J. MATER. CHEM., 1994, 4( 3), 393-397 Silicon-Germanium Films for Photomasking Applications Chi-Wing Liu,* James A. Cairns, Roderick A. G. Gibson, Andrew C. Hourd, Brian Lawrenson and Charles F. Leecet Department of Applied Physics and Electronic and Manufacturing Engineering, University of Dundee, Dundee, UK DDI 4HN Si-Ge films have been deposited by RF co-sputtering and have been compared to Cr films for use as opaque materials on photornasks. One of the potential attractions of Si-Ge is its ability to be patterned by dry etch, anisotropic processing. Optical absorption coefficients were measured from 347 to 820 nm using a modified spectrophotometer. The absorption coefficient of Si-Ge films (containing 32.6% Ge) was 1.5x 1O5 cm-' whereas the 'standard' Cr gave 5.6 x lo5crn-'.A simple scratch test was used to compare the adhesion of the films to fused silica substrates. Specially fabricated cantilevers were used to study interfacial stress between films and substrates. Cr and Si-Ge with similar thicknesses exhibited tensile and compressive stress, respectively. These observations have confirmed the potential of Si-Ge as a photomask opaque. Most modern integrated circuits are manufactured by the use of photomasks to define the circuit pattern to be transferred to the wafer. As integrated circuits continue to become ever more complex, the materials requirements of modern photo- masks become increasingly demanding. Every photomask is unique and must be free from physical defects, even down to sub-micrometre dimensions.These requirements impose a strict specification on all the materials and processes used in their manufacture, which starts from a highly polished fused silica substrate coated with a thin (100nm) layer of chromium, followed by a layer of organic resist. The latter is subjected to high-resolution radiation, using, for example, a focused electron beam, which changes its chemical nature, thereby allowing it to be removed selectively to reveal an exposed chromium layer. This, in turn, is removed by a liquid dissolu- tion process. The resultant chromium-patterned fused silica is then used to make integrated circuits by passing ultraviolet light through its clear areas to expose a resist-coated wafer. Each of the steps described above has the potential for departure from the strict specification required. As integrated circuits increase in complexity, the control of the dimensions provided by the chromium photomask becomes critical, and highlights the limitations of using chromium as a photomask opaque.For example, the process of removing chromium involves wet etching, which is an isotropic (non-directional) process. This can cause lateral etching of the chromium, leading to loss in definition of critical pattern features. Furthermore, there is a significant stress between the chro- mium film and the underlying fused silica, which can cause distortion of the photomask, and hence the final image. The stress also leads to uncertainty in the ability of sub-micrometre chromium features to retain adhesion to the fused silica, especially during the vigorous cleaning to which the photo- mask is ultimately subjected.Although chromium can be dry etched (an anisotropic or unidirectional process) to give the desired features, with little or no lateral etching, this requires the use of chlorine-containing process gases, which are extremely corrosive and environmentally undesirable. Therefore there is a strong motivation to identify a new opaque material to replace chromium. This should be opaque to deep UV, exhibit good adhesion to fused silica with minimum interfacial stress, and be capable of being patterned by dry etch processing. In addition, it should exhibit ideally an absorbance >3 at 7 Present address: Compugraphics International Limited, Eastfield Industrial Estate, Glenrothes, Fife, UK KY7 4NT.633 nm (corresponding to an optical absorption cotrfficient, a=7 x lo5cm-l) in order to be compatible with existing inspection and alignment systems. It is well known that silicon (Si) can be processed by dry etching to produce patterns with good edge and feature profiles. However, its absorbance at 633 nm is not high enough to meet the optical requirement detailed above. This can be overcome by introducing other elements into silicon during its deposition onto the fused silica substrates. Germanium was chosen because it has good optical absorption and is chemically compatible with Si. In addition, it exhibits similar etching characteristics.The importance of this approach is apparent from the considerable activity in the patent litera- ture.' Previous work by Kao et aL2 has shown that the incorporation of Ge into Si can cause a significant increase in the absorption coefficient. This was done by placing a known area of Si segments onto a Ge wafer target.3 The Si-Ge films were deposited onto optically flat glass substrates by RF sputtering in a gaseous mixture of argon-5% hydrogen at a pressure of 6 mTorr, with the result that hydrogen was incorporated into the substrates. In the present investigation, films of Si-Ge were produced by RF co-sputtering from separate Si and Ge targets, in the absence of hydrogen. Experimental The fused silica substrates were cleaned using a neutral detergent (TeePol, BDH) followed by rinsing with deionised water and immersion in isopropyl alcohol in an ultrasonic bath for 10 min.They were then dried in nitrogen and finally heated to 90 "C in air for 30 min, before being cooled, ready for use. Si-Ge films were deposited onto the fused silica substrates, using an RF co-sputtering unit (RF Applications Ltd.). The substrates were secured to a water-cooled turntable nkounted above the sputtering targets (Fig. 1).Argon gas, at a pressure of 60 mTorr, was fed constantly at 20 sccmt into the chamber during deposition. The sputtering targets were first sputter cleaned in an argon atmosphere to remove oxide. Throughout this process, shutters were used to protect the fused silica from contamination.During deposition of Si-Ge films, the magnetron sources were at the following powers: 100W for Ge, 150 W for Si. This resulted in an overall deposition rate of 10 nm min-'. -f Standardcm3 min-'. I J. MATER. CHEM., 1994, VOL. 4 water-cooled turntable Fig. 1 Preparation of Si-Ge films by RF co-sputtering Optical measurements of the samples were performed over the range 347-820 nm, using an optical spectrometer (PYE Unicam). This instrument was modified to perform both transmission and reflection measurements. The optical absorp- tion coefficient, SI was calculated from 100-R 1 =In (7)-a dcm -1 where T and R are the percentage transmission and reflectivity, respectively, and d is the alloy thickness measured by surface profilometry (DEKTAK, Sloan Technology Corporation).A chrome blank (type AR3, Hoya Corporation) from which the resist had been removed to reveal a continuous film of chromium, was also measured as a representative comparison of a standard photomask. The adhesion behaviour of the films was investigated by a simple purpose-built scratch test device, shown schematically in Fig. 2. This accommodated the substrates under test mounted on an alignment table incorporating two micro- meters. A load was applied to a diamond stylus perpendicular to the substrate. The substrate was moved against the station- ary stylus by means of the micrometers and examined by optical microscopy. The applied load was increased in small steps until the Si-Ge or chromium films began to lose adhesion.A novel method of observing the interfacial stress between the opaque layer and the underlying fused silica was performed by using cantilevers of SiO, fabricated on an Si wafer by photolithography4 (Fig. 3). Reactive ion etching (RIE) of the Si-Ge films was performed in a custom-built etcher (RF Applications Ltd.) using a CF4-5% 0, gas mixture. load mond counterbalanceIlls substrate microme Fig. 2 Simple purpose-built scratch test device Fig. 3 Specially fabricated cantilevers with zero stress Results and Discussion Deposition of Si-Ge Films Samples of Si, Ge and Si-Ge were produced. The pressure in the chamber had a significant effect on their uniformities, e.g.the non-uniform area of the sputter films was clearly visible on square substrates with dimension of 7.6 cm side, at sputter- ing pressures below 30 mTorr. Above 30 mTorr, the non-uniformity was much less pronounced, being seen as an array of interference fringes. Visibility of these fringes could be enhanced by depositing a layer of chromium onto the fused silica prior to Si-Ge sputtering. Satisfactory uniformity was obtained only at pressures of 60mTorr (and above), but higher pressures carry the disadvantage of low deposition rate. For this reason 60mTorr was adopted as the standard sputtering pressure. Deposition rates of Si, Ge and Si-Ge, at various powers are shown in Fig. 4. The film thicknesses of samples produced at different power settings are all directly proportional to the deposition time.As expected, the depos- ition rate increases as the sputtering power is increased. The deposition rate of Si is 22% higher than that of Ge, for the same power (100 W). The Ge composition in the Si-Ge films at a given power ratio could be estimated from the deposition rates of Si and Ge. For example, when the power ratio is 150:100, the deposition rates of Si and Ge are 6.28 and 3.03 nm min-', respectively, resulting in the Ge composition of the film being 32.6%. This has been confirmed by electron- induced X-ray analysis (EDX). Optical Absorption Coefficients The transmission and reflectivity of sputtered Si, Si-Ge and chromium (from a standard photomask) were measured and t c --.E UJ a,c x0 ._s 0 10 20 30 40 50 60 70 deposition time/min Fig.4 Deposition rates: (a) Si (100 W), (b)Si (1 50 W), (c)Ge (100 W), (d)Si-Ge(150W:lOOW) J. MATER. CHEM., 1994, VOL. 4 their optical absorption coefficients plotted on a linear scale against wavelength as shown in Fig. 5(a). Fig. 5(b)shows the optical absorption coefficient plotted logarithmically against photon energy. The standard chromium (Cr) and sputtered Si have the highest and the lowest absorption coefficients, respectively. The Si and Si-Ge are capable of absorbing deep UV radiation used during fabrication of circuit devices, but neither may be compatible with existing inspection and align- ment systems, which requires absorption coefficients of ca.7 x lo5cm-I for films 100 nm thick. The absorption coefficient of Si-Ge films (containing 32.6% Ge) is 1.5 x lo5cm-I whereas the standard Cr gives 5.6 x lo5cm-l. By increasing the Ge content in Si-Ge films, higher absorption coefficients can be achieved; however, initial chemical stability tests suggest that Ge-rich films are highly soluble in aluminium etch solution (a mixture of nitric, acetic and orthophosphoric acids) and certain alkaline chemicals. This could damage the Si-Ge films during photomask or integrated circuit pro- duction should such etches be encountered during processing. However, as a dry (anisotropic) etch process is to be used, vertical walls and thus good pattern definitions should be possible in thicker films of Si-Ge, allowing the absorbance to be increased by this means. Interaction between deposited Films and Fused Silica Substrates The simple scratch test involved moving a coated substrate against a diamond stylus and measuring the load applied to the stylus when loss of adhesion occurred. The load needed to remove 100nm of Cr (defined as 100%) was compared with that necessary to remove various thicknesses of Si-Ge.For Si-Ge layers of similar thickness to Cr, 80% of the load was required. This increased to 99% as the Si-Ge thickness was reduced to 50 nm. However, films having three times the a Cr 6 4 2 SiGe\ Cr thickness exhibited extremely low adhesion to the sub- strates. The resultant scratch tests on standard Cr and Si-Ge films are shown in Fig.6 and 7, respectively. The Cr shows a well defined scratch track which seems to have been ‘scooped’ up by the diamond stylus without any evidence of flaking of the coating. For Si-Ge of a similar thickness, flaking of the film is evident and the scratch track is not smooth and well defined. The flaking is thought to be due to lower elastic strain energy in the Si-Ge films: instead of a continuous film being peeled, a ‘hammering’ effect is induced in the stylus which shatters the film into small fragments; therefbre the direction of the scratch is altered by the recoil of the stylus and the resulting scratch track is not straight and smooth. In order to observe interfacial compatibility between deposited films and fused silica substrates, specially fabricated cantilevers were used., Fig. 8 and 9 show a comparison of the behaviour exhibited by Cr and Si-Ge films of similar th tckness.In the former case, the pronounced upward bending of the cantilever demonstrates significant tensile stress, whereas in the latter case the cantilevers show only a small downward movement, indicative of modest compressive stress. Etching Behaviour of Si-Ge Films Si-Ge films coated with photoresist were patterned by using photolithography and reactively ion etched using CF,-5% O2at 70 mTorr, 50 W and 60 sccm. The etch rate was 200 nm Fig. 6 Simple scratch test on 100 nm thick Cr showing well defined -\scratch track 0 I I I Si 1 1.5 2 2.5 3 3.5 4 photoenergy/eV Fig.5 (a) Optical absorption coefficients (linear) as a function of wavelength. (b) Optical absorption coefficients (logarithmic) as a Fig. 7 Scratch test on 96 nm thick Si-Ge showing poorly defined function of photon energy. scratch track J. MATER. CHEM., 1994, VOL. 4 Fig.8 Upward bending produced by the tensile stress between Cr and the SiOz cantilevers Fig. 9 Downward bending of cantilevers of deposited Si-Ge on SO,, showing modest compressive stress min-'. Etch depths were recorded by means of a Dektak profilometer. In order to observe the edge profiles of etched Si-Ge, samples were prepared in a similar manner to that used for the films described above but on an Si wafer substrate rather than on fused silica.The testing patterns could be cleaved cleanly along the wafer crystalline planes to reveal their cross- sections. A thin layer of gold was then evaporated onto the wafer to dissipate charging effect when the sample was examined under a scanning electron microscope (SEM). Preparation of Cr edge profiles was performed by a similar method except that wet etching was used to define the pattern. Fig. 10 shows the resultant edge profile of the wet-etched Cr (an isotropic process). This process attacks the Cr layer equally in all directions and results in undercut of the resist protecting the Cr with consequent narrowing of the features. Fig. 11 shows a sharp and well defined edge profile produced by RIE of Si-Ge. The process can be seen to cause etching significantly faster in the vertical direction than in the hori- zontal, in accordance with the anistropic nature of the RIE technique.Discussion Si-Ge films have the potential to replace Cr as a photomask opaque. However, a number of aspects of this material may require further improvements and investigations in order to Fig. 10 Edge profile of wet-etched Cr (isotropic process) Fig. 11 Edge profile produced by RIE of Si-Ge (anisotropic process) satisfy the requirements of current photomask manufacturers and users fully. Optical, physical and chemical experiments need to be carried out on the influence of Ge composition on Si-Ge films. In addition, the measurements involving chemical solu- bility and the strain gauge cantilevers should be quantified.An optimum Ge composition should be established so that the Si-Ge films have maximum optical absorption coefficients at 633 nm, the wavelength used in inspection and alignment systems, while maintaining chemical resistance to metal and organic etches. It is well known that the hydrogen incorporated into amorphous Si and Ge films can alter the distribution of defect states, and change the optical gap. This can reduce the interfacial stress when Si-Ge is deposited onto the fused silica substrates, but the optical gap will have a tendency to increase, thus reducing opacity of the film. Note that when reactive ion etching of Si-Ge was performed in the presence of poly(butene sulfone) (PBS), a widely used electron beam resist, poor selectivity was observed between the two.However, this situation can be improved by exposure of PBS to oxygen5 or nitrogen6 plasmas. Conclusions This study has confirmed the potential attractions of Si-Ge as a photomask opaque. However, it has highlighted also a number of features which require further investigation. J. MATER. CHEM., 1994, VOL. 4 397 The work has been supported under the Joint European Sub- micron Silicon Initiative (JESSI) to whom the authors are grateful. Thanks are also due to N. Holmes, Compugraphics International Ltd., for his constructive comments on this manuscript and to the Electron Microscopy Group in Dundee 2 3 4 5 6 K. C. Kao, R. D. McLeod, C. H. Leung, H. C. Card and H. Watanabe, J.Phys. D, 1983,16,1801. H. Watanabe and K. C. Kao, J. Vucuum SOC., Jpn., 1981,24,417. R. Keatch and B. Lawrenson, this conference. W. M. Mansfield, J. Am. Chem. SOC., 1987,346,317. W. A. Loong and H. W. Chang, Electronics Lett., 1991,21. 541. University for their assistance in measuring film compositions. Paper 3/04293F; Received 21st July, 1993 References 1 W. I. Lehrer, US Put., 3,830,686, 1974; D. B. Fraser, US Patent No. 3,975,252, 1976; Fuji Photo Film Co. Ltd., Br. Put. 1,481,623, 1977.