Amorphous materials

(134 Elliot SR, physics of amorphous materials)

Amorphous materials are not new; the iron-rich siliceous glassy materials recovered from the moon by the Apollo missions are some billions of years old, and man has been manufacturing glassy materials (principally from silica) for thousands of years. Preparation of amorphous materials There are at least a dozen different techniques that can be used to prepare materials in an amorphous state. Of these, five are commonly used in one form or another to produce most non-crystalline materials of commercial or academic interest. The different techniques use variously all three phases of matter, vapour, liquid and solid, as the starting materials in producing amorphous solids, but only deposition from vapour or liquid phases is really important.

Since the amorphous phase is less thermodynamically stable than the corresponding crystalline form (i.e. it possesses a greater free energy), the preparation of amorphous materials can be regarded as the addition of excess free energy in some manner to the crystalline polymorph. How this is done can vary widely, but it is a rule of thumb that the faster the rate of deposition (or cooling), the further the amorphous solid lies from 'equilibrium'. Thermal evaporation This technique is perhaps conceptually the most easy to understand, and is possibly the most widely used method for producing amorphous thin films. It is one of several ways of producing amorphous solids by deposition from a vapour. In essence, it is very simple; the starting compound is vaporized and the material is collected on a substrate. In practice, the evaporation is performed in vacuo to reduce contamination; a typical arrangement is shown in Fig. 1.3. An oil diffusion pump, together with a liquid nitrogen cooled trap (to eliminate water vapour and to a lesser extent hydrocarbon contamination) can produce a base pressure typically of the order of 10-6 Torr in the chamber. The starting compound can be heated to vaporization in one of two ways.

A 'boat' containing the powdered material acts as a resistive element between two electrodes for the passage of a large d.c. current; high melting-point metals, e.g. molybdenum or tungsten, are used as boat materials. This method is suitable for relatively low melting-point compounds; those materials with higher melting points are best vaporized by bombardment with high-energy electrons from an electron gun placed in the chamber. The purpose of the electron- beam bombardment is simply to heat the sample to its melting point; compare this with the mechanism that takes place in 'sputtering' which is the topic of the next section.

Fig. 1.3 Schematic illustration of a thermal evaporation chamber. The material to be evaporated lies in a high-melting point metallic boat

(B) through which a heating current is passed, and the evaporated film is collected on a substrate (S) attached to a heater (H). The entire apparatus is placed in a vacuum chamber pumped to a high vacuum (V). Many materials can be rendered amorphous using thermal evaporation. The essential feature is that atomic surface mobility is greatly diminished because of the cold substrate, causing the adatoms to be 'frozen' in the random positions at which they arrive. Examples include Si and Ge and the tetrahedrally bonded III-V alloys (e.g, GaAs), 'chalcogenide' materials (those containing wholly or mainly chalcogen group VI elements), and SiOx (where x can be varied between 0 and 2). These are all examples of solids containing principally covalent bonds. Certain metals can also be produced in an amorphous form by thermal evaporation without the need of stabilizing 'metalloid' atoms, but this must be done using substrates held at very low temperatures (4 K), otherwise polycrystalline films are produced. Amorphous Ga and Bi can be made in this way, although other elements invariably crystallize even at these low temperatures, and other methods must be employed. The principal disadvantage of thermal evaporation as a preparative technique lies in the variability in purity and composition of the resulting films (in the case of alloys). Many factors combine to produce this variability, and some are listed below: Substrate temperature. (b) Substrate-source separation and orientation. (c) Base gas pressure in the chamber. (d) Impurities from the evaporation boat, or desorbed gases from the chamber surfaces. (e) Boat or filament temperature. Some of these parameters the experimenter can control (and should then specify in a description of the material preparation), others, including several not listed, cannot be varied at will, and so can change in a random fashion from one preparation to another of ostensibly the same material. This was the principal reason why so many experimental data were apparently contradictory in the early days of the subject. Some of the variables listed above are self-evident, others are more subtle. The substrate temperature is important since the higher the temperature, the higher is the surface mobility which can lead to a material with considerably fewer structural defects. However, if the substrate temperature is too high, the material will crystallize.

The relative orientation of substrate with respect to source is of importance since evaporation at oblique angles can produce materials which are macroscopically homogeneous (Leamy et aL 1980). A 'shadowing' effect causes a columnar growth morphology to be formed with dense regions parallel to the source-substrate direction (see Fig. 1.4); this has been observed by scanning electron microscopy in e.g. thin films of a-Ge and a-Ge-Se alloys. In general, the column direction and the evaporant beam direction do not lie parallel to each other; rather, the columns lie more nearly perpendicular to the film surface. The respective angles of incidence, and , are found to obey the empirical equation (Leamy et aL 1980): tan = 2 tan .

Fig. 1.4 Schematic illustration of the formation of columnar growth morphology of thin films evaporated at oblique angles of incidence. Note that in general the columns do not lie parallel to the evaporant beam direction. A major problem with conventional thermal evaporation is differential evaporation and the consequent lack of control over the resulting composition of the deposited film. In conventional vacuum evaporation, the vapour is removed immediately from the vicinity of the molten source, thereby preventing the establishment of equilibrium. In this case the element in the source alloy with the highest vapour pressure evaporates preferentially, depleting the source and leading to compositional inhomogeneities in the deposited film. Flash evaporation can be used to circumvent this problem to a certain extent, as mentioned earlier.

A further complication which can arise in both flash and conventional evaporation is that the dominant (equilibrium) vapour species may not have the same composition as the source material. A case in point is that of arsenic chalcogenide alloys (e.g. As-S), in which the dominant vapour species are molecules of As4S4 even though the starting material may be stoichiometric As2S3. If the substrate is at a low enough temperature (and room temperature might be sufficient for this purpose) the deposited film will tend to consist of discrete As4S4 molecules bonded together by Van der Waals' forces. Annealing the film at a higher temperature will polymerize and cross-link these units yielding a fully coordinated random structure similar to that obtained by quenching from the melt. Alternatively, the substrate can be held at an elevated temperature and the annealing then occurs simultaneously with the deposition process. Certain vapour species (e.g. As4) have very low 'sticking coefficients' because of their shape, and consequently produce poor quality films. Sputtering The sputtering process is rather more complicated than thermal evaporation but enjoys the advantage of being far more flexible. In essence, it consists of the bombardment of a target by energetic ions from a low-pressure plasma, causing erosion of material, either atom by atom or as clusters of atoms, and subsequent deposition of a film on a substrate. The simplest way to induce sputtering is to apply a high negative voltage to the target surface, thereby attracting positive ions from the plasma. However, this d.c. sputtering process is only feasible for targets composed of metals, or at least consisting of material sufficiently electrically conducting that the target can act as an electrode; certain of the crystalline arsenic chalcogenides are sufficiently conducting to act in this way to produce amorphous films. A more common approach is to apply an r.f. field (at typically ~ 13 MHz), to the target (Fig. 1.5). In this case, both metallic and insulating materials may be used as targets; all that is required is for the r.f. voltage to be capacitatively coupled to the target surface. For metallic targets, this is achieved by connecting a capacitor in series with the target; insulating targets are normally bonded to a metal backing electrode which itself acts as a capacitive component. As in evaporation a residual gas pressure in the chamber of less than 10-6 Torr is desirable and achieved by similar means.

The sputtering gas, commonly an inert gas such as argon, is introduced at a higher pressure, say 1-20 mTorr. Application of the r.f. voltage causes the striking of a plasma and positive ions are attracted to the target during each negative half-cycle. However, since the mobility of electrons is higher than that of ions in the plasma, more electrons than ions are attracted to the front surface of the target in their respective half-cycles, resulting in the buildup of a negative bias. After some time, therefore, the ions are attracted from the plasma essentially by the bias potential, rather than the r.f. potential. The ejected material from the target is then carried to the substrate by the following half-cycle of the r.f. field. The material deposited on the substrate can form an amorphous film for the same reason as in evaporation; namely, adatom mobility is sufficiently low, because of the relatively low temperature of the substrate, that crystallization is prevented. However, the substrate must be cooled and the r.f. power limited so that heating due to electron bombardment is minimizedThis method generally produces reasonably homogeneous and uniformly thick samples; for 5% uniformity of film thickness the diameter of the target needs to be about 5 cm larger than the maximum linear dimension of the substrate. The process can be operated in either sputter-up or sputter-down modes, the former having the advantage of being able to use targets in powder form. Sputtering, particularly using r.f. fields, can produce amorphous samples of most materials of interest, Si, Ge, SiO2, chalcogenide glasses and amorphous metals being examples. Sputtering is significantly superior to evaporation for the production of multicomponent systems, For targets already in compound form this is because the sputtering rates for different elements do not vary widely, unlike melting points, vapour pressures and, therefore, evaporation ratesAs a consequence, sputtered films tend to preserve the stoichiometry of the starting material. In addition, amorphous alloys can be sputtered simply by using a target composed of the individual components of the alloy. For targets comprising, say, small circles of one component placed uniformly over the surface of the circular target of the predominant component, films possessing a good degree of compositional homogeneity can be achieved for sufficiently large substrate-target distances; for 'split' targets consisting of two semicircles of different components, on the other hand, films having a composition gradient (perpendicular to the division of the target) can be obtained. Usual sputtering rates for insulating target materials in an inert gas atmosphere (e.g. Ar) are in the range 1-10 s-1, and perhaps a little higher for metallic targets.

These rates, it will be noticed, are significantly lower than those encountered in evaporation. A further option offered by sputtering is the use of gases other than Ar which chemically react with the target, resulting in 'reactive' sputtering. This can significantly increase the sputtering rate, as well as incorporate chosen additives into the films. This has been used in producing hydrogenated amorphous silicon (by sputtering in an Ar:H2 mixture), doped amorphous silicon films (Ar:H2 :N2/PH3), and amorphous transition-metal oxide materials (Ar:O2). Most of the factors which cause variations in the films produced by thermal evaporation are also significant in the sputtering process. Additional factors which need to be considered are the following: (a) Sputtering gas pressure. (b) Ratio of partial pressures of reactive gas to inert gas (in reactive sputtering). (c) R.F. power applied to target. (d) Bias voltage of target or substrate. Close control over all these parameters is necessary to ensure that films with reproducible characteristics are producedChemical vapour deposition The technique of chemical vapour deposition (CVD) is analogous to the GD method in that both depend on the decomposition of a vapour species. The difference is that the CVD process relies on thermal energy for the decomposition (i.e. it is pyrolytic), and the applied r.f. field (if used) simply serves to heat up the substrate upon which the vapour decomposes. Temperatures of the order of 1000 K are commonly used. The CVD process is often used to produce polycrystalline material, particularly silicon from SiH4. However, under the appropriate conditions true amorphous hydrogenated silicon (a-Si:H) can also be deposited by the CVD technique from SiH4, and doped.

Melt quenching The oldest established method of producing an amorphous solid is to cool the molten form of the material sufficiently quickly; what is meant by 'quick cooling' will emerge later. Amorphous materials produced in this manner have often in the past been termed 'glasses'; we prefer here to define such materials as those which exhibit glass-transition phenomena (discontinuities in specific heat, etc.) and hence are not necessarily formed only by melt quenching, although the great majority of melt- quenched amorphous solids do indeed show glass-transition behaviour. The distinguishing feature of the melt-quenching process of producing amorphous materials is that the amorphous solid is formed by the continuous hardening (i.e. increase in viscosity) of the melt. In contrast, crystallization of the melt occurs as a discontinuous solidification, solid growth taking place only at the liquid-solid interface, with the result that crystallites grow in the body of the melt. An essential prerequisite for 'glass' formation from the melt, therefore, is that the cooling be sufficiently fast to preclude crystal nucleation and growth; the crystalline phase is thermodynamically more stable and crystal growth will always dominate over the formation of the amorphous phase if allowed to take place. A good discussion of crystal nucleation and growth kinetics can be found in Turnbull A969). The crystallization rate of an undercooled liquid depends on the rate of crystal nucleation and on the speed, w, with which the crystal-liquid interface moves. Both in turn are strongly dependent on the reduced temperature Tr=T/Tm, and the undercooling Tt = (Tm T)/Tm. It is expected that u should be proportional to some function of Tr (linear for small Tr), and inversely proportional to an average jump time, , of atoms in the interfacial region; in turn might be expected to scale as the viscosity. Thus, high values of viscosity reduce the crystallization-front velocity and hence the crystallization rate itself; this is in accordance with common experience in which it is often found that the liquids of 'easy' glass-formers have high viscosities for temperatures at and below Tm (e.g. the viscosity of silica = 107 poise at the melting point of crystobalite). What then is a condition for glass formation? It is that the nucleation irate In should be less than a certain value, say, 10-6 cm-3 s-1, not to be observable in practicable time-scales and liquid volumes. Simple nucleation theory dictates that crystal nucleation is promoted by the change in free energy on crystallization (dependent on the volume of the crystallite), and at the same time retarded by the work needed to form the crystal-liquid interface (proportional to the area times the surface tension y); the

interplay between these two competing factors ensures that only those crystals which contain more than some critical number of atoms can exist as crystal nuclei; all others remelt.

In

3 exp[ b n n / Tr (Tr ) 2 ]

Turnbull (1969) has shown that the homogeneous nucleation rate can be expressed in the form: where k is a constant, b is a constant dependent on the shape of the nucleus 16/3 for a sphere) n and n are dimensionless parameters defined as: 1

( NV 2 ) n H m

(where N is Avogadro's number, V is the molar volume of the crystal and Hm is the molar heat of fusion), and as:

H m S m RTm R

Fig. 1.8 Homogeneous nucleation rate as a function of reduced temperature Tr= T/Tm calculated for various values of an/?n1/3 (Turnbull 1969). The quantity n represents the number of monolayers per unit area of crystal which Would be melted at Tm by an enthalpy equal in magnitude to .

A plot of [1.2] is shown in Fig. 1.8, showing the nucleation rate In as a function of reduced temperature Tr for different values of the parameter . The value chosen for the pre-exponential factor in [1.2] was 1032 with a temperature independent value of = 10-2 poise; this factor can be written equivalently as the product of the atomic density (1021 cm-3) and a typical atomic jump frequency (1011 s-1). It can be seen from Fig. 1.8 that for values of the crystal nucleation rate becomes less than that which is readily observable, and these liquids would not crystallize (unless seeded), forming glasses if sufficiently under cooled. Using a more realistic model for the temperature dependence of the viscosity, it is found that the peak in the I-Tr relation moves to higher values of Tr and becomes lower and narrower for increasing values of the reduced glass-transition temperature Tg/Tm. The con- consequence of this is that glass-forming tendency should increase with Tg/Tm. In this respect, then, we may regard 'glass' formation as a kinetic phenomenon; a glass is simply a supercooled liquid in which no time for crystal growth has been allowed. In fact, 'glass' is considerably more complicated than this simple picture implies, and many factors contribute to whether a particular glass will form. We have seen that the cooling rate is often a critical factor in determining glass formation. Certain 'easy' glass-formers such as.B2O3, will form an amorphous solid even under conditions of very slow cooling (say 1 Ks-1), whereas other materials, notably metallic glasses, require very high rates of cooling indeed before they form an amorphous phase; otherwise polycrystalline material is produced. The most usual way of producing samples of 'easy' glass-formers is to seal a charge (1-10 g) in a fused silica ampoule under a good vacuum (10 -6 Torr), and keep the ampoule in a rocking or rotating oven at a sufficiently elevated temperature that the constituents become molten and can react; the rocking motion ensures that a thorough mixing of the mixture takes place. The melt can then be quenched, either slowly by simply switching off the oven, or more rapidly by bringing the ampoule into the air, or yet more rapidly by plunging the ampoule immediately from the oven into a liquid (preferably one with a high thermal conductivity and high latent heat of vaporization so that heat is conducted away from the sample as fast as possible without the formation of a thermally insulating vapour layer around the ampoule). In this manner cooling rates of the order10-2 to 103 K s-1 are achievable (see Table 1.1), and are sufficient for most purposes.

Some parameters which are important in the ampoule method of melt-quenching are the following: (a) Temperature of the oven. (b) Rate (or equivalently, method) of cooling. (c) Volume of charge in ampoule. (d) Thickness of the wall of the ampoule.

Material produced in this manner is often in the form of a 'plug' or rod, but can be made into thinner sections by either sawing and polishing or by heating the glass up to its softening point (glass-transition temperature) and then compressing it between two parallel plates or by blowing a thin film. For certain materials however, notably metals, even the fastest cooling rates achieved by the above methods are insufficient to prevent and still faster cooling methods are required. Such methods have been pioneered by Duwez who was the first to demonstrate, in 1959, that cooling rates in excess of 106 Ks-1 can be achieved. The technique relies on the rapid cooling afforded by a 'chill-block' in intimate contact with a thin film of the liquid, in which the heat is swiftly and continually transported away from the interface; chill-blocks of copper are commonly used. The liquid melt may be brought into contact with the chill-block in one of several ways. The first that was used projected small droplets of liquid at a Cu sheet ('splat cooling'); in this manner the first amorphous metal Au75Si25 was produced by melt quenching (Klement et al. 1960). This and a related technique, the aptly named hammer-and-anvil 'drop smasher', can give cooling rates of the order of 105 to 106 Ks-1, but produces thin foils of material of irregular area.

These amorphous metallic materials have many interesting properties (e.g. see Cahn 1980) which lend themselves to several technological applications, but material produced by the hammer-and-anvil or projection techniques is not generally in a usable form. However, there are certain chill-block cooling techniques more recently developed which do produce amorphous metals in a useful geometry, and with a faster cooling rate as an added bonus; these are the 'melt-spinning' and 'melt- extraction' techniques (Figs 1.9(a) and (b)). Both use a rapidly spinning copper disc as the chill-block. The former uses a jet of liquid impinging on the disc producing thin ribbons, whereas the latter removes liquid from a reservoir producing fine wires; in both cases, cooling rates of the order 106-108 Ks-1 can be achieved. These deceptively simple techniques yield cooling rates which depend in a complicated way on a variety of parameters, some of which for melt spinning as an example, are given below: (a) Rotation speed of Cu disc. (b) Coating and surface finish of disc. (c) Melt jet diameter, velocity and angle of attack. (d) Ribbon geometry and uniformity, particularly surface smoothness. (e) Nature and pressure of ambient gas. Despite this apparent complexity, ribbons up to 15 cm wide and 30 m thick can be produced by careful attention to conditions of preparation, although 1 cm by 15 m is more usual. Notice that in contrast to other melt-quenching methods, chill-block techniques cannot produce material in bulk form, but only in the form of thin ribbons (or wires). If attempts are made to produce ribbons of thickness greater than a few tens of microns, only polycrystalline material is produced, since beyond a critical thickness conduction of heat through the ribbon becomes the critical factor, and this rate cannot be made large enough to produce amorphous metals. A summary of various melt-quenching techniques and their cooling rates are given in Table 1.1 together for comparison with those processes normally regarded as annealingIt will be seen that thermal evaporation and sputtering (i.e. vapour quenching) are also included; the precise rates of'cooling' are impossible to quantify in these cases, but must certainly be well in excess of those achievable using chill- block methods. This makes understandable the fact that if a material cannot be rendered amorphous any other way, then some form of vapour deposition will almost certainly be successful.