CN115332440A - Ultra-compact non-volatile photonic neurosynaptic device - Google Patents
Ultra-compact non-volatile photonic neurosynaptic device Download PDFInfo
- Publication number
- CN115332440A CN115332440A CN202210883229.9A CN202210883229A CN115332440A CN 115332440 A CN115332440 A CN 115332440A CN 202210883229 A CN202210883229 A CN 202210883229A CN 115332440 A CN115332440 A CN 115332440A
- Authority
- CN
- China
- Prior art keywords
- phase
- ridge waveguide
- ultra
- film
- phase change
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Landscapes
- Optical Integrated Circuits (AREA)
Abstract
The invention discloses an ultra-compact non-volatile photonic neurosynaptic device, and belongs to the field of optical computing. The phase-change ridge waveguide structure comprises a substrate, a ridge waveguide, a phase-change film and a metal film, wherein the ridge waveguide is arranged above the substrate and covers the whole substrate area, the phase-change film is distributed above a ridge type area of the ridge waveguide, the metal film covers the phase-change film, the geometric parameters of the phase-change film are the same as those of the phase-change film, a Surface Plasmon Polariton (SPPs) mode of a basic mode in the ridge waveguide and the interface of the phase-change film and the metal film forms a superposed optical field, the geometric structure of the ridge waveguide influences the effective mode refractive index of the superposed mode, and different normalized intensity outputs are further realized by optimizing the geometric parameters of the periodic ridge waveguide and the length of a modulation area. By changing the phase state (crystalline state and amorphous state) of the phase change film, the photon nerve synapse has larger output intensity difference under the condition that the phase change film is in different phase states.
Description
Technical Field
The invention belongs to the field of optical computing, and particularly relates to an ultra-compact non-volatile photonic neurosynaptic device.
Background
The traditional electrical computing system can not meet the requirements of modern computing and communication systems gradually due to the problems of low computing speed, high energy consumption, approaching the mole limit of the semiconductor integrated process and the like. With the gradual reduction of the size of the transistor, the problems of power consumption and heat dissipation of the chip become more and more serious, which severely limits the development of large-scale integrated chips and the further improvement of the computational efficiency of the chip. To solve this problem, researchers have proposed solutions to the problems of energy consumption and computational rate using optical signal processing systems, such asSuch as an optical processing chip based on a neuromorphic computing network architecture. For example, the brain-like chip based on the neuromorphic computing technology can imitate the human brain working principle to realize quick learning, and can meet the complex information processing condition through continuous autonomous learning. Generally, a neuromorphic computing network consists of approximately 10 11 And each artificial neuron is used for adjusting the weight based on an adjustable neural synapse. Therefore, the tunability of a single neurosynaptic will greatly affect the computational efficiency and energy consumption of an optical neuromorphic network. Synapses are important structures for signal transmission and information exchange of neurons, and are also the basis of neuromorphic computing technology. Optical synapses and neurons may be designed using differences in crystalline and amorphous optical properties of Phase Change Materials (PCMs), and may be directly modulated by optical pulses. PCMs have excellent optical property contrast between covalently bonded amorphous and resonantly bonded crystalline states and have found widespread adoption in reconfigurable photonic applications, such as optical switches, optical routers, and hypersurfaces. Photonic synapses based on PCMs maintain the switching state without the need for a static power supply and have modulation speeds on the order of picoseconds are considered as key unit devices for designing optical neuromorphic networks.
However, when the PCMs are sputter deposited directly on the surface of the silicon waveguide, the output transmission intensity changes little at different phase change states. This is because the waveguide mode is mainly concentrated in the silicon-based waveguide region, and only a small part of the energy of the optical field can interact with the PCMs, so that the change of the phase state of the PCMs hardly has a large influence on the optical field. Therefore, researchers have designed nonvolatile photonic neurosynaptic devices that are more sensitive to phase change of phase change materials, such as nonvolatile optical switches based on mach-zehnder (MZ) structures and nonvolatile optical switches based on ring couplers. Such non-volatile photonic neural synapses utilizing interference and coupler effects are more sensitive to changes in the phase states of PCMs, but have large physical dimensions, which severely hamper large-scale integration of non-volatile optical neuromorphic networks. To address this problem, some reports have proposed designing special small-sized waveguide structures, such as periodic arrays of PCMs, land waveguides, and surface plasmon waveguides, to enhance the interaction effect between the waveguides and the PCMs. However, the intensity difference of the output transmittance between different PCMs is still very small, about 40%, which is not beneficial to designing multi-level adjustable neuro-photonic synapses and limits the development of non-volatile photonic neural networks.
Disclosure of Invention
In view of the defects of the prior art, the present invention aims to provide an ultra-compact non-volatile photonic neurosynaptic device, which aims to solve the problem that the normalized output intensity range of the photonic neurosynaptic is too small.
In order to achieve the above objects, the present invention provides an ultra-compact non-volatile photonic neurosynaptic device, which includes a substrate, a ridge waveguide disposed above the substrate and covering the entire substrate region, a phase change film disposed above the ridge region of the ridge waveguide, and a metal film covering the phase change film and having the same geometric parameters as the phase change film.
Input light enters the photonic nerve synapse device for transmission after passing through the coupling waveguide, and different cross-section geometrical parameters of the ridge waveguide influence the optical field distribution in the ridge waveguide. Meanwhile, the input light excites Surface Plasmon Polariton (SPPs) waves at the interface of the metal thin film and the phase change thin film, and the wavelength and optical field distribution of the SPPs waves are related to the refractive indexes of the metal thin film and the phase change thin film. Therefore, the optical field in the ridge waveguide and the SPPs optical field at the interface of the metal film and the phase change film finally form a superposed optical field, part of energy of the superposed optical field is distributed in the phase change film, and the interaction strength between the optical field and the phase change material is increased. When the phase state of the phase change film is changed, the output intensity of the photonic synapse will change significantly.
Furthermore, through the cross section geometrical parameters (height and width) of the ridge waveguide, the ultra-compact non-volatile photonic neurosynaptic device can realize the maximum optical output intensity regulation range when the phase change thin film is in different phases. The thickness of the flat plate layer of the ridge waveguide is more than or equal to 0, and the specific value is related to the practical optimization condition of the geometrical parameters of the cross section.
Preferably, the structure of the ultra-compact non-volatile photonic neurosynaptic device is a T-shaped structure, and the initial cross-sectional geometrical parameters are as follows: the height of the flat plate layer is 100nm, the height of the ridge region is 240nm, and the width of the silicon waveguide of the ridge region is 300nm. By changing the parameters, the difference value of the effective mode refractive index imaginary part of the photonic neural synaptic inner mode is larger when the phase change film is in different phase states, the regulation range of the phase change film on the output intensity of the photonic neural synapse is larger, and more optical field energy can be stably transmitted in the waveguide when the phase change film is in an amorphous state; when the phase-change film is in an amorphous state, most of the optical field energy needs to be lost in the transmission process, so that the purpose of regulating the output intensity of the photonic neurosynaptic is achieved.
Further, the material of the phase change film is Ge 2 Sb 2 Se 4 Te 1 (GSST)。
Further, when the phase change film is amorphous and crystalline, the real part of the refractive index is smaller and larger than the real part of the refractive index of the metal film, respectively. Preferably, when the phase-change film is amorphous, its imaginary refractive index should be approximately equal to 0 to reduce the insertion loss of the device.
Further, the metal film is selected to ensure that the SPPs excited at the interface between the phase change film and the metal film have a minimum transmission loss.
Further, the material of the ridge waveguide is silicon or other low-loss semiconductor material.
Further, the metal film structure further comprises a cladding layer, and the cladding layer is arranged above the metal film. The cladding may be air or other material such as silica.
Further, the cross-sectional geometrical parameters of the ridge waveguide, the phase change film and the metal film are constant or slowly changed in the propagation direction.
Further, the lengths of the phase change film and the metal film in the propagation direction are selected so that the SPPs waves generated at the incident end and the SPPs waves reflected at the interface of the emergent end satisfy the constructive interference condition.
Further, the ultra-compact non-volatile photonic neurosynaptic device can be processed on a silicon-on-insulator (SOI) platform with a height of 340nm and other semiconductor platforms with a height of 340nm.
Through the technical scheme, compared with the prior art, the invention has the following beneficial effects:
1. in order to realize larger output transmission intensity difference between different phase states of PCMs in a shorter modulation distance, the invention provides an ultra-compact non-volatile photonic neurosynaptic device on a silicon platform.
2. The modulation distance of the ultra-compact non-volatile photonic neurosynaptic device provided by the invention is about one micron, which is far smaller than the modulation distance required by the existing non-volatile photonic neurosynaptic device for realizing the modulation range of the maximum output intensity, and the large-range change of the normalized output intensity is realized.
3. On the basis that the output intensity difference of adjacent regulation levels exceeds 1%, the ultra-compact non-volatile photon nerve synapse device can realize coding exceeding 64 levels (6-bit), and is twice as large as other output intensity regulation type non-volatile photon nerve synapse devices.
4. The ultra-compact non-volatile photonic neurosynaptic device provided by the invention has an extremely small unit size, is beneficial to large-scale integration, and has very important significance and value in future optical neural network design and optical computing application.
Drawings
FIG. 1 is a two-dimensional cross-sectional schematic diagram of an ultra-compact non-volatile photonic neurosynaptic device.
FIG. 2 is a diagram of simulation results obtained from an ultra-compact non-volatile photonic neurosynaptic device, where (a) and (b) represent normalized electric field distributions of a conventional ridge waveguide and a hybrid waveguide proposed by the present invention, respectively, when the phase change film 3 is amorphous; (c) And (d) respectively represent the normalized electric field distribution of the traditional ridge waveguide and the hybrid waveguide provided by the invention when the phase change film 3 is crystalline.
FIG. 3 is a diagram of simulation results obtained from an ultra-compact non-volatile photonic neurosynaptic device, (a), (b) represent the normalized output intensity of the photonic neurosynaptic as a function of the width of the ridge region and the height of the slab of the ridge waveguide 2 when the phase-change thin film 3 is in the amorphous state and the crystalline state, respectively, and (c) represent the normalized output intensity difference between different phases as a function of the width of the ridge region and the height of the slab of the ridge waveguide.
FIG. 4 is a diagram of simulation results obtained from an ultra-compact non-volatile photonic neurosynaptic device, where (a) and (b) represent the distribution of transmission electric field when the phase-change film 3 is in amorphous state and crystalline state, respectively.
Fig. 5 is a graph of modulation distance versus normalized optical output intensity.
FIG. 6 is a graph of simulation results of multi-level tunability of an ultra-compact non-volatile photonic neurosynaptic device.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and do not limit the invention. In addition, the technical features involved in the embodiments of the present invention described below may be combined with each other as long as they do not conflict with each other.
The invention provides an ultra-compact non-volatile photonic neurosynaptic device which comprises a substrate 1, a ridge waveguide 2, a phase change film 3 and a metal film 4, wherein the ridge waveguide 2 is arranged above the substrate 1 and covers the whole substrate 1 area, the phase change film 3 is distributed above a ridge type area of the ridge waveguide 2, the metal film 4 covers the phase change film 3, and the geometric parameters of the phase change film 3 are the same as those of the phase change film 3.
The silicon waveguide width of the ridge waveguide 2 is 140nm, the height of the flat plate layer is 100nm, the height of the ridge region is 240nm, the thickness of the phase change film 3 and the metal film 4 is 50nm, and the length of the phase change film and the metal film is 1mm, and the silicon waveguide width, the flat plate layer and the ridge region are used as initial geometric parameters of the ultra-compact non-volatile photonic neurosynaptic device.
Specifically, the phase change thin film material is selected from Ge 2 Sb 2 Se 4 Te 1 (GSST)。
Specifically, the choice of metal thin film material is gold.
Specifically, the sum of the slab layer height of the ridge waveguide 2 and the ridge region height is always 340nm.
Examples
As shown in fig. 1, the multilevel dimmable switch of the present embodiment includes a substrate 1, a ridge waveguide 2, a phase change film 3, and a metal film 4; the ridge waveguide 2 covers the whole substrate 1 area, the phase change film 3 is distributed on the upper side of the ridge waveguide 2 partial area, the metal film 4 covers the upper side of the phase change film 3, and the geometric parameters of the metal film are the same as those of the phase change film 3. Input light enters the photonic neurosynaptic device for transmission after passing through the coupling waveguide, and different cross-sectional geometrical parameters of the ridge waveguide 2 influence the optical field distribution in the ridge waveguide 2. Meanwhile, the input light excites SPPs waves at the interface of the metal thin film 4 and the phase change thin film 3, and the wavelength and the optical field distribution of the SPPs waves are related to the refractive indexes of the metal thin film 4 and the phase change thin film 3. Therefore, the optical field in the ridge waveguide 2 and the SPPs optical field at the interface of the metal film 4 and the phase change film 3 finally form a superposed optical field, and part of energy of the superposed optical field is distributed in the phase change film, so that the interaction strength between the optical field and the phase change material is increased. When the phase state of the phase change film 3 is changed, the output intensity of the photonic synapse will change significantly.
For the conventional ridge waveguide, the mode energy distribution change caused by the phase change film 3 changing the phase state is not obvious, and the ultra-compact non-volatile photonic neurosynaptic device provided by the invention can realize the remarkable mode energy distribution change, as shown in fig. 2. The geometry of the cross-section of the hybrid waveguide, such as slab height, ridge region height, and ridge waveguide width, affects the real and imaginary components of the fundamental mode effective index, and thus the output transmittance strength of the hybrid waveguide. By scanning these parameters, a photonic neurosynaptic device with optimal multi-level tunability is determined. For the ultra-compact non-volatile photonic neurosynaptic proposed by the present invention, the structural optimization effect is analyzed and quantified by monitoring the normalized output intensity. The variation of normalized output intensity difference (Δ T) was obtained using three-dimensional time-domain finite difference method (FDTD) simulation to quantify the multi-level tunability of our non-volatile photonic neurosynaptic devices.
Specifically, the silicon waveguide width and slab layer height of the ridge waveguide 2 were changed to observe changes in the optical output intensity difference value when the phase change film 3 was in different phase states. The normalized output intensity as a function of the silicon waveguide width and slab height of the ridge waveguide 2 is shown in fig. 3. It can be seen that the normalized output intensity of the non-volatile photonic neurosynaptic increases slowly with increasing silicon waveguide width and slab layer height of the ridge waveguide 2 when the GSST is in the amorphous and crystalline states. And the normalized output intensity reaches a maximum value when the silicon waveguide width is 700nm and the slab layer height is 200 nm. The reason why the optical output intensity difference gradually increases in the entire modulation range is that when the silicon waveguide width and the slab layer height of the ridge waveguide 2 are small, the mode field cannot stably exist in the silicon waveguide region but radiates in the external environment, thereby reducing the interaction between the superimposed optical field and the phase change film 3, and therefore the transmission loss caused by the GSST material is reduced. In addition, it can be found that the difference between the normalized output intensities of the amorphous and crystalline phase-change films 3 reaches a maximum value of 87.9%, which is almost twice that of the existing structure, and the method is suitable for multi-stage adjustable non-volatile photonic neurosynaptic design in practical application.
Specifically, the effect of the modulation distance on the normalized output intensity of the proposed non-volatile photonic neurosynaptic device is observed by varying the length of the phase change thin film 3 and the metal thin film 4. In fact, part of the optical field energy present in the phase change film 3 will be reflected at the device-to-air interface, and when the input light interferes with the reflected light, the real part of the mode effective index has a large effect on the normalized output intensity of the output device. The distribution of the cross-sectional normalized transmission electric field of the phase-change film 3 in the amorphous and crystalline states under the initial geometric parameters is shown in fig. 4. When GSST is amorphous, the electric field is not uniformly attenuated in the direction of propagation, but periodically increases and decreases. This is due to the reflection of SPPs waves transmitted at the interface of the gold film and the output terminal GSST, the reflected signal interfering with the SPPs waves, resulting in a periodic increase or decrease in the optical field distribution. Thus, if the output transmittance intensity of the waveguide is modulated with both the real and imaginary parts of the mode effective index, the difference in output transmittance intensity of the waveguide will be much greater when the GSST is in different phases. Since the interference effect is closely related to the propagation distance, the invention analyzes the influence of the modulation distance on the normalized output intensity of the nonvolatile photonic neurosynaptic device in a simulation way. Fig. 5 is a graph of the effect of varying the modulation distance on the normalized output intensity of the device. As can be seen from fig. 5, the output transmittance intensity difference shows a periodic variation in the process of increasing the modulation distance, the period is about 245nm, and the maximum value is 89.2% at a modulation distance of 1010 nm.
Specifically, in the multi-level tunable test of the ultra-compact non-volatile photonic synapse, the optical refractive index of the phase-change film 3 may be changed by changing the degree of crystallinity of the phase-change film 3, wherein the relationship between the degree of crystallinity and the optical refractive index satisfies:
wherein p represents the degree of crystallization, e a And e c Preferably, the phase change material used in the present invention is G SST, and the refractive indices of the amorphous and crystalline states at 1550nm wavelength are 2.9817+0.005i and 4.9956+0.2938i, respectively. Simulation results show that a multi-level dimmable switch can control the light output intensity at 65 different levels, corresponding to approximately 6-bit programming resolution, as shown in fig. 6.
It will be understood by those skilled in the art that the foregoing is only a preferred embodiment of the present invention, and is not intended to limit the invention, and that any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the scope of the present invention.
Claims (10)
1. The ultra-compact non-volatile photonic neurosynaptic device is characterized by comprising a substrate (1), a ridge waveguide (2), a phase change film (3) and a metal film (4), wherein the ridge waveguide (2) is arranged above the substrate (1), the phase change film (3) is distributed above a ridge type region of the ridge waveguide (2), and the metal film (4) is distributed above the phase change film (3).
2. An ultra-compact non-volatile photonic neurosynaptic device according to claim 1, wherein the structure of the ridge waveguide (2) comprises a T-shaped structure.
3. Ultra-compact non-volatile photonic neurosynaptic device according to claim 1, wherein the material of the phase change thin film (3) is Ge 2 Sb 2 Se 4 Te 1 。
4. An ultra-compact non-volatile photonic neurosynaptic device according to claim 1, characterized in that the phase change thin film (3) has a real refractive index smaller and larger than the real refractive index of the metal thin film (4) when in the amorphous and crystalline states, respectively.
5. The ultra-compact non-volatile photonic neurosynaptic device according to claim 1, wherein the metal thin film (4) is gold.
6. An ultra-compact non-volatile photonic neurosynaptic device according to claim 2, wherein the material of said ridge waveguide (2) is silicon.
7. The ultra-compact non-volatile photonic neurosynaptic device according to claim 1, further comprising a cladding (5), said cladding (5) being disposed above said metal thin film (4).
8. An ultra-compact non-volatile photonic neurosynaptic device according to claim 1, wherein cross-sectional geometrical parameters of the ridge waveguide (2), the phase change thin film (3) and the metal thin film (4) are fixed or slowly varying in the propagation direction.
9. The device of claim 1, wherein the lengths of the phase change film (3) and the metal film (4) in the propagation direction are selected such that the SPPs waves generated at the incident end and reflected at the exit end interface satisfy the constructive interference condition.
10. The ultra-compact non-volatile photonic neurosynaptic device of claim 1, wherein the ultra-compact non-volatile photonic neurosynaptic device is fabricated on a silicon-on-insulator (SOI) platform or other semiconductor platform.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210883229.9A CN115332440A (en) | 2022-07-26 | 2022-07-26 | Ultra-compact non-volatile photonic neurosynaptic device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210883229.9A CN115332440A (en) | 2022-07-26 | 2022-07-26 | Ultra-compact non-volatile photonic neurosynaptic device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115332440A true CN115332440A (en) | 2022-11-11 |
Family
ID=83920371
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210883229.9A Pending CN115332440A (en) | 2022-07-26 | 2022-07-26 | Ultra-compact non-volatile photonic neurosynaptic device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115332440A (en) |
-
2022
- 2022-07-26 CN CN202210883229.9A patent/CN115332440A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Cheben et al. | Subwavelength integrated photonics | |
| CN115327703B (en) | Nonvolatile multistage adjustable photonic synapse device based on phase change material | |
| Li et al. | Free-space coupling of a light beam into a symmetrical metal-cladding optical waveguide | |
| Lee et al. | Highly efficient plasmonic interconnector based on the asymmetric junction between metal-dielectric-metal and dielectric slab waveguides | |
| CN102124395B (en) | Surface-plasmon-based optical modulator | |
| CN102231034A (en) | Light beam adjuster using adjustable photonic crystal auto-collimation effect and application thereof | |
| Quan et al. | Nonvolatile multi-level adjustable optical switch based on the phase change material | |
| CN115332440A (en) | Ultra-compact non-volatile photonic neurosynaptic device | |
| CN111929930A (en) | Programmable waveguide based on adjustable metamaterial | |
| Han | 1D photonic crystals: principles and applications in silicon photonics | |
| CN116027481A (en) | Ultra-low power consumption photon chip and neural network architecture | |
| Yan et al. | High optical contrast and multi-level storage of the ultracompact plasmonic device based on phase change materials | |
| CN113673677B (en) | Method, equipment and medium for realizing nonlinear activation function RELU | |
| Li et al. | High efficiency and compact vertical interlayer coupler for silicon nitride-on-silicon photonic platform | |
| CN211669401U (en) | Active polarization rotator realized based on mixed surface plasma groove waveguide | |
| Al-Tameemi et al. | All-optical universal logic gates at nano-scale dimensions | |
| Zhang et al. | Tunable optical power splitter based on directional coupler structure with phase change material Sb2Se3 | |
| CN115018041A (en) | Reconfigurable nonvolatile integrated three-dimensional optical diffraction neural network chip | |
| Tang et al. | Dynamic beam scanning metasurface with high reflectivity and independent phase control based on phase change materials | |
| CN111239896A (en) | Active polarization rotator realized based on mixed surface plasma groove waveguide | |
| CN108802900B (en) | Nanowire optical waveguide based on all-dielectric | |
| Sakurai et al. | Proposal of tunable hollow waveguide distributed Bragg reflectors | |
| Quan et al. | On-chip ultra-compact nonvolatile photonic synapse | |
| Okamoto et al. | Low loss plasmonic bragg gratings with a trench plasmonic waveguide | |
| CN111090147A (en) | Integrated optical waveguide integrated with periodic array of nano-structure |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |