Laser micromachining of semiconductors for photonics applications Marc Nantel*, Yuri Yashkir, Seong-Kuk Lee, Chas Mugford, Bernard Hockley

Photonics Research Ontario

ABSTRACT

For decades, precisely machining silicon has been critical for the success of the semiconductor industry. This has traditionally been done through wet chemical etching, but in the pursuit of integrating photonics devices on a single chip, other techniques are worth exploring. This quest opens up interest in finding a non-wet, non-contact, arbitrary-shape milling technique for silicon. In this paper, we present our latest work in the laser micromachining of silicon. A kilohertz-repetition-rate diode-pumped Nd:YLF laser (in infrared, green or ultraviolet modes) is focused on the surface of silicon wafers in a chlorine atmosphere for an enhanced magnitude and control of the etching rate. In the chlorine atmosphere, much less debris is deposited on the surface around the cut, sub-damage threshold machining is achieved for a better control of the etching depth, and etching rates ranging from 20-300,000 µm3/s have been measured. In particular, the use of an infrared laser beam is singled out, along with the advantages that it holds. Results of simulations highlight the particular characteristics of the various wavelength chosen for the machining.

Keywords: Laser micromachining, silicon, chlorine, ultraviolet, green, infrared, photonics, simulations, finite difference

1. INTRODUCTION

Machining silicon to very small details and high accuracy has enabled the integration of electronics into microelectronics, and has made possible the current tracking of Moore’s law, wherein computer processing power and memory doubles every 18 months or so. Microelectronics has arguably been the most important technology for enabling economic development and standard-of-life improvement in the second half of the 20th century, with its presence transforming all sectors from automotive to medical, from financial to entertainment. There is another technology with the potential to have the same importance and presence in the first half of the 21st century and beyond: photonics. The main hurdle that photonics has to overcome before becoming as ubiquitous as microelectronics lies in miniaturisation: photonics has to acquire the prefix “micro”. Presently, photonics devices are typically assembled by hand, under microscopes, one device at a time. This modus operandi lead to slow production, low yields, bulky devices and a critical reliance on difficult-to-find qualified personnel. This is especially true for the telecommunications/fibre-optics industry. Photonics needs to take the turn towards integrated manufacturing and self-assembly. Already, progress in photonics band gap (PBG) materials1 open the door to more integrated devices, from better notch filters to optical computers. But the devil is in the manufacturing and packaging details, and much progress has to happen in these areas for the bright promises of integrated microphotonics to see the light of day. Micromachining silicon may again be the answer to open up the manufacturing bottle-neck, this time for photonics. The techniques developed for microelectronics can play a significant role in helping this integration, and silicon is a platform of choice to bring together many photonics elements. Wet chemical etching of silicon is well established and its limitations well understood. It allows for bulk machining with high reproduceability and high yields, with minimum human intervention. On the other hand, it is a complex process, with several thin-film operations, projection or contact masks, excimer lasers and wet acids, and is limited in the shapes it can produce by the anisotropic response of the silicon to the etching solution. Since integrated photonics devices are likely to join optical fibers with waveguides, PBG materials, gratings, mirrors and detectors on a micrometer scale, a more versatile technique for their production is needed, at least in the prototyping stage of new devices. Faster and more flexible methods of micromachining silicon could make an important contribution, especially in such an area as photonics where time-to-market is a crucial concern. In this paper, we present improvements on such a method: the chlorine-assisted laser-machining technique. We particularly focus on the comparison of ultraviolet and green light laser machining on one hand, and infrared light on the other, with experimental data and finite-difference simulations illustrating the differences.

2. CHLORINE-ASSISTED LASER MICROMACHINING OF SILICON

The technique of chlorine-assisted laser micromachining was developed for silicon and other difficult-to-machine materials in the 1980s2,3,4. Figure 1 shows a schematic representation of the process. An ultraviolet (UV) or green laser beam is focused onto the silicon wafer through an atmosphere of molecular chlorine. The laser focus heats the silicon to near-melting or melting temperatures and the chlorine gas reacts with the heated silicon to form SiCl2 and SiCl4 gas, thereby removing silicon from the wafer. Only the silicon heated by the laser beam reacts with the chlorine, and a direct-write method of etching silicon results. The molecular chlorine gas can be replaced with HCl or other halogen gases. This technique, while requiring chlorine-handling systems, gives much superior results than laser ablation in air or in vacuum, as will be shown later. For this reason, in this paper the term “ablation” will be reserved to indicate the usual melting/vaporising/plasma-generating laser-matter interaction, while the term “etching” will refer to chlorine etching of laser-heated (but not ablated) silicon. Typically, the lasers used for chlorine-assisted laser etching are excimers (pulsed) or ion-argon (CW, fundamental or doubled). When UV light is used, much of it is absorbed by the chlorine gas, which disassociates into atomic chlorine. Nonetheless, UV has the advantage of being absorbed in the silicon in a very short layer (about 0.01 µm for 1/e drop in intensity at 351-nm wavelength) for a good control of the etching depth. Green light does not get substantially absorbed in the chlorine layer, and also deposits its energy in a thin layer of silicon (0.94 µm for 1/e drop at 534-nm wavelength). Trenches as narrow as 60 nm in width have been machined in silicon using this technique5.

Fig. 1: Principle of chlorine-assisted laser micromachining of silicon. The laser beam (here shown with a 351-nm wavelength) is focused through a transparent window and onto the silicon wafer bathed in chlorine gas. An XYZ positioning stage moves the part around for the laser focus to write the desired features in the silicon. The laser-heated silicon under the focal spot interacts with the gaseous chlorine to form SiCl2 and SiCl4 products, also in the gas phase, thereby resulting in an etching of the silicon.

2.1 Experimental set-up In our experiments, the traditional excimer or ion-argon laser has been replaced by a state-of-the-art diode-pumped solid-state (DPSS) Nd:YLF laser. This particular model, a Sigma-400 from GSI-Lumonics, has a pulse duration of 60-150 nanoseconds and a pulse-repetition-frequency of 5-35 kHz. This laser can operate in the green (526 nm) and the UV (351 nm) through frequency-doubling and -tripling, respectively, but we also used it at 1,053 nm, in the infrared (IR), something the excimer and ion-argon lasers cannot do. The parameters of the laser relevant to this paper are shown in Table 1. As will be shown later, access to an IR wavelength potentially increases the etching rates dramatically. Other advantages of diode-pumped solid-state lasers include a) the absence of dangerous gases (fluorine for excimer), b) the absence of large quantities of cooling water (excimer, argon-ion) because DPSS lasers are typically air-cooled or have self-contained water chillers, c) a much lower electricity consumption, with some DPSS lasers plugging straight into the normal 110-V socket, and d) durable pumping diodes needing only to be changed every 10,000 hours (for the same price or cheaper than an argon-ion plasma tube lasting typically 2,500 hours). All these advantages will make a difference in the cost of using this technique for prototyping or production.

The laser beam is expanded and relayed to the target, which is enclosed in a vacuum chamber. The laser focusing lens (f = 65 or 150 mm) is situated outside the chamber. Figure 2 shows a schematic drawing of our vacuum and chlorine-handling system (Fig. 2a) and a photo of the vacuum chamber on the XY micropositioning table (Fig. 2b). These Aerotech stages have 30-cm by 30-cm XY travel (horizontal plane) with 1-µm accuracy and repeatability, and 20-cm Z travel (vertical) with 5-µm accuracy and repeatability. Laser firing and sample positioning are computer-controlled, and a white-light source and CCD camera (both collinear with the laser beam path) provide vision.

IR Green UV Power (W) 1→10 0.1→2 0.005→1

Wavelength, λ (nm) 1,053 526 351 Pulse Width (ns) 150→80 120→60 120→60

Repetition Rate (kHz) 5→8 5→8 5→8 Shape Gaussian Gaussian Gaussian

Spot Size (µm) 10 8 5 Duty Cycle 100 100 100

Intensity (GW/cm2) (At 5kHz) 3→30 0.6→12.5 0.08→16 Table 1: Laser characteristics used in the experiments presented in this paper. 2.2 Ultraviolet light Ultraviolet lasers are used to mark silicon and other traditionally difficult-to-machine materials like glass and ceramics. Rough features are welcomed in such applications because they enhance the visibility of the markings. However, when precision machining is the desired result, much cleaner results are needed, as shown in Figure 3. In Fig. 3a are two trenches produced in vacuum, with the laser power at 16 mW (or an intensity of 108 W/cm2 at the focus). The ablation process is evident from the debris present on both sides of the trenches. As a comparison, Fig. 3b shows two trenches produced in a 100-Torr atmosphere of chlorine gas, with the same laser parameters. These trenches are much cleaner, with a full-width-at-half-depth of 5 µm, the size of the laser focus spot on the surface of the silicon. Cleaner features are not the only advantage of machining in a chlorine atmosphere. Figure 4 shows a graph of the machined features’ depth as a function of laser power. In the case where chlorine is used, the depth obtained and its increase as a function of laser power are both larger than in the case of vacuum. Also, the threshold for etching is much lower, at less than 4 mW (2.5 x 107 W/cm2 ), than it is for the ablation, at about 8 mW (5 x 107 W/cm2). With chlorine, less laser power is needed to start the etching and its efficiency increases faster with incremental power.

a) b) Fig. 2: Experimental set-up for chlorine-enhanced laser micromachining of silicon. a) Schematics of the vacuum and gas systems; b) photo of the chamber and accessories on the XY positioning stage.

a) b) Fig. 3: The difference between laser micromachining silicon a) in vacuum and b) in a 100-Torr atmosphere of chlorine gas. Laser parameters: 16-mW of λ = 351-nm light in 2-µJ pulses, 100-ns pulse duration, 8-kHz repetition rate, 5-µm focal spot diameter for an on-target intensity of 108 W/cm2. Photographs taken with an optical microscope.

Fig. 4: Depth of features obtained as a function of the laser power, for the case without chlorine (stars) and with chlorine (other symbols), for various chlorine pressures. With chlorine, the etching starts at a lower threshold (3-4 mW) and its efficiency increases faster with incremental power, as compared with the case without chlorine.

2 4 6 8 10 12 14 16 18 200.0

0.5

1.0

C

BAv=0.2mm/min

0 torr 7 torr 15 torr 20 torr 31 torr 40 torr 50 torr 102 torr 200 torr

Dep

th, µ

m

Laser power, mW

a) b) Fig. 5: The depth of the feature can be controlled through the number of passes the laser beam makes on the sample. a) cross-section of trenches created with 1, 2 and 4 passes; b) close-up on the trench created in one pass. Laser parameters: 16-mW of λ = 351-nm light in 2-µJ pulses, 100-ns pulse duration, 8-kHz repetition rate, 5-µm focal spot diameter for an on-target intensity of 108 W/cm2. Figure 5 shows depth profiles of trenches etched, with chlorine, through 1, 2 and 4 passes of the laser (Fig. 5a), with a close-up on the shallowest trench (Fig. 5b). The measurements were made using a WYKO white-light interferometric profilometer. The depth of the trenches can therefore be controlled quite straightforwardly through the number of laser passes to produce features of the type needed for fibre-optics alignment or PBG stacking (1). This type of features is also of interest to the pharmaceutical and molecular biology communities for their “labs-on-a-chip”, where the minute amounts of reagents needed for their experiments are channelled through such trenches and mixed in shallow “pools”6. Figure 6 shows such a “pool”: a 40-µm x 40-µm square 1.5-µm deep, micromachined in a 10-Torr atmosphere of chlorine with four laser passes. The pool was etched with this low chlorine pressure to increase the accuracy of the cut and the flatness of the bottom, which features +/- 0.1-µm roughness. The walls are fairly vertical, making a ~30-degree angle with the vertical. Whiles this compares well with state-of-the-art in the technique7, it was our first attempt at such a cavity, and much improvement can be achieved merely by tuning the laser, chlorine gas and feedrate parameters.

a) b) Fig. 6: Example of shape milling in silicon. a) Top view of a 40-µm by 40-µm by 1.5-µm deep excavation; b) profile at the centre of the trench. The surface roughness at the bottom of the excavation is +/-0.1 µm. Laser parameters: 8-mW of λ = 351-nm light in 1-µJ pulses, 100-ns pulse duration, 8-kHz repetition rate, 5-µm focal spot diameter for an on-target intensity of 5x107 W/cm2.

235 240 245 250 255 260 265

-1.0

-0.5

0.0

0.5

p=100 TorrP=16 mWv=0.2mm/min

h,µm

x, µm

230 240 250 260 270 280 290 300 310 320 330 340 350 360 370

-3.5

-3.0

-2.5

-2.0

-1.5

-1.0

-0.5

0.0

0.5

4

2

1

h,µm

x, µm

p=100 TorrP=16 mWv=0.2mm/min

The etching rates obtained with the 351-nm pulses can be as low as 20 µm3/s (0.0025 µm3/pulse) for a high control of the etching rate and can be increased to 10,000 µm3/s (1.67 µm3/pulse). All the above examples were machined at 8-kHz repetition rate. The choice of UV light for chlorine-assisted silicon laser micromachining is clearly for finesse work. For higher etching rates, however, it would be necessary to increase the chlorine pressure and the enhanced UV absorption by the chlorine would ensure diminishing returns. Figure 7 shows how critical the absorption can be, to the point of dropping the laser intensity reaching the target below the machining threshold shown in Fig. 4 above (about 3-4 mW). Another way to improve the etching with UV would be to use a continuous-wave beam. 2.3 Green light What is typically done to avoid the loss of efficiency due to the absorption of UV light by the chlorine is to use green light from an argon-ion laser (see <www.revise.com> for examples). In the case of a pulsed SSDP laser, there is more power available in green than in UV from the frequency conversion process, the 1/e depth of absorption of the green light in silicon is similar to that of UV (both are sub-micron), and the features machined have similar resolutions. These considerations have made it such that green is the colour of choice in our lab for chlorine-assisted laser machining of silicon for precision features. Etching used for maximum reproduceability and precision are typically 1,000 µm3/s. Higher rates have been obtained with green; the use of a continuous-wave beam would further increase the etching rate. Fig. 7: Depth of single-pass etching as a function of chlorine pressure, with UV light, for an 8.7-mm-thickness chlorine layer above the target. An initial increase is followed by a decrease corresponding to the lower UV power available at the substrate due to absorption of the light by the chlorine in the laser beam’s path (see insert for laser power as a function of chlorine gas pressure). At around 100 Torr, the laser power falls below the 3-mW threshold needed for etching, and the depth etched essentially drops to zero.

Fig. 8: Etch depth in 1 laser pass versus chlorine pressure, for three different laser powers. Laser parameters: λ = 1,053-nm light, 100-ns pulse duration, 5-kHz repetition rate, 10-µm focal spot diameter. 2.4 Infrared light

Where our work significantly departs from previous contributions in the field is by the introduction of an infrared wavelength. When a higher etch rate is needed – to remove larger amounts of material, for example – the use of infrared pulses can be helpful, for several reasons. First, since the laser’s IR output power is higher than its frequency-double green or frequency-tripled UV, there is more power available for etching. More importantly, from a physics point of view, at 1,053-nm the 1/e absorption length of the laser in silicon is 710 µm, or more than the thickness of a typical silicon wafer. This means that instead of heating only a thin “pancake” of silicon near the surface (for green) or an even thinner “crèpe” (for UV), IR actually heats all the silicon in the confocal parameter of the focusing lens. This results in a much larger volume of silicon ready to interact with the chlorine gas as the uppermost layer is etched away, and consequently in a higher etch rate. In fact, in comparing our UV and IR etch rates, we found that an increase in power by a factor 4.3 when going from UV to IR resulted in a 30-fold increase in the silicon etching rate. This improved rate cannot be explained solely through the power increase. This scenario will be further validated by the results of our finite-difference simulations presented in the next section. Unlike the UV case, IR light is not absorbed by the chlorine. Figure 8 shows the increase in etching depths as a function of chlorine pressure. This data was taken at three different laser powers, but at the same feedrate of 4-mm/min. Again, the depth of the features can be controlled through the number of passes, but much deeper features can be obtained. Indeed, chlorine-assisted IR laser micromachining has been used in our lab to cut silicon wafers right through. The

presence of chlorine is critical to both the quality of the cut and the etching rate, as is demonstrated through Figure 9. The annulus in Fig. 9b was cut with the same laser power and number of passes (3) as the one in Fig. 9a, but with an etching rate of 280,000 µm3/s with chlorine versus the less than 6,000 µm3/s obtained in the case without chlorine (Fig. 9a), a factor of almost 50 times faster. At the laser repetition frequency of 5-kHz that was used for the IR experiments, 280,000 µm3/s represents 56 µm3/pulse.

a) b) Fig. 9: Comparison of an annulus cut a) in vacuum and b) in 400 Torr of chlorine gas. Laser parameters: 2.6-W of λ = 1,053-nm light in 520-µJ pulses, 100-ns pulse duration, 5-kHz repetition rate, 10-µm focal spot diameter for an on-target intensity of 7x109 W/cm2. In three passes, the annulus in a) is 8-µm deep (6,000 µm3/s etching rate) and the one in b) is 380-µm deep (280,000 µm3/s etching rate). Photographs taken with an optical microscope.

3. NUMERICAL SIMULATIONS

In order to understand more quantitatively the difference between the UV/green-silicon interactions and the IR-silicon one, a two-dimensional (2D) heat flow model was implemented numerically through the finite-difference method. From the 1/e absorption length alone, we expect that the heat flow pictures be quite different in the two regimes. 3.1 The model The model has been developed to predict and understand etch rates, heat flow and general light-matter phenomena during laser-target interaction. In this version of the model we account for: 1) linear absorption of the laser pulse, 2) heating and heat diffusion, and 3) phase transition from condensed to liquid to gas state of the material. We do not model photon interactions and heat diffusion in the gas phase, and we consider the liquid-to-gas transition as removal of the material. The main equations going into the model is the diffusion of heat-energy density Q with a source of heat A(r,z,t):

),,(12

2

2

2

tzrAzQ

rQ

rrQ

tQC +

∂∂

+∂∂

+∂∂

=∂∂ κρ , (1)

where Q is the heat density, ρ is the material density, C is the heat capacity of the material, κ is the thermal conductivity, t is time, z is the depth in the target, and r is the radial co-ordinate away from the centre of the laser beam. The laser beam is modelled as having Gaussian profiles in time and radially:

2)2(~ τ

t

e−

, 2

0)(

~ Rr

e−

, (2)

where τ is the time duration of the pulse (at full-width-half-maximum) and R0 its spatial radius (also at half-maximum). The laser pulse energy is calculated using

0202

120

22

)2(0

2

IRyxeteIERyx

t

p ⋅⋅⋅=∂⋅∂∂= ∫∫∞

∞−

+−∞

∞−

−τππτ , (3)

and the laser pulse peak intensity is defined as

2/32

20

0πτR

pEI = . (4)

In this model, a spatial volume element – or a cell – is a ring of radius r with volume

)32(]][[ 2 −∆∆= kzrkjV π , (5)

where j is the z-direction index and k is the r-direction index. The finite-difference energy density calculation is then solved using

∆+

+−

∆

∆+

+−

∆

∆+

−

∆

∆+=

∆−−∆−

−−∆−

−+

−+−++

0

2

0

2)1(

))1(

()22(0

],1,[],,[],1,[2

]1,,[],,[]1,,[2]1,,[]1,,[2],,[],,1[

2

22

Rrk

Rrkti

kjikjikji

kjikjikjikjikjikjikji

eeteIQQQztD

QQQrtDQQ

krtDQQ

ατ

τ

α

(6)

where the diffusion coefficient D is defined as

CkD

ρ= . (7)

The temperature in the various cells is then obtained by

iI TCkjQT +=

ρ]][[

, (8a)

meltII TT = , (8b)

CL

TCkjQT f

iIII −+=ρ

]][[, and (8c)

VapIV TT = , (8d) where TI is the temperature below the melting point, TII is at the melting temperature (1,687 K for silicon), TIII is the temperature in the liquid phase and TIV is the vaporisation temperature (3,173 K for silicon). Ti is the substrate’s initial temperature (room temperature in the present simulations, or 20 C). Other thermal constants used for silicon are the

thermal conductivity κ =148 (W/m-K) and the heat capacity C = 705(J/Kg-K). The absorption coefficients in silicon used for the various laser wavelengths are αIR = 14 cm-1 for the infrared, αG = 1.06x104 cm-1 for the green and αUV = 1.08x106 cm-1 for the ultraviolet. When a cell reaches a sufficient energy density to vaporise, then it is set to a temperature of Tvap+1. Subsequent computations look for adjacent cells with this temperature and disregard them as possible heat sources or sinks. The absorption of energy of these cells is ignored and the cells that lower temperatures will absorb the corrected amount of energy. Other approximations that are made in the code include: - the density is not a function of temperature; - the heat capacity is not a function of temperature; - the thermal conductivity is a function of temperature; - the reflectivity is a function of the spatial profile of the material as well as the temperature; - no material flows in the liquid or gaseous phases; - there is no surface heat radiation; and - there is no plasma generated (vaporised cells just disappear). 3.2 Preliminary simulations results The code was written in C, and it is still in its preliminary phase. For example, it is strictly a heat diffusion code so far, and does not yet include the chlorine-silicon chemistry. Nevertheless, early simulations give an interesting look at why UV/green machining is different from IR machining. A qualitative agreement with our earlier statement about the heat distribution in the silicon can be seen in figure 10, where a single laser pulse was made to hit the solid silicon target. Figure 10 shows the temperature profiles as a function of the depth into the silicon sample (vertical axis labeled “Z Axis”) and the distance from the centre of the laser pulse hitting the sample (horizontal axis labeled “Radial Axis”), 300 ns after the laser pulse. All simulations are for 100-ns pulses, but Fig. 10a is for a 50-µJ UV pulse, Fig. 10b is for a 50-µJ green pulse, and Fig. 10c is for a 50-µJ IR pulse. While the temperature profiles for the UV and green interactions are quite similar (Figs. 10a and 10b), they are both quite different from those for the IR interaction for the same pulse energy (Fig. 10c). In the UV and green case, the laser energy was absorbed in thin layers near the surface of the target, thereby allowing the heat diffusion to happen in a three-dimensional fashion (in the z direction and radially), giving a characteristic hemispherical profile for the temperature gradient. In contrast, for the IR case the energy is absorbed much more gradually through the whole thickness of the silicon target along the laser propagation axis (the z axis). This only allows the heat diffusion to happen in a cylindrical 2-dimension fashion, as shown in Fig. 10c. Because of the lower amount of energy absorbed in the target (some just goes right through it), the much larger volume over which the absorbed laser energy is distributed and the initially larger area available for the heat diffusion, the temperatures for a 50-µJ IR pulse do not rise as high as for the UV and green pulses of the same energy. This is why “holes” can be seen in Figs. 10a and 10b – the dark area or removed material near the origin – while there is none in Fig. 10c. For the IR case, the temperature never got high enough for vaporisation. Higher pulse energies (in the mJ range) show holes going right through the 50-µm-thick simulation target for all three colours, but it is unwise to use the code at these high energies to study this problem without including the silicon-chlorine chemistry effects. While the preliminary results shown in Fig. 10 are phenomenologically sound and agree with what one would expect from the physics involved, much work has still to be done on the code for it to be used as a useful validation tool. The simulation results cannot readily be compared to the experimental results just yet. The main improvement needs to be the inclusion of the chemistry between the silicon and the chlorine to truly represent the physical processes at hand. That way, the silicon won’t have to get to the vaporisation temperature to be “etched”, but will combine with the chlorine at melt or near-melt temperatures. This should make quite a difference in the IR case, where it is more difficult for the temperature to rise.

4. CONCLUSION

We are using a diode-pumped solid-state Nd:YLF laser at 1,053-nm, 526-nm and 351-nm wavelength for chlorine-assisted laser micromachining of silicon. The use of a frequency-convertible DPSS laser makes possible several colours

for the machining and – most importantly – introduces an infrared wavelength to the process. We show that the IR greatly enhances the etching of the silicon (up to 300,000 µm3/s in our experiments) and explain it through its long absorption depth in the silicon. This is further validated by our 2D finite-difference heat-flow simulations. While UV and green radiation is absorbed in very thin slices at the surface of the silicon, IR heats the whole confocal volume and thus supplies a large amount of hot silicon for the chlorine to etch. It is therefore conceivable to design a chlorine-assisted silicon etching system with the three wavelengths, using the IR for the rough work, and the green and UV for finer work. The result is a simple, compact and quick direct-write method for the creation of fine silicon structures for applications to components integration for photonics and labs-on-a-chip for biomedical research. Work continues in our lab to further improve this technique and obtain flatter etched surfaces, more vertical walls and yet higher etching rates.

10 20 30 40 50

10

20

30

40

50

Temperature Profile 300ns after pulse100ns, 50µJ, λ=351nm

Radial Axis (µm)

Z Ax

is (µ

m)

2100 -- 2400 1800 -- 2100 1500 -- 1800 1200 -- 1500 900.0 -- 1200 600.0 -- 900.0 300.0 -- 600.0 0 -- 300.0

10 20 30 40 50

10

20

30

40

50

Temperature Profile 300ns after pulse100ns, 50µJ, λ=526.5nm

Radial Axis (µm)

Z Ax

is (µ

m)

2100 -- 2400 1800 -- 2100 1500 -- 1800 1200 -- 1500 900.0 -- 1200 600.0 -- 900.0 300.0 -- 600.0 0 -- 300.0

a) b)

10 20 30 40 50

10

20

30

40

50

Temperature Profile 300ns after pulse100ns, 50µJ, λ=1053nm

Radial Axis (µm)

Z Ax

is(µ

m)

173.6 -- 180.0 167.2 -- 173.6 160.8 -- 167.2 154.4 -- 160.8 148.0 -- 154.4 141.6 -- 148.0 135.2 -- 141.6 128.8 -- 135.2 122.4 -- 128.8 116.0 -- 122.4 109.6 -- 116.0 103.2 -- 109.6 96.80 -- 103.2 90.40 -- 96.80 84.00 -- 90.40 77.60 -- 84.00 71.20 -- 77.60 64.80 -- 71.20 58.40 -- 64.80 52.00 -- 58.40 45.60 -- 52.00 39.20 -- 45.60 32.80 -- 39.20 26.40 -- 32.80 20.00 -- 26.40

c) Fig. 10: Simulation results of laser-silicon interactions with various parameters, all taken as snapshots of the temperature profiles 300 ns after the laser pulse. a) UV pulse at 351-nm wavelength, 50 µJ of energy, b) green pulse at 526-nm wavelength, 50 µJ of energy, and c) IR pulse at 1,0530-nm wavelength, 50 µJ of energy.

ACKNOWLEDGEMENT

This work is supported by Zenastra, the University of Toronto, Photonics Research Ontario (PRO) and Materials and Manufacturing Ontario (MMO) through a PRO/MMO Industrial Collaborative Grant, and in part by the Canada Foundation for Innovation, and the Ontario Ministry of Energy, Science and Technology, the University of Toronto, McMaster University and 7 industrial partners through an Ontario Research and Development Challenge Fund grant. The authors would like to thank University of Waterloo co-op students Howard Siu and Nathan Douglas for their hard work on this project.

REFERENCES

1) G.A. Ozin and S.M. Yang, Adv. Funct. Mater. 11, 95 (2001). 2) D.J. Ehrlich, R.M. Osgood, and T.F. Deutsch, Appl. Phys. Lett. 38, 1018 (1981). 3) K.W. Beeson and V.H. Houlding, J. Appl. Phys. 64, 835 (1988). 4) W. Sesselmann, E. Hudeczek, and F. Bachmann, J. Vac. Sci. Technol. B 7, 1284 (1989). 5) M. Müllenborn, H. Dirac, and J.W. Petersen, Appl. Phys. Lett. 66, 3001 (1995). 6) S.C. Jacobson, R. Hergenröder, L.B. Koutny, and J.M. Ramsey, Anal. Chem. 66, 2369 (1994). 7) See the work from Revise at <www.revise.com>, for examples. * email: mnantel@pro.on.ca; phone: (416) 978-3932; fax: (416) 978-3936; http://www.pro.on.ca; Photonics Research Ontario, Suite 331, 60 St-George Street, Toronto, ON, Canada M5S 1A7