29.CA Agell.pdf

Journal of Non-Crystalline Solids 73 (1985) 1 17 1 North-Holland, Amsterdam

Section L Glass structure

SPECTROSCOPY SIMULATION AND SCA'ITERING, AND THE MEDIUM RANGE ORDER PROBLEM IN GLASS

C.A. ANGELL Department of Chemistry, Purdue University, West Lafayette, Indiana 47907, USA

Consideration is given to phenomena which we believe to be controlled by the fluctuations in the poorly understood intermediate range order in liquids and glasses. These include structural relaxation, crystal nucleation and incipient liquid-liquid phase separation. Techniques which show promise for investigating the intermediate range order are considered, and predictable or conceiva- ble developments in these techniques which may, by the year 2004 (N.J. Kreidl's 100th), greatly increase or even revolutionize our knowledge of the intermediate range order, are discussed. These include difference spectroscopy, difference scattering, and computer simulation techniques. Fi- nally, we consider possible developments in system preparation or system manipulation techniques which may lead to new insights into relations between physical properties and intermediate range order. An example of special interest is the preparation of noncommunicating microsystems of the same size as the clusters which many believe are the building blocks of the glassy state.

1. Introduction

This manuscript is devoted to a consideration of some of the problems and prospects concerning what might be called intermediate range order in glassy solids. The microscopic distance range 8-50 A or - 3-20 atomic diameters is the range within which are determined many important aspects of liquid and glass phenomenology (e.g. relaxation of stress, nucleation of crystals, pho- tochromism, incipient liquid-liquid phase separation, and possibly ultra-fast ion conduction in glasses) and yet it is this same range which is beyond the power of most structure-determining techniques to elucidate. We believe, with others, that it is here that the primary challenge in glass structure understanding currently resides, and that the problems posed are challenging enough to project, incompletely resolved, into the 21st century.

In this paper we will first give some lines to describing certain properties of glass-forming liquids which seem to depend on the structural arrangements well outside the first and second coordination shell of the most strongly coordinating cations. These are considered in subsections 1 5 below. With the importance of these regions (about which relatively little is known at the present time) established, we will then (a) review some related phenomenology in more detail for the case with which we are most familiar and then (b) discuss the possible techniques which might be used in elucidating more clearly the importance of medium range structure on material properties.

0022-3093/85/$03.30 Elsevier Science Publishers B.V. (North-Holland Physics Publishing Division)

2 C.A. A ngell / Spectroscopy simulation and scattering

1.1. Structural relaxation

Structural relaxation is of course the fundamental phenomenon underlying glass formation since it is the lengthening time scale for relaxation with decreasing temperature which leads to the glass transition and the generation of the "solid" state. The temperature dependence of the relaxation time near Tg determines the ease of glass-working as well as the danger of devitrification by growth of pre-existing nuclei. The temperature dependence of the average relaxation time as well as the detailed relaxation function seem to be closely connected with the nature of the intermediate range order. Thus, structural relaxation is the first of our "glass science problems" identified as involving the intermediate range order. We treat this matter in some detail below.

1.2. Crystal nucleation

The second phenomenon which is of vital importance and which involves distance scales of the intermediate range is the nucleation phenomenon. We believe that, except in extreme instability cases, nucleation in amorphous phases requires fluctuations extending over distances somewhat greater than those involved in structural relaxation, though, in most cases of practical significance, it is still a phenomenon which occurs within subsystem sizes of only tens of A in diameter. Techniques for probing the nature of structural fluctuations involved in nucleation, and the relation between critical nucleus structure and the structure of the thermodynamically stable phase into which it grows, are practically nonexistent at this time, though the significance of the question being addressed is obvious to all.

1.3. Incipient liquid-liquid phase separation

Another important characteristic of many glass-forming systems, which involve distance scales of the intermediate class is encountered in demixing systems. Two cases involving phase boundaries, or quasi phase-boundaries are important. One is the phase boundary existing between microscopic liquid droplets in pre-opalescent phase-separated glassy systems. This distance becomes macroscopic at the temperature and composition of critical demixing, but in the majority of cases where a two phase structure has either developed or only has a tendency to develop, the phase boundary or the preseparation fluctuation dimension is of microscopic dimensions, and their structures are of both academic and technological interest. The other, and the least understood though not the most important technologically, is the mean dimension of composition fluctuations in glasses which exhibit a tendency to phase separation, without actually achieving it. A current case in question is that of the AgI-, CuI-, and LiI-based fast ion conductors. The best conductors seem always to be those at the iodide-rich end of the glass-forming range where there are indications that the iodide is segregated in some sort of ramified

C,A. Angell / Spectroscopy simulation and scattering 3

prephase separation clusters within which the properties approach those of the pure iodide.

1.4. Density fluctuations

On a smaller distance scale, but one still classifiable as intermediate, are the configurational density fluctuations which determine the configurational part of the thermodynamic isothermal compressibility and which are a necessary feature of all vitreous systems. The magnitude of these fluctuations is known if the compressibility of the liquid at the temperature of vitrification, and the glass compressibility, are both known. In all but exceptional cases they will be of the order of a few angstroms. How closely this distance scale relates to that involved in structural relaxation is not clear at this time. Free volume theories for transport properties in which the expansion coefficient plays a vital role in the temperature dependence of transport, would suggest that density and entropy fluctuations are, in fact dominating the probability of molecular rearrangements.

1.5. Clusters

Finally, on a larger scale than the latter, both in physical size and in potential for scientific controversy, lies the cluster dimension of the paracrys- talline models of glass which are currently experiencing a resurgent tide of interest. Championed by Phillips [1] who musters a great deal of spectroscopic and scattering information in search of support for cluster models for silica glass ( - 30 ,~ crystobalite paracrystals with Si = 0 internal surfaces), models of this general type are currently being supported and rejected by roughly equal numbers of investigators.

2. Structural relaxation and the intermediate range order: "strength" and "fragility" in viscous liquids

We choose the structural relaxation problem for detailed discussion of intermediate range order effects because of its overriding importance in glass formation phenomenology, and of our own preoccupation with this subject area.

Fig. 1 shows the behavior of the structural relaxation time, as it is reflected by the shear viscosity, for four groups of liquids, one silica-based with various concentrations of modifying oxides, the second BeF2-based and modified by extra fluorides, the third is zinc chloride-based and modified by chloride ion additions, while the fourth is ZrF4-based and is a member of the heavy metal fluoride glasses currently under development for fiber optics applications. To permit more effective comparisons, the data in fig. 1 are plotted on a reduced inverse temperature scale using the temperature at which the viscosity reaches

4 C.A. Angell / Spectroscopy simulation and scattering

m

n

I

E

o

t~ o

Q.

I0

8

6

40.2 0.6 1.0 1.4 I.~3 I 2.2 T/Tcj

._. "~

a . c o B t-

B

~ 0 0 o o

~ ~

o

Tg/T

Fig. 1.Scaled viscosity data for glass-forming liquids showing range of behavior from "strong", characteristic of open tetrahedral network liquids, to "fragile" typical of ionic and molecular liquids. Except for the fluorides BeF 2 (strong) and ZrF 4 -BeF 2 based mixtures (fragile) a common point at the high T extreme is indicated. Note added in proof." New viscosity data by Pantano et al. for ZBLA extending to 1011 P suggest ZBLA should superimpose on 21~iCl 3. KC1.

1013 P as a normalizing parameter [2,3]. The common point at 1/T= 0 has been discussed elsewhere [4] and is to be understood in terms of compensating effects of bond strength on shear modulus and quasi-lattice vibration frequency in the expression ~/r~ oo = Goo%ib-

Fig. 1 shows, interestingly enough, that the two glass-forming systems currently under most active development as fiber optics materials find themselves at opposite extremes of this behavior pattern [4]. The technological consequence of this difference is that it is very much more difficult to work in

C.A. Angell / Spectroscopy simulation and scattering 5

a controlled fashion with the softened fluoride glass because of its much greater temperature dependence of relaxation time in the viscosity range where working is carried out. We believe this distinction is to be associated with the differences in stability of intermediate range order between the two systems.

The origin of the almost Arrhenius variation of the relaxation time in liquid SiO 2 (and its simple exponential character which is implied by the data on liquid GeO 2 which almost superimposes with that of SiO 2 in this reduced representation) is clearly in the extended three-dimensionally connected network structure. This structure, in which every silicon is connected to four nearest neighbor silicons through bridging bonds is self-reinforcing: it offers both resistance to the rupture of any single bond, and restrictions on the number of new configurations generated by the rupture of a single bond when it does occur. Such a structure is resistant to thermal degradation, and reflects this in the very small change of heat capacity which accompanies passage through the glass transformation region [5], i.e. even though the structure can change on the time scale of a heat capacity measurement, the additional contribution to heat absorption per degree of temperature rise is small. This can be demonstrated simply and semi-quantitatively with rudimentary models of glass excitations such as the "bond lattice model" [5] and more recent and refined models of the same ilk due to Brawer [6].

That the foregoing features are to be associated with the mutually support- ing three-dimensional network structure can be confirmed by comparing with silica the properties of aluminum-substituted silica structures when the charge shortfall due to aluminum is compensated by sodium ions on the one hand and by calcium ions on the other. In the former case (liquid albite NaA1SiBOs) the viscosity remains close to Arrhenius in character [7] and the change in heat capacity on glass transition is small [8]. In the calcium compensated case (anorthite CaA12Si2Os) on the other hand, the more competitive status of the charge compensating calcium leads to severe network distortions, as if pressure were applied [9], with the result that the viscosity temperature dependence becomes more strongly non-Arrhenius and the heat capacity change at T~ is distinctly increased. Characteristic of liquids in this intermediate region of the total range of behavior seen in fig. 1, the viscosity of anorthite remains non-Arrhenius all the way to the glass transition. In fact, it almost superimposes on the curve for zinc chloride [3,10]. Zinc chloride itself behaves rather like the hydrogen bonded liquid n-propanol for which a wide (reduced) range of data exist [4,11]. The high temperature data for propanol are used to guide the extrapolation of the ZnCI z and anorthite curves to 1/T = O.

A different and intermediate loss of self-reinforcing capability follows when alkali oxide is added alone to the silica structure. This, as the common wisdom has it, ruptures bridging bonds in proportion to the number of oxides added. At the disilicate composition the conditions for two dimensional sheet-like structure are achieved. Indeed X-ray studies by Imaoka et at. [12] show a distorted planar structure is achieved. This loss of bonding constraints in one of the three dimensions is accompanied by larger departures from Arrhenius


behavior than seen before, though the liquid is still less than halfway between the strong and fragile extremes, see fig. 1.

If two-dimensional ordering is related to relaxational character it might be expected that As2S 3 [13] would exhibit similar characteristics to those of Na20.2S iO 2. The data for this case [14] near Tg are included in fig. 1 to support this notion. The trend seems to be continued when additional alkali oxide is added to reach the metasilicate composition according to limited data from Endell and Hellbrugge [14]. At this composition, one-dimensional chain structures should be the first order description of the intermediate range order. Some M.D. simulations of this structure are discussed in ref. [9].

There are, unfortunately, no viscosity data on orthosilicate type liquids, although it is known that glasses can be formed by the quenching of a mixture of calcium and magnesium orthosilicates [5].

Some correlations with structure may be made. Although the first coordination shell remains fixed for Si-O, coordination no. = 4, the second nearest neighbor distribution g(Si-Si) obtained from MD calculations [16-19] shows strong variations, see fig. 2. The changes show up not in the Si-Si distance nor in the peak width at half height, but in the coordination number and in the second nearest Si neighbor position. Such a structural characterization is, however, crude and inadequate to give insight into either the origin of non-Arrhenius behavior itself or of the changes in the parameter fl in the detailed relaxation function O(t)= e ~tt/~l~) which seem to associated with it [20-221.

Zinc chloride, illustrated in fig. 1, is a weak network liquid at the outset. The size of the zinc ion relative to its chloride ligands is less favorable to four coordination than in the case of silica, and the result is a network in which the CI-Zn-C1 bond angles are rather acute, and the packing density is high [22,23]. This liquid has characteristics intermediate between the strong and fragile classes [4]. However, the effects of the (less demanding) medium range order imposed by the chloride bridging are still present as can be seen when the data for ZnC12 are compared (see fig. 1) with those of a zinc chloride-based system in which the bridging bonds have been broken by chloride ion additions in the form of pyridinium chloride Py+C1 [10].

In both the latter cases the restrictions on the first coordination number have remained in force, but relief of order constraints at the intermediate range has been achieved. If we relax the constraints further by removing restrictions on the first coordination shell or at least making interchanges between first and second shells simpler, a further increase in "fragility" occurs. These conditions can be seen in the behavior of the molten heavy metal fluorides where the coordination number of the most strongly coordinating species Zr 4+, is less well specified. Current studies [24,25] indicate that zirconium ions exist in a mixture of 7 and 8 coordination states even in the unmodified liquid fluoride, and that this indefiniteness and variability in the medium range order is enhanced by the addition of modifying fluorides.

Extending our considerations to non-ionic glass-forming liquids, we might

q.o00

3.000

~-" 2.000-

1.000-

.000 000

SILICON

C.A. Angell / Spectroscopy simulation and scattering

I I

2.000 R[ I , J ) (ANGSTROMS)

ABOUT SILICON [STANDARD NORMRLIZATION)

q.o00 6.000 8.000

20.00 -

,5.oo- I A J / i0.00- / / ev- /

5.00 - ~ / . / / /

I I / /

.00 I I I I 000 2 .000 4 .000 6 .000 8 .000

R(I,J) (ANGSTROMS) SILICON ABOUT SILICON

Fig. 2. Si-Si pair distribution functions for SiO 2 and alkali silicate composit ions showing the change in S i -S i coordinat ion numbers. The decrease implies a relaxation of the intermediate range order which is associated with the increasing fragility of the viscous liquid.

note that alcohols tend to occupy the center of the fig. 1 pattern [4], while molecular liquids fall at the fragile extreme. Again this is consistent with the restrictions on intermediate range order imposed by hydrogen bonding in the alcohol cases and the lack of anything but repulsion potential determined packing restrictions in the case of the van der Waals liquids.

3. Techniques for investigating intermediate range order

In this section we first discuss some ways of exploiting and extending known techniques which have the power to give structural information which is


relevant to the sort of problem outlined in the previous section. We then go to speculate about other possible avenues for exploring the physical importance and phenomenology of intermediate range order-determined properties, and to consider possible direct and indirect means of observing this order.

3.1. Spectroscopic techniques

In general it has been concluded [26] that spectroscopic techniques are primarily useful for elucidating the first nearest neighbor arrangements, but that with adequate care information on the nature of second nearest neighbor arrangements can be obtained. An example of the latter is the molten salt study of Smith et al. [27] in which systematic studies of this composition dependence of ligand field spectra (in the visible range) studies were used to establish that Ni 2+ in liquid KCI-LiC1 solutions existed in two coordination states and that in the octahedrally coordinated configuration it was Li cations which occupied the second nearest neighbor shell while in the tetrahedrally coordinated state it was potassium ions which occupied the second nearest neighbor position. To the best of this author's knowledge such systematic spectroscopic studies have not yet been performed in oxide glass systems, and the field could profit from the development of relevant studies of this type. Trivalent titanium, for instance, could conceivably be used as a probe for the behavior of Ga 3+, or even perhaps A13+ in glasses where the composition has been chosen to produce both tetrahedral and octahedral coordination states.

Alternatively Ni 2+, which has usefully distinct spectroscopic signatures in octahedral and tetrahedral sites, could be used as a probe for Mg z+ coordination in glasses where, particularly at high temperatures, tetrahedral as well as octahedral sites are anticipated. Nickel spectra obtained in high temperature melts have indicated [28] that except in the most acid conditions, tetrahedral coordination is the common state in the low viscosity high temperature liquid. It is of interest to determine whether, in such systems containing more than one type of alkali, there is a pairing of alkali type with coordination number type for more highly charged cations. Such studies could shed new light on the mixed alkali effect on which little specifically structural information is available.

This problem in intermediate range (at least second nearest neighbor) ordering would require the use of high temperature visible spectrophotometers (of which the CARY-17DHC is the outstanding commercial example) and computer-assisted data analysis to properly execute the spectroscopic subtrac- tions which are needed to clarify the second nearest neighbor arrangements. Useful, and currently unavailable, information on mutual cation ordering in glass forming liquids could be obtained in this manner. The importance of such information can be emphasized by referring to the striking differences in structural relaxation characteristics of NaA1Si308 (albite) and CaA1Si2010 (anorthite) illustrated in fig. 1, the difference clearly being associated with the difference in the aluminum second nearest neighbor relationships.


A powerful technique for exploring directly the latter second nearest neighbor problem, one which is just beginning to be exploited and will surely contribute a great deal of structural understanding in the next two decades, is solid state NMR. Not only does this technique distinguish with great clarity between octahedrally and tetrahedrally coordinated AI (Muller et al. [29], Ohtani et al. [29] but tetrahedral sites with different second nearest neighbors may be revealed in the right conditions. Furthermore, the presence of a microscopic crystalline fraction can be detected in partly devitirfied samples by the line shape due to the narrow band characteristic of the ordered phase. Finally, the fact that the spin lattice relaxation time is longer for ordered than disordered regions means that manipulation of the "dead" time experimental variable allows the investigator to bring ordered regions "into focus" in a sense .

In recent studies using the Si nucleus, Gerstein (Gerstein and Nicol [29]) has been able to detect 4 distinct Si resonances in an aluminosilicate glass, each being attributed to a particular combination (total 4) of AI + Si in the second neighbor shell. Si with 4 A1 in the second shell are, of course, a weak minority occurrence but nevertheless exist.

Since these network sites are well defined it must be expected that there will be difference in the potential energies of the alkali cations which charge compensate the AI 3 + cations in the network. Such differences may well be the origin of the different energy sites of alkali cations which current weak electrolyte theories of glass conductivity are obliged to postulate. By analyzing the thermal history dependence, and composition dependence of the Si A1 neighbor relationships and comparing with electrical conductance responses to the same two variables and with the 23Na+ solid state NMR spectrum itself, it might be expected that much progress in understanding the fast ion conductor problem will be made in the near future and the problem may well be solved by 2O04.

While spectroscopic techniques, such as the d d transition and solid state N MR spectroscopies discussed above, may elucidate important second nearest neighbor structure problems, they appear to be inadequate to research the nature of third and fourth nearest neighbors. Since it is in these larger distance ranges that much important physics, including the nucleation phenomenon, is determined, we must inquire as to suitable methods for obtaining information in this range.

3.2. Pair distribution functions from simulation and difference scattering techniques

3.2.1. Computer simulation techniques The application of fast computers, using sophisticated multiparticle system

simulation programs to obtain both structural and dynamic information on ionic liquid structures, has been exploited by several workers in recent years [16-19,29-32], and their power is generally appreciated. Although there are


FN.~leOe 6OOOK. z.99o

.500 -

1.500-

1.000-

.DO0 - .000

SODIUM

2.000 4.000 8,000 R(I,J) (ANGSTROMS)

ABOUT BLUMINUM

\ l

[STRNDQRD NQRMALIZQTI~N)

~o

c~

$

8.000

ZB 21:12 - 21ZRF4- I2BAF2 - ~00K.2 .990 .

8.000

6.000

2.000

.000

I c~

o o

4

8

o

o

~z cic

8.000 .000 2.000 q'.ooo 6.000 R( I . J ) (ANGSTROMS)

ZiRCON]UM ABOUT ZIRC@'!UM (STA',;D~RD NORHALIZATIO~,J)

Fig. 3. (a) Correlation of Na + and AI 3+ position in high temperature NaAISi206 melts at low presure. (b) Zr-Zr pair distribution function in quenched BaF 2-ZrF4 melts showing presence of two distinct Zr-Zr distances (associated with single and double fluoride bridges).

severe limitations on the information on experimental glassy states [16,33], these techniques are rather well suited for the study of hot fluid states where experimentation is particularly difficult. Examples of their output have already been given in fig. 2, and fig. 3 shows relationships between aluminum and alkali cations in some sodium aluminosilicate melts which have recently been studied as a function of increasing pressure [9]. Dropping the temperature quenches in the high temperature structure but reduces vibrational smearing so that such interesting features as the existence of two distinct Zr -Zr distances in


liquid (hence presumably also glassy) fluorozirconates can be revealed [25] (se fig. 3b). Similar features have been seen in preliminary simulation studies of silicate liquids in which some "oxide" has been replaced with "sulfide" (i.e. by doubly charged anions with an appropriately increased repulsive potential parameter) and, more prominently, in an all-sulfide melt of stoichiometry Li6SizS 7. Again, the new Si-Si distance is a short one, 3.5 A, also due to double sulfide bridges such as occur in SiS 2 itself.

The system represented in fig. 3 is relatively small (190 atoms) but with growing computer power and improved algorithms, the investigation of much larger systems is becoming feasible, Such simulations will give full detail on the intermediate range order out to distances of the order of 15 A, which should be sufficient to contain most of the important physics of microhomogeneous melt behavior. Algorithms for obtaining viscosities of simple liquids with good accuracy have been developed [34], and it should not be long before the viscosities of complex ionic liquids of different fragilities (see fig. 1) can be determined in the same way. Given the rate of development in power, and decrease in price, of computing hardware, it is reasonable to suppose that direct computer-simulation studies will be in a position to provide a great deal of insight into the relation between intermediate range order and physical properties by the year 2000. It, of course, must be demonstrated that the pair potentials used in these simulations are adequate for the purpose. In this respect it is necessary and important that experimental techniques be devised for obtaining equivalent information for at least a few important test cases. We consider how this will probably be done in the following paragraphs.

3.2.2. Anomalous X-ray scattering studies using small laboratory X-ray laser sources

The major problems encountered in attempts to unscramble total distribution functions obtained from simple X-ray scattering studies are well recognized. Major strides towards the refinement of our knowledge of pair functions has, however, been made with the development of differencing techniques. In particular the isotope replacement technique in neutron scattering for ionic and metallic liquid structure studies has been skillfully exploited by Enderby and colleagues [35], and Dupuy and co-workers [36] and a start has been made on its application to ionic glasses by Wright and co-workers [37]. The number of pair functions contributing to the final pattern is greatly reduced in this manner, and interpretation of structural features in multicom- ponent systems becomes possible. A limit is imposed by the statistical noise in the initial patterns. With one or two orders of magnitude increase in counting statistics 3rd and even 4th order differences could be taken and enormous structural detail revealed. We consider how such ameliorations of the basic idea involved in difference techniques may come about in the following paragraphs.

A more exciting and potentially "personalizable" technique for obtaining difference patterns involves the use of the anomalous X-ray scattering phenom-


enon, in which the differencing may be done using only a single sample, which is investigated by X-rays of different wavelength. If one of the wavelengths is chosen to fall on an absorption line for one of the nuclear species of interest, the resultant patterns can be subtracted to yield information on the pair functions involving only this species. While this technique is currently in its infancy [37,38], and suffers, at the moment, from the need to have access to a synchrotron, this position is possibly about to undergo a drastic change. Work by Rhodes and co-workers at the University of Illinois [39] has demonstrated the feasibility of generating extreme UV or soft X-ray quanta at intensities far in excess of those available from synchrotrons, and tunable over wide ranges. Plausible arguments have been given by Rhodes [40] to the effect that by the extension of current developments even hard X-ray laser sources with intensities far in excess of those available in synchrotron sources could be developed in the near future. Evidently [40] the hardware involved in the multupling (septupling is needed ultimately) of the initial laser frequencies to produce intense X-ray beams could be available to the individual investigator at costs of the order of $100000, in the not far distant future. Such instrumentation, clearly, would revolutionize the experimental investigation of intermediate range order in liquids of all types. Such "surrogate synchrotron" systems could conceivably be available to the individual investigator within the decade. The corresponding increase in information on intermediate range order could become a surfeit by the year 2004.

3.3. X-ray lasers and the relaxation of intermediate range structure

The advent of the ultra high intensity X-ray sources discussed in the previous section would also permit the investigation of important questions concerning the dynamics of intermediate range order. Experiments equivalent to the probe ion relaxation experiments (short range order relaxation) performed in this laboratory [41-43] will become feasible for following relaxation of various features of the structure factor. For instance, it will be possible to study the time development, following a temperature jump, of the small q feature of S(q) in As2S 3 which has been studied recently by Busse [44] as a function of temperature. This peak is related to the separation of raft-like features of the microstructure. Combination of relaxation studies of short and intermediate range order, with thermodynamic studies (e.g. volume relaxation) would do much to unravel such currently vexing problems as the origin of the e -[~/~)BI relaxation function characteristic of viscous liquids and glasses.

Time-dependent differencing techniques may also become a possibility with the high intensity sources (projected to be 10 6 more intense than current synchrotron sources, and fully tunable). By alternating short exposures, the time development of different pair functions following a perturbation should become possible. Limitations are set by the tendency of incident energies of these magnitudes to vaporize the sample.

Time-dependent small angle X-ray studies, which in principle could show


how the larger scale (20-200 A) structure develops during crystal nucleation or liquid-liquid phase separation, will also become feasible.

3.4. Other techniques

There are other techniques, such as transmission electron microscopy and lattice imaging from quasi-crystalline or highly correlated regions of the amorphous structure, which will be of the greatest importance in enhancing our understanding of intermediate range order. With resolution of the order of 3 A now becoming possible [45,46] it seems clear that in the next two decades studies in which the chemical components are chosen to maximize electron density differences between dense-packed and loose packed regions of the amorphous structure will contribute greatly to our conception of the essential features of amorphous packing, and in particular the relation of intermediate range order to paracrystallinity [1]. This author is not competent to comment on the most probable developments in this area but their potential impact on our thinking, given their "real space" information aspects, will be enormous.

Finally we should note the possibility of gaining fundamental insights from the study of microscopic model systems such as the latex or lucite microsphere (0,3 ~m diam.) suspensions of Ottewili and Pusey and possibly the new 5 nm diam. droplet microemulsions with vitrified droplets. These systems which can be obtained transparent, and interrogated by visible or UV light, may behave like hard, soft or attractive spheres (see sect. 4.2 and ref. [55]).

4. Short range order and physical properties

In this section we return to our starting point and speculate briefly on possible developments in techniques for probing the role played by intermediate range order in determining physical properties.

4.1. Dependence of "'fragili(v" on intermediate range order, bv time scale mixing

4.1.1. Fully amorphous phase studies If glasses that have two structural order parameters with very different

relaxation times can be obtained, and if the slow order parameter involves the intermediate range order via, for instance, a ring-chain conversion then it may be possible to monitor the relation between liquid fragility and structure rather directly. A possible case would be that of NaPO3 or its relatives in which small anion rings and long chains are both possible and interconversion from one structure to another can occur as a function of time. If this equilibration could be slowed down relative to local structure relaxation times, by, for instance, suitable choice of temperature, then their isothermal viscosity vs. time behavior referred to periodic Tg measurements could reveal the relationship between extended structure and the fragility.


Another possibility is to combine the observations of de Neufville and Rockstad [47] who correlated Tg and band-gap measurements with connectedness in related chalcogenide glasses, with those of Calemzcuk [48] who showed that irradiation of Se below Tg could greatly enhance the rate of enthalpy relaxation below Tg, to investigate in a dynamic mode the relation between bonding dimensionality and relaxation kinetics. For instance, a photon correlation determination of the relaxation function of a viscous liquid of a given connectedness could be determined in the presence and absence of an irradiat- ing beam at the band gap frequency to monitor simultaneously changes in the average relaxation time and associated changes in the /3 parameter of the relaxation function e I~'/~)~1. By variable choice of constituents, bands associated with in-plane order and between-plane order (layer crosslinks) could be selectively irradiated to establish the connection between structure relaxation time and nonexponentially. Some preliminary attempts along these lines using dielectric relaxation to detect changes have been made by Calemczuk [48], and they deserve to be extended in the next decade.

4.1.2. Fractionally crystallized glass studies By correct choice of composition it is possible to cause glasses to nucleate

and crystallize very slowly, with ultimate "crystallite" size far below the visible range. In fact, both our group [49] and Wright and colleagues at the ILL, Grenoble [50], have caused complete crystallization to occur under conditions well below Tg in which mean diffusion paths are less than 20 A. According to small angle neutron scattering studies, no structural inhomogeneities in excess of 10 A in diameter can be observed in these products [50] despite the fact that the glass ~ crystal phase change in essentially complete. These irreversible changes could be viewed as controllable time-dependent changes in the intermediate range order, and their effect on viscosity during the period of ultra-microcrystallite formation could be monitored using time scales adequate to ensure local liquid-like structural equilibrium.

4.2. Study of systems which are themselves microscopic

Another approach to the characterization of processes dominated by intermediate range order characteristics may be via the preparation of systems which themselves have the dimensions of the range of interest. This is now a possibility with the recognition and study of thermodynamically stable systems in which there is microscopic phase separation yielding more or less monodis- perse microphases of diameter 15-100 A. For instance, water can be obtained in microdroplets of this size range dispersed in decane using certain molecular or ionic surfactants [51]. Study of its spectroscopic characteristics have shown it to be in an essentially identical state to ordinary bulk water [52]. Microemul- sions of molecular organic liquids have likewise been obtained dispersed in an aqueous medium [53] and in at least one case the development of glass-forming microemulsion systems has been reported [54,55], see fig. 4.

CA. Angell / Spectroscopy simulation and scattering 15

S (surfactant) /~ween 80

/ : o o. / / oo,

(PG'3H20) Vol % Aqueous Phase (OiE,O-Xylene)

Fig. 4. Composition regions in the pseudo-ternary system (propylene glycol trihydrate)+(Tween 80 surfactant)+(o-ylene) in wl~h 20-50 ,~, nficrodroplets of o-xylene can be studied in the viscous liquid and glassy states in a matrix of glass PG. 3H20.

The glass transition temperature in the latter was found to be almost unchanged from that of the bulk material, and to be continuous between normal and microemulsion phases. In the finer emulsion and the microemulsion states, crystallization after glass transition does not occur [55]. Variations in the "oil" phase volume fraction between 0 and 50% are possible, see fig. 4, dotted line.

The possibility of using such systems for study of system size-dependent (indeed, wave vector-dependent) properties has just been recognized, and its exploitation in the next decade may be anticipated. We must emphasize that in such systems we have individual entire systems which are of the dimensions of a single cluster of clustermodeler's glasses and not much larger than the "amorphons" described by Hoare for packing of spherical particles [56]. The fact that Tg remains essentially unchanged under these circumstances [55] would argue that at least the transport properties of viscous liquids are not determined by the presence of clusters if they exist. It will likewise be of interest to pursue the existence and properties of secondary relaxations in such microsystems because of the tendency to think that they originate in the "tissue" material of microscopically inhomogeneous glasses.

A number of problems in interpretation of the behavior of such truly microscopic (better, nanoscopic) systems will have to be solved, and probably it will the features they show in common with bulk system rather than the converse which will be the most informative. A twenty year time scale is probably appropriate in this instance also.

16 C.A. A ngell / Spectroscopy simulation and scattering

Th is work has benef i ted f rom the suppor t of the NSF Sol id State Chemis t ry

Grant No. DMR 8007053.

References

[1] J.C. Phillips, J. Non-Crystalline Solids 34 (1979) 153; Phys. Stat. Sol. (b)101 (1980) 473; Sol. State Phys., in press.

[2] W.T. Laughlin and D.R. Uhlmann, J. Phys. Chem. 76 (1972) 2317. [3] (a) C.A. Angell and J.C. Tucker, in: Physical Chemistry of Process Metallurgy: the 1973

Richardson Conference, eds., Jeffes and Tait (Inst. Min. Met. Publ., 1974) p. 207. (b) C.A. Angell and W. Sichina, Ann. N.Y. Acad. Sci. (Proc. Workshop on Glass Transition and Nature of the Glassy State) 279 (1976) 53.

[4] C.A. Angell, Strong and Fragile Liquids, in: Proc. Workshop on Relaxation Processes, Blacksburg, Va, ed., K. Ngai (July 1983).

[5] C.A. Angell and K.J. Rao, J. Chem. Phys, 57 (1972) 470. [61 S.A. Brawer, J. Chem. Phys. 81 (1984) 954. [7] D. Cranmer and D.R. Uhlmann, J. Geophys. Res. 86 (1981) 7951. [8] P. Richer and J. Bottinga, Geochim. Cosmochim. Acta 48 (1984) 453. [9] C.A. Angell, P.A. Cheeseman and S. Tamaddon, Bull. Miner. 1/2 (1983) 87; Science 218

(1982) 885. [10] A.J. Easteal and C.A. Angell, J. Chem. Phys. 56 (1972) 4231. [11] A.C. Ling and J.E. Willard, J. Phys. Chem. 72 (1968) 1918. 112] M. Imaoka, H. Hasegawa and I. Yasui, Phys. Chem. Glasses 24 (1983) 72. [13] S.V. Nemilov, Sov. Phys. Sol. State (1964) 1075. [14] K. Endell and H. Hellbrugge, Angew. Chem. 53 (1940) 271. [15] (a) P.R. McMillan, J.-P. Coutures and B. Pirion, C.R. Acad. Sci. Paris, Ser. II 292 (1981) 195.

(b) P.F. McMillan and B. Piriou, Bull. Miner. 106 (1983) 57. [16] L.V. Woodcock, C.A. Angell and P.A. Cheeseman, J. Chem. Phys. 65 (1976) 1565. [17] T.F. Soules, J. Chem. Phys. 71 (1979) 4570; J. Non-Crystalline Solids 49 (1982) 29. [18] S.K. Mitra & R.W. Hockney, Phil. Mag. B48 (1983) 151. [19] S.H. Garofilini, J. Chem. Phys. 76 (1982) 3189; J. Non-Crystalline Solids 63 (1984) 337. [20] H. Tweer, J.H. Simmons and P.B. Macedo, J. Chem. Phys. 54 (1971) 1952. [21] K. Ngai, Solid State ionics 5 (1981) 27. [22] L.M. Torell and C.A. Angell, J. Chem. Phys. 78 (1983) 937. [23] J.A.E. Desa, A.C. Wright, J. Wong and R.N. Sinclair, J. Non-Crystalline Solids 51 (1982) 57. [24] R. Coup& D. Louer, J. Lucas and A.J. Leonard, J. Am. Ceram. Soc. 86 (1983) 523. [25] J. Lucas, C.A. Angell and S. Tamaddon, Mat. Res. Bull. 19 (1984) 945. [26] J. Wong and C.A. Angell, Glass: Structure by Spectroscopy (Dekker, New York, 1976) Ch.

12. [27] G.P. Smith, J. Brynestad, C.R. Boston and W.E. Smith, in: Molten Salts, ed., G. Mamantov

(Dekker, New York, 1969). [281 T.C. Lin and C.A. Angell, J. Am. Geram. Soc. 67 (1984) C-33. [29] D. Muller, G. Berger, I. Grunze, G. Ludwig, E. Itallus, and U. Haubenreisser, Phys. Chem.

Glasses 24 (1983) 37; E. Ohtani, F. Taulelle and C.A. Angell, Nature (1985) in print; B. Gerstein and A. Nicol, J. Am. Chem. Soc., to be published.

[30] Y. Matsui and K. Kawamura, Nature (Lond.) 285 (1980) 648; Y. Matsui, K. Kawamura and Y. Syono, Adv. Earth Planet. Sci. 12 (1982) 511.

[31] M. Saboungi, M. Blander and A. Rahman, J. Chem. Phys., to be published. [33] C.A. Angell, J.H.R. Clarke and L.V. Woodcock, Adv. Chem. Phys. 48 (1981) 397. [34] D. Fincham and D.M. Heyes, Chem. Phys. 78 (1983) 425. [35] J.E. Enderby D.M. North and P.A. Egelstaff, Phil. Mag. 14 (1966) 961; J.E. Enderby and

G.W. Neilson, in: Water: A Comprehensive Treatise, Vol. 6, ed., F. Franks (Plenum, New York, 1982).

C.A. A ngell / Spectroscopy simulation and scattering 17

[36] Y. Derrein and J. Dupuy, J. Phys. Paris 36 (1975) 191. [37] A.C. Wright, G. Etherington, J.A. Erwin Desa and R.N. Sinclair, J. Phys. Colloq. 43 (1982)

C9. [38] P.H. Fuoss, P. Eisenberger, W.K. Warburton and A. Bienenstock, Phys. Rev. Lett. 46 (1981)

1537. [39] T. Srinivasan, H. Egger, H. Pummer and C,K. Rhodes. IEEE J. Quantum Electronics QE-19

(1983) 1270. [40] C.K. Rhodes, private communication. [41] A. Barkatt and C.A. Angell, J. Phys. Chem. 82 (1978) 2622. [42] A. Barkatt and C.A. Angell, J. Chem. Phys. 70 (1979) 901. [43] C. Hunter and C.A. Angell, to be published. [44] L.E. Busse, Phys. Rev. B29 (1984) 3639. [45] J. Thomas, L.A. Bursill and K.J. Rao, Nature 89 (1982) 157; L.A. Bursill, and J.M. Thomas, J.

Phys. Chem. 85 (1981) 3007. [46] P.H. Gaskell, contribution to NATO ASI on: Glass, Current Aspects (Plenum, New York~ to

be published). [47] J.P. de Neufville and H.K. Rockstad, in: Proc. 5th Int. Conf. on Amorphous and Liquid

Semiconductors, ed. G. Lucovski (Taylor and Francis, New York, 1973) p. 420. [48] R. Calemczuk and E. Bonjour, J. Non-Crystalline Solids 43 (1981) 427; R. Calemczuk, private

communication. [49] C.A. Angell and D.R. MacFarlane, Advan. Ceram. 4 (1981) 66. [50] A.F. Wright, contribution to NATO ASI on: Glass Current Aspects (Plenum, New York, to

be published). [51] L.R. Angel, D.F. Evans and B.W. Ninham, J. Phys. Chem. 87 (1983) 538. [52] D.L. Fields and C.A. Angell, to be published. [53] J. Peyrelasse, C. Boned, J. Heil and M. Clausse, J. Phys. C15 (1982) 7097. [54] D.R. MacFarlane and C.A. Angell, J. Phys. Chem. 86 (1982) 1927, [55] C.A. Angell, R.K. Kadiyala and D.R. MacFarlane, J. Phys. Chem. 88 (1984) 4593: J.

Oubochet, J. Teixeira, R.K. Kadiyala, C.M. Alba, D.R. MacFarlane and C.A. Angell, J. Phys. Chem. 88 (1985) 6727 (Debye Memorial Issue).

[56] M. Hoare, Ann. N.Y. Acad. Sci. 279 (1976) 186.

29.CA Agell.pdf

Documents