Information processing device

Abstract

The present invention relates to an apparatus (12) for processing information by evaluating randomly moving tokens corresponding to particles or quasi-particles, the apparatus (12) comprising: an input interface (20) for receiving an input signal; a carrier (22) supporting a plurality of pathways (30) for the randomly moving tokens and at least one join (32) for connecting two pathways and allowing a first token in a first one of the two connected pathways to pass through the join when a second token from a second one of the two connected pathways is present within the join; an excitation unit (26) for applying a stimulus to the tokens in at least one of the plurality of pathways, the stimulus acting to increase the random motion of the tokens in a direction substantially parallel to the carrier; and an output unit (28) for outputting an output signal based on the position of the at least one token after a predetermined or dynamically adjusted time has elapsed. The present invention further relates to a sensor system (10) for detecting a physical phenomenon in an environment.

Description

[Technical field]

[0001] The present invention relates to a device for processing information by evaluating randomly moving tokens corresponding to particles or quasiparticles. The present invention further relates to microelectronic systems requiring information processing, for example sensor systems for detecting physical phenomena in the environment. [Background technology]

[0002] Thermal fluctuations are becoming an increasing problem in modern digital electronic computers as computing devices become smaller. As the size of these devices and circuits is scaled down, the relevant energy scale for computation becomes comparable to that of thermal fluctuations, which means that deterministic behavior is no longer guaranteed. Traditional strategies to combat the effects of thermal fluctuations, such as noise suppression and error correction, come at the expense of larger devices, significantly limited computation speed, and high power consumption. In recent years, unconventional computing approaches have been attracting attention for their novel ways of circumventing these problems.

[0003] One approach in this respect is the so-called token-based computing. Tokens are discrete signal carriers that move in a circuit with specific transition rules to perform calculations. In Brownian computing, this movement occurs due to the stochastic Brownian motion of particles or quasiparticles due to thermal fluctuations. One advantage of this Brownian computing (or token-based computing) is its low power consumption. It makes it possible to process information even with only small amounts of power available. This is especially interesting in application scenarios that rely on energy harvesting for self-sufficient operation. A possible scenario is a monitoring application where a sensor measures a quantity and acts according to changes in this quantity. For example, such a processing device could be used for indoor climate monitoring.

[0004] A promising candidate for a token is the magnetic skyrmion, a topologically stable magnetic vortex that exhibits quasiparticle behavior and has recently been shown to exhibit thermodiffusion at room temperature.

[0005] In this context, Jibiki et al.'s "Skyrmion Brownian circuit implemented in continuous ferromagnetic thin film" (2019) is relevant for ultra-low power Brownian computers. Specifically, a skyrmion Brownian circuit implemented in a continuous ferromagnetic film patterned with SiO2 capping to stabilize skyrmion formation is disclosed. The patterned SiO2 capping controls the saturation field of the ferromagnetic layer and forms a wire-like skyrmion potential well to stabilize skyrmion formation in the circuit. Furthermore, this patterned SiO2 capping is used to implement a Y-junction hub circuit that does not exhibit significant pinning sites at the junction, unlike conventional etched hubs. Efficient control of skyrmion-based memory and logic elements to move closer to the realization of a Brownian computer is disclosed.

[0006] "Brownian Circuits: Designs" by Lee et al. (2016) discusses reduced-complexity Brownian circuits and presents designs for circuits with functionality such as counting, conditional testing, memory, and arbitration for shared resources. Furthermore, the possibility of implementing Brownian circuits with single-electron tunneling techniques is discussed.

[0007] In "Brownian motion of skyrmion bubbles and its control by voltage applications" (2018) by Nozaki et al., the dynamics of skyrmion bubbles in a W / FeB / Ir / MgO structure is studied with the aim of using the thermally activated random walk of skyrmion bubbles for logic operations (i.e., token-based Brownian computing). In addition to observing the Brownian motion of skyrmion bubbles, it is shown to electrically control the diffusion constant by applying a voltage. The developed technique will be useful for various kinds of skyrmion-based spintronic devices as well as Brownian computing.

[0008] Sanz-Hernandez et al. (2017) in "Fabrication, Detection, and Operation of a Three-Dimensional Nanomagnetic Conduit" presents a 3D nanomagnetic system fabricated by 3D nanoprinting and physical vapor deposition that acts as a conduit for domain walls. Domains formed at the substrate level are injected into a 3D nanowire where they are controllably trapped by a vector magnetic field. Dark-field magneto-optical techniques are also presented to measure different regions of individual 3D nanostructures in parallel and separately.

[0009] One drawback of the Brownian computing (i.e. token-based computing) method is that information processing depends on the movement of particles or quasi-particles. On the one hand, this can lead to slow processing, which prevents application in fields that require a certain degree of calculation speed. Simulations have shown that even for a rather simple circuit such as a half adder, the average calculation time for one operation can be on the order of several minutes. On the other hand, the use of thermal fluctuations and / or random motion makes the processing time non-deterministic. The non-deterministic calculation duration, combined with the long calculation time, poses a challenge to the application of the Brownian computing (i.e. token-based computing) method in almost all application fields. Summary of the Invention [Problem to be solved by the invention]

[0010] In light of the above background, the present invention faces the problem of enabling the application of token-based computing schemes in further application scenarios, specifically, the problem of speeding up calculations in token-based computing schemes while avoiding excessive power consumption. [Means for solving the problem]

[0011] To solve this problem, one aspect of the invention relates to an apparatus for processing information by evaluating randomly moving tokens, corresponding to particles or quasiparticles, the apparatus comprising: an input interface for receiving an input signal; a carrier supporting a plurality of pathways for randomly moving tokens and at least one join that connects two pathways and allows a first token in a first of the two connected pathways to pass through the join when a second token from a second of the two connected pathways is present in the join; an insertion unit for inserting a token into a pathway based on an input signal; an excitation unit for applying a stimulus to the token in at least one of the plurality of pathways, the stimulus acting to increase random motion of the token in a direction generally parallel to the carrier; an output unit for outputting an output signal based on the position of the at least one token after a predetermined or dynamically adjusted time has elapsed; Includes.

[0012] Another aspect of the present invention relates to a sensor system for detecting a physical phenomenon in an environment, the sensor system comprising: A device as defined above, a sensor configured to detect a physical phenomenon and generate an input signal based on the physical phenomenon; Includes.

[0013] Preferred embodiments of the invention are defined in the dependent claims. It will be appreciated that the claimed devices and systems have similar and / or identical preferred embodiments to the claimed devices and systems, in particular the devices and systems defined in the dependent claims, and to the devices and systems disclosed herein.

[0014] The invention is based on the idea that the random movement of particles or quasi-particles representing the token in a token-based computing device is stimulated by an excitation unit. An external additional stimulus is applied that increases or induces the random movement of the token in a direction approximately parallel to the carrier, thereby making it possible to accelerate the process.

[0015] Pathways are arranged on the carrier along which particles or quasi-particles corresponding to the tokens can move randomly. Joins connecting the pathways allow the passage of the token only if another token is present. In particular, the joins can be so-called C-joins. The pathways are connected to form a network that performs the desired computing operation. An input signal is received, which contains the information to be processed. The input signal forms the basis for inserting the token into the network. The movement of the token through the network and the functionality of at least one join result in the processing result. That is to say, the position of the token after the processing time has elapsed represents the result of the information processing. An additionally applied stimulus amplifies the movement. This movement is accelerated. Or, if the token does not initially show any movement, for example due to environmental conditions, the movement of the token is increased by inducing this movement. In other words, the stimulus and the stimulated movement are independent of the thermal diffusion and may be additionally superimposed with the (potentially still present) thermal diffusion. Preferably, the carrier is approximately flat or shaped in the shape of a plane. The stimulus is applied parallel to this plane. The carrier may represent a substrate or other structure that holds the circuit formed by the pathways and at least one join. The carrier may in particular be made of a metal, an insulator, a semiconductor, a ceramic, Teflon, or a polymer. The carrier does not necessarily have to be conductive. Depending on the material, the pathways and the joins may be implemented, for example, by structuring or treating the material.

[0016] This additionally applied stimulus allows for faster processing compared to previous methods of token-based computing. It allows for information to be processed at a higher processing speed. The stimulus is applied in a direction approximately parallel to the carrier, allowing for energy-efficient stimulating of movement. Furthermore, it allows for the stimulus to be modified and / or adapted according to predefined or constantly updated criteria. The stimulus may be applied to the entire network or to parts of the network.

[0017] In contrast to the concept of using ratchets, as discussed in Lee et al., "Brownian Circuits: Designs" (2016), the excitation unit of the present invention is configured to apply a stimulus that acts to increase the random movement of the token. The ratchet corresponds to a component included in a specific position of the pathway, which accelerates the movement of the token in a given direction as it passes. As the token passes, the movement of this token is accelerated in the given direction of the specific ratchet. Thus, the ratchet does not increase the random movement, but rather induces or supports a directed movement. Furthermore, the ratchet is usually implemented as a passive element that does not require an external energy supply. On the other hand, the action of the excitation unit is not limited to a specific position, does not prefer a specific direction, and increases the random movement of the token in a direction approximately parallel to the carrier. For this purpose, the excitation unit preferably corresponds to an active element that applies energy.

[0018] The sensor system of the invention makes it possible to detect a physical phenomenon in the environment. In particular, a parameter representative of this physical phenomenon is measured. For example, light intensity, temperature, gas concentration, etc. may be detected. The output signal of this sensor is evaluated by the device of the invention. In particular, passive sensors may be used that do not require an external energy source. The device of the invention may also operate with very limited power consumption. This approach can therefore be more energy efficient than previous sensor systems.

[0019] In a preferred embodiment, the device is configured to evaluate particles or quasiparticles exhibiting Brownian diffusive motion as tokens. In particular, magnetic skyrmions may be used as tokens. Particles or quasiparticles exhibiting Brownian diffusive motion are particles that exhibit random fluctuations in their respective positions, and no external trigger is required for this motion. In particular, magnetic skyrmions may be used. These quasiparticles exhibit advantageous properties with regard to their motion as well as with regard to stimulability. Efficient stimulation can be achieved.

[0020] In a preferred embodiment, the excitation unit comprises a pulse unit for applying a current to the carrier and / or to at least one of the pathways. The carrier preferably comprises a conductive material layer enabling the spin torque effect (in particular the spin orbit torque effect), to which a current is applied. The pulse unit allows a voltage to be connected to the carrier, whereby a current flows through the carrier. It is also possible that a voltage is applied to at least one pathway, whereby a current flows through this pathway. To achieve the required conductivity, a conductive material may be used. If this conductive material is additionally susceptible to the spin torque effect or the spin orbit torque effect, this leads to an energy-efficient stimulation of the movement of magnetic particles or magnetic quasiparticles. The random movement of the tokens is efficiently multiplied. An option is given to efficiently perform the stimulation of the movement of the tokens. Furthermore, an energy-efficient stimulation is achieved.

[0021] In a preferred embodiment, the excitation unit comprises a magnetic field unit for applying a magnetic field or a magnetic field gradient to the carrier. This magnetic field unit preferably comprises at least two coils arranged on different sides of the carrier. The magnetic field unit is preferably configured to generate a magnetic field having components in at least two directions that are not parallel. In particular, the generated magnetic fields may be perpendicular to each other. Different sides of the carrier therefore specifically mean positions arranged in a direction parallel to the (approximately planar) carrier. A magnetic field or a magnetic field gradient is generated that stimulates the movement of the token. This magnetic field or magnetic field gradient specifically allows the excitation of magnetic particles or magnetic quasi-particles or particles or quasi-particles with magnetic properties. One efficient implementation of the magnetic field unit is the use of coils arranged to induce a magnetic field or a magnetic field gradient.

[0022] In a preferred embodiment, the excitation unit comprises an electric field unit for applying an electric field to the carrier. The carrier comprises an electrical insulating layer for propagating the electric field. Preferably, a high-K dielectric material layer may be used. The implementation of the unit for applying an electric field to the carrier requires little additional hardware. The electric field can in particular propagate through a specially designed electrical insulating layer. This layer forms part of the carrier and preferably comprises a high-K dielectric material. For example, a layer of a material comprising hafnium silicate and / or hafnium oxide may be used. Energy-efficient excitation of the tokens is achieved.

[0023] In a preferred embodiment, the excitation unit comprises an electromagnetic field unit for generating a high-frequency electromagnetic field. This electromagnetic field unit preferably comprises an antenna. Additionally or alternatively, it is possible to stimulate the tokens by a high-frequency electromagnetic field. This electromagnetic field may for example be generated by one or more antennas. The antennas are preferably arranged in the immediate spatial vicinity of the carrier. It is also possible that the one or more antennas are integrated with the carrier. The use of a high-frequency electromagnetic field has the advantage that the stimulation of the tokens can be efficiently performed over a wider area on the carrier.

[0024] In a preferred embodiment, the output unit comprises a magnetic sensor which detects the change in magnetization or the change in the magnetic field at the output location on the carrier. If magnetic particles / quasiparticles or particles / quasiparticles with magnetic properties are used, these can be detected by the magnetic sensor. This magnetic sensor makes it possible to detect whether a token is present at the output location. This output location corresponds to the location that the token can reach depending on the input signal, i.e. the location where the token was inserted. By using the magnetic sensor, it is possible to reliably detect the token at the output location with low energy consumption.

[0025] In a preferred embodiment, the carrier comprises a multi-layer thin-film system arranged on a semiconductor wafer. The wafer preferably comprises an insulating top layer. The thin-film system preferably comprises a magnetizable material. The pathways and at least one join are preferably manufactured in the multi-layer thin-film system in an etching process and / or a lithography process. When manufacturing a multi-layer thin-film system on a semiconductor wafer, existing manufacturing techniques may be applied. An efficient and effective manufacturing process may be achieved. In particular, standard etching and / or lithography processes may be used to manufacture the pathways and the joins in the multi-layer thin-film system.

[0026] In a preferred embodiment, the insertion unit comprises a nucleation unit which nucleates tokens at an input location connected to a pathway. In particular, magnetic skyrmions may be nucleated. This nucleation therefore corresponds to forming or generating particles or quasiparticles at a desired location on the carrier. The particles or quasiparticles are generated at the input location and move from there through the pathways and joins that form the network.

[0027] In a preferred embodiment, the input interface is configured to receive an input signal from a sensor. In particular, the device may be used to process information measured by the sensor. For example, the measurement signal may be processed in order to evaluate whether a certain condition is met. A possible application area is the monitoring of parameters such as illuminance, gas concentration, etc. The input is processed by the device, for example to check whether a certain condition is met. An energy-efficient monitoring of the parameters is feasible.

[0028] In a preferred embodiment, the input interface is configured to receive a stimulation signal. The excitation unit is configured to apply a stimulation based on the stimulation signal. The excitation unit is preferably configured to adapt the intensity of the stimulation based on the stimulation signal. That is to say, it is possible to apply the stimulation on demand. If an additional acceleration provided by the excitation unit is required, the stimulation signal is applied. If no further acceleration of the process is required or if no energy is available, no stimulation is applied. The excitation unit is configured to be activated or deactivated on demand. By activating the excitation unit and by accelerating the random movement of the tokens, the process is accelerated. It is thus possible to accelerate the process on demand. This control is performed by the stimulation signal. It is also possible that the intensity of the applied stimulation is adapted or controlled based on the stimulation signal. For example, if a very fast processing is required, a stronger stimulation may be applied. An optimization of the energy consumption of the device depending on the application scenario is achieved.

[0029] In a preferred embodiment, the excitation unit is configured to apply a stimulus based on the amount of energy present in an energy source connected to the energy harvesting unit, which generates electrical energy from the environment. In particular, it is advantageous to utilize a stimulus that leads to additional acceleration when additional energy is available. In an energy harvesting scenario, there may be more or less available power. When there is available power, a stimulus may be applied, which can accelerate the processing of information in the device. When there is no available power, no stimulus is applied and the processing is performed slower.

[0030] In a preferred embodiment, the sensor system includes an energy harvesting unit that generates electrical energy from the environment. The energy may be harvested from heat, light, or other phenomena. Typically, the energy harvesting unit is in each case configured to harvest energy depending on its availability. This energy may then be used to apply an additional stimulus to the excitation unit of the device of the invention.

[0031] In this specification, the tokens corresponding to particles or quasiparticles may in particular be magnetic skyrmions. However, other particles / quasiparticles may also be used, in particular colloids exhibiting Brownian motion or susceptible to movement due to excitation by electric or magnetic fields. It would also be possible to use the concept of the invention based on defect electrons, photons, magnons or excitons in semiconductors as tokens in general, even if they do not exhibit Brownian motion. The device of the invention is suitable for processing information, which in particular means that it is capable of processing digital signals and of carrying out calculations on the basis of said digital signals. The sensor of the invention is capable of detecting a physical phenomenon. The physical phenomenon may be any type of change in a parameter of the real world. The physical phenomenon may mean, for example, illumination, gas concentration, temperature, distance of an object, etc. This physical phenomenon may be measured by a sensor. The sensor in particular outputs a signal depending on this phenomenon. For example, this signal may be proportional to the phenomenon. The signal may be an analog or digital signal.

[0032] These and other aspects of the invention will be apparent from and elucidated with reference to the embodiments described hereinafter. [Brief description of the drawings]

[0033] [Figure 1] 1 shows a schematic diagram of a sensor system according to the invention; [Diagram 2] 1 shows a schematic diagram of an apparatus for processing information according to another aspect of the present invention. [Diagram 3] A schematic diagram of the pathway and join network is shown. [Figure 4] 1 shows a schematic diagram of a carrier and an excitation unit with a pulse unit. [Diagram 5] 13 shows a schematic diagram of an alternative embodiment of a pulse unit. [Figure 6] 1 shows a schematic diagram of a carrier and an excitation unit with a field unit. [Figure 7] 1 shows a schematic diagram of a carrier and an excitation unit with a magnetic field unit. [Figure 8] 1 shows a schematic diagram of a carrier and an excitation unit with an electromagnetic field unit. [Figure 9a] ~ [Figure 9b] A schematic diagram of two different configurations of pathways and joins on a carrier is shown. [Figure 10] 1 shows a schematic diagram of another embodiment of a half adder. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0034] FIG. 1 shows a schematic representation of a sensor system 10 for detecting a physical phenomenon in an environment. The sensor system 10 includes a device 12 for processing information and a sensor 14 for generating an input signal based on the physical phenomenon. The sensor system 10 may be used, for example, to monitor the carbon dioxide concentration in a building. The device 12 may process the input signal and provide an output signal indicating whether the building needs to be vented due to elevated carbon dioxide concentration. One requirement for these types of monitoring applications is low energy consumption. To achieve the desired functionality, the sensor system 10 further includes an energy harvesting unit 16 with an energy source 18. For example, a temperature gradient may be exploited by a Peltier element to charge a capacitor that serves as the energy source 18.

[0035] Fig. 2 shows a schematic diagram of an apparatus 12 for processing information according to the invention. The apparatus 12 comprises an input interface 20, a carrier 22, an insertion unit 24, an excitation unit 26 and an output unit 28. The apparatus 12 corresponds to a token-based computing device (i.e. a Brownian computing device) in which randomly moving tokens, corresponding to particles or quasiparticles, move through pathways 30 and joins 32 on the carrier 22 to perform computations. The tokens move through a network. A join 32 is connected to two (or more) pathways 30. In particular, the join 32 corresponds to a C-join connected to four pathways 30. A token arriving at a join 32 can only pass through this join 32 if, while the token is present at the join 32, another token from another pathway 30 also arrives at the join 32. The layout of the network of pathways 30 and joins 32 makes it possible to perform computational or information processing tasks.

[0036] Of course, in the illustrated example, the pathways 30 and joins 32, as well as the network of insertion units 24 and output units 28, are only diagrammatically illustrated. The network extends in two dimensions on the carrier 22, as indicated by dots in FIG. 2. The size of the network depends on the computations that need to be performed by the network or device, respectively. The network corresponds to a combinational logic or combinational circuit (a network of logic gates) whose outputs are processed to determine the corresponding inputs. A number of structures corresponding to logic gates are arranged on the carrier 22. In addition to the pathways 30 and joins 32, other components (such as hubs and ratchets) are usually also part of the network.

[0037] The use of randomly moving tokens allows for the implementation of a computing device that is very energy efficient. In particular, it is possible to exploit particles / quasiparticles exhibiting Brownian motion that do not require any external energy input to move through the network of pathways 30 and joins 32. Preferably, the joins 32 are the only active components that require (small amounts of) external energy to perform their operations.

[0038] In particular, a magnetic skyrmion may be used as a token. The input interface 20 is configured to receive an input signal. The input signal thus represents information to be processed. In particular, the received signal may be digital or analog. The input interface may for example correspond to a wired connection to a sensor or the like.

[0039] In the insertion unit 24, tokens are inserted into the pathways based on the content of the input signal. In particular, the nucleation unit 25 allows magnetic skyrmions to nucleate at an input location 34 connected to the pathway 30. Nucleation can thus be based, for example, on a magnetic field sweep or on an effective spin-polarized current in the defect. The tokens can then move from the input location 34 through the pathway 30 and the network of joins 32.

[0040] The output unit 28 is configured to determine an output signal based on the position of at least one token after a predefined or dynamically adjusted time. In particular, it is possible to determine whether a token is present at the output location 36 on the carrier 22. For example, if the token is a magnetic skyrmion, the change in magnetization or magnetic field at the output location 36 may be measured. This measurement may therefore be performed periodically or at dynamically adjusted time intervals. The movement of the token through the network is based on a random movement, so the time taken to perform the calculation is non-deterministic. For example, if the Brownian motion of particles is utilized, the calculation speed depends on the temperature and other parameters. For example, a magnetic sensor 38 mounted on the bottom of the carrier 22 may be used, as shown diagrammatically by the dashed line. It is possible to directly detect the magnetic field of the skyrmion. Additionally or alternatively, it is possible to detect the magnetization (for example in a tunnel magnetoresistance element or based on a magnetic microscope), in particular by utilizing the magneto-optical Kerr effect.

[0041] The excitation unit 26 is configured to apply a stimulus to the tokens in the network of pathways 30 and joins 32. This stimulus acts to increase the random movement of the tokens in the network in a direction substantially parallel to the carrier 22 shown in FIG. 2. The carrier 22 is substantially planar. The excitation unit 26 increases the movement of the tokens by accelerating or directing their motion in the x-direction and / or y-direction. This stimulus makes it possible to achieve faster processing of information. The movement of the tokens through the network is accelerated. The tokens move faster in the network.

[0042] FIG. 3 shows a schematic of an exemplary network of pathways 30 (lines) and joins 32 (squares, which may also be called C-joins). In particular, the functionality of a half adder is shown. Two bits are added. In this example, two bits are fed into the network via four token input positions. The first two input positions 34 on the right represent a bit. A token is inserted into one of the two input positions 34 on the right if that bit is a one (in:1) or if that bit is a zero (in:0). Similarly, a second bit is fed into the lower input position 34. Pathways 30 and joins 32 connect the input positions 34 to output positions 36. Again, the output position 36 on the left represents the first bit of the output signal, and the output position 36 on the top represents the second bit of the output signal. The joins 32 provide the functionality to capture the token for a predefined length of time when the token enters the join. If, within this predetermined time, another token enters the join 32 coming from another pathway connected to the join 32, then both tokens can pass through the join 32. If a second token does not enter the join 32 within this predetermined time, the first token can leave the join in the direction it entered.

[0043] In the illustrated example, there is also a hub 40 (circle) in the network, which allows movement of tokens entering the hub in all connected pathways 30. Additionally, the illustrated network includes an (optional) ratchet 42 (arrowhead), which corresponds to a component that accelerates the movement of tokens passing through it in one direction. In other words, the ratchet 42 prioritizes the movement of tokens in a particular direction.

[0044] As shown in Fig. 3, the stimulus is applied in a direction approximately parallel to the carrier. The four arrows in Fig. 3 indicate four different possible stimulus directions. Preferably, the stimulus is applied in a direction parallel to the arrangement of pathways on the carrier. Usually, pathways are arranged in two orthogonal directions (X direction and Y direction) for reasons of manufacturing efficiency. And it is preferable that the stimulus is also applied in these directions.

[0045] Therefore, stimulating the token movement is preferably done randomly, i.e. the direction of the stimulation (parallel to the carrier) varies randomly. For example, one stimulation pulse can be applied in a first direction and then a second stimulation pulse can be applied in another direction. These stimulation pulses therefore do not necessarily have to be of the same intensity. However, it is preferred that the direction in which the token movement is stimulated is chosen arbitrarily.

[0046] Fig. 4 shows a schematic representation of one implementation option of the excitation unit 26. Fig. 4 therefore shows a cross-sectional view of the carrier 22. In the example shown, the carrier 22 comprises a multi-layer thin-film system 50 arranged on a semiconductor wafer 52. The pathways and at least one join are implemented in the multi-layer thin-film system 50.

[0047] The multi-layer thin film system 50 is preferably magnetic. One example of such a multi-layer thin film system is Ta(5) / Co 20 Fe 60 B 20 (0.9) / Ta(0.08) / MgO(2) / Ta(5), where the numbers in parentheses are the thicknesses of each layer in nanometers. Semiconductor wafer 52 is preferably a silicon wafer with an oxidized surface representing the insulating top layer.

[0048] One way to fabricate structures (corresponding to walls forming the pathways) on the substrate is to remove (etch) the outer magnetic layer of the structure. Another way is to deposit material that geometrically shapes the pathways. The structuring can then be realized by standard lithography. For example, optical methods based on electron beams or other lithography processes can be used. Ratchets can be realized, for example, in the form of triangular geometries. For the joins, a vertically oriented electric field can be used to locally control the diffusion of the tokens.

[0049] FIG. 4 shows an embodiment in which the excitation unit 26 includes a pulse unit 54 for applying an electric current to the carriers 22. In particular, an electric pulse is applied to a conductive material layer 56 that enables the spin torque effect. A current is injected into this conductive layer. This conductive material layer 56 may in particular be implemented in the form of a continuous thin film of a heavy metal. For example, tantalum, platinum or tungsten may be used. Then, on top of this layer, a structuring of pathways and joints may be applied. The current injected into the conductive material layer 56 results in the stimulation of the movement of all tokens in all pathways, joints and other components.

[0050] Alternatively or additionally, it is also possible to inject a current into the pathways and / or other elements in which the token is located, as shown in Fig. 5. The excitation unit 26 comprises a pulse unit 54 connected to at least one of the pathways or other structures of the multi-layer thin film system 50 through which the token moves on the carrier 22. This allows to exercise further control over the stimulation, in particular to increase the stimulation exclusively for the movement of the token in selected pathways and joins, if desired.

[0051] 6 shows another example of the excitation unit 26. The excitation unit 26 includes an electric field unit 62 for applying an electric field to the carrier 22. In the illustrated example, an (optional) electric field-propagating electrical insulation layer 60 forms part of the carrier 22 and is arranged on the semiconductor wafer 52. This electrical insulation layer 60 preferably includes a high-K dielectric material. The electric field may therefore be applied in particular on the basis of electrodes arranged on two horizontally opposite sides of the carrier 22 or the multilayer thin-film system 50. The electric field unit 62 including the electrodes is configured to generate an electric field in a plane parallel to the carrier 22. In contrast to the application of a current, the application of an electric field does not lead to a current flow.

[0052] FIG. 7 shows a schematic representation of another example of the excitation unit 26. The carrier 22 includes a semiconductor wafer 52 and a multilayer thin film system 50 on which the pathways, joins and other structures are mounted. The excitation unit 26 includes a magnetic field unit 58 which applies a magnetic field or a magnetic field gradient to the carrier 22. The generation of the magnetic field may be performed, for example, by using local coils or striplines deposited on the substrate. By adapting the geometry of the coils, it is possible to generate a magnetic field gradient. In particular, the magnetic field unit 58 may include two coils. The coils are thus preferably arranged on different sides of the carrier 22, but are not aligned with respect to each other (i.e. they generate magnetic fields that are not parallel to each other). In particular, the coils may be configured to generate magnetic fields or magnetic field gradients oriented perpendicular to each other.

[0053] 8 shows yet another example of the excitation unit 26. The excitation unit 26 includes an electromagnetic field unit 64 that generates a high-frequency electromagnetic field. Thus, the multilayer thin-film system 50 is also placed directly on the semiconductor wafer 52 without any additional functional layers in between. The token movement is increased based on the high-frequency alternating electromagnetic field generated by the electromagnetic field unit 64. Thus, it is possible to stimulate the token movement based on a magnetic field component or an electric field component.

[0054] Naturally, it is also possible to combine the various options for the application of the stimuli and the implementations of the excitation unit 26 shown in Figures 4 to 8. A combination may allow a better control of the stimulation and an adaptation to specific requirements.

[0055] Figures 9a and 9b show diagrammatically two layouts of a half adder realized based on multiple pathways 30 and joins 32 (compare Figure 3). Both figures thereby show the same layout, but with different lengths of the pathways 30 in the x-direction.

[0056] As outlined above, the stimuli are applied randomly. Specifically, the movement of the tokens in the network is stimulated in randomly varying directions (approximately parallel to the carrier). In the illustrated layout, the pathways are oriented in either the x or y direction, and in this case the stimuli are preferably applied in one of four directions: left / right / up / down (±x and ±y directions in the figure). However, the probability of having a stimulus in each direction does not necessarily have to be the same. Specifically, the probability of having a stimulus in each direction: left / right / up / down does not necessarily have to be 25% / 25% / 25% / 25%, and may deviate from such an even distribution, for example, 20% / 22% / 30% / 28%. Then, the calculation is performed correctly, even if the increase in calculation speed is potentially reduced compared to the even distribution case. However, the implementation of the stimuli may be easier if all directions do not necessarily have to be completely even.

[0057] Moreover, depending on the layout, it may even be useful to adapt the application of stimuli depending on the layout. As shown in Figures 9a and 9b, the length of all pathways in the left-right direction (x-direction) is twice as long in Figure 9b as in Figure 9a. In terms of computational implementation, due to this larger left-right distance, it may be possible and useful to adapt the probability of stimulation in this direction (e.g. by choosing 33% / 33% / 17% / 17% for left / right / up / down).

[0058] FIG. 10 shows a schematic of another layout of a half adder, which may allow easier implementation since no crossing of pathways is required. The network includes pathways (lines), joins (squares), ratchets (arrowheads), and hubs (circles). On the left side of FIG. 10, the first bit of the input is calculated. On the right side, the second bit is calculated. In the illustrated layout with no crossings, any two joins on the left and right with the same Roman numeral are connected to each other (e.g., by striplines) and communicate with each other to perform the function of the join. That is, each of the two squares with the same Roman numeral performs the function of one join.

[0059] The invention described herein is applicable, for example, to the systems discussed in Raab et al., “Brownian reservoir computing using geometrically confined skyrmions,” (2022) (hereinafter Reference 1) (preprint available at https: / / download.klaeui-lab.de / skyrmion-reservoir-computing / ).

[0060] Specifically, in a corresponding embodiment of the device of the present invention, a circuit on a carrier, comprising a plurality of pathways and at least one join, may include one join connected to at least two (preferably three) pathways supported by the carrier. The width of the pathway may be significantly larger than the diameter of the randomly moving token. As a result, the pathways may partially overlap. Furthermore, at least one output unit may be located in the join connecting the pathways, or may be located essentially in the partially overlapping pathways.

[0061] The device in Reference 1 relies on two different types of movement of the token in a direction approximately parallel to the carrier for operation. First, movement by a bias potential, which may preferably be applied to the ends of the pathways connected by a join, as in Reference 1. The bias potential should not be confused with the excitation unit of claim 1, since it induces a directed movement, but not a random movement. Second, random movement of the token, which is achieved in Reference 1 by thermal diffusion. The device operates as the token moves, which is influenced by the bias potential, and thus the probability of the token being present in the different pathways connected by the join varies with the bias potential. As a result, the probability of the token being present in the output unit varies with the bias potential, and there is a relationship between the bias potential and the output.

[0062] Random movement of the token is necessary for two reasons. First, random movement is necessary for the token to explore the various pathways connected by the joins. As discussed in the supplementary material in Ref. 1, the lack of random movement would introduce ambiguity into the relationship between the output and the bias potential, thereby making the operation of the device unreliable. Second, random movement of the token is used as a reset mechanism to reset the probability of the token being present in the various pathways, and therefore in the output units, when the bias potential itself changes.

[0063] The invention may be applied to the above mentioned setup, in particular by increasing the random movement of the tokens in a direction substantially parallel to the carrier using an excitation unit that applies a stimulus to the tokens in at least one pathway, thereby making it possible to increase the operating speed of the device. Furthermore, the random movement by the excitation unit makes it possible to operate the device even when the thermal diffusion is insufficient to provide at least one of the two required functionalities mentioned above.

[0064] Therefore, what has been disclosed and described in the above discussion are merely exemplary embodiments of the present disclosure. As will be understood by those skilled in the art, the present disclosure may be embodied in other specific forms without departing from the spirit or essential characteristics of the present disclosure. Therefore, this specification is intended to be illustrative and not limiting of the scope of the present disclosure and other claims. This disclosure defines the scope of the terms in the above claims, including all readily recognizable variations of the teachings of this specification, to a certain extent so as not to provide the subject matter of the invention to the public.

[0065] In the claims, the word "comprising" does not exclude other elements or steps, and the indefinite articles "a" or "an" do not exclude a plurality. A single element or other minimal unit may fulfill the functions of several items recited in a claim. The mere fact that certain measures are recited in mutually different dependent claims does not indicate that a combination of these measures cannot be utilized.

[0066] To the extent that the embodiments of the present disclosure described thus far are implemented, at least in part, by software-controlled data processing apparatus, it will be appreciated that a non-transitory machine-readable medium (e.g., optical disk, magnetic disk, semiconductor memory, etc.) bearing such software is also considered to represent an embodiment of the present disclosure. Moreover, such software may be distributed in other forms (e.g., via the Internet or other wired or wireless communications systems). Methods in accordance with the present invention may be specifically implemented to control the operation of a software-defined radio. [Explanation of symbols]

[0067] 10 Sensor System 12 Equipment 14 Sensors 16 Energy Harvesting Unit 18 Energy Sources 20 Input Interface 22 Career 24 Insertion Unit 25 Nucleation Units 26 Excitation Unit 28 Output Unit 30 Pathways 32 Join 34 Input Position 36 Output Position 38 Magnetic Sensor 40 Hub 42 Ratchet 50 Multilayer thin film system 52 Semiconductor Wafers 54 Pulse unit 56 Conductive material layer 58 Magnetic Field Unit 60 Electrical insulation layer 62 Electric Field Unit 64 Electromagnetic Field Unit

Claims

1. 1. An apparatus for processing information by evaluating randomly moving tokens corresponding to particles or quasiparticles, comprising: an input interface for receiving an input signal; a carrier supporting a plurality of pathways for the randomly moving tokens and at least one join that connects two pathways and allows a first token in a first of the two connected pathways to pass through the join when a second token from a second of the two connected pathways is present within the join; an insertion unit for inserting a token into the pathway based on the input signal; an excitation unit for applying a stimulus to a token in at least one of the plurality of pathways, the stimulus acting to increase the random motion of the token in a direction substantially parallel to the carrier; an output unit for outputting an output signal based on the position of the at least one token after a predetermined or dynamically adjusted time has elapsed; An apparatus comprising:

2. The apparatus according to claim 1 , wherein the apparatus is configured to evaluate particles or quasiparticles exhibiting Brownian diffusive motion as tokens (in particular as magnetic skyrmions).

3. The excitation unit includes a pulse unit for applying a current to the carrier and / or to at least one of the plurality of pathways; The carrier preferably comprises a conductive material layer enabling a spin torque effect (in particular a spin-orbit torque effect), and the current is applied to the conductive material layer. An apparatus according to any one of claims 1 to 2.

4. the excitation unit includes a magnetic field unit for applying a magnetic field or a magnetic field gradient to the carrier; The magnetic field unit preferably comprises at least two coils arranged on different sides of the carrier. An apparatus according to any one of claims 1 to 3.

5. The excitation unit includes an electric field unit for applying an electric field to the carriers; the carrier includes an electrically insulating layer (preferably a high-K dielectric material layer) that propagates the electric field; An apparatus according to any one of claims 1 to 4.

6. The excitation unit includes an electromagnetic field unit for generating a high-frequency electromagnetic field; The electromagnetic field unit preferably comprises an antenna. An apparatus according to any one of claims 1 to 5.

7. The apparatus according to any one of claims 1 to 6, wherein the output unit includes a magnetic sensor for detecting a change in magnetization or a change in a magnetic field at an output location on the carrier.

8. Apparatus according to any one of claims 1 to 7, wherein the output unit is configured to periodically determine whether a token is present at an output location on the carrier.

9. the carrier comprises a multi-layer thin film system arranged on a semiconductor wafer, the wafer preferably comprising an insulating top layer, the thin film system preferably comprising a magnetisable material; The pathways and the at least one join are preferably produced in the multilayer thin film system in an etching process and / or a lithography process. An apparatus according to any one of claims 1 to 8.

10. The apparatus of any one of claims 1 to 9, wherein the insertion unit comprises a nucleation unit for nucleating a token at an input location connected to the pathway.

11. Apparatus according to any one of claims 1 to 10, wherein the input interface is configured to receive the input signal from a sensor.

12. the input interface is configured to receive a stimulus signal; the excitation unit is configured to apply the stimulus based on the stimulus signal; The excitation unit is preferably configured to adapt the intensity of the stimulation based on the stimulation signal. An apparatus according to any one of claims 1 to 11.

13. 13. The device of claim 1, wherein the excitation unit is configured to apply the stimulus based on an amount of energy present in an energy source connected to an energy harvesting unit that generates electrical energy from the environment.

14. 1. A sensor system for detecting a physical phenomenon in an environment, comprising: A device as defined in any one of claims 1 to 13, a sensor that generates the input signal based on the physical phenomenon in the environment; A sensor system comprising:

15. The sensor system of claim 14 including an energy harvesting unit that generates electrical energy from the environment.

Images