2.1.

Design

Fig. 1(a) shows the layout of the deformable MEMS grating. The optical grating is formed by silicon beams etched in the device layer of an SOI wafer. The grating is designed with a duty-cycle of 50% and different periods, Λ (6 µm, 9 µm and 12 µm). The grating beams, connected one to the other through grating springs, form a suspended structure that is anchored to the substrate via the oxide layer in 6 different points: the grating is held at either end at two points through flexible suspension beams which allow for a stretching motion of the grating; the grating is rigidly held in the middle by the central beam which is attached to the central anchor pads. The grating springs design is that of a leaf spring (Fig. 1b). During actuation these are the only elements where stress is concentrated and, thus, where deformation is present. They do not allow torsional movement to the grating beams around their longitudinal axes. The grating is stretched by two comb actuators. The device layer thickness is 10 µm. This guarantees sufficient stiffness to the wide suspended structure to prevent stiction during the process release step and to prevent buckling during operation.

Fig. 1.

(a) Overview of the deformable MEMS grating structure: it is possible to distinguish the beams forming the grating (inside the dotted area), the 6 anchor points at which the suspended structure is fixed and the comb actuators on both grating sides. The grating dimensions are on the order of 200 µm x 200 µm. (b) Detail of the deformable MEMS grating, in particular of the leaf hinge structures. The leaf hinge is formed by a thin flexible beam with two wide structures at its extremities: this structure provides the required spring constant without inducing any stress on the adjacent beams.

This design requires only one mask level for the definition of the grating, the springs and the actuators.

Finite element modeling (CoventorWare^TM) was used to simulate the in-plane grating motion and the resonant modes of the structure. From this modeling, the first resonant mode for a structure with a 9 µm periodicity is expected at 19.5 kHz and it is in-plane; the second and third modes are expected at 40 kHz and 45 kHz and are out-of-plane modes.

Fig. 2(a) shows the simulation of the first (in-plane) resonant mode seen from above. The relative motion of the beams can be seen as a cross section in Fig.s 2(b) and 2(c) which show the first and second (out-of-plane) resonant modes respectively.

Fig. 2.

(a) First (in-plane) resonant mode for the deformable MEMS grating with periodicity of 9 µm. (b) Cross-section A-A’ of the structure in Fig. 2(a) for the first (in-plane) resonant mode. (c) Cross-section A-A’ of the structure in Fig. 2(a) for the second (out-of-plane) resonant mode.

The simulations for the structure with a 12 µm periodicity give the first three resonant modes at 28.5, 60 and 68.5 kHz, respectively. As for the 9 µm period structure, the first mode is an in-plane mode, while the second and third ones are out-of-plane. The high frequencies of the resonant modes are obtained because of the particular structure design (in particular the spring region) and to the thickness of the device layer (10 µm).

2.2.

Fabrication

Fig. 3 shows the process flow for planar deformable MEMS gratings. The process starts with a low resistivity SOI (Silicon On Insulator) wafer (Fig. 3a) with a 2 µm buried oxide layer and a p++ 10 µm device layer. This thickness ensures that the structure is stiff enough to prevent stiction during the final release step. Photolithography (Fig. 3b) with 1 µm minimum feature size defines structures. It is followed by deep reactive ion etching (Fig. 3c) to define in one step grating beams, springs and comb actuators. A well controlled lithography and deep reactive ion etching are required in order to guarantee that all flexible structures have the same spring constant. The accuracy of the positioning of each grating beam depends on such a control. The device is then released by removing the sacrificial oxide in HF vapor (Fig. 3d) [10]. A final coating step can be added (e.g. a gold coating) to give the required reflection properties for any specific spectral region.

Fig. 3.

Process flow for the deformable MEMS grating fabrication.

Gratings with period of 6 µm, 9 µm and 12 µm were fabricated on the same wafer. Fig. 4(a) and 4(b) show the top view of an entire device and a cross-section of the spring region, respectively, for a device with 6 µm period on a Silicon test wafer.

Fig. 4.

Fabricated devices on a Si test wafer. (a) Top View of an entire device (optical microscope). (b) Cross-section SEM image of the spring region for one test structure with 6 µm period.

2.3.

Mechanical characterization

The deformable gratings were tested by measuring the displacement as a function of the applied voltage and by measuring the resonant frequencies of the structure using a Doppler Interferometer. The displacement was measured in a probe station by applying a voltage to the comb actuators. With 100 V applied each end of the grating moves by 3.1 ¼m, resulting in an overall stretching of the structure of 6.2 ¼m. This is equivalent to a relative tuning of the grating period ΔΛ/Λ of approximately 3% or to a tuning range of 150 nm at the wavelength of interest, 5 μm.

In this measurement, maximum displacement was limited by the high voltage supply: in principle the present structure can have a relative tuning ΔΛ/Λ up to 10%, the limiting factor being the comb dimension.

Fig. 5 shows the measured displacement of the comb actuator versus the applied voltage. Experimentally, a linear behavior of the displacement is found at higher voltages. A quadratic behavior is normally expected from this kind of structure, but FEM simulations of the entire structure agree experiment.

Fig. 5.

Measured (dots) and simulated (solid line) displacement versus applied voltage for the 9 µm period grating. Maximum displacement at each grating end is 3.1 µm. Good agreement is obtained with simulated behavior. A deviation from ideal quadratic behavior (dotted line) is related to the design of the anchor points.

The deviation from the ideal quadratic behavior is related to the mechanical properties of the flexible anchor beams, on which the structure is suspended (Fig. 1). In fact the flexible anchor beams are not free to rotate around the axis that is perpendicular to the suspended structure plane and that passes through the anchor point itself. Fig. 6(a) shows that with the current design, the stress is concentrated in the beam, but not in the anchor support. This linear behavior reduces the range of displacement but could help in simplifying any controlling algorithms for the actuation.

Fig. 6.

(a) Doppler Interferometer measurement of the first and second resonant mode frequencies for the structure with periodicity of 9 µm. They are located at 13.8 kHz and 28.1 kHz, respectively. (b) Experimentally measured resonant mode of the structure when resonating at 28.1 kHz.

If needed, the mechanical constraint can be removed by adding a right-angle in the beam at the anchor support, enabling the full exploitation of the applied voltage.

The frequencies for the first two resonant modes of the structure with periodicity of 9 µm were measured using a Doppler Interferometer that is able to measure the out-of-plane displacement of a structure driven with sinusoidal voltage at a frequency spanning a desired range.

Fig. 6(a) reports the out-of-plane displacement as a function of frequency. Two resonances can be seen, the first at 13.8 kHz (f1) and the second at 28.1 kHz (f2). The first mode should not normally be present in an out-of-plane displacement (Fig. 2a and 2b).

However, a small coupling of the in-plane resonant mode to vertical displacement makes it appear in the Doppler measurement. Fig. 6(b) shows the out-of plane motion of the whole structure for the second resonant mode (28.1 kHz). This corresponds to the second resonant mode of the structure (out-of-plane mode) as simulations have shown (Fig. 2b).

The experimental values for the first two modes (f1=13.8 and f2=28.1 kHz) can be compared with the simulated ones (19.5 and 40 kHz). The relative positions of resonant modes agree, even if a discrepancy exists in the absolute values.

Such discrepancy is probably due to the tolerances in the process that were not taken into account in the simulations and will be further investigated. While the resonant frequencies are lower than expected they are still high enough to allow sub-millisecond tuning of the grating over its full range.

2.4.

Optical characterization

The fabricated structures were tested optically with a He-Ne laser (632.8 nm) in a static condition.

Fig. 7(a) shows the diffraction intensity as a function of angle for a grating with period 12 µm at normal incidence. Good agreement is found between the experimental results and the optical simulations in which the geometrical parameters of the grating are derived from direct measurement: in particular, the grating duty cycle is not 0.5 as designed, but 0.45 (Fig. 7b).

Fig. 7.

(a) Intensity of diffracted orders for the 12 µm period grating at normal incidence at the He-Ne laser wavelength of 632.8 nm. Good agreement is obtained between experiments and optical simulations using data derived from geometrical characterization. (b) Geometrical characterization of the 12 µm period grating (SEM cross-section in a test wafer).

In order to use such a grating to tune external cavity lasers, diffraction efficiency can be improved by blazing the grating beams.

3.1.

Design

For the blazed gratings, the basic layout is the same used for the planar ones. Only the process flow is modified with the introduction of extra fabrication steps and a second mask. The challenge for blazed gratings is to obtain an optically smooth surface. Anisotropic KOH etching is well known for its ability to stop at {1 1 1} Silicon planes, e.g. silicon v-grooves. This technique was chosen, which provides optically smooth surfaces but fixes the blaze angle at 54.74°.

The second mask is introduced in order to apply the steps needed to obtain the blazed features only in the beam region: this region correspond to the dotted region in Fig. 1(a).

A commercial KOH etch simulation tool was used to verify the expected the structure of the etched beam and to verify how much the KOH etches the beam in the mask border region (Fig. 8a). The particular shape of the etched beam is a consequence of the KOH angle and the exposure angle at the lithographic step (see 3.2). The second small tilted side is not desired, but at the same time it doesn’t represent a problem since the grating will be operated in the Littrow condition (Fig. 8b), that is the input angle (in respect to the vertical plane) is equal to the blazing angle (in respect to the horizontal plane). In such a case the second small tilted side is shadowed completely by the adjacent beam.

Fig. 8.

(a) Simulated structure of the blazed beam after KOH etch. (b) The blazed grating is supposed to be operated in the Littrow condition.

From Fig. 8(b), moreover, the need of a safety margin of about 10 µm can be deduced. This has been introduced between the mask border region and the spring region as seen in Fig. 1(a).

3.2.

Fabrication

Fig. 9 shows the process flow that has been used for the fabrication of the blazed deformable MEMS gratings. Steps (d) to (h) were added to the process used for the planar deformable MEMS grating presented in 2.2.

Fig. 9.

Process flow for the blazed deformable MEMS gratings. This process adds steps (d) to (h) to the process for the planar deformable MEMS gratings (Fig. 3). Using KOH etch, an optically smooth surfaces can be obtained, corresponding to {1 1 1} Silicon planes. The entire process requires two masks.

The proposed process is thus fully compatible with the process for the planar gratings, giving the possibility to use the same fabrication run to obtain both the planar and the blazed deformable MEMS grating, even on the same wafer.

After the deep reactive ion etching that defines grating beams, springs and comb actuators (c), a thermal oxide is grown and a resist spinned on the wafer (d): the oxide, conveniently patterned, will act as mask during the KOH etch.

In order to pattern the oxide only in the region of interest, a mask is used at an exposure angle of 30 degrees (e). After resist development (f), a BHF etch (g) is applied to remove the oxide specifically in the regions where the KOH bath has to etch the Silicon beams (h). The device is then exposed to HF vapor (i) to remove the thermally grown oxide and to liberate the grating, by removing the sacrificial oxide.

The feasibility of the process has been verified. Fig. 10 presents preliminary fabrication results: in (a) the top view of a grating test structure, in (b) the lateral view. It was possible to obtain the desired structure as shown by the good agreement with simulations results. In (b) the lateral residuals of the oxide after the KOH step are still visible. They will be removed by the final HF vapor release step.

Fig. 10.

(a) Top view of a grating test structure with periodicity 12 μm. (b)Lateral view of the same test structure.