Issue



A close look at laser marking of silicon wafers


07/01/1997







A close look at laser marking of silicon wafers

Jim Scaroni, Lumonics Corp., Oxnard Operations, Oxnard, California Terry McKee, Lumonics Inc., Kanata, Ontario, Canada

Since the introduction of the first laser wafer marking system (WaferMark 345) based on the neodymium-doped yttrium aluminum garnet (Nd:YAG) laser, the technology of wafer processing has changed dramatically, and so have the requirements for applying the mark used to record batch numbers, process parameters, etc. In the early days of large feature size lithography, fabs were satisfied with the 5-20-?m deep "hard" laser mark consisting of a series of dots for each character. Back then, they were primarily concerned with human readability and durability of the marks, rather than the considerable splatter and debris that the marks generated.

As the feature size of device designs dropped into the submicron levels and the number of process steps increased, industry specifications for mark depth, splatter, and debris became more stringent. Figure 1 shows the improving quality of dots used to keep pace with the demands of silicon suppliers and device manufacturers.

The dot in Fig. 1c is made with the SuperSoftMark technology. SuperSoftMark makes a debris-free mark with a dot diameter of 70 ?10 ?m and dot depth of 2.6 ?0.4 ?m. It is used to apply alphanumeric characters in OCR (optical character readable) format, Bar Code 412, or 2D symbology (such as ID Matrix`s data matrix) in specified areas (Fig. 2) to the right or left of the fiducial axis on flatted or notched wafers, as defined in SEMI M12-92, T1-93, and T2-93. Figure 3 shows how the dots are used to form the alphanumerics, bar codes, and two dimensional codes covred by the SEMI specifications.

Click here to enlarge image

Figure 1. SEM views of three different dots on silicon wafers made with laser marking systems; a) the "hard" dot, ca. 1979; b) dots made using SoftMark, ca. early 1980s; c) dot made with today`s SuperSoftMark.

Click here to enlarge image

Figure 2. WaferMark codes on silicon wafers.

The laser marking phenomenon

The high intensity, focused laser light is absorbed very close to the surface. The extent of absorption is determined by the wavelength of the laser used, pulse length, number of pulses fired, passes made, etc. When marking silicon, gallium arsenide, or indium phosphide, the focused laser pulse heats up the material until it melts. A crater forms with a circular ridge or bulge around its heat-affected zone.

Use of the proper laser marking process for the material being marked will control or essentially eliminate debris in the area immediately surrounding the crater. Debris results when the liquefied material is ejected as droplets and propelled by the expansion of the vapor formed. Some of this liquid may adhere to the material and solidify (and is generally called recast). The vapor fraction condenses, typically as a submicron-sized powder. The recast and powder can be damaging if they are transported to the patterned area on the wafer or interfere with subsequent polishing operations.

The ideal time to mark the wafer to achieve a debris-free impression is at its base silicon state, before the layering process begins. However, today`s fabs prefer marking after one or more wafer processing steps have been completed (e.g., after epi deposition, coating with nitride). Once the wafer is marked, it is usually dedicated to a specific chip design and/or customer. By waiting to mark the lot number on the wafer, the fab can be more flexible with its production lines, thus keeping work-in-progress inventory down. Unfortunately, it is more difficult to obtain a debris-free mark due to the absorption differentials of the various deposited layers. Each deposited material absorbs the laser light differently. The result, trapped gas and outgassing, can affect the quality of the mark.

Interaction of YAG laser light with silicon

YAG laser marking on silicon involves some interesting time-dependent physics models. The absorption depth of silicon at 1.064 ?m - the wavelength of the YAG laser - is relatively large (about 250 ?m [1]). Initially, it seems likely that the YAG laser will make a mark of comparable depth. However, since 1.064 ?m is very close to the band gap edge of silicon, the absorption spectrum in that region is highly dependent on temperature [2]. For example, the absorption depth decreases by a factor of 7, to 36 ?m, at 300?C (Fig. 4). Silicon melts at 1410?C. Depositing laser energy in short time scales to raise the surface temperature of silicon to the melting temperature controls melting within a shallow absorption depth of about 2 ?m. This allows the melt/flow cycle and subsequent cooling cycle to occur before any debris is ejected from the interaction region. A melted silicon mark of accurately controlled depth and size with essentially no debris formation results. The physical model of this interaction must include the temperature dependence of silicon absorption at 1.064 ?m. The variation in thermal conductivity with temperature, the dependence of reflectivity at 1.064 ?m on temperature and phase transition to liquid, and the pulse shape of the YAG laser [2] are also considered.

Much of the credit for understanding the silicon marking process comes from the work of F. Kuhn-Kuhnenfeld, J. Kramler, and H.A. Gerber of Wacker-Chemitronic, who patented their discoveries in 1985 (US patent 4,522,656). They pioneered a way to make the fine adjustments to the laser parameters that were necessary to accommodate the small variations in absorption and conductivity due to differing silicon purity and laser marking levels.

Product application of laser marking principles

In the WaferMark SuperClean laser marking system (Fig. 1), each series of uniform dots (of (70 ?10)-?m diameter and (2.6 ?0.4)-?m depth, Fig. 5) on the wafer is made by applying the carefully controlled pulsed laser melting principles described previously. Each dot is formed using a single laser pulse from the system`s Q switched continuous wave (CW) lamp-pumped YAG laser operating at a repetition rate of 1000 pps to yield marking speeds of over 250 wafers/hr.

Click here to enlarge image

Figure 3. How laser-made dots are used in alphanumeric, bar codes, and 2D symbols; a) BC 412 data character construction; and b) border rows and columns for 2D dot matrix code.

Click here to enlarge image

Figure 4. Measured absorption coefficients near and above the fundamental absorption edge for pure Ge, Si, and GaAs [2].

Recently, Lumonics determined that continuous lamp pumping does not provide sufficient flexibility. Moreover, pulse energy during the Q-switching process is dependent on repetition rate, gain, and a host of other parameters. Despite a variety of techniques used to address these issues, it has become difficult to extend the technology further to meet present demands.

Today`s technology

The CW lamp-pumped Q-switched YAG laser technology has now been superseded by the newly developed semiconductor diode-pumped yttrium lithium fluoride (Nd:YLF) laser, commonly referred to as DPL, which also delivers SuperSoftMark. High power diodes [4], cousins to the lower power versions common in fiber optic telecommunications, incorporate the reliability and compactness of semiconductor electronics with dramatically improved optical stability (versus that of lamps).

Using a diode laser to replace the lamp for the pumping function has proved to be an important step in providing better control of the laser marking parameters, such as pulse energy, pulse length, and peak power. Replacing the Nd:YAG crystal with the Nd:YLF laser medium (whose absorption spectrum is closely matched to the laser diode emission wavelength for efficient pumping) results in a more stable laser source that is less sensitive to thermal variables. YLF, being a fluoride material, has substantially better thermal conduction properties than YAG. The laser can thus have much more stable performance (better pulse-to-pulse and pulse length stability) as pumping powers are increased or decreased.

The future in laser marking

Work is progressing on making this technology available for the new 300-mm wafers. Further advances will allow increased throughput and retooling DPL technology to the 300-mm platform size. There is increased pressure from fabs demanding smaller dot and character size to free up wafer space for more die. Some fabs are also pushing for greater dot-to-dot depth and diameter consistency. These "wish list" items are driven primarily by increasing fab sophistication and the 6-s mentality found in the industry.

Click here to enlarge image

Figure 5. WYKO surface profile measurement showing depth of SuperSoftMark on silicon wafer.

As a result of high production rates, more wafer laser marking systems will be integrated with data-matrix readers and vision-mark systems for catching the marker`s own mark defects. The software available now allows the vision system to determine the characteristics when the mark is applied, eliminating the need to use a "golden" marked wafer to teach the system what mark qualities to look for.

Current investigations include interfacing the laser marking system and some form of sealed SMIF pod containing the wafer cassette to support an industry move toward portable Class I minienvironments - the so-called "islands of cleanliness."

With the unrelenting march toward finer and finer linewidths, even more cleanliness will be necessary in the next century. At present, the state of the art in particle density/wafer resulting from laser marking is 0.02 particles/cm2 at a particle size of 0.17 ?m. Is it possible to hold that particle count but cut the particle size by half - as is being encouraged by some fabs - for the next generation of laser markers? It will depend on the progress made in the areas of marking processes, handling schemes, and fab designs.

Acknowledgment

WaferMark, SoftMark, and SuperSoftMark are trademarks of Lumonics Inc.

References

1. C.J. Nonhof, Material Processing with Nd-Laser, Electrochemical Publications, p. 244, 1988.

2. S.M. Sze, Physics of Semiconductor Devices, Wiley and Sons, NY, pp. 53-55, 1969.

3. T.J. McKee, Physics in Canada,, pp. 107-114, March/April 1995.

4. J. Boneberg, O. Yavas, B. Mierswa, P. Leiderer, Proceedings, SPIF (Society of Photonics & Instrumentation Engineers), 1991.

5. J.J. Ewing, Laser Focus World, pp. 105-110, Nov. 1993.

JIM SCARONI is general manager of Lumonics Corp., Oxnard Operations. He has more than 16 years of experience in marking semiconductor wafers. He was the cofounder of Laser Identification Systems, which introduced the WaferMark into the semiconductor industry in 1980. Lumonics/Oxnard Corp. Operations, 130 Lombard St., P.O. Box 9010, Oxnard, CA 93031-9010; ph 805/485-5559, fax 805/485-3310.

TERRY MCKEE received his PhD degree in physics from Yale University. He is senior scientist and patent officer with Lumonics Inc. He manages an applications lab devoted to laser marking processes for advanced materials including polymers, ceramics, semiconductors, glasses, and composites.