1.Introduction- Molding the Light

Light has several properties, namely amplitude, phase and polarization. If we are able to manipulate these properties as we like, we would have the power to control the light. Conceptually this is a simple story, but in real life we need suitable technologies and devices to get it. In a famous book by J. Joannopoulos [1] we read how we can mold the “flow” of light by making the light propagate though suitably structured materials, i.e. photonic crystals, having some spatial periodicity of the dielectric permittivity along certain directions. The physics behind is rather simple: optical beams can self-sustain and propagate (or self-inhibit and be stopped) depending on the constructive (or destructive) interference among different time-delayed portions of the same beam. So, we can engineer allowed and forbidden directions of propagation for the light and, at least within certain limits, control its speed, provided that we can build such periodic structures with an accuracy well below the optical wavelength.

A similar concept applies if we want to mold the “shape” of light, that is if we want to design light beams with suitable amplitude, phase, and polarization characteristics. In principle, this control can still be operated by exploiting self-interference of different portions of the same beam. But we need a suitable technology. An option is offered by spatial light modulators (SLMs), that are kind of optical screens where each pixel can be individually controlled to introduce a controllable local phase shift in the wavefront of the beam [2]. SLMs can have a very high resolution (thousands or millions of pixels) and provide a reliable commercial tool for applications where size, cost and speed are not a big issue. However, there are a bunch of emerging applications that require fast manipulation of free-space optical beams, together with integrability with other optical functionalities, small size, low power-consumption, and low-cost.

In this scenario, programmable photonic integrated circuits (PICs) are likely to be a promising technology. Here we illustrate the main properties of programmable PICs and we explain why they can be a game-changer for the generation, control, and detection of free-space optical beams.

Figure 1. (a) Schematic of a programmable PIC made of a mesh of cascaded tuneable MZIs. (b) The mesh is integrated in a photonic chip bridged to a CMOS ASIC hosting the electronics for the monitoring and control of the programmable PIC.


2. Programmable Photonic Circuits 

Programmable PICs are general-purpose meshes of elementary 2×2 tunable beam couplers implemented by using integrated optical interferometers [3]. To some extent, they are the photonic (analog) counterpart of electronics (digital) FPGA [4]. Fig. 1(a) shows an example of programmable PIC made of a mesh of diagonally cascaded Mach-Zehnder interferometers (MZIs), but other topologies can be designed with either feedforward or recursive configurations [1]. Each MZI can control the amplitude and the phase of the optical field at its output ports, in such a way that the light can be coherently addressed to the output ports of the mesh with arbitrary-controllable amplitudes and phases. Mathematically, a mesh with N diagonal rows of M cascaded MZIs can implement any arbitrary N × M linear transformation [5]. High-index-contrast photonic platforms, like silicon photonics, enable the integration of programmable PICs with hundreds of MZIs on a few-square-mm chip. As unique property, these circuits can self-configure with simple algorithms, which automatically set the MZIs to their optimum working point [1,6]. The MZI tuning is operated by means of integrated phase shifters, with a time response ranging from a few ns (thermal actuation) down to ns time-scale (electrooptic actuation). Each MZI (or a subset of MZIs) is monitored by a photodetector integrated in the photonic chip and is controlled by a local feedback loop, without the need for global optimization  techniques. The front-end electronics can be integrated in a CMOS ASIC bridged to the photonic chip [see Fig 1(b)], which can also host the analog control loops for the self-stabilization of many MZIs [7]. Thanks to the  modular topology and parallel implementation of the control system, programmable PICs can be scaled up to circuits with many optical I/Os without increasing the complexity of the overall control.

3. Beaming “Perfect” Beams Through Obstacles   

If we want to generate, or receive, arbitrarily-shaped free-space optical beams, we need a programmable optical antenna. A possible implementation exploits an optical antenna array controlled by a programmable PIC. In integrated photonics, an optical antenna array can be realized by using a set of grating couplers arranged in 1D [Fig. 2(a)] or 2D [Fig. 3(a)] configurations. By connecting each antenna of the array to an output waveguide of the programmable mesh, the amplitude and the phase of the light emitted (or captured) by each element of the array can be controlled independently.

From here on we consider the case of beam generation (transmitter side), but the same considerations apply for the antenna array employed at the receiver side. The spatial resolution of the radiated beam is related to the number of optical antenna elements integrated in the photonic circuit, that is to the port count of the programmable PIC, which provide the pixels, or equivalently the spatial harmonics, available for designing our beams.

Besides the large flexibility in generating arbitrary field profiles, controlling optical antenna array with programmable PICs enables the creation of “perfectly” shaped beams even if the antennas are not perfectly fabricated [8]. For instance, Fig. 2(b) shows that the far-field pattern generated by a 1D array of four identical antennas (top) is dramatically distorted by the presence of amplitude and phase errors due to fabrication tolerances (center); this effect can be automatically compensated by the programmable PIC (bottom) that redistributes the optical intensity of the excitation field among the antennas and introduces a suitable relative phase shift to recover the desired beam pattern.

Further, light beams can be suitably shaped to bypass obstacles in the free-space path, which introduce wave-front distortion or scattering effects. Fig. 2(c) shows that the field pattern measured without any obstacle (top) is substantially distorted when the beam propagates through a diffusive medium (center), but this effect can be mitigated by the action of the programmable PIC that automatically self-configures to recover the desired beam shape (bottom). The goodness of beaming through obstacles improves with the number of optical antennas employed to shape the beam [in Fig. 2(c) only four antennas are used]. Notably, the tuning procedure of the programmable PIC does not require any prior information on the nature of the obstacle, and the resulting tuned settings can effectively be used for the identification and recognition of unknown obstacles. Because the MZI mesh can generate any output amplitudes and phases, obstacles with arbitrary amplitude and phase profile can be bypassed and/or recognized, provided that they maintain some degree of transparency. For example, we could imagine readout systems for multilevel (amplitude and phase) bar codes as well as generators of optical beams that are tailored to propagate almost unaffected though particle systems, fog, or turbulent environments.


Figure 2. (a) A programmable PIC can be employed for the control of the optical field radiated by an array of optical antennas. Panel (b) shows, respectively, the far-field pattern radiated in case of perfect antennas (top), the effects of amplitude and phase errors due to imperfections in the optical antennas (center) and the desired beam automatically recovered by self-calibration of the programmable PIC (bottom). Panel (c) shows the pattern observed in the far field without any obstacle in the free-space optical path (top), the distortion introduced by propagation through a diffusive medium (center), and the beam that is automatically recovered by the programmable PIC (bottom) (adapted from [8]).


4. Beaming More Beams to Beam More Information 

Light beams can carry information. This is obvious thinking about fiber-optic communications. And looking at the recent evolution of fiber-optic communications towards ultrahigh capacity systems exploiting space division multiplexing (SDM) [9], we understand that, if we are able to simultaneously handle many light beams with a single device, we can hugely increase the amount of information that we can transmit. So, it exciting to realize that programmable PICs can be used to simultaneously create, manipulate and detect multiple free-space optical beams. And, more important, these beams are inherently orthogonal, meaning that they can be processed with no mutual cross-talk.

Figure 3. (a) An optical antenna array controlled by a programmable PIC can simultaneously radiate and receive several orthogonal beams with same wavelength and polarization. Example of beam pairs radiated in (b) different directions or (c) with different shape (reproduced from [10]). (d) Real time compensation of time-varying effects, due for instance to dynamic changes in the optical path, can be achieved by exploiting the self-adaptive properties of programmable PICs used at the transmitter and/or receiver side of the optical link (reproduced from [11]).


Figure 3 shows the optical beams than can be generated by a 2D array of 9 grating couplers arranged in a 3×3 square optical antenna array and controlled by a silicon photonic programmable PIC. For example, the mesh can be configured to transmit beams that are steered to different directions (b) as well as beams that are shaped according to different modes (c) [10]. Notably these beams share the same wavelength (1550 nm) and polarization (transverse electric, TE), so their distinguishability is only related to their spatial orthogonality. In principle, with an array of M optical antennas, up to M orthogonal beams can be simultaneously handled, this feature enabling the implementation of high capacity FSO SDM links. In this scenario, the self-adaptive nature of programmable PICs enables the possibility of automatically establishing optimum FSO communication channels [11] and of compensating for dynamic changes in the link [see Fig. 3(c)] caused by, for instance, moving obstacles or atmospheric turbulence. Finally, the implementation of the programmable PIC on a standard photonic platform enables integration with many other devices for functions like wavelength filtering, (de)multiplexing, and fast time-domain modulations, thus enabling implementation of sophisticated and advanced all-optical processing of FSO beams.

5.Final Remarks and Perspectives

In a world where any kind of device will be ever more connected and need to exchange huge amounts of data, flexible, robust and wideband wireless optical connections are expected to play a key role. The capability of shaping and manipulating free-space optical beams with arbitrary shapes and controllable properties is essential in a variety of rapidly emerging applications from autonomous drones to wearable technology. Working with a multi-disciplinary team of researchers, supported by the EU Future and Emerging Technologies Program Super- Pixels [12], we are exploring the application this technology for microscopy, quantum communications, environmental monitoring and endoscopy. The way we see and sense the world around us is evolving rapidly. And the possibility of molding light beams with small photonic chips is something that that can facilitate and speed up such a transformational process.

[1] J. D. Joannopoulos; R. D. Meade; J. N. Winn, Photonic crystals: molding the flow of light,
Princeton University Press, 1995.
[2] G. Lazarev, A. Hermerschmidt, S. Krüger and S. Osten, “LCOS Spatial Light Modulators: Trends
and Applications,” Optical Im- aging and Metrology: Advanced Technologies, pp. 1-29, 2012.
[3] W. Bogaerts et al., “Programmable photonic circuits,” Nature, vol. 586, pp. 207-216, 2020.
[4] D. Pérez et al., “Multipurpose silicon photonics signal processor core.” Nature
Communications, vol. 8, 636 (2017).
[5] D. A. B. Miller, “Self-configuring universal linear optical com- ponent,” Photonic Research,
vol. 1, pp. 1-15, 2013.
[6] A. Annoni et al., “Unscrambling light – Automatically undoing strong mixing between
modes,” Light: Science and Applications, vol. 6, p. e17110, 2017.
[7] F. Toso et al., “Self-Stabilized Silicon Mach-Zehnder Interferom- eters by Integrated CMOS
Controller,” IEEE 18ᵗʰ International Conference on Group IV Photonics, 2021.
[8] M. Milanizadeh et al. “Coherent self-control of free-space optical beams with integrated
silicon photonic meshes,” Photonics Re- search, vol. 9, pp. 2196-2204, 2021.
[9] D. Richardson, J. Fini and L. Nelson, “Space-division multiplex- ing in optical fibres.,”
Nat. Photonics, vol. 7, p. 354–362, 2013.
[10] M. Milanizadeh et al., “Multimode Free Space Optical Link En- abled by SiP Integrated Meshes,”
European Conference on Optical Communication (ECOC), 2021
[11] M. Milanizadeh et al., “Establishing free-space optical commu- nication channels through a
reconfigurable silicon mesh,” IEEE Photon. Soc. Summer Topicals Meeting Series, 2021.
[12] https://www.superpixels.org