Micromachined phase-shifted array-type Mirau interferometer for swept-source OCT imaging: design, microfabrication and experimental validation

: OCT instruments permit fast and non-invasive 3D optical biopsies of biological tissues. However, they are bulky and expensive, making them only affordable at the hospital and thus, not sufficiently used as an early diagnostic tool. Significant reduction of system cost and size is achieved by implementation of MOEMS technologies. We propose an active array of 4x4 Mirau microinterferometers where the reference micro-mirrors are carried by a vertical comb-drive microactuator, enabling the implementation of the phase-shifting technique that improves the sensitivity and eliminates unwanted interferometric terms. We focus on the design of the imaging system, the microfabrication and the assembly of the Mirau microinterferometer, and the swept-source OCT imaging.

frequency of 110 kHz. The incident light beam is collected by an array of 4x4 Mirau microlenses and directed towards the sample to be measured. For each single-channel interferometer the collected light passes through a thin beam-splitter plate that reflects a part of it back to a moving reference micro-mirror integrated on top of the MEMS actuator, while the rest of light is transmitted towards the sample to be measured. The beams reflected by the sample and by the reference micro-mirror interfere, generating an interference pattern, and are directed by a 4x4 array of microlenses towards a high-speed camera after the transmission by a cube beam-splitter. The square-pixel size of the camera is 12 µm and the frame rate is set to 4 kfps. The vertical actuation of Mirau reference micro-mirrors allows a phase shifting increment to be achieved. Being actuated at the resonance frequency by sinusoidal driving signal, the MEMS-based system enables a rapid measurement of the interferograms.
The architecture of the active Mirau interferometer is shown schematically in Fig. 1(b). The device consists of a series of vertically stacked components: a doublet of plano-convex microlenses, an electrostatically driven vertical comb-drive actuator carrying the array of Mirau reference micro-mirrors, a spacer and a planar beam splitter plate. In the following discussion, W1 represents the microlens doublet matrix, W2 represents the Z microscanner, and W3 is the spacer while W4 represents the beam-splitter wafer. To build the Mirau interferometer, the doublet of microlenses has been selected instead of single microlenses [14]. Because of lower optical aberrations, the optical performance achieved by the lens doublet is better than for an individual lens having an equivalent numerical aperture. The assembly of two glass microlens arrays is made by anodic bonding at the wafer-level thanks to their silicon frame resulting from each array fabrication. Such microlens doublets are then well aligned and robust. The diameter of an individual microlens is 1.9 mm and the array pitch is 2 mm. The focal length of a microlens doublet is 7.5 mm, giving a numerical aperture of 0.1.
The transverse resolution of OCT imaging is defined by the effective numerical aperture of the focusing lens and is limited to 6.3 μm. Moreover, the axial resolution of OCT imaging is 6.2 µm that is determined by the optical performance of the light source, i.e. the central wavelength and the swept range of the laser source. The OCT configuration allows a depth of penetration of 0.6 mm with a sensitivity in the range of 70-85 dB.
The key element of Mirau interferometer is the vertical microscanner W2. The design and the fabrication of this silicon-on-insulator (SOI)-based scanner are described in Ref. 15. The microscanner is designed to generate a vertical displacement of a large platform with an array of 4x4 reference micro-mirrors of the Mirau interferometer, as shown in Fig. 1(c). Here, the micro-mirror platform is represented with micro-mirrors 1-4 whose mechanical characterization is provided latter. The vertical motion of the whole array of the 4x4 reference micro-mirrors at the resonance frequency of actuator is used to implement the sinusoidal phase-shifted imaging where the displacement can be controlled precisely by in situ position capacitive sensor. In our case, the microscanner is designed to work in transmission, i.e. the areas around the micro-mirror are transparent. The platform is thus formed by 16 apertures made in a 40-μm thick device layer of the SOI wafer and is structured with small holes in order to decrease its weight. A honeycomb structure is chosen for its good trade-off between mass and stiffness, preventing the possible vertical deformation of the platform (see the image of Fig. 1c). The size and the shape of the micro-mirrors are defined by the optical design of the Mirau microinterferometer as a 400x400 μm 2 square surface. Important design issue concerns the minimization of the obscuration created by the reference micro-mirror in the center of the light beam.

Results and discussion
To validate performances of the MOEMS-based Mirau interferometer in OCT imaging both optical and mechanical characterizations have been performed.
First, the lateral resolution has been evaluated experimentally by imaging a 1951 USAF resolution target. The linewidth of smallest line pair on the USAF target, which is well resolved, is 6.2 µm. Figure 7(a) shows two cross sections of the intensity Point Spread Function (IPSF) of one selected Mirau interferometer after assembly [19]. The image at the left of Fig. 7(a) represents the axial irradiance distribution (XZ slice) measured along the optical axis. The image on the right of Fig. 7(a) corresponds to the YX slice measured at the focal point. The FWHM value of the IPSF at the center of the focal plane is 5 µm, which confirm the previous optical characterization of the lateral resolution. Moreover, the focal length of the doublet equals 7.65 mm and the depth of field is around 400 µm.
The actuation performance of the Z microscanner has been tested using the Doppler vibrometry module MSA500 from Polytec. We measured the actuator static displacement as a function of applied voltage between 0 and 40 V. For 40 V the measured amplitude of the static displacement is around 1.75 µm. The first resonance frequency of the Z microscanner is 485 Hz, corresponding to a piston mode motion of the scanner. This mode is precisely used for the actuation of the vertical microscanner to perform the phase modulation. The quality factor of the vertical microscanner is around 64, giving thus an excellent basis for robust and stable actuation. Figure 7(b) shows the micro-mirror displacement under a sinusoidal excitation, demonstrating the targeted peak-to-peak amplitude of 352 nm. This function is obtained by applying a voltage V S = 2.5 + 1.5sin(ω 0 t) V for all the actuators, where ω 0 is the first resonance pulsation. As the electrostatic force is proportional to the square of the voltage, the applied voltage is a sum of a DC force, a force proportional to sin(ω 0 t) and a force proportional to sin(2ω 0 t + π/2).
For the implementation of phase shifting algorithm, it is important to know the displacement amplitude of all the reference micro-mirrors in the same array. The dynamic displacement of the platform at resonant frequency was investigated at different points of the platform, i.e. at the position of the four micro-mirrors of Fig. 1(c), forming a quarter of the platform. The vertical displacement of the reference micro-mirrors was measured under dynamic driving voltage at the center position of the four micro-mirrors (Mirror 1-Mirror 4). Here, the measured displacement is much larger than the targeted one (352 nm), demonstrating the potential of this microscanner for medical imaging applications, for which small voltages are particularly important since the device is in contact with the patient body. The displacement of the micro-mirror in the middle of the platform (Mirror 2) is 70 nm higher than the one next to the spring (Mirror 3) because of the vertical mechanical deformation of the suspended shift efficienc Fig. 7 axis a excita Standard S overlapping m the measured shifting it is p of the present interferometer eliminate the phase modula actuated surfa The demo surface of a m plot of this sa to the spectra other artefact which is cente the lack of an generating an depth corresp The back sur also responsib space. Its com d platform. W cy, especially in 7. Opto-mechanica and XY slice at th ation obtained at re SS-OCT suffer micro-mirror im sample by de possible to rem t Section is to d r as a phase parasitic term ators [23], ele ace-micromach onstration of p mirror with one ample obtained l signal. The tr terms such as ered onto the z ntireflective co n autocorrelatio ponding to the rface of the be ble for a replic mplex conjugat e demonstrated n case of sinus reference mic A-scan obtain the acquisition two shifted s d its replica SR lation terms, a Fig. 8(c) show We used a class e true OCT sig rithm improves h a 2-frame phase-cro-mirror ned after n of two spectra is R*) while as well as ws the Asical fourgnal (S) is s both the The first experimental demonstration of SS-OCT imaging has been made using a scotch tape stick onto a silicon wafer. Figure 9(a) shows the acquired cross-section image where the three different layers appear. This image is completed by two A-scan plots, which are extracted from the center and from the edge of the image. As the OCT image was obtained without the phase-shift, artifact peaks appear around the three main ones, indicated by orange dots. The B-scan reconstruction was retrieved along 300 µm lateral field of view. A small decrease of the sensitivity can be seen at the edge of the cross-sectional image. The decrease of the signal-to-noise ratio at the border is estimated to be −5 dB, compared to the main peak, corresponding to the signal from the wafer layer. Despite this signal loss, the two diopters generated by the scotch-tape and the surface of the wafer can be retrieved. The measured height of the scotch-tape is 22 µm. The sensitivity of this OCT image is in the range of 60 dB.
Second, we selected a diffusive sample optically similar to the sample of pig skin, which is representative of what is desired in dermatology. For a dermatologist a good OCT imager is capable of revealing the fine details of skin tissue microstructures with a lateral resolution of less than 10 µm and an imaging depth of less than 1 mm. Figure 9(b) represents a crosssectional OCT image of a layer of scratched paint covered with varnish. Here, the sample was mounted on a mechanical stage, insuring a linear displacement on 4 mm with a precision of 1 µm. Scratches under the varnish are clearly visible. The thickness of the varnish's layer is about 300 nm. The estimated sensibility is around 80 dB.
Finally, we measured OCT images of an onion slice to demonstrate the feasibility for 3D imaging of biological specimens. Figure 9(c) shows the microscopic structure of an onion slice where a 3D volumetric image of 300 × 300 × 600 μm 3 is reconstructed. The 3D reconstruction reveals inner sample structure corresponding to the different separated layers. We estimate the sensitivity of this image to be in the range of 35-40 dB.

Conclusions and perspectives
MOEMS technology combining MEMS and micro-optics is well suited for manipulating light with different ways to scan, steer or modulate the beams, offering batch-fabricated microsystems at lower cost. A number of MOEMS technology demonstrators have been developed for imaging needs of the medical market.
In this paper, we reviewed the design, the fabrication process and experimental results of an active micromachined array-type Mirau interferometer. The Mirau interferometer includes the electrostatic vertical microscanner carrying an array of reference micro-mirrors, being the key element of a SS-OCT imager. The technology of each functional part of this miniature multi-channel Mirau microinterferometer and the integration schema are described and discussed. In particular, the 3D assembly offers an integration platform for complex MOEMS and allows the effective integration of various heterogeneous technologies, disposed in vertically stacked building blocks (glass microlens, MEMS actuator, beam splitter) in a minimum space. The presented results demonstrate experimentally the proof-of-concept of our approach.
Thanks to the optimized design of this vertical architecture, several original technologies are proposed, offering the integrity of MEMS microactuators assembled with microoptics. This approach offers a low level of residual stress and provides a miniaturized and low-cost solution to create highly accurate microsystem for OCT imaging, composed by several hybrid components. The proposed OCT microsystem, presenting the lateral resolution of 6.3 µm, is well suited for the early diagnosis of cutaneous pathologies. The OCT microsystem is rapid because of the array-type architecture, and has the attractive attributes of simplicity and low cost.
When operating working in the regime of full-field OCT, realizing direct imaging with the help of the complete array of Mirau interferometers, the microsystem has a field of view of 8x8 mm 2 . This field of view is larger than the field of view of commercially available OCT microscopes, widely used in dermatology. Thus, the field of view of Vivo Sights system from Michelson Diagnostic is 6x6 mm 2 while that of Afga HealthCare Skinell is 1.8x1.5 mm 2 . In future works, the full benefit of full field OCT imaging will be obtained by the improvement of the illumination block with the implementation the Köhler illumination. Other step of development will include the improvements of optical quality of Mirau components, including in particular the design and fabrication of an improved version of focusing microlenses with AR coatings.