1. Introduction

Conventional Si etching can be performed by dry plasma etching [1], wet anisotropic etching [2], laser ablation using Nd:YAG [3-6], Excimer [6,7] or femtosecond lasers [8,9] with wavelengths between 1.06 μm and 193 nm. No CO₂ laser with a 10.6 μm wavelength was directly used for Si etching.

In terms of plasma or wet etching process, they all need photolithography to define the pattern geometry and select the suitable mask materials of the resist or hard mask to meet the different requirements of chemistry endurance and etching depth. Their etching rates are generally in the range of 0.5-3 μm/min, dependent on the chemistry, process conditions and sample geometry. Such etching rates are still low, especially for a through Si wafer etching of 525 μm, and it will take several hours. The equipment cost of dry plasma etching is high and the etched geometry for wet etching is limited to crystallography and directionality.

In terms of laser ablation, Nd:YAG lasers with wavelengths of 1.06 μm, 532 nm and 355 nm, or KrF, ArF Excimer and femtosecond lasers with shorter wavelengths were used for direct Si ablation due to the adsorption in Si. But the cost of above lasers is much higher than CO₂ laser at the same laser power. Although a Si wafer is normally almost transparent to the CO₂ laser, Wang et al. [10] reported that periodic surface structure of Si with a thin Au film coating was achieved by pulsed CO₂ laser through a melting-and-resolidification process. Yang et al. [11] reported that the thin native oxide of Si surface could absorb pulsed CO₂ laser to result in the surface damage, not the bulk etching. Both above processes were related to material melting and resolidification. Chung et al. [12] reported an interesting phenomenon of Si etching involved in the melting-and-evaporation process with the help of glass in an ambient atmosphere. It is different from the old techniques using CO₂ laser assisted Si etching in the Fluorine-bases gases [13] or NaOH liquid [14] chemical etchants.

In this paper, we present more experimental evidences to confirm this repeated Si deep etching behavior and establish its mechanism with sufficiently semi-quantitative discussions referred to the principles of the laser-material interaction, semiconductor physics and process [15-19]. Also, this technology is defined as Glass Assisted CO₂ LAser Processing (GACLAP).

2. Experimental procedures

Figure 1 shows the schematic diagram of GACLAP setup. It includes a computer, a continuous wave (CW) CO₂ laser source, a reflective mirror, a focal lens and an experimental sample with a 525 μm thick Si wafer or chip on the top of a glass wafer or plate. The single-side polished p-Si(100) wafers were used with an average roughness around 1 nm and resistivities of 1- 20 × 10^-2 ohm-m, corresponding to a boron doping concentration around 7×10²⁰- 10²² m^-3. The glass can be a Pyrex glass or others which can absorb CO₂ laser energy. A pure Si sample was used as a reference for comparing the Si etching behavior with the glass assistance. The polished surface of the Si wafer was contacted to the bright smooth glass surface with the help of another heavy glass on the top of Si edges. The sample preparation for the contact of Si and glass is easy in this study. We used the clean-room class Kimberly-Clark wiper to clean the surfaces of Si and glass, and then slid two samples slowly to contact by hand. Thus both Si and glass could be closely contacted by electrostatic force and easily separated after experiments. The purpose of another heavy glass put on Si edges was used to fix the sample for the following laser processing.

Commercial desktop air-cooled CO₂ laser equipment (VL-200, Universal Laser system Inc., U.S.A.) was used with a maximum laser power of 30 W. The CO₂ laser uses a sealed-off, RF excited, slab design and a multi-pass, free space resonator for a very good quality beam. The fastest scanning speed of laser was 1140 mm/s. The largest working area was 409 × 304 mm² and the focal length of the lens is 38.1 mm. The smallest beam spot size offered by the commercial technical data could reach at 76 μm in FWHM definition. The laser spot was focused on the top surface of Si during GACLAP. A computer aided design program of CorelDraw software was available to set the experimental parameters of laser power, scanning speed and number of beam passes for auto-controlled processing. Two primary kinds of linear and circular patterns were drawn to study the Si etching behavior. The linear length of 5 mm and circular diameter of 300 μm were used in the experiments. At a constant power of the laser beam, the laser scanning speed could be adjusted between 0% and 100%. Because higher laser power and slower scanning speed were good for the observation of GACLAP phenomena, we controlled the laser power between 15 and 30 W and scanning speed between 2.3 and 11.4 mm/s. In order to see the pronounced GACLAP effect on Si etching, a nearly maximum laser power of 30 W and a low scanning speed of 2.3 mm/s were used to perform the etching of six linear patterns respectively with 10, 20, 30, 40, 60, and 80 passes. The sample was then cross-sectioned in order to evaluate the etching depth. The circular pattern was used for examining the hole drilling capability. Each condition was done at least two times for the repeatability. The optical microscopy (OLYMPUS BX51 M, Japan) and digital camera were used to examine the cross-sectional and planar-view images of the etching results. The morphology and profile of trenches was examined by the alpha-step (α-step) profiler (Kosaka Lab, ET3000, Japan).

 

Fig. 1. The schematic diagram of glass assisted CO₂ laser processing setup

Download Full Size | PDF

3. Results and discussion

Figure 2 (a) shows the optical micrograph of the Si sample on a glass with six evident etching lines from right to left using the CO₂ laser at a fixed 30 W power and a 2.3 mm/s speed for 10-80 passes. The Si etching behavior by CO₂ laser is changed as it is put on a Pyrex glass plate compared to pure Si normally transparent to CO₂ laser where no etching occurs even at more passes input [12]. The sample was cleaved from the middle of trenches for the cross-sectional observation to examine the etching behavior. Figures 2(b)-(c) show the representative Si trenches etched at 10 and 80 passes, respectively. The cross section of the trench exhibits a V-like shape subjected to the Gaussian distributed laser energy. The occurrence of asymmetrical profile is due to the scattering effect of photon at high temperature in this new mechanism which leads to the non-uniform absorption of Si etching involved in a melting-and-evaporation process. The maximum roughness on the rim of each trench measured by the alpha-step profiler was in the range of 1 to 20 μm, roughly increasing with pass number. Figure 2 (d) shows the overall relationship between the etching depth and laser passes. The etching depth measured was about 208, 236, 300, 354, 408 and 415 μm corresponding to 10, 20, 30, 40, 60, and 80 passes, respectively. The depth of trench non-linearly increases with increasing laser passes at the fixed power and scanning speed. The variation of depth in the repeated experiments exhibits the similar trend of etching, that is, the etching depth increases with the pass number. The depth may have tens μm difference in the repeated experiments attributed to the variation of contact between the Si and glass surfaces in our simple setup. The closer is the surface contact, the higher the energy transfer and the etching efficiency. The non-linear etching rate increases first at low-medium passes, then decreases at higher passes. It is because of the absorption of the laser beam at the sides of the trench near the surface that prevents absorption of the laser energy deeper inside the sample. This would also lead to the enlargement of the V-shaped trench near the surface at higher passes (Figures 2(b)-(c)).

 

Fig. 2. The optical micrograph of: (a) Si on a glass for 10- 80 passes with six evident etching lines. The representative cross-sectional images of Si trenches at: (b) 10 and (c) 80 passes. And (d) the etching depth as a function of laser passes.

Download Full Size | PDF

Figure 3(a) shows the optical micrograph of the circular Si pattern in a 300 μm diameter etched by GACLAP technology at a 24 W power and a 5.7 mm/s speed for 20- 120 passes. Six visible holes including three blind ones had been achieved from right to left corresponding to the 20, 40, 60, 80, 100 and 120 passes drilling. The transparent holes with visible white light penetration indicate that Si had been etched through wafer of 525 μm thick after 80 or more passes drilling. All the magnified images of holes had been shown in Figures 3-4 of reference 12. The diameters of holes increase rapidly with pass number in the blind holes and then tend to a saturated value in the open holes. The range is 350-550 μm corresponding to the 20-120 passes. As blind hole is etched through and pass number is too large, the morphology around the hole becomes much rough [12]. The diameter is larger than the design value of 300 μm due to the lateral removal of Si from the accumulated energy in the continuous processing and the laser beam with a 76 μm spot size in a circular motion. The etching geometry affects the energy distribution and the etching rate. The accumulated energy enhances the vertical etching rate of a small-size circle (Figure 3) compared to the long-lines etching for trenches (Figure 2). That is, the 300 μm circular pattern is etched through 525 μm depth at a lower 24 W power and a higher 5.7 mm/s speed compared to the 5 mm long line which is only etched 415 μm at the 30 W power and 2.3 mm/s speed at the same 80 passes. Also, the accumulated energy influences the diameter variation of blind holes in the photothermal process. Figure 3(b) shows the diameter of the blind holes as a function of passes in the laser power of 15-27 W at a constant 5.7 mm/s speed. The diameter of hole roughly increases with increasing laser power and pass number. It is related to the total accumulated absorption energy of Si for the lateral removal, which is proportional to the laser power and passes and affected by the efficiency of lateral heat transportation. The variation is large at low pass number.

 

Fig. 3. (a) The optical micrograph of the etched circular Si pattern by GACLAP for 20- 120 passes from right to left and (b) The diameter of the blind holes as a function of passes in the laser power of 15-27 W at a constant 5.7 mm/s speed.

Download Full Size | PDF

From Figures 2 and 3, the Si etching from the top surface toward the interface implies that Si must absorb the laser energy with the assistance of glass to modify the Si absorption through some mechanisms to make etching possible. The phenomenal mechanism of Si etching goes through the following steps: (1) absorption modification before etching, (2) low-pass absorption and shallow etching, (3) middle-pass absorption and deeper etching, and (4) high-pass absorption to through-wafer etching. Figure 4 shows the proposed model for GACLAP mechanism. Glass initially absorbs CO₂ laser energy near the interface and then heats Si to change its absorption behavior extending to CO₂ laser wavelength of 10.64 μm for etching to start from the top surface. As we put our sample in water to etch Si like liquid-assisted laser processing, no etching occurs. It proves that the high temperature is a key factor for the modified high CO₂ laser absorption of Si. Since Si has a high thermal conductivity of 150 W/(m-K) for solid (300 K) and 450 W/(m-K) for liquid (1685 K) [15], the heat can rapid transport to the surface to result in high temperature at surface to change the Si absorption behavior. The microstructure will be changed at high temperature due to the thermal (/phonon) induced Si atom rearrangement from single crystal to polycrystalline or amorphous states with high entropy and defects in both the heat affected zone (solid) beneath surface and the molten zone at surface (liquid). The new mechanisms of the modified CO₂ laser absorption in Si with varied microstructure may be involved in two main reactions: one is the variation of Si energy band structure at high temperature [16,17] for promoting the absorption of photon-electron interaction between the Si and CO₂ laser. The other is the increased Si oxidation at high temperature [13,15]. The thin oxide on the surface can directly absorbs the CO₂ laser for further etching, which may be the origin of modified absorption. The above two kinds of mechanism may occur simultaneously during GACLAP and interact each other for the high etching rate process, whose average rate larger than 240 μm/min at the condition in Figure 2. The thin oxide layer formation changes the light absorption from the interface to top surface of Si. The surface oxide is etched away after absorbing CO₂ energy and fresh oxide is formed immediately during continuous laser scanning to keep the Si at very high temperature with varied band structure to absorb CO₂ laser for deeper Si etching. The estimated temperature at the local surface of Si from the heating effect of glass and the isolation of the surface oxide will be between 1685 K (Si’s melting point) and 3173 K (Si’s boiling point) because the CO₂ photothermal etching mechanism is related to the melting and evaporation process for high etching rate.

 

Fig. 4. The schematic diagram of the proposed model for GACLAP mechanism.

Download Full Size | PDF

In terms of the variation of Si energy band structure at high temperature, it may include the band gap narrowing [16], the formation of new defects states in the band gap [17] near/toward the valence band due to highly increased entropy of point defect configuration, electronic polarization [18] and the rapidly increased intrinsic free electrons concentration [16]. In principle, light radiation is absorbed in the nonmetal materials by valence band-conduction band electron transitions and electron transitions involved in defect levels lying within the bandgap and electronic polarization [18]. The bandgap energy of single crystal Si at 300 K is 1.12 eV corresponding to an absorbable wavelength of about 1.11 μm. However, the 10.6 μm CO₂ laser has a 0.12 eV photon energy. The occurrence of band gap narrowing and new defect states near the Si valence band at the local etching area at the high temperature will be very important for enhancing the photon absorption. The Si bandgap energy, E_g(T), is a function of temperature in a universal formula (1) from 0 K to the melting point of material [16]:

where E_g(0), α and β constants for Si are 1.170, 4.73 × 10^-4 and 636, respectively. The bandgap energy decreases with increasing temperature. For examples, E_g(300) is 1.12 eV for 27 °C while E_g(1685) equals to 0.59 eV for Si’s melting point. If we extrapolate the data to higher temperature of liquid Si, E_g= 0.10 eV can be obtained at 2773 K. Although the value is just as a reference for the trend prediction, it implies that Si at very high temperature can absorb the CO₂ laser energy. In case the reduced E_g is not less than 0.12 eV, the bulk defect levels within the band gap may introduce valence band-defect level electron transition for enhancing the optical absorption [13,18]. The absorption increases with increasing temperature. Also, the absorption by electronic polarization is important at the light frequencies in the vicinity of the atomic relaxation or vibration frequency [18,19]. The CO₂ laser with a wavelength of 10.6 μm corresponds to a frequency of 2.8 × 10¹³ Hz. The average number of the occupied quantum state for photon or phonon particles following the Bose-Einstein distribution function ($\frac{1}{\exp \left(\mathit{hv}/\mathit{kT}\right)-1},$ h: Planck constant, k: Boltzmann constant, v: frequency of particle, T: temperature) at a frequency of 2.8 × 10¹³ Hz will be one at 1940 K [19]. The average number of occupation at 1685 K is 0.82. It is manifested that the electronic polarization contributes CO₂ absorption of Si at high temperature. The absorption increases with increasing temperature. In addition, the intrinsic concentration of Si will dominate the free carrier concentration at above 277-327 °C in our samples (10²¹-10²² m^-3 doping) and the free carrier concentration rapidly increases with temperature in log relationship [16] to 10²⁵-10²⁶ m^-3 at a temperature larger than 1685 K. These free carriers (electron and holes) make more collision with atoms to create excited electrons and holes, which increase the absorption probability. The above variations in energy band structure lead to the strong interaction between photons, electrons and phonons of materials for enhancing high absorption of CO₂ laser in Si.

4. Conclusions

Si etching has been achieved using GACLAP technology which is to put a Si sample on a glass plate for the CO₂ laser irradiation. The glass initially absorbs CO₂ laser energy near the interface of Si and glass and then heats Si to change its absorption behavior. The new mechanism may be involved in two main reactions: one is the variation of Si energy band structure at high temperature for promoting the photon-material interaction. It may include the band gap narrowing, new defect states formation, electronic polarization and rapidly increased intrinsic free carrier concentration. The other is the increased Si oxidation at high temperature. The thin oxide can directly absorbs CO₂ laser and keep high temperature for further Si etching. The above two reactions may occur simultaneously during GACLAP for high absorption of CO2 laser to etch Si. The etching depth of Si increases with increasing laser passes and power. The etching geometry will affect the energy distribution and accumulation for the etching rate. The etching of the small-size circle is faster than the linear shape at the same laser energy.