2.1

Flow Cytometry

Modern flow cytometry allows a fast identification and separation of cells according to their size, shape or type and has become one of the most widely used techniques in life sciences¹⁴. In the typical configuration, a flow cytometer consists of a core stream of a sample solution (~20 µm diameter) containing the cells to be analyzed. A surrounding sheath fluid of saline solution (~200 µm diameter) produces a hydrodynamic focusing of the core stream and creates a flow of cells in a single line. In this configuration, the cells are interrogated one by one as they pass through a laser beam. When recording scattered and fluorescent responses simultaneously, biologists receive comprehensive and quantitative information about the examined specimen that allows precise identification and sorting.

The critical requirements for the illuminating beam in flow cytometry are power stability, low noise level at high frequency and high spatial shape quality. In regards to the last point, an efficient excitation requires elliptical spots of around 60 µm in the direction perpendicular to the cell flow and 10 µm in the direction in line with the flow (see Figure 1a). The wider dimension of the elliptical beam along its major axis (60 µm) provides an almost uniform illumination field across the width of the flow (~ 20 µm) which assures that the optical response (scattering or fluorescence) does not fluctuate if cells stray from the center of the illuminating beam. On the other hand, the thinner dimension of the elliptical beam along its minor axis (10 µm) allows cells to quickly pass through the illumination point and avoids the simultaneous excitation of more than one cell at a time. Typical target cells have dimensions of 5 – 15 µm and flow in the core stream at a velocity of around 1 m/s. As cells pass through the laser beam at this speed, they generate scattering and fluorescent signals with durations in the order of several microseconds.

Figure 1.

Drawing of a multi-color flow cytometer and details of the required illumination pattern allowing time-domain resolution of different fluorescent signals from excited cells. b) Schematic of TOPTICA’s MLE for super-resolution applications (iChrome SLE) including several free-space optics components such as dichroics, an acousto-optic tunable filter, a fiber switch and several lenses.

Flow cytometers based on MLEs use a number of excitation wavelengths that can be focused on a single spot or at different points along the stream. The latter configuration is sometimes preferred because it enables time-domain discrimination of fluorescent signals from fluorochromes exhibiting different excitation wavelengths but similar emission spectra. As the cells go across each color beam sequentially, they emit independent pulses that can be well resolved in the time domain with a fast photodetector.

In typical flow cytometers the lasers are used in Continuous Wave (CW) operation and benefit from variable optical attenuators (VOA) to adjust the excitation power for each color. In contrast to direct laser attenuation, an external VOA allows setting the laser in the optimal operating conditions, which can lead to a better performance in terms of noise and thermal stability. Flow cytometry is not very demanding in terms of power with levels of around 15 mW at the illumination point. However, in order to benefit from multi-laser excitation with separated spots, the output beams need to be very well defined in shape and spacing.

In a PIC, the lithographical definition of the waveguides enables a modification of the beam shape and a very precise control of the spacing between the different colors. Furthermore, the implementation of integrated VOAs with active phase tuners allows the replacement of expensive AOMs. Thus, a MLE with PIC provides a cost effective way to control the output beam that can be refocused into the flow cell of the cytometer.

2.2

Confocal Laser Scanning Microscopy

Confocal Laser Scanning Microscopy (CLSM) is a very popular technique in life sciences, semiconductor inspection and material science. In bioimaging, CLSM is commonly used in combination with fluorescence microscopy: The laser beam is raster scanned over the biological sample and the fluorescent response is recorded to create a two-dimensional image. For many applications, excitation with multiple wavelengths allows the resolution of different structures with respect to each other. This is for instance the case of the Brainbow technique for the study of neural connections that uses red, green and blue fluorescent proteins to flag individual neurons and track their pathways with distinctive colors¹⁵.For such multicolor experiments, commercial MLEs provide the combination of all wavelengths in a single output fiber. The wavelength combination is typically implemented with dichroic mirrors whose alignment needs to be very accurately maintained for an efficient coupling into the output fiber.

For CLSM typical required power levels are on the order of µW for imaging and mW for bleaching. Excitation at µW levels requires the laser diodes to operate around the threshold, which leads to increased noise and stability issues when the lasers are directly modulated. Moreover, when the mirror returns to the beginning of the line (fly-back) during the raster scanning operation, a fast switching signal is typically used to blank the lasers. This blanking signal avoids an unnecessary exposure of the probe to the laser excitation and minimizes photobleaching. Other microscopy techniques that require intensity modulation and rapid switching are Controlled Light Exposure Microscopy (CLEM) and Fluorescence Recovery After Photobleaching (FRAP). The required fast switching operation is challenging, since it causes major difficulties for direct modulation as the laser does not reach thermal equilibrium. Furthermore, concomitant mode hoping between longitudinal laser modes results in spectral instability and creates artifacts in the confocal image. Additionally, CLSM usually requires particular excitation wavelengths, such as 561 nm, that are only available as DPSS lasers and therefore cannot be directly modulated at the required speeds.

All these reasons make external modulation a required feature for most commercial MLE solutions targeting confocal microscopy applications. Typical implementations are based on AOMs, which are expensive components increasing the cost of the final product. Therefore, the replacement of AOMs by integrated modulators on a PIC-based solution together with the on-chip implementation of the wavelength combiner can lead to a significant cost reduction.

2.3

Super-Resolution Microscopy

On the other end of the broad spectrum of MLE applications are power intensive imaging techniques such as super-resolution microscopy. They allow imaging with a resolution higher than the diffraction limit and can be separated into two major groups: deterministic super-resolution (e.g. STED) and stochastic super-resolution (e.g. PALM, STORM). All these methods have in common that they demand relatively high excitation powers. Furthermore, in order to reach maximum efficiency, they require a few wavelengths that are often only available from DPSS but not from laser diodes. Here again, intensity modulation of these lasers requires the incorporation of external modulators.

Moreover, novel illumination techniques have brought the requirement of having multiple fiber outputs. This feature requires a discrete fiber switch that is often realized with a MEMS or Galvo Scanner. All these components are expensive, add transmission losses and increase power instability due to the longer beam paths.

For MLE targeting super-resolution applications, PIC technology has the potential to make additional external modulation and switching obsolete. Once the laser beams are coupled into the waveguides, these functionalities can be added in a cost-effective manner without incurring the stability problems arising as a result of direct modulation. This also allows building more compact systems than what it is possible with current mainstream technology.

On the other hand, the required output power levels also represent an important challenge for PICs. All integrated components need to be designed with tight specifications in regards to allowable insertion losses. Furthermore, high power densities at required near UV wavelengths can cause long term degradation at the facets of the chips. Flexibility is also limited with PICs when a requirement of adding additional wavelengths arises after chip design and fabrication.

The following table shows a summary of the different use cases:

Table 1.

Summary of MLE requirements for biophotonic applications.

4.1

Beam shaping for PIC-to-fiber coupler and flow cytometry illumination pattern

In current commercial MLEs all combined colors are typically coupled into a single mode (SM) and polarization maintaining fiber. The implementation of a similar interface for a solution based on a PIC requires the design of an edge coupler that simultaneously adapts the output beam for each color to the respective mode profile in the SM fiber. SM fibers exhibit mode profiles with circular symmetry. Furthermore, the 1/e² Mode Field Diameters (MFDs) in both axes typically scale with the wavelength (λ) according to the linear approximation:

In the PIC, the greater refractive index contrast and the resulting greater spectral variation of modal confinement hinders the design of a device with excellent performance for all colors. Thus, it is only possible to design the output width of the edge coupler as the result of a trade-off between the coupling losses at the different wavelengths. Figure 4a shows the calculated coupling efficiencies for edge couplers with respective output widths of 0.8 and 1 µm and for both polarizations as a function of the operating wavelength. For TM polarization and 0.8 µm, the mode expands outside the waveguide core and creates a circular profile with an excellent matching to the fiber mode profile for green light. At the peak wavelength (532 nm – 580 nm) the device achieves a maximum coupling efficiency close to 90%. However, at lower (higher) wavelengths, the mode confinement increases (decreases) faster than in the fiber, which results in increased coupling losses due to mode mismatch. Nevertheless, at 405 nm and TM polarization, the PIC-to-fiber coupling efficiency is still higher than 50%

Figure 4.

a) Simulated PIC-to-fiber coupling efficiencies considering a standard SM fiber (MFD given by eq. 1) as a function of wavelength for edge coupler widths of 1 µm (blue) and 0.8 µm (green) and for TM polarization (solid lines) and TE polarization (dashed lines). b) Angular divergences (full angle at 1/e² intensity) in the in-plane (solid) and out-of-plane (dashed) directions for TE polarized modes as a function of the output waveguide width. c) Ratio between in-plane and out-of-plane angular divergences as a function of the output waveguide width.

For the selected values of edge coupler output width at TE polarization, the required modal delocalization outside of the core is only reached at the longer wavelengths. As a result, the PIC-to-fiber coupling efficiency is penalized by more than 50% for wavelengths below 550 nm.

On the other hand, the higher confinement of TE polarized modes makes it possible to obtain larger divergence angles in the out-of-plane direction. Figure 4b shows the calculated angular divergence in the far-field as a function of the waveguide width for the different colors. The angular divergence is defined here as a full angle and refers to the 1/e² of the peak intensity in air. It can be seen that it is possible to reduce the divergence in the in-plane direction by increasing the waveguide width. However, in the out-of-plane direction the divergence angle increases and saturates for waveguides wider than 3 µm as a consequence of the increased confinement. For small waveguide widths, the angular divergences in both directions tend to take the same value as is expected from a beam with circular symmetry.

The adjustment of the divergence angle by the definition of the waveguide width enables the implementation of an MLE output optical interface optimized for flow cytometry featuring the required beam sizes at the illumination points (60 µm by 10 µm) and a compact separation between the different colors. In the far-field, the spot size of the radiated beams (MFDλ) scales linearly with the divergence angles (θ_λ) and the propagation distance (z) as:

In order to create the elliptical beams with the required aspect ratio, we first selected values for the waveguide widths at the PIC output facet that achieve a ratio of 6:1 between the angular divergence in the out-of-plane direction (perpendicular to the flow) and the in-plane direction (parallel to the flow). With TE polarization at all colors, these values correspond to 4.9 µm, 7.3 µm, 9.8 µm and 13 µm respectively for 405, 488, 561 and 640 nm wavelengths (marked points in Figures 4b and c). As depicted in the schematic in Figure 5a, adiabatic tapers in the in-plane direction assure the gradual conversion from the compact modes of single mode interconnection waveguides to the elliptical modes targeted at the output interface of the chip.

Figure 5.

a) Representation of the optical interface of a PIC generating an illumination pattern for flow cytometry. The PIC is tilted at an angle φ relative to the direction of the flow cell for compensation of the large spectral dependency of the vertical angular divergence of the emitted beams. b) MFDs (in the direction perpendicular to the PIC surface) for the different color beams as they diverge during propagation in the direction perpendicular to the cell flow. c) Normalized intensity at the illumination plane (direction parallel to the flow) of the simulated beam spots.

The large spectral dependency of the angular divergence θ_λ in the out-of-plane direction does not allow getting the required spot size for all colors simultaneously at a given propagation distance. For example, due to its smaller angular divergence the red beam (640 nm) needs to propagate over a larger distance than the other colors before the spot size increases to the targeted 60 µm in the out-of-plane direction. In order to compensate for the differences in divergence angles and to achieve the required spot sizes at the same illumination plane, the red color needs to begin free space propagation at a length L1 = 15.6 µm before the beam at the 561 nm wavelength. Similarly, the beam at 488 nm has to start radiating a length L2 = 14.4 µm after the beam at 561 nm. Finally, the violet beam (405 nm) has to remain guided a distance L₃ = 16.3 µm longer than the blue beam (488 nm). In our design, as is illustrated in Figure 5a, the required differences in the free space optical path lengths (L1 - L₃) are easily achieved by tilting the PIC at an angle φ relative to

the direction of the flow and by setting the lateral separation between the waveguides at the output facet (D1 - D₃) accordingly. The relation between the waveguide separation and the tilt angle can be readily calculated as:

Furthermore, the tilt angle and the corresponding relative positions of the output waveguides also determine the separation between the centers of the different beams (S1-₃) according to the relation:

In this scheme, the waveguides reach the output facet at an angle φ_in that is related to the PIC tilting angle φ in air (n = 1) through the Snell’s law of refraction:

with ne_ff the effective refractive index of the modes in the SiN slab (-1.5). For the selected polarization (TE), increasing φ_in to the Brewster angle value (~ 34 degrees) has the benefit minimizing reflections at the PIC-air interface. However, the selection of such a steep angle also results in small values of S1-₃, leading to a significant overlap (> 5%) between the adjacent beams at the illumination plane. In the flow cytometry application, a high overlap value is detrimental because it hinders the differentiation and resolution of fluorescent signals coming from adjacent spots. In our final design, the tilt angle has been fixed to φ = 39.3 degrees (φ_in = 25 degrees) for an overlap of less than 0.02%. Thus, the waveguides at the output facet are separated with values D1 = 24.6 µm, D₂ = 22.7 µm and D₃ = 25.7 µm and the laser beams at the illumination plane are separated by S1 = 19 µm, S₂ = 17.6 µm and S₃ = 19.9 µm, extending a total length of less than 80 µm. Figure 5c shows the normalized intensities of the different spots in the axis parallel to the flow and at the illumination plane. The slight deviations in MFDs in relation to the target value (10 µm) are derived from a near-field divergence due to the large Rayleigh range of the beams for the in-plane direction (103 µm for the beam at 640 nm) and the proximity of the projection plane (112 µm distance from the 640 nm waveguide at the PIC output facet).

4.2

Variable Optical Attenuators and Switches

In the proposed PIC, the variable optical attenuators (VOAs) and the switches are implemented by Mach-Zehnder Interferometers (MZI). In these devices, the input light beam is equally divided in two waveguides by means of an integrated splitter (see Figure 6). The two waveguides constitute the arms of the interferometer and comprise phase tuners that allow an active adjustment of the optical phase difference (ΔΦ). At the output of the phase tuners the two light paths are again recombined and interfere in an integrated combiner.

Figure 6.

a) Field propagation (TM pol. Ey field amplitude) inside the 1x2 MMI designed for a wavelength of 561 nm. b) Building block diagram of the Mach-Zehnder interferometer for the implementation of the VOA and switch components. c) Field propagation (TM pol. Ey field amplitude at 561 nm) in the 2x2 MMI combiner for a phase shift difference of +90 degrees (top) and -90 degrees (bottom).

In our design, the splitter and combiner components are respectively implemented by means of 1x2 (Fig. 6a) and 2x2 (Fig. 6c) Multimode Interference couplers (MMIs). Depending on the phase difference, it is possible to adjust the power on one of the 2x2 MMI output ports (VOA functionality) or completely redirect the light beam to one port or the other (switch functionality). For the VOAs, the additional output can be used as a reference monitor for the stabilization of the power at the main port. In our demonstrator, the phase tuners leverage the thermo-optic effect in Silicon Nitride¹⁶ and are implemented by thermal resistors made of a metal layer placed above the top SiO₂ cladding of the interferometer waveguides. Faster modulation speeds (in the MHz range) are also possible in TriPleX technology by means of the stress-optic effect induced by a piezoelectric lead zirconate titanate thin film¹⁷.

Since the phase difference directly depends on the operating wavelength, it is necessary to implement independent VOAs and switch units optimized for every single color. In comparison with a MLE solution based on a single fiber switch, the PIC configuration allows a more versatile distribution of the laser lines between the output fibers.

4.3

Wavelength Combiners

The spectral variation in modal confinement is used for the implementation of the wavelength combiner. As represented in Figure 7a, the device is comprised of three cascaded directional coupler (DC) stages. In the first DC (WLC1), the optical field at 488 nm is evanescently coupled into the adjacent bus waveguide that transports the 405 nm beam. Similarly, in the second DC (WLC2) the wave at 561 nm is coupled into the bus waveguide with the already combined waves at 405 nm and 488 nm. Finally, in a third DC (WLC3) the 640 nm light beam is also added to the other colors. The waveguide width of each DC is independently designed in such a manner that the evanescent field of the optical modes at the wavelength to be injected (higher wavelength with lower modal confinement) is significantly larger than the evanescent field at the lower wavelengths already present in the bus waveguide. Thus, by the selection of an appropriate DC gap it is possible to achieve, at the longer wavelength, a total transfer of power from one waveguide to the other, whereas only a small amount of light coupling occurs at the lower wavelengths (see Figure 7b). In a DC, the power at the output cross-port (P_cross) can be described with the expression:

Figure 7.

a) Schematic of the wavelength combiner based on three cascaded directional coupler stages (WLC). b) Field propagation (Ey amplitude, TM pol.) in the first WLC for the combination of the wavelengths 405 nm and 488 nm. c) Simulated transmission spectra of the different WLC in the bar port (dashed lines) and the cross port (solid lines).

with L the coupling length of the DC and L_π,λ the half-beat length of the directional coupler.

In order to achieve complete light coupling into the bus waveguide, the length of each DC is set to the half-beat length (L_λ) at the injected wavelength (longer wavelength λ).This length is adjusted in simulations to also account for the coupling that occurs in the S-bends at the input and the output of the DC. Moreover, the gap and the waveguide width are designed to be sufficiently large in order to get a L_π,λ that is at least seven times shorter than the half-beat length for the immediately shorter wavelength(L_π,λ=7•L_π,λ-1). This relation assures an insertion loss of less than 5% for each of the colors already propagating in the bus waveguide. Figure 7c shows the simulated spectral response of each WLC corresponding to the cross-port (solid line) and the bar-port (dashed lines).