As regulations and environmental factors converge to create a powerful driving force, the electric vehicle industry and its various segments of the value chain are creating a blossoming field of innovation. Today's electric vehicle (EV) battery packs are running at increasingly higher voltages, sometimes as high as 800 V. The voltage of the battery packs is increasing, but the voltage of the battery packs is also increasing.
The benefits of higher voltage include more horsepower, higher efficiency, longer range, and shorter charging times. Inside the vehicle, power electronics convert the high DC voltage into the different forms required by various systems. For example, traction motors require three-phase AC power. At the same time, vehicle chargers dynamically adjust the current and voltage.
Silicon is currently widely used in many areas of consumer and power electronics, but it has also become a bottleneck for their upgrading. Power electronics based on conventional silicon integrated circuits (ICs) cannot operate properly at high voltages, high temperatures and high switching frequencies. As a result, manufacturers must turn to alternative semiconductor materials to take full advantage of high-voltage battery packs for electric vehicles. The most promising alternative semiconductor is silicon carbide (SiC). This material possesses properties that make it ideal for EV power electronics, so that SiC is key to improving EV performance and range as EVs become more popular.
However, manufacturing SiC devices has its own unique challenges. The mechanical, chemical, electronic, and optical properties of SiC differ significantly from those of silicon in areas where mature processes and established protocols predominate. For example, SiC is one of the hardest materials known, comparable to diamond, making it difficult to scribe wafers using traditional mechanical methods such as sawing, and it is also a brittle material that breaks easily when sawed. In addition, SiC rapidly wears out saw blades, including those made from hard diamond, requiring frequent replacement of this expensive consumable. The sawing itself is a relatively slow process, and the heat generated tends to negatively affect material properties.
The combination of these issues creates a number of obstacles for electric vehicle manufacturers, as many established IC manufacturing processes are different, or even the opposite, of those used for SiC.
Single-crystal dicing, or wafer slicing, is a prime example; mechanical sawing is the primary method for single-crystal dicing of silicon wafers, but it is not universally effective for SiC, and while laser single-crystal dicing is promising, replacing the material means at least changing the process parameters. End users must also determine the optimal light source for SiC single-crystal dicing compared to traditional methods using silicon.
Image.
A close-up of a microscope shows that UV picosecond pulses in burst mode produce excellent edge quality without major chipping. Conventional mechanical sawing cannot achieve such results.

Quality ablation with picosecond lasers
SiC devices are fabricated in the same way as conventional silicon microelectronics: a large number of individual integrated circuits are made on a single wafer, which is then single-crystallized and cut into individual chips, which are then ready for packaging.
When cutting brittle SiC wafers, it is important to reduce or completely eliminate edge chipping from mechanical sawing. Single-crystal dicing should also minimize mechanical changes in the material. Priority should also be given to minimizing the kerf width to limit the size of the "space" (i.e., the empty area between adjacent circuits) to maximize the number of chips on each wafer.
Engineers must weigh these factors against dicing speed, throughput, and other cost-impacting determinants. The use of consumables, such as the use of coolants and cleaning fluids during the dicing process, also needs to be factored in.
Ultrashort pulse lasers in the picosecond and femtosecond pulse width ranges can be used for high-precision cutting and ablation of many different materials, including hard, transparent and/or brittle materials. Benefits of processing with ultrashort pulse widths include minimal overall heating of the material and a negligible heat affected zone (HAZ). These sources also provide improved edge quality and reduced debris generation compared to other types of lasers.
The infrared output of most picosecond lasers can be frequency-doubled to provide visible green or ultraviolet light, while the ultraviolet wavelength is typically used for demanding applications. Sources operating in this optical band can often achieve smaller focal spot sizes and increased depth of focus or Rayleigh range for a given spot size.
These characteristics make UV picosecond lasers the preferred choice for producing high aspect ratio features and thinner kerf widths because of the more precise depth control that can be achieved. In addition, the larger depth of focus makes these sources easier to apply to wide-field galvanometer scanning systems. The limited penetration of UV light further reduces the heat affected zone (HAZ).
Detailed configuration of the analyzed experiment
However, achieving higher yields with short pulse widths and short wavelengths is difficult in any environment. To ensure reproducible SiC single crystal dicing results, different system designs and parameters must be tested. mks/Spectra-Physics conducted a series of dicing experiments to evaluate the prospects for advantages of UV picosecond lasers, such as smaller focal spot sizes and larger focal depths. These trials also sought to achieve greater processing ease, and a smaller heat affected zone (HAZ). Finally, in addition to gauging the technical and economic feasibility of the process, the trials were designed to investigate how various burst settings might affect the results.
In the first round of tests, a 340 µm thick 4H-SiC wafer sample was processed using a 50 W, 355 nm picosecond laser. The laser has a maximum pulse energy greater than 60µJ and delivers an average power of 50 W at repetition frequencies from 750 kHz to 1.25 MHz, with a maximum operating frequency of 10 MHz. Tests were conducted at repetition frequencies from 200 to 400 kHz to ensure that all pulse output formats maintained similar pulse energy and average power levels, allowing direct comparison of results.
The picosecond laser is used with a dual-axis galvanometer scanner and a 330 mm focal length f-theta objective. The focal spot size on the working plane is approximately 30 µm (1/e2 diameter). The scanner operates at speeds ranging from 2 to 4 m/s, with multiple passes per scribe, and net cutting speeds ranging from 12.5 to 25 mm/s. The laser used in these tests supported a wide range of applications.
The lasers used in these tests support pulse trains: the laser emits a series of closely spaced sub-pulse trains, followed by the next pulse train after a time interval. It has been well documented that pulse trains can increase ablation rates and reduce surface roughness in many material processing situations.

In addition, the laser used in the test supports programmable bursts. This means that the number of pulses in the burst, as well as the amplitude and time interval of each pulse in the burst, is controllable. In addition, the timing time jitter of the pulse train is very low, enabling direct placement and positioning on the work surface with high accuracy, even at very fast scanning speeds. These flexible pulsing capabilities allow us to explore a wide range of process space during testing.
Analysis of results
Figure 2 below shows the scribe depth values as a function of average laser power for various pulse string configurations ranging from a single pulse to 12 pulses. In each test, a total of 80 strokes were made at the same location on the material. The position of each pulse train on the working surface (total pulse overlap) was strictly controlled. In this case, the effective spatial overlap of the pulses was approximately 84%.

Figure 2. shows the depth of scribing as a function of power through four passes at 25 mm/s for a single pulse (a, top panel) and various pulse string configurations (b-d, middle and bottom panels). The data show how the pulse string improves ablation rates.
These results show that the use of the pulse string greatly increased the ablation rate. This result was expected and is consistent with results obtained using picosecond laser pulse string processing in other materials. Again, the ablation threshold decreases (essentially logarithmically) with the number of pulses contained in each pulse train. This suggests that many materials typically "accumulate" under multi-pulse irradiation.
Both 3D and 2D surface topography tools are used to accurately measure scribe depth and edge quality. Images obtained with a scanning white light interferometer show further details of the scribing (Figure 3). Because the surface is smooth and free of debris, the picosecond UV laser also achieves another desired result: a high quality cut.

Figure 3. The scribing results obtained by scanning the white light interferometer confirm that the picosecond UV laser is capable of making clean, chip-free cuts.
A further qualitative assessment of the scribing can be seen in Figure 4 below. A single image shows a series of 25 µm deep grooves that were generated sequentially with 1, 4, 8 and 12 pulse trains. The average power was adjusted as necessary to obtain the best results in each case. The four images in the top row are focused on the top surface of the wafer. The four images in the bottom row are focused on the bottom surface of the scribe. Figures 4e-h show a clear comparison and progression of cut quality as a function of the number of pulses in each pulse train.

Figure 4. Close-up images of the top (bottom, a-d) and bottom (e-h) of the 25-µm deep notch. As the number of pulses in the burst increases, the different values of notches show a steady improvement in the quality of the cut.
The discoloration around the scribe line indicates a change in the surface or substrate material, which disappears as the number of pulses increases. The higher the number of pulses, the faster the feed rate and the better the results. This suggests that the process can be used to ensure adequate throughput and good quality at the same time.
Figure 5 below shows a series of high magnification views of the scribed bottom surfaces, all scribed under the same laser operating conditions at an average power of 16 W and a net processing speed of 25 mm/s. The results of this process are shown in Figure 5 below. The scribing depths for each condition range from 8 to 25 µm at different pulse values. This higher resolution view highlights the improvement in smoothness as the number of pulses increases. Adjusting the pulse output increases the scribing depth by a factor of three while keeping the average power and overall processing speed constant.

Figure 5. Processing with a picosecond UV laser results in excellent edge/surface quality, highlighting the advantages of higher pulse count strings (a-d)
Perfecting the technology
In advancing from theory to practice, the potential for applying UV picosecond lasers to scribe SiC wafers is demonstrated by the ability to utilize pulse string output to improve processing quality and increase processing speed. Further exploration is needed to measure and evaluate the parameters and results of complete dicing of 340 µm wafers.
In the meantime, we are investigating the use of mechanical saws, traditionally used for scribing silicon wafers, for SiC. published results show that this method still suffers from limited feed rates and generates large amounts of debris, for example in chips larger than 10 µm.
Nonetheless, mechanical sawing is still a commonly used method in the semiconductor industry, and any alternative technology would need to demonstrate significant advantages in terms of throughput, yield, and operating costs to gain industry acceptance. Although the UV picosecond results obtained need to be further improved in terms of complete dicing, further sustained improvements are possible as an alternative technology.





