Browse

• Abstract

SECTION I

## INTRODUCTION

LITHIUM NIOBATE $({\rm LiNbO}_{3})$ is an extensively studied material mainly because of its piezo-electro-optical properties [1]. It is commonly employed in applications involving acoustic and optical signal processing and nowadays it is adopted in the emerging field of acousto-fluidics [2] opening also new opportunities in opto-fluidics [3].

Despite their wide diffusion, commercial devices based on ${\rm LiNbO}_{3}$ rarely take advantage of the great potential offered by surface micromachining. Precise control of relief microstructures shape and roughness could be a technologic breakthrough for a new generation of ${\rm LiNbO}_{3}$ devices.

In particular it is known that the sidewalls roughness affects the power loss of ridge optical waveguides and the liquid flow behaviour through micro/nanofluidic channels while absence of verticality in sidewalls imposes a lower aspect ratio in microreliefs and broader gaps among them thus limiting the integration of different structures. For this reason a mature etching technique of ${\rm LiNbO}_{3}$ should aim to produce vertical walls with optical quality. Many microstructuring technologies have been proposed in the last ten years (Table I).

Dry-etching methods such as RIE or ICP-RIE [4] have slow etching rate and produce quite smooth sidewalls and trapezoidal cross sections. Combining these methods with other techniques, such as proton exchange [5], [6], can significantly improve both smoothness and etching rate allowing higher aspect ratios. Other dry-etching methods are mechanical dicing of planar ${\rm LiNbO}_{3}$ [7] and ultrafast laser machining [8], which allow to obtain trenches as deep as several hundred micrometres. These methods appear to be quick but may have some limitations. In fact, mechanical dicing produces only straight geometries with optically smooth sidewalls whereas ultrafast laser machining allows almost any geometry but can hardly reach a low roughness [7]. Wet-etching methods usually involve chemical etching with HF or ${\rm HF/HNO}_{3}$ solutions. Etching rates, except for the ${\rm Z}^{-}$ face, are negligible in ${\rm LiNbO}_{3}$ due to its strong chemical inertia. Therefore, on Z-cut wafers, surface microstructuring can be achieved by selective chemical etching of ${\rm Z}^{-}$ face patterned by polarity inverted regions [9] or with Cr or Al masking layers [10]. As before the cross sections are trapezoidal. A further electric field-assisted proton exchange step has been proposed [11] to improve the control of the etched region shape.

Other processes, like FIB and ion implantation [12], introduce structural defects in ${\rm LiNbO}_{3}$ so that any exposed surface can be micromachined with significant etching rates. D. M. Gill et al. [13] fabricated 3.5 $\mu{\hbox {m}}$ high ridge waveguides with indented sidewalls damaging with a single high energy/high fluence oxygen implant and etching with a ${\rm HF/HNO}_{3}$ mixture. The cross section of these structures shows an overetch at their base correlated to the irregular vertical damage profile [14], [15].

The maximum chemical selectivity between virgin and implanted ${\rm LiNbO}_{3}$ is reached with amorphization. Therefore, from a practical point of view, production of suitable amorphous regions is essential for ${\rm LiNbO}_{3}$ micromachining. In [14] a new ion implantation-assisted wet-etching technique for very smooth ridge waveguides fabrication was proposed. It is based on a high energy/high fluence Cu implantation performed to produce a fully disordered Lithium Niobate from the surface to the ions end-of-range followed by the amorphized material removal in a HF solution with an etching rate of 100 nm/s. Such procedure is further investigated in this work where three different masking layers, properly designed to stop the impinging ions, are characterized and qualitatively compared in terms of roughness, shape stability and damage resistance under irradiation.

The resulting microstructures were also characterized and, in those showing the best features, optical waveguides were produced by a multi-step ion implantation process [16] in order to explore a possible application of this technology. The optical loss analysis of the ridge waveguides allows to quantitatively evaluate the efficiency of the proposed technique.

SECTION II

## EXPERIMENTALS

### A. Relief Microstructures Fabrication in $LiNbO_{3}$

Three 4${^{\prime\prime}}$ wafers of X-cut congruent Lithium Niobate, optically polished, were covered on the ${\rm X}^{-}$ side by a sputtered 25 nm thick Cr adhesion layer and by a sputtered 50 nm thick Au layer. This Au layer is meant as a growth seed or adhesion layer for the subsequent masking processes.

TABLE I MAIN FEATURES OF STATE-OF-THE-ART LITHIUM NIOBATE ETCHING TECHNIQUES

Three different masking layers were patterned with conventional photolithographic techniques and tested for ${\rm LiNbO}_{3}$ micromachining process based on ion implantation. The first one consisted of a 3 $\mu{\hbox {m}}$ thick galvanically grown Au layer, the second one was a 10 $\mu{\hbox {m}}$ thick AZ 4562 positive photoresist (PR) layer and the third one was a 10 $\mu{\hbox {m}}$ thick SU-8-3005 negative PR layer (SU-8). The masking layers patterning was performed by Oclaro North America Inc., in San Donato Milanese.

Their thicknesses were chosen in order to preserve the underlying ${\rm LiNbO}_{3}$ from the high energy Cu ion induced damage ensuring that the ion end-of-range (simulated by the Binary Collision Monte Carlo program SRIM version 2008.04 [17]) lies inside the masking layer.

The Au and Cr in the unmasked regions were then removed by wet etching in “Gold etch” and in “Chromium etch” (standard etchant solutions by Fujifilm) respectively.

The wafers were processed, when possible, with two steps of implantation-assisted etching in order to obtain 4.6 $\mu{\hbox {m}}$ thick relief structures. The ion implantation was performed at the 1.7 MV Tandetron accelerator of CNR-IMM Laboratory in Bologna.

Each step included a 10 nm thick Al sputtering onto the whole wafer surface to avoid superficial charge accumulation under irradiation, a 5 MeV Cu ion implantation at a fluence of $1\times 10^{15}\ {\hbox {cm}}^{-2}$ to induce the amorphization of almost 2.3 $\mu{\hbox {m}}$ of ${\rm LiNbO}_{3}$ in the unmasked region [14], [15] and two wet etching processes, one in ${\rm H}_{3}{\rm PO}_{4}:{\rm CH}_{3}{\rm COOH}:{\rm HNO}_{3}:{\rm H}_{2}{\rm O}$ (16:1:1:2) and the other in HF 49%wt for one minute to remove Al and amorphous ${\rm LiNbO}_{3}$ respectively.

During the implantation process the wafers were clamped onto a massive copper holder kept below room temperature by liquid nitrogen and oriented perpendicular to the ion beam in order to prevent both excessive heating and shadow effects.

The masking layers were then removed and the wafers were cleaned with a Piranha solution.

The wafers status was monitored after each process phase with an optical microscope. SEM observations of the samples were made before and after the masking layer removal. The heights of the reliefs were measured with an Alphastep profilometer. Finally the structures were sawed and front-end polished for further SEM characterizations.

### B. Optical Waveguides Fabrication

The chips exhibiting the best quality structures underwent a multi-step carbon ion implantation (see Table II and [16] for details) followed by an annealing at 280°C for 30 minutes. As shown in our previous work [16] this process creates an optical waveguide with a step-like enhanced extraordinary refractive index $({\rm n}_{\rm e})$ profile inside the ridges extending from the surface to a depth of 2.75 $\mu{\hbox {m}}$. The plateau value of ${\rm n}_{\rm e}$ is 2.2094 at 632.8nm.

TABLE II MULTI-STEP ION IMPLANTATION RECIPE

During this implantation process the chips were clamped onto a massive copper holder kept below room temperature by icy water and oriented perpendicular to the ion beam in order to prevent both excessive heating and shadow effects.

### C. Optical Waveguides Characterization

Optical loss characterization of ridge waveguides was performed@660 nm using the cut-back method.

Two sets of 6.1 cm long waveguides (six waveguides 7.1 $\mu{\hbox {m}}$ wide and two 5 $\mu{\hbox {m}}$ wide) lying on the same chip were characterized in terms of total power losses for both coupling facets. The chip was then cut in two parts 3 cm long (after the sawed facets were polished) and the new short waveguides were also characterized in terms of total power losses.

A monomode fiber with 4 $\mu{\hbox {m}}$ mode field diameter, controlled by a 3 axis nanopositioning stage, was coupled in an end-fire configuration to inject light into the waveguides. A near field imaging system (infinity corrected 100x objective with NA 0.95, tube, field lens and CCD monochrome camera) mounted on a 3 axis nanopositioning stage was used to detect the light intensity distribution on both fiber and waveguide facets.

Input power (light from fiber) and output power (light from waveguide) were measured integrating the respective intensity pixels map from CCD camera with an estimated uncertainty of ±2%. Only the extraordinary polarization was taken into account since it is the only one transmitted by the waveguide.

The following measurement protocol was performed. First the mean reference power from fiber was measured in a given time interval (one minute). Then, the coupling between the fiber and a given facet (facet 1) of a waveguide was optimized by maximizing the output power. The mean output power from the waveguide was measured and the total loss for the waveguide coupled through the specific facet was obtained by subtracting the waveguide output power form the reference power. This measurement is repeated once again after resetting the chip. The measurements are also performed for the other facet (facet 2) of the waveguide and for each 6.1 cm long waveguide. After cutting the chip and polishing the sawed facets the protocol was repeated for each of the obtained 3 cm long waveguides considering only the facet 1 and 2 of the original 6.1 cm long waveguide.

The propagation loss of a waveguide was calculated by subtracting the total loss for a specific facet of the 6.1 cm long waveguide from the one for the same facet of the corresponding 3 cm long waveguide and by dividing the result by 3.1 cm (the length of the missing part).

The coupling losses variability, mainly due to unavoidable slight differences in fiber-waveguide alignment and non identical facets polishing is the main cause of the uncertainty in the cut-back method. However with the above measurement protocol two total loss measurements for each of the considered facets were performed and a maximum percentage variability of 8% was found.

In order to improve the measurement accuracy two strategies were therefore adopted. The first one, the most important, was to compare, as explained above, the total losses of the long and short waveguides considering the same two facets used in the long waveguides measurements thus reducing the uncertainties related to the surface finishing. The second one was averaging on multiple propagation loss measurements performed on a series of waveguides with identical design.

SECTION III

## RESULTS AND DISCUSSION

TABLE III MAIN FEATURES OF THE MASKING LAYERS
Table III summarizes the main characteristics of the three masking layers.

Galvanically grown Au was initially chosen because of its known high stopping power and the negligible deterioration during the ion bombardment as we verified in other ion implantation experiments. In Fig. 1 the Au top view is reported while in Fig. 2(a) the obtained ${\rm LiNbO}_{3}$ reliefs are shown. As expected the ${\rm LiNbO}_{3}$ top surface maintained the original optical quality of the polished wafer since it was completely preserved by ion implantation damage. However the sidewalls exhibit a large roughness strictly connected to the masking layer morphology (see Fig. 1) faithfully reproduced onto ${\rm LiNbO}_{3}$. The half height discontinuity observable on the sidewalls (Fig. 2(a) and Fig. 2(b)) is probably due to the ions lateral straggling close to the end-of-range during the first implantation step which causes a little indent after the HF etching.

Fig. 1. SEM top view of the Au masking layer showing its surface roughness and edges irregularity.
Fig. 2. a) Lateral view of two ridges obtained with the Au masking layer showing the half-height discontinuity. The ridges height is 4.6 $\mu{\hbox {m}}$. b) Detail of a relief structure corner.

### B. Positive PR Masking Layer

This masking layer, obtained by a standard photolithographic process, has a much lower roughness compared to the Au one, potentially enabling a simple and cost effective micromachining of smooth reliefs. However, as a positive PR gets easily damaged by irradiation, its functionality could be affected by ion implantation. In fact Fig. 3 reports a SEM image of the masked wafer after the first implantation-assisted etching.

Fig. 3. SEM image of a ridge terminal before the removal of the positive PR showing the shrunk PR layer and the slide end.

The PR progressively shrunk under the ion beam producing slide structures on the edges clearly visible especially at the ridge ends. Moreover, a serious and inhomogeneous degradation occurred making the positive PR masking layer not suitable for this specific process.

For this reason the second implantation step was not performed. Fig. 4(a) and Fig. 4(b) show the 2.3 $\mu{\hbox {m}}$ high reliefs after PR removal and wafer cleaning. Despite its low resistance to ion damage this masking layer allowed the fabrication of very smooth ridge structures.

Fig. 4. a) SEM view of a ridge terminal after the PR removal showing the excellent smoothness of the surfaces and the rounded slide end. The ridge height is 2.3 $\mu{\hbox {m}}$. b) A frontal view of the ridge terminals.

As shown in the past sections the main features of the masking layer affecting the transferred pattern quality are its roughness, its shape stability and damage resistance during implantation.

SU-8 [18] appeared to be the best choice in order to obtain structures reproduced in a reliable way onto the crystal with optical grade sidewalls [19]. In fact it combines an improved resistance to irradiation compared to positive PR, vertical and very smooth surfaces and high aspect ratios required for narrow and deep features.

Fig. 5. a) SEM image of a ridge cross section after SU-8 removal. The ridge height is 4.6 $\mu{\hbox {m}}$. b) Tilted image of a ridge. In the inset a detail of its top and lateral surfaces is shown. The optical quality of the two surfaces cannot be distinguished. The observed roughness is due to the 10 nm thick sputtered Au film used for SEM observations.

Both implantation-assisted etching steps were performed with no masking layer degradation. However a remarkable SU-8 volume change induced by the irradiation process has been observed. Nevertheless the resist shrinking resulted to be homogeneous and repeatable on the entire wafer area. In Fig. 5(a) and Fig. 5(b) two of the obtained ridges are shown. Their actual widths (7.1 $\mu{\hbox {m}}$ and 5 $\mu{\hbox {m}}$), compared to the original mask widths (respectively 10 $\mu{\hbox {m}}$ and 7 $\mu{\hbox {m}}$), suggest that a 25 $\div$ 30% linear contraction occurred in that direction. This roughly corresponds to the SU-8 vertical compression estimated by the Alphastep. Rounded corners contour and slight sidewalls asymmetry have to be ascribed to the SU-8 reshaping during implantation. On the other hand the obtained sidewalls are vertical and their optical quality cannot be distinguished from the polished surface one (inset of Fig. 5(b)—the roughness of the surfaces is due only to the 10 nm thick sputtered Au film used for SEM observations). Therefore, this technology may be ideal for integrated optics and acousto/opto-fluidics. In order to prove that, structures like those in Fig. 5(a) and Fig. 5(b) were further processed for waveguiding purpose.

Fig. 6 shows the near field characterization of the waveguide generated in a ridge structure 5 $\mu{\hbox {m}}$ wide. In Fig. 6(a) the first propagating mode is clearly shown together with its intensity map. In Fig. 6(b) the measured vertical profile for the first mode relative intensity is reported. The very good agreement between the measured profile and the result of a simulation (see [16] for details) denotes the very good control of the ${\rm n}_{\rm e}$ tailoring.

Fig. 6. a) Near field image of the output facet for a 5 $\mu{\hbox {m}}$ wide ridge waveguide obtained with the SU-8 masking layer. In the inset the intensity map with the actual scale is reported. b) Vertical profile of the first propagating mode relative intensity for the same waveguide. Simulated and measured profiles are reported.

The cut-back measurement for the fabricated waveguides yielded a mean propagation loss of 0.23 ± 0.09 dB/cm@660 nm. The almost identical propagation loss values for different waveguides reveals the homogeneity and robustness of the technological process over a large area. Also no differences, within the uncertainties, can be appreciated in the propagation losses of the two sets of ridge waveguides.

Though these waveguides were not monomodal@660 nm, their multimodal behavior did not introduce uncertainty in the measure as the optimized output power was always associated with the first propagating mode. Nevertheless further studies on monomodal ridge waveguides will be carried out using a more accurate method like Fabry–Perot in order to complete the analysis especially regarding the effect of waveguide narrowing on scattering losses.

SECTION IV

## CONCLUSIONS

Relief microstructures were successfully manufactured in X-cut ${\rm LiNbO}_{3}$ by ion implantation-assisted wet etching using Cu ions at 5 MeV and a fluence of $1\times 10^{15}\ {\hbox {cm}}^{-2}$, thus validating at higher energies the cut-independent amorphization recipe proposed in [14]. Three masking layers were patterned and tested in these implantation conditions.

Galvanically grown Au demonstrated, as expected, the highest stability under irradiation but, because of its intrinsic roughness, it can be adopted only when the optical quality of sidewalls is not mandatory.

Positive PR, although got seriously damaged by ion implantation, allowed the micromachining of extremely smooth reliefs, suggesting that this PR could be properly employed at lower energies for high quality thinner structures fabrication.

SU-8 negative PR appeared to be a good trade-off between Au and positive PR, combining an improved resistance to ion implantation (compared to positive PR) with smooth surfaces and high aspect ratios. The main drawback of SU-8 is its shape instability under irradiation that must be accounted in mask design to obtain the desired pattern in ${\rm LiNbO}_{3}$. The SU-8 shrinking appeared to be highly repeatable and homogeneous over the whole wafer area. Further analysis has been planned in order to investigate the SU-8 reshaping as a function of process parameters such as ion energy, masking layer thickness, geometry and baking treatments.

Optical waveguides inside the high quality relief structures, obtained with the SU-8 masking layer, were fabricated by tailoring ${\rm LiNbO}_{3}$ ${\rm n}_{\rm e}$ with a multi-step carbon ion implantation process. The low waveguide losses (0.23 ± 0.09 dB/cm)@660 nm denotes the excellent quality of the obtained surfaces, which makes this technology highly promising for integrated optics and acousto/opto-fluidics.

## Footnotes

This work was supported in part by the Operative Program ERDF 2007-2013 of Emilia-Romagna Region-Activity I.1.1.

P. De Nicola, S. Sugliani, G. B. Montanari, A. Menin, and A. Nubile are with the Laboratory of Micro and Submicro Enabling Technologies of Emilia-Romagna region S.c.r.l. (MIST E-R S.c.r.l.), Bologna, I-40129 Italy (e-mail: denicola@bo.imm.cnr.it; sugliani@bo.imm.cnr.it; montanari@bo.imm.cnr.it; menin@bo.imm.cnr.it; nubile@bo.imm.cnr.it).

P. Vergani, A. Meroni, M. Astolfi, M. Borsetto, G. Consonni, and R. Longone are with the Oclaro North America Inc.-San Donato Office, San Donato Milanese (MI) I-20097, Italy (e-mail: Paolo.Vergani@oclaro.com; Andrea.Meroni@oclaro.com; Marco.Astolfi@oclaro.com; Marco.Borsetto@oclaro.com; Guido.Consonni@oclaro.com; Roberto.Longone@oclaro.com).

M. Chiarini is with the CGS S.p.A.-Bologna Office, Bologna, I-40129 Italy, and also with the MIST E-R S.c.r.l., Bologna I-40129, Italy (e-mail: mchiarini@cgspace.it).

M. Bianconi and G. G. Bentini are with the Institute for Microelectronics and Microsystems, National Research Council, (CNR-IMM), Bologna, I-40129 Italy and also with the MIST E-R S.c.r.l., Bologna I-40129, Italy (e-mail: bianconi@bo.imm.cnr.it; bentini@bo.imm.cnr.it).

Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org.

## References

No Data Available

## Cited By

No Data Available

None

## Multimedia

No Data Available
This paper appears in:
No Data Available
Issue Date:
No Data Available
On page(s):
No Data Available
ISSN:
None
INSPEC Accession Number:
None
Digital Object Identifier:
None
Date of Current Version:
No Data Available
Date of Original Publication:
No Data Available