Direct laser writing lithography using a negative-tone electron-beam resist

1. Introduction

The drive for smaller and faster industrial technologies has produced an increasing demand for fabrication techniques that allow the production of nanoscale features [1–5]. Electron beam (e-beam) lithography has been widely used in the field of nanofabrication owing to its high resolution and maskless processes, which are based on direct writing [6,7]. However, e-beam lithography is relatively costly and requires a high-vacuum environment. Sometimes, it may also cause sample damage, particularly if an e-beam with a large kinetic energy is used [8,9]. The direct laser writing (DLW) technique, based on photochemical processes, is a simple, low-cost, mask-free, and path-directed method which can provide large-area structures with micrometer resolution [10–12]. Conversely, achieving a sub-micrometer resolution by conventional DLW techniques is more challenging than by e-beam lithography [13]. Recently, novel techniques such as the two-photon absorption (TPA) [14–17], stimulated emission depletion (STED) [18–20], and photothermal effects have been introduced to achieve nanoscale resolution [21–23]. The TPA technique requires femtosecond laser light sources and can hardly produce high-aspect-ratio patterns. Conversely, the STED technique uses two laser sources with different wavelengths, available on limited photo-resists. The photothermal effect is based on single-photon absorption; however, the removal of the crosslinked polymers after the lithographic process is difficult, making this process impractical for various applications [21,22].

Polymethylmethacrylate (PMMA) is a positive-tone e-beam resist that has been widely used in nanoscale lithography for fabricating a variety of high-resolution structures [24,25]. In contrast, ma-N 2400 (Micro Resist Technology GmbH) has been developed as an excellent negative e-beam resist (N-ER), capable of fabricating structures down to 50 nm resolution [26,27]. The ma-N 2400 series is composed of a phenolic resin (Novolak) as a polymeric binder and a bisazide as a photoactive compound (PAC). When exposed by the e-beam, the azide groups (R-N₃) release nitrogen (N₂), and the reactive nitrenes (R-N:) initiate the cross-linking of the resist [28,29]. Until now, the ma-N 2400 series has not been used as a photoresist in DLW lithography with sub-micrometer resolution, whereas the use under deep-ultraviolet illumination has been implied by the manufacturer [30].

In this work, we used the negative-tone e-beam resist as a photo-resist in the direct-laser writing lithography system, by which we were able to produce high-resolution metal nano-patterns. We investigated the linewidth and height of the resist after the developing processes as a function of the exposed dose and the baking temperature. Finally, we compared the achieved lithographic contrast and aspect ratio with those of conventional photoresists.

2. Experimental results and discussion

Direct laser writing lithography began with the spin coating of the N-ER (ma-N 2405) on a silicon substrate at 3000 rpm for 60 s. The thickness of N-ER was 750 nm. This was followed by annealing at the baking temperature of T_bake = 70–150 °C for 90 s, as shown in Fig. 1(a). For direct laser writing, we used a microscope equipped with galvanometer scanning mirrors (Thorlabs, Inc.) and objective lens (Olympus, ×50, NA = 0.8). After exposure to light, the sample was developed using ma-D 525 (typically for 70–90 s), which is the developing solution used for conventional e-beam lithography processes with the N-ER. The metal pattern can be formed by conventional lift-off or etching processes, followed by the removal of the N-ER using an acetone solution. For the light source, we used a laser diode (Thorlabs, Inc.) at 405 nm with a typical power range of 3–200 µW, which is an affordable and cost-effective source available in the market. A relatively high intensity (more than two orders of magnitude higher than that of the photo-resists) is required because the N-ER is less sensitive to near-UV light exposure, as shown in the absorbance data of Fig. 1(b). Therefore, the N-ER is advantageous because it is less sensitive to ambient light and allows stable operation under laser power variations. In this setup, the N-ER was illuminated using a diffraction-limited laser focal spot with a scanning speed of 50 µm/s and a step size of 50 nm. The full width at half maximum of the focused spot was 310 nm; in other words, the radius of the spot size was 260 nm. In general, we scanned the line twice to generate a single narrow line pattern.

 

Fig. 1. (a) Schematic of the direct laser writing process. (b) Absorbance spectrum for ma-N 2405. (inset) Semi-logarithmic plot of absorbance. (c) AFM image of array patterns fabricated on Au films with periodicities of 1 µm (top) and 600 nm (bottom).

Download Full Size | PPT Slide | PDF

In Fig. 1(c), we show typical Au metal patterns fabricated by direct laser writing followed by the developing and etching processes under a dose condition of D = 3.2 J/cm². Here, D denotes the effective dose with a factor of two increase considering that we scanned twice to generate the single line pattern. The baking temperature was 90 °C. We fabricated one-dimensional array patterns on a gold film with a thickness of 50 nm. After the development process, we used reactive ion etching (RIE) with argon gas for 70 s at a forward power of 350 W and a chamber pressure of 50 mTorr. We demonstrate the atomic force microscope (AFM) images of two arrays with different periodicities (1 µm, top, and 600 nm, bottom) in Fig. 1(c). A linewidth of less than 300 nm was achieved successfully, whereas the rough edge morphology appeared because we used an etching process with Ar bombardment. In other words, the line-edge roughness (LER) of the metal pattern reached 51 nm, whereas that of N-ER was less than 20 nm in general as will be shown later [31]. As a result, we found our DLW technique very useful in the fabrication of optical and electronics devices such as the metamaterials with nanogaps and the field-effective transistors with a narrow channel width [32–34].

To find the optimal condition, we investigated the linewidth and height of the N-ER after the developing processes as a function of the exposed dose. Figure 2(a) shows a series of scanning electron microscope (SEM) images of the N-ER for different dose conditions (D = 7.5–15 J/cm²). We used a silicon substrate; therefore, we needed higher doses (about twice) compared to those used in Fig. 1(c), in which laser lithography was performed on the Au film. We clearly observed narrowing of the linewidth as we decreased the exposed dose. The linewidth increased up to 4 µm in the high-dose condition, but decreased to as low as 150 nm, as shown in the inset of Fig. 2(a), for D = 7.5 J/cm². In other words, we could achieve a linewidth close to or even smaller than the diffraction limit. This is a resolution that is not easily accessible by conventional photoresists with both positive and negative tones. In addition, the LER of N-ER was 14 nm which is very small compared to the linewidth of patterns. In Fig. 2(b), we plotted the linewidth as a function of the dose, extracted from the SEM images (red squares), and compared it with the results obtained by the representative positive photoresist (P-PR) and negative photoresist (N-PR). The P-PR (GXR 601) and N-PR (ma-N 1410) delivered the thinnest linewidths (typically of 550 nm), which were still more than three times larger than the linewidths achieved by the N-ER (and also much more sensitive to light exposure). We note that the N-ER has been developed for the e-beam lithography capable of producing linewidths of less than 50-nm [27]; this enabled us to achieve the higher resolution in DLW than the photoresist cases. Although, as aforementioned [21,22], the photothermal effects enabled resolutions down to 50 nm based on the single-photon absorption on the photoresist, our technique enabled the lift-off process in addition to chemical and physical etching, thereby boosting the versatility for fabricating both positive and negative structures.

 

Fig. 2. (a) SEM images of the N-ER after the developing processes with different doses (D = 15–7.5 J/cm² from left to right). (b) Linewidth versus dose for different e-beam and photo-resists. P-PR (black triangles), N-PR (blue circles), and N-ER (red squares) refer to GXR 601, ma-N 1410, and ma-N 2405, respectively.

Download Full Size | PPT Slide | PDF

Next, we evaluated the height of the patterned N-ER (red squares) and N-PR (blue circles) as a function of dose in Fig. 3(a) for the series of results shown in Fig. 2. We extracted the height by taking AFM images of the photo- and e-beam resists. For both resists, the height decreased together with the linewidth as we reduced the exposed dose. In other words, in high-intensity conditions, the height of the N-ER was almost the same as the initial N-ER film thickness (750 nm). However, the thickness decreased to as low as 80 nm for D = 7.5 J/cm², that is when the linewidth was only 150 nm. This highlights the existence of a trade-off between the linewidth and height of the resist pattern. From the slope of the height-dose curve in Fig. 3(a) we evaluated a lithographic contrast (γ) of 4.8 [35,36], which was approximately 3.4 times higher than that achieved by the N-PR (1.4). We noted that the enhanced lithographic contrast of the N-ER relative to the N-PR is consistent with the enhanced resolution of the former; however, the mechanisms on the photo-induced polymerization responsible for the improved contrast necessitates future investigation. Conversely, sensitivity is defined by the dose required for achieving a thickness of 50%. The calculation yielded 9.3 J/cm² and 120 mJ/cm² for the N-ER and N-PR, respectively. This implies that the N-ER requires approximately 80 times higher doses compared to the N-PR, as mentioned previously.

 

Fig. 3. (a) Resist height versus dose for the two different negative resists of N-PR (blue circles) and N-ER (red squares). The solid lines represent sigmoidal interpolations of the data. (b) Aspect ratio as a function of dose for N-PR and N-ER.

Download Full Size | PPT Slide | PDF

The optimal condition, potentially useful for typical optical and electronic device fabrication, can be found at D = 9.9 J/cm², which corresponds to a linewidth of 330 nm and a height of 470 nm. In this condition, the pattern has the largest aspect ratio (1.4), as illustrated in Fig. 3(b) (red squares). We also note that the highest aspect ratio in the N-ER is approximately 2.5 times higher than that in the N-PR (0.6; blue circles).

Finally, we evaluated the linewidth and height of the N-ER resist as a function of the baking temperature. We varied T_bake from 70 °C to 150 °C as shown in Fig. 4. The developing times also varied from 70 s (for 70 °C) to 90 s (for 150 °C). The data for T_bake = 90 °C are reported in Figs. 2 and 3. As shown in Fig. 4(a), the linewidth-dose curves did not change noticeably with T_bake from 70 °C to 110 °C, whereas they deviated significantly for T_bake = 150 °C, which yielded linewidths as large as 1.3 µm. In addition, we found a similar tendency of the height to increase from 0 to 750 nm as we increased the dose as shown in Fig. 4(b). The contrast was the highest (∼5.0) for T_bake = 70 °C but decreased to as low as 1.0 for T_bake = 150 °C, as shown in the inset of Fig. 4(b). On the other hand, the aspect ratios for different T_bake (Fig. 4(c)) reached ∼1.4 for both the 90 °C and 110 °C cases. Notably, the aspect ratio for T_bake = 70 °C was relatively low (1.1), even though its contrast was the highest (Fig. 4(b)). Therefore, we identified 90 °C as the optimal baking condition, which coincides with the temperature suggested for the e-beam lithography procedure.

 

Fig. 4. (a) Linewidth, (b) height, and (c) aspect ratio as a function of dose for different baking temperatures (70–150 °C). The inset in (b) shows the contrast as a function of the baking temperature.

Download Full Size | PPT Slide | PDF

3. Conclusion

We used a negative-tone e-beam resist for direct laser writing based on a simple single-photon absorption process. Using a laser source at a wavelength of 405 nm, we achieved a linewidth of less than 300 nm on the metal film in low-dose condition. To find the optimal condition, the linewidth and height of the resist after the developing processes were studied as a function of the exposed dose. The linewidth of the N-ER patterns could be as low as 150 nm, that is 3.7 times thinner than those achievable by the N-PR. Conversely, this required a laser power more than 80 times that required for the N-PR. The lithographic contrast of the N-ER reached 4.8 (with a maximum aspect ratio of 1.43) as compared to 1.4 of the N-PR, confirming the superior resolution achievable by the N-ER. Finally, the linewidth, contrast, and aspect ratio of the N-ER were studied as a function of the baking temperature, and an optimal temperature of 90 °C was determined.