2.1.

Approach Using Spray Coating and Angled Exposure

The fabrication starts from SOI wafer. After preparing the p-n junction on the device layer of n-type Si (1 to $20\mathrm{\Omega }\text{-}\mathrm{cm}$) by diffusing boron, the cells are separated from each other using the deep RIE. The cell islands are thermally oxidized covering the surface with 200-nm-thick ${\mathrm{SiO}}_2$. The subsequent process needs 3D patterning for opening contact holes on the ${\mathrm{SiO}}_2$ cover and for defining sputter-deposited Al metal pattern, since the sample is no longer planer.

Figure 2 shows the schematic drawing of the spray coating. The spray flow is irradiated from the nozzle to the sample. Different from the standard spin coating, the gas flow carries the photoresist as a microparticle. The particles are expected to enter inside the trench. Although the trench is open upside, the actual deposition frequency or the film thickness distribution is influenced by the gas flow and the trench shape. The sprayed resist film becomes thinner at the deeper position inside the trench compared to that at the top surface. This can be explained theoretically considering that the spray flow is blocked against entering inside the trench by the confined air inside the trench.²¹ The subsequent patterning technique is required to allow the resultant nonconformal thickness of the resist.

Fig. 2

(a) Schematic drawing of the spray coating of the photoresist on the trench sample. The photo indicates one example of the resist profile. (b) Photo of the spray coater used.

For irradiating a UV light pattern on the vertical sidewall, the angled exposure is indispensable. Figure 3(a) shows the actual setup. Notice that the angled light generates the reflection, which patterns other surfaces in addition to the first incident surface as shown by the dotted smaller arrows in Fig. 3(b). This reflection generates the problem of patterning the secondary incident surface. Since the sprayed resist film is thinner at the deeper region, the light energy adjusted for patterning the top surface is overdosed for the deeper region. Figure 3(b) includes the thickness information of the sprayed positive resist (Tokyo Ohka Kogyo, TMMR P-W1000). The resist thicknesses at the top and the bottom surfaces are about 6 and $2\mu \mathrm{m}$, respectively. The appropriate exposure energies are 300 and $150\mathrm{mJ}/{\mathrm{cm}}^2$ for the top and bottom surfaces, respectively. The exposure dose is adjusted to $300\mathrm{mJ}/{\mathrm{cm}}^2$ for patterning the top surface. This should be decreased to $150\mathrm{mJ}/{\mathrm{cm}}^2$ on the bottom of the trench. Therefore, the absorbent liquid immersion is applied.²⁰ The trench is filled with the water with black ink for attenuating the incident light power. Since the trench is $50\text{-}\mu \mathrm{m}$ wide and $25\text{-}\mu \mathrm{m}$ deep, the attenuation ratio can be decided by the concentration of the black ink as the design parameter. The light intensity decreases while UV propagates in the trench. The overdose problem as well as the reflection problem can be removed.

Fig. 3

Schematic drawing of the angled exposure with the absorbent liquid immersion. The setup is for opening contact holes. Schematic drawing of the angled exposure with the absorbent liquid immersion. The setup is for opening contact holes.

Figure 3(b) includes the information of the optical setup for opening two contact holes (size: $15\times 50\mu {\mathrm{m}}^2$). One is on the top surface (p-type region). Another is at the sidewall bottom (n-type region). Different from the planer lithography, the resist thickness ($6\mu \mathrm{m}$ here) has to be taken into account in the mask design since the vertical position shift of the mask causes the lateral position shift of the light pattern. The hole pattern is incident not only on the sidewall, but also on the bottom as shown in Fig. 3(c). This gives the lateral alignment margin of $\pm 4\mu \mathrm{m}$. The wiring Al film is sputter-deposited. The thickness on the top surface is about $1\mu \mathrm{m}$. This thick film is for covering over the corrugations of the scallop on the sidewall. Next is preparing the wiring pattern. Again the angled exposure is applied to pattern the metal lines as shown in Fig. 3(d). Notice that some parts are inevitably in the shadow caused by the neighboring cell, and that the Al layer remains there. The design allows the remained Al for realizing the metal line connecting neighboring cells as shown in Fig. 3(d).

2.2.

Approach Pasting a Sheet Including Resist and PVA Layers

Figure 4 illustrates the principle for patterning the vertical sidewall using the sheet. The resist film is originally pasted over the trench making the bridge. The resist is patterned as the cantilever. The left side in Fig. 4 is the fixed point. The right side is the free end. This shape is formed when the developing process proceeds in the liquid developer. When the rinsing water decreases upon being dried, the surface tension of the water pulls the cantilever film inside the trench. Eventually, the film adheres on the vertical sidewall. Once the resist film is on the sidewall, the additional post-baking will strengthen the adhesion of the photoresist to the substrate. Patterning the vertical sidewall is possible using UV irradiation normal to the substrate. The maximum depth of the pattern is the width of the trench corresponding to the aspect ratio of 1. For realizing the submount with the required height, the trench width becomes the parameter to be designed. Notice that this method is effective for patterning the one side of the sidewall and not for both sides connecting lines as shown in Fig. 3(d).

Fig. 4

Patterning the vertical sidewall by bending a part of the planer pattern. This patterning is for the metal line connecting the top side to the vertical sidewall. Drawings show the conditions (a) before bending (b) after bending.

Figure 5 shows the fabrication sequence starting from the spin coating of the photoresist on the newly introduced sheet (SO sheet, Aicello Corp.) which consists of a PVA layer and a PET substrate. (1) Since the sheet shape is planer during spinning, a uniform thin film is obtained. (2) The resist/PVA layers are pasted on the substrate making the bridge over the trench. The pasting temperature of 90°C is somewhat lower than the glass transition temperature of the resist. (3) The patterning defines the cantilever shape to the resist bridge. The mask alignment adjusting to the sample trench is exactly the same operation as that of the standard aligner. (4) PVA is dissolved in water, and the resist film gets the pattern shape being developed. (5) After replacing the developer in the rinsing water, the sample is slowly dried. The surface tension of water attracts the resist cantilever to the trench inside. (6) Finally, the resist film adheres on the vertical sidewall. Notice that all process steps proceed in parallel on the substrate indicating the high throughput.

Fig. 5

Fabrication sequence. In this drawing, the negative photoresist is supposed.

As shown in Fig. 6, the PVA layer in SO sheet can be peeled from the PET sheet manually. The PVA thickness used is about $35\mu \mathrm{m}$. Table 1 lists the characteristics of PVA compared with that of the photoresist. PVA is water-solvable and stable against organic solvents (e.g., acetone).²² The photoresist is acetone-solvable and stable against water. So, these two materials are opposite in their solvents. The photoresist liquid can be spin-coated on the PVA layer without suffering from the organic solvent. PVA can be selectively removed from the photoresist. This is useful for the process design.²³ In addition, PVA is transparent. Figure 7 shows the absorption spectrum. The absorption is only 4% at the wavelength of 400 nm near to the i/g lines used for the patterning. This transparency allows the patterning of the photoresist through the PVA layer.

Fig. 6

PVA film on PET sheet. This photo shows the peeled PVA from the PET sheet.

Table 1

Characteristics of PVA and photoresists.

Fig. 7

Absorption spectrum of PVA. PVA is transparent at UV wavelengths. The patterning through PVA film is possible.

Figure 8 shows the mask design for patterning the vertical sidewall showing 8 units. A $\mathrm{\Gamma }$-shape pattern has $400\text{-}\mu \mathrm{m}$ width for all units. The right-side I-shape thinner pattern has widths of 10, 25, 40, and $100\mu \mathrm{m}$. The maximum width examined is $350\mu \mathrm{m}$. The cantilever length is $315\mu \mathrm{m}$ (the length from the trench edge). The trench width is $350\mu \mathrm{m}$. Sixteen parallel trenches are prepared by the half-depth dicing of the substrate (size: $1\times 2{\mathrm{in.}}^2$). Thus, the sidewall surface is relatively rough. The trench width and depth are both $350\mu \mathrm{m}$ in design. One corner is chamfered using the bevel blade. The substrate material is alumina or Si.

Fig. 8

Photomask pattern relative to the trench. The number on each element means the line width at its right bottom, which will bend to the vertical sidewall.

The positive photoresist used is AZ-1500 (AZ Electronic Materials), 38 cp. The spin-coating is at 3000 rpm giving a thickness of about $2.4\mu \mathrm{m}$. The exposure dose is $100\mathrm{mJ}/{\mathrm{cm}}^2$. The negative photoresist used is TMMR SA390N (Tokyo Ohka Kogyo Co., Ltd.), 100 cp. This resist is softer and adhesive to the sample. As shown in Fig. 9, Young’s modulus is 250 MPa, which is only 1/20 compared to that of the positive resist. The value before the cross-linking reaction will be smaller than that. The spin-coating is at 2000 rpm. The resist thickness is about $5\mu \mathrm{m}$. The exposure dose is $200\mathrm{mJ}/{\mathrm{cm}}^2$. The negative photoresist without the exposure dissolves in the developer. This is good for removing the residue and useful for the thick film.

Fig. 9

Characteristics of the photoresists used.

3.1.

Pattern Obtained by Spray Coating

Figures 10(a) and 10(b) show the whole device of the photo cells connected in series. The $5\times 5$ photo cell array is inside the area of $1.5\times 1.5{\mathrm{mm}}^2$. The single cell size is $250\times 250\mu {\mathrm{m}}^2$ with the height of $25\mu \mathrm{m}$. The Al pattern on the top surface is $35\times 70\mu {\mathrm{m}}^2$, which is 3.9% of the top area of the photo cell. The aspect ratio of the separation trench is 0.5. The outside larger islands with numbers are electrodes for probing. The voltage and the current can be measured for each row. The pattern on the vertical sidewall is $70\text{-}\mu \mathrm{m}$ wide for all cells. The connection between columns is the rounding path on the oxide bottom layer. This design is for simplifying the angled exposure. This rounding path on the bottom is $20\text{-}\mu \mathrm{m}$ wide at the center of the $50\text{-}\mu \mathrm{m}$-wide trench. The smaller 5 square ($100\times 100\mu {\mathrm{m}}^2$) cells for each column are the bypass diodes. Figure 10(c) shows the magnified view. The angled exposure is single UV irradiation from one side. The resist pattern for Al etching is on one sidewall, and the Al film remains on the other sidewall as indicated in Fig. 10(d). Figures 10(e) and 10(f) show the sidewall patterns viewed from the opposite side from others [Figs. 10(b)–10(d)]. Figure 10(f) shows the close-up view of the contact electrodes at the sidewall and the top surface. The contact holes are covered by Al film. The electrical connection from end to end is successfully obtained with a yield of about 50% at present.

Fig. 10

SEM images of the photo cells with 3D wiring across vertical side walls. (a), (b) Whole image of $5\times 5$ cells. $1\times 5$ cells are one group. (c), (d) Magnified view of photo cells and the bypass diode. (e), (f) Close-up views from the opposite angle observing contact holes.

Figure 11 shows the $I\text{–}V$ curve for 10 cells in $5\times 5$ arrays. The light illumination is from the white light-emitting diode (measured to be about $4\mathrm{mW}/{\mathrm{mm}}^2$ supposing a 550-nm wavelength). Under the light illumination, the average open voltage is $0.241\mathrm{V}/\text{cell}$, which indicates that there is a margin for improving the quality of the p-n junction.

Fig. 11

Typical $I\text{–}V$ curves of 10 cells connected in series from $5\times 5$ array. The short-circuit current $I_{\mathrm{sc}}$ is $-5.89\mu \mathrm{A}$. The open-circuit voltage $V_{\mathrm{oc}}$ is 2.51 V. The fill-factor defined by $V_{\mathrm{mp}}$ $I_{\mathrm{mp}}/V_{\mathrm{oc}}$ $I_{\mathrm{sc}}$ is 0.827. $I_{\mathrm{mp}}$ and $V_{\mathrm{mp}}$ are the current and the voltage operating point, respectively, which give the maximum power output, where $V_{\mathrm{mp}}$ is 2.37 V and $I_{\mathrm{mp}}$ is $-5.16\mu \mathrm{A}$.

3.2.

Pattern Obtained by Sheet

3.2.1.

Positive resist

Figure 12(a) shows SEM images of the patterning result obtained by the sheet using the positive resist. The vertical pattern is obtained on the sidewall. The pattern on the top surface is stable and working as the fixed point for the cantilever. The fixed point [indicated by the white arrow in Fig. 12(b)] is stable even if its area ($40\times 60\mu {\mathrm{m}}^2$) is smaller compared to that of the cantilever. Figure 12(b) shows the film divided from the corner edge. This is the main failure mode. This can be explained by the relatively thin edge and the material fragility. Figure 12(c) shows the resist film fixed with the wrinkle. This indicates the importance of the quiet water condition. Figure 12(d) shows the remaining resist at the corner edge of the trench. This will be caused by the locally thickened resist and the underdosed UV exposure for patterning the positive photoresist. Figure 13 shows the illustration for explaining the above phenomena. The resist material can migrate at the corner from the surrounding area during the heating step for pasting the film (step 2 in Fig. 8).

Fig. 12

(a) Patterned positive resist connecting from the top surface to the vertical sidewall. (b) Resist film divided from the corner edge. (c) Resist film pasted on the sidewall with the wrinkle. (d) Corner edge with the local resist residue.

Fig. 13

Comparison of the positive and negative photoresists inside the bridge for avoiding the residue.

3.2.2.

Negative resist

Figure 14(a) shows the pattern prepared using the negative resist. The pattern from the top surface to the sidewall is well obtained. After the rinsing water is removed from the dish, the water remains in the trench. The substrate is continuously dried at room temperature for about 5 min. The rinse water drying speed is necessarily slow, especially for the narrower pattern. The wider pattern (e.g., $400\text{-}\mu \mathrm{m}$ width) is stable in the direction to the bottom. Since each cantilever separates individually, the narrower pattern (e.g., 10- or $25\text{-}\mu \mathrm{m}$ width) tends to be slanted as shown in the white arrows in Fig. 14(a). The condition of the water in the trench may influence the result. The undermost trench in Fig. 14(a) has the vertical patterns with wrinkles. This may be the influence of the local water flow in this trench. Figure 14(b) is the magnified image of $100\text{-}\mu \mathrm{m}$-width pattern at other position in the substrate. Figure 14(c) shows the pattern having the connecting line between the free ends of the cantilevers. The pattern direction is stabilized to be straight vertical even for the $10\text{-}\mu \mathrm{m}$-wide line. There is the design variation of the connecting line width to be 15, 30, and $45\mu \mathrm{m}$. For all line widths, the stabilization effect seems to be same. Figure 14(d) shows the magnified corner with a $10\text{-}\mu \mathrm{m}$-wide line. There is no residue at the corner. This is in contrast to Fig. 12(d), which is the case using the positive resist. The resist thickness seems to be same at the top surface, the corner, and the sidewall. This is also in contrast to the inset of Fig. 2(a) obtained using the spray coating.

Fig. 14

Patterned negative resist connecting from the top surface to the vertical sidewall. (a), (b) Pattern having the design of the individual cantilever. (c), (d) Pattern having the design with connecting line between cantilever ends. This line width is $45\mu \mathrm{m}$. The photos are obtained at different positions from the same substrate.

3.2.3.

Stencil mask using positive resist

Based on the fact that the positive resist is relatively rigid for folding, this material is rather useful for the stencil mask. The process revision is changing the mask tone making the holes in the resist bridge and increasing the resist thickness to about $7.1\mu \mathrm{m}$. Figure 15 shows the obtained pattern. The clear hole pattern is well obtained. Since this resist film contacts on the sample surface, this is ideal for minimizing the blur of the metal deposition at the subsequent process step. The membrane over the trench somewhat deflects. This can be explained by the deformation due to the surface tension of the rinsing wafer. Inside the resist membrane, some cracks are observed as shown in Fig. 15(b). They will be removed with the thicker resist membrane or the gentle drying condition reducing the surface tension. There is no residue of the photoresist as seen in Fig. 12(d). The metal deposition from the angled direction using such resist mask will give the 3D wiring across the vertical sidewall.

Fig. 15

Patterned positive resist as the stencil mask over the trench. (a) Reversed tone compared to Fig. 14. (b) Another part around the corner edge of the trench with the cracks in the film.