JOURNAL OF MATERIALS SCIENCE 28 (1993) 6261-6273 Review Wet chemical etching of silicate glasses in hydrofluoric acid based solutions G. A. C. M. SPIERINGS Philips Research Laboratories, P.O. Box 80 000, Eindhoven, The Netherlands The etching of silicate glasses in aqueous hydrofluoric acid solutions is applied in many technological fields. In this review most of the aspects of the wet chemical etching process of silicate glasses are discussed. The mechanism of the dissolution reaction is governed by the adsorption of the two reactive species: HF and HF 2 and the catalytic action of H + ions, resulting in the breakage of the siloxane bonds in the silicate network. The etch rate is determined by the composition of the etchant as well as by the glass, although the mechanism of dissolution is not influenced. In the second part of this review, diverse applications of etching glass objects in technology are described. Etching of SiO= and doped Si02 thin films, studied extensively for integrated circuit technology, is discussed separately. 1. Introduction Silicate glasses, because of their unique properties, are widely used in everyday life. The optical transparancy and the ability to withstand many environmental influences can be considered as the most striking properties in this respect. This is combined with the relative ease with which the material can be given the desired shape, due to its unique viscosity-temperature relation. Silicate glasses are resistant to most gases and liquids. At room temperature they are only readily dissolved by hydrofluoric acid or other HF containing aqueous solutions [1]. Controlled dissolu- tion in HF-based etchants can be applied to remove material from glass objects for a variety of appli- cations. Wet chemical etching of silicate glasses in aqueous HF solutions is a subject which has been studied over many years. The first report originates from the dis- covery of HF by Scheele in 1771 [2]. However, this subject has never been studied intensively and publi- cations on this subject are wide apart in time and scattered over a range of technical journals. In integ- rated circuit (IC) technology wet chemical etching of SiO2 and doped SiO2 thin films was and is being studied more extensively, because these films are widely applied as the dielectric isolation material in IC devices. In this paper the etching of vitreous SiO2 and multicomponent silicate glasses, in bulk as well as thin films, is reviewed and an attempt is made to combine the results from the fields of glass technology and IC fabrication. First, the etching mechanism is described. This is followed by a description of the effect of etchant and glass composition. Then the etching of thin SiO 2- based films is discussed. Finally, most of the important applications are reviewed. 2. The etching process 2.1. Reaction mechanism The dissolution of vitreous SiO2, chemically the most simple silicate glass, into an aqueous HF solution can be described by the overall reaction SiO2 + 6HF --* HzSiF 6 4- 2H20 (1) This equation is a simplification of the reactions occurring during the heterogeneous SiO 2 dissolution. Vitreous (as well as crystalline) SiO 2 consists of tetrag- onal SiO4 units connected at all four corners with four other SiO 4 units by covalent -Si-O-Si- (siloxane) bonds. In this way a covalently interconnected three- dimensional silicate network structure is formed. It is necessary to break all four siloxane bonds in order to break down the network and release a silicon from the glass. HF, dissolved in water, is a weak acid and its solutions contain H +, F- and HF 2 ions and un- dissociated HF molecules. Their concentrations are related by Equations 2a and b: K1 = [H+] "[F-]/EHF] (2a) K= = [HF].[F-]/[HF2] (2b) At 25~ Kl=6.7x10-4moll -a and K2=0.26 moll -1 [3]. Values for K 1 and K 2 at other temper- atures are summarized in [4]. In concentrated solu- tions higher H,F~-+t polymeric species have been observed [5]. The dissolution of vitreous SiO 2 is a hetero- geneous reaction, which makes it difficult to study the mechanism(s) governing the dissolution process. The breaking of all the chemical bonds which results in Equation 1 will require several reaction steps. One of these steps will be the slowest and its rate constant will 0022-2461 (cid:14)9 1993 Chapman & Hall 6261
`
`Raytheon2028-0001
`
`Sony Corp. v. Raytheon Co.
`IPR2015-01201
`
`

`
`determine the reaction rate. The specific property of HF-containing solutions to attack the glass is related to the presence in solution of the fluorine-containing species: F-, HF and HF2. As it is well established that solutions of, for example, NaF or NHgF do not attack SiO 2, so the reactivity of F- ions can be considered to be negligible [6, 7]. The insensitivity of the etch rate to agitation of the solution [8] and the activation energy Ea, which is in the 25-40 kJ mol- 1 range, (to be discussed in Section 3) both indicate that the dissolution reaction is kinetically controlled. Consequently, the adsorption or chemisorption of the reactive species and the effect of this adsorption on the siloxane bonds at the glass surface dominate the dissolution process. The dissolution mechanism, in particular the role of the various fluorine-containing species, has been studied in more detail by Judge [6], Prokopowicz- Prigogine [9-11], Kline and Fogler [12, 13], Spierings [14], Kikuyama et al. [15, 16] and Proksche et al. [17]. The most detaikM model has been presented by Pro- kopowicz-Prigogin~e~[ll], in which the adsorption processes of HF molec'Me~s, HF~ and H + ions deter- mine the reaction rate. The't--I~2 ions are adsorbed on surface silanol groups, the HP molecules on vicinal silanol groups and tt + ions on surface bridging oxygens in siloxane units. Fluorine adsorption com- plexes have been observed at hydrated SiO2 surfaces in gaseous HF by infrared spectroscopy. These are trans- formed into surface groups such as =Si-F and =Si-O-SiF3 [18]. The adsorption of HF and HF~- increases the electronic density on the bridging oxygen in the siloxane unit. This in turn makes these oxygens more basic, so more H § ions are adsorbed, which leads to more siloxane bonds being broken per time unit, i.e. a kind of catalytic effect. The rate-determining step is then the breakage of the siloxane bond by the com- bined action of the adsorbed species. The catalytic action of H + ions on breaking siloxane bonds also occurs in the dissolution of glasses in acidic and weakly alkaline solutions [19]. This effect had already been observed in early HF etching studies by Palmer [20]. The above model results in a complex Equation 3 in which the etch rate VE is determined by the concen- tration of H +, HF~ and HF [11]: VF = kx'O (H+)'{k2"O(HF~ -) + k3"O(HF)} + k4'O(H +) (3) where k i are constants which contain the reaction rate and adsorption equilibrium constants as well as the number of adsorption sites per unit surface area. 19 expresses the degree of coverage of active adsorp- tion sites, which can be assumed to follow Langmuir's isotherm, e.g. | = b'[HF]/(1 + b'[HF]) (4) where b is the ratio between the rate constants of adsorption and desorption, respectively. Etch rate data in solutions with varying HF and HF 2 concentrations show a substantially higher re- activity of the HF~- compared to that of HF [11]. Previously this higher reactivity of the HF~- ion was reported by Judge [6], who obtained for thin film SiO 2 etched in HF-NH4F-containing solutions Vr = a'[HF] + b'[HF~-] + c (5) At 25 ~ a = 0.250, b = 0.966 and c = - 0.014 with VE in nm s-1. Recent studies of the etching in HF solutions buffered with NH4F [16, 17] also report the relation in Equation 5 at low NH4F concentrations. Although the above investigations all report that HF 2- ions have a higher reactivity than HF molecules, Kline and Fogler [12, 13] came to the opposite conclusion, that only HF molecules are reactive. The catalytic role of H § ions was also observed by these authors, which results in the etch rate equation VE = kl"| + k2"| (6) An etch rate versus .concentration (Equation 7) relation was proposed recently which follows the theoretical treatment of Kline and Fogler [12, 13] but with the difference in reactivity between HF and HF~- and the catalytic effect of H + all taken into account: VE = k,'(k2[HF2] + k 3"[HF])'(1 + k 4"[H+]) (7) This relation gives a satisfactory explanation of the observed etch rate of a multicomponent silicate glass in HF-strong acid mixtures [14]. It is clear from the above discussion that the exact reaction mechanism at the molecular level, and parti- cularly the rate determining step, is not completely understood. The discrepancy, particularly in the re- activity of the HF and HFf, has to be investigated in more detail. The mechanism(s) proposed for the dissolution of vitreous SiO2 described above can also be applied to the dissolution of glasses with more complex composi- tions [21] and for compositionally related crystalline silicates such as m-quartz [22] and feldspars [12, 13, 23]. However, the etch rate varies widely. 2.2. Determination of the etch rate Different methods have been developed to determine the etch or dissolution rate of glasses. The weight loss of dispersed powders with known surface areas was particularly used in early studies [24], and more recently using a slurry reactor [12]. A second group of methods is based on partially masking the surface of a glass body, e.g. a disk, by photoresist or wax [8, 14, 25-27]. The glass is locally dissolved and the quantity of material which has dissolved can be determined either by measuring the weight loss [8] or the depth of the recessed etched region [14, 26, 27]. The etch rate for thin glass films deposited on a substrate material can be measured by monitoring the decrease in film thickness during etching by ellipso- metry [28-31] or other optical interference methods [6, 31, 32], such as colour evaluation [33]. Kern developed a method for thin glass films on silicon wafers, which makes use of the hydrophobic/hydro- philic transition occurring when the glass film is completely removed [34, 35]. Since the dissolution 6262
`
`Raytheon2028-0002
`
`

`
`rate of silicate glasses is kineticalty controlled, it is only slightly affected by agitation of the etchant by rotation, stirring or ultrasound [-8, 32, 35]. 2.3. Surface morphology The etching process of a glass not only removes surface material, but it also changes the surface morphology of the glass. When a smooth, e.g. polished, surface is etched the surface roughens slightly and cusp-like structures are formed [7, 21, 36]. A SEM micrograph (Fig. 1) of the surface of a polished soda lime glass etched in a 2.5 wt % HF + 4.6 wt % HNOa solution [-14] shows this structure. Its formation is related to the presence of flaws in the surface before etching. The size or depth of these flaws varies from microcracks (a) 'l'] "l 'l II (b) (c) Figure 1 SEM micrograph showing the cusp-like surface obtained after etching a polished soda lime silicate glass surface [14]. Figure 2 Transformation of a surface with closed microcracks or flaws into a cusp-like glass surface by wet chemical etching. (a) Initial surface. (b) After etching 0.2 time units. The dashed line indicates the initial glass surface. (c) After etching 1 time unit. (d) After etching 3 time units. Figure 3 SEM micrographs of the surface of a soda lime silicate glass after particle erosion and etching in 2 wt % HF. (a) After 2 min, surface view showing the opened microcracks. (b) After 2 rain, cross section. (c) After 30 min surface view showing the Cusp-structure. (d) After 30 rain, cross section. 6263
`
`Raytheon2028-0003
`
`

`
`present after polishing to large cracks formed by mechanical actions such as grinding or particle erosion [36, 373. Fig. 2 shows how a closed flaw or microcrack is etched open and this surface defect is gradually transformed into a cusp as etching proceeds. This process indicates a diffusion of fluorine species over the glass surface into the closed crack, which is fast compared to the dissolution rate. When inhomogenei- ties are present in the bulk of the glass;e.g, in fused silica prepared by melting quartz sand, these will also give rise to a cusp-like surface structure on etching [383. Etching glass surfaces with very small microflaws results in a surface containing very small cusps (,~ i gin), as are shown in Fig. 1. When a glass surface severely damaged by particle erosion is etched, the surface reaction process opens all cracks which are present in large numbers at the surface. Initially this leads to the formation of a three-dimensiorml labyrin- thian surface (Fig. 3a and c). This is also gradually transformed into a cusp-like structure (Fig. 3b and d), but with much larger cusps (~ 10 gin) than those shown in Fig. 1. 3. Effect of etchant composition The reaction mechanisms discussed in Section 2,1 all indicate a difference in reactivity between the HF molecule and the HF 2 ion and a catalytic effect of H + ions. The concentration of these species which can be modified by adding, for example, fluorides or strong acids will determine the dissolution rate of a glass. These effects will be discussed in this section. 50 10 5 1 0.5 0.1 i i i flirl ~ ~ J 5 10 50 HF concentration (M) 0.5 1 5 10 - 28 t I,,,,I , I,,,,I , , I,,,,I 0.5 1 HF content (wt %) Figure 4 Collected etch rate data of SiO 2 in HF aqueous solutions at 23 -i- 2 ~ as a function of the HF content of the etchant [6, 16, 32, 39-453. 3.1. H F concentration Fig. 4 shows the etch rate V~ of vitreous SiO2 in HF solutions versus concentration of HF, compiled from a large number of literature sources. The curve shows that in the middle concentration range (l-d0 wt % HF) the etch rate is approximately linearly dependent on the HF concentration. At higher HF contents the etch rate increases more rapidly, an effect which can be explained by assuming that higher polymeric H,F~-+ 1 ions are present in the etchant [5], which are more reactive towards the siloxane bonds [6, 24]. Fig. 5 shows that the activation energy for the dissolution of vitreous SiO 2 is dependent on the HF content. At low concentrations Ea increases signific- antly with HF content, to the level of 30-35 kJ mol- 1 at about 10 wt % HF, with a tendency to decrease again in concentrated HF solutions. This concentra: tion-dependent E a can be explained by assuming the presence of more than one reactive species [39]. At higher HF contents this could be caused by the assumed presence of H,F~-+ 1 (n > 1) ions. The cause for the more drastic increase at small HF contents is not clear. 3.2. Addition of strong acids By adding strong acids such as HC1, HNO 3 and H2SO4 to HF solutions, the concentration of the more 6264 E E 0 > < 50 40 30 2o 10 oi 't~3 E E r <3 b3 (cid:14)9 0 /o. oO\o o/o/~ ] I I I I I I I 0.05 0.1 0.2 0.5 1 2 5 10 50 HF content (wt %) Figure 5 Activation energy of the dissolution reaction of SiO 2 as a function of the HF content in: (3 HF etchant [6, 8, 17, 40, 45, 46]; (cid:12)9 Bog etchant [6, 47, 48].
`
`Raytheon2028-0004
`
`

`
`reactive HF2 ion in the etchant is lowered, following Equations 2a and b. At H + concentrations larger than 1-2M the etch rate is enhanced [8, 9]. This is due to the catalytic role of the H30 + ions in the dissolution process [13-15], which more than compensates for the reduction in the HFz concentration which is already NH + low in these etchants. For alkali calcium silicate glasses Na+ no difference in etch rate enhancement was observed K+ Mga + between the three acids mentioned above [14]. Only in Ca2+ etchants containing both HF and concentrated Sa2+ H2SO 4 were the etch rates higher than for the HF/HC1 Zn 2+ or HF/HNO3 etchants, an effect which can be ascribed pb2 § to the formation of the strong fluorine-containing A13+ HSO3F acid. 3.3. Addition of N H4F The addition of NH4F to an HF solution shifts Equilibria 2a and b, resulting in an increase in the HF~ concentration as well as in the pH. Etching solutions prepared by mixing a concentrated 40 wt % NH4F solution with a concentrated 49 wt % HF solution in various ratios (usually from 6:1 to 10:1) are used extensively for etching SiO2 and doped SiOa thin films in IC processing (see Section 5.2) [15-17]. They are often referred to as buffered oxide etches (BOEs) or buffered HF (BHFs). Due to the dependence of the etch rate on the HF and HF 2 concentrations, the addition of NH4 F affects the etching behaviour significantly [17]. Adding NH4F to a 6 wt % HF solution first increases the etch rate VE up to about 10-15 wt% NH4F, where a maximum in VE is observed. At larger NH4F contents VE is found to decrease again. It is thought that at low NH4F contents the etch rate is enhanced because the concentration of the more reactive HF2 ions increases. Furthermore, cations such as NH~ (as well as Li + and Na +) also have a catalytic effect on the dissolution reaction of SiO 2 [13]. At higher NH4F contents the reaction with HF becomes increasingly more import- ant, because the HF~- ions are inactivated by com- plexation with NH + ions [17]. The activation energy for SiO 2 dissolution in BOEs is significantly higher than for HF etchants with the same HF content (Fig. 5). This could be caused by the higher activation energy for the reaction involving HF~- ions which predominate in BOEs as compared to that of HF [6], which are mostly present in the HF solutions. At high NH4F concentrations the danger of (NH4)zSiF 6 precipitation, particularly at the glass surface, increases, due to its limited solubility [15, 50] (see also Table I). Its solubility is lowered from 20 wt % in a 6 wt % HF solution to about 1 wt % in a 40 wt % NH4F+6 wt % HF solution [15]. 3.4. Solubility of fluorides and hexafluorosilicates An overview of the solubility of the most relevant bifluorides and hexafluorosilicates in a 30% HF solu- tion [50] is presented in Table I. The low solubility of alkaline earth and lead hexafluorosilicates can cause the precipitation of these compounds, particularly TABLE I Solubility (g per 100g solvent) of the most relevant bifluorides and hexafluorsilicates in 30 wt % HF solutions at 25 ! 2 ~ [50] HFy SiF6 z- 75.8 14.4 3.7 1.2 32.8 0.6 0.0072 Not formed 0.0056 Not formed 0.084 Not formed 11.44 5.8 0.022 Not formed 18.02 6.07 when multicomponent glasses incorporating these bivalent cations are etched in concentrated HF solu- tions. For instance, precipitations are observed when a soda lime silicate glass (with about 15 wt% Na20 and 10 wt % CaO) is etched in HF solutions with contents larger than 30 wt %. 3.5. Other additives Factors such as wetting of the glass and foam forma- tion on the etch solution become of prime importance when an extreme control of the wet chemical etching process has to be realized. This is the case for etching holes in SiO2 film with submicrometre dimensions, as is required in the VLSI and ULSI IC technologies. It then becomes necessary to add HF-resistant additives, such as surfactants [ 15]. The effect of many surfactants on the etch behaviour has been reported [403. 4. Effect of glass composition The incorporation of other oxides in vitreous SiOz modifies its interconnected three-dimensional siloxane structure. In this respect two types of oxides can be distinguished: network modifiers and network- forming oxides. The effect of the incorporation of both oxide types is discussed in this section. 4.1. Effect of network-forming oxides SiO 2 films doped with the network-forming oxides P2Os, B20 3 and As20 a are widely applied in the IC industry, and therefore they have been studied exten- sively. These network-forming oxides are incorporated into the silicate network structure or they form a separate network which is intimately mixed with the silicate network. When a network-forming oxide AxOr is added to SiO2, =Si-O-A- and -A-O-A- bonds are formed [51], which also have to be broken in order to dissolve the material. The chemistry of the breakage of these bonds is not necessarily the same as for the siloxane bonds and the etch rate will be affected. Fig. 6 shows the effect of the incorporation of P205, B203 and As20 3 into SiO2 thin films on the etch rate in 10:1 buffered oxide etch (10NH4F (40 wt %):1 HF (49 wt %)). For P20 s [26, 52] and As20 3 [53] the etch rate increases, indicating that the rate of breakage of P-O and As-O bonds is high compared to that of the 6265
`
`Raytheon2028-0005
`
`

`
`4 ~ E r 3 e- ~2 LU o 0 ), jo J I I 6 12 18 I I 24 30 Mol % dopant in glass Figure 6 Effect ofB20 a (0) [33, 54], P205 ( X ) [33] and As20 a ((cid:14)9 [29] content on the etch rate (at 23 _ 2 ~ of annealed doped SiO 2 films in BOE (10:1). Si-O bond. To a first approximation the etch rate is determined by the SiO 2 content of the glass. The etch rate for SiO 2 doped with B20 3 in a buffered oxide etch (10: 1) first decreases significantly with the B203 con- tent (Fig. 6), indicating that the etch rate in buffered HF is primarily controlled by the breakage of =B-O-Si= bonds [34, 54]. At high B=O 3 contents the =B-O-B= bonds become more abundant [51] and these bonds are readily attacked by water, as indicated by the increased hygroscopicity. The effect of B203 on the etch rate is found to depend on the type of etchant. In contrast to buffered oxide etches, in HF- or HF-strong acid-based etchants with low pHs the etch rate increases with B20 3 content [25, 34]. In these etchants the dissolution rate is limited by the SiO2 content of the glass, because the =B-O-Si= and =B-O-B= bonds are more rapidly attacked by H § ions, which are almost absent in the buffered oxide etch. Three component all-network-forming oxide glasses, such as B203-P2Os-SiO 2 [55, 56] and AI203-B203-SiO2 [34] glasses, have also been studied. The former glasses, knownas BPSG glasses, are used widely in IC technology. At the relatively low temperature of 800 ~ these materials have a low viscosity compared to SiO2, so they will flow over sharp topography steps, smoothing the surface of a processed silicon wafer [30, 56]. The etch rate of BPSG glasses is determined by the boron and phosphor content. In BOE B203 has a decreasing effect, while P20 5 enhances the etch rate [55, 56]. Implantation of ions (At +, P§ N + and B +) into a vitreous SiO 2 partially destroys the network structure. Consequently, the etch rate is enhanced, increasingly so with higher implantation doses [57]. This property has been used to determine implantation profiles of B § [27] and P + ions [27, 57]. SiO 2 prepared by implanta- tion of large doses of O § ions into Si and subsequent annealing has an etch rate slightly lower than that for thermal SiO2 [44], which is thought to be due to the presence of small concentrations of -Si-Si_ = bonds in this material. 4.2. Effect of network-modifying oxides Network modifiers such as Na20, K20, CaO and BaO are incorporated by breaking a siloxane bond, forming non-bridging oxygens by the reaction: Na20 + -=Si-O-Si = --* 2 -SiO-'Na § In this way the three-dimensional silicate network structure is partially broken down. The presence of already broken bonds will strongly increase the etch rate for the glass, as is clearly observed in Fig. 7, which shows the increase in etch rate as NazO is introduced into SIO2. Cations from the network-modifying oxides are bonded ionically to the silicate network and therefore they are relatively mobile. In aqueous solutions (such as HF etchants) these mobile monovalent and bivalent ions (such as Na § and Ca z+) are leached out of the glass and are replaced by ion-exchanged H30 + ions, which form =SiOH groups [19, 58]. The formation of this hydrated silica film preceeds the dissolution of the material in the HF etchant [21, 59]. Due to its open structure and the presence of already broken siloxane bonds it will dissolve more rapidly than vitreous SiO z and increasingly so as the network modifier content is larger. Furthermore, there are indications [21] that fluorine species also diffuse into this hydrated material attacking Si-O-Si bonds positioned within the hydra- ted material. The small effect on the etch rate by partially replacing Na20 by CaO in a sodium calcium silicate glass [21] is an indirect proof for this mech- anism. The formation of such a hydrated film is also supported by microhardness .measurement data of etched sodium calcium glass surfaces. This data indic- ates the existence of a relatively soft surface layer extending to a depth of approximately 1 ixrn [60]. 2O 10 E 5 2 0.5 0 10 20 30 Mol % Na20 Figure 7 Effect of Na20 content on the etch rate ofxNa20. (1 - x) SiO 2 glass at 25 ~ in 5 wt % HF [21]. 6266
`
`Raytheon2028-0006
`
`

`
`4.3. Siliconoxynitrides With chemical vapour deposition (CVD) techniques it is possible to prepare thin films of vitreous siliconoxy- nitrides. The incorporation of nitrogen considerably lowers the etch rate for these materials [61, 62], because the Si-N bonds are not attacked by HF solutions. For example, in a 48 wt % HF solution the etch rate for a SiO2 film prepared by high-temperature CVD is 70 nm s- 1, for SizONz it is 8 nm s- 1 and for Si3N 4 it is reduced to 0.2 nm s- 1 [62]. 4.4. Multicomponent silicate glasses In multicomponent glass systems the compositional dependence of the etch rate becomes more complex as the number of components increases. In these glasses different network-forming and network-modifying oxides can be present, resulting in complex glass network structures. The etching rate of technologically important multicomponent glasses has been sys- tematically studied by Honigmann [50], Molchanov [63] and Spierings [21]. The etching of lead silicate glasses Was studied by Hasegawa [64]. From these paPers the following picture arises. (i) The etching rate of the glass increases as the SiO z content of the glass is' lowered, because less 'silicate' network structure has to be broken by the reaction with HF. (ii) Replacement of monovalent alkali ions by biva- lent alkaline earth ions has a insignificant effect on the etch rate and this is in keeping with the view that the 3D network structure is unchanged. This also supports the view that a hydrated silica film is formed by ion exchange. (iii) Introduction of other network-forming oxides (B203, A1203) at small concentrations lowers the etch rate. At larger concentrations the etch rate increases again [63]. The etch rate has been linked to the degree of linkage of the network structure and the concentration of SiO2 [21, 64]. Ostaf'ev et al. [653 linked the etch rate of multicomponent optical glasses to their annealing temperatures and the expansion coefficients, however, both of these properties are also related to the network structure and the SiO2 content of the glass. Phase separation, as occurs, for instance, in sodium borosilic- ate glasses, modifies the network structure and con- sequently changes the etch rate [66]. 5. Applications Applications of silicate glass etching are found in many areas of technology, as well as in daily life. In this section the most important applications are discussed. 5.1. Etch technology The most simple and widely applied way to etch glass objects is by immersion into the HF-containing solution. Although the glass etch rate is not limited by the diffusion from and to the glass surface, the homo- geneity and effectiveness of etching can be increased by stirring, bubbling and moving the object or by apply- ing ultrasonic agitation of the etchant. For planar substrates such as glass wafers the homogeneity of etching can be increased by rotating the substrate and spraying the etchant onto the rotating wafer [67]. This process has some advantages as compared to immer- sion etching. Fresh etchant is supplied and reaction products are continuously removed, which results in better process control and uniformity. Aerosol jet etching with HF solutions has also been reported and is used to increase the anisotropy of etching [68]. When a glass object has to be etched locally, the glass is partially protected by covering it with a material which is resistant to the etchant. This can be a wax applied manually, but when dimensions of pat- terns in the submillimetre or even submicrometre range have to be realized, photolithographic tech- niques have to be applied which are more appropriate to planar substrates. In IC technology, etching of SiO2 glass-based films is predominantly used to pattern the film, e.g. for contact hole etching [69, 70], in order to make contact with the underlying silicon. The glass film is partly masked, before etching, with a photoresist which is patterned using lithographic techniques. After etching the photoresist mask is removed, e.g. by dissolution in acetone. Due to the small dimensions, it is necessary to work in clean rooms using specially developed particle-free chemicals. In pure HF solutions delamin- ation of the photoresist occurs due to attack at the glass-photoresist interface. This can be avoided by using HF solutions which are buffered with NH4F to increase the pH to about 5-6 [26, 713. In ultra large scale integration (ULSI) of ICs the size of the contact holes is reduced into the submicrometre range. Consequently, the requirements for wet etching become very stringent [153. In order to apply the buffered oxide etch (BOE), the etchant has to be optimized with respect to the NH4F concentration in order to avoid the precipitation of (NH4)2SiF 6 at the SiO2 surface [15, 72]. Furthermore, surfactants have to be added, such as aliphatic amines and atiphatic alcohols [153. The photoresist delamination is a problem which occurs when long etching times or concentrated HF solutions are used. In such cases, a second HF- resistant film, such as chromium, can be applied as the HF etch mask. The chromium film is patterned with a photolithographic procedure and the patterned chromium film is, in turn, used as the etch mask for the glass [14]. This technique can be used to etch deep grooves into a borosilicate glass such as Pyrex (Fig. 8). 5.2. Silicate glass etching in IC technology The deposition of SiO 2 and doped SiO 2 thin films on silicon wafers is widely applied in IC technology [69, 70]. Their most important use is for the intermetallic dielectric isolation barrier between electrically con- ducting parts of the IC. Many methods have been developed to prepare such thin films. With the high temperature (800-1200 ~ oxidation of mono-crystal- 6267
`
`Raytheon2028-0007
`
`

`
`Figure8 Semicircular groove etched into Pyrex glass using a chromium mask which was structured using photolithographic means. line silicon in 0 2 or H20 , a SiO 2 film is formed on the surface of the silicon [73]. The etch rate of this so- called thermal oxide film is very similar to that of fused silica [74], which indicates a large similarity in struc- ture. The thermal oxide layer serves as an etch rate standard for thin films prepared by other methods. Techniques for the preparation of SiO 2 and doped SiO 2 films include various chemical vapour deposition (CVD) methods. When the process is performed at atmospheric pressure it is referred to as APCVD, similarly at reduced or low pressure it is known as LPCVD and when deposition is enhanced using a plasma the process is referred to as PECVD. For all these CVD methods many variant processes have been developed with respect to starting materials, pressures, additives, temperatures and equipment design. Other methods for SiO2 film deposition include sol-gel processing [75], sputtering [74, 76], E-beam evapor- ation [74] and anodic oxidation [77]. Overviews of these deposition processes are given in IC technology handbooks [69, 70]. It has to be mentioned at this point that in present- day IC technology with sub-micrometre dimensions, plasma etching processes have largely replaced wet- etching with HF-containing solutions. However, wet- chemical etching still remains a simple and cheap method to etch SiO2-based films in this field. 5.2. 1. Etching of as-deposited films Prior to annealing, as-deposited SiO z films are struc- turally and compositionally imperfect compared to fused silica or thermal SiO 2 films. These imperfections, which include non-stoichiometry, porosity and impu- rities, are related to the type of deposition technique and post-deposition processing [67, 68, 74, 78]. As- deposited SiO= films prepared by APCVD using Sill4 + O2 or via sol-gel routes are stoichiometric, but are also porous and incorporate ----SiOH groups and adsorbed H/O [74]. PECVD films can contain Si-H groups [79, 86], while reactively sputtered films can be oxygen deficient [80]. The etch rate is found to be sensitive to all these imperfections. In general they lead to a higher et