The effects of wet-etch parameters on wafer-thinning etchants
09/01/2005
Designed experiments determine the effects of wet-etch parameters on the performance of several different types of wafer-thinning etchants. Based on the use of a single-wafer spin-processor for 200mm wafers, the studies investigate the effects of temperature, chuck speed, flow rate, and dispense profile on etch rates, etch uniformity, and surface roughness during wafer-thinning process steps.
Flexible media applications such as smart cards are driving die thickness to below 100μm [1]. Grinding, chemical mechanical polishing (CMP), plasma etching, and chemical wet etching can be employed to thin wafers [2]. Thinning by grinding or CMP can introduce surface microscratches that will decrease the mechanical strength of a thinned wafer. A wet chemical etch is employed after grinding or CMP to remove these defects.
Grinding or CMP cannot, however, thin a wafer to a thickness of <120μm. To obtain thinner wafers, a wet etch must be employed. Several types of wafer-thinning etchants are available. Bulk silicon etchants are designed to rapidly etch silicon in order to quickly thin wafers or remove grinding-induced scratches. Polish etchants are slower, and are designed to impart a very smooth wafer surface. Texture etchants are designed to uniformly roughen the silicon surface and thus facilitate bonding to that surface.
For a repeatable and highly selective backside etch process, it is necessary to utilize an extremely consistent etchant. Honeywell Electronic Materials, in collaboration with SEZ America Inc., has completed experiments designed to investigate the effects of equipment parameters on wafer etch performance. An SEZ SP 203 single-wafer spin-processor, configured for 200mm wafers, was used in these evaluations. The effects of process temperature, chuck rotational speed, etchant flow rate, and dispense profile on the silicon etch rate, etch nonuniformity, and surface roughness were determined for a bulk silicon etchant, a polish etchant, and a texture etchant.
Designed experiments
Each of the three four-factor, three-level, Box-Behnken-type designed experiments employed 20 runs and 20 wafers. The equipment process parameter settings for both the bulk silicon and the silicon polish etchant chemistries were:
- temperature = 22, 25, and 28°C;
- chuck speed = 400, 600, and 800rpm;
- flow rate = 1.6, 1.8, and 2.0L/min; and
- dispense profile = 70, 75, and 80mm.
The factor settings for the texture etchant were:
- temperature = 50, 55, and 60°C;
- chuck speed = 300, 450, and 600rpm;
- flow rate = 1.0, 1.3, and 1.6L/min; and
- dispense profile = 40, 45, and 50mm.
The single-wafer spin-processor dispenses chemicals onto the surface of a wafer rotating on a process chuck. Chuck rotation speed, chemical flow rate, temperature, and the dispense profile (i.e., the track of the chemical dispense nozzle relative to the speed of the track across the wafer) are all highly controlled parameters throughout processing to achieve specific etch characteristics on the wafer. Various combinations of these parameters dictate etch characteristics, such as etch rate, uniformity, and wraparound on the bevel edge to the wafer frontside.
The average silicon loss (Δd) was determined by averaging the difference in wafer thickness at 29 locations on each 200mm wafer prior to and after etching. The silicon etch rate was calculated by dividing the average silicon loss by the etch time. The percent silicon etch nonuniformity, N(%), was calculated by employing the following equation:
where σ is the standard deviation of the etch measurements. If the wafer center etched faster than the wafer edge, N was assigned a positive value. If the wafer center etched slower than the wafer edge, N was assigned a negative value. The post-etch surface roughness average (Ra) of a wafer was determined by measuring the center of that wafer after etch using a KLA-Tencor P2 profilometer.
Bulk silicon etchant results
Figures 1a and 1b present surface and contour plots of the silicon etch rate of a Honeywell bulk silicon etchant as a function of chuck speed and temperature when flow rate and dispense profile were held at their middle settings. The silicon etch rate is primarily determined by the chuck rotational speed. For instance, the silicon etch rate increases from ~27 to 38µm/min as the chuck increases from 400 to 800rpm, while the temperature is held constant at 25°C.
The etchant temperature has less of an effect than the chuck speed on the silicon etch rate, which only increases from 31 to 33µm/min as the temperature increases from 22 to 28°C, while the chuck speed is held constant at 600rpm. The flow rate and dispense profile have a small effect on the silicon etch rate.
Surface and contour plots of the silicon etch nonuniformity of the bulk silicon etchant are presented in Fig. 2a and 2b, respectively, as a function of chuck speed and temperature when flow rate and dispense profile are held at their middle settings. The silicon etch nonuniformity is primarily determined by the chuck rotational speed. The silicon etch nonuniformity decreases from approximately +6% (center fast) to -4% (center slow) as the chuck speed increases from 400 to 800rpm, while the temperature is held constant at 25°C. A chuck speed of 750rpm and temperature of 25°C yield an absolutely uniform (N = 0) etch.
The etchant temperature also has a large effect on the silicon etch nonuniformity when the chuck speed is high. The silicon etch nonuniformity decreases from +2% (center fast) to -7% (center slow) as the temperature increases from 22 to 28°C, while the chuck speed is held constant at 800rpm. The etch nonuniformity, however, remains near +6% (center fast) as the temperature is increased from 22 to 28°C, while the chuck speed is held constant at 400rpm. The flow rate and dispense profile have a small effect on the silicon etch nonuniformity.
Silicon polish etchant results
Figures 3a and 3b show surface and contour plots of the post-etch Ra of a Honeywell silicon polish etchant as a function of chuck speed and dispense profile when flow rate and temperature are held at their middle settings. When the dispense profile is large, Ra increases with increasing chuck speed. For example, Ra increases from ~70 to 180Å as the chuck increases from 400 to 800rpm, while the dispense profile is held constant at 80mm. When the dispense profile is small, Ra decreases with increasing chuck rotational speed, such as from ~200 to 0Å as the chuck increases from 400 to 800rpm, while the dispense profile is held constant at 70mm. High chuck speed coupled with a small dispense profile will produce the smoothest surface (i.e., lowest Ra value). A high chuck speed coupled with a large dispense profile, or a low chuck speed coupled with a small dispense profile, will produce the roughest surface (i.e., highest Ra value).
Surface and contour plots of the silicon etch rate using the silicon polish etchant as a function of chuck speed and temperature are presented in Fig. 4a and 4b, respectively, when flow rate and dispense profile are held at their middle settings. The silicon etch rate is primarily determined by the chuck rotational speed. For example, the silicon etch rate increases from ~12 to 16.5µm/min as the chuck increases from 400 to 800rpm, while the temperature is held constant at 25°C. The etchant temperature has less of an effect on the silicon etch rate, which only increases from 16.5 to 17µm/min as the temperature increases from 22 to 28°C, while the chuck speed is held constant at 800rpm.
null
Texture etchant results
Surface and contour plots of the post-etch Ra of a Honeywell silicon texture etchant are presented in Fig. 5a and 5b, respectively, as a function of dispense profile and temperature when chuck speed and flow rate are held at their middle settings. A calculated maximum Ra of 2250Å occurs at a temperature of 52°C and a dispense profile of 46mm. Changing either the dispense profile or temperature settings will decrease the surface roughness. At low temperature, the surface roughness is high, and the dispense profile has little effect on the surface roughness. At higher temperatures, surface roughness is low and decreases with increasing flow rate. For instance, Ra decreases from 2000 to 160Å as the dispense profile increases from 40 to 50mm, while the temperature is held constant at 60°C. When the dispense profile is large, Ra decreases from ~2200 to 1600Å as the temperature increases from 50 to 60°C, while the dispense profile is held constant at 50mm. A large dispense profile coupled with a high temperature will produce the smoothest surface (i.e., lowest Ra value).
Figures 6a and 6b show surface and contour plots of the silicon etch rate of the silicon texture etchant as a function of flow rate and temperature when dispense profile and chuck speed are held at their middle settings. Low temperatures coupled with low flow rates, or high temperatures coupled with high flow rates, yield low silicon etch rates. High temperatures coupled with low flow rates yield high etch rates. The silicon etch rate increases from ~5 to 11.5µm/min as the temperature increases from 50 to 60°C, while the flow rate is held constant at 1.0L/min.
null
Summary
Designed experiments have been completed to investigate the effects of etch parameters on performance for several wafer-thinning etchants. These evaluations have shown that the etch rates of the bulk silicon etchant and silicon polish etchant are functions primarily of the chuck rotational speed. For these etchants, the etch rate increases with increasing chuck rotational speed. On the other hand, the silicon etch rate of the texture etchant increases with increasing temperature. The etch nonuniformity of the bulk silicon etchant decreases (i.e., the etch becomes more uniform) with increasing chuck rotational speed. The surface roughness produced by the bulk silicon etchant increases with increasing chuck speed. The surface roughness produced by the silicon polish etchant, however, decreases with increasing chuck speed. Finally, the surface roughness produced by the texture etchant decreases with increasing temperature and decreasing dispense profile.
Acknowledgments
The authors gratefully acknowledge and thank Aaron Bicknell, a field applications engineer, and Gale Hansen, a metrology engineer, as well as the rest of the staff of SEZ America Inc. research lab in Phoenix, AZ, for their valuable support.
References
- M. Reiche, G, Wagner, Advanced Packaging, Vol. 12, No. 3, p. 29, 2003.
- W. Sievert, K.-U. Zimmermann, J. Starzynski, European Semiconductor, to be published.
For more information, contact John S. Starzynski at Honeywell Laboratories, Mail Stop MN14-2B45, 12001 Highway 55, Plymouth, MN 55441; ph 763/954-2841, e-mail [email protected].
Deborah Yellowaga, John McFarland, Ben Palmer, Honeywell Electronic Materials, Chandler, Arizona
Scott N. Drews, SEZ America Inc., Phoenix, Arizona