Z = ρ_i ν_i

ρ_i = density of the ith layer

ν_i = velocity of sound in the ith layer.

Since process variations can result in changes to the measured film density, monitoring the measured density value has proven to be a reliable indicator of process excursions in head manufacturing.⁶ Additional process information can be obtained from the shape and size of the picosecond ultrasonic signal. If a layer in the stack adheres weakly or has delaminated, the sound wave becomes trapped in the layers above the poor interface, and little or no signal is detected from layers underneath. The roughness of the interface also affects the measured signal, as illustrated in Figure 1. Echoes from a smooth interface are sharp, while echoes from a rough one are broad.

Figure 1: Effect of interface condition on picosecond ultrasonic echoes: (a) smooth interface, and (b) rough interface.

Excellent results have been obtained by comparing multilayer picosecond ultrasonic metal-film-thickness measurements with measurements derived from transmission and scanning electron microscopy (TEM and SEM). However, unlike TEM and SEM, the picosecond ultrasonic technique is fast and nondestructive.

Work on read/write heads has also shown that picosecond ultrasonic measurements correlate very well (R² > 0.99) with magnetic flux measurements taken with a looper and magnetometer measurements. The excellent correlation between picosecond ultrasonic thickness measurements and atomic force microscopy (AFM) measurements for a single TiW layer is shown in Figure 2. Good results were also obtained by correlating picosecond ultrasonic measurements with x-ray fluorescence (XRF) and focused ion beam (FIB) measurements.

Figure 2: Correlation between picosecond ultrasonic and AFM thickness measurements of a TiW single-layer film (red line, R2 = 0.98; black line, R2 = 1.0).

Many disk-head manufacturing processes rely on four-point-probe sheet-resistance measurements. For single-layer blanket films, picosecond ultrasonic and four-point-probe measurements correlate well. Figure 3 illustrates the excellent correlation between picosecond ultrasonic physical thickness measurements and four-point-probe sheet resistance measurements of a single-layer Ta sheet film.

Although the four-point-probe technique is essentially blind to the thin underlying layers in multilayer metal film stacks, it has historically been used to monitor such stacks. The sheet resistances shown in Figure 4 were measured on monitor wafers using a four-point-probe technique on a multilayer magnet stack covered with a thick (900–1100-Å) capping layer. Collected over a period of several months, these data were compared with picosecond ultrasonic thickness measurements of the metal film stack's capping layer from product wafers manufactured on the same process equipment. The correlation between the two techniques was found to be good.

Figure 3: Correlation between four-point-probe sheet-resistance measurement and picosecond ultrasonic thickness measurement of a single-layer Ta film (R2 > 0.99).

Although the data from the two techniques correlate well, the picosecond ultrasonic measurements provide more-detailed information about the stacks than the four-point-probe measurements, such as the thicknesses of underlying layers. In addition, since the picosecond ultrasonic measurements are from product wafers, they can be used to identify process excursions more quickly than the four-point-probe technique, and accurately reflect the thicknesses of the layers in the finished heads. Finally, the sheet-resistance measurements did not vary significantly with a relatively wide range of thickness variation. When sheet resistance was used for process control, the thickness variation in the Ta layer was several hundred angstroms. The picosecond ultrasonic technique was used to reduce the amount of process variation.

Figure 4: Production data comparing four-point-probe sheet-resistance measurements of multilayer magnet stack films on monitor wafers and picosecond ultrasonic measurements of similar films on product wafers. The data were gathered over a period of several months.

Applications

Picosecond ultrasonic technology is used at Seagate Technology to perform process control measurements.⁷ Three typical applications are in the areas of shield fabrication, thin multilayer stacks used in the read structure, and thicker multilayer stacks used in the write structure.

Shield Thickness Measurements. The shield in an MR or GMR head is a thick layer of NiFe that protects the read sensor from stray magnetic fields coming from adjacent tracks on the disk. The shield is fabricated during the initial stages of head building. Process control measurements of the shield thickness are critical, since variations in shield thickness affect the performance of the completed head.

Figure 5: Picosecond ultrasonic measurement of a 2.3-µm-thick lower shield on a product wafer.

Traditionally, the thickness of the shield could not be measured directly on product wafers because the small pattern made opaque metrology techniques such as XRF or four-point-probe technology imprac-tical. Therefore, its thickness was monitored using a reflectometer, which measured the thickness of the transparent oxide layer adjacent to the shield itself. However, this method did not provide reliable results because measurements were made after a chemical-mechanical polishing step, which polished NiFe and oxide layers at rates that differed, but not always by the same amount.

The measurement-spot size for picosecond ultrasonic technology is 15 X 30 µm or smaller, allowing for the direct measurement of shield thickness on product. A typical example of shield measurement is shown in Figure 5. The peak at approximately 900 picoseconds is generated by the echo returning from the bottom of the NiFe film. The thickness of this film layer was 2.3 µm.

Figure 6: Shield thickness measurements: (a) correlation between high-resolution metrology and picosecond ultrasonic measurements of the NiFe shield, and (b) correlation between high-resolution metrology and reflectometer oxide-thickness measurements.

Figure 6 compares picosecond ultrasonic measurements of shield thickness (a) and reflectometer measurements of oxide thickness (b) with high-resolution metrology results. The correlation between picosecond ultrasonic and high-resolution metrology measurements is better than that for reflectometer and high-resolution metrology measurements.

Picosecond ultrasonic measurements were also found to be a good predictor of final head quality. In Figure 6, it can be seen that the measured picosecond ultrasonic thickness of the shield ranged up to 2.5 µm, while the oxide thickness ranged up to only 2.35 µm. Picosecond ultrasonic measurements capture the shield protrusion above the surrounding oxide that results from the different polishing rates of NiFe and oxide. These data indicate that picosecond ultrasonic measurements provide more-precise control of shield thickness and, consequently, higher yields than reflectometry.

Figure 7: Measurement of the gold layer in a trilayer metal stack. There is good correlation between the sheet resistance of the gold layer measured on monitor wafers and the actual thickness measured on product wafers using the picosecond ultrasonic technique.

Thin Multilayer Film-Stack Measurements. Read-head builds are complex structures of very thin multilayer stacks. To achieve high yields, it is critical that the layers have the appropriate deposition thickness, that the film thickness is uniform across the wafer, and that all layers of the stack are present. The traditional method of using monitor wafers to measure each individual film layer assumes that if film thickness is correct and uniform on the monitor wafers, subsequent product wafers will all have the same characteristics. However, that is not necessarily a good assumption. Film layers on product wafers can vary when process drifts occur between monitor-wafer runs, when underlying layers are not uniform, or when film layers interact with one another.

Figure 7 compares four-point-probe sheet-resistance measurements from monitor wafers with picosecond ultrasonic thickness measurements of buried films on product wafers. The stack consists of three layers: an upper metal layer 100 Å thick, a middle gold layer approximately 2400 Å thick, and an underlying metal layer 100 Å thick. The correlation between the two techniques is good. This test indicates that picosecond ultrasonic thickness measurements of buried films can be used to tune the deposition process, reduce variations, and virtually eliminate the need for monitor wafers.

Figure 8: Measurements of a six-layer magnetic film stack: (a) monitor-wafer measurements using the four-point-probe technique, and (b) product wafer data from the picosecond ultrasonic technique. Product wafer data are less symmetrical and, therefore, more representative of actual film-thickness distribution.

Similar results can be achieved with multilayer magnet stacks having up to six layers. Figure 8 compares a wafer map containing sheet-resistance measurements from monitor wafers and a wafer map with actual thickness measurements from product wafers for a six-layer stack. While the monitor-wafer data are perfectly symmetrical, the picosecond ultrasonic data show the more characteristic distribution of film thicknesses that would be expected from a product wafer. Such results obtained from product wafers have enabled Seagate to reduce process variations and identify missing and misprocessed layers in the process flow.

Figure 9: Picosecond ultrasonic signal from a correctly processed bilayer metal stack. The 0.24-µm-thick spacer metal is the sharp peak at approximately 100 picoseconds, and the echo from the 1.7-µm magnetic metal occurs at approximately 750 picoseconds.

Thick Multilayer Film-Stack Measurements. In the write build, thick multilayer film stacks >2 µm are required. Conventionally, such layers were monitored for uniformity and misprocessing with step-height measurements. Other techniques such as XRF or FIB cannot be used to monitor these layers on product wafers because of the small pattern site. Disadvantages of the step-height technique include a loss of cycle time, because each layer must be measured after deposition, and a lack of information on the quality of the interface between layers. In contrast, pico-second ultrasonic technology can measure multiple thick layers simultaneously at process speeds and can detect processing errors.

Figure 9 is a characteristic picosecond ultrasonic signal from a correctly processed bilayer metal stack. The 0.24-µm-thick spacer metal is seen as a sharp peak at approximately 100 picoseconds. The echo from the 1.7-µm magnetic metal occurs much later, at approximately 750 picoseconds. That result is expected, since the sound wave takes significantly longer to travel through the much thicker magnetic metal than through the spacer metal.

Figure 10: (a) Picosecond ultrasonic signal from a misprocessed bilayer metal stack revealing that the spacer metal peak is very weak and occurs earlier than expected, and (b) cross-section TEM of the area, confirming that the spacer metal was incorrectly processed and thinner than it should be.

In contrast to Figure 9, Figures 10 through 12 present examples of misprocessing detected by the picosecond ultrasonic technique. Figure 10a shows a signal from the same nominal stack structure as that in Figure 9, with a magnetic peak that appears to have the same shape and occurs at approximately the same time. However, the spacer metal peak is very weak, and the signal peak occurs earlier than expected. A cross section TEM of the area in Figure 10a confirms the picosecond ultrasonic result: the spacer metal was incorrectly processed and is thinner than its nominal thickness.

Figure 11: Picosecond ultrasonic signal from a misprocessed bilayer metal stack. The peak for the magnetic metal layer is missing, indicating that the magnetic layer is too thick.

The measured picosecond ultrasonic signal in Figure 11 indicates that the magnetic layer is too thick. Echoes from the magnetic metal do not appear at 750 picoseconds, as expected from a 1.7-µm thick film, but begin to return at approximately 990 picoseconds, indicating that the film is about one-third thicker than it should be.

Figure 12 shows the picosecond ultrasonic signal resulting from a bilayer film that has delaminated. The delamination eliminates or blurs the echo peaks that appear in multilayer films with smooth interfaces. By monitoring these thick stacks on process wafers, process excursions can be identified almost immediately, and the process can be corrected before additional wafers are put at risk.

Figure 12: Picosecond ultrasonic signal from a misprocessed bilayer metal stack. The normal peaks (as shown in Figure 9) are absent, indicating that the metal layers have delaminated.

Conclusion

Because of its ability to measure metal films and film stacks on product wafers nondestructively, picosecond ultrasonic technology has improved the yields of read/write heads at Seagate Technologies. By measuring directly on product wafers, the costs and lost production capacity associated with the use of monitor wafers has been reduced or, in some cases, eliminated.

With picosecond ultrasonic technology, it is possible to monitor the thicknesses of individual layers on product wafers and to identify missing or misprocessed layers in the completed stack. The technology can also enable manufacturers to get direct feedback from product wafers about the condition of their deposition tools, allowing them to identify drifting processes before yields are affected and providing information that can improve tool-to-tool performance matching. Such information can be used to tune process equipment, reducing process variations and improving overall yields. Picosecond ultrasonic technology can also identify process problems in-line, enabling rapid root-cause analysis and eliminating further misprocessing.

Acknowledgments

The authors would like to acknowledge Markus Michel from Seagate Technology in Springtown, Northern Ireland, for his help with this project.

References

1. RM White, "Magnetic Storage Industry Continues to Grow and Grow," in APS News Online, March 2001 (cited 4 April 2003); available from Internet: www.aps.org/apsnews/0301/ 030111.html.

2. V Lauter-Pasuy, H Lauter, and B Toperverg, "Explaining Giant Magnetoresistance," in Exploring Matter with Neutrons: Highlights in Research at ILL (Grenoble, France: Institut Laue Langevin, 2002 [cited 4 April 2003]), 22–23; available from Internet: www-ucjf.troja.mff.cuni.cz/cejnar/prednasky/AR_pop.pdf.

3. "Giant Magnetoresistive (GMR) Heads," (cited 4 April 2003); available from Internet: www.pcguide.com/ref/hdd/op/heads/ techGMR-c.html.

4. B Dodrill and B Kelley, "Detect GMR Defects Early with In-Line Metrology," Data Storage 7, no. 2 (February 2000): 39–46.

5. GJ Collins, "Measuring and Characterizing Opaque Multilayer Metal Film Stacks on Product Wafers," MICRO 18, no. 6, (2000): 93–106.

6. J Xu et al., "Improving Process Control in Read/Write Head Manufacturing," Data Storage 8, no. 5 (May 2001): 25–32.

7. Rudolph Technologies, "PULSE Technology," Technical Report (Flanders, NJ: Rudolph Technologies, 1997).

Arun Natarajan, PhD, is a staff engineer in the wafer metrology (wafer process engineering) area of Seagate Technology (Bloomington, MN). He received a PhD in materials science and engineering from Johns Hopkins University in Baltimore. (Natarajan can be reached at 952/402-8743 or arun.natarajan@seagate.com.)

Darrell Louder, PhD, is senior manager of wafer production metrology at Seagate Technology. He received a PhD in analytical chemistry from Colorado State University in Fort Collins. (Louder can be reached at 952/402-8048 or darrel.r.louder@seagate.com.)

Matthew Dietz was a staff engineer at Seagate Technology and is now the engineering director of the suspension development group at Innovex (Maple Plain, MN). With more than 20 years of semiconductor and thin-film-head experience, he has worked at Texas Instruments, Digital Equipment, and Quantum. He received a BS in electrical engineering from the University of Minnesota in St. Paul. (Dietz can be reached at 763/479-5328 or mdietz@innovexinc.com.)

Peter Weyandt is a senior manager of development engineering at Seagate Technology. He has more than 22 years of experience in semiconductor and thin-film-head development and manufacturing. He received BS degrees in chemical engineering and metallurgical engineering from the University of Minnesota Institute of Technology in Minneapolis. (Weyandt can be reached at 952/402-7825 or peter.t.weyandt@seagate.com.)

Joseph Ivanecky, PhD, works in the wafer metrology area (wafer process engineering) at Seagate Technology. He received a PhD in chemistry from the University of Wisconsin in Madison. (Ivanecky can be reached at 952/402-5937 or joseph.ivanecky@seagate.com.)

Chris Morath, PhD, is director of product development at Rudolph Technologies (Flanders, NJ). He received a PhD in condensed-matter physics from Brown University in Providence, RI. (Morath can be reached at 973/448-4362 or cmorath@rudolphtech.com.)

Guray Tas, PhD, is manager of advanced systems development at Rudolph Technologies. He received a PhD in condensed-matter physics from Brown University in Providence, RI. (Tas can be reached at 973/448-4372 or gtas@rudolphtech.com.)

Jana Clerico is marketing communications manager at Rudolph Technologies. She received a BS in electrical engineering from Stevens Institute of Technology in Hoboken, NJ, and an MBA from Fairleigh Dickinson University in Teaneck, NJ. (Clerico can be reached at 973/448-4316 or jclerico@rudolphtech.com.)