News

Computing trends: quantum computers, neuromorphic computers and tap-proof telephony

How Saxony's engineers look to the future

Over the past decades, the semiconductor industry has pushed supposed boundaries of physics and technology further and further: With clever tricks such as double patterning, wet lithography, 3D gate architectures, multi-core processors and structure genesis by means of extreme ultraviolet beams, one nanoworld barrier after another that was believed to be irrefutable fell.

Companies like TSMC and Intel will soon be scratching the 2-nanometer structure mark. The most modern EUV imagesetters with high numerical apertures (High-NA) from the Dutch supplier AMSL - whose equipment also includes Saxon know-how - are expected to help.

Nevertheless, it has been clear for a long time that classical semiconductor technology cannot continue to develop linearly through mere miniaturization. What was still science fiction yesterday is now within reach in the search for alternatives: Computers that are "knitted" similarly to the human brain, computers that can crack almost any code, and the like.

We have taken a closer look at some of the most important computing trends and also want to shed light on what Saxon engineers and researchers are currently working on.

Neuro-electronics: copied from the brain

To this day, the "Von Neumann architecture," which strictly separates signal processing and memory, dominates the world of digital computers. To achieve higher performance, engineers are packing the components in these computers ever more densely and clocking them ever higher. The disadvantages include high power consumption - which could overtax the batteries of future autonomous electric cars - and cumbersome "back and forth" movement of data between the computing unit and memory.

Evolution has "worked out" a more economical and efficient principle: The brain uses its neurons equally to store information and to process it.

The neurons are connected by analog communicating and changing synapses, where the signals are transmitted by pulses. The conductivity of the synapses changes throughout a person's life depending on the number of refreshing impulses - which is why we sometimes forget what we once learned. The brain drastically reduces the energy consumption of network structures that are not needed at the moment. All these principles ensure learning ability, flexibility and energy efficiency: Some tasks that a human brain can solve within the blink of an eye with a power consumption of 20 watts, today's digital computers often manage only approximately as well - and often suck up 1000 watts or more in the process.

Neural networks

simulate these brain principles, at least in part, via software on classic silicon digital computers. They are used primarily for the training and work of "artificial intelligences" (AI). For example, when we start an Internet search query today, a Google neural network is usually pondering in the background.

Neuromorphic networks

go one step further: They reproduce brain structures - at least to some extent - at the hardware level. Here, artificial neurons and synapses connect according to the pulse principle. The European "Human Brain" project, in which the TU Dresden is playing a major role, has even set itself the goal of completely rebuilding the human brain. In practice, we are still a long way from this point in terms of technology. So far, hybrid neuromorphic networks dominate, combining, for example, classic silicon technologies and concepts from nature.

The Saxon contributions

A prominent example is the "SpiNNaker" supercomputers by neuromicroelectronics professor Christian Mayr from TUD: Together with British partners and GlobalFoundries Dresden, he has designed and built brain-like computers based on ARM computer cores. Another example is the neuromorphic Loihi test chips, which Intel first presented in 2018.

A particularly hot research topic, however, is the attempt to reproduce and network artificial neurons and synapses even at the lowest hardware level, following the example of the brain. The synapses, for example, could consist of analog-switched memristors. These are components that combine memory and resistance (memory + resistor) and whose resistance depends on how much current has previously flowed through them. In other words, they are "capable of learning" from the outset.

Ferroelectric transistors, on the other hand, are candidates for both artificial synapses and neurons. In the case of the latter, it is above all the ability to "accumulate switching" that is important. In other words, they must be able to send a new control pulse to their neighboring neuron only after they themselves have received a certain number of suitable input signals.

The TU Dresden, NaMLab gGmbH and the Fraunhofer Nanoelectronics Center CNT, among others, are working on such concepts in Saxony. Contributions to the functioning of an AI in neural networks can also be expected from the Infineon Development Center in Dresden, the "Center for Explainable and Efficient AI Technologies" (CEE AI) in Dresden, the AI research center "Scads AI" Leipzig/Dresden, and other players in the now quite broad AI scene in the Free State.

Recognize patterns in large amounts of data

Applications for neural and neuromorphic electronics include analysis of large amounts of medical or research data, environment recognition for automated and autonomous cars, and other pattern recognition tasks. Hybrid solutions are already emerging, particularly in automotive applications. This could be, for example, some preprocessing of camera images and other sensor data by artificial neurons, which then forward the already abstracted information to the digital on-board computer in the car or to a supercomputer on the roadside (edge cloud) to relieve the binary computers.

Quantum computing: ion traps, diamonds and very low temperatures

When digital computers have to crack a code, simulate a complicated traffic situation, analyze a black hole in space or forecast the future world climate, they more or less stubbornly try out all the possibilities one after the other in their binary cells, which are counted in bits. This can take days, months or sometimes even years. Quantum computers, on the other hand, are composed of "qubits," or "quantum bits." Cells that can assume multiple quantum mechanical states at once. Put simply, they can run through multiple solutions to a computational task simultaneously. This makes them far superior to their digital brethren in some - not all - disciplines: for example, they are capable of cracking encryption based on prime number factorization very quickly. They are also particularly well suited for optimization tasks and simulations.

To date, there are only quantum computers with limited performance and still comparatively few qubits. This is due to the costly error correction, the very unique way of programming them, but most of all due to the elaborate design for the hardware. Several approaches are possible here, for example, cryogenic ion traps, manipulated diamonds, 2D materials, cryogenic chips with current loops more. The quantum computer from technology pioneer IBM, for example, is made of multiple superconducting materials that conduct electricity without resistance at very low temperatures. Others, such as the Leipzig University spin-off "SaxonQ," use doped diamonds in which the spins (quantum mechanical angular momentum) of nitrogen imperfections serve as qubits - without any need for deep cooling at all. In addition to SaxonQ in Leipzig, other players in Saxony are working on quantum computing. These include the new microelectronics research center "Center for Advanced CMOS & Heterointegration Saxony" (Cachs) in Dresden, which is planning its own pilot line for silicon-based quantum chips. In addition, Infineon Dresden, the Fraunhofer Institute for Photonic Microsystems (IPMS) and other partners are involved in projects such as "Quasar", "GEQCOS" and "PIEDMONS", which aim to build their own German quantum processors and computers on various technology paths. And GlobalFoundries Dresden is also involved in some process steps for quantum chips, specifically for the Californian project partner "Psiquantum".

Quantum Communication

Closely related to quantum computing and also dependent on novel electronics and optoelectronics is quantum communication. It is intended to enable tap-proof and tamper-proof data connections and telephony based on quantum entanglements. Saxony has recently caught up noticeably in this segment. For example, the announcement by the new quantum tech lab in the Dresden branch of the Fraunhofer Institute for Integrated Circuits (IIS) "Development of Adaptive Systems" (EAS) that an in-house quantum communication test track was already operating there caused a sensation. In addition, TU Dresden is collaborating on the federal projects "Quiet" and "QD-CamNetz", which aim at quantum networks for the "Internet of Things" (IoT) as well as hybrid 5G and quantum data networks based on the campus network principle.

Chiplets & Co.: The Systems Approach

In addition to such leap innovations, which tend to have a long-term impact, microelectronics engineers are also looking at solutions for the near future. These include the systemic approach. This approach applies Moore's Law*, according to which the complexity, number of transistors and power of integrated circuits doubles every one to two years, to the overall electronic system. In addition to new component architectures, the focus is now also shifting to everything that the industry used to classically assign to the "back end" and relocate to Asia. One focus topic is packaging technology, which contacts and interconnects individual chips and protects them from external influences.

One example is chiplet technology, which makes it possible to manufacture several circuits or circuit segments in differently sophisticated and expensive processes and even in different factories, but then organically combine them into a whole. This also includes the three-dimensional contacting of several circuits, sensors, actuators and other elements to form a system or the connection of superimposed wafers from different production chains. In Saxony, the Fraunhofer facilities IPMS, Assid, the Center for Advanced CMOS & Heterointegration Technologies Saxony and other players are working on these technologies.

In terms of economic policy, this trend has a very special dimension: If the boundaries between the core processes ("frontend") and the packaging and final assembly ("backend") in microelectronics dissolve, this could lead to some wafers processed here no longer being transported by airmail to Singapore, Malaysia or other locations specializing in backend processes, but being processed immediately in Europe. In other words, the industry may relocate part of the backend value chain back here.

Automotive

As an important automotive state and pioneer for electromobility, more and more research projects in Saxony are focusing on the future of mobility. These include innovative radar sensors and AI concepts, on which Infineon Dresden and its Development Center are working, among others, or the car factory of the future networked with its own factory cloud, on which the VW Research Center at the Universal Works Dresden is conducting research. Teams from the TU Chemnitz, the Fraunhofer Institute IWU and the HTW Dresden, among others, are targeting innovative digital manufacturing processes. The Fraunhofer Institute IVI and TU Dresden are among those working on the networked traffic of the future.

Sensors

What just existed only on the TV spaceship "Enterprise" is becoming reality in Saxony: With massively miniaturized near-infrared spectrometers, it is now possible to determine the composition and quality of drugs, freshly brewed beer or T-shirts without contact from a distance with "tricorder"-like devices. What was initiated years ago by the micromirror development at Fraunhofer-IPMS has meanwhile resulted in microelectromechanical systems (MEMS) ready for series production and several spin-offs such as Hiperscan or Senorics. It is to be expected that, in addition to pharmacies, brewers and household appliance manufacturers, quite a few other industries will discover this innovation from Saxony for themselves.
However, the micro spectrometers are just one example of many of the upheavals that are in the offing in the sensor sector and especially in Saxony's value chains. Other examples of innovation include the aforementioned AI-supported radar sensors from Infineon or the "Universal Sensor Platform" (USeP), which is currently picking up speed. USeP was developed in a joint project by GlobalFoundries, the Fraunhofer institutions EAS, Assid, Enas, the IPMS, the spin-off Sensry and other partners for SMEs that want to use it to upgrade their own IoT and "Industrie 4.0" products. In the meantime, the first products based on this technology have reached the market - including an electronic athlete tracker and systems for "predictive maintenance".

Energy systems of tomorrow

Circuits that operate in a particularly power-saving manner have gone from being a niche product to a growth driver, and not just since the proclamation of the German energy turnaround. And Saxony has gained a competitive edge in this segment. Since the launch of the "Cool Silicon" cluster in 2009, research into energy-efficient microelectronics has been stepped up here. An important building block for this is now the 22FDX technology from GlobalFoundries Dresden, which is meeting with growing interest from more and more industries. The special funding for "Important Projects of Special European Interest" (IPCEI) helped and is helping to open up new applications for it.
One technology path that Saxon players have taken quite successfully on the way to tomorrow's energy systems is power electronics. Better rectifiers and inverters as well as other semiconductors for high voltages and strong currents are, in fact, increasingly needed to make the transition of the automotive industry from "burners" to "electric cars". This includes the power semiconductors Infineon manufactures in its Dresden factory, the gallium nitride line from X-FAB in Dresden, automotive electronics from Bosch in Dresden, but also the gallium arsenide diodes from 3-5-PE.

All of the technologies and innovations mentioned are still in the start-up curve, but they will undoubtedly play a growing role in "Silicon Saxony" in the coming years.

Even beyond microelectronics in the narrower sense, many engineers and scientists in the Free State are working on forward-looking energy systems whose production plays a role in parts of circuits, sensors, automation solutions and other high technologies "Made in Saxony". One example is the high-temperature electrolysers from Sunfire. In the future, they will produce eco-hydrogen on a large scale throughout Europe using solar and wind power - and do so highly efficiently with efficiencies of over 80 percent. In order for these innovative, albeit still quite expensive systems to become established on the market, Sunfire has to cut costs. High-automation technologies from Xenon and other partners, which are currently being developed, will help.

The costs and efficiency of a new generation of flywheel energy storage systems, which TU researchers are testing in Boxdorf, also stand or fall with mechanical engineering, plant engineering, hydraulics, vacuum technology, electrical engineering and sensor technology from the Free State. Here and elsewhere, Saxony is constantly breaking new technological ground - and, just as in microelectronics in the past, pushing the boundaries of what is possible.


This article was first published as part of our NEXT magazine "In the spotlight: Microelectronics". https://bit.ly/3UuYoy6

To the complete edition of the magazine

Related articles:

A journey across the periodic table - What the chips of the future are made of

Germany‘s microelectronics landscape - Europe‘s semiconductor heart beats here

European Chips Act: Good, but good enough?