rus
Issue #2/2025
A. S. Davydov, A. S. Chernyshov, V. M. Kirichenko
Domestic CAD System for Engineering Analysis of Photonic Integrated Circuits from T1 Integration
Domestic CAD System for Engineering Analysis of Photonic Integrated Circuits from T1 Integration
DOI: 10.22184/1993-7296.FRos.2025.19.2.88.100
The new Russian system of computer-aided design and analysis of components of photonic integrated circuits of T1 Integration company (T1 IT holding) has been presented. The article analyzes the capabilities of domestic CAD, including in comparison with its foreign equivalent, which is widely used in enterprises of the Russian Federation. Special attention is paid to the applicability of the methods proposed by Russian developers.
The new Russian system of computer-aided design and analysis of components of photonic integrated circuits of T1 Integration company (T1 IT holding) has been presented. The article analyzes the capabilities of domestic CAD, including in comparison with its foreign equivalent, which is widely used in enterprises of the Russian Federation. Special attention is paid to the applicability of the methods proposed by Russian developers.
Domestic CAD System for Engineering Analysis of Photonic Integrated Circuits from T1 Integration
A. S. Davydov, A. S. Chernyshov, V. M. Kirichenko
TS Integration LLC, Moscow, Russia
The new Russian system of computer-aided design and analysis of components of photonic integrated circuits of T1 Integration company (T1 IT holding) has been presented. The article analyzes the capabilities of domestic CAD, including in comparison with its foreign equivalent, which is widely used in enterprises of the Russian Federation. Special attention is paid to the applicability of the methods proposed by Russian developers.
Key words: BPM-method, FDTD-method,
FDFD-method, EME-method, CAD system for photonic integrated circuits
The article received on: February 20, 20205
The article accepted on: March 10, 2025
Introduction
Over the past 50 years, fiber-optic communication has become the dominant technology for long-range data transmission due to low signal intensity losses. Thanks to this, about 10 years ago, free wireless Internet became available in the Moscow Metro when using fiber. Today, optical fiber is already used to conduct the Internet in apartment buildings and over short distances inside data centers. At even shorter distances, photonics phenomena can be used to provide communication between individual IP blocks of the chip. For example, based on photonic networks on a chip, there are already attempts to optimize GPUs [1].
In order to provide domestic companies with the opportunity to develop photonic integrated circuits and optical coprocessors in the face of sanctions pressure from foreign countries, the T1 Integration team, as part of its work on engineering analysis CAD system for electronics, is developing a computer-aided design and analysis system for passive photonic components, which will then be expanded to active components, such as modulators and lasers, and will also provide connectivity between individual components for designing complete photonic integrated circuits. A photonic integrated circuit is a device with lasers, detectors, waveguides, modulators, and other passive and active components that work with each other directly on a single chip.
The T1 Integration team, which has more than twenty years of experience in creating software for the engineering analysis of electronic cooling, provides an opportunity for design engineers to take advantage of waveguide photonics, with which they can produce devices characterized by higher data transfer rates and virtually no losses. Due to the fact that photons are bosons and not fermions like electrons, they can be in the same place at the same time. This makes it possible to use methods such as wavelength-division multiplexing [2], which was first mentioned more than 30 years ago. Multiplexing is the combination of multiple signals at different wavelengths in a single fiber to increase bandwidth. And amplitude, phase, or polarization can be used to encode information.
TauLIGHT product from Т1
Currently, the product provides the ability to perform engineering analysis of passive components, such as rectangular waveguides, Y-dividers, multimode interferometers, ring resonators, and other components. Components can be imported in GDSII format or in widely used CAD formats. Fig. 1 shows the product interface from T1 Integration.
After performing calculations, the product makes it possible to obtain distributions of electromagnetic fields in arbitrary geometry sections and other important data, such as the effective refractive index, reflection factor, transmission factor, scattering matrices and other parameters.
Overview of implemented methods
Algorithms for Modeling Passive Photonic Elements
Currently, the TauLIGHT product implements algorithms for modeling passive photonic elements in which light propagation occurs in an environment with a constant or space-dependent refractive index. Active elements in which light interacts with electrons play an important role in modern photonics (LEDs, laser diodes, etc.), and are the subjects of our further developments.
A wide variety of different algorithms have been developed for modeling passive photonic elements, but only a few of them have shown good results in engineering calculations. The product currently implements four main known methods:
Beam Propagation Method (BPM);
Finite-Difference Frequency Domain method (FDFD) for modal calculation;
Finite-Difference Time Domain method (FDTD);
Eigenmode Expansion Method (EME).
This set of methods allows for a full-fledged engineering analysis of passive components, starting with the search for supported modes and ending with the export of scattering matrices. Next, the results obtained with the help of TauLIGHT will be presented in comparison with the foreign equivalent widespread in the Russian Federation and with the published data.
Beam Propagation Method (BPM)
The beam propagation method is the first widely used algorithm in production. There are two main implementations of the algorithm: FFT-BPM (based on fast Fourier transform) and FD-BPM (based on finite difference method). BPM in its initial implementation uses the paraxial approximation and assumes that light propagation occurs predominantly in one direction. All the main algorithm variants are implemented in TauLIGHT: scalar, semi-vector, vector and wide-angle approximations. Possible boundary conditions include Dirichlet, Neumann, periodic, and transparent conditions. Added the ability to calculate the fundamental mode using BPM in the planar case.
Many examples show that the solver works correctly. An example of modeling a two-dimensional Y-divider using BPM with parameters taken from the article [3] is shown in Fig. 2. A beam of light with a profile corresponding to an even TE0 mode at the beginning of the Y-waveguide is launched into the Y-divider. The dimension of the space has been reduced to 2D, because the waveguide can be considered flat. The computational domain is constructed in such a way as to fully describe the Y-waveguide and take into account the evanescent field. In order to avoid reflections, transparent boundary conditions were applied at the boundaries of the computational domain. The geometry was divided by a grid with a step of 0.5 λ/2π in the X direction and 20 λ/2π in the Z direction. The intensity of the electric field in each of the arms of the Y-waveguide is approximately equal to 50% of the incoming intensity. Leaks are noticeable in the shoulder straightening area, which is caused by radiation modes. The implemented method calculates the intensity distribution of the electric field in the Y-waveguide well, since the results are consistent with what is described in the literature. Figure 3 shows the results of the three-dimensional modeling of the Y-divider presented in the article [4]. The error in the calculations turned out to be insignificant. BPM can be used not only to find the distribution of the electromagnetic field inside the structure, but also to find the fundamental and first antisymmetric mode. The results for a multimode planar waveguide are shown in Fig. 4. The validity of the wide-angle approximation for BPM, according to [5], is shown in Fig.5.
To determine the type and number of modes that can propagate in the cross-section of the waveguide, TauLIGHT implements the finite-difference frequency domain method (FDFD). As an example of using FDFD, Figure 6 shows the profiles of electromagnetic modes that can propagate in the cross-section of an optical fiber. Figure 6 shows modes significantly higher than the fundamental one. The calculated geometry consists of a fiber-optic core with a diameter of 6 microns, with a refractive index of 1.45; a wavelength of 1.5 microns. The core is surrounded by air with a refractive index of 1. All modes supported by this fiber-optic core were found, as well as their effective coefficients of electric field refraction and intensity distribution. Our implementation of the finite-difference frequency domain method has proven its operability, which makes it possible to use the developed product as a research tool in problems with arbitrary geometry.
Finite-Difference Time Domain Method (FDTD)
The Finite Difference Time domain (FDTD) method is perhaps the most widely used algorithm for solving Maxwell’s equations. FDTD makes it possible to solve any problems of electrodynamics where there is the necessary computing power. TauLIGHT implements the FDTD method with perfectly matched layers (PML), Dirichlet and Neumann conditions. The ability to specify various types of sources, including modal sources, is enabled.
The results of 3D FDTD calculations for the geometric structure in Fig.3 are presented in Fig. 7 for various time points in the longitudinal section of the waveguide.
Eigenmode Expansion Method
To speed up the calculation of relatively long passive components and with a wide range of light propagation angles, the EigenMode Expansion method has been implemented. It is designed to combine high speed and precision calculations.
The peculiarity of this method is that it requires defining planes that divide the calculated structure into sections with a constant cross-section. These planes serve as places for calculating the full mode basis and makes it possible to calculate scattering matrices for coupling and for propagation. By obtaining the scattering matrices for each of the sections, it is possible to obtain a combined matrix for the entire structure.
Currently, the solver is provided for structures whose cross-section does not change constantly. Additional developments are underway for structures whose cross-section is constantly changing, such as a tapered waveguide.
Methods Under Development
For active devices, the Discontinuous Galerkin Time Domain method (DGTD) is being developed, which requires a transition to finite element grids and will allow multidisciplinary calculations.
Comparison of methods
The FDTD method, due to the requirement of a sufficiently small grid element size, is not very effective for large structures compared to the wavelength of the source. If the first order of approximation is used in the calculation, then the correct solution by the FDTD method often requires at least 18–20 nodes per wavelength. For applications with geometries where one of the dimensions is much larger than the others (for example, the length relative to the thickness and width of the device under study), a huge number of grid nodes may be required.
This problem has several solutions, the first of which is the transition to 2.5D calculation, which will require additional refinement of the FDTD algorithm and the use of so-called efficient material. The second solution involves switching to the use of distributed memory in calculations, which requires impressive computing resources. The third one involves the use of subgrid fragmentation, which is technically quite difficult to implement due to the resulting artificial reflections at the grid boundaries. Nevertheless, this method works well with high-contrast structures (large difference in refractive indices between individual parts).
The BPM method is effective for structures with large longitudinal dimensions and does not work well with high-contrast elements. A large difference in refractive indices leads to a discrepancy between experimental and calculated data, as conditions arise for a wide range of light propagation angles.
The FD-BPM method copes well with large longitudinal sections, however, it is bad for waves propagating at large angles, which is why the computational engineer is forced to use a wide-angle approximation. On the contrary, the FDTD method is effective for any possible propagation angles, but is poorly suited for structures with a large longitudinal dimension.
The implemented EigenMode Expansion method complements the existing set of solvers. The method is intended to be optimal for structures with high contrast and large longitudinal size, which would require an increased amount of computing resources when using other methods. The price for speed and accuracy is the requirement for a more thoughtful approach when setting the task.
Comparing Results
with Competitive Products
During product development, solver validation uses not only analytical solutions and information from published articles, but also calculation results from foreign products from world leaders in software development for the engineering analysis of photonic integrated circuit components. This makes it possible to compare solvers on much more complex geometry, relative to academic papers.
Comparing Results Obtained Using Fdfd
Figure 8 shows the distribution of the actual part of the electric field strength in the cross-section of a waveguide with a refractive index of 3.48. The refractive index of the shell is 1.44. The wavelength is 1.53 microns. A qualitative comparison of the results indicates a good match. The effective refractive indices are 2.380 for the TauLIGHT and 2.378 for the foreign equivalent, which is widely used in enterprises of the Russian Federation. The computational grid is the same in all cases and consists of approximately 122 thousand nodes. The boundary conditions are the same in all products, Perfectly Matched Layers were used. There are good matches between the results obtained for different products.
Comparing Results Obtained Using BPM
Figure 9 shows the distribution of the actual part of the electric field strength in the longitudinal section of a multimode interferometer of the T1 Integration product and its foreign equivalent. The refractive indices are similar to those in the previous section. The setting is three-dimensional and without wide-angle zoom. The calculated grid is the same in both products and was obtained after a grid convergence study, the number of calculated nodes was about 154 million. Transparent boundary conditions were used to prevent back reflection into the calculated area.
Scattering matrices were exported using both products. The results for the real part of the Samp scattering parameters are shown in the table. Samp for Sij shows how much of the power passed through port i, launched from port j. Since the calculation was performed using BPM, which performs calculations only in one direction, it was possible to obtain scattering parameters only for output ports 3 (S31) and 4 (S41), which corresponds to the left and right terminals in the upper part of Fig. 9.
A comparison of the results for the multimode interferometer allowed us to conclude that the results from the TauLIGHT product and its foreign equivalent are very close, and have a maximum difference of 7% for Samp for output port number 3.
A. S. Davydov, A. S. Chernyshov, V. M. Kirichenko
TS Integration LLC, Moscow, Russia
The new Russian system of computer-aided design and analysis of components of photonic integrated circuits of T1 Integration company (T1 IT holding) has been presented. The article analyzes the capabilities of domestic CAD, including in comparison with its foreign equivalent, which is widely used in enterprises of the Russian Federation. Special attention is paid to the applicability of the methods proposed by Russian developers.
Key words: BPM-method, FDTD-method,
FDFD-method, EME-method, CAD system for photonic integrated circuits
The article received on: February 20, 20205
The article accepted on: March 10, 2025
Introduction
Over the past 50 years, fiber-optic communication has become the dominant technology for long-range data transmission due to low signal intensity losses. Thanks to this, about 10 years ago, free wireless Internet became available in the Moscow Metro when using fiber. Today, optical fiber is already used to conduct the Internet in apartment buildings and over short distances inside data centers. At even shorter distances, photonics phenomena can be used to provide communication between individual IP blocks of the chip. For example, based on photonic networks on a chip, there are already attempts to optimize GPUs [1].
In order to provide domestic companies with the opportunity to develop photonic integrated circuits and optical coprocessors in the face of sanctions pressure from foreign countries, the T1 Integration team, as part of its work on engineering analysis CAD system for electronics, is developing a computer-aided design and analysis system for passive photonic components, which will then be expanded to active components, such as modulators and lasers, and will also provide connectivity between individual components for designing complete photonic integrated circuits. A photonic integrated circuit is a device with lasers, detectors, waveguides, modulators, and other passive and active components that work with each other directly on a single chip.
The T1 Integration team, which has more than twenty years of experience in creating software for the engineering analysis of electronic cooling, provides an opportunity for design engineers to take advantage of waveguide photonics, with which they can produce devices characterized by higher data transfer rates and virtually no losses. Due to the fact that photons are bosons and not fermions like electrons, they can be in the same place at the same time. This makes it possible to use methods such as wavelength-division multiplexing [2], which was first mentioned more than 30 years ago. Multiplexing is the combination of multiple signals at different wavelengths in a single fiber to increase bandwidth. And amplitude, phase, or polarization can be used to encode information.
TauLIGHT product from Т1
Currently, the product provides the ability to perform engineering analysis of passive components, such as rectangular waveguides, Y-dividers, multimode interferometers, ring resonators, and other components. Components can be imported in GDSII format or in widely used CAD formats. Fig. 1 shows the product interface from T1 Integration.
After performing calculations, the product makes it possible to obtain distributions of electromagnetic fields in arbitrary geometry sections and other important data, such as the effective refractive index, reflection factor, transmission factor, scattering matrices and other parameters.
Overview of implemented methods
Algorithms for Modeling Passive Photonic Elements
Currently, the TauLIGHT product implements algorithms for modeling passive photonic elements in which light propagation occurs in an environment with a constant or space-dependent refractive index. Active elements in which light interacts with electrons play an important role in modern photonics (LEDs, laser diodes, etc.), and are the subjects of our further developments.
A wide variety of different algorithms have been developed for modeling passive photonic elements, but only a few of them have shown good results in engineering calculations. The product currently implements four main known methods:
Beam Propagation Method (BPM);
Finite-Difference Frequency Domain method (FDFD) for modal calculation;
Finite-Difference Time Domain method (FDTD);
Eigenmode Expansion Method (EME).
This set of methods allows for a full-fledged engineering analysis of passive components, starting with the search for supported modes and ending with the export of scattering matrices. Next, the results obtained with the help of TauLIGHT will be presented in comparison with the foreign equivalent widespread in the Russian Federation and with the published data.
Beam Propagation Method (BPM)
The beam propagation method is the first widely used algorithm in production. There are two main implementations of the algorithm: FFT-BPM (based on fast Fourier transform) and FD-BPM (based on finite difference method). BPM in its initial implementation uses the paraxial approximation and assumes that light propagation occurs predominantly in one direction. All the main algorithm variants are implemented in TauLIGHT: scalar, semi-vector, vector and wide-angle approximations. Possible boundary conditions include Dirichlet, Neumann, periodic, and transparent conditions. Added the ability to calculate the fundamental mode using BPM in the planar case.
Many examples show that the solver works correctly. An example of modeling a two-dimensional Y-divider using BPM with parameters taken from the article [3] is shown in Fig. 2. A beam of light with a profile corresponding to an even TE0 mode at the beginning of the Y-waveguide is launched into the Y-divider. The dimension of the space has been reduced to 2D, because the waveguide can be considered flat. The computational domain is constructed in such a way as to fully describe the Y-waveguide and take into account the evanescent field. In order to avoid reflections, transparent boundary conditions were applied at the boundaries of the computational domain. The geometry was divided by a grid with a step of 0.5 λ/2π in the X direction and 20 λ/2π in the Z direction. The intensity of the electric field in each of the arms of the Y-waveguide is approximately equal to 50% of the incoming intensity. Leaks are noticeable in the shoulder straightening area, which is caused by radiation modes. The implemented method calculates the intensity distribution of the electric field in the Y-waveguide well, since the results are consistent with what is described in the literature. Figure 3 shows the results of the three-dimensional modeling of the Y-divider presented in the article [4]. The error in the calculations turned out to be insignificant. BPM can be used not only to find the distribution of the electromagnetic field inside the structure, but also to find the fundamental and first antisymmetric mode. The results for a multimode planar waveguide are shown in Fig. 4. The validity of the wide-angle approximation for BPM, according to [5], is shown in Fig.5.
To determine the type and number of modes that can propagate in the cross-section of the waveguide, TauLIGHT implements the finite-difference frequency domain method (FDFD). As an example of using FDFD, Figure 6 shows the profiles of electromagnetic modes that can propagate in the cross-section of an optical fiber. Figure 6 shows modes significantly higher than the fundamental one. The calculated geometry consists of a fiber-optic core with a diameter of 6 microns, with a refractive index of 1.45; a wavelength of 1.5 microns. The core is surrounded by air with a refractive index of 1. All modes supported by this fiber-optic core were found, as well as their effective coefficients of electric field refraction and intensity distribution. Our implementation of the finite-difference frequency domain method has proven its operability, which makes it possible to use the developed product as a research tool in problems with arbitrary geometry.
Finite-Difference Time Domain Method (FDTD)
The Finite Difference Time domain (FDTD) method is perhaps the most widely used algorithm for solving Maxwell’s equations. FDTD makes it possible to solve any problems of electrodynamics where there is the necessary computing power. TauLIGHT implements the FDTD method with perfectly matched layers (PML), Dirichlet and Neumann conditions. The ability to specify various types of sources, including modal sources, is enabled.
The results of 3D FDTD calculations for the geometric structure in Fig.3 are presented in Fig. 7 for various time points in the longitudinal section of the waveguide.
Eigenmode Expansion Method
To speed up the calculation of relatively long passive components and with a wide range of light propagation angles, the EigenMode Expansion method has been implemented. It is designed to combine high speed and precision calculations.
The peculiarity of this method is that it requires defining planes that divide the calculated structure into sections with a constant cross-section. These planes serve as places for calculating the full mode basis and makes it possible to calculate scattering matrices for coupling and for propagation. By obtaining the scattering matrices for each of the sections, it is possible to obtain a combined matrix for the entire structure.
Currently, the solver is provided for structures whose cross-section does not change constantly. Additional developments are underway for structures whose cross-section is constantly changing, such as a tapered waveguide.
Methods Under Development
For active devices, the Discontinuous Galerkin Time Domain method (DGTD) is being developed, which requires a transition to finite element grids and will allow multidisciplinary calculations.
Comparison of methods
The FDTD method, due to the requirement of a sufficiently small grid element size, is not very effective for large structures compared to the wavelength of the source. If the first order of approximation is used in the calculation, then the correct solution by the FDTD method often requires at least 18–20 nodes per wavelength. For applications with geometries where one of the dimensions is much larger than the others (for example, the length relative to the thickness and width of the device under study), a huge number of grid nodes may be required.
This problem has several solutions, the first of which is the transition to 2.5D calculation, which will require additional refinement of the FDTD algorithm and the use of so-called efficient material. The second solution involves switching to the use of distributed memory in calculations, which requires impressive computing resources. The third one involves the use of subgrid fragmentation, which is technically quite difficult to implement due to the resulting artificial reflections at the grid boundaries. Nevertheless, this method works well with high-contrast structures (large difference in refractive indices between individual parts).
The BPM method is effective for structures with large longitudinal dimensions and does not work well with high-contrast elements. A large difference in refractive indices leads to a discrepancy between experimental and calculated data, as conditions arise for a wide range of light propagation angles.
The FD-BPM method copes well with large longitudinal sections, however, it is bad for waves propagating at large angles, which is why the computational engineer is forced to use a wide-angle approximation. On the contrary, the FDTD method is effective for any possible propagation angles, but is poorly suited for structures with a large longitudinal dimension.
The implemented EigenMode Expansion method complements the existing set of solvers. The method is intended to be optimal for structures with high contrast and large longitudinal size, which would require an increased amount of computing resources when using other methods. The price for speed and accuracy is the requirement for a more thoughtful approach when setting the task.
Comparing Results
with Competitive Products
During product development, solver validation uses not only analytical solutions and information from published articles, but also calculation results from foreign products from world leaders in software development for the engineering analysis of photonic integrated circuit components. This makes it possible to compare solvers on much more complex geometry, relative to academic papers.
Comparing Results Obtained Using Fdfd
Figure 8 shows the distribution of the actual part of the electric field strength in the cross-section of a waveguide with a refractive index of 3.48. The refractive index of the shell is 1.44. The wavelength is 1.53 microns. A qualitative comparison of the results indicates a good match. The effective refractive indices are 2.380 for the TauLIGHT and 2.378 for the foreign equivalent, which is widely used in enterprises of the Russian Federation. The computational grid is the same in all cases and consists of approximately 122 thousand nodes. The boundary conditions are the same in all products, Perfectly Matched Layers were used. There are good matches between the results obtained for different products.
Comparing Results Obtained Using BPM
Figure 9 shows the distribution of the actual part of the electric field strength in the longitudinal section of a multimode interferometer of the T1 Integration product and its foreign equivalent. The refractive indices are similar to those in the previous section. The setting is three-dimensional and without wide-angle zoom. The calculated grid is the same in both products and was obtained after a grid convergence study, the number of calculated nodes was about 154 million. Transparent boundary conditions were used to prevent back reflection into the calculated area.
Scattering matrices were exported using both products. The results for the real part of the Samp scattering parameters are shown in the table. Samp for Sij shows how much of the power passed through port i, launched from port j. Since the calculation was performed using BPM, which performs calculations only in one direction, it was possible to obtain scattering parameters only for output ports 3 (S31) and 4 (S41), which corresponds to the left and right terminals in the upper part of Fig. 9.
A comparison of the results for the multimode interferometer allowed us to conclude that the results from the TauLIGHT product and its foreign equivalent are very close, and have a maximum difference of 7% for Samp for output port number 3.
Readers feedback
