WO1989003899A1

WO1989003899A1 - Etching process using metal compounds

Info

Publication number: WO1989003899A1
Application number: PCT/AU1988/000414
Authority: WO
Inventors: Christopher Max Horwitz; Mark Gross
Original assignee: Unisearch Limited
Priority date: 1987-10-23
Filing date: 1988-10-24
Publication date: 1989-05-05

Abstract

A method and apparatus for dry process etching wherein an rf plasma etching and deposition chamber (11) is provided with a metallic source (18) which is etched during processing of the target (24) to provide a source of metal ions in the discharge. Deposition of this metal on the surface of the target (24) during processing results in an etch at a controlled angle into the substrate.

Description

ETCHING PROCESS USING METAL COMPOUNDS

The present invention relates generally to the field of rf (radio frequency) techniques for use in integrated circuit manufacture and similar microfabrication techniques. In particular the present invention provides an apparatus and method for improved dry process etching. Vacuum etching processes are generally a mix of two types of etching: directional and isotropic. Modern high-density circuitry can tolerate very little isotropic etching due to line width loss, but fully directional etching often results in poor step coverage in overlaying layers.

The prior art has shown that the angle of an etch can been controlled in five ways:- (a) Isotropic ("plasma") etching combined with directional (anisotropic, reactive-ion etching) etching can give good control of etched profiles in materials which are easily etched by isotropic (purely chemical) processes. However, these chemical processes are strongly material-, temperature-, and gas-dependent with consequent careful process adjustments required.

(b) Directional etching can be combined with controlled erosion of a sloping mask sidewall to yield a sloped etch profile. The most common approach uses heated photoresist as a mask, but requires careful process control to retain consistent resist "rounding" under heat, and also requires poor etch selectivity since the resist must be etched. However, this process is suitable for Si0₂ films which are not easily etched by isotropic processes.

(c) A variant of the preceding method operates with an elevated substrate temperature during the etch, resulting in resist swelling, and narrowing of the etched apertures during the etch process. The result is again a sloped sidewall, with strong dependence on photoresist and etch conditions. An advantage of this process is that high etch selectivity can be obtained. (d) An undercut trilevel resist structure, combined with a high-pressure (partly isotropic) etch, has given good angled etching in Si0₂ substrates. At least two etch steps are needed in this process.

(e) The mask can be coated during the etch, resulting in narrowing of the etched aperture and an angled sidewall. The coating (and mask) must either be desired components of the circuit under construction, or be easily removed.

(f) Angled ion beams have also been used with some success.

The applicant has found that by adding various metals to the discharge plasma, an etched angle which is roughly proportional to the quantity of metal injected will result. The observed angle is substantially independent of the material.being etched, and in addition is often independent of the selectivity of the etch process. It is shown that low pressure operation is desirable for good uniformity of the etched angle, and that the substrate temperature should be maintained at a low value. Integrated circuit processes require a mix of film control capabilities. For instance, isotropic film deposition combined with directional (or an-isotropic) etching is required in the LDD (Lightly Doped Drain) field effect gate structure. On the other hand aluminium metallisation will not reliably make contact through a directionally etched contact (or "via") hole; sloped etch sidewalls are required. In addition an isotropic or angled etch component is required to prevent "stringer" formation during polysilicon gate etching which could short out devices. The present invention provides an apparatus and method for forming etched angles without dependence on the etched material, the mask layer or selectivity of the etch, or on mask sidewall buildup. According to a first aspect of the present invention there is provided a dry process etching chamber comprising: a vacuum chamber; at least one pair of radio frequency electrodes disposed within said chamber; at least one source of radio frequency potential connected to said electrodes; a metallic source within said chamber; inlet means for introducing a gas into chamber, which gas becomes chemically reactive when ionised by a radio frequency discharge so as to transport metal from said metallic source and perform an etching process upon target material positioned on one or more said electrodes.

According to a second aspect of the present invention there is provided a method of dry process etching in a vacuum chamber comprising the steps of: positioning target material on a radio frequency electrode, said electrode being one of a pair of radio frequency electrodes positioned within said chamber; providing a metallic source; applying radio frequency potential between said pair of electrodes; and introducing a gas into said chamber, which gas when excited by a radio frequency energy, forms a chemically reactive plasma, said plasma transporting metal from said metallic source and performing an etching process on said target material.

In different embodiments of the invention the metallic source may take various forms. For example, the metallic source may be a metallic target which is etched by the reactive plasma, or alternatively the metal may be introduced as a component of the gas introduced into the chambe .

A variant of this method allows etching of an asymmetric profile which is suitable for the construction of blazed diffraction gratings. Opposing groove angles can be formed at 90° and 20° from the substrate surface, allowing the blazing of holographic master gratings with high precision.

In order that the nature of the invention may be Q.» better understood experimental results are described hereunder with reference to accompanying drawings in which:-

Fig. 1 diagrammatically illustrates mechanisms of angle control using a surface coating; 5 (a) Conformal surface coating,

(b) Effective coating thickness along ion etch direction is larger on angled surfaces,

(c) At the angle where effective etched thickness equals deposited thickness in any given time period, 0 an equilibrium state is reached;

Fig. 2 illustrates the effect of upper target metal on etched angles in Si0 and Si at 1.7kVp-p in a 2.2 Pa, 0.006 PaM 3s-1CF. discharge, where samples are heatsunk and have a O.δu top layer of thermal SiO- and 5 the MgF₂ mask removed, (a) Si, (b) W, (c) Cu, (d) Ta,

(e) Mo, (f) Al.

Fig. 3 illustrates the effect of gas chemistry on etched angles at 1.7kVp-p with a Mo upper target, where samples are heatsunk and have a 0.6um top layer of thermal 0 Si0₂ and the MgF₂ mask removed, (a) CF, and (b)

Cl₂ at 2.2 Pa, (c) SF_g at 1.1 Pa pressure.

Fig. 4 schematically illustrates methods of incorporating a metal source in hollow cathode discharges.

(a) metal source at chamber side, (b) Metal source on 5 opposing-target, (c) Small metal chamber electrode, and (d) Trielectrode hollow cathode;

Fig. 5 schematically illustrates a prior art sputter etching chamber wherein the rf electrode is modified to form a hollow cathode; Fig. 6 schematically illustrates an apparatus embodying the present invention for the fabrication of blazed optical diffraction gratings (Distorted sheath hollow cathode grating etch system) .

Fig. 7 illustrates SEM photographs of etched profiles for various values of the sample position d from the upper target upright and various widths of exposed aluminium on the upper target of Fig. 6;

Fig. 8 graphically illustrates the dependence of the grating blaze angle (i.e., the angle of the larger sawtooth face from the horizontal) on the distance d from the target upright, with the exposed aluminium width as the parameter;

Fig. 9 graphically illustrates the etch rate of Si0₂ as a function of the sample position d for various widths W of exposed aluminium;

Fig. 10 illustrates a SEM photograph of an etch profile of a 0.5 micron period grating. Some Cr mask remains on the tops of most of the grating lines;

Fig. 11 illustrates simulated etch profiles of various ion incidence angles. The metal widths quoted are calibrated from =2-6 in Fig. 8 with the metallic compound coating rate assumed to be linearly proportional to W;

Fig. 12 graphically illustrates the simulated blaze angle dependence on d corresponding to Fig. 8. The ion incidence angles were determined from the results of Fig. 7 for =0, with the metal compound coating rate assumed to be proportional to the target metal width ;

Fig. 13 illustrates simulated etch profiles for various values of the deposit coating rate D. The deposit etch rate and substrate etch rate are constant at 0.5 and 2 units/cycle, respectively. Each profile is calculated over 10 cycles;

Fig. 14 illustrates simulated etch profiles of grooves with different widths after (a) 10 cycles (b) 18 cycles (c) 25 cycles;

Fig. 15 illustrates simulated etch profiles of a previously existing groove as a function of etch time, (a) Initial trench, (b) After 8 cycles, (c) 50 cycles, (d) 70 cycles; Fig. 16 illustrates angle etch profiles for various values of the ion incidence angle and deposit coating rate D,

Fig. 17 graphically illustrates the expected variation of the net effective substrate etch rate as a function of the sidewall angle for several assumed values of the angular deposit etch yield;

Fig. 18 graphically illustrates the experimental variation of the etch rate of a Si0₂ substrate as a function of the sidewall angle. The theoretical curve corresponds to the best fit of the data to Equation (2); Fig. 19 graphically illustrates the experimentally determined yield function of the deposit;

Fig. 20 illustrates a SEM photograph of etch profiles for angled etching grooves of different widths; Fig. 21 illustrates SEM photographs of the angled etch of a previously etched groove, (a) First angled etch, (b) After second angled etch, (c) After cleaning with acetone.

Fig. 22 illustrates a SEM photograph of a profile etched in Si0₂ on Si at an ion incidence angle of approximately 25°; and

Fig. 23 illustrates (a) an experimental etch profile for 10 ion incidence angle showing slowly curved sidewalls and tilted groove bases, (b) a simulated etch profile for an ion incidence angle of 10 , under the conditions similar to Fig. 16, (c) the same profile as in Fig. 23(b) except that 2% of the etched material is redeposited on the groove sidewalls.

The coating layer deposited from the etching discharge is responsible for angle control. This coating is a thin film which is continuously re-etched from the substrate surface, and so remains•as a surface monolayer. Later processing can easily remove this thin film in a variety of ways, if necessary. The present invention discloses the formation of such angle-control films containing aluminium and molybdenum and other metals with CF. , SF_β or Cl discharges. These films appear to have little effect on, and are little affected by, the "polymer" films which are also present in typical etching discharges and which are responsible for etch selectivity between Si0₂ and other materials. Thus these metal-containing films allow independent control of etched angles and of other discharge parameters such as etch selectivity. From the range of metal and gas combinations which have been found to give good etch angle control it is expected that any halogen-containing gas combined with most metals will give good results if their product compound does not have a high vapour pressure. In experimental tests two types of etching systems were used. Limited work was performed in a planar sputter etching system with inserts of metal placed near the substrates, on the target electrode. Most work was then carried out in a hollow cathode system with metal inserts placed on the Si targets. MECHANISMS

Etched angles must be controlled in many applications. One example is in via hole and contact hole etching, where angles of 10 - 20 from the vertical permit good metal coverage (J.S. Chang, Solid State

Technol. 27, 214 (April 1984)). Here we discuss a method of etched angle control employing a metal compound of low volatility which is coated onto surfaces while they are etched. This metal compound is normally present only in monolayer quantities and is compatible with standard device processing. As will be seen the process is applicable to a wide variety of etcher configurations, although it has mostly been used with the hollow cathode.

If a conformal uniform-thickness layer is being continuously applied over a surface which is also subject to directional ion-assisted etching, horizontal areas will be etched to completion faster than angled areas, where the effective layer thickness is greater (Fig. 1) . Defining the angle between the substrate and the incident ion direction as θ, the effective layer thickness is proportional to l/sinθ, hence rises as the angle θ falls. At large angles θ (i.e. for horizontal surfaces, normal to the ion direction), this coating typically etches rapidly and so does not greatly subtract from the substrate etch rate. However, at a sufficiently small angle θ to the ion direction, the effective layer thickness will be so large that the substrate etch rate is reduced to zero. At this angle the conformal deposition rate of this layer is precisely equal to its directional, etch rate, resulting in the formation of a stable, very thin equilibrium layer. At this equilibrium angle no further etching of the substrate can take place.

The above mechanism implies that the etched angle is dependent on the nature of the coating material and on its deposition rate, but not on the material being etched. This surprising result is borne out by experiment, where both Si and Si0₂ layers are etched to very nearly identical angles in the presence of metallic compound deposition. Another useful attribute of this angle-control mechanism is its independence of other discharge parameters, allowing independent control of etch selectivity and etched angles. It has been found that a wide variety of metal and gas combinations result in useful angle control properties, as illustrated in Figs. 2 and 3; in general the metal compound must have a low vapour pressure to have a significant angle-control effect, but not so low that its range in the discharge is small, affecting uniformity.

Metal compounds may be added to the discharge gas in a variety of ways. If the operating pressure is low, or the compounds in use have a moderately low vapour pressure (as with Mo-F combinations), a source of metal from a subsidiary discharge 30 at the side of the etching chamber will result in good angle uniformity (Fig. 4(a)). Sources of metal closer to the etched substrates permit a wider choice of materials and operating pressures; Fig. 4(b) shows that an opposing hollow cathode target may have a metallic insert 31. However this permits etched angle control only through target changes. Another relatively inflexible- method relies on a small (chamber:target) area ratio, causing ion bombardment of the chamber to a lesser extent than the target, but nevertheless sufficient to sputter a controlled amount of chamber metal material into the discharge (Fig. 4(c)).

The "trielectrode" hollow cathode (Fig. 4(d)) has three main electrodes; central targets 32, metallic ring targets 33, and the chamber walls 34. This configuration provides a source of metal 33 close to the targets, while permitting etched angle control by adjustment of the metal ring target discharge current. These ring targets also contribute to discharge confinement and hence to discharge power efficiency. Although the rf generator 35 is shown connected only to the central targets 32, and the dc supply 36 is only connected to the ring targets 33, in fact a fraction of the central target rf voltage appears on the ring targets 33. External capacitors can be connected between the ring targets and the chamber or central electrodes to adjust the value of this ring rf voltage; a high ring voltage results in low input power density for a given etch rate, but also in a high rate of metal supply to the discharge.

The hollow cathode system used for some experiments is shown schematically in Fig. 5. Fig. 5 shows a hollow cathode structure 15 placed on a stainless steel rf electrode plate 10. The rf electrode 10 is mounted in an earthed chamber 11, the electrode being supplied with rf power via an rf input 12 and the substrates 24 to be etched being located on the target surface 16. The reactive gas is supplied to the chamber through an inlet 14 in the top of the chamber. The lower target 16 rests directly on an aluminium block 17 which in turn rests on the rf electrode 10 while the upper target 18 is held against an aluminium disc 19, which forms the upper electrode of the hollow cathode structure 15. This disc 19 is bolted to aluminium pillars 21 which are in turn bolted to an aluminium ring 22 resting on the rf electrode 10. A quartz ring 23 serves to confine the discharge and in the present embodiment was placed on the rf electrode 10, allowing a 10mm gap between the ring and the upper target. The quartz ring 23 could be replaced with an anode surface of aluminium in some experiments (connected to the chamber) and in others was not used.

Similar results were obtained in both the planar and hollow cathode systems, however, the high etch rates obtained in the hollow cathode made it the preferred machine.

Etch rates, when needed, were obtained from stylus traces over etched features using a DEKTAK IIA profilometer. All etched angle measurements were obtained from silicon wafer substrates cleaved perpendicular to etched grating pattern lines. Quartz (Si0₂) films were prepared by thermal oxidation of silicon. Polycrystalline silicon wafer surfaces have also been etched with identical results; the crystalline properties of the underlying wafer have no effect on the etched angle obtained. Grating mask patterns were formed by Cr or MgF₂ lift-off after trilevel photoresist processing.

RESULTS

A. Si targets. With a pure Si target or targets in a CF. discharge, directional etching without angle control is obtained, however, profile control can be obtained by introducing an aluminium anode surface near to, and confining, a hollow cathode discharge. Such a confining electrode will etch, albeit at a lower rate than the target electrodes, if the electrode: arget area ratio is suitably small. The etched aluminium is able to form an angle control film on the target substrates, resulting in an etched angle independent^' of selectivity on cool substrates. On non-heat sunk substrates such etched angle control is rarely observed, indicating that the metal containing deposit only forms, or only has an angle control effect, at low temperatures. A target temperature of 100°C leaves the angle unchanged so fairly high temperatures may be present in the non-heat sunk cases. B. Si Target with Mo metal insert. With a piece of molybdenum metal placed on a silicon target in a simple planar sputter etch system, a CF₄ discharge results in angled etching. Target coating with a dark-coloured water-absorbing film occurs when large areas of molybdenum are placed on the target indicating that compounds such as MoFg, MoOF. may be present and related to the observed angle control.

C. Mo inserts in a hollow cathode system. Small Mo clips installed in a Si hollow cathode etcher have given dramatic angle control effects. The clips were installed on an upper electrode and caused angles which varied with CF. gas flow rate to be etched in the substrates. Without the clips present, or with uncooled substrates, no angle control was evident. Otherwise, higher CF₄ flow resulted in larger angles away from vertical being etched, up to about 45 from vertical in both Si and in Si0₂. These etched angles were roughly independent of gas pressure (between 5 and 20 Pa) and of applied radio frequency voltage (between 700 and 1000 Vp-p) . D. Opposing Si and Si Al targets. Useful results have been obtained with a hollow cathode system comprising 100mm dia electrodes, one of which was a 100mm Si-wafer, the other a 76mm patterned Si wafer surrounded by an Al ring. An etched angle on the 76mm wafer of about 10° away from the vertical (i.e., 80° angle) was obtained in Si0₂ and resulted in zero metallisation step coverage failures in about 20,000- contacts, indicating satisfactory angled etching on real device wafers. Yield and long term reliability tests on CMOS devices have given excellent results. The etch gas was a selective CF./H₂ mixture, followed by a CF. or SF_β post-etch cleanup.

In addition, a similar composite electrode composed of a lower 76mm dia-patterned Si wafer and an upper 100mm Si wafer partially coated with Al has given similarly good angle control. This method suffers, however, from the need to periodically recoat the upper Si wafer.

We have found that Al, in contrast to Mo, has a fairly short "lifetime" in the discharge. This is evidenced by the ability of Mo to produce similar angle control up to at least 20 Pa pressure. Al begins to show greater angle slopes facing the source of Al than facing away from it at a CF. pressure of 1.5 Pa; at 4Pa pressure the Al angle control fails to uniformly reach 50mm from an Al source ring. E. Ring Electrodes Around a Main Target. Aluminium rings placed around the target electrodes in a hollow cathode system have given angled etching with the angle controlled by the ring power applied. The rings were installed similarly to the grounded "guard" electrodes in a normal sputtering apparatus, and so some capacitance existed between target and ring. This capacitance resulted in some radio frequency power being diverted to the rings and this alone induced some etching of the ring and angle-control film formation. This effect was enhanced with additional power coupled directly to the ring at a different radio frequency from the target frequency of 13.56 MHz. Frequencies of lower than 11.5 MHz were found to work well; closer frequencies to 13.56 MHz resulted in excessive coupling, through the matching networks, between the two rf generators. It is expected that frequencies suitably higher than 13.56 MHz would work equally well.

We have found that a DC power supply is simpler to operate than .a second rf generator and works equally well. We couple the DC power supply to the ring electrode(s) through a radio frequency choke and operate at 1.5Pa CF. pressure with hydrogen added for selectivity.

Additional advantages of. the ring electrodes described above are that they can be self-cleaning, preventing the polymer buildup often experienced in selective Si0₂ etching processes; a reduction in the target radio frequency power required for a given etch rate is observed; and the uniformity of etching can be adjusted with control of the radio frequency coupling between target and ring.

F. Asymmetric Angle Formation and Blazed Gratings. The "undesirable" asymmetry in etched profiles sometimes noted with Al metal sources or at the edges of the target electrodes can be used to advantage in the fabrication of blazed optical diffraction gratings. Such gratings yield excellent spectral purity if produced by holographic means, but holographic processes do not easily yield the asymmetric "blazed" profiles required for high diffraction efficiency.

Many methods have been employed to combine the accuracy of a holographically recorded diffraction grating pattern with the diffraction efficiency of a mechanically ruled grating. The mechanical ruling process forms asymmetrically angled flat surfaces which directionally reflect light at preferred "blaze wavelengths". However, groove period anomalies ("ghosts") decrease the optical signal-to-noise ratio of ruled gratings. Interferometric ruling control is one approach to ruled ghost reduction (T. Harada, T. Kita, M. Itou and H. Taira, Nuclear Inst. and Methods A246,272 (1986)).

Other methods improve on the limited diffraction efficiency of holographic gratings recorded in photoresist (I.J. Wilson, L.C. Botten and R.C. McPhedran, J. Optics (Paris), 8,217 (1977)) with exposure by asymmetrically angled laser beams (N.K. Sheridon, Appl. Phys. Lett. 12, 316 (1968)), or by multiple interfering beams (S.J. Zhu, in Proc. SPIE, Vol. 339, Edited by P.J. Rogers and R.E. Fischer p. 329 (1983), SPIE, Bellingham, WA -and- M. Breidne, S. Johansson, L.E. Nilsson, H. Ahlen, Optica Acta 26, 1427 (1979)). All of these techniques yield blazed gratings with excellent first-order diffraction efficiency. The developed photoresist profiles follow the incoming light intensity distribution, and so have rounded profiles. This rounding results in poorer high-order diffraction efficiency, limiting the applicability of such gratings (e.g., see Breidne et al) . A blazed profile may be formed using angled ion beam etching (L.F. Johnson and K.A. Ingersoll, Appl. Phys. Lett. 35,500 (1979) -and- S. Somekh and H.C. Casey, Appl. Opt. 16,126 (1977)). However, such etching develops an asymmetric rounded profile from the rounded photoresist original, and so is again limited in available high-order diffraction efficiency. In addition, these methods are dependent on ion flux and photoresist exposure uniformity, often resulting in efficiency variations across gratings (J.M. Lerner, J. Flamand and A. Thevenon, SPIE Vo. 353 - Industrial and Commercial Applications of Holography, p. 68 (1983) SPIE Bellingham, WA) . The hollow cathode permits high-rate, low pressure, high efficiency processing (CM. Horwitz, Appl. Phys. Lett. 43,977 (1983) -and- CM. Horwitz, S. Boronkay, M. Gross and K. Davies, J. Vac. Sci. Tech. A6,1837 (1988)). In this hollow cathode, a combined reactive etch and metal compound deposition process enables accurate control of the etched angle in a wide variety of materials (CM. Horwitz, Appl. Phys. Lett. 44,1041 (1984)). The above symmetric etch process can be modified by varying the ion incidence angle to yield asymmetric blazed profiles. One angled ion incidence method angles the substrate in a field-free target region, but is limited in substrate size (G.D. Boyd, L.A. Coldren and F.G. Storz, Appl. Phys. Lett. 36,583 (1980)). Another method angles the ion beam by field distortion (e.g., at the edge of a target - (H.W. Leh ann and R. Widmer, J. Vac. Sci. Tech. 15, 319,

(1978))). In this method the substrate size is again limited, this time by the area of usefully distorted field. However, substrates can be scanned past this area to process a larger area. This paper describes such a scanned method, applied to a field-distorted, controlled-angle hollow cathode etch system. The experimental results are compared with our angle-control theory and the possible limits to the method are discussed. Four blazing methods have been used here: (a) the asymmetric metal source mentioned above, which has yielded opposing face angles from the substrate surface of roughly 70° and 90 ;

(b) angling of the substrate in the discharge, yielding opposing angles of about 20° and 90° (over a limited area) ;

(c) provision of a metallic step on the target near the etched region of the grating which ^'we believe both acts as a directional metal source, and bends the discharge "dark space" with a consequent change in the directions of incident bombarding ions from the discharge;

(d) provision of an upper target support 54 which bends the discharge dark space as in (c) (Fig. 6).

Blazed gratings with well-defined facet angles can thus be etched in such modified hollow cathode reactive sputtering systems. Electric field distortion and metal compound deposition are combined to give asymmetrically etched, controlled-angled structures. The grating blank is scanned past the active discharge area to form a large-area sample with uniform properties. Computer models of the etch and deposition processes agree with our experimental data in showing that blaze angles of about 25° with apex angles of less than 90 are possible using this method.

In Fig. 6 there is shown schematically a preferred apparatus which employs method (c) above. Again a hollow cathode geometry is used. A vacuum chamber 40 is provided with a reactive gas inlet 41 and a gas outlet 42 adapted for connection to a vacuum pump. An rf generator 43 supplies rf power to the stepped aluminium target electrode 44 and to the grating blank 45 which is being etched, via a matching network 51. Silicon target pieces 46, 47 and 54 are mounted on the target electrode. The second target piece 47 is provided with a slot 48, located near to upright electrode 54, under which the grating 45 is moved in the direction of arrow A. A guard electrode 49 is also provided to minimize etching of unwanted target areas and this guard electrode is held at the same potential as the chamber 42.

The apparatus shown in Fig. 6 was used with CF, etch gas and a 13.56 MHz rf generator 43 to etch gratings on blanks 45 mounted in a moveable slide 52. The slide 52 rests in electrical contact with the lower cathode electrode 54 which is insulated from the remainder of the chamber wall by insulators 55. The grounded guard electrode 49 ensures that the only surfaces exposed to energetic ion bombardment are the Si target surfaces 46 and 47, the grating blank 45, and the (Al) metallic area 53 of width W on the upper target block 44. An idea of scale can be obtained from the 30mm height of the vertical Si-coated target surface 54. This surface acts to bend the discharge dark space electric field lines, resulting in angled ion incidence on the grating surface 45, especially at small distances d from the vertical surface. The 60mm width of target area was designed to accommodate a 50mm-wide grating blank with good uniformity. The grating blank 45 would normally be Si0₂-coated low expansion glass, with metallic Cr masking. However, in this study, most of our measurements have been made on Cr-masked oxidized Si wafers, which are easily cracked for SEM analysis. Fine patterns were prepared holographically using a multiple liftoff process (E.H. Anderson, CM. Horwitz and H.I. Smith, Appl. Phys. Lett. 43,874 (1983)); the coarser patterns of 4 micron period were prepared by contact trilevel lithography and Cr liftoff. A matrix study was performed at 1.5kVp-p target voltage and 4 Pa CF. pressure, with a 0.035 Pa m 3s-1 gas flow rate. Both the metal width W (controlling the average sidewall angle) and the distance d (controlling the profile asymmetry) were varied. With W=0 directional etching is obtained, resulting in parallel etched sidewalls (Fig. 7, top). However, the angle of these parallel etched sidewalls follows the ion incidence angle, resulting in highly sloped etching at the edges of the exposed target region at d=2mm. The addition of metal to the discharge forces a slope onto all etched sidewalls which results in blaze formation close to the d=2mm target edge. The dependence of blaze angle on distance d is shown in Fig. 8, with metal width W as a parameter. For these conditions, a blaze angle between 25 and 40° can be formed with good process tolerance. Smaller blaze angles require very small values of d which in turn yield very small etch rates (Fig. 9). This introduces the possibility of surface coating rather than etching, with consequent process control difficulties in.this instance. Our apparatus exhibits varying properties as Al metal width W is increased. While maintaining constant target voltage (1.5kV p-p) , the input power to the discharge rises from 53 Watts (for W=0) to 58 Watts (W=6mm) , 88 Watts (W=16mm) and 138 Watts (W=26mm) . These large power increases may be related to the nonmonotonic variation of angle with metal width W which we observe in Fig. 8, in contrast with what is predicted by a simple model described hereinafter.

The above work was performed with a nonerodable mask to clarify our etch processes. Such a mask is unsuitable in actual grating fabrication since it would give an unacceptably high efficiency loss to zero-order reflection. The profile of a 0.5-micron period grating, scanned through an etching slit and with conditions adjusted to erode the mask during etching is shown in Fig. 10. The profile shows a small mask residue but is acceptably blazed and demonstrates the validity of this approach.

Previously obtained angled etchings are well described by a model assuming isotropic metal-compound deposition combined with directional ion-assisted etching. It has been shown in previous work that tilting the ion incidence angle produced asymmetric "sawtooth" profiles of the type required for blazed grating formation.

Some examples of simulated profiles are shown in Fig. 11 for various ion incidence angles and metallic-compound coating rates. These simulations assume that all other etching parameters are held constant and that the coating rate is proportional to the width of metal, W, exposed on ^•the upper target in Fig. 6. The simulations compare well with the SEM results in Fig. 7. Fig. 12 shows how the simulated data compares with the experimental data of Fig. 8. For this graph the ion incidence angles were determined from the sidewall slopes in the case of W=0mm in Fig. 7. The exposed metal widths, W, were related to the metallic compound coating- rate by calibrating the simulations against the experimental results in Fig. 8 for W=0 and W=26. It can be seen that for increasing W one expects the blaze angle to decrease for any given position, d, from the target upright. Furthermore, the curves for each value of W should not overlap. However, as observed in the previous section, the total discharge power increases by almost 40% between W=6 and W=16. It is most likely that the etching conditions between W=6 and W=16mm are sufficiently different to account for the difference between the experimental and simulated results in this case.

The machine described above allows precise blazed grating formation from a non-blazed grating pattern. A combination of electric field distortion and metallic compound deposition during the etch process allows blaze angles of between 25 and 40° to be produced. Smaller blaze angles require longer process times and are more troublesome to control due to the competitive coating process. A simple computer model agrees well with our observed etched angles.

At present a 50mm x 50mm grating has an etch time of roughly half an hour to form a blaze angle of 25 , and no significant problems in scaling this process up are envisaged.

MODELLING OF SLOPED SIDEWALLS FORMED BY SIMULTANEOUS ETCHING AND DEPOSITION In order to model the basic process of angled etching we assume that two competing processes take place.

1. Continuous formation of a conformal film.

2. Ion-enhanced directional etching of this film and the underlying substrate.

A conformal film could result from isotropic deposition from the etch gas followed by surface migration of the film. The source of the deposited material could be from target etch products, or from a deposit-forming etch-gas mixture. Such films have been observed during trench etching of silicon (M. Sato and Y. Arita, J. Electrochem. Soc. 134, p. 2856, ^" (1987)) . Ion-enhanced etching involving highly directional ion bombardment is a characteristic of RIE.

The thickness of a conformal film is the same everywhere when measured in a direction normal to the substrate surface (Fig. 1(a) . However, if the film thiclcness is measured along the direction of etching, then on surfaces sloped with respect to this direction the thickness of the conformal film will appear greater than on surfaces normal to the etching direction, as shown in Fig. 1(b) . This leads to the concept of an effective deposition rate which is measured along the direction of etching. From Fig. 1(c) this rate can be seen to equal ϋ/sin(θ), where D is the conformal film deposition rate and θ is the angle of the surface with respect to the direction of etching. The equilibrium condition is defined when the etch rate of the deposited film, E,, is equal to the effective deposition rate, that is,

In this case no net change in the surface will occur. Thus, from a mask edge, an initially flat substrate will develop a surface having a slope θ = sin^" (D/E,) away from the etching direction. Following from this, several points may be noted. The first is that in order for large etched angles θ to develop under this scheme, the film deposition rate must be a large fraction of the film etch rate, suggesting that the overall substrate etch rate may be reduced compared with the rate for small values of θ. Second, the process involves continuous removal of the deposited film on all etched surfaces and hence roughly a monolayer of deposited film would be present at any one time. Finally, the angle is determined only by the etch and deposition rates of the conformal film and should therefore be independent of the type of substrate.

SIMULATION A computer program has been written to simulate the above process. Initially, we make a number of simplifying assumptions:

1. The deposited film is conformal and independent of the developing surface profile.

2. the etching is monodirectional and proceeds in a direction parallel to the ion angle of incidence. 3. The properties of the deposited film are independent of the type of substrate. 4. There is no dependence of the etch rate on the angle of the incidence of the ions.

As with similar simulations (H.W. Lehmann, L. Krausbaer and R. Widmer, J. Vac Sci. Tech. 14, p. 281, 1979 -and- J. Maa and B. Halon, J. Vac. Sci. Tech. B4, P822, (1986)) we assume that the processes of directional etching and conformal deposition, although simultaneous, can be calculated separately and then added to produce the net result. One complete process cycle consists, therefore, of conformal deposition of a given thickness, followed by etching of the film and substrate at given rates. The process is completed after a given number of cycles. Some examples of simulated etch profiles in a single material substrate are shown in Fig. 13. The surface is described by the line segments joining adjacent points in an array, which are moved along their normal direction by an amount corresponding to the deposited coating rate and then vertically by the deposit and/or substrate etch rate. The substrate etching time is measured in units of one such deposition and etch cycle. In order to minimize computation time, the number of points was chosen to be the minimum needed to give a satisfactorily smooth profile, and was usually 30. The choice of substrate etch rate was determined by step size considerations and was set at 2 units/cycle. Since only the ratio of the deposition rate to the deposit etch rate determines the sidewall slope, the absolute values of these parameters are unimportant. In the absence of any experimental values the deposit etch rate was set at 0.5 units/cycle. The simulations clearly show an increase in sidewall angle with increasing deposition rate, as expected, and that the ratio of the deposition to the etch rate of the deposited film is equal to the sine of the sidewall angle.

In Fig. 14, the development of a structure having two unmasked areas of different width is shown. It can be seen that the smaller groove ceased etching after reaching its equilibrium surface after 18 cycles, while the depth of the larger groove continues to increase. The smaller groove maintains its equilibrium state without further net coating or etching.

In Fig. 15 the etching of a surface that was initially a trench with a slightly sloped sidewall is shown at various times in the process sequence. The etching and deposition conditions were chosen so that a sidewall angle considerably greater than the slope of the initial sidewall should develop. At the base of the trench the substrate is etched at the desired angle, while the trench sidewalls are coated with deposit. In time the trench fills to the point where a new equilibrium surface, with the larger slope, is formed from the mask edge.

The above examples have assumed that the initial substrate surface is normal to the direction of ion incidence. The effect of varying the angle of ion incidence on an initially flat substrate, is shown in Fig. 16 for various etch and deposition conditions. The etch time was sufficient in all cases for the grooves to reach their equilibrium surfaces. The pronounced asymmetry of the etched grooves is evident, in some cases resembling saw-tooth grating profiles.

An important observation from the results of Fig. 13 is that the effective substrate etch rate along the direction of ion incidence decreases with increasing deposition rate. The amount by which the substrate etch rate is reduced due to deposition of the film material can be quantified. In the simple case where the deposit etch rate and the substrate etch rate are the same, the net substrate etch rate E.n = (^vEo - D)' , where Eo_Λ is the substrate etch rate in the absence of deposition. In the more general case where the deposit and substrate etch rates are different, E may be written as

hence using Equation (1),

E_n = E_o(l-sin(θ)). • (2)

Thus, as the etched angle θ increases, the net effective substrate etch rate must decrease.

Generally, we do not expect experimental agreement with Equation (2) since the etch rates always have some dependence on the angle of incidence of the ions. This depedence is not straightforward, being a function of the material being etched and the type of etching (physical sputtering, or reactive sputter etching) . If our previous assumption of constant_^etching yield as a function of angle of incidence is relaxed, we must modify Equation (1) since the deposit etch rate E, will then vary with the sidewall angle. If the etch yield is f(θ), such that E,(θ) = E,(0).f(θ), where E,(0) is the deposit etch rate on a surface normal to the ion direction, then Equation (1) becomes

D = sin(θ)

E_d(0)f(θ) (3)

and hence

^En ⁼ E₀(l-f(θ)sin(θ)) (4)

Equation (4) provides a measure of the expected reduction in etch rate for a given angle, and is plotted in Fig. 17 for some previously published values of the etch yield (which have been normalized to the same etch rate at 6=o°) (S. Somekh, J. Vac. Sci. Tech., 13, p. 1003, 1976 -and- T.M. Mayer, M.S. Ameen, E.L. Barish, T. Mizutani and D.J. Vitkavage, J. Vac. Sci. Tech. B3, p. 1373, 1985). These examples, of course, can serve only as a guide since the properties of the deposit material have yet to be determined. It should be noted that for the two physical sputtering examples from Somekh, the maximum of each yield curve leads to- maximum possible sidewall angle, beyond which the net substrate etch rate becomes negative, implying net deposition instead of etching.

EXPERIMENTAL RESULTS The hollow-cathode RIE apparatus used to obtain all the following results has been described previously herein. In all cases, two 10cm dia. silicon targets were used with a spacing of 4cm.

Fig. 18 shows the results of an experiment in which grating samples of thermal SiO_? on Si, masked with Cr, together with samples of Si0 , were etched in CF. at ^"a pressure of 7 Pa and a flow rate of 0.06 Pa m 3sec-1

The samples were placed on the bottom target and heatsunk to it with diffusion pump oil. The rf input voltage was kept constant at 1.4 kV The deposition rate of the metallic film was varied by placing pieces of Mo foil on the top and bottom targets. Despite the increasing area of metal (from 1cm 2 to about 60cm2) the input power and bias voltage varied by less than 2%, suggesting constant etching conditions. The Si0₂ etch rate was determined by stylus, profilometry, and the sidewall angle from SEM photographs of the gratings sectioned after etching. The trend toward lower substrate etch rate with increasing angle is obvious. No experimental points corresponding to angles greater than about θ=50 could be obtained since it was found that, using the above method for controlling the deposition rate, increasing the Mo coverage on the targets beyond that corresponding to the maximum angle usually resulted in the deposition of a thick, hygroscopic film over part or all of the targets - 25 -

and samples. This could be due to the deposition rate becoming a critical function of the Mo coverage on the targets and may only indicate a lack of sufficient process control. However, as shown in Fig. 6, a maximum possible sidewall angle exists for physical sputtering forms of the deposit etch-yield function. Moreover, the departure of the results in Fig. 18 from the simple predictions of

Equation (3) may indicate some dependence on the deposit etch rate on the angle. We may estimate this dependence by replotting Fig. 18 using Equation (4), since for each point f(θ) = (1 - E /E )/sin(θ). The results, shown in Fig. 19, are qualitatively similar to the etch yield resulting from a physical sputtering mechanism and suggest that this is the dominant form of etching of the deposit material in this case.

The etching of the grooves of different widths is shown in Fig. 20. The same etching conditions as above were used, but with a grating sample of variable line spacing. As can be seen the groove of smaller width has reached an equilibrium surface, while the larger groove has continued to etch. This agrees with the simulation of

Fig. 14. Furthermore, although the smaller groove has

"stopped" etching there appears to be no visible residue on its surface. The angled etching of an existing groove is shown in

Fig. 21. An initially flat Cr masked Si0₂ on Si substrate was etched for 7 minutes using the conditions

2 above, with 5cm of Mo metal on the target. The resulting sample (Fig. 21a) was removed from the chamber, cleaned to remove any possible residues, then re-etched

2 for 2 minutes, this time with about 60cm of Mo on the target. The resulting profile (Fig. 21b) shows an increased sidewall angle at the base of the groove similar to the expected profile of Fig. 15. Furthermore, the groove has been completely filled with apparently deposited material, although its surface profile bears no resemblance to the etched sidewall. However, as noted above, the deposit which can form on the targets is hygroscopic. Since the groove deposit is likely to be the same material, one would expect it to absorb water when the sample is exposed to the atmosphere. The probable swelling of the deposit could cause destruction of the surface profile. This may also account for the "frothy" appearance of the deposited material in the groove. The deposited material is easily removed in acetone, leaving the etched substrate shown in Fig. 21c.

The above results have tested situations involving normal ion incidence to the initial substrate plane. By mounting samples at an angle to the target surface, providing the samples are small enough to lie within the sheath region, ions will be incident on the sample at the mounting angle. The results of etching a grating sample at a mounting angle of 25 is shown in Fig. 22. The result is in good qualitative agreement with the simulated profiles of Fig. 16.

Earlier work on the etching of some group III-V semiconductors in halogen based gases has suggested that the presence of low volatility halides of the group-Ill metal as a surface layer (either adsorbed, deposited or formed in-situ) can limit the substrate etch rate (V.M. Donnelly, D.L. Flam , C.W. Tu and D.E. Ibbotson, J. Electrochem. Soc. 129, p. 2533, 1982 -and- R.J. Contolini, J. Electrochem. Soc. 135, p. 292, 1988). Such work has often also been associated with observations of overcut, or angled etching. The work of Kay and Dilks (E. Kay and A. Dilks, J. Vac. Sci. Tech. 16, p. 428, 1979) has shown that during fluorocarbon plasma etching of molybdenum, MoFg is formed and is deposited on the targets. Moreover, exposure to atmospheric oxygen results in hydrolysis of the film. We have observed the suppression of etch rates and have related it to the sidewall angle. Furthermore, we have observed the formation of hygroscopic films following fluorocarbon etching with Mo. Our work has indicated that those metals which most readily produce angled sidewalls in a given halogen based etch gas are those with the least volatile halide. Metals with very high volatility halides rarely produce noticable etched angles. This observation is also supported by the work of Klinger and Greene (R.E. Klinger and J.E. Green, J. Appl. Phys. 54, p. 1595, 1983) who reported that in etching GaAs the angle of the sidewall is greater using etch gases with a high F to Cl ratio. The volatility of gallium chloride - is significantly higher than that of gallium fluoride. Together, these results strongly suggest that the deposition of metal halide films during directional etching can cause angled etching.

During our work we have observed a maximum practical etched sidewall angle θ, which correlates with the maximum practical etched sidewall angles theoretically obtainable for some physical sputtering mechanisms (Fig. 17, Somekh data) , implying that the deposit material may be etched predominantly by a physical, non-reactive process. This would have an impact on the exact sidewall shape since the possibility then exists for the redeposition of the deposit material from one sidewall to another, as in inert-ion sputter etching (Lehmann et al) . However, if the etching mechanism were dominantly ion-induced volatilization or chemically enhanced physical sputtering, as has been suggested for the etching of gallium fluoride (Klinger & Green) , then the chemical component of etching would dominate over the physical component and redeposition would be much less of an issue.

In the case where extremely low volatility metal halides such as aluminium fluoride are involved in the etch process, the physical etching component is probably more important. This may be used to explain some anomolous experimental results which we have observed, An example is shown in Fig. 23(a) of etching with a non-normal ion direction in CF. with Al metal present, where slight rounding of the sidewalls has occurred and the base of the groove appears to be tilted with respect to the initial substrate surface. Assuming that physical sputtered material is emitted from a sidewall surface with a cosine distribution. (P.G. Gloersen, Solid State Tech. p. 68, April 1976 -and- Lehmann et al) and that some of this etched material is redeposited on the sidewalls of the developing groove, then an apparently higher sidewall deposition rate will result. Our earlier etching model can be easily modified to account for this. Fig. 23(b) shows a simulation under the same conditions as in Fig.

16. In Fig. 23(c), 2% of the etched deposit is assumed to be redeposited in the manner described above. The resulting etched profile is in qualitative agreement with the experimental result in Fig. 23(a). Naturally, precise agreement is unlikely since the sputter redistribution may well be non-cosinusoidal and other possible effects such as corner trenching have been omitted. Redeposition may also account for the surface roughness that is also sometimes observed, especially in discharges containing Al. We have shown that angled etching may be readily achieved by the deposition of a metal-halide film during directional etching. However, we do not suggest that this is the only such mechanism, since according to the simple model we have described any directional etching process ought to produce angled sidewalls if the rate of deposition is sufficient. In fact, this is supported by other work in our laboratory which has provided evidence for angled etching of cooled silicon samples etched in SiCl./θ₂ gas mixtures.

Claims

THE CLAIMS DEFINING THE INVENTION ARE AS FOLLOWS:

1. A dry process etching chamber comprising: a vacuum chamber; at least one pair of radio frequency electrodes disposed within said chamber; at least one source of radio frequency potential connected to said electrodes; a metallic source within said chamber; inlet means for introducing a gas into chamber, which gas becomes chemically reactive when ionised by a radio frequency discharge so as to transport metal from said metallic source and perform an etching process upon target material positioned on one or more said electrodes.

2. The chamber of claim 1 wherein the metallic source comprises a metallic target within the chamber arranged to be etched by the reactive gas and thereby source metallic ions.

3. The chamber of claim 1 wherein the metallic source comprises a metal compound as^* a component of the gas introduced into the chamber.

4. The chamber according to any one of the preceding claims wherein the radio frequency electrodes comprise an anode and a hollow cathode having two substantially opposed cathode electrodes.

5. The chamber according to any one of the preceding claims wherein the target material is a grating block to be etched with grating lines and the grating block is slidably located on a first cathode and covered by a target material having a slot for etching of a portion of the grating line in the grating block, the chamber also including a second stepped cathode electrode located adjacent to the slot.

6. The chamber according to Claim 5 wherein the stepped electrode includes a metal surface for sourcing metal.

7. A method of dry process etching in a vacuum chamber comprising the steps of: positioning target material on a radio frequency electrode, said electrode being one of a pair of radio frequency electrodes positioned within said chamber; providing a metallic source; applying radio frequency potential between said pair of electrodes; and introducing a gas into said chamber, which gas when excited by a radio frequency energy, forms a chemically reactive plasma, said plasma transporting metal from said metallic source and performing an etching process on said target material.

8. The method of claim 7 wherein a metal target is provided in the etching chamber to source the metal.

9. The method of claim 7 wherein a metal compound is included in the gas introduced into the chamber as the source of metal ions.

10. The method according to any one of the preceding claims wherein asymmetrical etch profiles are achieved by selected placement of the metal source.

11. The method according to any one of the preceding claims wherein a slotted mask is located over the material to be etched such that grating lines are etched in the material.

12. The method of claim 11 wherein asymmetrical etching is achieved by selected placement of the metal source.

13. A process of etching at controlled angles into substrates wherein metal is deposited onto the substrate simultanteously with the etching whereby the etched angle is independent of the etched material, the mask material and the etch selectivity.