MEMS TECHNOLOGY

Evaluating the use of hard-mask films during bulk silicon etching

Ken Goldman, K. Sooriakumar, Cindy Ray, and Mark Schade, Motorola

Bulk-micromachined silicon pressure sensors typically consist of a diaphragm that deflects when subjected to a pressure load, a boron-implanted piezoresistive transducer that translates strain to a differential voltage, and metal pads that are used to interface with other system components. Signal-conditioning circuitry for calibration or amplification is optional, but is often included as well. The piezoresistive transducer is strategically placed near the edge of the diaphragm, since that is a high-strain lo-cation, and each sensor is designed such that its output voltage is linearly proportional to the applied pressure in its operating range.^1,2 Figure 1 shows perspective cross-sectional views of a typical bulk-micromachined sensor.

Figure 1: Perspective view of a typical bulk-micromachined pressure sensor (a) and its cross section (b).

Figure 2: Schematics of the cavities-last process (a) and cavities-middle process (b).

During sensor fabrication the diaphragm is formed either prior to or following metallization. Known as the cavities-middle and cavities-last processes, respectively, these alternatives are illustrated in Figure 2. Because the silicon wafer becomes fragile after cavity etch, the cavities-last process is preferable from a manufacturing standpoint. Cavities-last etching also lends itself very well to etch-stop technology for diaphragm thickness control. However, the frontside of the wafer has to be protected during the etching process to prevent metal interconnects and bond pads from being etched. Deposition of a hard-mask film is one method of shielding frontside metallization from the corrosive etchant. Because exposure to high temperatures must be limited following metallization, LPCVD silicon nitride or thermal oxide (SiO^₂) cannot be used for this purpose; either films that can be deposited at low temperatures or spin-on coatings are required. The study presented here investigated the use of two such films — LPCVD silane-based silicon oxide and PECVD silicon nitride — as metallization overcoats. These materials are deposited using the following chemical reactions, respectively.³

Experimental Setup

The structure of the silicon wafers used for this study, shown in Figure 3, represents the frontside topography of a typical pressure sensor. After a thermal oxide was grown, aluminum was sputtered and patterned on the wafer surface. Using PECVD, a passivation layer of nitride was then deposited and patterned over the bond pads. Lastly, silane-based silicon oxide and silicon nitride were deposited via LPCVD and PECVD, respectively.

Figure 3: Device structure used for the DOE study.

The design of experiments (DOE) technique was used to set up a simple 2² factorial screening experiment to determine the effectiveness of the hard-mask films for frontside metal protection during cavity etch. The independent variables were LPCVD oxide thickness (10 and 30 kÅ) and PECVD nitride thickness (0 and 4 kÅ), while the response was cavity etch process yield. The film thicknesses were chosen based on existing design rules, topography, and manufacturing processes.

Because of its high selectivity to dielectrics, tetramethylammonium hydroxide (TMAH) was chosen as the cavity etchant. It has been reported that by controlling the pH of the etchant solution, the TMAH etch rate of aluminum can be controlled to the point where it becomes insignificant.⁴ Control of the pH is typically accomplished by either dissolving silicon in the solution or adding acid to it prior to etching. The hard-mask scheme being evaluated was intended for use as a protection barrier in conjunction with pH-controlled TMAH. During the testing the etch process was conducted at a temperature of 95°C in a 25 wt% TMAH solution for various lengths of time (30—480 minutes) and the solution was not pH controlled. No cavities were actually formed since a diaphragm on the wafer was not required to determine the effects of the etch on the frontside metal. It was also assumed that any change in TMAH pH that might be caused by dissolved silicon would not affect the outcome of the experiment.⁴ The wafers exposed to TMAH were examined for damage using scanning electron microscopy (SEM).

Varying the etch time created an additional independent variable, giving the experiment a 2 x 2 x N design, where N is the number of levels of time. However, the length of time to completely etch such cavities is essentially constant — approximately 8 hours for a 15-mil-thick substrate under the conditions described above. The purpose of varying the time increments was simply to determine the time to failure for the hard mask, as will be explained below.

Results and Discussion

When the first set of test wafers were etched in the 25 wt% TMAH solution at 95°C for 8 hours, the typical cavity etch time, the result was catastrophic: no good die were found on any of the wafers. In other words, all DOE cells exhibited zero yield. SEM micrographs indicated that the primary failure mechanism was lifting and cracking of the hard-mask films, which permitted etching of the metal pads underneath (see Figure 4). In theory, this metal etching should not have occurred because the metal was covered by dielectric films with high selectivities in TMAH. However, holes such as that visible in the figure allowed TMAH to reach the metal. Traditionally, the primary failure mechanism for cavity etch has been etch pits in the substrate, but etch pits, which are commonly caused by defects in the silicon substrate such as scratches and particles, were not observed on the test wafers. The implication is that the lower thermal oxide layer protected the substrate even though the hard mask failed. (This result was expected since the TMAH etch rate of thermal oxide is <10 Å/min.)

Figure 4: SEM micrographs of a failure site after TMAH etch at two different magnifications.

A second set of replicate wafers, which had been inspected and found to be free of holes, were used to determine the rate at which the failure mechanism of random hole formation occurs. The wafers were immersed in TMAH for 30 minutes and were then examined under a microscope. This visual inspection indicated that holes in the mask were already abundant after the 30-minute etch. Several failure spots were then cross-sectioned using a focused ion beam (FIB) and examined via SEM. As seen in Figure 5, the TMAH had begun etching the metal exposed by the holes in the hard mask. Further etching of the metal resulted in a freestanding layer of hard-mask film (inadvertent surface micromachining), which could crack and break during spin-rinse drying, like the wafer shown in Figure 4.

Figure 5: SEM micrographs of an FIB cross section of a failure site at two different magnifications.

Careful inspection indicated that some of the failure locations were more susceptible to TMAH etching than others. Figure 6 shows an SEM micrograph of an unetched spot where the hard-mask films are conforming to an underlying irregularity. The micrograph reveals that the film structure at this spot is much different than that in surrounding regions, implying that some type of residual contamination is present. The residue was cross-sectioned using an FIB and was subsequently examined via energy-dispersive x-ray spectroscopy (EDS). The EDS graph of the cross section, presented as Figure 7, indicates a strong presence of silicon and oxygen, which was not surprising since one of the hard-mask films was silane-based silicon oxide.

Figure 6: SEM micrograph showing an unetched area where the hard mask is overcoating an embedded residual contaminant.

Figure 7: EDS graph of the residue shown in Figure 6.

It is well known that the reaction between SiH^₄ and excess oxygen produces silicon oxide particles that adhere to reactor walls as a white powder and may also contaminate wafer surfaces.⁵ Closer examination of additional test wafers prior to TMAH etch revealed that particle contamination resembling the residue seen in Figure 6 was present throughout the wafers. Because embedded particles create weak regions in film, this particle contamination was assumed to be the root cause of the failures observed on both sets of DOE test wafers.

In order to verify the failure mechanism, a separate DOE study was conducted using a PECVD tetraethoxysilane-based oxide as the hard mask. In this experiment, the aluminum pads were fully protected and no evidence of failures were seen after TMAH etch. Therefore, it was concluded that an oxide hard mask is sufficient to withstand TMAH if the oxide film does not contain embedded particle contamination.

Conclusion

When the DOE technique was used to evaluate the use of deposited silane-based silicon oxide and silicon nitride as a hard mask for aluminum during silicon pressure sensor etching in TMAH solutions, all of the experimental cells had a process yield of zero. After SEM analyses and further investigation, the cause of the yield failure was attributed to particle contamination created by the reaction of silane and excess oxygen. Planarization techniques to reduce the required protective oxide thicknesses, increased preventive maintenance (PM) to prevent particle formation in the process chamber, and improved SPC to optimize PM scheduling will likely improve the yields for such a process, but will not be a cure-all and will result in higher manufacturing costs.

Acknowledgments

The authors would like to thank the Motorola MEMS-1 production fab team members for their support and Bruce Huling for encouraging us to write this article. We also would like to thank Demetre Kondylis and Dan Wallace for supporting this project.

References

1. Tufte ON, Chapman PW, and Long D, "Silicon Diffused-
Element Piezoresistive Diaphragms," Journal of Applied Physics, 33:3322—3327, 1962.

2. Samaun S, Wise K, and Angell JB, "An IC Piezoresistive Pressure Sensor for Biomedical Instrumentation," IEEE Transactions on Biomedical Engineering, BME-20(2):101—109, 1973.

3. Ghandhi SK, VLSI Fabrication Principles, New York, Wiley, pp 422—428, 1983.

4. Tabata O, "pH-Controlled TMAH Etchants for Silicon Micromachining," presented to the 8th International Conference on Solid-State Sensors and Actuators, Stockholm, Sweden, June 1995.

5. Wolf S, and Tauber RN, Silicon Processing for the VLSI Era, Sunset Beach, CA, Lattice Press, pp 183—184, 1986.

Ken Goldman has been an electrical engineer in the micromachining development group for Motorola's sensor products division in Phoenix, AZ, since 1994. The coauthor of four technical articles, he has five patents pending. Goldman received his BS and MS degrees in electrical engineering and applied physics (1992 and 1994, respectively) from Case Western Reserve University, specializing in micromachining. (Goldman can be reached at 602/244-4089; e-mail, RP3217@email.sps.mot.com)

Kathirgamasundaram Sooriakumar is a technical member of Motorola's sensor process development group, which he joined in 1993. He previously worked at Ford Motor Co.'s electronic division and at Ford Micro Electronics (FMI) in Colorado Springs. During his career, Sooriakumar has codeveloped many sensor products, including silicon nozzles for fuel injectors, capacitive and piezoresistive pressure sensors for automotive and biomedical applications, and accelerometers for air bag applications. He also has coauthored many technical papers and holds patents in the silicon micromachining field. He received a BS in physics from the State University of New York at Brockport in 1986 and an MS in electrical engineering and applied physics from Case Western Reserve University in 1989.

Cindy Ray is a process technician in Motorola's sensor process development group. Since joining the company in 1980 as a production operator she has worked in radio-frequency test and assembly, and in the zener and bipolar fabs. She contributed to developing and transferring electrochemical etching processes for pressure sensors.

Mark Schade is senior staff scientist at Motorola's chemical and surface analysis laboratory in Phoenix, where his responsibilities include specialized material characterization studies using scanning electron microscopy, Auger electron spectroscopy, and atomic force microscopy. Prior assignments have included metallurgical characterization studies and process engineering/production line support duties. A member of ASM International, the American Vacuum Society, the Microscopy Society of America, and the Materials Research Society, Schade has published more than 10 technical papers and was recently granted a Special Patent Engineering Award. He has a BS in chemical engineering (1982) from Michigan State University and an MS in materials science engineering (1995) from Arizona State University.

MicroHome | Search | Current Issue | MicroArchives
Buyers Guide | Media Kit