One of the key issues in implementing the transition from photolithography to projection e-beam

Transcription

1 Mark topography for alignment and registration in projection electron lithography Reginald C. Farrow, Masis Mkrtchyan, Kevin Bolen, Myrtle Blakey, Chris Biddick, *Ljnus Fetter, Harold Huggins, Regine Tarascon, Steven Berger AT&T Bell Laboratories, Murray Hill, NJ *AT&T Bell Laboratories, Holmdel, NJ ABSTRACT We have studied two mark geometries for possible use in a projection e-beam lithography system using SCALPEL (SCattering with Angular Limitation in Projection Electron Lithography).' These are V-grooves and vertically etched geometries -- pedestals or trenches. We report results of measurements of backscattered electron (BSE) contrast from topographic marks of varying size and as a function of energy up to 100 kv. The marks were fabricated on silicon wafers. The measurements were taken both in a scanning electron microscope and in an experimental SCALPEL machine operating in focused probe mode. The V-grooves ranged from 1.0 to 30 im wide. The vertical etched features ranged from 2to 30 im wide and 0.6 and 50 im depth. The results depended not only on the feature width and depth, but also on whether the features were isolated or in line and space patterns. Using a BSE ratio of 1.05 as a criterion for acceptable contrast from an alignment mark, V- grooves and vertical etched features had acceptable contrast with exception of the smallest and shallowest features for both geometries. Keywords: alignment marks, mark detection, topographic marks, backscattered electrons, e-beam lithography, projection electron lithography, SCALPEL, V-grooves, alignment and registration 1. NTRODUCTON One of the key issues in implementing the transition from photolithography to projection e-beam lithography (SCALPEL) is the adaptability of current processes for alignment mark fabrication. This is especially important in the context of a mix and match strategy, where only critical layers will be exposed by projection e-beam lithography. The mark detection system for SCALPEL uses backscattered electrons to detect the overlap between the image of a mask mark and the corresponding mark on a wafer (the mask image features are referred to as "beamlets").2'3 The BSE contrast from the mark has to be sufficient to provide a signal that will allow for mark detection under the proposed operating conditions of a high throughput lithography These operating conditions would require that during mark detection for alignment and registration, the BSE signal from the mark must be acquired in 10 jisec with a probe current of ia at 100 kv. Earlier we reported that a sufficient requirement for the mark under these conditions is that the ratio of the BSE signal between aligned and misaligned conditions be at least This requirement was established after analysis of experiments with the proposed mark detection system in an experimental SCALPEL machine. One possible wafer mark fabrication process is the deposition of a material such as a metal. For deposited materials the BSE contrast results from the difference in BSE coefficient between the mark and the underlying substrate material (i.e. silicon). Using a single scattering theory for thin films deposited on a bulk substrate6, the BSE coefficient is proportional to Z2t, where Z is the atomic number and t is the film thickness.7 This type of signal is referred to as compositional contrast or Z contrast in scanning electron microscopy. We will, therefore, refer to marks that exploit this contrast mechanism as compositional marks. Previously reported experiments and theoretical calculations showed that there is sufficient BSE contrast for mark detection from deposited films of most of the materials being considered for conductive links ULS circuits.5 Compositional mark contrast is not significantly degraded by resist overlayers. As an example, an alignment mark composed of 100 nm of tungsten on a silicon substrate (patterned appropriately) would provide sufficient BSE contrast. However, there was not sufficient contrast from 500 nm of aluminum on silicon because of the small difference in atomic number. One possible disadvantage to using compositional marks is that their fabrication may require the deposition of a material that is not one of the circuit materials if the normal circuit materials would not generate sufficient BSE signal for mark detection. O /96/$6.OO SP1E Vol / 143

2 Another wafer mark fabrication process is to etch marks into the wafer or overlayer. The BSE signal is then dependent on the cross sectional geometry of the mark features or topography. We will refer to this type of mark as a topographic mark. Two versions of topographic marks that have been studied in e-beam lithography are V-grooves and vertically etched features -- pedestals or trenches. V-grooves are fabricated on <001> oriented silicon using an anisotropic etch in KOH. The KOH etch exposes the (1 1 1) planes which are at 54.7 angles from the surface normal. Vertically etched features are fabricated on silicon by using a plasma etch process. They can also be fabricated with an anisotropic etch using <1 10> oriented silicon. n this case the (11 1) planes are perpendicular to the surface normal. Topographic marks are advantageous for C process integration since they do not require the deposition of a special scatterer for BSE contrast. n this case the BSE contrast derives from the angular dependence of BSE scattering and is also dependent on the BSE detector location and solid angle. Studies have shown that both geometries can be used as alignment marks in direct write e-beam 89 The results reported here will emphasize topographic marks for SCALPEL proposed operating conditions. However, the results are applicable to all e-beam lithography systems. n the present study V-grooves and vertically etched features were fabricated on silicon and the BSE contrast was measured. Dependence of BSE contrast on the dimensions of the marks was measured and for the V-grooves the energy dependence was measured. Formerly we have qualitatively discussed the topographic contrast generation in the case of V-grooves and This qualitative level of understanding of the problem, apparently appropriate for the relatively large size marks (compared to the electron range), is not capable of explaining the experimental results obtained here for a wide range of variation of mark size and incident electron energy. A comprehensive model of the topographic contrast generation requires the consideration of the complex variety of interaction effects of the electrons with a solid target including elastic and inelastic, single (large angle) and multiple (small angle) scattering. The difficulties encountered in the corresponding mathematical task are significant and simpler analytical theories, such as Everhart's singlescattering10 and Archard's diffusion'1 theories, or Monte Carlo (MC) method9 are often used to describe the electron backscattering.12 We have applied these simple theories to simulate the BSE contrast generation from an isolated topographic mark. Unlike MC method, these theories are capable of revealing of the basic trends in the contrast generation when the parameters involved vary in a wide range. We have found that a combined model which accounts for both large-angle single-scattering and diffusion (multiple scattering) fractions of electron backscattering has the potential to explain the experimental results presented here. A detailed discussion of this model and the results obtained for different marks will be published later EXPERMENT The V-grooves were fabricated on <100> silicon wafers in a pattern of lines ad spaces and isolated lines which were oriented perpendicular to the wafer flat. The pattern was produced using e-beam lithography and the V-grooves were formed by a KOH etch. The features ranged from 1.0 to 30 im wide which corresponds to 0.7 to 20 im depth. The vertical etched features were fabricated on <110> silicon wafers in a similar pattern using e-beam lithography. These were etched in KOH for a series of time durations to give features of different depths from.6 to 50 tim. The depths were measured with a surface profilometer. For the V-grooves measurements of BSE contrast were recorded in an SEM and in an experimental SCALPEL machine. The SEM was a JEOL 6300F (field emission source). The BSE detector was an annular type with two semi-circular regions. Measurements were taken with the samples at a working distance of 12 mm. The detector solid angle was 0.39 sr. The SEM measurements were performed by recording the signal from the BSE detector amplifier on a digital oscilloscope.. The image of the feature of interest was centered and oriented on the SEM monitor and line profiles were recorded from 15 to 30 kv. For the V-grooves the BSE ratio was defined as the ratio between the BSE signal away from the feature and the minimum signal (at the trough of the V-groove). 144 SPE Vol. 2723

3 The experimental SCALPEL machine has been described elsewhere.14 t is a modified JEOL 4000 transmission electron microscope (TEM). The TEM camera apparatus was replaced with a computer controlled wafer translation stage. The BSE detector was an EG&G surface barrier detector mounted at a 45 angle at a distance of 39 mm from the samples. The detector solid angle was 0.29 sr. Measurements were taken by 15 operating the SCALPEL machine in an SEM mages of the sample were recorded in an image acquisition system and analyzed on a computer in a similar way as the SEM recorded lineprofiles. 3. RESULTS AND DSCUSSON Examples of BSE lineprofiles collected from V-grooves are shown in Fig. 1. The BSE signal increases as the electron beam approaches the edge of the V-groove. This is a consequence of the range, R, of the electrons. The excess signal is from backscattered electrons that exit the sample through the walls of the V-groove. This type of signal is often referred to as edge enhancement. The electron range dependence in the edge enhancement is evident from Fig ic. The onset of the excess BSE signal occurs closer to the V-groove at lower energy where R is smaller. Compare the 15 kv (R =2.7 rim) to the 25 kv lineprofile (R = 6.2 rim). We have used the Bethe range, RB, to estimate R for the purposes of qualitatively describing the data. At constant energy the maximum amplitude of the excess BSE signal is a function of the depth, h, of the V-groove and the takoff angle, i1ii, of the detector. The iji dependence is most clearly evidenced in the 30 im V-groove at 100 kv (see Fig. la). n the 100 kv measurements the takeoff angle differs on each of the inside faces of the feature because of the location and orientation of the detector. This causes a non-symmetric BSE lineprofile. Also, the maximum amplitude of the excess BSE signal reaches a plateau when the depth of the V-groove reaches and exceeds the range of the backscatter electron (see Fig. ib). 100 kv 30kV c) 3 tm ci) C,) Displacement (rim) Fig. 1 BSE lineprofiles from V-grooves at a) width. U U U 100 kv, b) 30 kv, c) 3 im SPE Vol / 145

4 The BSE signal decreases from the maximum excess signal near the edge of the V-groove and reaches a minimum when the beam is at the bottom of the feature. The behavior of the BSE signal in this region is dependent on h, R, and iy. One interesting effect is the appearance of an additional inflection point in the lineprofile when the depth of the feature becomes larger than R (see Fig ib, 11 rim, and Fig. ic, 15 kv). The minimum BSE signal also reaches a plateau. n Fig. 2 are plotted the BSE ratio as a function of V-groove depth for 30 kv. The BSE ratio reaches a plateau at approximately half of the electron range (R = 8.2 tim). 30kV V-Groove Depth (tim) Fig. 2 BSE ratio as a function of depth for V-grooves at 30 kv The energy dependence of the BSE ratio for the 3 jim V-groove is shown in Fig. 3 compared to theoretical calculations based on a combined single-scattering diffusion theory. t should be noted that the measurements are from two different experimental apparatus. The BSE detector for the 100 kv measurements (SCALPEL machine) has a different solid angle and ipr than the SEM measurements. The plot illustrates the expected behavior when compared to the calculated BSE ratio for the annular detector used with the SEM measurements. This will be discussed in more detail in a separate paper.'3 C 5.' 2.6' '' 1.8' 1.6' SEM SCALPEL Machine Theory - Hemispherical Detector Theory - SEM Annular Detector Energy (kev) Fig. 3 Energy dependence of BSE ratio for 3 jim wide V-groove compared to theory /SP!E Vol. 2723

5 n Fig. 4 are plotted examples of lineprofiles from pedestals of 4 jim width that are separated by 4 jim (i.e. lines and spaces) for a series of depths at 100 kv. Here again the takeoff angle and detector location cause non-symmetric lineprofiles. The edges of the features are not resolved. This is because of range of the electrons at 100 kv (R = 62 jim). The interior pedestals have reduced contrast compared to the outside pedestals and become comparable as h approaches approximately R12. n Fig. 5 are plotted the BSE ratio vs. h for 4 jim and 16 jim lines and spaces and for a 30 jim isolated pedestal. The V-groove data is also plotted on the same figure (see Fig. 5). Notice that the 30 jim isolated pedestal has less signal when compared to the 16 jim line and space and the V-groove. For both V-grooves and pedestals the BSE ratio is acceptable for mark detection with the exception of the smallest and shallowest features. 4 p.m Lines and Spaces ȧ-i ri) Ci) ' Displacement (jim) Fig. 4 Plot of BSE signal vs. displacement for 4im pedestals etched in silicon at different depths at 100 kv. 100 kv A 1.0 p_! Depth (jim) A Fig. 5 Plot of BSE ratio vs. depth for vertically etched features and V- grooves at 100 kv. U 4 im Pedestals (Lines & Spaces) 16 im Pedestals (Lines & Spaces) A 3OimPedestal (solated Line) V-groove (solated) Detection Criterion SPE Vol / 147

6 A further analysis is warranted to determine how the topographic mark contrast is manifested using the beamlet mark detection system.5 Unlike the present measurements, the beamlets that are scanned over the mark features are typically the same lateral size (e.g. width) as the feature. Therefore the detected BSE signal lineprofiles will be broader and have less contrast than those for a focused probe measurement. For the V- groove topography the contrast is reduced more than either a pedestal or trench topography. This effect is simulated in Fig. 6 for a 3 jim pedestal, trench, and V-groove. Assuming that the electron probe penetration is much smaller than the feature size, a focused probe will generate a BSE lineprofile that matches the geometry of the feature (the dashed lines in Figs. 6a and 6b and the V shaped feature in Fig. 6c). T the lineprofiles from the focused probe simulations are convoluted with a "top hat" function of unit amplitude, the effects of a beamlet probe can be simulated. The resulting BSE lineprofiles would have a V shape for the pedestal and trench (see Figs. 6a and 6b). The maximum contrast would be the same as for a focused probe in the case of the pedestal and trench topography. f the beamlet is scanned over the V-groove the BSE lineprofile is sinusoidal and the maximum contrast is half of the corresponding focused probe lineprofile (see Fig. 6c) TJ) Displacement (jim) Fig. 6 Comparison of simulated BSE signal lineprofiles for focused probe and beamlet detection schemes. a) pedestal, b) trench, c) V-groove, d) 3 jim V-groove at 30 kv convoluted with 3 jim "top hat" function (see text). Taking into account penetration of the electron probe and multiple scattering effects will further reduce the BSE contrast with all marks. Beam penetration and multiple scattering effects were discussed earlier in the context of the excess BSE signal near the edge of the V-groove. Multiple scattering effects can be simulated by convoluting measured BSE lineprofiles (focused probe) with a "top hat" function of unit amplitude. The resulting lineprofile is shown in Fig. 6d, where the 3 jim V-groove lineprofile at 30 kv was used. The beamlet lineprofile has less than half the contrast of the original lineprofile (see Fig. 6d). The BSE ratio is reduced to 1.19 compared to 2.06 for the original lineprofile. 148/SPE Vol. 2723

7 4. CONCLUSON We have measured BSE signal lineprofiles from topographic marks of both V-grooves and pedestals as a function of width and depth. The lineprofiles show dependencies of mark depth, electron range, and detector location and takeoff angle. The energy dependence of the V-grooves agrees with predictions using a combined single scattering and diffusion theory. Using a BSE ratio of 1.05 as a criterion for acceptable contrast from an alignment mark, V-grooves and vertical etched features had acceptable contrast with exception of the smallest and shallowest features for both geometries at 100 kv. An analysis of the BSE signal using the proposed mark detection system for SCALPEL, shows that the contrast for a V-groove geometry will be reduced to half of that expected from the vertical etched features. These factors will have to be considered in the design of a topographic mark process for alignment and registration under high throughput operating conditions. 5. ACKNOWLEDGMENTS This work was supported by ARPA under contract MDA C SCALPEL is a registered trademark of AT&T Corporation. 6. REFERENCES 1. S. D. Berger, J. M. Gibson, R. M. Camarda, R. C. Farrow, H. A. Huggins, J. S. Kraus, J. A. Liddle, J. Vac. Sci. Technol. B 9, 2996 (1991). 2. R. C. Farrow, J. A. Liddle, S. D. Berger, H. A. Huggins, J. S. Kraus, R. M. Camarda, C. W. Jurgensen, R. R. Kola, And L. Fetter, J. Vac. Sci. Technol. B 10, 2780 (1992). 3. J. Frosien, B. Lischke, and K. Anger, J. Vac. Sci. Technol. 16, 1827 (1979). 4. J. A. Liddle, and S. D. Berger, J. Vac. Sci. Technol. B 10, 2776 (1992). 5. R. C. Farrow, J. A. Liddle, S. D. Berger, H. A. Huggins, J. S. Kraus, R. M. Camarda, R. G. Tarascon, C. W. Jurgensen, R. R. Kola, and L. Fetter, J. Vac. Sci. Technol. B 11, 2175 (1993). 6. For a review, see H. Niedrig, J. Appl. Phys. 53, R15 (1982). 7. P. B. DeNee, in Proceedings of Scanning Electron Microscopy 1978 (AMF O'Hare, L, 1978), Vol., H. Friedrich, H.-U. Zeitler, and H. Bierhenke, J. Electrochem. Soc., 124, 627 (1977). 9. D. Stephani, J. Vac. Sci. Technol. 16, 1739 (1979). 10. T. E. Everhart, J. Appl. Phys. 31, 1483 (1960). 11. G. D. Archard, J. App!. Phys. 32, 1505 (1961). 12. H. Niedrig, J. App!. Phys. 53, R15 (1982). 13. M. M. Mkrtchyan and R. C. Farrow, EPB-96, (1996) (to be published). 14. R. C. Farrow, S. D. Berger, J. M. Gibson, J. A. Liddle, J. S. Kraus, R. M. Camarda, and H. A. Huggins, J. Vac. Sci. Technol. B 9, 3582 (1991). 15. M. B. Heritage, J. Vac. Sci. Technol., 12, 1135, (1975). SPE Vol / 149