Mitsumasa Koyanagi
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Mitsumasa Koyanagi has a total of 65 co-authors. Here are the top co-authors by number of shared publications:
- Takafumi Fukushima
Tohoku University
18 shared publications - Tetsu Tanaka
Tohoku University
18 shared publications Kang-Wook Lee
New Industry Hatchery Center, Global Integration Initiative, Tohoku University, Sendai, Japan
7 shared publications- Mariappan Murugesan
Tohoku University
6 shared publications - Hisashi Kino
Frontier Research Institute for Interdisciplinary Sciences, Tohoku University, Sendai 980-8579, Japan
5 shared publications - Kang-Wook Lee
Tohoku University
4 shared publications
Latest Publications
- Open Access
- 20 May 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal on Emerging and Selected Topics in Circuits and Systems
- Vol. PP  (21563357) , 1
Die-Level Transformation of 2D Shuttle Chips into 3D-IC for Advanced Rapid Prototyping using Meta Bonding
Three-dimensional integrated circuit (3D-IC) technology, often referred to as through-silicon via (TSV) formation technology, has been steadily maturing and is increasingly used in advanced semiconductor devices such as 3D complementary metal-oxide-semiconductor image sensors, high-bandwidth memory, and static random-access memory on central processing units (commonly known as 3D V-Cache). However, the initial development costs remain prohibitively high, primarily due to the substantial investment required for TSV formation at the wafer level. Meanwhile, conventional system-on-chip designs are transitioning from fin field-effect transistor to gate-all-around architectures using the latest sub-3 nm technology nodes and incorporating extreme ultraviolet lithography along with other cutting-edge techniques. Simultaneously, the academic community is fostering an environment that supports technology node utilization from legacy 180 to 7 nm, enabling designers to develop two-dimensional IC (2D-IC) chips with novel architectures at reduced costs. Despite these advancements, foundry shuttle services utilizing TSVs remain largely inaccessible, making proof-of-principle demonstrations and functional verification using 3D-ICs extremely challenging. This study introduces recent technological advancements that enable the transformation of 2D-ICs into 3D-ICs using shuttle chips from multi-project wafers, ranging from small to large-scale implementations. The discussion primarily focuses on die-level, short-turnaround-time 3D-IC fabrication, emphasizing key enabling technologies such as multichip thinning and TSV/microbump formation. In addition, the study explores the effectiveness of Meta BondingTM techniques, including fine-pitch microbump, direct bonding, and hybrid bonding, for future high-performance 3D-IC prototyping.
- 20 May 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal on Emerging and Selected Topics in Circuits and Systems
- Vol. PP  (21563357) , 1
Open AccessDie-Level Transformation of 2D Shuttle Chips into 3D-IC for Advanced Rapid Prototyping using Meta Bonding
Three-dimensional integrated circuit (3D-IC) technology, often referred to as through-silicon via (TSV) formation technology, has been steadily maturing and is increasingly used in advanced semiconductor devices such as 3D complementary metal-oxide-semiconductor image sensors, high-bandwidth memory, and static random-access memory on central processing units (commonly known as 3D V-Cache). However, the initial development costs remain prohibitively high, primarily due to the substantial investment required for TSV formation at the wafer level. Meanwhile, conventional system-on-chip designs are transitioning from fin field-effect transistor to gate-all-around architectures using the latest sub-3 nm technology nodes and incorporating extreme ultraviolet lithography along with other cutting-edge techniques. Simultaneously, the academic community is fostering an environment that supports technology node utilization from legacy 180 to 7 nm, enabling designers to develop two-dimensional IC (2D-IC) chips with novel architectures at reduced costs. Despite these advancements, foundry shuttle services utilizing TSVs remain largely inaccessible, making proof-of-principle demonstrations and functional verification using 3D-ICs extremely challenging. This study introduces recent technological advancements that enable the transformation of 2D-ICs into 3D-ICs using shuttle chips from multi-project wafers, ranging from small to large-scale implementations. The discussion primarily focuses on die-level, short-turnaround-time 3D-IC fabrication, emphasizing key enabling technologies such as multichip thinning and TSV/microbump formation. In addition, the study explores the effectiveness of Meta BondingTM techniques, including fine-pitch microbump, direct bonding, and hybrid bonding, for future high-performance 3D-IC prototyping.
- 1 October 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-4
Design for 3-D Stacked Neural Network Circuit with Cyclic Analog Computing
An analog-based CMOS neuron with crossbar configuration multiplier-adder circuit for three-dimensional chip stacking (3-D) cyclic signal processing deep neural network is presented. The circuit designed to verify the proposed concept is a total of four-layer stacked chips, consisting of two neuron chips that mainly perform analog multiply-accumulate (MAC) operations and two memory chips that store neuron potential (activation) and weight data. According to papers and conference presentations on accelerators for the neural network circuit, the analog MAC operation has the potential to achieve high efficiency by using the analog processing of the crossbar configuration. However, analog MAC circuit with compensation technique is required to overcome the decrease in computing accuracy due to electrical noise and element nonlinearity. In this study, we have designed and verified a highly linearity multiplier-adder circuit for highly efficient 3-D cyclic analog processing. In this paper, we focus on the current adder circuit with switched capacitor correlated double sampling (CDS) technique, to minimize non-linearity, noise, and circuit area. The proposed the analog adder circuit was designed with 180-nm and operation power supply voltage 1.8V CMOS technology, and the fabricated neuron chip that integrates 64 neurons and $64 \times 64$ synaptic connections. As analysis results, it was confirmed that the CDS technique apply analog adder circuit reduced the noise output voltage less than 0.1mV and the range of voltage linearity was over 1.6V.- 1 October 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-4
Design for 3-D Stacked Neural Network Circuit with Cyclic Analog Computing
An analog-based CMOS neuron with crossbar configuration multiplier-adder circuit for three-dimensional chip stacking (3-D) cyclic signal processing deep neural network is presented. The circuit designed to verify the proposed concept is a total of four-layer stacked chips, consisting of two neuron chips that mainly perform analog multiply-accumulate (MAC) operations and two memory chips that store neuron potential (activation) and weight data. According to papers and conference presentations on accelerators for the neural network circuit, the analog MAC operation has the potential to achieve high efficiency by using the analog processing of the crossbar configuration. However, analog MAC circuit with compensation technique is required to overcome the decrease in computing accuracy due to electrical noise and element nonlinearity. In this study, we have designed and verified a highly linearity multiplier-adder circuit for highly efficient 3-D cyclic analog processing. In this paper, we focus on the current adder circuit with switched capacitor correlated double sampling (CDS) technique, to minimize non-linearity, noise, and circuit area. The proposed the analog adder circuit was designed with 180-nm and operation power supply voltage 1.8V CMOS technology, and the fabricated neuron chip that integrates 64 neurons and $64 \times 64$ synaptic connections. As analysis results, it was confirmed that the CDS technique apply analog adder circuit reduced the noise output voltage less than 0.1mV and the range of voltage linearity was over 1.6V.- 1 October 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-4
Implementation of a Chaotic Neural Network Reservoir on a TSV/$\mu\text{Bump}$ Stacked 3D Cyclic Neural Network Integrated Circuit
A TSV/ $\mu\text{Bump}$ stacked 3D cyclic deep neural network integrated circuit architecture was proposed. Furthermore, a technique for embedding a chaotic neural network reservoir into the proposed architecture was devised. A proof-of-concept neural network chip, and a weight memory chip have been designed and fabricated to confirm the feasibly of the proposed architecture. In this study, the neuron chip configured as a cyclic chaotic neuron circuit is evaluated by measuring circuit building blocks, and constructing a dedicated MATLAB circuit emulator based on the measurement results from the chip. Some of the results from the MATLAB emulator are illustrated.- 1 October 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-4
Implementation of a Chaotic Neural Network Reservoir on a TSV/$\mu\text{Bump}$ Stacked 3D Cyclic Neural Network Integrated Circuit
A TSV/ $\mu\text{Bump}$ stacked 3D cyclic deep neural network integrated circuit architecture was proposed. Furthermore, a technique for embedding a chaotic neural network reservoir into the proposed architecture was devised. A proof-of-concept neural network chip, and a weight memory chip have been designed and fabricated to confirm the feasibly of the proposed architecture. In this study, the neuron chip configured as a cyclic chaotic neuron circuit is evaluated by measuring circuit building blocks, and constructing a dedicated MATLAB circuit emulator based on the measurement results from the chip. Some of the results from the MATLAB emulator are illustrated.- 1 January 2021
- Published by Institute of Electronics, Information and Communications Engineers (IEICE) in Nonlinear Theory and Its Applications, IEICE
- Vol. 12  (3) , 309-322
Cyclic reservoir neural network circuit for 3D IC implementation
Reservoir computing is a computational model inspired by the information processing of the brain. In particular, it shows high performance in time-series processing using recurrent neural network dynamics despite its simple structure. Furthermore, a simple learning algorithm only in the output layer is sufficient for training the entire network. Therefore, its efficient hardware implementation is highly expected. However, it is important for a reservoir network to have a rich variety of dynamics to deal with complex time-series information. To introduce rich dynamics in the reservoir network without degrading the network stability, a chaotic neural network reservoir was proposed. In this paper, we propose a cyclic reservoir neural network circuit suitable for a stacked three-dimensional (3D) integrated circuit (IC). Through 3D IC fabrication technology, in which several semiconductor substrates are vertically stacked and connected by through-silicon vias (TSVs), we can efficiently integrate the chaotic neural network reservoir circuit. We designed and fabricated a prototype IC chip of the proposed circuit with a TSMC 180 nm CMOS semiconductor process. We verified its operation through SPICE and MATLAB simulations and preliminary experiments with the fabricated prototype chip.
- 1 January 2021
- Published by Institute of Electronics, Information and Communications Engineers (IEICE) in Nonlinear Theory and Its Applications, IEICE
- Vol. 12  (3) , 309-322
Cyclic reservoir neural network circuit for 3D IC implementation
Reservoir computing is a computational model inspired by the information processing of the brain. In particular, it shows high performance in time-series processing using recurrent neural network dynamics despite its simple structure. Furthermore, a simple learning algorithm only in the output layer is sufficient for training the entire network. Therefore, its efficient hardware implementation is highly expected. However, it is important for a reservoir network to have a rich variety of dynamics to deal with complex time-series information. To introduce rich dynamics in the reservoir network without degrading the network stability, a chaotic neural network reservoir was proposed. In this paper, we propose a cyclic reservoir neural network circuit suitable for a stacked three-dimensional (3D) integrated circuit (IC). Through 3D IC fabrication technology, in which several semiconductor substrates are vertically stacked and connected by through-silicon vias (TSVs), we can efficiently integrate the chaotic neural network reservoir circuit. We designed and fabricated a prototype IC chip of the proposed circuit with a TSMC 180 nm CMOS semiconductor process. We verified its operation through SPICE and MATLAB simulations and preliminary experiments with the fabricated prototype chip.
- 24 September 2018
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Components, Packaging and Manufacturing Technology
- Vol. 9  (1) , 181-188
High-Thermoresistant Temporary Bonding Technology for Multichip-to-Wafer 3-D Integration With Via-Last TSVs
In this study, a high thermoresistant temporary bonding/debonding system using spin-on glass (SOG) and hydrogenated amorphous silicon (a-Si:H) layers is proposed for multichip-to-wafer (MCtW) 3D integration based on a vialast/ backside-via through-silicon via (TSV) approach. The shear strengths of chips bonded to wafers through the SOG layers are evaluated. In addition, the debonding performance of the chips from the wafers are investigated by using KrF excimer laser irradiation. Finally, a via-last/backside-via MCtW 3D integration process using the temporary bonding/debonding system is demonstrated to show high feasibility with successful interconnect formation of a Cu-TSVs daisy chain (10 μm in diameter and 50 μm in depth) with a SiO2 liner dielectric deposited by O3-TEOS CVD at 350°C .
- 24 September 2018
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Components, Packaging and Manufacturing Technology
- Vol. 9  (1) , 181-188
High-Thermoresistant Temporary Bonding Technology for Multichip-to-Wafer 3-D Integration With Via-Last TSVs
In this study, a high thermoresistant temporary bonding/debonding system using spin-on glass (SOG) and hydrogenated amorphous silicon (a-Si:H) layers is proposed for multichip-to-wafer (MCtW) 3D integration based on a vialast/ backside-via through-silicon via (TSV) approach. The shear strengths of chips bonded to wafers through the SOG layers are evaluated. In addition, the debonding performance of the chips from the wafers are investigated by using KrF excimer laser irradiation. Finally, a via-last/backside-via MCtW 3D integration process using the temporary bonding/debonding system is demonstrated to show high feasibility with successful interconnect formation of a Cu-TSVs daisy chain (10 μm in diameter and 50 μm in depth) with a SiO2 liner dielectric deposited by O3-TEOS CVD at 350°C .
- 21 July 2018
- Published by Springer Nature
- p. 719-733
A 3D-VLSI Architecture for Future Automotive Visual Recognition
- 21 July 2018
- Published by Springer Nature
- p. 719-733
A 3D-VLSI Architecture for Future Automotive Visual Recognition
- 8 November 2017
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Electron Devices
- Vol. 64  (12) , 5065-5072
Self-Assembly and Electrostatic Carrier Technology for Via-Last TSV Formation Using Transfer Stacking-Based Chip-to-Wafer 3-D Integration
A self-assembly and electrostatic (SAE) carrier technology is developed for high-precision and high-throughput chip-to-wafer 3-D integration. In this paper, water surface tension-driven chip assembly is combined with electrostatic adhesion to keep high alignment accuracies obtained by the capillary self-assembly process. The self-assembled chips can be firmly fixed on an SAE carrier wafer by electrostatic adhesion, and then, the chips can be readily detached from the carrier by discharging and transferred to another carrier with a temporary adhesive. This paper describes the impact of chip clamping forces and electrical reliability of the SAE carrier on chips to be 3-D stacked in chip-to-wafer configuration. Through-Si via formation is demonstrated by using a via-last 3-D integration process based on the SAE carrier. The demonstration shows that the SAE carrier maintains higher chip alignment accuracies than does conventional carrier without electrostatic adhesion.
- 8 November 2017
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Electron Devices
- Vol. 64  (12) , 5065-5072
Self-Assembly and Electrostatic Carrier Technology for Via-Last TSV Formation Using Transfer Stacking-Based Chip-to-Wafer 3-D Integration
A self-assembly and electrostatic (SAE) carrier technology is developed for high-precision and high-throughput chip-to-wafer 3-D integration. In this paper, water surface tension-driven chip assembly is combined with electrostatic adhesion to keep high alignment accuracies obtained by the capillary self-assembly process. The self-assembled chips can be firmly fixed on an SAE carrier wafer by electrostatic adhesion, and then, the chips can be readily detached from the carrier by discharging and transferred to another carrier with a temporary adhesive. This paper describes the impact of chip clamping forces and electrical reliability of the SAE carrier on chips to be 3-D stacked in chip-to-wafer configuration. Through-Si via formation is demonstrated by using a via-last 3-D integration process based on the SAE carrier. The demonstration shows that the SAE carrier maintains higher chip alignment accuracies than does conventional carrier without electrostatic adhesion.
- Open Access
- 10 October 2016
- Published by MDPI AG in Micromachines
- Vol. 7  (10) , 184
Oxide-Oxide Thermocompression Direct Bonding Technologies with Capillary Self-Assembly for Multichip-to-Wafer Heterogeneous 3D System Integration
Plasma- and water-assisted oxide-oxide thermocompression direct bonding for a self-assembly based multichip-to-wafer (MCtW) 3D integration approach was demonstrated. The bonding yields and bonding strengths of the self-assembled chips obtained by the MCtW direct bonding technology were evaluated. In this study, chemical mechanical polish (CMP)-treated oxide formed by plasma-enhanced chemical vapor deposition (PE-CVD) as a MCtW bonding interface was mainly employed, and in addition, wafer-to-wafer thermocompression direct bonding was also used for comparison. N2 or Ar plasmas were utilized for the surface activation. After plasma activation and the subsequent supplying of water as a self-assembly mediate, the chips with the PE-CVD oxide layer were driven by the liquid surface tension and precisely aligned on the host wafers, and subsequently, they were tightly bonded to the wafers through the MCtW oxide-oxide direct bonding technology. Finally, a mechanism of oxide-oxide direct bonding to support the previous models was discussed using an atmospheric pressure ionization mass spectrometer (APIMS).
- 10 October 2016
- Published by MDPI AG in Micromachines
- Vol. 7  (10) , 184
Open AccessOxide-Oxide Thermocompression Direct Bonding Technologies with Capillary Self-Assembly for Multichip-to-Wafer Heterogeneous 3D System Integration
Plasma- and water-assisted oxide-oxide thermocompression direct bonding for a self-assembly based multichip-to-wafer (MCtW) 3D integration approach was demonstrated. The bonding yields and bonding strengths of the self-assembled chips obtained by the MCtW direct bonding technology were evaluated. In this study, chemical mechanical polish (CMP)-treated oxide formed by plasma-enhanced chemical vapor deposition (PE-CVD) as a MCtW bonding interface was mainly employed, and in addition, wafer-to-wafer thermocompression direct bonding was also used for comparison. N2 or Ar plasmas were utilized for the surface activation. After plasma activation and the subsequent supplying of water as a self-assembly mediate, the chips with the PE-CVD oxide layer were driven by the liquid surface tension and precisely aligned on the host wafers, and subsequently, they were tightly bonded to the wafers through the MCtW oxide-oxide direct bonding technology. Finally, a mechanism of oxide-oxide direct bonding to support the previous models was discussed using an atmospheric pressure ionization mass spectrometer (APIMS).
- 23 June 2016
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Components, Packaging and Manufacturing Technology
- Vol. 6  (7) , 1000-1006
Heterogeneous 3-D Integration Using Self-Assembly and Electrostatic Bonding
To overcome various concerns caused by scaling down the device size in future Large Scale Integrated Circuits (LSIs), it is indispensable to introduce a new concept of heterogeneous 3-D integration in which various kinds of device chips with different sizes, different devices, and different materials are vertically stacked. To achieve such heterogeneous 3-D integration, a key technology of self-assembly and electrostatic bonding has been developed. Exploring new devices for the Internet of Things, we have fabricated several kinds of heterogeneous 3-D LSIs called superchip by stacking compound semiconductor device chip, photonic device chip, and spintronic device chip on CMOS device chips using self-assembly and electrostatic bonding. Furthermore, a new system integration technology using a large-area substrate has been developed to reduce the cost of 2.5-D/3-D system module. A 3-D-stacked image sensor system module for automatic driving vehicle has been fabricated using this technology.
- 23 June 2016
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Components, Packaging and Manufacturing Technology
- Vol. 6  (7) , 1000-1006
Heterogeneous 3-D Integration Using Self-Assembly and Electrostatic Bonding
To overcome various concerns caused by scaling down the device size in future Large Scale Integrated Circuits (LSIs), it is indispensable to introduce a new concept of heterogeneous 3-D integration in which various kinds of device chips with different sizes, different devices, and different materials are vertically stacked. To achieve such heterogeneous 3-D integration, a key technology of self-assembly and electrostatic bonding has been developed. Exploring new devices for the Internet of Things, we have fabricated several kinds of heterogeneous 3-D LSIs called superchip by stacking compound semiconductor device chip, photonic device chip, and spintronic device chip on CMOS device chips using self-assembly and electrostatic bonding. Furthermore, a new system integration technology using a large-area substrate has been developed to reduce the cost of 2.5-D/3-D system module. A 3-D-stacked image sensor system module for automatic driving vehicle has been fabricated using this technology.
- 4 November 2015
- Published by Institute of Electrical and Electronics Engineers (IEEE) in Journal of Microelectromechanical Systems
- Vol. 25  (1) , 91-100
Impact of Chip-Edge Structures on Alignment Accuracies of Self-Assembled Dies for Microelectronic System Integration
The self-assembly of known good dies on hosting substrates using liquid surface tension is a promising technology to create highly integrated 3-D and heterogeneous microelectronic systems. In this paper, we investigate the effects of the edge structures of self-assembled chips on alignment accuracies. Nine types of 100-μm-thick Si chips (3 mm × 3 mm) with and without step geometries on their hydrophilic or hydrophobic peripheries are self-assembled onto hydrophilic assembly sites formed on planar- and plateau-type host substrates. When hydrophobic peripheries with step geometries are applied to both the edges of chips and assembly sites formed on substrates, the resulting average alignment accuracy is 300 nm. Total accuracy variation within 2 μm is realized by using either chip or substrate having 10-μm-height step structures with hydrophobic edges. We obtain a high tolerance for initial offsets indicating positioning misalignment prior to chip release, with the plateau-type substrates and the chips having hydrophobic step structures at the edges. These chips are precisely self-assembled, even under a large initial offset of 1.5 mm in a horizontal direction to both the substrates. The extremely large offset is comparable with 50% of the side length of the 3-mm-square chip. On the other hand, the chips formed by an accurate saw dicing that gives high chip-size accuracies as designed exhibit high alignment accuracies and tolerances when compared with the chips with the hydrophobic step structures and the chips formed by plasma dicing, which offer a large pseudo step with a height of 100 μm.
- 4 November 2015
- Published by Institute of Electrical and Electronics Engineers (IEEE) in Journal of Microelectromechanical Systems
- Vol. 25  (1) , 91-100
Impact of Chip-Edge Structures on Alignment Accuracies of Self-Assembled Dies for Microelectronic System Integration
The self-assembly of known good dies on hosting substrates using liquid surface tension is a promising technology to create highly integrated 3-D and heterogeneous microelectronic systems. In this paper, we investigate the effects of the edge structures of self-assembled chips on alignment accuracies. Nine types of 100-μm-thick Si chips (3 mm × 3 mm) with and without step geometries on their hydrophilic or hydrophobic peripheries are self-assembled onto hydrophilic assembly sites formed on planar- and plateau-type host substrates. When hydrophobic peripheries with step geometries are applied to both the edges of chips and assembly sites formed on substrates, the resulting average alignment accuracy is 300 nm. Total accuracy variation within 2 μm is realized by using either chip or substrate having 10-μm-height step structures with hydrophobic edges. We obtain a high tolerance for initial offsets indicating positioning misalignment prior to chip release, with the plateau-type substrates and the chips having hydrophobic step structures at the edges. These chips are precisely self-assembled, even under a large initial offset of 1.5 mm in a horizontal direction to both the substrates. The extremely large offset is comparable with 50% of the side length of the 3-mm-square chip. On the other hand, the chips formed by an accurate saw dicing that gives high chip-size accuracies as designed exhibit high alignment accuracies and tolerances when compared with the chips with the hydrophobic step structures and the chips formed by plasma dicing, which offer a large pseudo step with a height of 100 μm.