effect of electron cyclotron resonance plasma parameters on the aspect ratio of trenches in HgCdTe, The
Stoltz, A J
(Received November 13, 2002; accepted March 27, 2003)
Values of the aspect ratio for trenches etched into HgCdTe by an electron cyclotron resonance (ECR) plasma containing hydrogen and argon are limited by the phenomenon of etch lag. Modeling this plasma as an ion assisted, reactive-etching process leads to a set of conditions that greatly reduces etch lag. Use of these new process conditions produces trenches with aspect ratios greater than 3, widths less than 3 [mu]m, and depths in excess of 15 [mu].
Key words: Electron cyclotron resonance (ECR), reactive ion etching (RIE), plasma, focal plane array (FPA), HgCdTe (mercury cadmium telluride), etch lag, aspect ratio
INTRODUCTION
We have been developing an electron cyclotron resonance (ECR) plasma technology to etch trenches with high aspect ratios in HgCdTe for third-generation, infrared detector arrays. Certain array architectures1-3 require trenches with widths less than 3 [mu]m and depths greater than 15 [mu]m. For complementary metal oxide semiconductor (CMOS) materials, such as Si, SiO^sub 2^, and Si^sub 3^N^sub 4^, it has been found that the more narrow the trench width, the more difficult it is to achieve high aspect ratios. This phenomenon is known as etch lag.
Two factors have been identified that contribute to the aspect ratio of HgCdTe trench etching. The first is the type and processing of the photoresist mask. New developments in resist technology will be the focus of a companion study by Benson et al.4 The second are the composition and processing parameters of the plasma. An investigation into plasma effects is the subject of this paper. We will report on chemical, crystallographic, and electrical properties of high aspect ratio HgCdTe trenches in a future report.
EXPERIMENTAL
A Model 357 PlasmaQuest (Nexx Systems LLC, Wilmington, MA) ECR reactor with an ASTeX 2.45 GHz microwave source supplying power through a quartz window was used in these experiments. The heated wave resonance zone is 25.3 cm from the sample. An upper electromagnet produced an 873 G field to create the resonance condition. A capacitively coupled, 40.68 MHz (rf) generator was used to supply a direct current (DC) bias potential using between O W and 300 W of input power. Self-bias could be varied between 0 V and -635 V (0 W to 300 W of input power). The higher self-bias enhances the contribution of Ar^sup +^ bombardment.
The argon to hydrogen ratio was set at 4:1, 80 sccm of argon and 20 sccm of hydrogen, with the H^sub 2^ being injected downstream to give a 2 mtorr process pressure. Excited species generated in the ECR zone dissociate H^sub 2^ while minimizing hydrogen-ion formation, in part, by taking advantage of the fact that the bonding energy of H^sub 2^ (4.5 eV) is less than the ionization energies of H^sub 2^ and H (15.4 eV and 13.6 eV, respectively). The active etching species are, therefore, H and Ar^sup +^.^sup 5-12^ Atomic hydrogen is expected to react with Te in the manner described by Hirsch et al.12 This ion assisted process was chosen to minimize the selective removal of host atoms in the ternary compound. Argon ions are used to facilitate physical etching. To promote stoichiometry, a DC self-bias is applied to accelerate positive ions produced in the ECR zone to extract Cd. The Hg volatilizes from the surface without any ion assistance. At 2 mtorr, some H^sub 2^ will likely reach the ECR region where a small density of H^sub 2^^sup +^ and H+ will be produced. The downstream injection is intended to reduce these concentrations and thereby minimize disruption of the HgCdTe crystal.
At 60 W, DC bias-input power, long-wave HgCdTe, with an infrared absorption cutoff of approximately 10 [mu]m, etches at a rate of approximately 1 [mu]m/min. At 180 W, DC bias-input power, the same material etches at approximately 1.5 [mu]m/min. These rates are for large, open areas, as features with small dimensions etch at lower rates.
Both (211) and (111) HgCdTe oriented epilayers were deposited by molecular beam epitaxy (MBE) at the Night Vision and Electronic Sensors Directorate and by MBE and liquid phase epitaxy (LPE) at Raytheon Vision Systems (RVS). Substrates for MBE were (211) oriented Si or CdZnTe and, for LPE, (111) CdZnTe. Resist strips, approximately 1 mm wide and several millimeters long, were fabricated on the HgCdTe.4 Spacings between these strips varied from 1 [mu]m to 500 [mu]m. This pattern allowed the simultaneous etching of trenches with various widths in a single ECR etch,13,14 thus facilitating an examination of the effect of trench width on the etching process.
A Plasma Device 1-Dimensional Bounded Electrostatic Code (PDP1) version 4.1 for X-Windows15 was used to model the plasma in the ECR reactor.
RESULTS I
In a first series of experiments, the resist patterns were applied to MBE and LPE epilayers, and trenches were formed by etching. A scanning electron micrograph of a cleaved edge is shown in Fig. 1. Trench widths vary from 2 [mu]m on the left of the micrograph to 7 [mu]m on the right. It is clear that the trench depth decreases as the width decreases.
To quantify these observations, we define W^sub e^ and D as the trench width and the trench depth, respectively, and D^sub open^ as the depth etched in an unmasked portion of the epilayer during the same process. The aspect ratio, AR, of a trench is defined by the ratio D/W^sub e^. Data from Fig. 1 are plotted in Fig. 2. The ratio of the depth for a given trench and the depth in the unmasked region is displayed on the ordinate. The abscissa is 1/AR, defined here to be the inverse aspect ratio (IAR). For IAR values >0.77, D/D^sub open^ is unity, and no etch lag is evident. For IAR values
THEORY
A great many theories have been proposed to account for the plasma etching of CMOS related materials.16-34 Many of these models failed to explain the etch lag results measured in HgCdTe. However, one particular model, that of Jansen et al.16 and Elwenspoek and Jansen17 is consistent with measurement made in HgCdTe. This model states a plasma consists of three species: (1) charged particles that bombard and remove material by a momentum transfer process, (2) radicals that react chemically with a surface and form volatile species that are then pumped away, and (3) inhibitors that either neutralize etching ions or radicals or are etch resistant products on the semiconductor. For plasmas reported in the literature as etchants for HgCdTe, examples of these three would be Ar^sup +^; H that reacts with tellurium to form TeH^sub 2^, which is volatile and pumps off;5,6,12 and N^sub 2^, used with plasmas containing methane to slow the etch rate,5,6 respectively. Such a plasma process can be described as ion induced, reactive ion etching (RIE). Ionic paths acquire a preferred directionality because of the various self and applied biases, while radicals and inhibitors are directionally random. The ion angular distribution (IAD) is predicted by Jansen and Elwenspoek to have a strong effect on the shapes of etched features. This effect is described in Fig. 3, where the aspect ratio of a trench is described by an angle [theta]. The critical angle [theta]^sub c^ is the angle [theta] that is approximately equal to the maximum of the IAD,17 labeled as [straight phi] in Fig. 3. In this model, trenches with [straight phi] [theta]^sub c^ do show etch lag because fewer directional ions reach the bottom of the trench.
Thus, the path to achieving values of IAR^sub c^ is reduced to reducing the value of [straight phi] of the plasma.
SIMULATION
To reduce [straight phi], we chose to vary the DC bias. The heated wave resonance zone of the PlasmaQuest 357 is 25.3 cm from the sample; therefore, changes in DC bias are expected to vary the IAD of our plasma while having a minimal effect on other plasma parameters, such as species type, electron temperature, and plasma density. Ion bombardment energy will, however, be increased with this configuration.35 The results of a PDP1 simulation of the IAD are shown in Fig. 4. Ions with angles greater than [straight phi] have less energy and, therefore, contribute little to physical etching.36,37 Increasing the DC bias reduces the value of [straight phi] and, therefore, [theta]^sub c^.
RESULTS II
In Fig. 5, we show the results of exposing HgCdTe to plasmas with three values of DC bias. It can be seen at once that the region with no etch lag is extended to IAR values less than the 0.77 value in Fig. 2.
The experimental results in Fig. 5 can be compared to the Jansen and Elwenspoek model. In Fig. 5, the equation for the etch-lag region has the mathematical form:
The experimental value of m is obtained from the slope and is equal to the critical aspect ratio, AR^sub c^. This compares to the relationships given by
The Jansen and Elwenspoek model predicts a trend that is approximately linear but with a value for the slope m that is twice the observed value. This disagreement is reasonable in view of the fact that their model was developed for CMOS processes that are quite different from the one considered here. Also, the value of AR for our process is reduced because the resist etch rate in our plasma is not negligible. Values for AR^sub c^ found experimentally, calculated from Eq. 7 (the Jansen and Elwenspoek model) and calculated with Eq. 1 from the simulation, results in Fig. 4 and are shown in Table I.
The Elwenspoek and Jansen theory, the experimental data, and the PDP1 simulation all show the same trend: as the DC bias increases, AR^sub c^ increases. The simulation results yield AR^sub c^ values that are much larger than either experimental values or those calculated from Eq. 7. These increased AR^sub c^ values are believed to result from errors in the simulation of the plasma because small changes in chamber dimensions cause large changes in [straight phi]. Also, the values of experimentally measured AR^sub c^ are closer to values that Elwenspoek and Jansen17 reported for other RIE processes.
APPLICATION
Increasing the DC bias input power has allowed much higher aspect ratios to be obtained, as shown in the scanning electron micrographs in Fig. 6. These are trenches from the samples whose data appear in Fig. 5. Each started with a nominal 3 [mu]m resist opening. Note that etch lag also appears to be independent of the thickness of the resist as discussed in a companion paper.4
At smaller DC biases, Fig. 6a, the etch becomes more isotropic or more “wet chemical like” and AR 1. As the DC bias increases, the etch becomes more anisotropic and AR > 1. This improved behavior can be coupled with the use of hardened resists to make a dramatic change in the final trench profile. An example of a trench produced with conventional resist geometries and conventional plasmas is shown in Fig. 7a, with an AR = 2.69. An example of a trench produced with improved resist sidewalls and improved plasma etching parameters is given in Fig. 7b. For this trench, W^sub e^ and D are measured to be 2.9 [mu]m and 15.7 [mu]m, respectively, and AR = 5.4.
In a future paper, we will document the effect of various masks and plasmas on the chemical and electro-optical characteristics of etched HgCdTe. While obtaining the scanning electron micrographs reported here, however, we used a variant on the scanning electron micrograph technique to map the location of the host chemical species in HgCdTe exposed to the plasma.
Figure 8a-c are the energy dispersive x-ray (EDX) maps of the same trench shown in Fig. 7b. Figure 8d is an electron micrograph with the same spatial registration as Fig. 8a-c. The effective spatial error, caused by spot size and detection geometry, is 0.75 [mu]m. Figure 8 shows that there is only a very thin Hg depletion region at the sidewall of the etched region (less than 0.75 [mu]m). In addition, Fig. 8b and c shows a possible redeposition of Cd and Te on the sidewall of the resist, which may be due to the high DC bias input power of 180 W.
CONCLUSIONS
When trenches are etched in HgCdTe epilayers by exposing them to an ECR hydrogen/argon plasma through a photoresist mask, the achievable depths and aspect ratios are limited by the phenomenon of etch lag. If a correction factor is applied to the model of Elwenspoek and Jansen, predictions are in good agreement with the HgCdTe data. Etch lag can be decreased by increasing the DC bias on the plasma. Trenches with widths less than 3 [mu]m and depths greater than 15 [mu]m have been achieved by applying 180 W of bias input power. This width enables the delineation of the small, closely spaced pixels required for high-resolution detector arrays. This depth enables the optical isolation of pixels in the multilayered structures needed for two- and three-color detectors and avalanche photodiodes.
REFERENCES
1. W.E. Tennant et al., J. Electron. Mater. 30, 590 (2001).
2. R.D. Rajavel et al., J. Electron. Mater. 27, 747 (1998).
3. T.J. de Lyon et al., J. Cryst. Growth 201, 980 (1999).
4. J.D. Benson, A.J. Stoltz, A.W. Kaleczyc, M. Martinka, L.A. Almeida, P.R. Boyd, J.B. Varesi, and J.H. Dinan, in this issue.
5. R.C. Keller, H. Zimmerman, M. Seelmann-Eggebert, and H.J. Richter, J. Electron. Mater. 25, 1270 (1996).
6. R.C. Keller, H. Zimmerman, M. Seelmann-Eggebert, and H.J. Richter, J. Electron. Mater. 26, 542 (1997).
7. P. O’Dette, G. Tarnowski, V. Lukach, M. Krueger, and P. LoVecchio, J. Electron. Mater. 28, 821 (1999).
8. A. J. Stoltz, J.D. Benson, M. Thomas, P.R. Boyd, M. Martinka, and J.H. Dinan, J. Electron. Mater. 31, 749 (2002).
9. A.J. Stoltz, M.R. Banish, J.H. Dinan, J.D. Benson, D.R. Brown, D.B. Chenault, and P.R. Boyd, J. Electron. Mater. 30, 733 (2001).
10. J.D. Benson, A.J. Stoltz, A.W. Kaleczyc, M. Martinka, L.A. Almeida, P.R. Boyd, J.H. Dinan, J. Electron. Mater. 31, 822 (2002).
11. M.A. Lieberman and A.J. Lichtenberg, Principles of Plasma Discharges and Materials Processing (New York: John Wiley & Sons, Inc., 1994), pp. 472-511.
12. L.S. Hirsch, Z. Yu, S.L. Buczkowski, T.H. Myers, and M.R. Richards-Babb, J. Electron. Mater. 26, 534 (1997).
13. J.D. Benson, A.J. Stoltz, A.W. Kaleczyc, M. Martinka, L.A. Almeida, P.R. Boyd, and J.H. Dinan, Proceedings of the SPIE-The International Society for Optical Engineering, vol. 4795, no. 1 (2002), pp. 129-135.
14. J.D. Benson, A.J. Stoltz, A.W. Kaleczyc, and J.H. Dinan, XIX Proc. SPIE Adv. Resist Technol. Processing 4690, 1224 (2002).
15. J.P. Verboncoeur, M.V. Alves, V. Vahedi, and C.K. Birdsall, J. Comp. Phys. 104, 321 (1993).
16. H. Jansen, M. de Boer, R. Wiegerink, N. Tas, E. Smulders, C. Neagu, and M. Elwenspoek, Microelectron. Eng. 35, 45 (1997).
17. M. Elwenspoek and H.V. Jansen, in Silicon Micromachining (Cambridge, UK: Cambridge University Press, 1998), pp. 331-381.
18. D. Keil and E. Anderson, J. Vac. Sci. Technol. B19, 2082 (2001).
19. J.C. Arnold and H.H. Sawin, J. Appl. Phys. 70, 5314 (1991).
20. J.W. Coburn and H.F. Winters, Appl. Phys. Lett. 55, 2730 (2730).
21. K.P. Giapis, G.R. Scheller, R.A. Gottscho, W.S. Hobson, and Y.H. Lee, Appl. Phys. Lett. 57, 983 (1990).
22. J. Matsui, N. Nakano, Z.L. Petrovic, and T. Makabe, Appl. Phys. Lett. 78, 883 (2001).
23. M. Watanabe, D.M. Shaw, and G.J. Collins, Appl. Phys. Lett. 79, 2698 (2001).
24. G. Kokkoris, E. Gogolides, and A.G. Boudouvis, J. Appl. Phys. 91, 2697 (2002).
25. E.S.G. Shaqfeh and C.W. Jurgensen, J. Appl. Phys. 66, 5314 (1989).
26. T. Ohiwa, A. Kojima, M. Sekine, I. Sakai, S. Yonemoto, and Y. Watanabe, Jpn. J. Appl. Phys. 37, 5060 (1998).
27. S. Samukawa and T. Mukai, Thin Solid Films 374, 235 (2000).
28. M.J. Buie, J.T.P. Pender, and P.L.G. Ventzek, Jpn. J. Appl. Phys. 36, 4838 (1997).
29. J.-H. Ting, J.-C. Su, and S. Su, Microelectron. Eng. 54, 315 (2000).
30. Y-S. Kim, P.T.-C. Wei, G.R. Tynan, R. Charatan, and D. Hemker, Jpn. J. Appl. Phys. 37, 327 (1998).
31. D.J. Resnick, S.V. Pendharkar, W.J. Dauksher, K.D. Cummings, W.A. Johnson, and C. Constantine, Microelectron. Eng. 30, 221 (1996).
32. S.G. Ingram, J. Appl. Phys. 68, 500 (1990).
33. S. Fang, C. Chiang, D. Fraser, B. Lee, P. Keswick, M. Chang, and K. Fung, J. Vac. Sci. Technol. A14, 1092 (1996).
34. S. Tanaka, K. Rajanna, T. Abe, and M. Esashi, J. Vac. Sci. Technol. B19, 2173 (2001).
35. M.A. Lieberman and A.J. Lichtenberg, Principles of Plasma Discharges and Materials Processing (New York: John Wiley & Sons, Inc., 1994), pp. 412-449.
36. J. Liu, G.L. Huppert, and H.H. Sawin, J. Appl. Phys. 68, 3916 (1990).
37. J. Janes, J. Vac. Sci. Technol. A12, 97 (1994).
A.J. STOLTZ,1 J.D. BENSON,1 P.R. BOYD,2 M. MARTINKA,1 J.B. VARESI,3 A.W. KALECZYC,1 E.P.G. SMITH,3 S.M. JOHNSON,3 W.A. RADFORD,3 and J.H. DINAN1
1.-U.S. Army CECOM RDEC Night Vision and Electronic Sensors Directorate, Ft. Belvoir, VA 22060. 2.-U.S. Army Research Laboratory, Adelphi, MD 20873. 3.-Raytheon Infrared Operations, Goleta, CA 93117.
Copyright Minerals, Metals & Materials Society Jul 2003
Provided by ProQuest Information and Learning Company. All rights Reserved
