Voltage Noise Thermometry in Integrated Circuits at Millikelvin Temperatures

G. Ridgard*, M. Thompson, L. Schreckenberg, N. Deshpande, A. Cabrera-Galicia, O. Bourgeois, V. Doebele, J. Prance · 2025

Read Voltage Noise Thermometry in Integrated Circuits at Millikelvin Temperatures by G. Ridgard*, M. Thompson, L. Schreckenberg, N. Deshpande, A. Cabrera-Galicia, O. Bourgeois, V. Doebele, J. Prance (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2502-16661.txt.

Introduction

A team of eight authors showed in 2025 how to measure temperature inside a computer chip at near-absolute-zero conditions, using the chip's own voltage noise. Knowing the local temperature inside cryogenic chips matters for quantum computing, where heat at the boundary between classical and quantum parts can spoil qubit performance.

Their method uses cross-correlation to read thermal noise without adding heat. It matched the refrigerator temperature across a wide range, and it agreed with an independent in-chip method and a simple thermal model as power varied. Because the test device is a simple resistor, many such sensors could be placed across a chip to map its thermal profile with little added complexity.

Catalog record

CREATORG. Ridgard*, M. Thompson, L. Schreckenberg, N. Deshpande, A. Cabrera-Galicia, O. Bourgeois, V. Doebele, J. Prance
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE30 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2502-16661

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Voltage Noise Thermometry in Integrated Circuits at Millikelvin Temperatures Preprint: AIP/123-QED G. Ridgard*, M. Thompson, L. Schreckenberg, N. Deshpande, A. Cabrera-Galicia, O. Bourgeois, V. Doebele, J. Prance Abstract This paper demonstrates the use of voltage noise thermometry, with a cross-correlation technique, as a dissipation-free method of thermometry inside a CMOS integrated circuit (IC). We show that this technique exhibits broad agreement with the refrigerator temperature range from 300 mK to 8 K. Furthermore, it shows substantial agreement with both an independent in-IC thermometry technique and a simple thermal model as a function of power dissipation inside the IC. As the device under test (DUT) is a resistor, it is feasible to extend this technique by placing many resistors in an IC to monitor the local temperatures, without increasing IC design complexity. This could lead to better understanding of the thermal profile of ICs at cryogenic temperatures. This has its greatest potential application in quantum computing, where the temperature at the cold classical-quantum boundary must be carefully controlled to maintain qubit performance. I Introduction At deep cryogenic temperatures (< 4 K), power dissipation in electronic components can induce significant localised heating Blagg, Castagnède, and Singh 2022 . The primary cause is the reduction in the thermal conductivity of both silicon Glassbrenner and Slack 1964 and the insulating materials in its packaging Chawner et al. 2019 , coupled with the increase of the thermal boundary resistances Swartz and Pohl 1989 . This issue is particularly important in Quantum Information Processing (QIP). The optimal performance of many types of qubit Brandl et al. 2016 ; Córcoles et al. 2011 ; Veldhorst et al. 2017 occurs at deep cryogenic temperatures due to diminished thermal noise. These qubits require ancillary classical electronics for control and read-out. As the qubit count scales, so does the requisite number of wires. To mitigate the unsustainable proliferation of wires into the cryostat, additional classical electronics are now co-located with the quantum components Gonzalez-Zalba et al. 2021 ; Pauka et al. 2021 . Consequently, monitoring the local temperature at the qubit-classical electronics interface becomes crucial for maintaining consistent qubit operation. While thermometry at deep cryogenic temperatures has been studied for many years, some existing solutions are inadequate for determining the temperature within an integrated circuit (IC). The most prevalent thermometry technique, temperature coefficient of resistance (TCR) thermometry Courts 2014 ; Courts and Krause 2008 ; Dechaumphai and Chen 2014 , is unsuitable because the thermometer will be poorly thermalised to the IC interior due to the low thermal conductivity of silicon and the thermal boundary resistance between the IC and the thermometer. Thermometry methods have been developed for use within ICs to address this issue Huizinga et al. 2022 ; Noah et al. 2024 . Methods such as Gate Resistance Thermometry (GRT) Triantopoulos et al. 2019 ; Artanov et al. 2022 , quantum dot thermometry Chawner et al. 2021 ; de Kruijf et al. 2024 ; Singh et al. 2024 , and diode thermometry Choi et al. 2021 ; Bebitov et al. 2023 each have their individual merits and specific applications, but all have associated power dissipation inside the IC. Consequently, all will influence the internal temperature of the IC. One such technique which is both dissipationless and primary is noise thermometry Qu et al. 2019 . This has been well established in a broad range of environments Casey et al. 2014 ; Ivashchenko, Korobchenko, and Bondarenko 1975 ; Brixy and Kakuta 1996 . However, it has several barriers to entry as a technique for measuring the internal temperature of an IC at temperatures that QIP platforms are located. The first problem is that in the case of voltage noise thermometry the device under test (DUT) resistor must have a very high resistance to make the noise measurable above the typical readout amplifier noise. High resistance values result in low cut off frequencies, making measurements extremely slow and susceptible to 1/f noise. This paper demonstrates the use of the cross-correlation technique to remedy this issue. Cross-correlation allows the use of lower value resistors by allowing the measurement of noise values below the noise of the amplifiers used in the setup. In this paper we demonstrate this by using cross-correlation to accurately measure the temperature of our DUT, a \mathrm{45~kOhm} resistor, from 300 mK to 8 K. Another problem in performing noise thermometry in an IC is the potential for non-thermal noise contributions. The resistors are typically made of doped silicon, which as a semiconductor has many more potential sources of noise Wilamowski and Irwin 2018 and the close proximity of power dissipating devices (on the order of \mathrm{100~\mu m} ) may provide a source of cross-talk in the spectral measurements. In this paper, we resolve this problem by showing that even if the resistors have non-thermal noise sources, if this is appropriately characterised then its contribution can be subtracted from the spectral measurements, enabling accurate noise thermometry. We demonstrate this by first showing that our DUT has an elevated noise level at the fridge base temperature, which cannot be attributed to thermal noise. Next, from the base temperature measurement, we extract an offset noise (the non-thermal contribution). We then calibrate the noise temperature against the refrigerator temperature. It was found that when the offset noise is subtracted from the spectral measurements, the noise thermometry shows good agreement with the temperature of the refrigerator from 300 mK to 8 K. This was further tested by comparing the noise temperature against a different in-IC thermometry technique (TCR of the resistor) as a function of power. In Sec II, we present our measurement setup and protocol of voltage noise thermometry inside an IC. Next, in Sec III, we examine the noise characteristics of the resistor selected for the experiment and its calibration against the temperature of the Mixing Chamber (MXC) in a refrigerator. Finally, we compare this technique to another internal thermometry method and a thermal model in Sec IV. II Set up Figure 1 illustrates the IC configuration. The IC, a 2 mm x 2 mm x 0.8 mm chip, is affixed to a silicon interposer, mounted on a ultra-high-purity copper mount, which in turn is secured to the puck connected to the MXC. The device is interconnected via gold bond wires to gold traces on the interposer, and subsequently bonded to the printed circuit board (PCB). The IC contains a \mathrm{350~\Omega} power resistor, R_{P} , and a \mathrm{45~k\Omega} measurement resistor, R_{M} . These components are fabricated using an industrial 65 nm bulk CMOS process. The entire assembly is cooled by an Oxford Instruments Triton 400 dilution refridgerator. Figure 1: Illustration of the setup: PCB board composed of standard FR4. The GE varnish was lightly diluted to aid its application. The copper mount was constructed from a 6 mm thick piece of C110 oxygen free copper which has a minimum of 99.99~\% purity. The silicon interposer was 550\mathrm{\mu m} thick and was lightly boron doped with gold patterned on top to allow for bonding between IC, interposer and PCB. Two Niobium Nitride (NbN) based thermometers were placed on the copper and silicon interposers to monitor their temperature as a function of power. Details on their operation and calibration can be found in

Read the complete text