BY MARTIN GEHERAN
As electronic devices get smaller, components that go into them get smaller as well. A few years ago, many electronic packages were made with ceramic and in some cases were covered with gold lids. Recently there has been a trend toward plastic packaging, and in some current applications, the devices have no covering at all. Many products have moved to packaging that leaves the backside of the silicon chip exposed, with more applications moving in this direction. There are three major types of exposed silicon packages: chip scale packaging (CSP), direct chip attach (DCA) and flip chip technology. These packaging innovations create some interesting challenges for marking solutions.
Today, there are two major technologies used for marking of electronic devices – pad printing and direct laser marking. Many exposed silicon packages are used in memory applications, and direct laser marking has been widely accepted in this part of the industry as the best solution for these devices. Because memory devices are low-value products, manufacturers are very focused on cost, so the lower cost of ownership (COO) of laser marking compared to traditional ink marking is appealing. The speed, lack of consumables and compatibility with silicon package handling contribute to the lower COO for laser marking. In addition, the non-contact nature of direct laser marking is attractive to manufacturers because there are no static issues.
Laser Marking Developments
There are two types of laser markers that are used for marking on silicon devices. The most common type has been a near infrared (NIR) laser that produces laser energy in the 1,064 nm range of the spectral field. The most common type of NIR laser used in marking is the Nd:YAG (neodymium yttrium aluminum garnet) laser. Recently, second harmonic generation or green lasers have become available for this application. Green lasers offer some significant advantages over NIR lasers in silicon marking applications.
Figure 1. Diode pumped green laser marking system.
Second harmonic generation, sometimes called “frequency doubling,” involves a process where light from a longer wavelength laser enters a non-linear crystal and is converted to shorter wavelength (i.e., higher frequency) laser light. In the case of a green laser (Figure 1), YAG infrared (1,064 nm) laser radiation is converted to 532 nm laser light. Energy with a wavelength of 532 nm is optimal for marking on silicon substrates for several reasons.
Because spot size scales linearly with wavelength, for the same scanning system and mode structure, the spot size of a 532 nm laser is one-half the diameter of a 1,064 nm NIR laser. This is important because most silicon-based packages are very small and the marks required are also very small. As a result, the green laser is capable of creating characters that are one-half the size of those created by an NIR laser. The second harmonic generation laser can make a character as small as 0.075 mm in height.
In many applications, spot sizes have been measured at one third of the NIR spot sizes used to generate the green laser energy; this results in the capability of producing characters that are three times smaller than the smallest produced by an NIR laser. This becomes important in applications where larger field sizes are needed. For a given character height required, the green laser is capable of marking a substantially larger field area than an NIR laser.
There are two types of marks that are typical with silicon surfaces. The first is an etched mark. In this case, the laser is used to ablate the silicon and leaves an etch typically 12 to 25 microns deep in the surface. This is often referred to as a “hard mark.” The second is an annealed mark. In this case, the laser is set to a lower power and/or taken out of focus to a point where the energy does not ablate the surface but rather rearranges the molecules on the surface of the part. This molecular metamorphosis manifests itself as a change in reflectance that creates a contrast and results in a marked surface. In this case, molecular rearrangement is controlled to one to two microns of depth in the silicon.
Silicon is a very sensitive surface for marking. Minor changes in laser power applied to a silicon surface can cause differences in mark appearance and consistency. Applying such energy so close to active components creates a concern with manufacturers about device damage. As a result, specifications for etch depth or annealing depth have very tight tolerances.
Power consistency is gained through power stability and is key in marking silicon and keeping within these tight industry tolerances. Diode pumped lasers have superior power stability as compared to flash lamp pumped lasers. There are both lamp and diode pumped green and NIR lasers available in the industry.
Figure 2. Laser beam modes and resulting marks in each axis.
It has been frequently observed that power stability of a 532 nm laser beam is much tighter than that of a 1,064 nm NIR beam. When power fluctuation is measured for NIR compared to green, it is found that NIR has significantly higher variation than green does as a percentage of its total power. This has to be quantified as a percentage of total power because, in most cases, NIR lasers have a much higher energy output than green lasers. Although we do not fully understand why there is such a difference between green and NIR, this power fluctuation in NIR or YAG lasers causes inconsistencies in etch depth or mark contrast depending upon whether you are making an annealed or ablated mark. Power stability guarantees etch and anneal depth consistency in the direction of the beam path.
Modes of Operation
YAG lasers operate in multiple modes. There are four common modes that the beam takes on and randomly switches between. These are commonly referred to in the laser industry as “donut,” “bull's-eye,” “split-beam” and “TEM00” (Figure 2). As a YAG laser changes between each of these modes, the etch depth or anneal contrast will change. This modal variability is not desired especially on silicon because of the differences it causes in mark appearance and consistency. Because there is no shadow marking or double marking, TEM00 is the cleanest beam and is required for the very small legends that are marked on silicon substrates. TEM00 mode guarantees etch and anneal depth consistency perpendicular to the direction of the laser beam. Green lasers operate in TEM00 mode as part of the design. Most companies that make NIR lasers modify their markers to attain TEM00 mode for marking on silicon. YAG lasers typically control this through the use of one-millimeter apertures. With this process, most NIR lasers lose more than 80 percent of their power. This means that to generate 10 watts of TEM00 NIR laser energy, it's necessary to start with a 60-watt YAG laser. Because most silicon substrates will mark with three watts of green laser energy, there are significantly higher operating costs associated with red laser marking on silicon.
Potential Device Damage
One of the main concerns during marking for manufacturers of silicon-packaged components is device damage. With silicon-based packaging, there is no plastic, metal or ceramic to protect the silicon chip from the outside. As a result, device manufacturers must be much more cognizant of forces that could damage the device.
One specific area of concern to many device manufacturers is the use of laser marking equipment on direct silicon substrates. One issue is micro-fracturing of the silicon chip itself. Generally though, this type of failure is fatal at the time of mark, so it is not a reliability issue down the road. The other area of concern is in absorption of the laser energy within the silicon. Some manufacturers have reported device damage or failure with the use of NIR (1,064 nm) laser light on silicon. They attribute the failures to the poor absorption of the NIR laser light by silicon. Much of the laser light passes through the silicon and causes hot spots on the underlying active components of the device. In some cases, these hot spots can cause immediate failure. In many other cases, the damage is not always fatal at test or burn-in and the device will fail early in its use in the field, which is a worse problem than a yield loss during manufacturing.
Figure 3. Laser beam irradiance vs. laser penetration depth.
Green laser energy (532 nm) is readily absorbed by the silicon very close to its surface, making it a better solution for marking silicon devices. As a result of this, much less energy is required with green compared to NIR, allowing the use of a much lower powered laser. Because green energy is absorbed by the silicon in the very top layer of material, the creation of hot spots on the active components does not occur.
Lambert's Law tells us about light absorption properties of materials and how key physical properties play a role in the amount of light absorbed by that material.
I = Ioe-kx
I = final irradiance of light (watts/cm2)
Io = original irradiance of light (watts/cm2)
k = absorption coefficient of material (cm-1)
x = material thickness (cm)
It considers the specific absorptive properties of a material as a function of wavelength and thickness.
The absorption coefficient for silicon at 1,064 nm is approximately 10 cm-1 while for 532 nm it is approximately 9,500 cm-1. When these numbers and varying thicknesses of silicon are applied to Lambert's Law, the graph in Figure 3 is generated. The graph shows that for a half-millimeter thick silicon chip, only 40 percent of the NIR laser light that enters the chip is absorbed by the silicon. The remaining 60 percent passes through the silicon and can be absorbed by components on the other side of the silicon. As the graph also shows, nearly all of the green laser energy is absorbed in the top 2 percent of the silicon surface. Also, as electronic devices get thinner and thinner, the issue of energy absorption during laser marking will become a more significant issue.
Based on these issues – spot size, power stability and absorption of green laser light – a green laser is an appealing marking solution for silicon packages. The next question, of course, is: What features are important to look at in a green laser?
Not All Green Lasers Are the Same
When selecting a green laser for a manufacturing environment, cost of ownership and reliability are two key attributes that are measured. In addition to capital cost of equipment, there are other factors of cost that should be considered.
Cost of energy: Flash lamps and chillers draw massive amounts of energy. Many companies do not consider this to be part of the overhead of the plant and it therefore does not always get included in the product cost, but it does affect the overall cost to the manufacturer.
Cost of consumable parts: For laser markers, this would be flash lamps or laser diodes. Generally, the more powerful a laser is to start with, the higher the replacement costs of these consumable parts. Flash lamps must be replaced quite often (usually every month or two), and can cost as much as $35,000 in five years.
Reliability: Reliability should be considered because not all lasers and their consumable parts have the same life expectancy. Short-lived parts will add considerably to the COO, as will short intervals between failures of the equipment and the need to send equipment back to the laser manufacturer for repair.
Green crystal stability: There are many factors that enter into the equation of crystal life and stability. These include the crystal material, the cleanliness of the crystal enclosure and the environmental stability of the crystal.
There are two materials used to make green crystals. The most common is potassium niobate (KTP). The other is lithium triborate (LBO). LBO is nearly perfect for producing second harmonic generation laser energy. KTP has been found not to be as reliable a material as LBO for green crystal manufacture, because it has a trait known as “gray tracking.” This is caused by absorption of NIR energy in the process of creating green light and results in power loss over time. Particulate matter, which may collect on the crystal, and instability of the crystal's operating environment can cause premature failure and require replacement of the green crystal.
Other factors to consider when selecting a green laser include safety, serviceability, ease of use, ease of integration and percent of NIR energy converted to green laser energy.
Green lasers offer unique advantages over NIR laser markers for silicon applications. Power stability, spot size, low energy consumption and no possible device damage are some of the advantages of green laser markers. AP
Martin J. Geheran, senior product manager of industrial marking and coding, can be contacted at MARKEM, 150 Congress Street, P.O. Box 2100, Keene, NH 03431-7100; 603-357-4255; Fax: 603-352-0479; E-mail: [email protected].