a wavelength causes destructive interference of light and the light is absorbed at the metal surface.Along with current trends to eversmaller, more versatile devices, thin and reconfigurable infrared absorbers which do not require cumbersome nanofabrication are desirable, for example, to fabricate ultra-thin, effective light detectors. This is especially true for the mid-infrared spectral range, where optical imaging devices are important for many technological applications in key areas like thermography, surveillance, automotive safety, and astronomy. [8] It has been shown that ultra-thin absorbers can consist of just an imperfect conductor as a reflector and a very thin, lossy dielectric as the absorbing film because of non-trivial phase shifts at the interfaces. [13] Originally, the absorber functionalities are fixed after fabrication, limiting their application range. To achieve more versatile absorbers, research has been directed towards realizing tunable absorption functionality. [8,[14][15][16][17][18][19][20][21][22] Phase-change materials (PCMs) are especially interesting for this because they offer non-volatile tuning (in contrast to the volatile tuning with transparent conducting oxides or phasetransition materials) by switching between their amorphous and crystalline structural phases whose optical properties differ significantly, because of a unique bonding mechanism, referred to as metavalent bonding. [23][24][25][26] They can be switched between their phases by thermal, [27] electrical, [28] or optical heating [29] on time-scales down to a few nanoseconds [30] and they have already been successfully commercialized in products like re-writable DVDs [29] and PC-RAM. [31] Crystallization leads to an approximately twofold increase of the refractive index in the infrared spectral range for many PCMs [32,33] like Ge 2 Sb 2 Te 5 and its stoichiometric neighbors Ge 3 Sb 2 Te 6 and Ge 8 Sb 2 Te 11 . In addition, these PCMs feature low losses in the infrared, which makes them suitable for many nanophotonics applications [34][35][36] like waveguides, [37,38] photonic memory, [39,40] polaritonics, [41][42][43][44] tunable metasurfaces, [45][46][47][48][49][50][51][52][53] and tunable metasurface absorbers. [8,14,18,[54][55][56][57] But the same low-loss optical properties become disadvantageous for constructing ultra-thin absorbers with thicknesses well beyond the quarterwavelength limit. While their use is still viable in the visible and near-infrared spectral range [58][59][60] where they have significant optical losses, they are no longer suitable at infrared frequencies. Absorbersfor infrared light are important optical components in key areas like biosensing, infrared imaging, and (thermal) light emission, with special need for thin and reconfigurable devices. Here, the authors demonstrate ultra-thin, switchable infrared absorbers based on thin layers of chalcogenide phase-change materials (PCMs) with high optical contrast between a lossless amorphous and an exceptionally lossy crystalline phase (Ge 3 Sb ...

Integrated neuromorphic photonic circuits aim to power complex artificial neural networks (ANNs) in an energy and time efficient way by exploiting the large bandwidth and the low loss of photonic structures. However, scaling photonic circuits to match the requirements of modern ANNs still remains challenging. In this perspective, we give an overview over the usual sizes of matrices processed in ANNs and compare them with the capability of existing photonic matrix processors. To address shortcomings of existing architectures, we propose a time multiplexed matrix processing scheme which virtually increases the size of a physical photonic crossbar array without requiring any additional electrical post-processing. We investigate the underlying process of time multiplexed incoherent optical accumulation and achieve accumulation accuracy of 98.9% with 1 ns pulses. Assuming state of the art active components and a reasonable crossbar array size, this processor architecture would enable matrix vector multiplications with 16,000 × 64 matrices all optically on an estimated area of 51.2 mm2, while performing more than 110 trillion multiply and accumulate operations per second.

The number of systems that are investigated for computation in the physical domain has increased substantially in the recent past. Optical and photonic systems have drawn high interest due to their potential for carrying out energy-efficient linear operations and perceived advantages in latency and general computation speed. One of the main challenges remains to scale up integrated photonic designs to integration densities required for meaningful computation, in particular for matrix-vector multiplications. To address upscaling for photonic computing, here we propose an on-chip scheme for dimension reduction of the input data using random scattering. Exploiting tailored disorder allows us to reduce the incoming dimensionality by more than an order of magnitude, which a shallow subsequent network can use to perform image recognition tasks with high accuracy.

Integrated optical processing networks enable high computation speeds combined with low energy consumption. We present here a scheme for dimension reduction for optical neural networks, by orders of magnitudes, while still reaching high classification accuracies.