The Key Role of Quantum Computing! The Challenges Facing Silicon Photonics in Advanced Computing

Silicon Photonics is an innovative field that applies optical communications technology to integrated semiconductor circuits. The operating principle of the silicon photonics chip is basically to replace traditional “electrical signals” with “optical signals”, resulting in excellent features such as a large bandwidth, high computing density, low power consumption and anti-interference capabilities. These advantages are the reason that this technology is regarded as the best solution for achieving high efficiency computing. In recent years, the rapid growth of cloud computing and artificial intelligence (AI) application markets has served to promote silicon photonics technology, making it the focus of global industry and academic circles, which, in turn, has led to many research breakthroughs.

Intel was the first manufacturer to realize the commercial application of silicon photonics. They established a dedicated business department to invest in its strategic layout as early as 2012 and began making mass shipments in 2016. In June of 2022, Intel also successfully demonstrated the results of their advanced research in the integration and control of an eight wavelength laser array on a silicon chip. IBM, on the other hand, developed a 90nm nano silicon photonic integrated circuit chip at the end of 2012, where they integrated optical and electrical circuits on a single chip. In 2017, the well-known silicon photonics company Luxtera announced that it would be developing the next generation of silicon photonics technology jointly with TSMC, a leading foundry. In addition, according to foreign media reports, the Dutch government will be investing 1.1 billion Euros to promote the development of new generation silicon photonics technology companies in preparation to build the next ASML. Yole Developpement estimates that the silicon photonics optical module market will grow from about 455 million USD in 2018 to about 4 billion USD in 2024 with a CAGR of up to 44.5%.

According to Open AI’s statistics, starting from 2012, the computing power required by AI applications will double every three to four months. However, the current development of semiconductor chips is approaching the limits of Moore’s Law. They cannot meet the requirements of future high performance computing. However, the high costs and large sizes of traditional optical modules make them unsuitable for large scale applications. Under the premise of lower costs, Silicon photonic chips can effectively improve the computer clusters in data centers as well as the communication speed and efficiency between servers and chips, thus providing the enormous computing power needed. At SEMICON TAIWAN in November of 2022, the CEO of ASE, Tianyu Wu, predicted 4 major technology trends in the future, one of which was “Silicon Photonics”. According to industry analysis and predictions, data centers will be the main competitive market for the first wave of silicon photonics applications and development. It is expected that, over the next 3 years, silicon photonics chips will become widely used in high speed information transmission in large data centers. To gain technological leadership in the silicon photonics market, renowned manufacturers such as Intel, IBM, Google, Facebook, Cisco, Marvell, Huawei, Mellanox, Luxtera, Acacia, Finisar and Avago, etc. have increased their R&D investments and production capacity in recent years as they strive to become cornerstones of the future silicon photonics era.

As the development of silicon photonics technology advances in multiple directions, its fields of application also continue to expand. In addition to its use in data centers, silicon photonics chips can also be applied to optical radars (LiDAR), fiber optic gyroscopes, biomedical sensors, AI systems and other products and devices that require complex optical circuits. The increasingly popular quantum computing is also expected to be an important area of development for silicon photonics in the future. It is anticipated that the application of silicon photonics technology in quantum optics will promote the practical progress of quantum computing hardware and solve the technical bottlenecks of “manipulation” and “reading” faced by Ion Trap qubits operating at room temperature. The use of silicon photonic chips that can be integrated with ion trap qubit chips and their surrounding CMOS circuits could not only eliminate the jitter, drift and electrical noise issues of optical components in free space but also ameliorate problems such as those caused by overly long optical fibers, the mutual alignment of complex optical components and more. Its great potential for future developments such as expanding the number of qubits and improving the reading fidelity of ion trap qubits has made this a highly anticipated technology.

MA-tek R&D Center Director Chris Chen 2023/8/30

The Key Role of Quantum Computing! The Challenges Facing Silicon Photonics in Advanced Computing

Institute of Electronics, National Yangming Chiao Tung University

Professor Pei-Wen Li

  (This article was provided by Professor Pei-Wen Li and edited by MA-tek)

-

Silicon photonics technology combines two of the most important inventions of 20th century semiconductor technology: CMOS integrated circuits and semiconductor lasers. These can be manufactured using mature silicon mass production technologies and applied to a diverse range of application fields, including 5G wireless communications, automobiles, medical care, and even multifunctional sensors for the Internet of Things, such as LiDARs and gyroscopes, etc. [1]. Another reason that integrated silicon photonics technology has attracted so much attention is that it can use optical connections to make up for the serious delays and energy consumption problems caused by having highly miniaturized metal electrical connections within a chip. For a long time now, the characteristics (length/width) of transistors, or metal electrical connections, within a microchip and the measures for increasing integration density, such as Dennard Scaling, have been facing severe, physical bottlenecks. Even though parallel computing by multi-core processors can be used to improve the performance of computer systems, it will eventually become difficult to break through the Energy Efficiency Wall [2].

Ideally, silicon photonics integrated optical circuits would homogeneously or heterogeneously integrate light sources, functional sensing components, optical/electrical signal conversion components, and optical waveguides, etc. all on one silicon platform to realize multiple modules such as optical interconnect chips, processors and sensors. In particular, integrating silicon photonic chips and silicon CMOS integrated circuits into optoelectric hybrid integrated chips is expected to significantly increase data processing speeds, reduce power consumption, decrease chip areas, save data bit costs and improve reliability [3].

Because silicon photonics integrated circuit technology has already revealed several excellent business opportunities in data centers, 5G communications and other fields. It has attracted the interest of many European and American companies and research institutions, including IBM, Intel, Cisco, STM and CEA / Leti, IMEC, AIM Photonics, and IME, etc., and many have already begun carrying out the research and development of and laying out their technical strategies for silicon photonics chips, seeking to continue to improve the speed and bandwidth of Intra-Chip data processing. Google, Apple, Meta, Amazon, Microsoft and other companies are also actively developing high efficiency silicon photonic connection technology, building optical interconnections between short-distance data center interiors (850nm lasers and multi-mode fibers) and long-distance data centers (1310nm lasers and single-mode fibers).

Silicon photonics technology can not only improve the speed of traditional computing and enhance the efficiency of data center transmissions but is also expected to promote the development of quantum computing hardware technology. Take, for instance, Ion Trap qubits that can operate at room temperature. Generally, light or microwave signals on the optical table are used to remotely control the ion trap qubits located in the vacuum chamber [4]. When reading the quantum state information of the ion trap qubit, light is focused to the single photon detector through a high numerical aperture lens on the optical table. It has been verified that ion trap qubits can function in a normal temperature, vacuum environment. This is a promising sign for practical quantum computing. However, ion trap quantum technology still faces many substantial technical difficulties and challenges in terms of expanding the number of qubits, and improving the fidelity of actual initialization, control and detection of quantum states [4].

The optical components/optoelectric components required to “control” and “read” ion trap qubits and the technical bottlenecks they face are briefly described below:

1. It is necessary to use radio frequencies and DC high voltages to control the trapping potential for ion movement. Therefore, the power injected will inevitably dissipate in the trapping electrode, heating the ion trap qubit and destroying the fidelity of the quantum state.

2. At present, the operating situation for ion trap qubits is that the optical components and electronic circuits for control and reading are placed on an optical table at room temperature while the ion trap qubit chip is placed in a low temperature vacuum chamber, and the two are connected via complex optical fibers and cables. Due to mechanical vibration/thermal noise and surrounding environmental disturbances, the accuracy of optical component alignments is often disturbed by low frequency jitters and temperature drifts, negatively impacting the long-term stability and fidelity of qubit chip operations.

3. A large, high numerical aperture lens is needed to focus the laser light on a single ion at a certain point or to collect the small number of photons from the spontaneous scattering of a specific ion into a single photon detector. However, the lenses, lasers and light detectors located on the optical table can be disturbed by alignment and coupling noise/time drifts, etc.. As such, the control/reading fidelity still needs improvement.

In summary, most optical components currently available, such as light sources, optical fibers and single photon detectors, are placed on free space optical tables. However, the alignment between complex optical components and optical fibers and environmental disturbances such as mechanical vibration/thermal noise generate many additional noise sources, greatly limiting the number and fidelity of ion trap qubits. Therefore, there is an urgent need to develop silicon photonic chips (such as optical waveguides, optical modulators, and optical detectors) and peripheral CMOS circuits that can be integrated with ion trap qubit chips [3].

Ideally, the laser light source would be connected directly to the ion trap qubit in a vacuum environment through an optical waveguide so that the ion trap can be directly controlled and cooled. At the same time, the optical waveguide would be used to connect a single photon detector that reads the quantum state information of the nearby ion trap qubit and then output that information directly to the CMOS integrated circuit for subsequent signal processing. In this way, not only the jitter or drift of optical components in free space but also the ubiquitous electrical noise can be eliminated. It would also eliminate the alignment problems brought about by long optical fibers and complex optical components. It would be a great help to “expanding” the number of qubits and “improving” the reading fidelity of ion trap qubits.

The research and development of silicon photonics component technology has been going on for more than thirty years. Initially, this research was aimed mainly at applications such as communications and classical computing. The silicon photonic connection technology developed (components such as optical waveguides, optical modulators, optical detectors and even light sources) focused on processing high speed, high bandwidth, high responsivity or high wattage optical/electrical signals for communications applications. Silicon photonics components that control or read ion trap qubits, however, must be able to handle low noise, low dark current and very low wattage (a few photons) near ultra violet or visible light signals. They must even be able to operate stably in low temperature environments. The following examines the application of ion trap qubits and how it relates to the key technical requirements and challenges facing silicon photonic components.

The wavelength range of laser light sources commonly used to control ion trap qubits is approximately 300-2000 nm. This covers the near ultra violet, visible, and near infrared spectrums. Unfortunately, the silicon optical waveguides currently commonly used in silicon photonics technology have high absorbance in the ultra violet and visible light bands and large optical losses, making them unsuitable for ion trap qubit technology. In contrast, the silicon nitride (Si3N4) optical waveguide is transparent in the ultra violet-visible wavelength range. There is no light absorption and no optical loss [5,6]. Fortunately, silicon nitride is a common material in the insulating, partition, and protective layers in CMOS process technology and can be deposited using conventional chemical vapor deposition (CVD) processes. The chemical vapor deposition process recipe can also be fine-tuned according to actual application needs, adjusting the chemical composition of silicon nitride (such as SixNy or even SiOxNy film) and refractive index. This helps modulate the mode number, light localization and transmission loss of silicon nitride optical waveguides, etc.. In recent years, because silicon nitride optical waveguides can make up for the visible light (400-1000nm) band that cannot be covered by silicon optical waveguides, they have come to be seen as the best platform for various experimental chips, especially in quantum communications/computing [7].

In 2020, ETH Zurich [8] reported on the use of ion trap quantum logic gates to integrate ion trap chips and silicon nitride optical waveguides. A single-mode fiber was used to inject 729 nm visible light into a silicon nitride optical waveguide, which passed it to the ion trap chip located in a low temperature, vacuum environment. This method can eliminate the problems of optical alignment, mechanical vibration and beam point drift on the optical table and improve the fidelity of quantum logic gates. However, the ion trap chip demonstrated has not yet been integrated with silicon photonic active components such as optical modulators and single photon detectors because the crystallization nucleation dormancy time of germanium and silicon germanium on silicon nitride film is very short. It is difficult to find an epitaxial growth method that can grow high quality, single crystal germanium or silicon germanium films on silicon nitride films, which makes it difficult to continue to produce silicon photonic active components. Although (1) Wafer Bonding can be used after bonding SOI on a silicon nitride platform to epitaxially grow a photoactive layer—germanium or silicon germanium film [13]. Another method is to imitate (2) STM, IHP and the University of Toronto by producing silicon germanium modulators and germanium photo detectors on the SOI platform. The silicon nitride film is then deposited via PECVD and smoothing via CMP to fabricate top layer silicon nitride optical waveguides [9-11]. However, the latter method proposed for the top layer silicon nitride optical waveguide makes it difficult to proceed with the following high temperature annealing process of silicon nitride dehydrogenation or densification, and the number of defects inside the silicon nitride optical waveguide cannot be reduced. The difficulty stems from the lattice relaxation of silicon germanium in the photoactive area at the bottom of the germanium epitaxial films, causing performance degradation of optical active components. Regarding silicon nitride optical platforms, there has been very little literature published so far on single-unit integrated germanium/silicon germanium high speed optical modulators [9], high speed optical detectors [10,11] or laser light sources [12]. Therefore, the manufacturing of optical active components and the integration of optical active/passive components on silicon nitride platforms are important research topics.

Quantum state qubits are weak and susceptible to interference by noise from the surrounding environment. Therefore, it is necessary to have a “built in” single photon detector near the ion trap qubit to quickly and accurately read and detect changes in the small number of photons in said qubit. It is best to connect the single photon detector directly to the ion trap quantum chip with a silicon nitride optical waveguide to minimize crosstalk in the collection and detection of photons as well as to further expand and measure the feasibility of large ion trap qubit arrays. Typically, ion trap qubits emit photons with wavelengths of about 300–500 nm. Currently, however, the most mature silicon avalanche photo detector detects light at 850 nm, so it is unable to directly detect ion trap qubit states.

NIST researchers in America used a homemade, built-in “superconducting” single photon detector that did not require either imaging lenses or cameras and were able to read the quantum state of Beryllium Ions with almost perfect accuracy (reading accuracy exceeding 99.9%) [13]. However, “superconducting” single photon detectors can only operate normally in an environment that is close to absolute zero. According to the NIST report, in order to effectively improve detection efficiency and reduce the dark count rate, there is an urgent need for near-UV silicon-based single photon detectors compatible with CMOS technology. Furthermore, there is a need to integrate silicon nitride optical waveguides/gratings with silicon-based single photon detectors in a single unit to further reduce coupling losses and noise and expand the number of ion trap qubits.

In addition to the single photon detector that can be integrated with the ion trap quantum chip, the visible light source coupled with the silicon nitride optical waveguide is a key component for controlling the ion trap quantum chip. However, realizing a light source on a silicon substrate has always been the biggest obstacle facing silicon photonics technology. Realizing a visible light source that can be integrated with a silicon nitride optical waveguide is an even greater challenge because silicon itself is an indirect bandgap material with a very low luminous efficiency. For a long time, scientists and engineers have been trying to grow indium phosphorus or germanium thin films on silicon wafers using epitaxy or wafer bonding techniques. However, due to limitations such as lattice constant match and thermal budget, it is impossible to produce thermally stable luminescent materials with a high crystalline quality.

More and more reports are demonstrating that the use of germanium nanostructures such as quantum wells, quantum wires and even quantum dots can effectively alleviate the problem with defects that occurs when growing single crystal germanium films on silicon wafers. In particular, by relying on the quantum confinement effect, the strong overlapping coupling of electron-hole wave functions in small germanium quantum dots greatly enhances the optical transition oscillation intensity of those germanium quantum dots. This makes it possible to break the curse that demands that germanium bulk materials adhere strictly to energy-momentum (E-k) conservation. Moreover, the luminescence energy gap can be modulated by adjusting the diameter of the single material germanium quantum dot in order to make it emit light of different wavelengths and thereby overcoming the need to use different bulk materials to produce light sources of different wavelengths. However, a single light-emitting quantum dot is small and needs to be placed in a resonance chamber. When the laser light irradiates the quantum dot resonance chamber, the Purcell Effect causes the number of light-excited photons in the quantum dot to increase rapidly, improving the overall luminous quality factor. Common quantum dot resonance chamber structures include photonic crystals, microdisks and microrings. Placing germanium quantum dots in photonic crystals can result in extremely high luminous efficiency and quality factors [14], however, the structure of the photonic crystal chamber, such as its template thickness, hole diameter, period and defect mode or mode structural design, is complex. Furthermore, an advanced electron beam lithography system must be used to expose photonic crystal arrays with sub-micron holes (diameter or period). In addition, quantum dot photonic crystal lasers are usually vertical cavity surface emitting lasers, which is not conducive to On-Chip planer integration.

In comparison, it is relatively easy to design and fabricate micron-level microdisk or microring resonance chambers, and the In-Plane light emitted can be coupled with adjacent bus waveguides, which is beneficial to integration on chips. The microdisk resonance chamber mainly confines the light field in a disc-shaped, optically dense medium and achieves resonance along the edge of the resonance chamber in the radial direction, generating the Whispering Gallery Mode (WGM). Compared to the photonic crystal resonance chamber’s complex structural design, which requires the use of advanced electron beam lithography technology, the microdisk resonance chamber structure offers more flexibility and cost advantages for the design and production of electrodes and waveguides.

In recent years, assorted European and American research institutions have successively demonstrated Micro Lasers by embedding various quantum dots (such as silicon, germanium, and CdSe) in suspended silicon, germanium, silicon dioxide or silicon nitride microdisk resonance chambers. The French CNRS-Univ. published a series of papers [15] demonstrating a light-excited germanium micro laser. First, a 300-nm thick n+- layer of germanium was grown epitaxially on a gallium arsenide substrate. After the suspended germanium microdisks were produced via photo-lithography, the germanium microdisks were coated with silicon nitride, forming a tensile deformation n+-Ge active light-emitting layer. However, the use of germanium on gallium arsenide (Ge-on-GaAs) is very difficult to migrate onto a silicon platform. Tokyo Metropolitan University proposed that P-I-N germanium quantum dot microdisk diodes [16] can be coupled with adjacent waveguides to stimulate light electrically. However, the vast majority of quantum dot microdisks are produced on SOI platforms, so they are not suitable for visible light sources. There is an urgent need to develop quantum dot/silicon nitride microdisk visible light sources for smooth integration with ion trap quantum chips.

At the 2022 international IEDM conference, Professor Li’s research team presented a monolithic integrated silicon nitride waveguide (with grating coupler and waveguide taper) which can be used with germanium quantum dot microdisk light sources, photon detectors and other components for near UV-visible light ion trap sensing applications, as shown in Figure 1. They developed a method to fabricate germanium quantum dots compatible with CMOS processes using single-step selective oxidation. In this process, microscopic polycrystalline silicon germanium pillars located on a silicon nitride film are converted into spherical germanium quantum dots embedded with silicon nitride. The most important feature of these quantum dots is that they are prepared via thermal oxidation at 900oC. Therefore, they have the advantage of high thermal stability, as shown in Figure 2. This inherent thermal stability opens up the possibility for wave coupling to top or bottom silicon nitride waveguides for germanium quantum dot photo detectors and photo emitters.

From a component manufacturing and integration perspective, the top waveguide coupling structure provides flexibility in component (photo detectors and emitters) design and 3D integration material selection. The top waveguide coupling structure eliminates the need for the “waveguide” and “substrate” to be of the same material. This silicon nitride self-organization into germanium quantum dot array structure method provides flexibility for the integration of silicon nitride microdisk light emitters, PIN photo detectors and top and bottom silicon nitride waveguides. It also makes PIC integration feasible. Furthermore, the germanium quantum dot production technology developed by Professor Li’s research team makes direct use of CMOS process technology. As such, it has excellent process control and engineering advantages in component design and can directly produce qubits, single electron transistors, and optoelectric transistors, etc.. It has practical and industrial potential and can help in the development of quantum computing, optical connection and other technologies.

References：

[1]L. Vivien et al., “Building blocks of silicon photonics,” Semiconductors and Semimetals vol.101, 0080-8784 (2019) .
[2]B. Marr, B. Degnan, P. Hasler and D. Anderson, "Scaling Energy Per Operation via an Asynchronous Pipeline," in IEEE Transactions on Very Large Scale Integration (VLSI) Systems, vol. 21, no. 1, pp. 147-151, Jan. 2013, doi： 10.1109/TVLSI.2011.2178126.
[3]H. Hogan, “Data centers and more for Si photonics,” https：//www.photonics.com/Articles/Data_Centers_and_More_for_Silicon_Photonics/a64879
[4]K. Brown, J. Chiaverini, J. Sage, et al. Materials challenges for trapped-ion quantum computers. Nat Rev Mater 6, 892–905 (2021) . https：//doi.org/10.1038/s41578-021-00292-1
[5]D.J. Blumenthal et al., “Silicon Nitride in Silicon Photonics,” Proc. IEEE, vol.106, 12 (2018) .
[6]P. Munoz et al., “Silicon nitride photonic integration platforms for visible, near-infrared and mid-infrared applications,” Sensors, vol.17, 2088 (2017) .
[7]A. Orieux et al., “Recent advances on integrated quantum communications,” J. Opt., vol. 18, 083002 (2016) .
[8]Mehta, K.K., Zhang, C., Malinowski, M. et al. Integrated optical multi-ion quantum logic. Nature 586, 533–537 (2020) . https：//doi.org/10.1038/s41586-020-2823-6
[9]F. Boeuf et al., “A Silicon photonics technology for 400 Gbit/s applications,” IEDM Tech. Digest, pp. 775 (2019)
[10]S. Lischke et al., “Silicon nitride waveguide coupled 67GHZ Ge photodiode for non-SOI PIC and ePIC platforms,” IEDM Tech. Digest, pp. 779 (2019)
[11]W. Sacner et al., “Monolithically integrated multilayer silicon nitride-on-Si waveguide platforms,” Proc. IEEE, vol. 16, 2232 (2018)
[12]S. Bao et al., “Low-threshold optically pumped lasing in highly strained germanium nanowires,” Nat Comm. vol. 8, 1845, (2017)
[13]S. L. Todaro, V. B. Verma, K. C. McCormick, D. T. C. Allcock, R. P. Mirin, D. J. Wineland, S. W. Nam, A. C. Wilson, D. Leibfried, and D. H. Slichter, Phys. Rev. Lett. 126, 010501 (2021) . https：//doi.org/10.1103/PhysRevLett.126.010501
[14]J. S. Xia, Y. Ikegami, and Y. Shiraki, “Strong resonant luminescence from Ge quantum dots in photonic crystal microcavity at room temperature,” Appl. Phys. Lett. vol. 89, 201102 (2006) .
[15]A. Ghrib, M. El Kurdi, M. de Kersauson, M. Prost, S. Sauvage, X. Checoury, G. Beaudoin, I. Sagnes, and P. Boucaud, “Tensile-strained germanium microdisks,” Appl. Phys. vol. 102, 221112 (2013)
[16]X. Xu, T. Maruizumi, and Yasuhiro Shiraki, “Waveguide-integrated microdisk light-emitting diode and photodetector based on Ge quantum dots,” Optics Exp., vol. 22, 3905 (2014)
[17]C. H. Lin, P. Y. Hong, B. J. Lee, H. C. Lin, T. George, and P. W. Li, “Monolithic integration of top Si3N4-waveguided germanium quantum-Dots microdisk light emitters and PIN photodetectors for on-chip ultrafine sensing,” IEDM Tech. Dig. pp. 451454, Dec. 2022.