Simon Thomann
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
Research Subjects
Simon Thomann has a total of 40 co-authors. Here are the top co-authors by number of shared publications:
- Hussam Amrouch
Technical University of Munich
14 shared publications - Albi Mema
TUM School of Computation, Information and Technology, Chair of AI Processor Design, Munich Institute of Robotics and Machine Intelligence (MIRMI), Technical University of Munich, Munich, Germany
6 shared publications - Yogesh Singh Chauhan
Indian Institute of Technology (IIT) Kanpur
4 shared publications - Paul R. Genssler
TUM School of Computation, Information and Technology, Chair of AI Processor Design, Munich Institute of Robotics and Machine Intelligence, Technical University of Munich, Munich, Germany
3 shared publications - Girish Pahwa
International College of Semiconductor Technology, National Yang Ming Chiao Tung University, Hsinchu, Taiwan
2 shared publications Hussam Amrouch
School of Computation, Information and Technology, Techical University of Munich, Munich, Germany
2 shared publications
Latest Publications
- Open Access
- 4 September 2026
- Published by Springer Nature in Microsystems & Nanoengineering
- Vol. 12 (1) , 313
High-precision hyperdimensional computing enabled by in-memory computing using high-polarization ferroelectric Hf0.5Zr0.5O2 capacitors
We demonstrate high-precision hyperdimensional computing using an in-memory computing (IMC) architecture based on ferroelectric Hf0.5Zr0.5O2 (HZO) capacitors. By exploiting the high polarization charge density of CMOS back-end-compatible HZO, we achieved 32 well-separated and linearly programmable intermediate states in 10-nm-thick capacitors making them suitable as capacitive IMC elements. In recent times, capacitive IMC emerged as a promising energy- and latency-efficient route for data-intensive computing tasks. However, compute-in-memory elements require non-volatile, reproducible, and multi-bit operation. In this work, we show that through optimized device fabrication without vacuum break between oxide and nitride depositions and tailored thermal engineering, the HZO capacitors can exhibit high remanent polarization (2Pr = 75 µC/cm²). Structural studies highlight a high orthorhombic phase fraction and clean HZO/TiN interface. The intermediate polarization states exhibit controllable, linear, and reproducible capacitance modulation via voltage-driven polarization switching, enabling reliable multi-bit device operation and non-destructive readout. Leveraging these 5-bit ferroelectric capacitors, it is possible to store 15-bit numerical values using only three capacitors to implement high-precision capacitive IMC in a hyperdimensional computing task, achieving improved inference accuracy of 92.3% and 2.3x reduced areal footprint compared to binary encoding. These results highlight the importance of advanced materials engineering to achieve high bit-precision and state linearity in ferroelectric capacitors for scalable capacitive in-memory computing.
- 4 September 2026
- Published by Springer Nature in Microsystems & Nanoengineering
- Vol. 12 (1) , 313
Open AccessHigh-precision hyperdimensional computing enabled by in-memory computing using high-polarization ferroelectric Hf0.5Zr0.5O2 capacitors
We demonstrate high-precision hyperdimensional computing using an in-memory computing (IMC) architecture based on ferroelectric Hf0.5Zr0.5O2 (HZO) capacitors. By exploiting the high polarization charge density of CMOS back-end-compatible HZO, we achieved 32 well-separated and linearly programmable intermediate states in 10-nm-thick capacitors making them suitable as capacitive IMC elements. In recent times, capacitive IMC emerged as a promising energy- and latency-efficient route for data-intensive computing tasks. However, compute-in-memory elements require non-volatile, reproducible, and multi-bit operation. In this work, we show that through optimized device fabrication without vacuum break between oxide and nitride depositions and tailored thermal engineering, the HZO capacitors can exhibit high remanent polarization (2Pr = 75 µC/cm²). Structural studies highlight a high orthorhombic phase fraction and clean HZO/TiN interface. The intermediate polarization states exhibit controllable, linear, and reproducible capacitance modulation via voltage-driven polarization switching, enabling reliable multi-bit device operation and non-destructive readout. Leveraging these 5-bit ferroelectric capacitors, it is possible to store 15-bit numerical values using only three capacitors to implement high-precision capacitive IMC in a hyperdimensional computing task, achieving improved inference accuracy of 92.3% and 2.3x reduced areal footprint compared to binary encoding. These results highlight the importance of advanced materials engineering to achieve high bit-precision and state linearity in ferroelectric capacitors for scalable capacitive in-memory computing.
- Open Access
- 11 May 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Solid-State Circuits Letters
- Vol. 9, 166-169
FeFET-Based CMOS Current Starvation Programmable Delay Element
Programmable delay elements (PDEs) are crucial circuit building blocks that enable precise timing adjustments of signal transitions. They are used in various critical applications, like time-to-digital converters. They are often used to mitigate aging by clock skew tuning post-fabrication. However, typical CMOS PDE designs require many transistors while still offering a limited delay range. This work implements a PDE based on the current starvation concept and utilizes Ferroelectric FETs (FeFETs) to realize delay programmability in a non-volatile manner. We fabricated 28 nm FeFET (both n-type and p-type) and measured their characteristics. We then design, simulate, and tape-out the proposed PDE with a ring oscillator (RO) as circuit-under-test to evaluate and demonstrate our concept. The frequency of the RO is successfully controlled via the polarization of the FeFETs and was measured via a wafer probe setup. Our proposed design offers a wide range of modulation from 2.8 MHz (when the FeFETs are programmed to the high VTH state) up to 7.47 MHz (when the FeFETs are programmed to the low VTH state). This in turn corresponds to 223.27 ns difference on the period between the two ferroelectric polarization states that put the underlying transistor into two distinguished low and high resistive states.
- 11 May 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Solid-State Circuits Letters
- Vol. 9, 166-169
Open AccessFeFET-Based CMOS Current Starvation Programmable Delay Element
Programmable delay elements (PDEs) are crucial circuit building blocks that enable precise timing adjustments of signal transitions. They are used in various critical applications, like time-to-digital converters. They are often used to mitigate aging by clock skew tuning post-fabrication. However, typical CMOS PDE designs require many transistors while still offering a limited delay range. This work implements a PDE based on the current starvation concept and utilizes Ferroelectric FETs (FeFETs) to realize delay programmability in a non-volatile manner. We fabricated 28 nm FeFET (both n-type and p-type) and measured their characteristics. We then design, simulate, and tape-out the proposed PDE with a ring oscillator (RO) as circuit-under-test to evaluate and demonstrate our concept. The frequency of the RO is successfully controlled via the polarization of the FeFETs and was measured via a wafer probe setup. Our proposed design offers a wide range of modulation from 2.8 MHz (when the FeFETs are programmed to the high VTH state) up to 7.47 MHz (when the FeFETs are programmed to the low VTH state). This in turn corresponds to 223.27 ns difference on the period between the two ferroelectric polarization states that put the underlying transistor into two distinguished low and high resistive states.
- Open Access
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Circuits and Systems I: Regular Papers
- p. 1-13
First Demonstration of 28-nm Fabricated FeFET-Based Nonvolatile 6T SRAM
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Circuits and Systems I: Regular Papers
- p. 1-13
Open AccessFirst Demonstration of 28-nm Fabricated FeFET-Based Nonvolatile 6T SRAM
- Open Access
- 2 June 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 105141-105150
ML-Driven Transistor Self-Heating Analysis From TCAD to Large IP Circuits
To mitigate the increasing short-channel effects of miniaturization, the channel is wrapped increasingly with gate metal and insulating oxides to improve electrostatic control while worsening self-heating. Hence, self-heating has become a reliability concern and will increase in severity with more advanced nodes in the future (e.g., nanosheets), accelerating aging and shortening the transistor’s lifetime. Although standardized compact models include self-heating, their lumped nature abstracts the temperature gradient over the transistor’s cross-section. To this end, we have used TCAD to generate a distributed channel temperature map data set. We propose a technology-agnostic methodology for extracting localized temperature data on the circuit level using the trained machine learning model (with a mean error of 0.1 K). We explore our method to study self-heating in complex large circuits featuring over 3700 transistors. This includes foundries IPs like a complete high-performance SRAM-based register file. Our method provides the thermal map for every transistor and time step. The voltage and thermal traces are processed with an aging framework to estimate the expected aging more realistically. Using the proposed methodology reduces the conservative constant 50 mV threshold voltage shift to 31 mV, improving circuit performance through less pessimistic aging estimations. This unprecedented detailed analysis of self-heating at the large scale of the circuit will open new doors for designers to evaluate how self-heating propagates from transistors to circuits and impacts their reliability and lifetime.
- 2 June 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 105141-105150
Open AccessML-Driven Transistor Self-Heating Analysis From TCAD to Large IP Circuits
To mitigate the increasing short-channel effects of miniaturization, the channel is wrapped increasingly with gate metal and insulating oxides to improve electrostatic control while worsening self-heating. Hence, self-heating has become a reliability concern and will increase in severity with more advanced nodes in the future (e.g., nanosheets), accelerating aging and shortening the transistor’s lifetime. Although standardized compact models include self-heating, their lumped nature abstracts the temperature gradient over the transistor’s cross-section. To this end, we have used TCAD to generate a distributed channel temperature map data set. We propose a technology-agnostic methodology for extracting localized temperature data on the circuit level using the trained machine learning model (with a mean error of 0.1 K). We explore our method to study self-heating in complex large circuits featuring over 3700 transistors. This includes foundries IPs like a complete high-performance SRAM-based register file. Our method provides the thermal map for every transistor and time step. The voltage and thermal traces are processed with an aging framework to estimate the expected aging more realistically. Using the proposed methodology reduces the conservative constant 50 mV threshold voltage shift to 31 mV, improving circuit performance through less pessimistic aging estimations. This unprecedented detailed analysis of self-heating at the large scale of the circuit will open new doors for designers to evaluate how self-heating propagates from transistors to circuits and impacts their reliability and lifetime.
- Open Access
- 30 January 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Circuits and Systems for Artificial Intelligence
- Vol. 2 (2) , 114-125
Capacitor Minimization in Analog Synapses for Efficient and Compact Binarized SNNs
Accelerating the immense workloads of Neural Networks (NNs) is a critical challenge, and analog computing-based approaches present a promising solution. Among these, the Integrate-and-Fire (IF) Spiking Neural Network (SNN) stands out for its potential in efficient neural computation. However, achieving high inference accuracy through precise multiply-accumulate (MAC) operations necessitates a large membrane capacitor—224 to 547 times larger than the synapse array at the 14nm technology node—resulting in prohibitive area costs. Additionally, the large capacitor size introduces higher energy consumption and longer access times. Thus, a key research challenge is maintaining accuracy while mitigating the costs associated with the large membrane capacitor. In this work, we propose a HW/SW Codesign method, called CapMin, for capacitor size minimization in analog computing IF-SNNs. CapMin minimizes the capacitor size by reducing the number of spike times needed for accurate operation of the HW, based on the absolute frequency of MAC level occurrences in the SW. To increase the operation of IF-SNNs to current variation, we propose the method CapMin-V, which trades capacitor size for protection based on the reduced capacitor size found in CapMin. In our experiments, CapMin achieves more than a 14× reduction in capacitor size along with a 34% reduction in energy consumption over the state of the art, while CapMin-V achieves increased variation tolerance in the IF-SNN operation, requiring only a small increase in capacitor size.
- 30 January 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Circuits and Systems for Artificial Intelligence
- Vol. 2 (2) , 114-125
Open AccessCapacitor Minimization in Analog Synapses for Efficient and Compact Binarized SNNs
Accelerating the immense workloads of Neural Networks (NNs) is a critical challenge, and analog computing-based approaches present a promising solution. Among these, the Integrate-and-Fire (IF) Spiking Neural Network (SNN) stands out for its potential in efficient neural computation. However, achieving high inference accuracy through precise multiply-accumulate (MAC) operations necessitates a large membrane capacitor—224 to 547 times larger than the synapse array at the 14nm technology node—resulting in prohibitive area costs. Additionally, the large capacitor size introduces higher energy consumption and longer access times. Thus, a key research challenge is maintaining accuracy while mitigating the costs associated with the large membrane capacitor. In this work, we propose a HW/SW Codesign method, called CapMin, for capacitor size minimization in analog computing IF-SNNs. CapMin minimizes the capacitor size by reducing the number of spike times needed for accurate operation of the HW, based on the absolute frequency of MAC level occurrences in the SW. To increase the operation of IF-SNNs to current variation, we propose the method CapMin-V, which trades capacitor size for protection based on the reduced capacitor size found in CapMin. In our experiments, CapMin achieves more than a 14× reduction in capacitor size along with a 34% reduction in energy consumption over the state of the art, while CapMin-V achieves increased variation tolerance in the IF-SNN operation, requiring only a small increase in capacitor size.
- 5 December 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Electron Devices
- Vol. 71 (1) , 213-222
ML-TCAD: Accelerating FeFET Reliability Analysis Using Machine Learning
Physics-based simulations using technology computer aided design (TCAD) offer high accuracy while suffering from exceedingly slow computations and significant license costs (especially when high parallelism is inevitable for Monte Carlo analysis), rendering large-scale design-space explorations infeasible. Therefore, we employ machine learning (ML) algorithms, trained on accurate datasets produced by TCAD, in order to massively accelerate multidomain ferroelectric FET (FeFET) simulations and take TCAD out of our framework’s loop. Part of this study explores approaches to predicting I – V characteristics using pure ML means or augmenting it with simple physics models. The huge speedup (13 600 $\times$ ) obtained through our ML-based modeling enabled unprecedented analysis of dtd and ctc variability for FeFET technology. Furthermore, we demonstrate how computationally infeasible analysis that would take years using TCAD (e.g., read disturbance pulses with 200 million time steps) became feasible for the first time.
- 5 December 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Electron Devices
- Vol. 71 (1) , 213-222
ML-TCAD: Accelerating FeFET Reliability Analysis Using Machine Learning
Physics-based simulations using technology computer aided design (TCAD) offer high accuracy while suffering from exceedingly slow computations and significant license costs (especially when high parallelism is inevitable for Monte Carlo analysis), rendering large-scale design-space explorations infeasible. Therefore, we employ machine learning (ML) algorithms, trained on accurate datasets produced by TCAD, in order to massively accelerate multidomain ferroelectric FET (FeFET) simulations and take TCAD out of our framework’s loop. Part of this study explores approaches to predicting I – V characteristics using pure ML means or augmenting it with simple physics models. The huge speedup (13 600 $\times$ ) obtained through our ML-based modeling enabled unprecedented analysis of dtd and ctc variability for FeFET technology. Furthermore, we demonstrate how computationally infeasible analysis that would take years using TCAD (e.g., read disturbance pulses with 200 million time steps) became feasible for the first time.
- 13 October 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Electron Devices
- Vol. 71 (1) , 368-373
Comprehensive Modeling of Switching Behavior in BEOL FeFET for Monolithic 3-D Integration
In this article, we have developed a comprehensive modeling framework to explain the switching characteristics of BEOL-compatible ferroelectric field-effect transistor (FeFET) with an amorphous IGZO channel. Our TCAD-based modeling framework, calibrated against measurement data, jointly incorporates: 1) the distributed channel; 2) a physics-based nucleation-limited switching dynamics model for multidomain ferroelectric polarization (P $_{\textbf{FE}}$ ); and 3) the domain-domain interaction. To our knowledge, this is the first demonstration of a physics-based comprehensive model of BEOL-compatible FeFET. Our model reproduces and explains the experimentally observed abrupt current jumps in the reverse and forward dc sweeps. Furthermore, our model is capable of processing arbitrary input waveforms such as quasi-dc and different kinds of pulse trains used in neuromorphic applications. This comprehensive modeling framework would enable researchers to explore the BEOL FeFET applications and guide device optimization and development.
- 13 October 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Electron Devices
- Vol. 71 (1) , 368-373
Comprehensive Modeling of Switching Behavior in BEOL FeFET for Monolithic 3-D Integration
In this article, we have developed a comprehensive modeling framework to explain the switching characteristics of BEOL-compatible ferroelectric field-effect transistor (FeFET) with an amorphous IGZO channel. Our TCAD-based modeling framework, calibrated against measurement data, jointly incorporates: 1) the distributed channel; 2) a physics-based nucleation-limited switching dynamics model for multidomain ferroelectric polarization (P $_{\textbf{FE}}$ ); and 3) the domain-domain interaction. To our knowledge, this is the first demonstration of a physics-based comprehensive model of BEOL-compatible FeFET. Our model reproduces and explains the experimentally observed abrupt current jumps in the reverse and forward dc sweeps. Furthermore, our model is capable of processing arbitrary input waveforms such as quasi-dc and different kinds of pulse trains used in neuromorphic applications. This comprehensive modeling framework would enable researchers to explore the BEOL FeFET applications and guide device optimization and development.
- 3 October 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Electron Devices
- Vol. 71 (1) , 287-293
FeFET Reliability Modeling for In-Memory Computing: Challenges, Perspective, and Emerging Trends
Ferroelectric FET (FeFET) is a singularly attractive emerging technology with a rich feature set. Boasting high versatility, it has already been implemented in a host of applications, like dense memory arrays and in-memory computing (IMC). Nevertheless, being an emerging technology, FeFETs are faced with reliability concerns that warrant analysis and mitigation. As a first step, modeling the FeFET variability can help to ascertain its present applicability. For instance, binary neural networks (BNNs) are more error resilient than classical deep-neural networks implementations and thus, can utilize this emerging yet unreliable hardware. However, conventional machine learning (ML) still relies on large datasets for training that need to be processed many times by the algorithm, exposing the memory wall of the von Neumann architecture. This necessitates the pursuit of new computing paradigms, like hyperdimensional computing (HDC). HDC is an emerging brain-inspired ML algorithm that can learn from very little data, using pattern recognition and randomness as its core working principles. Thus, it is inherently robust against noise and errors in the employed computing hardware. Hence, the robust HDC and the efficient yet unreliable FeFET are fiercely symbiotic. In this work, we will cover the challenges of the emerging FeFET technology and how to model them on a device level. We will then make use of these models and inject them into applications to study their error resiliency. For this, we will use the aforementioned applications of BNNs and HDC, and showcase their response to the unreliable hardware.
- 3 October 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Electron Devices
- Vol. 71 (1) , 287-293
FeFET Reliability Modeling for In-Memory Computing: Challenges, Perspective, and Emerging Trends
Ferroelectric FET (FeFET) is a singularly attractive emerging technology with a rich feature set. Boasting high versatility, it has already been implemented in a host of applications, like dense memory arrays and in-memory computing (IMC). Nevertheless, being an emerging technology, FeFETs are faced with reliability concerns that warrant analysis and mitigation. As a first step, modeling the FeFET variability can help to ascertain its present applicability. For instance, binary neural networks (BNNs) are more error resilient than classical deep-neural networks implementations and thus, can utilize this emerging yet unreliable hardware. However, conventional machine learning (ML) still relies on large datasets for training that need to be processed many times by the algorithm, exposing the memory wall of the von Neumann architecture. This necessitates the pursuit of new computing paradigms, like hyperdimensional computing (HDC). HDC is an emerging brain-inspired ML algorithm that can learn from very little data, using pattern recognition and randomness as its core working principles. Thus, it is inherently robust against noise and errors in the employed computing hardware. Hence, the robust HDC and the efficient yet unreliable FeFET are fiercely symbiotic. In this work, we will cover the challenges of the emerging FeFET technology and how to model them on a device level. We will then make use of these models and inject them into applications to study their error resiliency. For this, we will use the aforementioned applications of BNNs and HDC, and showcase their response to the unreliable hardware.
- Open Access
- 17 September 2023
- Published by Association for Computing Machinery (ACM)
- Vol. 71, 5-6
Tutorial: The Synergy of Hyperdimensional and In-memory Computing
Breakthroughs in deep learning consistently drive innovation. However, DNNs tend to overwhelm conventional computing systems. Hyperdimensional Computing (HDC) is rapidly gaining prominence as a potent method for rapid learning from a relatively small amount of data. It also holds the promise of offering energy-efficient lightweight computation. This tutorial will provide a comprehensive overview of the major shortcomings of existing von Neumann architectures and the growing need for innovative designs that fundamentally reduce memory latency and energy consumption by enabling data processing within the memory itself. Additionally, the tutorial will delve into the immense potential of beyond-von Neumann architectures, which utilize both emerging beyond-CMOS devices like Ferroelectric Field-Effect Transistors (FeFET) and conventional CMOS devices.
- 17 September 2023
- Published by Association for Computing Machinery (ACM)
- Vol. 71, 5-6
Open AccessTutorial: The Synergy of Hyperdimensional and In-memory Computing
Breakthroughs in deep learning consistently drive innovation. However, DNNs tend to overwhelm conventional computing systems. Hyperdimensional Computing (HDC) is rapidly gaining prominence as a potent method for rapid learning from a relatively small amount of data. It also holds the promise of offering energy-efficient lightweight computation. This tutorial will provide a comprehensive overview of the major shortcomings of existing von Neumann architectures and the growing need for innovative designs that fundamentally reduce memory latency and energy consumption by enabling data processing within the memory itself. Additionally, the tutorial will delve into the immense potential of beyond-von Neumann architectures, which utilize both emerging beyond-CMOS devices like Ferroelectric Field-Effect Transistors (FeFET) and conventional CMOS devices.
- Open Access
- 29 August 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Open Journal of Circuits and Systems
- Vol. 4, 258-270
Cryogenic In-Memory Computing for Quantum Processors Using Commercial 5-nm FinFETs
Cryogenic CMOS circuits that efficiently connect the classical domain with the quantum world are the cornerstone in bringing large-scale quantum processors to reality. The major challenges are, however, the tight power budget (in the order of milliwatts) and small latency (in the order of microseconds) requirements that such circuits inevitably must fulfill when operating at cryogenic temperatures. In-memory computing (IMC) is rapidly emerging as an attractive paradigm that holds the promise of performing computations efficiently where the data does not need to move back and forth between the CPU and the memory. Hence, it overcomes the fundamental bottleneck in classical von Neumann architectures, which provides considerable savings in power and latency. In this work, for the first time, we propose and implement an end-to-end approach that investigates SRAM-based IMC for cryogenic CMOS. To achieve that, we first characterize commercial 5 nm FinFETs from 300 K down to 10 K. Then, we employ the first cryogenic-aware industry-standard compact model for the FinFET technology (BSIM-CMG) to empower SPICE to accurately capture how cryogenic temperatures alter the electrical characteristics of transistors (e.g., threshold voltage, carrier mobility, sub-threshold slope, etc.). Our key contributions span from (1) carefully calibrating the cryogenic-aware BSIM-CMG against commercial 5 nm FinFET measurements in which SPICE simulations come with an excellent agreement with the experimental data, (2) exploring how the robustness of SRAM cells against noise (during the hold, read, and write operations) changes at extremely low temperatures, (3) investigating how the behavior of SRAM-based IMC circuits changes at 10 K compared to 300 K, and (4) modeling the error probabilities of IMC circuits that are used to calculate the Hamming distance, which is one of the essential similarity calculations to perform classifications.
- 29 August 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Open Journal of Circuits and Systems
- Vol. 4, 258-270
Open AccessCryogenic In-Memory Computing for Quantum Processors Using Commercial 5-nm FinFETs
Cryogenic CMOS circuits that efficiently connect the classical domain with the quantum world are the cornerstone in bringing large-scale quantum processors to reality. The major challenges are, however, the tight power budget (in the order of milliwatts) and small latency (in the order of microseconds) requirements that such circuits inevitably must fulfill when operating at cryogenic temperatures. In-memory computing (IMC) is rapidly emerging as an attractive paradigm that holds the promise of performing computations efficiently where the data does not need to move back and forth between the CPU and the memory. Hence, it overcomes the fundamental bottleneck in classical von Neumann architectures, which provides considerable savings in power and latency. In this work, for the first time, we propose and implement an end-to-end approach that investigates SRAM-based IMC for cryogenic CMOS. To achieve that, we first characterize commercial 5 nm FinFETs from 300 K down to 10 K. Then, we employ the first cryogenic-aware industry-standard compact model for the FinFET technology (BSIM-CMG) to empower SPICE to accurately capture how cryogenic temperatures alter the electrical characteristics of transistors (e.g., threshold voltage, carrier mobility, sub-threshold slope, etc.). Our key contributions span from (1) carefully calibrating the cryogenic-aware BSIM-CMG against commercial 5 nm FinFET measurements in which SPICE simulations come with an excellent agreement with the experimental data, (2) exploring how the robustness of SRAM cells against noise (during the hold, read, and write operations) changes at extremely low temperatures, (3) investigating how the behavior of SRAM-based IMC circuits changes at 10 K compared to 300 K, and (4) modeling the error probabilities of IMC circuits that are used to calculate the Hamming distance, which is one of the essential similarity calculations to perform classifications.