QUANTUM COMPUTER AND QUANTUM COMPUTING USING ISING MODEL
专利汇可以提供QUANTUM COMPUTER AND QUANTUM COMPUTING USING ISING MODEL专利检索，专利查询，专利分析的服务。并且For each pair of a plurality of slave lasers B for which injection synchronization is performed by a master laser M, by controlling the intensity of light exchanged between two slave lasers B and an optical path length between the two slave lasers B using a slave-to-laser intensity control unit IA and an inter-slave laser optical path length control unit IP, the magnitude and the sign of pseudo ising interaction Jij between the two slave lasers B are implemented. After the plurality of slave lasers B arrive at a steady state, by measuring relative values of the oscillation phases of the plurality of slave lasers B with respect to the oscillation phase of the master laser M by using an oscillation phase measuring unit PM, pseudo ising spins σi of the plurality of slave lasers B are measured.，下面是QUANTUM COMPUTER AND QUANTUM COMPUTING USING ISING MODEL专利的具体信息内容。
What is claimed is:
1. Field of the Disclosure
The present disclosure provides a computer and computing capable of easily solving an NP-complete problem mapped into the ising model by easily solving the ising model.
2. Discussion of the Background Art
The ising model has been researched originally as a model of a magnetic material but recently attracts attention as a model mapped in an NP-complete problem or the like. However, it is very difficult to solve the ising model when the number of sites is large. Thus, a quantum annealing machine and a quantum adiabatic machine in which the ising model is implemented are proposed.
In the quantum annealing machine, after ising interaction and Zeeman energy are physically implemented, the system is sufficiently cooled so as to realize a ground state, and the ground state is observed, whereby the ising model is solved. However, in a case where the number of sites is large, the system is trapped into a metastable state in the process of being cooled, and the number of the metastable states exponentially increases with respect to the number of sites, whereby there is a problem in that the metastable state is not easily mitigated to the ground state.
In the quantum adiabatic machine, transverse magnetic field Zeeman energy is physically implemented, and then the ground state of the transverse magnetic field Zeeman energy is realized by sufficiently cooling the system. Then, the transverse magnetic field Zeeman energy is gradually lowered, ising interaction is physically implemented gradually, the ground state of the system that includes the ising interaction and vertical magnetic field Zeeman energy is realized, and ground state is observed, whereby the ising model is solved. However, when the number of sites is large, there is a problem in that the speed of gradually lowering transverse magnetic field Zeeman energy and physically implementing the ising interaction in a gradual manner needs to be exponentially decreased with respect to the number of sites.
In a case where the NP-complete problem or the like is mapped into an ising model, and the ising model is implemented as a physical spin system, there is a problem of a natural law that ising interaction between sites that are physically located close to each other is large, and ising interaction between sites that are physically located far from each other is small. The reason for this is that, in an artificial ising model in which the NP-complete problem is mapped, there may be cases where ising interaction between sites that are physically located close to each other is small, and ising interaction between sites that are physically located far is large. The difficulty in mapping into a natural spin system also makes it difficult to easily solve the NP-complete problem or the like.
Non-Patent Literature 1: Tim Byrnes, Kai Yan, and Yoshihisa Yamamoto, Optimization using Bose-Einstein condensation and measurement-feedback circuits, [online], Jan. 26, 2010, arXiv.org, [searched on Oct. 9, 2012], the Internet URL:http://arxiv.org/abs/0909.2530
The configuration of an ising model computer that is disclosed in Non-Patent Literature 1 for solving some of the above-described problems is illustrated in
The Bose-Einstein condensing units B1, B2, and B3 are systems in which almost all the Bose particles are in the ground state at a very low temperature and are configured by exciton polaritons included in semiconductor micro-cavities, neutral atoms each having an unpaired electron, or the like. The Bose-Einstein condensing units B1, B2, and B3 are respectively applied with magnetic fields B1, B2, and B3 to be described later and are configured by Bose particles having mutually-different spin directions denoted by white circles and black circles illustrated in
The spin measuring units D1, D2, and D3 output currents I1, I2, and I3 that are in proportion to sums of all the spins respectively included in the Bose-Einstein condensing units B1, B2, and B3. Here, a sum Si of all the spins inside each site is represented as below. Here, a, represents the spin of each Bose particle of each site, and N represents a total number of Bose particles of each site.
The feedback control circuit F receives the currents II, I2, and I3 as inputs respectively from the spin measuring units D1, D2, and D3 and outputs feedback signals to the ising interaction implementing units I1, I2, and I3. The ising interaction implementing units I1, I2, and I3 input the feedback signals from the feedback control circuit F and applies magnetic fields B1, B2, and B3 to the Bose-Einstein condensing units B1, B2, and B3 respectively. Here, the magnetic fields B1, B2, and B3 are represented as below. In addition, J1 represents an ising interaction coefficient between an i-th site and a j-th site, and M represents the number (three in
Hamiltonian H of all the Bose-Einstein condensing units B1, B2, and B3 is represented as below. In other words, the ising interaction is implemented.
When the ising model computer illustrated in
When the ising model computer illustrated in
In the ising model computer illustrated in
In the ising model computer illustrated in
However, in the ising model computer illustrated in
Here, the determined spin state of the whole system is not always the ground state. Thus, the spin state of the whole system needs to be determined over and over until the spin state of the whole system is settled to the ground state, and, in a worst case, spin states of 2M kinds of the whole system need to be determined. In other words, the computation time is in proportion to 2M/N, and thus, even when Bose-Einstein condensing is applied, the exponential divergence of the computation time cannot be suppressed.
Thus, in order to solve the above-described problems, an object of the present disclosure is to provide a computer and computing for suppressing exponential divergence of the computation time of an NP-complete problem or the like mapped into an ising model.
In order to achieve the above-described object, for each pair of a plurality of coherent oscillators for which injection synchronization is performed by a master oscillator, by controlling the intensity of light exchanged between two coherent oscillators and an optical path length between the two coherent oscillators, the magnitude and the sign of the pseudo ising interaction between the two coherent oscillators are implemented. Then, after the plurality of coherent oscillators arrive at the steady state, by measuring relative values of the oscillation phases of the plurality of coherent oscillators with respect to the oscillation phase of the master oscillator, pseudo ising spins of the plurality of coherent oscillators are measured.
Specifically, according to the present disclosure, a quantum computer using an ising model includes: a plurality of coherent oscillators that oscillate light having polarization polarized in a same direction determined in advance in correspondence with a plurality of sites of the ising model; a master oscillator that performs injection synchronization for the plurality of coherent oscillators and oscillates the light having polarization polarized in the same direction determined in advance; a master oscillator-to-coherent oscillator optical path unit that is arranged between the master oscillator and each one of the coherent oscillators; an inter-coherent oscillator optical path unit that is arranged between two coherent oscillators for each pair of the plurality of the coherent oscillators; an oscillation frequency control unit that is arranged in each master oscillator-to-coherent oscillator optical path unit and controls an oscillation frequency of each one of the coherent oscillators so as to be an oscillation frequency of the master oscillator; an inter-coherent oscillator intensity control unit that is arranged in each inter-coherent oscillator optical path unit for each pair of the plurality of the coherent oscillators and implements a magnitude of pseudo ising interaction between two coherent oscillators by controlling an intensity of light exchanged between the two coherent oscillators; an inter-coherent oscillator optical path length control unit that is arranged in each inter-coherent oscillator optical path unit for each pair of the plurality of the coherent oscillators and implements a sign of the pseudo ising interaction between two coherent oscillators by controlling an optical path length between the two coherent oscillators; and an oscillation phase measuring unit that measures pseudo ising spins of the plurality of the coherent oscillators by measuring relative values of oscillation phases of the plurality of the coherent oscillators with respect to the oscillation phase of the master oscillator after the plurality of the coherent oscillators arrive at a steady state.
According to the present disclosure, a quantum computing using an ising model includes: starting oscillation of a plurality of coherent oscillators that oscillate light having polarization polarized in a same direction determined in advance in correspondence with a plurality of sites of the ising model and an oscillation stat step of starting oscillation of a master oscillator that performs injection synchronization for the plurality of coherent oscillators and oscillates light having polarization polarized in the same direction determined in advance; an oscillation frequency control step of controlling an oscillation frequency of each one of the coherent oscillators so as to be an oscillation frequency of the master oscillator; implementing the magnitude of pseudo ising interaction between two coherent oscillators by controlling the intensity of light exchanged between the two coherent oscillators for each pair of the plurality of the coherent oscillators and an inter-coherent oscillator intensity optical path length control step of implementing the sign of the pseudo ising interaction between two coherent oscillators by controlling an optical path length between the two coherent oscillators for each pair of the plurality of the coherent oscillators; and an oscillation phase measurement step of measuring pseudo ising spins of the plurality of the coherent oscillators by measuring relative values of oscillation phases of the plurality of the coherent oscillators with respect to the oscillation phase of the master oscillator after the plurality of the coherent oscillators arrive at a steady state.
According to such a configuration, the oscillation phase of light emitted from the coherent oscillator is associated with the pseudo spin direction of the coherent oscillator, and the pseudo spin state of the whole system is measured. Until the pseudo spin state of the whole system is settled to the ground state, the quantum coherence of the whole system is not broken, and accordingly, the problem of exponentially increasing of the computation time in accordance with the number of all the sites is solved, whereby the computation time can be shortened to a large extent.
Thus, for any pair out of M sites, through light exchanged between two sites, not only ising interaction between sites physically located close to each other but also ising interaction between sites located far from each other can be implemented. Accordingly, regardless of a physical distance between sites, an ising model in which any NP-complete problem or the like can be mapped can be solved.
In addition, for a number of pseudo spins in each site, whether one of the upward circulation and the downward circulation is more than the other is determined based on the rule of majority. Accordingly, even when there is a pseudo spin that is leaked from the ground state to the excited state, the probability of acquiring a correct solution is markedly higher in a case where there are a number of pseudo spins than that in a case where there is one pseudo spin.
Here, as the whole system transits from the initial state to the steady state, the oscillation phase of each coherent oscillator deviates from that of the master laser, but the oscillated polarized light of each coherent oscillator is constantly the same as that of the master laser.
Thus, for each coherent oscillator, a differential equation of the temporal developments of only three including the oscillation intensity, the oscillation phase, and the inverted population number difference of carriers may be formulated. Each coherent oscillator might have the direction of single oscillated polarized light, and accordingly, even in the case of a semiconductor surface emission laser having in-plane anisotropy, the in-plane anisotropy does not matter. As above, the quantum computer using the ising model can be easily designed and manufactured.
In the quantum computer using an ising model according to the present disclosure, each inter-coherent oscillator optical path length control unit implements the sign of pseudo ising interaction between two coherent oscillators by performing control such that an optical path length between the two coherent oscillators is a half-integer multiple or an integer multiple of an oscillation wavelength of the injection synchronization.
In the quantum computing using an ising model according to the present disclosure, the inter-coherent oscillator intensity optical path length control step implements the sign of pseudo ising interaction between two coherent oscillators by performing control such that the optical path length between the two coherent oscillators is a half-integer multiple or an integer multiple of an oscillation wavelength of the injection synchronization.
According to such a configuration, for each pair of coherent oscillators, when the ising interaction is positive, a deviation between oscillation phases of two coherent oscillators is π (the deviation may be zero depending on the definition of the sign of the ising interaction), and the oscillation mode can be easily started. On the other hand, when the ising interaction is negative, a deviation between the oscillation phases of two coherent oscillators is zero (the deviation may be π depending on the definition of the sign of the ising interaction), and the oscillation mode can be easily started.
Above all, in the whole ising model computer, one oscillation mode is configured to be integrally started. Thus, in each pair of coherent oscillators, when there is a case where the above-described oscillation mode is actually started, there is also a case where the oscillation mode is not necessarily started.
In the quantum computer using an ising model according to the present disclosure, the oscillation phase measuring unit determines that directions of pseudo ising spins of two coherent oscillators to be the same when oscillation phases of both the coherent oscillators lead before or lag behind the oscillation phase of the master oscillator for each pair of the plurality of the coherent oscillators and determines that the directions of the pseudo ising spins of two coherent oscillators to be different from each other when an oscillation phase of one coherent oscillator leads before the oscillation phase of the master oscillator and the oscillation phase of the other coherent oscillator lags behind the oscillation phase of the master oscillator for each pair of the plurality of the coherent oscillators.
In the quantum computing using an ising model according to the present disclosure, in the oscillation phase measurement step, directions of pseudo ising spins of two coherent oscillators are determined to be the same when oscillation phases of both the coherent oscillators lead before or lag behind the oscillation phase of the master oscillator for each pair of the plurality of the coherent oscillators, and the directions of the pseudo ising spins of two coherent oscillators are determined to be different from each other when an oscillation phase of one coherent oscillator leads before the oscillation phase of the master oscillator, and the oscillation phase of the other coherent oscillator lags behind the oscillation phase of the master oscillator.
According to such a configuration, the sign of a relative value of the oscillation phase of the coherent oscillator with respect to the oscillation phase of the master oscillator can be configured to be in correspondence with the direction of the pseudo spin.
The quantum computer using an ising model according to the present disclosure further includes: a master oscillator-to-coherent oscillator intensity control unit that is arranged in each master oscillator-to-coherent oscillator optical path unit and implements the magnitude of pseudo Zeeman energy in each coherent oscillator by controlling the intensity of light injected into each coherent oscillator; and a master oscillator-to-coherent oscillator phase control unit that is arranged in each master oscillator-to-coherent oscillator optical path unit and implements the sign of the pseudo Zeeman energy in each coherent oscillator by controlling the phase of light injected into each coherent oscillator.
The quantum computing using an ising model according to the present disclosure further includes implementing the magnitude of pseudo Zeeman energy in each coherent oscillator by controlling the intensity of light injected into each coherent oscillator and a master oscillator-to-coherent oscillator intensity phase control step of implementing the sign of the pseudo Zeeman energy in each coherent oscillator by controlling the phase of light injected into each coherent oscillator, in parallel with the inter-coherent oscillator intensity optical path control step.
According to such a configuration, Zeeman energy can be applied to a system in which only ising interaction is present, whereby the energy degeneracy in ground state can be resolved.
In the quantum computer using an ising model according to the present disclosure, each master oscillator-to-coherent oscillator phase control unit implements the sign of the pseudo Zeeman energy in each coherent oscillator by controlling lead or lag of an injection phase for each coherent oscillator with respect to the oscillation phase of the master oscillator.
In the quantum computing using an ising model according the present disclosure, in the master oscillator-to-coherent oscillator intensity phase control step, the sign of the pseudo Zeeman energy in each coherent oscillator is implemented by controlling lead or lag of an injection phase for each coherent oscillator with respect to the oscillation phase of the master oscillator.
According to such a configuration, for each coherent oscillator, when the Zeeman energy is positive, the oscillation phase of the coherent oscillator lags behind the oscillation phase of the master oscillator (the oscillation phase may lead depending on the definition of the sign of the Zeeman energy), and accordingly, the oscillation mode can be easily started. On the other hand, when the Zeeman energy is negative, the oscillation phase of the coherent oscillator leads before the oscillation phase of the master oscillator (the oscillation phase may lag depending on the definition of the sign of the Zeeman energy), and accordingly, the oscillation mode can be easily started.
Above all, in the whole ising model computer, one oscillation mode is configured to be integrally started. Thus, in the coherent oscillator, when there is a case where the above-described oscillation mode is actually started, there is also a case where the oscillation mode is not necessarily started.
In the quantum computer using an ising model according to the present disclosure, each oscillation frequency control unit controls the oscillation frequency of each coherent oscillator so as to be the oscillation frequency of the master oscillator by performing control such that an interference intensity of oscillated light of each coherent oscillator and oscillated light of the master oscillator is an extreme value with respect to a change in the oscillation frequency of each coherent oscillator.
According to such a configuration, since the oscillation frequency of each coherent oscillator is controlled so as to be the oscillation frequency of the master oscillator, as the whole quantum computer using the ising model, one oscillation mode integrally can be started.
In the quantum computer using an ising model according to the present disclosure, each inter-coherent oscillator optical path length control unit controls an optical path length between two coherent oscillators so as to be a half-integer multiple or an integer multiple of an oscillation wavelength of the injection synchronization by performing control such that an interference intensity of oscillated light of the two coherent oscillators is an extreme value with respect to a change in the optical path length between the two coherent oscillators.
According to such a configuration, for each pair of coherent oscillators, the optical path length is controlled so as to be a half-integer multiple or integer multiple of the oscillation wavelength of the injection synchronization, and, when the ising interaction is positive, the oscillation mode where a deviation between oscillation phases of two coherent oscillators is π (the deviation may be zero depending on the definition of the sign of the ising interaction) can be easily started. On the other hand, when the ising interaction is negative, the oscillation mode where a deviation between the oscillation phases of two coherent oscillators is zero (the deviation may be π depending on the definition of the sign of the ising interaction) can be easily started.
In the quantum computer using an ising model according to the present disclosure, control of the oscillation frequency of each coherent oscillator in each oscillation frequency control unit, control of the optical path length between two coherent oscillators in each inter-coherent oscillator optical path length control unit, and measurement of the pseudo ising spins of the plurality of the coherent oscillators in the oscillation phase measuring unit are performed in the mentioned order.
According to such a configuration, as the whole quantum computer using the ising model, one oscillation mode is assured to be integrally started, and the ising interaction can be correctly implemented, whereby the ising model can be correctly computed.
The quantum computer using an ising model according to the present disclosure further includes a pumping current control unit that performs gradual increase control of pumping currents of two coherent oscillators for the two coherent oscillators performing pseudo ising interaction through each inter-coherent oscillator optical path unit in a state in which the magnitude and the sign of the pseudo ising intersection between the two coherent oscillators are implemented to be fixed values and performs fixing control of the pumping currents of the two coherent oscillators at a time point when the plurality of the coherent oscillators arrive at one integrally oscillation mode for the first time, wherein the oscillation phase measuring unit, after the plurality of the coherent oscillators arrive at one integrally oscillation mode so as to arrive at a steady state, measures the pseudo ising spins of the plurality of the coherent oscillators by measuring relative values of the oscillation phases of the plurality of the coherent oscillators with respect to the oscillation phase of the master oscillator.
According to such a configuration, the state of the computer can correctly arrive at the ground state of the ising model without being incorrectly trapped into the quasi-stable state of the ising model.
The quantum computer using an ising model according to the present disclosure further includes an adjacent ising spin direction fixing unit that is arranged in each master oscillator-to-coherent oscillator optical path unit, and, for two coherent oscillators performing pseudo ising interaction through each inter-coherent oscillator optical path unit, by controlling the intensity and the phase of light injected into the two coherent oscillators when relative values of the oscillation phases of the two coherent oscillators with respect to the oscillation phase of the master oscillator are not measured to be meaningful, fixes the directions of pseudo ising spins of the two coherent oscillators to be different from each other when the sign of the pseudo ising interaction between the two coherent oscillators is implemented to be positive and fixes the directions of the pseudo ising spins of the two coherent oscillators to be the same when the sign of the pseudo ising interaction between the two coherent oscillators is implemented to be negative.
According to such a configuration, between two coherent oscillators performing pseudo ising interaction, the frustration of a pseudo spin can be resolved.
The quantum computer using an ising model according to the present disclosure further includes a peripheral ising spin direction fixing unit that is arranged in each master oscillator-to-coherent oscillator optical path unit, and, for two coherent oscillators performing pseudo ising interaction through each inter-coherent oscillator optical path unit, by controlling the intensity and the phase of light injected into an adjacent coherent oscillator performing pseudo ising interaction with the two coherent oscillators through another inter-coherent oscillator optical path unit when relative values of the oscillation phases of the two coherent oscillators with respect to the oscillation phase of the master oscillator are not measured to be meaningful, fixes the direction of the pseudo ising spin of the adjacent coherent oscillator to a direction of a current time point during computation.
According to such a configuration, between two coherent oscillators performing pseudo ising interaction, when the frustration of the pseudo spin is resolved, unintentional flipping of the pseudo spin can be prevented in the coherent oscillators that are adjacent to the two coherent oscillators in a pseudo manner.
In the quantum computer using an ising model according to the present disclosure, a delay time of the pseudo ising interaction between two coherent oscillators performing pseudo ising interaction through each inter-coherent oscillator optical path unit is shorter than the reciprocal of the injection locking width in the master oscillator and the two coherent oscillators.
According to such a configuration, the quantum computer using the ising model can be operated in a stable manner.
In the quantum computer using an ising model according to the present disclosure, the plurality of the coherent oscillators are a plurality of slave lasers.
According to such a configuration, the whole system can be operated not at a low temperature but at a room temperature.
In the quantum computer using an ising model according to the present disclosure, the plurality of the coherent oscillators are a plurality of Bose-Einstein condensates.
According to such a configuration, a quantum computer and quantum computing using an ising model can be realized using the Bose-Einstein condensates.
In the quantum computer using an ising model according to the present disclosure, the master oscillator is a master laser.
According to such a configuration, a quantum computer and quantum computing using an ising model can be realized using the master laser.
The present disclosure can provide a computer and computing for suppressing exponential divergence of the computation time of an NP-complete problem or the like mapped into an ising model.
Embodiments of the present disclosure will be described with reference to the accompanying drawings. The embodiments to be described hereinafter are examples of embodiments of the present disclosure, and the present disclosure is not limited to the following embodiments. In this specification and the drawings, like reference numerals denote like elements.
An NP-complete problem can be substituted by an ising model of a magnetic body, and the ising model of a magnetic body can be substituted by a network of a laser.
Here, in the ising model of a magnetic body, in a pair of atoms interacting with each other, the directions of spins tend to be oriented in opposite directions (in the case of interaction of antiferromagnetism) or in the same direction (in the case of interaction of ferromagnetism) such that the energy of spin alignment is the lowest.
On the other hand, in a network of lasers, in a pair of lasers interacting with each other, the phases of light emission tend to be opposite phases (in the case of interaction of antiferromagnetism) or the same phase (in the case of interaction of ferromagnetism) such that the threshold gain of the oscillation mode is the lowest.
As above, in a system configured by one pair of lasers, the phases of light emission can be optimized such that the threshold gain of the oscillation mode is the lowest. In a system configured by many pairs of lasers, when, in a case where the phases of light emission are optimized for a “certain” pair of lasers, the phases of light emission cannot be optimized for “the other” pairs of lasers, a “point of compromise” of the phases of light emission as a “whole” of the network of lasers is searched for.
In a case where the phases of light emission are optimized as a whole of the network of lasers, it is necessary to achieve synchronization between lasers such that individual oscillation modes for pairs of lasers are not started, but one oscillation mode is started as a whole of the network of lasers.
More specifically, by injecting light having polarization (for example, vertically linear polarized light |V>) that is polarized in a predetermined direction from a master (control side) laser to each slave (controlled side) laser (a constituent element of the network of lasers), injection synchronization among lasers is achieved.
Then, it is measured whether the phase of the polarized light (for example, vertically linear polarized light |V>), which is polarized in the above-described predetermined direction, emitted from each slave laser leads before or lags after the phase of the polarized light (for example, vertically linear polarized light |V>), which is polarized in the above-described predetermined direction, emitted from the master laser. In addition, it is determined whether, in each pair of slave lasers, a state in which both lasers have lead phases or lag phases, or a state in which one laser has a lead phase and the other laser has a lag phase, is realized.
At last, the direction (upward or downward) of the spin at each atom position corresponding to each slave laser is determined, and, finally, the NP-complete problem is solved. Here, in each pair of slave lasers, when both lasers have lead phases or lag phases, in each corresponding pair of atoms, the directions of spins are the same. On the other hand, in each pair of slave lasers, when one laser has a lead phase but the other laser has a lag phase, in each corresponding pair of atoms, the directions of spins are opposite to each other.
Next, a computer will be described. The configuration of an ising model computer is illustrated in
The slave lasers or the Bose-Einstein condensing units B1, B2, and B3 oscillate light having polarization (for example, vertically linear polarized light |V>) polarized in the same direction determined in advance in accordance with a plurality of sites of the ising model. The master laser M performs injection synchronization for the slave lasers or the Bose-Einstein condensing units B1, B2, and B3 and oscillates light having polarization (for example, vertically linear polarized light |V>) polarized in the same direction determined in advance.
Here, a photon or an exciton polariton is a Bose particle. As the slave lasers B1, B2, and B3 using photons, semiconductor surface-emitting lasers may be applied. In addition, as the Bose-Einstein condensing units B1, B2, and B3 using the exciton polaritons, semiconductor micro-cavities can be applied. Hereinafter, while the operations of the slave lasers B1, B2, and B3 using photons will be described, the operations of the Bose-Einstein condensing units B1, B2, and B3 using the exciton polaritons are similar thereto.
The master-to-slave optical path unit ZL1 is arranged between the master laser M and the slave laser B1. In addition, the master-to-slave optical path unit ZL2 is arranged between the master laser M and the slave laser B2. The master-to-slave optical path unit ZL3 is arranged between the master laser M and the slave laser B3.
The inter-slave optical path unit IL12 is arranged between the slave lasers B1 and B2. In addition, the inter-slave optical path unit IL23 is arranged between the slave lasers B2 and B3. The inter-slave optical path unit IL13 is arranged between the slave lasers B1 and B3.
Here, the master-to-slave optical path unit ZL and the inter-slave optical path unit IL may share a path. In such a case, beam splitters may be arranged at branching positions of the master-to-slave optical path unit ZL and the inter-slave optical path unit IL.
The oscillation frequency control unit FC1 is arranged in the master-to-slave optical path unit ZL1 and controls the oscillation frequency of the slave laser B1 so as to be the oscillation frequency of the master laser M. In addition, the oscillation frequency control unit FC2 is arranged in the master-to-slave optical path unit ZL2 and controls the oscillation frequency of the slave laser B2 so as to be the oscillation frequency of the master laser M. The oscillation frequency control unit FC3 is arranged in the master-to-slave optical path unit ZL3 and controls the oscillation frequency of the slave laser B3 so as to be the oscillation frequency of the master laser M. The oscillation frequency control units FC1, FC2, and FC3 will be described later with reference to
The inter-slave intensity control unit IA12 is arranged in the inter-slave optical path unit IL12 and implements the magnitude of pseudo ising interaction J12 between two slave lasers B1 and B2 by controlling the intensity of light exchanged between the two slave lasers B1 and B2. In addition, the inter-slave intensity control unit IA23 is arranged in the inter-slave optical path unit IL23 and implements the magnitude of pseudo ising interaction J23 between two slave lasers B2 and B3 by controlling the intensity of light exchanged between the two slave lasers B2 and B3. The inter-slave intensity control unit IA13 is arranged in the inter-slave optical path unit IL13 and implements the magnitude of pseudo ising interaction J13 between two slave lasers B1 and B3 by controlling the intensity of light exchanged between the two slave lasers B1 and B3. As the inter-slave intensity control units IA12, IA23, and IA13, for example, attenuation plates or the like may be applied.
The inter-slave optical path length control unit IP12 is arranged in the inter-slave optical path unit IL12 and implements the sign of the pseudo ising interaction J12 between the two slave lasers B1 and B2 by controlling the optical path length between the two slave lasers B1 and B2. In addition, the inter-slave optical path length control unit IP23 is arranged in the inter-slave optical path unit IL23 and implements the sign of the pseudo ising interaction J23 between the two slave lasers B2 and B3 by controlling the optical path length between the two slave lasers B2 and B3. The inter-slave optical path length control unit IP13 is arranged in the inter-slave optical path unit IL13 and implements the sign of the pseudo ising interaction J13 between the two slave lasers B1 and B3 by controlling the optical path length between the two slave lasers B1 and B3. The inter-slave optical path length control units IP12, IP23, and IP13 will be described later with reference to
The master-to-slave intensity control unit ZA1 is arranged in the master-to-slave optical path unit ZL1 and implements the magnitude of pseudo Zeeman energy λ1 of the slave laser B1 by controlling the intensity of light injected to the slave laser B1. In addition, the master-to-slave intensity control unit ZA2 is arranged in the master-to-slave optical path unit ZL2 and implements the magnitude of pseudo Zeeman energy λ2 of the slave laser B2 by controlling the intensity of light injected to the slave laser B2. The master-to-slave intensity control unit ZA3 is arranged in the master-to-slave optical path unit ZL3 and implements the magnitude of pseudo Zeeman energy λ3 of the slave laser B3 by controlling the intensity of light injected to the slave laser B3. As the master-to-slave intensity control units ZA1, ZA2, and ZA3, for example, attenuation plates or the like may be applied.
The master-to-slave phase control unit ZP1 is arranged in the master-to-slave optical path unit ZL1 and implements the sign of the pseudo Zeeman energy X of the slave laser B1 by controlling the phase of the light injected to the slave laser B1. In addition, the master-to-slave phase control unit ZP2 is arranged in the master-to-slave optical path unit ZL2 and implements the sign of the pseudo Zeeman energy λ2 of the slave laser B2 by controlling the phase of the light injected to the slave laser B2. The master-to-slave phase control unit ZP3 is arranged in the master-to-slave optical path unit ZL3 and implements the sign of the pseudo Zeeman energy λ3 of the slave laser B3 by controlling the phase of the light injected to the slave laser B3. As the configuration of the master-to-slave phase control units ZP1, ZP2, and ZP3, for example, any one of configurations illustrated as in
After a plurality of the slave lasers B1, B2, and B3 arrive at a steady state, the oscillation phase measuring unit PM measures relative values of the oscillation phases of the plurality of the slave lasers B1, B2, and B3 with respect to the oscillation phase of the master laser M, thereby measuring pseudo ising spins σ1, σ2, and σ3 of the plurality of slave lasers B1, B2, and B3. The oscillation phase measuring unit PM and the pumping current control unit PC will be described later with reference to
Next, the calculation principle will be described. Ising Hamiltonian is acquired using Numerical Expression 4.
Here, in Numerical Expression 4, a sign attached prior to Jij is “+”. Accordingly, when is positive, antiferromagnetic interaction works between Si and Sj. On the other hand, when is negative, ferromagnetic interaction works between Si and Si. However, the sign attached prior to Jij may be “−”. In such a case, when is positive, ferromagnetic interaction works between Si and Si and, when is negative, antiferromagnetic interaction works between Si and Si.
In addition, in Numerical Expression 4, the sign attached prior to is “+”. Accordingly, when λi is positive, it is easy for Si to take a downward spin “−1”. On the other hand, when λi is negative, it is easy for Si to take an upward spin “+1”. However, the sign attached prior to λi may be “−”. In such a case, when λi is positive, it is easy for Si to take an upward spin “+1”, and, when λi is negative, it is easy for Si to take a downward spin “−1”.
As the above-described direction of the polarized light that is determined in advance, vertically linear polarized light |V> is selected. In each one of the slave lasers B1, B2, and B3, rate equations for oscillation intensity AVi(t), an oscillation phase φVi(t), and an inverted population number difference Nci(t) of carriers are as in Numerical Expressions 5 to 8.
Here, φ is an oscillation frequency, and Q is a Q factor of resonator of the master laser M and each slave laser B. nM is the number of photons supplied from the master laser M. P is the number of electrons injected to each slave laser B per second for realizing inverter population, in other words, the pumping rate. −(½)(ω/Q)AVi(t) illustrated in Numerical Expression 5 represents the decreasing rate of the oscillation intensity AVi(t) according to elapse of time that is caused by a resonator loss.
τsp is an electron lifetime according to spontaneous emission toward another oscillation mode other than a laser oscillation mode. β is a coupling constant of all the light that is spontaneously emitted for the laser oscillation mode and is about 10−4 to 10−5 in the case of a semiconductor surface-emitting laser. (½)ECVi(t)AVi(t) illustrated in Numerical Expression 5 represents the increasing rate of the oscillation intensity AVi(t) according to elapse of time that is caused by stimulated emission. In addition, ECVi(t) illustrated in Numerical Expression 5 represents the increasing rate of the oscillation intensity AVi(t) according to elapse of time that is caused by spontaneous emission.
The injection phase of each slave laser B deviates from the oscillation phase of the master laser M. In light injected to each slave laser B, it is assumed that the intensity of a component having the same phase as the oscillation phase of the mater laser M is in proportion to ζ, and the intensity of a component having a phase deviating from the oscillation phase of the master laser M by π/2 is in proportion to ηi.
A term to which ζ and ηi relate in Numerical Expression 5 is a term relating to Zeeman energy. (ω/Q)√nM{ζ cos φVi(t)−ηi sin φVi(t)} illustrated in Numerical Expression 5 represents the change rate of the oscillation intensity AVi(t) according to elapse of time in the i-th site when light is injected from the master laser M to the i-th site.
A term to which ζ and ηi relate in Numerical Expression 6 is a term relating to Zeeman energy. (1/AVi(t))(ω/Q)√nM{−ζ sin φVi(t)−ηi cos φVi(t)} illustrated in Numerical Expression 6 represents the change rate of the oscillation phase φVi(t) according to elapse of time in the i-th site when light is injected from the master laser M to the i-th site.
In addition, terms to which ζ relates in Numerical Expressions 5 and 6 relate to injection for injection synchronization of laser before the start of computation (t<0) and after the start of computation (t>0) but do not directly relate to the magnitude and the sign of the Zeeman energy. Furthermore, terms to which ηi relates in Numerical Expressions 5 and 6 relate to injection for implementing Zeeman energy after the start of computation (t>0).
A term to which ξij relates in Numerical Expression 5 relates to mutually injected light between slave lasers B for implementing ising interaction. −(ω/Q)(½)ξijAVi(t)cos {φVj(t)−φVi(t)} illustrated in Numerical Expression 5 represents the change rate of the oscillation intensity AVi(t) according to elapse of time in the i-th site when light is injected from the j-th site to the i-th site. Σ(j≠i) illustrated in Numerical Expression 5 represents contributions made from all the sites (the j-th site) other than the i-th site in the i-th site.
A term to which relates in Numerical Expression 6 relates to mutually injected light between slave lasers B for implementing ising interaction. −(1/AVi(t))(ω/Q)(½)ξijAVi(t)sin {φVj(t)−φVi(t)} illustrated in Numerical Expression 6 represents the change rate of the oscillation phase φVi(t) according to elapse of time in the i-th site when light is injected from the j-th site to the i-th site. Σ(j≠i) illustrated in Numerical Expression 6 represents contributions made from all the sites (the j-th site) other than the i-th site in the i-th site.
FAV, FφV, and FN represent noises in the oscillation intensity, the oscillation phase, and the inverted population number difference of the carrier in the i-th site.
In the steady state, Numerical Expression 5 can be represented as Numerical Expression 9.
By transforming Numerical Expression 9 with FAV being ignored, Numerical Expression 10 is formed.
Here, when σi of the ising model takes −1 or +1, sin φVi represented in Numerical Expression 10 takes a value in the range of −1 to +1. Thus, when it is set σi=sin φVi in consideration of the similarity between the ising model and the laser system, Numerical Expression 10 can be represented as Numerical Expression 11.
By adding Numerical Expression 11 for all the M sites, Numerical Expression 12 is acquired and is represented as a threshold gain ΣECVi of the whole laser system.
Here, since the master laser M performs injection synchronization for each slave laser B, it can be set AVi(t)˜AVj(t)˜√nM. Then, in the above-described definition σi=sin φVi, σi=±1, and accordingly, it can be set φVi˜φVi˜±π/2. In such a case, Numerical Expression 12 is formed as Numerical Expression 13.
Here, in a case where each slave laser B has a uniform medium, an oscillation phase state {σi} implementing a minimum threshold gain ΣECVi as the whole laser system is selected. In other words, one specific oscillation mode is selected as the whole laser system. Then, owing to competition between oscillation modes, one specific oscillation mode suppresses the other oscillation modes.
In other words, ΣECVi illustrated in Numerical Expression 13 is minimized as the whole laser system. On the other hand, (ω/Q)M represented in Numerical Expression 13 is constant as the whole laser system. Accordingly, Σηiσi+Σξijσiσi represented in Numerical Expression 13 is minimized as the whole laser system.
By putting ξij=Jij and ηi=λi, when Σηiσi+Σξijσiσj is minimized as the whole laser system, Σλiσi+ΣJijσiσj is minimized as well. In other words, a ground state minimizing the Ising Hamiltonian is realized.
In order to improve the computation precision, it is necessary to configure a difference between a minimal threshold gain and a next minimal threshold gain to be sufficiently larger than β(ω/Q)(1/R) that is a difference between a saturation gain ECV determined based on the spontaneous emission rate and the photon attenuation rate w/Q for a laser oscillation mode of the whole laser system. Here, R=I/Ith−1 is a standardized pumping rate, and I and Ith are an injection current and a threshold of the laser oscillation thereof. Accordingly, by decreasing β and increasing R, the computation precision can be improved.
Next, the principle of the computation will be summarized. The principle of the ising model computer according to the present disclosure is illustrated in
For each pair of slave lasers B, when the ising interaction Jij is positive, it is advantageous from the viewpoint of energy that the pseudo spins a of the two slave lasers B have mutually different signs. In other words, an oscillation mode, in which the oscillation phases φV of the two slave lasers B have mutually-different signs and have a deviation of π therebetween, is configured to be easily started. Thus, as will be described later with reference to
For each pair of slave lasers B, when the ising interaction Jij is negative, it is advantageous from the viewpoint of energy that the pseudo spins a of the two slave lasers B have mutually the same sign. In other words, an oscillation mode, in which the oscillation phases φV of the two slave lasers B have the same sign and have a deviation of zero therebetween, is configured to be easily started. Thus, as will be described later with reference to
For each slave laser B, when the Zeeman energy λi is positive, it is advantageous from the viewpoint of energy that the pseudo spin σ of the slave laser B is −1. In other words, an oscillation mode, in which the oscillation phase φV of the slave laser B is −π/2, is configured to be easily started. Accordingly, for components (terms to which ηi represented in Numerical Expressions 5 and 6 relates) contributing to the implementation of the Zeeman energy λi among light injected to each slave laser B, each master-to-slave phase control unit ZP performs control such that the injection phase for the slave laser B lags by π/2 with respect to the oscillation phase of the master laser M.
For each slave laser B, when the Zeeman energy λi is negative, it is advantageous from the viewpoint of energy that the pseudo spin σ of the slave laser B is +1. In other words, an oscillation mode, in which the oscillation phase φV of the slave laser B is +π/2, is configured to be easily started. Accordingly, for components (terms to which ηi represented in Numerical Expressions 5 and 6 relates) contributing to the implementation of the Zeeman energy λi among components of light injected to each slave laser B, each master-to-slave phase control unit ZP performs control such that the injection phase for the slave laser B leads by π/2 with respect to the oscillation phase of the master laser M.
Above all, in the whole ising model computer, one oscillation mode is configured to be integrally started. Thus, in each pair of slave lasers B, when there is a case where the above-described oscillation mode is actually started, there is also a case where the oscillation mode is not necessarily started.
The computing sequence of the ising model is as follows. First, when an NP complete problem or the like is mapped into the ising model, Jij and λi are determined. Next, when the ising model is physically implemented, the magnitude and the sign of ξij are implemented in accordance with by using each inter-slave intensity control unit IA and each inter-slave optical path length control unit IP. In addition, the magnitude and the sign of ηi are implemented in accordance with λi, using each master-to-slave intensity control unit ZA and each master-to-slave phase control unit ZP.
Next, after each slave laser B reach the steady state, by measuring the oscillation phase φVi of each slave laser B using the oscillation phase measuring unit PM, the pseudo ising spin σi of each slave laser B is measured. Finally, the ground state of the ising model, which has been acquired, is mapped into a solution for the NP-complete problem or the like again.
In the above-described embodiment illustrated in
As a new modified example illustrated in
Above all, by applying the above-described embodiment illustrated in
In the above-described embodiment illustrated in
Next, simulation results of a second embodiment will be described. The simulation results of the temporal development of the oscillation phase of each site are illustrated in
In
In
In a case where α= 1/25 (
In a case where α= 1/200 (
As described above, the oscillation phase of light emitted from the slave laser B is associated with the pseudo spin direction of the slave laser B, and the pseudo spin state of the whole system is measured. Until the pseudo spin state of the whole system is settled to the ground state, quantum coherence of the whole system is not destroyed, and accordingly, the problem of exponentially increasing the computing time in accordance with the total number of sites is solved, whereby the computing time can be shortened to a large extent.
When the number of sites is M, it is sufficient to arrange M(M−1)/2 inter-slave optical path units IL, M(M−1)/2 inter-slave intensity control units IA, and M(M−1)/2 inter-slave optical path length control units IP. Accordingly, the number of required optical devices does not increase to 2M but is scaled to M2/2 with respect to the number M of sites. In a case where slave lasers B are applied instead of Bose-Einstein condensates, the whole system can be operated not at a low temperature but at a room temperature.
Thus, for an arbitrary pair out of M sites, through light exchanged between two sites, not only ising interaction between sites physically located close to each other but also ising interaction between sites located far from each other can be implemented. Accordingly, regardless of a physical distance between sites, an ising model in which any NP-complete problem or the like can be mapped can be solved.
In addition, for a number of pseudo spins in each site, whether one of the upward circulation and the downward circulation is more than the other is determined based on the rule of majority. Accordingly, even when there is a pseudo spin that is leaked from the ground state to the excited state, the probability of acquiring a correct solution is markedly higher in a case where there are a number of pseudo spins than that in a case where there is one pseudo spin.
Here, as the whole system transits from the initial state to the steady state, the oscillation phase of each slave laser B deviates from the oscillation phase of the master laser M, but the oscillated polarized light of each slave laser B is constantly the same as the oscillated polarized light of the master laser M.
Thus, for each slave laser B, a differential equation of the temporal developments of only three including the oscillation intensity AVi(t), the oscillation phase φVi(t), and the inverted population number difference NCi(t) of carriers may be formulated. Each slave laser B may have the single direction of oscillated polarized light, and accordingly, even in the case of a semiconductor surface emission laser (there is a slight difference in the frequency characteristics between oscillated light having vertically linear polarization and oscillated light having horizontally linear polarization) having in-plane anisotropy, the in-plane anisotropy does not matter. As above, the quantum computer using the ising model can be easily designed and manufactured.
The present disclosure can be applied to either a system in which only ising interaction is present or a system in which ising interaction and Zeeman energy are present together. In addition, according to the present disclosure, as will be described later with reference to
Each oscillation frequency control unit FC performs control of the interference intensity between oscillated light of each slave laser B and oscillated light of the master laser M to have an extreme value with respect to a change in the oscillation frequency of each slave laser B, thereby performing feedback control of the oscillation frequency of each slave laser B to be the oscillation frequency of the master laser M. Details thereof will be described as below.
The configuration of the slave laser B is illustrated in
Here, while reflected light of the frequency ωM, which is reflected from the light reflecting mirror BR arranged on the master laser M side, receives a phase change of π at the time of being reflected, outgoing light of the frequency ωS therefrom does not receive any phase change at the time of being transmitted. Accordingly, in a case where the frequencies ωM and ωS are synchronized with each other, destructive interference between the reflected light of the frequency ωM and the outgoing light of the frequency ωS may easily occur compared to a case where the frequencies ωM and ωS are not synchronized with each other. Therefore, by searching for a minimum of the total intensity of the reflected light of the frequency ωM and the outgoing light of the frequency ωS while sweeping the oscillation frequency ωS of the slave laser B, the oscillation frequency ωS of the slave laser B is controlled so as to be the oscillation frequency ωM of the master laser M.
The vertical axis illustrated in
Here, Δω and Δωmax on the horizontal axis illustrated in
Here, F0 is the amplitude of incident light, and A0 is the amplitude of oscillated light of a slave laser. A range of −1 to +1 will be considered as Δω/Δωmax. The reason for this is that injection synchronization cannot be achieved when the frequency ωS deviates from the frequency ωM by Δωmax or more.
As the state transits from a to d represented in
The configuration of the oscillation frequency control unit is illustrated in
The oscillation frequency control unit FC is configured by an optical splitter FCS, a photodiode FCC, and a mixer circuit FCM. A local oscillator ML performs frequency modulation of a carrier signal for the master laser M. A laser control unit BC performs temperature control or current control for the slave laser B, thereby controlling the oscillation frequency ωS.
The optical splitter FCS is arranged in the master-to-slave optical path unit ZL and splits reflected light and outgoing light, which are reflected and exits from the slave laser B, from incident light that is incident to the slave laser B. The photodiode FCC performs photoelectric conversion for the output of the optical splitter FCS. The mixer circuit FCM performs synchronous detection illustrated in
When the frequency ωS is lower than the frequency ωM, as the frequency ωS increases, the output of the photodiode FCC decreases. Accordingly, the output of the photodiode FCC has the same frequency and the opposite phase compared to the output of the local oscillator ML, whereby a result of the synchronous detection is negative.
On the other hand, when the frequency ωS is higher than the frequency ωM, as the frequency ωS increases, the output of the photodiode FCC increases. Accordingly, the output of the photodiode FCC has the same frequency and the same phase as those of the output of the local oscillator ML, whereby a result of the synchronous detection is positive.
In addition, when the frequencies ωS and ωM are the same, the output of the photodiode FCC is a minimum with respect to a change in the frequency ωS. Accordingly, the output of the photodiode FCC has a frequency that is twice the frequency of the output of the local oscillator ML, whereby a result of the synchronous detection is zero.
The laser control unit BC performs temperature control or current control for the slave laser B such that the result of the synchronous detection is zero, thereby controlling the oscillation frequency ωS. In addition, the frequency modulation width ΔωM of the master laser M is sufficient small, compared to the above-described injection locking width Δωmax (ΔωM/Δωmax˜0.03). Accordingly, upon performing frequency modulation of the master laser M, there is no influence on the precision of the injection synchronization.
The injection locking widths of the master laser and the slave laser are illustrated in
Here, <χr> is a real part of the electric susceptibility of a gain medium, <χi> is an imaginary part of the electric susceptibility of the gain medium, and <NCO> is the number of carriers in an active layer.
When the parameter α is zero, the injection locking width is as represented in Numerical Expression 17. On the other hand, when the parameter α is greater than zero, the injection locking width is as represented in Numerical Expression 18.
ΔωL illustrated in
Each inter-slave optical path length control unit IP performs control of the interference intensity of oscillated light of two slave lasers B to have an extreme value with respect to a change in the optical path length between the two slave lasers B, thereby performing feedback control of the optical path length between the two slave lasers B to be a half-integer multiple or integer multiple of the oscillation wavelength of the injection synchronization. Details thereof will be described as below. Here, the injection synchronization has been completed.
The configuration of the inter-slave optical path length control unit is illustrated in
A local oscillator BL performs phase modulation of a carrier signal for a slave laser Bi. The phase modulating unit IPP is arranged in the inter-slave optical path unit IL and controls an optical path length between slave lasers Bi and Bj based on a result of the synchronous detection to be described later.
The optical splitter IPS is arranged in the inter-slave optical path unit IL and splits outgoing light, which exits from the slave laser Bj, from incident light that is incident to the slave laser B. Here, the outgoing light exiting from the slave laser Bj receives a phase change of π. The photodiode IPC performs photoelectric conversion of the output of the optical splitter IPS. The mixer circuit IPM performs synchronous detection illustrated in
When the optical path length is less than π rad, as the optical path length increases, the output of the photodiode IPC increases. Accordingly, the output of the photodiode IPC has the same frequency and the same phase as those of the output of the local oscillator BL, whereby a result of the synchronous detection is positive.
On the other hand, when the optical path length is more than π rad, as the optical path length increases, the output of the photodiode IPC decreases. Accordingly, the output of the photodiode IPC has the same frequency and the opposite phase compared to the output of the local oscillator BL, whereby a result of the synchronous detection is negative.
In addition, when the optical path length is π rad, the output of the photodiode IPC is a maximum with respect to a change in the optical path length. Accordingly, the output of the photodiode IPC has a frequency that is twice the frequency of the output of the local oscillator BL, whereby a result of the synchronous detection is zero.
The phase modulating unit IPP performs control of the optical path length between the slave lasers Bi and Bj such that a result of the synchronous detection changes from a positive value to a negative value through zero as the optical path length increases.
When the optical path length is less than 0 rad, as the optical path length increases, the output of the photodiode IPC decreases. Accordingly, the output of the photodiode IPC has the same frequency and the opposite phase compared to the output of the local oscillator BL, whereby a result of the synchronous detection is negative.
When the optical path length is more than 0 rad, as the optical path length increases, the output of the photodiode IPC increases. Accordingly, the output of the photodiode IPC has the same frequency and the same phase as those of the output of the local oscillator BL, whereby a result of the synchronous detection is positive.
In addition, when the optical path length is 0 rad, the output of the photodiode IPC is a minimum with respect to a change in the optical path length. Accordingly, the output of the photodiode IPC has a frequency that is twice the frequency of the output of the local oscillator BL, whereby a result of the synchronous detection is zero.
The phase modulating unit IPP performs control of the optical path length between the slave lasers Bi and Bj such that a result of the synchronous detection changes from a negative value to a positive value through zero as the optical path length increases.
In addition, the phase modulation of the slave laser B is performed based on current modulation control. The current modulation control (˜10 MHz) causes phase modulation but does not cause a frequency shift. Accordingly, upon performing phase modulation of the slave laser B, there is no influence on the precision of the injection synchronization.
As above, the oscillation phase measuring unit PM measures an ising spin. In performing this process, the oscillation frequency control unit FC synchronizes the whole quantum computer, and the inter-slave optical path length control unit IP implements ising interaction.
Here, the oscillation frequency control unit FC adjusts the oscillation frequency of the slave laser B, and the inter-slave optical path length control unit IP adjusts the optical path length between slave lasers B. However, the oscillation phase measuring unit PM measures the ising spin in a state in which the oscillation frequency of the slave laser B and the optical path length between the slave lasers B are fixed. Accordingly, an oscillation frequency control step and an inter-slave optical path length control step cannot be performed simultaneously with an oscillation phase measuring step. The sequence of each step will be described as below.
In the inter-slave laser optical path length control step, in consideration of characteristics of the device, a time required for phase modulation of the slave laser B per each time is ˜10−5 s, and a time during which the optical path length is maintained to be stabilized is ˜104 s. In the oscillation frequency control step, in consideration of the characteristics of the device, a time required for frequency modulation of the master laser M per each time is ˜10−3 s, and a time during which the frequency is stabilized is ˜10−2 s.
As above, in the oscillation phase measuring step, the inter-slave laser optical path length control step, and the oscillation frequency control step, the time scales are separated. In other words, when the oscillation phase measuring step is performed after the oscillation frequency control step and the inter-slave laser optical path length control step, there is no change in the oscillation frequency and the inter-slave laser optical path length during the measurement of the oscillation phase. Accordingly, for the whole quantum computer using the ising model, one oscillation mode is assured to be integrally started, the ising interaction can be correctly implemented, and the ising model can be correctly computed.
For a case where the oscillation phase of the slave laser B is initialized to zero, the elapse of time in the ising model computation process is illustrated in
Thus, the pumping current control unit PC is applied. The pumping current control unit PC, for two slave lasers B performing pseudo ising interaction through each inter-slave optical path unit IL, performs gradual increase control of pumping currents of the two slave lasers B in a state in which the magnitude and the sign of the pseudo ising interaction between the two slave lasers B are implemented to have fixed values and performs fixing control of the pumping currents of the two slave lasers B at a time point when all the slave lasers B arrive at one integrally oscillation mode “for the first time”.
After all the slave lasers B arrive at one integrally oscillation mode “for the first time” so as to arrive at the steady state, the oscillation phase measuring unit PM measures relative values of the oscillation phases of all the slave lasers B with respect to the oscillation phase of the master laser M, thereby measuring pseudo ising spins of all the slave lasers B. The oscillation mode that has arrived at “for the first time” corresponds to the ground state.
For a case where gradual increase control of the pumping current of the slave laser B is performed, the elapse of time in the ising model computation process is illustrated in
Before pumping, the inverted population of carriers is not implemented, and any oscillation mode is not implemented. In the pumping step, the inverted population of carriers is implemented, and, in accordance with a gradual increase in the pumping current, an oscillation mode in which the inverted population number difference ENc, of carriers is minimal is selected, in other words, an oscillation mode in which the threshold gain ΣECVi is minimal is selected, and finally, the state is settled into the ground state.
As above, by using the pumping current control unit PC, the ground state of the ising model can be correctly arrived at without the state incorrectly being trapped into a quasi-stable state of the ising model.
While the quantum computer has been applied to the ising model including the ising interaction and the Zeeman energy as above, in this section, the quantum computer is applied to a problem of MAX-CUT-3. Here, the problem of MAX-CUT-3 will be described. In a graph configured by V nodes and E edges, each node has three adjacent nodes, and the edges have the same weighting or the same length. When V nodes are divided into two, the number of divided edges is configured to be minimal. The problem of MAX-CUT-3 is proved to be the NP-complete problem and is mapped into an ising model as represented in Numerical Expression 19.
Here, a problem of MAX-CUT-3 in which four sites form a graph of a regular tetrahedron will be considered. When Jij =+1 at adjacent sites, in order to minimize the Hamiltonian, it is preferable that the directions of spins at adjacent sites be opposite to each other. However, even though the direction of the spin can be opposite at a certain adjacent site, the problem of the spin frustration, that the direction of the spin is inevitably the same direction at another adjacent site occurs.
In a case where the problem of spin frustration occurs, a state is implemented in which a plurality of ground states are superimposed. Here, in order to statistically measure many photons, the oscillation phase measuring unit PM does not measure one ground state out of the plurality of ground states but measures an average state of the plurality of ground states. Therefore, “0 spin” occurs.
However, since any of the plurality of ground states leads a correct result, any one of the plurality of ground states may be selected. Thus, by fixing the direction of the spin at a specific site, candidates for the direction of the spin at another site may be narrowed. Here, in order to fix the direction of the spin at a specific site, Zeeman energy may be applied at the specific site by injecting light having vertically linear polarization |V>, for which phase modulation has been performed, from the master laser M to the slave laser B corresponding to the specific site through the master-to-slave intensity control unit ZA and the master-to-slave phase control unit ZP corresponding to the specific site.
The spin frustration is illustrated in
A case where the spin frustration is incorrectly resolved is illustrated in
Then, the direction of the pseudo spin of the slave laser B1 is determined to be +1, and the directions of pseudo spins of the slave lasers B2 to B4 are determined to be −1. However, (σ1,σ2,σ3,σ4)=(+1,−1,−1,−1) is an incorrect result, and a correct result is any one of (σi,σ2,σ3,σ4)=(+1,+1,−1,−1), (+1,−1,+1,−1), and (+1,−1,−1,+1). The reason for the acquisition of the incorrect result is that the oscillation phase measuring unit PM measures an average state of the above-described three ground states so as to measure many photons in a statistical manner.
A case where the spin frustration is correctly resolved is illustrated in
As illustrated in
Then, the direction of the pseudo spin of the slave laser B1 is determined to be +1, and the direction of the pseudo spin of the slave laser B2 is determined to be −1. However, the directions of pseudo spins of the slave lasers B3 and B4 are not determined. The reason for the occurrence of “0 spin” is that the oscillation phase measuring unit PM measures an average state of two ground states (σ1, σ2,σ3,σ4)=(+1,−1,+1,−1) and (+1,−1,−1,+1) so as to measure many photons in a statistical manner.
As illustrated in
Then, as illustrated in
When the description presented above is taken together, the master-to-slave intensity control unit ZA and the master-to-slave phase control unit ZP correspond to an adjacent ising spin direction fixing unit. Here, for two slave lasers B performing pseudo ising interaction through the inter-slave laser optical path unit ZL, there are cases where relative values of the oscillation phases of the two slave lasers B with respect to the oscillation phase of the master laser M are not measured to be meaningful.
Thus, the master-to-slave intensity control unit ZA and the master-to-slave phase control unit ZP control the intensity and the phase of light injected to each one of the two slave lasers B. Then, when the sign of the pseudo ising interaction between the two slave lasers B is implemented to be positive, the master-to-slave intensity control unit ZA and the master-to-slave phase control unit ZP fix the directions of the pseudo spins of the two slave lasers B to be different from each other. Alternatively, when the sign of the pseudo ising interaction between the two slave lasers B is implemented to be negative, the master-to-slave intensity control unit ZA and the master-to-slave phase control unit ZP fix the directions of the pseudo spins of the two slave lasers B to be the same. Therefore, the frustration of the pseudo spin between the two slave lasers B performing pseudo ising interaction can be resolved.
Here, when the directions of spins are fixed in sites adjacent to each other, there are cases where in a site adjacent to the sites adjacent to each other, the direction of the spin, which has not been fixed yet, is flipped. Accordingly, the frustration of a pseudo spin cannot be resolved by only fixing the directions of spins at the sites adjacent to each other. However, by fixing the direction of the spin at the site adjacent to the sites adjacent to each other, the frustration of the pseudo spin can be resolved.
Here, a problem of MAX-CUT-3 formed by six sites will be considered. The problem of flipping the direction of the spin of a peripheral site, which occurs when the directions of the spins of the adjacent sites are fixed, is illustrated in
Before the directions of pseudo spins of the slave lasers B5 and B6 are fixed to −1 and +1, the directions of pseudo spins of the slave lasers B1, B2, B3, and B4 are provisionally determined to be −1, +1, +1, and −1.
Here, as illustrated in
However, as illustrated in
When the description presented above is taken together, the master-to-slave intensity control unit ZA and the master-to-slave phase control unit ZP correspond to a peripheral ising spin direction fixing unit. Here, for two slave lasers B performing pseudo ising interaction through the inter-slave laser optical path unit ZL, there are cases where relative values of the oscillation phases of the two slave lasers B with respect to the oscillation phase of the master laser M are not measured to be meaningful.
Thus, the master-to-slave intensity control unit ZA and the master-to-slave phase control unit ZP control the intensity and the phase of light injected to each one of the adjacent slave lasers B performing pseudo ising interaction with the two slave lasers B through another inter-slave laser optical path unit ZL. Then, the master-to-slave intensity control unit ZA and the master-to-slave phase control unit ZP fix the directions of the pseudo spins of the adjacent slave lasers B to the directions of the current time point during the computation process. Accordingly, when the frustration of the pseudo spin between two slave lasers B performing pseudo ising interaction is resolved, it can be prevented that pseudo spins are unintentionally flipped in the slave lasers B that are pseudo-adjacent to the two slave lasers B.
A noise with respect to the oscillation phase of the slave laser is illustrated in
Here, in the slave laser B, since the oscillation phase φVi is associated with a spin σi, the influence of a noise FφV on the oscillation phase φVi is received much, but the influence of a noise FN on the oscillation intensity AVi and the inverted population number difference No of carriers is hardly received. Thus, the noise FφN on the oscillation phase φVi will be described.
First, the quantum computer is operated for a predetermined time using the phase noise FφV (Step S1). Here, the predetermined time is a time during which the direction of a pseudo spin can be determined when the frustration of the pseudo spin is dissolved.
When there is no “0 spin” (No in Step S2), the frustration of the pseudo spin is dissolved, and the quantum computing ends. On the other hand, when there is “0 spin” (Yes in Step S2), the frustration of the pseudo spin is not dissolved, the quantum computing is continued, and the direction of the pseudo spin of “0 spin” is fixed.
When there is a “0 spin pair” (Yes in Step S3), as described with reference to
On the other hand, when there is no “0 spin pair” (No in Step S3), all the “isolated 0 spins” and spins on the periphery thereof are fixed (Step S5).
Regardless of fixing all the “isolated 0 spins” and the spins on the periphery thereof (Step S5), there are cases where λi is not updated. It is assumed that the reason for this is not that there is an incorrect solution but that a difference in the photon numbers of light having oscillation phases φVi˜±π/2 is small, whereby the signal-to-noise ratio S/N is low.
When there is an update of λi (Yes in Step S6), the process is returned to Step S1. On the other hand, when there is no update of λi (No in Step S6), by increasing the magnitude of λi, the signal-to-noise ratio S/N is improved by increasing a difference in the photon numbers of light having oscillation phases φvi˜±π/2 (Step S7), and the process is returned to Step S1.
A delay time of pseudo ising interaction between two slave lasers B performing pseudo ising interaction may have influence on the stability of the quantum computer.
The simulation of the temporal development as illustrated in
Thus, in order to stably operate the quantum computer, it is necessary that td <6×10−10 seconds. In other words, it is necessary to set the optical path length between two slave lasers B performing pseudo ising interaction to be shorter than 3×108 m/s×6×10−1° s˜10 cm. In addition, it is necessary that the delay time of the pseudo ising interaction between two slave lasers B performing pseudo ising interaction is shorter than the reciprocal (˜ 1/30 GHz) of the injection locking width of the master laser M and the two slave lasers B.
Methods for mounting an ising model computer are illustrated in
The ising model computer is configured by slave lasers B1, B2, . . . , BM, a silicon waveguide SW, and a micro electro mechanical system (MEMS) mirror XC. The silicon waveguide SW is connected to the MEMS mirror XC and has oscillated light control units C1, C2, . . . , CM and a master laser M built therein. The MEMS mirror XC is connected to the silicon waveguide SW and has inter-slave optical path units IL12, IL1M, IL2M, and the like and a mirror array MEM built therein.
The oscillated light control units C1, C2, . . . , CM are applied to the slave lasers B1, B2, BM. The master laser M is applied in common to the slave lasers B1, B2, . . . , BM.
The inter-slave optical path units IL12, IL1M, and IL2M are arranged between the slave lasers B1 and B2, the slave lasers B1 and BM, and the slave lasers B2 and BM. The mirror array MEM has light being incident thereto from the output end of a certain slave laser B and reflects light to the input end of another slave laser B.
In addition, the oscillated light control unit C has the master-to-slave phase control unit ZP, the master-to-slave intensity control unit ZA, the oscillation frequency control unit FC, and the master-to-slave optical path unit ZL built therein as portions implementing Zeeman energy. In addition, the oscillated light control unit C has the inter-slave optical path length control unit IP, the inter-slave intensity control unit IA, and the inter-slave optical path unit IL built therein as portions implementing ising interaction. Here, the problem of MAX-CUT-3 is considered to be solved, and only three systems of the portions implementing the ising interaction are mounted.
As above, by applying the silicon waveguide SW, amplitude modulation and phase modulation can be performed at high speed. In addition, by applying the MEMS mirror XC, patterns of two-body interaction different for each ising model can be mounted.
In the description presented above, while the slave laser B is applied as a system having a pseudo spin, in the present modified example, an exciton polariton included in a semiconductor micro cavity or the like may be applied. A ground state in which the kinetic energy of all the exciton polaritons of the whole system is minimum is realized, and, by measuring the oscillation phase generated by each exciton polariton, the pseudo spin of each exciton polariton is measured. In order to improve the computation accuracy, for the energy of all the exciton polaritons of the whole system, an energy difference between the ground state and the first excited state needs to be sufficiently larger than kBT (here, T is the temperature of the whole system).
As a Bose-Einstein condensate, even when other than the exciton polariton, for example, a Bose-Einstein condensate of a photon may be applied, pseudo ising interaction between sites may be implemented through exchange of photons between sites, and pseudo Zeeman energy may be implemented by injecting photons into each site from the master laser M.
The quantum computer and the quantum computing using the ising model of the present disclosure are appropriate for solving an NP-complete problem or the like mapped into an ising model easily and fast. For example, in a case where an ising model having the number of sites M=140 is to be solved, in a general computer or a quantum computer, while the computation time is 1042 times longer than that of a case where M=2, according to the quantum computer and the quantum computing using an ising model of the present disclosure, the computation time is only 104 times longer than that of a case where M=2 in the worst case. In other words, the computation time is shortened by 38 digits.
- B1, B2, B3: Bose-Einstein condensing unit
- D1, D2, D3: spin measuring unit
- F: feedback control circuit
- I1, I2, I3: ising interaction implementing unit
- B, B1, B2, B3, B4, B5, B6, Bi, Bj, BM: slave laser
- M: mater laser
- ZL, ZL1, ZL2, ZL3: master-to-slave optical path unit
- IL, IL12, IL23, IL13, IL1M, IL2M: inter-slave optical path unit
- FC, FC1, FC2, FC3: oscillation frequency control unit
- IA, IA12, IA23, IA13: inter-slave intensity control unit
- IP, IP12, IP23, IP13: inter-slave optical path length control unit
- ZA, ZA1, ZA2, ZA3: master-to-slave intensity control unit
- ZP, ZP1, ZP2, ZP3: master-to-slave phase control unit
- PM: oscillation phase measuring unit
- PC: pumping current control unit
- BR: light reflecting mirror
- BG: amplified medium
- BC: laser control unit
- BL: local oscillator
- ML: local oscillator
- FCS: optical splitter
- FCC: photodiode
- FCM: mixer circuit
- IPS: optical splitter
- IPC: photodiode
- IPM: mixer circuit
- IPP: phase modulating unit
- SW: silicon waveguide
- C, C1, C2, CM: oscillated light control unit
- XC: MEMS mirror
- MEM: mirror array
| 标题 | 发布/更新时间 | 阅读量 |
|---|---|---|
| 基于量子真随机数矩阵实现一次一密的IP加密方法 | 2020-05-11 | 285 |
| 基于真随机流异或加密的高安全IP保密通信方法 | 2020-05-16 | 268 |
| 以统一场论和全息电视为基础实现量子计算机的方法 | 2020-05-14 | 839 |
| 一种机器人创意动作评估生成系统 | 2020-05-15 | 396 |
| 基于身份的抗量子计算数字签名方法和系统 | 2020-05-14 | 185 |
| 基于非对称密钥池对和密钥卡的外网接入身份认证方法和系统 | 2020-05-14 | 977 |
| 基于一次性非对称密钥对和可移动身份识别装置的抗量子计算门禁方法和系统 | 2020-05-13 | 455 |
| 一种标准模型下格上基于身份的盲签名方法 | 2020-05-12 | 234 |
| 基于非对称密钥池的量子通信服务站身份认证方法和系统 | 2020-05-16 | 557 |
| 基于公钥池和签名偏移量的抗量子证书颁发方法及系统 | 2020-05-08 | 396 |
高效检索全球专利专利汇是专利免费检索,专利查询,专利分析-国家发明专利查询检索分析平台，是提供专利分析，专利查询，专利检索等数据服务功能的知识产权数据服务商。
我们的产品包含105个国家的1.26亿组数据，免费查、免费专利分析。
专利汇分析报告产品可以对行业情报数据进行梳理分析，涉及维度包括行业专利基本状况分析、地域分析、技术分析、发明人分析、申请人分析、专利权人分析、失效分析、核心专利分析、法律分析、研发重点分析、企业专利处境分析、技术处境分析、专利寿命分析、企业定位分析、引证分析等超过60个分析角度，系统通过AI智能系统对图表进行解读，只需1分钟，一键生成行业专利分析报告。
