Issue



Observations on the chemistry and physics of STI etch in Cl2-Ar plasmas


12/01/2000







overview
The detailed mechanisms of a plasma etch process explain its ability to improve the taper profile for shallow trench isolation. A two-gas etch chemistry is used, and the controlled redeposition of etch products along the sidewall of the etched feature is a key part of the process.

Paul Werbaneth and John Almerico, Tegal Corp., Petaluma, California

Shallow trench isolation (STI) schemes have replaced LOCOS (local oxidation of silicon) isolation for device manufacturing at the 0.18µm process node. Inter-related factors among the major processing steps of silicon etch, dielectric formation, and chemical mechanical planarization determine the actual performance of the isolation structures in the STI process module.


Figure 1. STI feature local etch environment.
Click here to enlarge image

For STI plasma etch, etch depth uniformity and the sidewall profile characteristics of the etched silicon trench are among the important responses that must be optimized. The need for precise control of the trench profile has become even more important as feature sizes decrease. It has been observed, however, that STI etch processes can have significant etch profile variations (microloading) between narrow and wide features [1]. This profile microloading has been reported for various plasma etch reactor configurations, using several combinations of reactive chemistries, for etch structures with either photoresist or silicon nitride as the masking material. Generally, there seems to be a significant difference in STI structure profile microloading results for plasma etch systems using globally polymerizing etch chemistries, like Cl2-HBr-O2 families, and those using locally polymerizing etch chemistries, where Cl2 is the sole reactive component.

An explanation that accounts for reported etch profile microloading differences involves a look at the very local environment within the etched trench. There, the interaction of energetic bombardment by ions from the plasma with species adsorbed on the silicon surface is proposed as the mechanism that allows for control of STI profile angles with Cl2-Ar chemistries while keeping etch profile microloading to a minimum. This control over local etch conditions can be obtained from high-density plasma sources using dual RF frequencies, where ion bombardment at the wafer using low-frequency RF bias power is independent of the source plasma conditions.

Silicon etching in chlorine plasmas
The etching of silicon in a chlorine plasma environment, where the silicon surface is subject to ion bombardment, is determined by the action of three components: a sputter etch component, a thermal or "spontaneous" etch component, and an ion-enhanced etch component [2]. Typically, plasma etching of silicon for applications such as STI formation is carried out at moderate (20-80°C) temperatures, at which the spontaneous etch rate of silicon by atomic chlorine is negligible [3]. Ion-assisted or ion-enhanced etch is the dominant mechanism by which the etch front proceeds into the silicon surface in any plasma reactor in which there is an adequate supply of halogen reactant. Physical sputtering of silicon can also account for the removal of an appreciable volume of silicon at higher ion energies, although care must be taken to avoid the damage that might result from high-energy ions impinging on the device wafer.


Figure 2. HRe- dual-frequency reactor.
Click here to enlarge image

The volatile products evolved during the etch of silicon in a Cl2-Ar plasma will be some combination of silicon and the reactive halogen atom, along with atomic silicon sputtered from the mixed layer of adsorbed chlorine and silicon from the substrate. The silicon chloride etch product, SiClx, will be distributed among several possible forms: SiCl, SiCl2, SiCl3, and SiCl4, for example, with SiCl4 being the fully saturated product molecule.

The exact outcome will be a function of several factors. Chiefly, for the system considered here (silicon etching in a Cl2-Ar plasma), the distribution of etch products is a function of incoming ion bombardment energy and the neutral/ion flux ratio at the etch surface. As ion bombardment energy is decreased, SiCl4 production is favored over SiCl2 [4]; or, as ion bombardment energy is increased sufficiently, the etch product desorbed from the silicon surface will transition further from SiCl2 to SiCl and even Si [5]. (SiCl3 does not seem to be observed much.)

The neutral/ion flux ratio at the silicon surface has an effect similar to ion bombardment. Running with a very low neutral/ion flux ratio will favor the path of silicon removal from the etch surface by sputtering. Increasing the ratio will favor in turn the evolution of SiCl, SiCl2, and SiCl4 etch product species [6]. The precise mixture of etch product species can have a profound effect on STI etch profile evolution. Etch product redeposition along the evolving feature sidewall can result in tapering of the trench as it develops into the silicon [7]. Consider the local environment depicted in Fig. 1.

As etch product evolves along the silicon etch front and desorbs from the surface under the influence of ion bombardment, there will be a possibility that, depending on the "stickiness" of the product molecule SiClx (or of the sputtered Si), there will be redeposition on the wall of the etched feature. Stickiness is represented as the sticking coefficient, SP, which can range from a value of 0 (no propensity to stick) to 1 (guaranteed to stay in place). At wafer temperatures typical of commercial plasma etch reactors, 20-80°C, the sticking coefficient of a sputtered silicon atom is near 1. SiCl and SiCl2 sticking coefficients range between 0.1 and 0.3 [8, 9], and SiCl4 has a sticking coefficient estimated at <0.002 [9].

STI profile control
Control of the STI etch profile is one of the important process capabilities for STI etch module users. The ideal profile of the etched trench is a positive taper, which may range from a target value of anywhere from 75° to 89°. A traditional approach to trench etch profile control has been to employ combinations of Cl2, HBr, and O2, which are generally recognized as being a family of gases that can produce polymer deposits when used for silicon etch [10]. These deposits, originating globally from the gas phase, can be used to taper etch profiles for the STI feature, although a drawback may be that the depositing material will coat other surfaces within the plasma reactor, which may lead to issues with process stability and particle levels within the tool.


Figure 3. Silicon profile angle and etch rate vs. wafer bias power.
Click here to enlarge image

A second approach to STI profile control utilizes the depositing nature of the silicon etch product. If etch conditions within the local environment are carefully controlled, then "sticky" etch product evolved during the trench etch will redeposit on the trench sidewall. The amount of redeposition will determine the degree of taper. A dual-frequency high-density plasma reactor was used to develop STI etch processes using Cl2-Ar gas mixtures. A representation of the high-density reflected electron (HRe-) process chamber is shown in Fig. 2.


Figure 4. Silicon profile angle vs. Ar flow and wafer bias power.
Click here to enlarge image

A notable feature of the type of reactor in Fig. 2 is the use of low-frequency RF power for wafer bias. It has been observed that low-frequency biasing (600kHz vs. 1MHz) of the wafer electrode in a plasma etch system has a powerful influence over silicon etch rates in Cl2 plasmas. The mechanism that accounts for this is control of ions accelerated across the plasma sheath within one cycle of the RF electric field (as opposed to control of the time-average acceleration response that occurs with higher RF bias frequencies) [11].


Figure 5. Silicon etch rate vs. wafer bias power and pressure.
Click here to enlarge image

An initial feasibility test was performed in the HRe- reactor to etch shallow silicon trenches using Cl2-Ar chemistries. In this test, silicon nitride hard masks, previously defined, acted as the pattern for the etched trench feature. Wafer bias power was changed from a low value (45W) to high (75W), with a centerpoint. SEM images of a cross-sectioned feature post-etch (and post-HF clean) were used to determine trench etch profile and etch depth, from which the silicon etch rate was calculated. Figure 3 is a graph of the silicon profile angle and silicon etch rate vs. the applied RF bias power.


Figure 6. Etch profile microloading results for magnetically enhanced reactive ion etching, and two high-density reflected electron etching processes.
Click here to enlarge image

Notable in Fig. 3 is the strong response of both profile angle and silicon etch rate to changes in wafer bias power in this single-factor experiment. A design of experiments (DOE) approach was used to create a screening experiment with two factors, Ar gas flow and wafer bias (kHz) power levels, and it shows the same trend for trench profile angle (Fig. 4). As the kHz power level used for wafer bias is increased from 10W to 100W, the silicon profile angle decreases from 89° to 81°.

A second two-factor DOE, now with wafer bias power and pressure as variables, shows that the silicon etch rate increases significantly with increasing wafer bias power. These results are shown in Fig. 5. Our interpretation of these observed results is based on our understanding of the silicon etch mechanisms described above. We reason as follows:

  1. Generally, when source power, Cl2 flow, and pressure are held constant, as is the case in the single-factor experiment, the reactive components of the plasma are assumed to be present in fixed amounts.
  2. The removal of SiClx etch product from horizontal surfaces on the wafer is accomplished by the action of perpendicular bombardment by ions accelerated across the plasma sheath.
  3. Increased ion bombardment (increased bias power) enhances the formation and the desorption of SiClx from the silicon surface. The observed effect is a net increase in silicon etch rate with increasing bias power, given the excess of reactant under these experimental conditions.
  4. Unsaturated SiClx etch product will tend to redeposit on adjacent surfaces, subject to the stickiness of the SiClx species. The significant angular dependence of ion driven silicon etch in Cl2 plasmas suggests that very little etching will occur on surfaces tapered in the range of 75° to 89°, meaning a tapered etch profile can propagate down into the silicon feature.
  5. Therefore, the observed increase in the ion driven silicon etch rate, at constant reactant flow, favors the production of sticky SiCl2 or SiCl (by either the path of ion energy dependence or neutral/ion ratio dependence). This production leads to an increase in etch product redeposition on the trench sidewall, resulting in the observed decrease in trench profile angle.

A practical benefit to encouraging local redeposition of etch product for trench profile control is the minimal profile microloading observed as a result. Trench profile angles were obtained for small and large features etched in the dual-frequency reactor with either an 80° (HRe- 80) or 89° (HRe- 89) target profile angle using a silicon nitride hard mask. These results compare favorably to published STI results for a magnetically enhanced plasma etch tool using a Cl2-HBr-N2-He-O2 chemistry (Fig. 6) [1].

Conclusion
Effective control of silicon profile taper for STI etch has been demonstrated using a simple two-gas etch chemistry. The proposed mechanism for profile control in the dual-frequency reactor involves the evolution of unsaturated SiCl2 and SiCl etch products which, being sticky, have the tendency to redeposit along the silicon sidewall within the etched feature. Low-frequency RF power is used here for wafer bias. The ability to accelerate ions from the plasma to the wafer within one RF cycle is a key feature of this trench profile taper control scheme. An advantage of establishing local redeposition of etch products as the means for changing trench taper arises when etch profile microloading results are compared to other schemes. Significantly less profile microloading is observed for the Cl2-Ar chemistry than for other reported reactant families. Reduced profile microloading has an impact on future device generations, where the need for real estate-consuming dummy features to ameliorate profile control issues will be increasingly unattractive. n

References

  1. Chung-Kyu Yeon, Hyuk-Joon You, "Deep-submicron Trench Profile Control Using a Magnetron Enhanced Reactive Ion Etching System for Shallow Trench Isolation," J. Vac. Sci. Technol. A, Vol. 16, No. 3, p. 1502, May/June 1998.
  2. Jane P. Chang, Herbert H. Sawin, "Kinetic Study of Low Energy Ion Enhanced Polysilicon Etching Using Cl, Cl2, and Cl+ Beam Scattering," J. Vac. Sci. Technol. A, Vol. 15, No. 3, p. 610, May/June 1997.
  3. J.P. Chang, et al., "Kinetic Study of Low Energy Argon Ion-enhanced Plasma Etching of Polysilicon with Atomic/Molecular Chlorine," J. Vac. Sci. Technol. A, Vol. 15, No. 4, p. 1853, Julu/Aug. 1997.
  4. Rebecca A. Rossen, Herbert H. Sawin, "Time-of-flight and Surface Residence Time Measurements for Ion-enhanced Si-Cl2 Reaction Products," J. Vac. Sci. Technol. A, Vol. 5, No. 4, p. 1595, July/Aug. 1987.
  5. R.S. Goodman, N. Materer, S.R. Leone, "Ion-enhanced Etching of Si(100) with Molecular Chlorine: Reaction Mechanisms and Product Yields," J. Vac. Sci. Technol. A, Vol. 17, No. 6, p. 3340, Nov./Dec. 1999.
  6. J.W. Coburn, "Ion-assisted Etching of Si with Cl2: The Effect of Flux Ratio," J. Vac. Sci. Technol. B, Vol. 12, No. 3, p. 1384, May/June 1994.
  7. J.M. Lane, et al., "Feature Evolution During Plasma Etching. II. Polycrystalline Silicon Etching," J. Vac. Sci. Technol. A, Vol. 18, No. 1, p. 188, Jan./Feb. 2000.
  8. J.P. Chang, A.P. Mahorowala, H.H. Sawin, "Plasma-surface Kinetics and Feature Profile Evolution in Chlorine Etching of Polysilicon," J. Vac. Sci. Technol. A, Vol. 16, No. 1, p. 217, Jan./Feb. 1998.
  9. M. Tuda, K. Ono, K. Nishikawa, "Effects of Etch Products and Surface Oxidation on Profile Evolution During Electron Cyclotron Resonance Plasma Etching of Poly-Si," J. Vac. Sci. Technol. B, Vol. 14, No. 5, p. 3291, Sept./Oct. 1996.
  10. G.D Tipton, M.G. Blain, "Optimization of an Electron Cyclotron Resonance Plasma Etch Process for n+ Polysilicon: HBr Process Chemistry," J. Vac. Sci. Technol. B, Vol. 12, No. 1, p. 461, Jan./Feb. 1994.
  11. Seiji Samukawa, Hiroto Ohtake, "Effect of Supplied Substrate Bias Frequency in Ultrahigh-frequency Plasma Discharge for Precise Etching Processes," J. Vac. Sci. Technol. A, Vol. 14, No. 6, p. 3004, Nov./Dec. 1996.

Paul Werbaneth received his BS in chemical engineering from Cornell University. In 1983, he joined Tegal Corp., where he is currently a staff process engineer. He has also worked for Hitachi America Ltd. and Intel Corp. He is a member of the American Vacuum Society, the Electrochemical Society, and the Materials Research Society. Tegal Corp., 2201 S. McDowell Boulevard, Petaluma, CA 94955-6020; ph 707/765-5608, fax 707/773-3015, e-mail [email protected].

John Almerico received his BS degree in chemical engineering at the University of California, Berkeley, and he is currently a senior marketing manager for advanced products at Tegal. He joined Tegal in 1984 and has worked in a variety of individual and management positions in customer process support, customer applications, and technical marketing.