1. Introduction

Direct-write laser-induced chemical vapor deposition (LCVD), in which a laser beam (photons) is used to initiate the localized decomposition of precursor molecules either photothermally, photochemically, or by a combination of both, is a well-established add-on technique [1]. Among its many positive attributes, it can pattern materials incompatible with conventional photolithography, reduce processing steps to minimize costs and/or cross-contamination, and fabricate functional structures three-dimensionally [2]. In addition, a wide variety of precursors have been developed for LCVD to deposit materials spanning from metals, semiconductors, to insulators, rendering LCVD a very versatile tool for rapid prototyping [3]. However, some limitations of LCVD do exist. As is true for other add-on writing techniques, LCVD is a serial process that is slow compared to the parallel approach of convectional photolithography. In addition, feature size obtained by LCVD is limited by diffraction to be around half of the wavelength. Higher spatial resolution can be achieved by exciting precursor molecules either at near field [4] or using shorter wavelength. However, the former suffers from low throughput of light whereas the latter requires light source at short wavelength which are not easily available. Another way to achieve higher spatial resolution is to utilize nonlinear process during deposition, such as diffusion and nucleation of radicals on the surface [5], temperature dependence of thermal reaction rate [6], and multiphoton absorption-induced decomposition of precursors [7]. For example, Haight el. al. recently demonstrated direct writing of grating-like planar structures of chromium on various substrates using LCVD employing a 100-femtosecond 400-nm pulsed laser in a line-by-line fashion [7]. Line width as small as 100 nm (corresponding to λ/4, where λ is the vacuum wavelength of the laser light) was obtained as a result of a combination of multiphoton dissociation of precursors and tight focusing of incident light (NA = 0.9) which makes the further reduction of feature size very difficult.

In this article, we report a novel direct-write LCVD process that is capable to generate feature size of λ/5 and to process multiple features simultaneously. By gently focusing a single 400-nm 150-femtosecond laser beam onto the surface of substrates situated in a vacuum chamber at room temperature with tungsten hexacarbonyl [W(CO)₆] as precursors, we observe spontaneous formation of one-dimensional grating-like structures of tungsten that are grown heterogeneously on top of the underlying substrates without subject to any beam shaping such as holography or masks [8]. In sharp contrast to Ref. [7] which employed similar laser system as ours, our nanograting is formed simultaneously and has even smaller feature size. We also demonstrate tungsten nanograting (TNG) with excellent long-range order simply by translating the sample with respect to the laser focus. TNGs are observed on top of a wide range of insulators, semiconductors, and metals, suggesting it could be a universal phenomenon. Considering the universality of our observation and the choice of materials that can be deposited using LCVD, our finding could provide a flexible, simple, and cost-effective means to pattern periodic structures.

2. Experimental

 

Fig. 1. Schematic drawing of the experimental setup for the LCVD system. See texts for details. The inset illustrates nanogratings induced by a single 400-nm fs laser beam without any beam shaping.

Download Full Size | PDF

Experiments were carried out in a home built LCVD system as shown in Fig. 1. The tungsten precursor used is tungsten hexacarbonyl [W(CO)₆] (Aldrich, 99.9+%). Ultra high purity nitrogen gas was used as a carrier gas to transfer the precursor from a reservoir into a reaction chamber pumped by a mechanical pump. The reaction chamber was mounted on a piezoelectric stage and had a glass viewport through which a laser beam was introduced into the chamber. The laser source is a 400-nm linearly-polarized femtosecond (fs) laser generated by frequency doubling the 800-nm, 90-fs pulses from a mode-locked 80-MHz Ti:sapphire oscillator (Kapteyn-Murnane Labs), and is delivered to the reaction chamber via multiple wavelength-selective (high reflection at 400nm and high transmission at 800 nm) mirrors to effectively filter out the fundamental beam. The laser beam was focused by a 0.55-NA microscope objective, through the glass viewport, and incident normally onto the substrate surface. The objective is underfilled (effective NA ∼ 0.3) and the full width at half maximum (FWHM) spot size at focal point is calculated to be about 0.8 µm in diameter; actual spot size on substrate surface was larger due to aberration caused by the 1-mm-thick glass viewport. Laser power was adjusted using a neutral-density filter and values quoted here are corrected for all losses before substrates. Laser polarization was changed either using a half wave plate to rotate its linear polarization state or a quarter wave plate for circular polarization. Two growth modes were employed in this study: in the stationary mode, the substrate is exposed to the laser beam for a certain time administrated by a mechanical shutter; and in the scanning mode, the substrate was translated with respect to the laser beam at a desired scanning speed and direction. Both exposure time and stage scanning were controlled and synchronized by a computer. Different substrates were used in the experiments, including c-plane sapphire [Al₂O₃(0001)], glass, z-cut quartz, calcium fluoride (111), gallium nitride (0001), gold, copper, and palladium. Substrates were ultrasonically cleaned using acetone and methanol for 10 min each, followed by blow dry with dry nitrogen gas before they were loaded into the reaction chamber. Samples were investigated by scanning electron microscopy (SEM) using a Raith 150 system. Contact mode atomic force microscopy (AFM) images were taken with a Veeco NanoMan system. Energy-dispersive X-ray spectroscopy (EDX) spectra were collected on an Oxford Instruments INCA microanalysis system equipped on a JEOL JSM 6480 system.

3. Results and discussion

The majority of characterization is carried out on sapphire substrates. Figure 2(a) shows a SEM image of a typical TNG under the stationary growth mode with a laser power of 19 mW and an exposure time of 4 sec, and Fig. 2(b) shows an AFM image of a typical TNG grown under similar conditions along with its cross section profile. Each tooth in the grating resembles willow leaves with long strips and pointed ends [Fig. 2(a)]. The orientation of the grating teeth was found always parallel to the laser polarization; when the input linear polarization was rotated, the orientation of nanogratings rotated accordingly. Control experiments involving circular polarized light produce no gratings except random deposits. The AFM profile [Fig. 2(b)] indicates the height of the grating teeth, maximally around 20-30 nm, decreases monotonically from the center to the edge, consistent with the Gaussian intensity profile of the focused laser beam. The grating teeth have maximum linewidth about 80 nm (λ/5) and average periodicity about 160 (λ/2.5) -180 (λ/2.2) nm. AFM profile also indicates clearly there is negligible background deposition. Also noticeable in the AFM image is a shallow ring (< 5 nm) encompassing TNG which could be tungsten deposition due to the first diffraction maximum of the focused laser beam [9]. We do not, however, observe such ring pattern on other substrates.

In scanning growth mode, the substrates were translated with respect to the laser beam and this resulted in TNGs with excellent long-range spatial order. By controlling the laser polarization with respect to substrate translation direction, long strip of TNGs with various grating teeth orientation can be obtained, as exemplified in Figs. 2(c) and 2(d) which represent a longitudinal and transverse grating, respectively. Both gratings were grown with a laser power of 20 mW and a scanning speed of 0.4μm/sec. The small angle (∼ 3°) deviation from perfect parallel or perpendicular alignment between the scanning direction (indicated by the black arrow) and the laser polarization (indicated by the white arrow) was mainly due to coupling between orthogonal axes in the piezoelectric stage. Since multiple lines were formed by a single laser beam at one time, the nominally serial nature of direct-write LCVD turns into a parallel approach without adding any instrumental complexity. Such unique property offers capability to fabricate unique-shaped periodic structures in a single step. Fig. 2(e) demonstrates a TNG with linear teeth embedded in a ring pattern produced with a laser power of 21 mW by translating the substrate at a speed of 1.0 μm/sec in an annular fashion while the laser polarization was held linearly and constant. The length of grating teeth is short when the scan direction is perpendicular to the laser polarization and becomes longer as the scan direction turns parallel to the laser polarization. It is interesting to point out that the grating teeth at the northwest and southeast portions of the circle fall on the same line, as indicated by the dashed lines in Fig. 2(e). In another word, the spatial coherence between two physically separate TNGs could be preserved in they are connected through a third party. Such unique property could be utilized to generate large-area gratings. The defect at the top corresponds to where the fabrication starts and ends, and is due to the hysteresis of the translational stage which is expected to be corrected if a close-looped stage is used.

 

Fig. 2. Tungsten nanogratings on sapphire substrate: (a) SEM image of a typical nanograting grown under stationary growth mode. (b) AFM image and its cross section profile of a typical nanograting grown under similar conditions. (c-e) SEM images of different grating patterns grown under scanning growth mode: (c) longitudinal, (d) transverse, and (e) circular grating pattern. The white and the black arrows indicate the laser polarization and the scanning direction, respectively. Black dashed lines in (e) are for illustration purpose only and please refer to text for details.

Download Full Size | PDF

The AFM profile [Fig. 2(b)] clearly shows the nanograting structure is grown on top of the substrate surface. Control experiments using similar laser exposure without precursors produce no detectable patterned surface structures, indicating the nanograting is due to laser-induced deposition rather than surface roughening. In order to determine the composition of the deposited nanostructure, EDX study was carried out on bare sapphire, TNGs on sapphire, and tungsten microdots on sapphire (obtained by overexposure for 2 minutes). Samples were sputtering coated with a ∼5.5 nm thick layer of Au before the analysis to eliminate the charge effect. All spectra were taken with electron beam energy of 20 KeV and a working distance of 10mm. The EDX spectra along with insets of corresponding SEM images are shown in Figs. 3(a)-3(c). Figure 3(a) clearly indicates background peaks of O, Mg and Al from the substrate and Au from the coating. Figure 3(b) shows, in addition to those above-mentioned peaks, has a small yet clear signature of W. For comparison, these two spectra were taken at the same SEM magnification of 45K. To further verify this additional peak is due to W, EDX was also carried out on a micron size dot with a deposition time of ∼2 min [Fig. 3(c)] at a SEM magnification of 20K. Much strong W peaks are observed, confirming the extra peak in Fig. 3(b) is indeed tungsten. A small carbon peak was also found in this case and is due to the formation of coordinatively unsaturated metal carbonyls and the dissociation of CO, which is commonly seen in photolytic LCVD of metal carbonyls [10]. The weak W peak in Fig. 3(b) is mainly due to the fact that the thickness of teeth in TNGs are only about of 20-30 nm as indicated from AFM study, and that TNGs only cover partial portion of the detection area during EDX measurement. The W peak, although very weak, is still stronger than the Au peak which arises from a 5.5-nm-thick continuous Au layer. Comparison of these EDX spectra indicates the dominate presence of tungsten in these nanostructures and thus confirms the deposition of tungsten. The above data provide strong evidence that the observed surface pattern is a surface-relief nanograting formed by the deposition of tungsten.

 

Fig. 3. EDX spectra on (a) bare sapphire substrate, (b) on tungsten nanogratings, and (c) on tungsten micron-sized dot. Insets are corresponding SEM images.

Download Full Size | PDF

The dependence of morphology of TNGs on exposure conditions were studied systematically in order to gain insight of TNG formation process. Under stationary growth mode, the morphology of TNGs on sapphire depends strongly on laser power and exposure time. A laser power threshold around 17 mW was found, below which no TNGs was observed even for an exposure time more than 10 sec. For laser power equal to or higher than 17 mW, TNGs were well established in 2 sec [Fig. 4(a)]. At constant laser power of 19 mW, five teeth were found in TNGs for exposure time between 2 to 6 second [Figs. 4(a)-4(c)], during which the length of the grating teeth increased steadily with increasing exposure time before a flake-like structure appeared. These flakes were much easier to be scratched away by AFM tips than individual teeth of TNG. Continued exposure for several minutes led to a tungsten dot with diameter of several microns [see Fig. 3(c)]. At constant exposure time of 4 sec, beside teeth length, the number of teeth also increased steadily as power increased: from slightly more than 3 teeth at 17 mW [Fig. 4(d)], 5 at 19 mW [Fig. 4(b)], 6 at 22 mW, to almost 7 teeth at 24 mW [Figs. 4(e)-4(f). Flakes appeared at 22 mW and grew much more at 24 mW [Figs. 4(e)-4(f). Similar trend was observed for TNGs grown under scanning growth mode, as shown in Fig. 5 for longitudinal grating and Fig. 6 for transverse grating. A power threshold of 17 mW was also found. At low laser power and/or shorter dwell time (e.g. faster scanning speed), TNGs were clean (lower and left corner on Figs. 5 and 6). As laser power increased or dwell time became longer, more parallel lines or longer teeth were developed for longitudinal grating or transverse gratings, respectively. At even higher laser power and longer dwell time, TNGs were, again, covered with flakes (upper-right corner in Figs. 5 and 6). One noticeable difference between longitudinal and transverse gratings is that the former became broken and disconnected at scanning speed of 0.8 μm/sec (last row in Fig. 5) and above (not shown), whereas transverse TNGs stayed ordered for scanning speed up to 1.0 μm/sec [Figs. 6(j)-6(l)]. This difference is most likely due to the nature of longitudinal gratings: to form longitudinal grating require the teeth of TNGs connected along their longer axes as the substrate scans across the laser beam. If the displacement is longer than the grating teeth, then the teeth developed over adjacent exposure area won’t be connected and appear broken.

 

Fig. 4. Top: SEM images of tungsten nanogratings on sapphire grown with laser power of 19 mW at exposure time of (a) 2 sec, (b) 4 sec, and (c) 6 sec. Bottom: nanogratings grown with the same exposure time of 4 sec at laser power of (d) 17 mW, (e) 22 mW, and (f) 24 mW. The scale bar applies to all images.

Download Full Size | PDF

 

Fig. 5. SEM images of longitudinal Tungsten gratings on sapphire substrates grown at various laser powers and scanning speeds. The scale bar is 500nm and applies to all images. Please refer to text for details.

Download Full Size | PDF

 

Fig. 6. SEM images of transverse Tungsten gratings on sapphire substrates grown at different laser powers and scanning speeds. The scale bar is 500nm, and applies to all images. Please refer to text for details.

Download Full Size | PDF

Similar TNGs were also observed on substrates other than sapphire, including insulators (glass, fused and crystalline quartz, and calcium fluoride), wide bandgap semiconductor (gallium nitride), and metals (gold, copper, and palladium). Some of these results are shown in Fig. 7. These materials have large difference in their material properties such as dielectric constant, light absorption, and electronic bandgap etc. Although variation in growth conditions (laser power and exposure time/scanning speed) and morphology exist, the fact that TNG is present in all these substrates suggests it could be a universal phenomenon. To this end, the TNG reported here is somewhat in resemblance to laser-induced periodic surface structure (LIPSS [11]) during LCVD of metal film [10, 12-14] but with noticeable differences. In LCVD-LIPSS, a deep ultra-violet light (244 nm) was used to induce deposition of metallic film on dielectric substrates. For film thickness larger than hundreds of nanometers, the surface of the film exhibited periodic structures with varied periods (∼ λ/1.2 to λ/1.7) and has an orientation perpendicular to the laser polarization [12]. Such periodic features are attributed to the stimulated growth of a particular spatial component of small surface roughness that is selected by the phase-matching condition during light scattering at the metal surface. The scattering of the incident light via the surface roughness generates surface plasmon wave (SPW), which interferes with the incident light and produces standing wave pattern on the surface. Such surface standing wave induces periodic deposition of metal that serves as grating components to scatter more incident light into SPW. A positive feedback loop is thus established between the initial surface roughness and SPW, resulting in the spontaneous formation of surface periodic structures [12]. Although this picture can account for the periodic nature of our TNGs, it fails to explain some of their key attributes: TNGs grow heterogeneously on all substrates without an underlying thick metal film and has an orientation parallel to laser polarization, both of which suggest SPW may not be involved during the formation of TNGs. Therefore, it is likely that TNGs are resulted from the stimulated growth of tungsten induced by the interference between the incident light and its scattered waves on or near the surface. The nature of the scattered waves and its generation mechanism, however, are still unknown and deserve more in-depth study in the near future.

 

Fig. 7. SEM images of tungsten nanogratings on (a) glass with laser power 22 mW and scanning speed 0.2 μm/sec, (b) gallium nitride with 26 mW and 0.8 μm/sec, and (c) gold with 22 mW and exposure time 1s. The scale bar applies to all images.

Download Full Size | PDF

Our preliminary data indicates the threshold of TNG formation on glass and quartz is similar to that on sapphire. We also measured the transmission of 400-nm light through sapphire substrates under similar focusing conditions as in the actual experiments but without the presence of precursors. Negligible absorption in sapphire was found with laser focused either on or above the substrates, indicating very little substrate heating if any. These facts suggest the photodissociation of tungsten precursors is most likely induced by direct absorption of incident photons rather than by photo-excited hot carriers in the substrates. Since the total dissociation energy of W(CO)₆ to produce ground-state W is around 11 eV [15], it is expected that such reaction will require at least four photons at 400 nm (3.1 eV) to complete. The nature of the precursor decomposition is therefore nonlinear in laser intensity. This is also consistent with the observation that the formation of TNGs was found very sensitive to the laser focusing condition. Giving the same laser input power and exposure time/scanning speed, different morphology could be observed if the laser focusing was different from run to run, or even in the same run had the surface height with respect to laser focus varied considerably during scanning.

After the TNGs are developed on the substrate surface, continued exposure of light results in flakes on top of the TNGs. The Flakes, on the other hand, grows much faster than the TNGs, as shown by the significant increase in the deposited tungsten by comparing Figs. 4(d)-4(f). This indicates the flakes undergo a different and more efficient growth mechanism from the TNGs. Indeed, the presence of underlying metal surface can significantly increase the rate of photochemistry by providing other channels for photodissociation of precursors than direct photoexcitation. Not like insulators or semiconductors, metals can absorb the light significantly due to the large quantities of free carriers. Especially, the short pulse duration of femtosecond pulsed laser enables efficient generation of large density of excited carriers [16] and/or photoelectrons in the substrates [17, 18], which are known to facilitate surface photochemistry via multiple resonant tunneling [19] or electron impact [20]. Alternatively, these hot carriers could results in substantial heating of the substrate electrons near the surface which could enhance thermal dissociation of adsorbates [21]. More studies are needed in order to shed more light into the nature of flakes formation.

Many others techniques exist to fabricate periodic one-dimensional nanostructures. Conventional photolithography has continued to minimize its feature sizes using liquid immersion technique in which a high-index fluid is inserted between the lens and photoresist to increase the effective refractive index of medium and hence smaller wavelength of light. For example, Raub et. al. demonstrate 45-nm half-pitch lines generated by 213-nm light using NA = 1.18 with water as immersion fluid (n = 1.42) [22]. When normalized to the vacuum wavelength, the feature size (corresponding to λ/4.7) and period of the grating (λ/2.4) as demonstrated are comparable to our results. Although photolithography has potential for massive parallel processing, this method requires many processing steps such as deposition, spin coating, exposure and etching etc, many of which involve harsh temperature and/or chemistry. Other direct add-on techniques such as e-beam lithography [23] and scanning probe-based methods [24], although providing very fine spatial resolution, are truly serial process in nature. Although processing using multiple probes has been demonstrated [25], it increases system complexity significantly. Our method, on the other hand, generates multiple parallel lines with subwavelength resolution simultaneously using single gently-focused laser beam (effective NA ∼ 0.3). If successfully applied to other precursor systems, our approach could offer a simple and cost-effective technique for rapid prototyping of optical gratings for applications in various areas of photonics such as subwavelength metallic-grid polarizer, grating-coupled waveguide [26], and enhanced light extraction of LED [27], just to name a few. Metal or semiconductor gratings could also serve as ordered catalysts for the growth of one-dimensional nanostructures. In addition to rapid prototyping, our uniquely nanostructured surface could find other interesting applications in sensing. For example, sharp metallic tips such as the apexes of the TNGs are expected to enable strong field enhancement as optical antenna for enhanced sensing and optical nonlinearity [28]. The flakes significantly increases the effective surface area of the substrates and could be used to enhance catalytic reaction. It is interesting to point out the possibility of larger area patterning of TNGs or flakes beyond what is presented in this work. As suggested by Figs. 4(d)-4(f), higher laser power leads to longer grating teeth and flakes. If higher power laser is available, it is possible to use a low-NA objective and proper beam shaping to produce a large flat-top laser beam to pattern large area of either clean TNGs or flakes, depending on the total laser power.

4. Conclusion

In summary, we have demonstrated a novel LCVD growth of tungsten nanogratings induced by a single gently-focused linearly-polarized femtosecond laser beam without any masks. AFM and EDX studies confirm they are surface-relief features due to tungsten deposition. In conjunction with linear scanning, our method produces TNGs with excellent long-range order. Continued exposure on the as-formed TNGs leads to rapid growth of flake-like structures. TNGs possess subwavelength feature size (λ/5) and grating period (< λ/2); when normalized to the laser wavelength, these results are comparable or better than what can be achieved using immersion lithography and tightly-focused laser direct writing. We adopt the conventional theory of LIPSS to explain the periodic structures of TNGs, yet the nature of scattered wave is not clear. Our preliminary data suggest that the growth of TNGs is initiated by multiphoton absorption in precursors near the substrate surfaces, whereas the growth of flakes is most likely mediated by photoexcited electrons from the as-deposited tungsten. The combined capabilities of subwavelength resolution and parallel processing make our method a very promising tool for rapid prototyping of optical gratings as well as other uniquely shaped nanostructures (such as sharp tips and flakes) for applications in photonics and sensing.