DC Cryogenic Modeling of Open-Source SkyWater 130 nm MOSFETs at 77 K Using BSIM4

This paper presents a publicly available, SPICE-compatible BSIM4 model for SkyWater 130 nm MOSFETs at 77 K, developed through DC characterization to enable reliable, low-power circuit design for high-energy physics applications such as liquid argon detectors.

The Gist

The Big Picture: Building a "Winter Coat" for Computer Chips

Imagine you have a very sophisticated, high-speed race car (a computer chip) designed to drive on a sunny, 70°F (21°C) day. Now, imagine you want to drive that same car in the middle of a blizzard at -320°F (-196°C).

If you just put the car in the freezer, the engine would seize up, the tires would get brittle, and the fuel would freeze. The car wouldn't work at all.

This is exactly the problem scientists face with High-Energy Physics (HEP) experiments. They use massive detectors filled with liquid argon (which is super cold, around 77 Kelvin or -320°F) to catch particles from the universe. To read the data from these detectors, they need tiny electronic brains (CMOS chips) to sit inside the freezing liquid.

The problem? Standard computer chips are designed for room temperature. If you put them in liquid argon, they behave unpredictably. Engineers need a "map" or a "recipe" to tell computer simulation software how these chips will act when they are freezing cold.

This paper is about creating that map for a specific, open-source chip design called "SkyWater 130nm" (SKY130).

---

The Characters in Our Story

1. The Chip (SKY130): Think of this as a popular, open-source "Lego set" for building electronics. Unlike expensive, secret Lego sets owned by big corporations, anyone can download these blueprints for free. It's the "Linux" of chip design.
2. The Temperature (77 K): This is the temperature of liquid nitrogen. It's cold enough to freeze your breath instantly, but not as cold as deep space (4 K). It's the sweet spot for many particle detectors.
3. The Model (BSIM4): This is the "instruction manual" or the "simulator" that engineers use to design circuits before they build them. It's a set of mathematical rules that predicts how electricity flows through a transistor.
4. The Problem: The current instruction manual only works for "room temperature" (300 K). If you try to use it for "freezer temperature" (77 K), the predictions are wrong. It's like using a map of a city in summer to navigate it in winter; the roads might be covered in snow, and the traffic rules change.

---

What Did the Scientists Do?

The team from the University of Texas at Arlington and Fermi Lab decided to test these chips in the freezer and write a new instruction manual specifically for 77 K.

1. The "Freezer Test" (Experimental Setup)

They took a chip containing 170 different types of tiny transistors (the switches that make up the computer brain) and put it in a special vacuum chamber cooled by liquid nitrogen. They measured how much electricity flowed through them at different voltages.

The Analogy: Imagine testing a car engine at 70°F and then again at -320°F. You measure how fast the engine spins, how much fuel it burns, and how hard it is to turn the key.

2. The "Physics of Freezing" (What Happens to the Chips?)

When things get super cold, the rules of physics change in funny ways:

• The Threshold Voltage (The "Key"): At room temperature, a transistor needs a little push (voltage) to turn on. At 77 K, it gets "stiff" and needs a much bigger push to start working. The scientists had to update the manual to say, "Hey, you need to push harder to start this engine."
• Mobility (The "Speed"): Usually, cold makes things sluggish. But for electrons in these chips, cold actually makes them faster because the atoms they bump into are vibrating less. It's like running on a frozen lake vs. a muddy field; you slide much faster on the ice. The manual needed to reflect this new speed.
• Resistance (The "Traffic"): Some parts of the chip get harder for electricity to pass through when cold, while others get easier. It's like some roads getting icy (slippery but fast) while others get blocked by snowdrifts.

3. The "New Manual" (Modeling)

Using a powerful computer tool called Mystic, the team adjusted the numbers in the instruction manual (the BSIM4 model) to match their freezer test results.

Instead of rewriting the whole manual from scratch, they used a clever trick: they created "multipliers."

• Old Manual: "The engine speed is 100 mph."
• New Manual: "The engine speed is 100 mph multiplied by 1.5 (because it's cold)."

This way, they kept the original logic of the chip but just tweaked the numbers to fit the cold environment. They created 18 different versions of this manual to cover all the different sizes of transistors they tested.

---

The Results: Did it Work?

Yes, and it's pretty good!

• Accuracy: The new "Freezer Manual" predicted the chip's behavior with about 20% accuracy. In the world of complex physics simulations, this is a huge success. It's close enough that engineers can now design circuits that will actually work in the real world.
• No Surprises: The model worked consistently whether the electricity was flowing a little or a lot. It didn't break down when the voltage changed.
• Open Source: The best part? They didn't hide the manual. They put the new "Freezer Instruction Book" on GitHub (a website for sharing code) for everyone to use for free.

---

Why Does This Matter? (The "So What?")

1. Democratizing Science: Before this, only big labs with millions of dollars could figure out how to make chips work in the cold. Now, any university or small startup can download these free models and design their own detectors for particle physics, astronomy, or quantum computing.
2. Better Detectors: By putting the electronics inside the cold liquid (instead of outside with long wires), we get much clearer signals and less noise. It's like listening to a whisper in a quiet room vs. trying to hear it through a noisy hallway.
3. Future Proofing: This is the first step. Now that we have a map for 77 K, we can start building better, more powerful tools to explore the universe.

The Bottom Line

This paper is like taking a car that was only designed for summer, testing it in a blizzard, and then publishing a free, updated owner's manual so anyone can drive it in the snow. It opens the door for cheaper, better, and more accessible scientific discoveries.

Technical Summary

1. Problem Statement

High-energy physics (HEP) experiments, particularly those utilizing Liquid Argon Time Projection Chambers (LArTPCs), require low-power, reliable CMOS electronics that can operate directly inside cryostats at liquid nitrogen temperatures (77 K) or liquid argon temperatures (77–89 K). Placing readout ASICs inside the cryostat reduces noise, minimizes cable length, and simplifies system integration.

However, a significant barrier exists: standard CMOS Process Design Kits (PDKs) are typically validated only at room temperature (300 K). While some research exists for cryogenic operation at 4 K, the intermediate temperature range of 77 K remains largely unexplored for open-source technologies. Existing models (room temperature or 4 K) fail to accurately predict transistor behavior at 77 K due to the non-linear, temperature-dependent nature of key device parameters (e.g., threshold voltage, mobility, and subthreshold swing). Furthermore, proprietary PDKs limit accessibility for the broader HEP community.

2. Methodology

The authors developed a systematic approach to characterize and model the SkyWater 130nm (SKY130) open-source process at 77 K using the BSIM4 compact model framework.

• Device Characterization:

  • Fabrication: Devices were fabricated via the Efabless Open Multi-Project Wafer (MPW) program. The study included 120 nMOS and 50 pMOS transistors with varying geometries (Gate Length: 0.15–100 $\mu$m; Width: 0.42–100 $\mu$m).
  • Measurement Setup: Experiments were conducted at Fermi National Accelerator Laboratory (FNAL) using a single-stage cryocooler to reach 77 K. The setup utilized Keithley 237 Source-Measure Units (SMUs) for DC I-V measurements in a vacuum chamber (2 mTorr).
  • Data Collection: Output ($I_D$-$V_D$) and transfer ($I_D$-$V_G$) characteristics were measured for 22 representative transistors at both 300 K and 77 K.

• Modeling Strategy:

  • Isothermal Approach: Instead of a temperature-dependent model, the authors created an isothermal BSIM4 model specifically for 77 K. The TNOM parameter was changed from 300 K to 77 K.
  • Parameter Extraction: Using Synopsys' Mystic™ optimizer, the team extracted specific BSIM4 parameters to fit the 77 K data. They employed a "nominal parameter" strategy, where multiplicative factors were applied to original BSIM4 parameters to preserve the PDK's underlying dynamics (e.g., mismatch and process corners) while scaling values for cryogenic operation.
  • Key Parameters Extracted:
    • $V_{TH0}$: Threshold voltage (capturing the shift due to incomplete ionization and bandgap widening).
    • $U_0$: Low-field mobility (accounting for reduced phonon scattering).
    • $NFACTOR$: Subthreshold swing factor.
    • $VSAT$: Saturation velocity.
    • $RDSW$: Source/drain series resistance (critical for Lightly Doped Drain regions which may freeze out).
    • $ETA_0$ & $DELTA$: Drain-Induced Barrier Lowering (DIBL) and linear-to-saturation smoothing parameters.
  • Binning: Due to the SKY130 PDK's length/width binning, the team generated 18 distinct cryogenic models covering the tested geometry bins.

3. Key Contributions

• First Systematic 77 K Characterization: This work presents the first systematic DC characterization and modeling of the SKY130 130nm node specifically at 77 K.
• Open-Source Cryogenic PDK: The authors made the extracted models and measurement data publicly available on GitHub, democratizing access to accurate cryogenic design tools for the HEP community.
• Novel Extraction Methodology: The use of multiplicative "nominal parameters" allows the new cryogenic models to retain the statistical variations and process corner behaviors of the original foundry-verified PDK, which is often lost in direct parameter re-extraction.
• Comprehensive Dataset: The paper provides a detailed analysis of how specific physical phenomena (mobility, threshold voltage, series resistance) manifest in SKY130 devices at 77 K compared to 300 K.

4. Results

• Device Behavior:
  • Threshold Voltage ($V_{TH}$): Increased significantly at 77 K (e.g., scaling factors of 1.06 to 4.85 depending on the bin) due to Fermi potential shifts and incomplete dopant ionization.
  • Mobility ($\mu$): Generally increased at 77 K due to reduced phonon scattering, though the effect varied by geometry.
  • Series Resistance ($R_{SD}$): Showed complex behavior; while heavily doped regions decreased in resistance, Lightly Doped Drain (LDD) regions exhibited freeze-out, leading to increased resistance in some bins.
  • Subthreshold Swing: Improved (sharper turn-on) at 77 K, consistent with thermal scaling, though non-idealities were observed.
• Model Accuracy:
  • The developed 77 K models showed excellent agreement with measured data.
  • Error Metrics: The average Relative Root-Mean-Square (RRMS) error was on the order of 20%.
  • Bias Independence: The model accuracy showed no strong dependence on drain voltage bias, indicating robust performance across operating regions.
  • Comparison: The 77 K models significantly outperformed room-temperature models when simulating 77 K data, particularly in the subthreshold and saturation regions.

5. Significance

This work is a critical enabler for the next generation of High-Energy Physics experiments. By providing a validated, open-source, 77 K-compatible BSIM4 model for the SKY130 node, the authors:

1. Lower Barriers to Entry: Enable researchers to design mixed-signal ASICs for cryogenic environments without relying on expensive proprietary foundry data.
2. Facilitate LArTPC Upgrades: Allow for the direct integration of readout electronics inside liquid argon detectors, reducing noise and improving signal fidelity.
3. Establish a Baseline: Create a foundation for future work, including noise modeling, capacitance extraction, and temperature-dependent modeling across a wider range (e.g., 4 K to 150 K).

In summary, this paper successfully bridges the gap between open-source semiconductor technology and cryogenic high-energy physics applications, providing a reliable design toolkit for 77 K operation.