1. Introduction

The optical properties of diamond include a high refractive index and low absorption over a very wide spectral band. It also has impressive thermal, mechanical and chemical properties: a small thermal expansion, extreme hardness and thermal conductivity, and inertness to most chemicals. This makes diamond an interesting candidate for many optical applications in challenging environments, be they in space, high power optics, corrosive environments or cryogenic cooling. In the past few decades etching of diamond by reactive plasma etching has been widely studied; mainly with oxygen plasma and often with small amounts of SF₆ added to improve etch rate or smoothness [1–3]. Little has however been done to control the sidewall angle of the etched diamond structures. In one study [3], small tips of nanocrystalline diamond were fabricated by etching away a silicon oxide mask while etching the diamond in an inductively coupled plasma (ICP). In the present study we use a similar method, but with a thick Al-mask, to demonstrate good control over sidewall angles in chemical vapor deposited diamond of optical quality. The sputter yield of Al is strongly angle dependent, a dependence that appears even more pronounced with oxygen present, allowing the edges of the mask to be etched faster than flat horizontal areas.

Binary antireflective (AR) structures, which suppress reflections in narrow wavelength bands, have previously been demonstrated in diamond [4]. Here we expand this to broadband AR structures [5,6] by etching pyramidal shapes using the developed etch process. Optically, binary subwavelength AR structures behave like a solid film with a refractive index in between those of the two media present. Continuous AR structures, on the other hand, behave like a gradient film and as such can be designed to work over a larger spectral band and for a wider range of incident angles. Such AR surfaces have previously been fabricated in diamond by growing it on wet-etched silicon substrates [7] and by laser ablation [8], neither of which could achieve the high aspect ratios required for very low reflectivity.

Diamond waveguides have been previously demonstrated and may find a number of applications in high power/frequency applications and quantum computing [9]; however, the use of focused ion beam milling, a slow serial process, in the fabrication limits the scalability to large numbers. In this paper we demonstrate that sufficiently smooth beveled edges for coupling visible light into a waveguide can be produced by ICP etching.

2. Fabrication of inclined sidewalls in diamond

2.1 Substrate preparation and masking

Two kinds of diamond samples were used: 300 µm thick solid polycrystalline diamond substrates of optical quality (Element Six Ltd.) and 13-15 µm thick diamond film on top of 200 nm Si₃N₄ and 2 µm SiO₂ on Si (Diamond Materials GmbH). The solid diamond substrates were 1 cm diameter disks, while the diamond film was grown on a 4” wafer and diced into 1x1 cm² squares. The diamond film samples were used for the waveguides as well as for test etching (having Si as a carrier substrate made it easier to break cleanly for cross section images of the etched diamond structures) and the solid diamond pieces were used for the AR structures. Before processing all samples were cleaned in hot piranha solution (H₂O₂ and H₂SO₄) and isopropanol. Three masking layers were deposited in a Von Ardenne CS 730S magnetron sputter system: 1.7 µm Al, 400 nm Si and another 130 nm of Al. 200 nm of nanoimprint polymer (mr-I 8020E, Micro Resist Technology) was spin coated on top of the masking stack and patterned using a soft silicone stamp (polydimethylsiloxane, Wacker Elastosil RT601, cast on a laser patterned master) at 170° C and 10 bar for 300 s in a nanoimprinter (Obducat NIL-6).

To transfer the polymer pattern into the metal layers, the samples were dry etched in an ICP etcher with two chambers from PlasmaTherm. The two chambers allow etching with chlorine to be kept separate from other etching. In this kind of etch equipment the sample bias and plasma density can be controlled separately through the capacitively- and inductively coupled power, respectively. For brevity, these will be called RF-C, for capacitive power, and RF-I, for inductive power, from here on. Al was etched in BCl₃/Cl₂ plasma and Si was etched in SF₆/Ar plasma. The mask stack was etched one layer at a time, using the layer above as etch mask. This allowed a thick Al-mask to be micro structured from a thin nanoimprinted polymer film. Over-etching in the various etch steps also helped remove small defects and allowed for fine tuning of the pattern width. After etching the thick Al-layer, remaining Si was etched away. The final masks had squares of Al, around 2.3 µm wide and 1.7 µm thick, in a square grating with 4 µm period.

2.2 Diamond etching

Diamond was etched in the same plasma system as the mask materials, using O₂/Ar/SF₆ chemistry. To achieve sufficiently high etch rates a high bias power was required. The corresponding strong ion bombardment led to an etch mechanism dominated by ion-surface interactions. This in turn gave rise to two distinct angles of the lower and upper parts of the etched sidewalls (see Fig. 1 ). The lower part had an inclination that grew outward from the mask before it was fully faceted, due to effects such as ions being redirected in glancing hits to the wall, causing grooves to get narrower with depth. As the edge of the mask was fixed during this part of the etching, roughness in the mask gave rise to striations in the diamond sidewall. The angle dependency of the etch rate of Al eventually caused the mask to be fully faceted, and the edge of the mask started receding. A less steep upper slope in the diamond wall developed as diamond was etched while the mask withdrew, the angle determined by the rate of withdrawal of the mask compared to the etch rate in diamond. As the mask receded, the edge was smoothed out, making the upper slope much smoother than the lower.

 

Fig. 1 The progression of etching and the origin of the two angles of inclination of the sidewall. (a) The initial mask profile is vertical. (b) and (c) In early etching, before the mask is fully faceted, the edge of the mask does not move. The sidewall etched in diamond is slightly inclined because some ions are deflected instead of etching the wall. (d) Once the mask is fully faceted the edge starts receding. The angle etched in diamond then depends on the etch rate ratio between the mask material and diamond.

Download Full Size | PDF

In an attempt to produce structures without the lower, rougher part of the wall, the Al mask was exposed to ion bombardment in a pure Ar plasma for ten minutes prior to the diamond etching with the following etch parameters: RF-C: 120 W, RF-I: 800 W, gas flow: 50 standard cubic centimeters per minute (sccm) of Ar, process pressure: 5 mTorr. This produced a mask that was already faceted at the start of diamond etching. The initial mask shape and the effect it had on diamond etching can be seen in Fig. 2 . The parameters for the diamond etching of the samples seen in Figs. 2(d) and 2(e) were as follows; RF-C: 220 W, RF-I: 900 W, gas flows: 20 sccm Ar, 40 sccm O₂,and 1 sccm SF6, pressure: 5 mTorr. The pre-etched mask started receding nearly from the start of diamond etching, causing only a small steep portion to be formed at the base of the slope. Plasma etching with pure Ar (i.e. physical etching) gave rise to less steep faceting in Al compared to when O₂ was present in the plasma (i.e. combination of physical and chemical etching). This caused the mask to be thinner and it receded faster during diamond etching, resulting in a slightly more inclined upper slope in diamond.

 

Fig. 2 SEM micrographs showing the pre-etching of the mask and its effect on the diamond etching. (a) Al mask before, (b) after five minutes and (c) after ten minutes of etching in pure Ar plasma. (d) and (e) show pyramidal structures in diamond etched with masks like those shown in (a) and (c) respectively. The sample in (e) is the same as that shown in cross section in Fig. 3(b). The sample is tilted 30° in all micrographs and the magnification is consistent between them.

Download Full Size | PDF

Test etching of pyramid like diamond structures was carried out in order to see how much the angles could be varied in an O₂/Ar/SF₆ plasma. After etching, the samples were cracked and imaged edge-on in a scanning electron microscope (SEM) to measure the angles of the etched pyramids. The results of test etchings on samples pre-etched with Ar can be seen in Fig. 3 . The angle of the lower part was found to depend mainly on the bias power during etching, and could be varied from around 2° at 220 W to 5° at 100 W. The height of the lower part was reduced with higher SF₆ content, nearly disappearing altogether at the highest concentrations. The angle of the upper part could most easily be controlled by changing the gas mixture, with lower SF₆ concentrations leading to steeper angles. A higher bias power also led to a somewhat steeper upper slope and, as mentioned above, a thinner and more inclined mask led to a somewhat shallower slope. In the process regime investigated we were able to produce angles between 13° and 31° from vertical for the upper part. At the lowest SF₆ concentrations, redeposition of forward sputtered Al from the mask would stick to the top of the opposite wall and gave rise to a more rough and striated lower wall. Fissures also often formed in the upper part with these etch recipes (Fig. 3(d) and 3(h)). The mass flow controllers in our ICP equipment could not control flows below 1 sccm with good precision, so for the lowest concentrations we instead controlled the SF₆ ratio by using a 1 sccm flow and turning the flow off and on in a cycle during etching. We also tried varying the pressure between 5 and 15 mTorr with very little effect on the angles, the main difference was a much slower etch rate but also somewhat smoother surfaces at higher pressure.

 

Fig. 3 Test etchings for controlling the sidewall angle. The Al masks were pre-etched with Ar, such as that shown in Fig. 2(c). Flows of O₂ (40 sccm) and Ar (20 sccm) as well as pressure (5 mTorr) and ICP power (900 W) were the same for all the etchings in the figure. The samples in the upper row (a-d) were etched with a bias power of 220 W and in the lower (e-h) with a power of 100 W. The SF₆ flow was 5 sccm in (a) and (e) and 1 sccm in (b) and (f). In (c) and (g) the SF₆ flow was 1 sccm for 2 s, then turned off for 8 s. Finally in (d) and (h) the flow was 1 sccm for 2 s and then off for 28 s.

Download Full Size | PDF

3. Optical applications

Here we demonstrate two practical applications of the developed etch process described above: a broadband diamond antireflective surface in the mid- to far infrared wavelength regime, and diamond waveguides with inclined sides for enabling light injection, from above, into and out of the waveguide.

3.1 Broadband antireflective structure

With the knowledge of what shapes could be reliably produced, rigorous coupled wave analysis (RCWA) with Pavel Kwiecien’s rcwa-2d code for MATLAB (sourceforge.net/projects/rcwa-2d/), based on [10], was used to design pyramid like structures for reducing unwanted surface reflection. The structures were approximated as square pyramids with a steeper lower part and a shallower upper part, similar to the famous Bent Pyramid in Egypt. A grating period of 4 µm was chosen. For an AR structure not to disturb the incoming optical wavefront, only the zeroth diffraction order should be allowed to propagate [6]. Whether a diffraction order propagates or not can be determined by the grating equation. In the case of diamond in air and the condition that only the zeroth order be allowed to propagate, the following inequality can be derived: λ > (n + sinθ) Λ, where λ is the illuminating wavelength, n is the refractive index of diamond (around 2.38 in the mid-IR), Λ is the grating period and θ is the maximum incidence angle allowed. With a grating period of 4 µm, this is fulfilled for wavelengths above 10 µm at normal incidence and for wavelengths over 14 µm even when the incidence angle is high. At shorter wavelengths higher diffraction orders propagate, disturbing the wave front and causing the transmission to drop dramatically. We chose to focus on a wavelength regime between 10 and 50 µm, which falls between this zeroth order limit and the longest wavelength measurable in our spectrophotometer.

The RCWA simulation showed that with a grating depth of 14 µm and normal incidence a transmission over 98.8% could be reached for the whole wavelength regime investigated, with an average transmission of 99.45%. At higher incidence angles, average transmission remained better than 95% up to 60° incidence angle, with transmission only falling below 90% at the lowest wavelengths (for which higher diffraction orders may propagate at that incidence angle). For wavelengths longer than 50 µm, transmission falls off smoothly, reaching 90% at 100 µm for normal incidence. To reach these high simulated transmission values it was preferable to have no gap between the bases of pyramids and having the top part as steep as we could make it.

To achieve these structures experimentally in diamond a two-step etch process was used without Ar pre-etching of the Al mask. First we employed a step with low bias and no SF₆,to form a deep lower portion that joins at the base. This etch step also faceted the Al mask and left some redeposited mask material on the upper part of the sidewall. A second step with a higher bias and a moderately low SF₆ ratio was used to slowly etch away the mask and redeposited material from the first step until a sharp peak was formed. The masking procedure was the same as described in section 2.1 above. A longer Si etch step was employed to produce somewhat smaller squares than for the test etchings. The full diamond etch parameters were as follows. For the first step: RF-C: 90 W, RF-I: 750 W, 30 sccm Ar, 30 sccm O₂, 10 mTorr, 90 min (initialized with 3 sccm SF₆ for 15 s to avoid microvilli formation). For the second step: RF-C: 220 W, RF-I: 900 W, 20 sccm Ar, 40 sccm O₂, 1 sccm SF₆ in a 3/11 duty cycle (3 s on, 8 s off), 15 mTorr, 44 min. The resulting structures, after cleaning off remaining redeposited material in piranha and HF:HNO₃ solution, can be seen in Fig. 4(a) . The angle of the sides was 5° from vertical for the lower part and 16° for the upper. A slightly lower etch rate for the redeposited material than for the original mask gave rise to a small (around 100 nm wide) step below the upper part. The tip width, determined by how precisely the etch process was stopped as the last of the mask was etched away, was as low as 50 nm on some samples. This could certainly be improved with more careful etching, but our simulations showed that a very small tip radius was not critical for good optical performance. The pyramids came together nicely at the base, which was 1-1.5 µm deeper at the corners than in between.

 

Fig. 4 Broadband AR structures in diamond. (a) SEM micrographs of the etched structures, tilted 30° and side view of cracked sample. (b) Measured and calculated transmission of a sample structured on both sides. (c) Pyramid shape approximating the structure on one side of the sample for the RCWA calculation.

Download Full Size | PDF

For measuring the transmission characteristics, AR-structures were etched on both sides of a diamond substrate. A thick (~1.7 µm) layer of Al was sputtered onto the first side after etching, this to protect it while processing the other side. The fabricated AR-structures were not identical on both sides. There were small differences in the tip radius (150 and 300 nm) and in the width of the upper part of the pyramid (2.1 and 1.8 µm). Transmission calculations were carried out individually for approximations of the two sides and then combined for a calculated total transmission. The result of this calculation as well as the actual transmission spectrum (measured with a Perkin Elmer 983 spectrophotometer) is found in Fig. 4(b) and the structure used for calculating the transmission from one side of the sample is seen in Fig. 4(c). The measured transmission spectrum agrees well with the calculated, but is consistently lower, particularly at the shortest wavelengths, possibly due to scattering on imperfections in the pattern. In the range between 10 and 50 µm the lowest transmission value was above 93%, corresponding to a 3.6% reflection per side. The average transmission for the whole range was 96.4% (1.8% reflection per side). For comparison, a smooth diamond surface reflects 17% over the evaluated wavelength interval, leading to a transmission of 71% for a two-sided sample.

3.2 Coupling light through waveguides

Free-hanging waveguides were fabricated to demonstrate the coupling of light through diamond waveguides. Since the light is coupled from the top into (and out of) the waveguide inclined surfaces were needed to ensure total internal reflection (TIR). A more inclined side is advantageous as it gives a wider target for the incoming light. Diamond has a refractive index between 2.38 and 2.46 in the visible-IR regime which means that the critical angle for TIR is about 25°. If the sidewall has an inclination between 12.5° and 50° from vertical, light can be coupled into the waveguide from both below and above. The top level masking on our waveguide structures was achieved by standard UV-lithography using Shipley S1813 photoresist, but otherwise the preparation was as for the AR-structures. Diamond etching was performed with the same parameters as for the sample in Fig. 3(b): RF-C: 220 W, RF-I: 900 W, 20 sccm Ar, 40 sccm O₂, 1 sccm SF₆, 5 mTorr. Etching was continued, even after the diamond layer was etched through, for as long as the mask lasted. This to produce an upper slope as wide and smooth as possible. In total the diamond was etched for 80 minutes. The rougher, steeper lower part of the sidewall was almost completely removed, as can be seen in Fig. 5 . The silicon substrate was etched through from the back, using standard deep reactive ion etching with Al masking to leave enough silicon for easy handling. Finally, hydrofluoric acid was used to remove the exposed oxide and nitride layers from the diamond. The coupling of light into and out of the waveguides was tested using HeNe laser in a Raman spectrometer microscope (NT-MDT NTEGRA Spectra), emitting light at 632.8 nm (Fig. 5). Up to 250 µm wide waveguides were used, as wider ones would not fit in the field of view. As can be seen in Fig. 5, the etching was smooth enough with the light still concentrated in a small spot when exiting the waveguide.

 

Fig. 5 Waveguides. Left: SEM micrograph of the end of a 100 µm wide free-hanging waveguide. Upper right: optical micrograph of laser light going through a 250 µm wide waveguide. Lower right: schematic showing the dimensions of the waveguide.

Download Full Size | PDF

4. Discussion and conclusions

We have demonstrated a simple way of controlling the sidewall angle when dry etching diamond. Angles between 13 and 31° have been achieved. The etch process uses an etch mask with a square profile as its starting point; further, since the mask is etched in several steps, the thickness of the initial polymer mask is not critical. Thus the etch process is suitable for use with most patterning methods. We have used nanoimprint lithography and photolithography, which allow for rapid patterning of the entire sample. All other process steps used are also suitable for batch processing. A limitation of the described etch process is that the achievable width of the sloped sidewall depends on the thickness of the Al mask.

We successfully fabricated two optical structures to demonstrate the usefulness of the etch process: broadband diamond antireflective structures and smooth angled surfaces for coupling light into and out of diamond waveguides. The antireflective structures drastically reduced reflection over a very wide wavelength band in the mid- to far infrared (10-50 µm). Even when working with narrow wavelength bands, pyramidal structures have the advantage over binary AR gratings since their performance is less sensitive to variations in etch depth and angle of incidence. The beveled edges of the waveguide also performed well and should prove useful for coupling visible and IR light into and out of diamond waveguides.

It is worth noting that the possibility to manufacture sharp tips in diamond might be useful in other areas than optics. They might for example be useful as tips for scanning probe microscopy [11].