Expectations for quantum computers are high: They are supposed to outperform digital computers and pave the way for solutions that go far beyond the capabilities that artificial intelligence already delivers. They are predicted to crack unbreakable codes, find new materials for superconductors, and help develop medicine for the next pandemic. These are only some of the envisioned outcomes. So far, the digital computers of today have succeeded in adding integers. These computers do leverage gates to do NAND or XOR operations, for example, but in the end, it is all about integers and their processing in a few specific types of electronic gates. Collaborators on the German quantum computing effort “PhoQuant” are addressing system architecture, operability, and integration in their development of a photonic quantum computer. The effort involves 14 parties, spanning industry, academia, and R&D. Courtesy of Fraunhofer IOF. To be sure, digital computers can do complex calculations. But, at the end of the day, adding integers is what they do: One plus one equals two. Quantum computers are fundamentally different, starting with their bits, known as qubits, which are the quantum analog to classical bits. Like classical bits, qubits can take the values “one” or “zero.” But unlike classical bits, the actual value remains uncertain until it is measured. This means that the qubit is in a state of coherent superposition of the values it can take. The probability that a qubit takes one possible value, or the other, is continuously changing. This becomes a particularly powerful feature when qubits controllably interact with other qubits, creating entanglement and thereby exponentially increasing the dimensionality of quantum computing systems. It allows for the parallel computation of all possible outcomes. Even though the state of a qubit collapses to only a single one at the moment it is measured, dedicated quantum computing algorithms are in position to extract probability information. This makes computations using these algorithms more efficient and more powerful than classical computers achieve. It is obvious to systems designers and engineers that we should leave behind what we know from digital computers and start from scratch when it comes to building quantum computers. Qubits can be mapped to different quantum systems: ions, atomic energy levels or photons. For this reason, different platforms are currently being proposed. One of these proposals is photonic quantum computers. What is a quantum computer made of? One-hundred years ago, physicist Erwin Schrödinger carved out his piece of quantum theory. Schrödinger was looking at electrons, and, more precisely, at their behavior as waves and particles. He came up with an equation to explain how electrons move through space and time. Schrödinger and Paul Dirac were later awarded the Nobel Prize for their findings. At the time, it is doubtful that Schrödinger was thinking about quantum computing. Nevertheless, his equation for the evolution of a wave function is a mathematical way to describe how qubits behave in time and space. Today, people create quantum states with electrons, photons, and other quantum objects; allow them to interact according to the theories of Schrödinger (and others); and measure the result. The function of a quantum computer is to prepare quantum states, controllably apply transformations/gates on them both individually and collectively, and measure the output quantum state. A quantum computer requires a complex setup to perform such processes. Figure 1 shows a scheme in which the actual quantum device is at the lowest level of the overall system, with several layers of software and hardware above it. The top layer in this scheme is the cloud access layer. This refers to the fact that today's quantum computers are designed for remote operation, where a user feeds a specific task into it. There, software is used to translate the problem into parameters that can be sent to the quantum control unit — a hardware module that controls the input of the quantum unit. This control describes, for example, how to prepare the qubits and apply operation in single or multiple qubits, as well as the output of the quantum unit, for example, the registration of the measurement of the results. Level 1 in Figure 1 refers to hardware that is specific to the type of qubits. Figure 1. A real quantum computer needs several layers to transform a task from a user in the cloud (Level 4) into an algorithm (Level 3). A special control system (Level 2, FPGA/ASIC) is required to manage the actual quantum system, including quantum state preparation, quantum processing (Level 1, interferometer), and measurement of the result (Level 1, detector). The result is fed back to the conventional hardware and requires additional software for interpretation. Courtesy of Fraunhofer Institute for Applied Optics and Precision Engineering IOF. What is a photonic computer made of? A digital computer adds and compares numbers based on a hardware element that mainly consists of transistors and capacitors. It relies on currents and electrical charges to make computations. A photonic computer, as its name suggests, uses photons instead. Why photons? Photons currently offer one of the viable pathways to a quantum computer at room temperature. The generation and processing of photons and photonic qubits is routinely done without excessive cooling. And, photonic integrated circuits (PICs) is happening at a fast pace, further clearing a path to scalable integrated circuits for photonic quantum computers with up to millions of qubits. Any photonic computer consists of three major parts: a light source; a processing unit that is most often a multichannel interferometer; and a detection system (Figure 2). For a photonic quantum computer, the light source must produce quantum states or qubits. It should be noted that the qubit in this case is an information unit that propagates through the system and interacts there with other qubits. The photons are measured after such operations, and a result is calculated from these measurements (Figure 2). Figure 2. Scheme of a photonic quantum computer. Courtesy of Fraunhofer Institute for Applied Optics and Precision Engineering IOF. The PhoQuant project Backed by €50 million ($54 million) in funding from the German Federal Ministry of Education and Research (BMBF), a group of German companies and research institutions have joined forces in the PhoQuant project. The effort aims to build a photonic quantum computer, made in Germany, that can be accessed worldwide via cloud services. Funding for the initiative runs through 2026. From a long-term perspective, the project focuses on the development of photonic quantum computing chips, control components, test infrastructure, software, and novel algorithms. The computational core system is planned to work at room temperature. Currently, the detector system is a superconducting nanowire single-photon detector (SNSPD) that works in a small cryostat. All other parts of the system, the laser system, source, demultiplexer, interferometer, control unit, and data acquisition mechanism, are functional at room temperature. While the science behind the planned system is quite clear, its components need to be optimized to meet the project’s ambitious goals. In addition, system integration toward a PIC-based solution presents its own distinct challenge. To accomplish both targets, the project partners installed two setups: one with off-the-shelf components for system testing at the University of Paderborn, and a hybrid platform to enable step-by-step miniaturization and integration of the components into a PIC at Fraunhofer Institute for Applied Optics and Precision Engineering IOF (Fraunhofer IOF) in Jena. The physics of the design chosen by the PhoQuant researchers differs from the qubit model explained earlier; the researchers adopted Gaussian Boson Sampling (GBS) as the main principle of the photonic QC. GBS is based on so-called qumodes, and, more specifically, on squeezed states of light. “Squeezed light” earns its name given that the uncertainty of the field amplitude is squeezed along one of the field quadratures and stretched in the orthogonal quadrature (as is to be expected from the Heisenberg uncertainty principle). Due to such properties, these non-classical states have been used in interferometric measurement, beating the shot noise limit. The concept of squeezed light can also be expressed via a Fock state, or on a “Photon number” basis. For example, the quantum state could have zero, one, two, or N number photons, each of them with a certain probability. In the GBS scheme, N squeezed states are prepared and sent to a multi-mode interferometer. Following the interference, each output will have a different number of photons coming out. Computing the probability of each configuration becomes computational intractable as the number of photons and number of modes in the interferometer increase. This sampling problem can, for example, be mapped to problems in quantum chemistry, to simulate the vibrionic spectra of molecules, to determine an airport gate assignment, optimize a financial portfolio, or to typical pick-up/delivery problems. Finally, the system uses a phase-stabilized dual-frequency (780/1560 nm) pulsed laser system to generate the quantum states. Its signals are fed to a second stage where squeezed states are generated in a periodically poled potassium titanyl phosphate crystal. Processing and measuring the qumodes For this quantum computing mode, the basic components for manipulating the input quantum states are simple: One is a beam splitter and the other a phase shifter. Qubit processing takes place in Mach-Zehnder interferometers, where a beam is first split into two replicas. Each of these passes through a phase shifter. Finally, the two beams reunite at another beamsplitter, with a controlled phase delay between the two replicas. In this way, the outcome of the interference is controlled by the electrical signals sent to the phase delays. An array of Mach-Zehnder interferometers and phase shifters can in principle perform any arbitrary unitary transformation on the input state of the interferometer. Figure 3. Photonic integrated chip (PIC) with 4-mode-interferometers. Courtesy of Q.ANT. To put such an array on a PIC, all components must be transferred to the PIC itself. This raises the question of the right material for the photonic processor. The PhoQuant team’s material of choice is lithium niobate on-insulator (LNOI). This material has several advantages. First, it can be processed in conventional optoelectronic fabs. Second, it has a strong electro-optic coefficient, which allows extremely fast switching (up to GHz rates) of the phase shifters. This is silicon nitride (SiN), which is often used for PIC architectures, requiring the heating of certain elements to achieve phase shift. Heating is slower than electro-optical switching and consumes more power, which also requires post-process cooling. Obviously, the selection of material is much more complex than these considerations alone. The PhoQuant project teams considered both SiN and LNOI for the project, with extensive simulations and testing. The final decision favored LNOI because of its potential in terms of achieving low losses, high switching speeds, and excellent processability. Currently, the group at Fraunhofer IOF is using a PIC platform (Figure 3) from industrial quantum technology and photonics solutions provider (and project partner) Q.ANT, with 4-mode interferometers. An upgrade up to 100 modes is planned. And, to measure the number of photons coming from each output of the interferometer, the team selected to SNSPDs, shown in Figure 4 as a prototype(s) with 20 channels for detection. Figure 4. Prototype of a 20-channel-photon-number-resolving (PNR)-detector. The element uses space multiplexing of superconducting nanowire single-photon detectors (SNSPDs). Courtesy of the University Heidelberg. The future of photonic quantum computers As mentioned, expectations for quantum computers are undoubtedly high. Once they achieve supremacy over digital computers, they may advance to break new ground in materials science, mathematics, or biology. However, this supremacy will be limited to very specific problems. Therefore, it seems likely that quantum computing will be part of a computing system that is comparable to other special purpose chips — such as a GPU or the new iteration of neural engines. Under the PhoQuant initiative, the collaborators are on their way to an integrated photonics-based quantum computer. Individual efforts not only for the quantum system itself, but also for the hardware and software to control and read-out the quantum states. Future computing devices for complex computations will require not only such PICs, but also sophisticated solutions for high-speed data transfer and low power consumption — a well-known advantage of optical computing technology. This may ultimately point to a wider implementation of optoelectronic solutions, as currently developed in projects such as PhoQuant.