. Appl. Phys. 1985,58,663; J. Phys. Chem. 1985,89, 3474. (3) Dubochet, J.; McDowell, A. W. J. Microsc. 1981, 124, RP3-RP4. (4) Mishima, D.; Calvert, L. E.; Whalley, E. Nature 1984, 310, 393395. (5) Johari, G. P.; Hallbrucker, A.; Mayer, E. Nature 1987, 330, 552553. (6) Hallbrucker, A.; Mayer, E.; Johari, G. P. Philos. Mag. B 1989,60, 179-187. (7) Johari, G. P.; Astl, G.; Mayer, E. J. Chem. Phys. 1990,92,809. (8) Mug, D. D.; Handa, Y. P. J . Phys. Chem. 1988,92, 3323. (9) Speedy, R. J. J. Phys. Chem. 1992,96,2322. (10) (a) Angell, C. A. In Relaxations in Complex Systems; Ngai, K., Wnght, G. B., Eds.;NationalTechnicalInformationService,U.S.Department of Commerce: Springfield, VA, 1985; p 1. (b) Angell, C. A. J. Non-Crysr. Solids 1991, 131-133, 13. (11) (a) Vuillard, G. Ann. Chim. 1%7,2, 233. (b) Sam, E. J.; Angell, C. A. J. Chem. Phys. 1970,52, 1058-1068. (12) Chaudhari, P.; Turnbull, D. Science 1978,199, 11-21. (13) Chen, H. S.;Turnbull, D. J. Chem.Phys. 1968,48, 2560. (14) Two, S. S.;Spaepen, F. Aero Merall. 1985, 33, 881. (15) Drehman, A. J.; Greer, A. L.; Turnbull, D. Scr. Metall. 1977, 11, 367. (16) Chen, L.-C.; Spaepen, F. Philos. Mag. B 1991,63, 585-586. (17) Taub, A. I.; Lubonky, F. E. Acta Metall. 1981, 29, 1939. (18) MacFarlane, D. R.; Angell, C. A. J . Phys. Chem. 1984, 88, 759.
Letters (19) (a) Bressel, R. D. Ph.D.Thesis,PurdueUniversity, 1972. (b) Angell, C. A.; Bressel, R. D.; Green, J. L.; Kanno, H.; Oguni, M.;Sare, E. J. Int. J . Food Sci., in press. (20) Moynihan, C. T.; Balitactac, N.; &ne, L.; Litovitz, T. A. J . Chem. Phys. 1971,55, 3013. (21) At the same time we must sound a note of caution since the whole scenario we have developed depends on the correctness of the glass transition temperatures of ASW and water in hydrogelsbasedon theannealing procedure developed by Johari and co-worker~.~*~ While these have won a considerable degree of acceptance:.23 it has not been universal, at least in application to the parallel case of metallic glasses (Ram, S.;Johari, G. P. Philos. Mag. B 1990,61,299 and Chen, L.-C.; Spaepen, F. Philos. Mag. B 1991,63, 585586). Furthermore, the full phenomenologyof enthalpic relaxation,particularly the aspect related to nonlinearity which is of special importance in this application, has not yet been worked out (Moynihan, C. T.; Crichton, S.N.; Opalka, S . M. J. Non-Cryst. Solids 1991,131-133,420; Hodge, I. M., Ibid., in press). Therefore, alternative origins of the effects interpreted as glass transitions cannot be completely excluded, though the coincidence of their temperatures with the temperatures at which the dielectric relaxation times are about 200 s (see ref 26 discussed below) certainly offers strong support for their validity. (22) Narten, A. H.; Venkatesh, C. G.; Rice, J. A. J. Chem. Phys. 1976, 64, 1106.
The Journal of Physical Chemistry, Vol. 97, No. 24, 1993 6341 (23) Poole, P. H.; Sciortino, F.; Essmann, U.; Stanley, H. E. Nurure 1992, 360, 324328. (24) Poole, P. H. Private communication. (25) Hofer, K.; Mayer, E.; Johari, J. P. J. Phys. Chem. 1990, 94,2689. (26) Pathmanathan, K.; Johari, G. P. J . Polym. Sci., Polym. Phys. 1990, 28. (27) DeBolt, M. A.; Easteal, A. J.; Macedo, P. B.; Moynihan, C. T. J. Am. Ceram. Soc. 1976,59, 16-121. (28) Bhmer, R.; Angell, C. A. Phys. Rev. B 1992,45, 10091. (29) Bhmer, R.; Ngai, K. L.; Angell, C. A.; Plazek, D. J. J. Chem. Phys., in press. (Note that the base value of m, corresponding to D- 2, is smaller, 16, if the variable is a relaxation time rather than a viscosity. The relation between m and D given in Angell, C. A.; et al. Am. Inst. Phys. Conf. Proc. 1992, 256, 3 is in error.) (30) Vilgis, T. A. Phys. Rev. B 1993, 47,2882. (31) (a) Angell, C. A. J . Phys. Chem. 1971,75,3698. (b) Angell, C. A.; Rao, K. J. J. Chem. Phys. 1972, 57, 470. (32) (a) Nadler, W.; Krausche, T. Phys. Rev. A 1991, 44, 7888. (b) Krausche, T.; Nadler, W. Z . Phys. B Condens. Matter 1992,86, 433. (33) Johari, G. P. J. Chem. Phys., in press. (34) Green, J. L.; Fan, J.; Angell, C. A. To be submitted.
