Cryogenic and Vacuum Infrastructure

Cryogenic and vacuum infrastructure controls the environment seen by a quantum device. It removes heat and residual gas, but its wires, windows, supports, pumps, and shields also transmit noise, light, and vibration. The temperature at a refrigerator plate need not equal the temperature inferred from a qubit; the pressure at a remote gauge need not describe collisions at a trap.

This page connects heat and gas-flow models to those local observables. Vacuum Technology introduces pumps, gauges, mean free paths, and chamber practice; the discussion here concentrates on their consequences for quantum devices and arrays.

Required background. Hardware Overview introduces device operations; Control, Readout, and Calibration explains signal paths; Thermal Density Operators provides equilibrium populations.

Thermal occupation and staged refrigeration

A mode population is not a plate temperature

For a bosonic mode of angular frequency $\omega$ in equilibrium at temperature $T$,

$$\bar n_{\mathrm{th}} = \frac{1} {\exp\!\left(\hbar\omega/k_BT\right)-1}.$$

At microwave frequencies, kelvin-scale radiation contains many photons. A $5\,\mathrm{GHz}$ mode has $\hbar\omega/k_B\approx0.240\,\mathrm K$: its equilibrium occupation is about $16$ at $4\,\mathrm K$, about $0.10$ at $100\,\mathrm{mK}$, and about $6\times10^{-6}$ at $20\,\mathrm{mK}$. For a two-level system with splitting $\hbar\omega$, the equilibrium excited population is instead

$$p_e=\frac{1}{e^{\hbar\omega/(k_BT)}+1} =\frac{\bar n_{\rm th}}{2\bar n_{\rm th}+1}.$$

The harmonic mode can contain many excitations, whereas $p_e\le1/2$ at positive temperature. The two quantities agree approximately only in the low-temperature limit. Millikelvin operation suppresses thermal population, but the relevant energy gap may be a qubit splitting, a spin or valley gap, or a reservoir broadening scale rather than the same frequency for every device.

An equilibrium population law does not establish that a device has equilibrated. A mode coupled weakly to a cold package but strongly to a warmer control line can have an effective occupation far above the value inferred from the mixing-chamber thermometer. Conversely, active reset can prepare a qubit colder than a simple equilibrium model would suggest. Population thermometry, sideband asymmetry, resonator dephasing, or another device-relevant observable is needed to infer the actual state.

For an oscillator weakly coupled to independent Markov thermal reservoirs, with number-decay rates $\kappa_j$ and occupations $n_j$ at its frequency,

$$\begin{aligned} \frac{d\langle n\rangle}{dt} &=-\sum_j\kappa_j(\langle n\rangle-n_j), \\ n_{\rm ss}&=\frac{\sum_j\kappa_j n_j}{\sum_j\kappa_j}. \end{aligned}$$

A weak hot connection can dominate a very cold reservoir’s occupation. Assigning an effective temperature to one transition does not prove a common temperature for the entire device. Thermal and Vacuum Noise distinguishes thermal excitations from vacuum fluctuations.

Cooling power is a function, not a label

A modern dry dilution refrigerator commonly uses a pulse-tube cooler to provide tens-of-kelvin and few-kelvin stages, then a circulating $^3\mathrm{He}$–$^4\mathrm{He}$ mixture to reach millikelvin temperatures. Typical named stages include:

Approximate stage	Representative infrastructure role
$300\,\mathrm K$	room-temperature control, pumps, compressors, laser and data systems
$30$–$80\,\mathrm K$	first radiation shield, initial cable interception
$3$–$5\,\mathrm K$	second shield, cryogenic amplifiers, circulators, cryopumps, superconducting magnets
$0.5$–$1\,\mathrm K$	still or intermediate heat interception
$50$–$200\,\mathrm{mK}$	cold plate, further line thermalization and filtering
$5$–$30\,\mathrm{mK}$	mixing chamber, quantum processor, cold attenuators and filters

These are representative functions, not universal temperature bands. The installed options, helium circulation, magnetic field, orientation, cabling, optical load, and upstream stage temperatures change the load curve.

At temperatures well below $0.1\,\mathrm K$, the usual low-temperature enthalpy approximation for a dilution unit gives (Pobell)

$$\dot Q_{\mathrm{mix}} = \dot n_3 \left( 95T_{\mathrm{mix}}^2 - 11T_{\mathrm{in}}^2 \right),$$

with temperatures in kelvin, molar $^3\mathrm{He}$ flow $\dot n_3$ in $\mathrm{mol\,s^{-1}}$, and the numerical coefficients in $\mathrm{J\,mol^{-1}\,K^{-2}}$. If the incoming concentrated phase is well precooled so that $T_{\mathrm{in}}\approx T_{\mathrm{mix}}$, the familiar low-temperature estimate is

$$\dot Q_{\mathrm{mix}} \approx 84\,\dot n_3 T_{\mathrm{mix}}^2.$$

Real performance also depends on heat exchangers, boundary resistance, circulation stability, parasitic heat leaks, and the thermal resistance between the mixing chamber and device. A quoted base temperature is usually measured near zero added load. Scaling must use the full load curve and an engineering margin.

Heat balance and thermal response

At an actual steady temperature, extracted heat equals the total incoming load. For a specified maximum acceptable stage temperature, compare that load with the available heat lift at that temperature:

$$\begin{aligned} \dot Q_{i,\mathrm{load}} &= \dot Q_{i,\mathrm{cond}} +\dot Q_{i,\mathrm{rad}} +\dot Q_{i,\mathrm{diss}} +\dot Q_{i,\mathrm{opt}} +\dot Q_{i,\mathrm{other}} .\\ \dot Q_{i,\mathrm{load}} &<\dot Q_{i,\mathrm{avail}}(T_{i,\max}). \end{aligned}$$

The margin should remain positive during simultaneous operation, not only at idle. Do not subtract a parasitic load twice if the quoted available power already excludes it. Fast pulse sequences, readout amplifiers, heaters, switches, magnetic coils, optical beams, mechanical motion, and regeneration cycles can create bursty or slowly varying loads. A stage can satisfy its average budget while exceeding a transient temperature or recovery-time requirement.

For a small temperature excursion $\delta T$ of a well-mixed device with heat capacity $C_T$ and effective differential thermal conductance $G_T>0$,

$$\begin{aligned} C_T\frac{d\delta T}{dt}+G_T\delta T&=\delta P(t), \\ \tau_T&=\frac{C_T}{G_T}. \end{aligned}$$

Here $G_T$ includes the slope of net heat extraction about a stable operating point. A short rectangular pulse of extra power $\delta P$ and duration $t_p$ raises the temperature by

$$\delta T_{\rm peak}=\frac{\delta P}{G_T} \left(1-e^{-t_p/\tau_T}\right).$$

For $t_p\ll\tau_T$, this is deposited energy divided by heat capacity. Repeated pulses can accumulate heat before recovery. A stage thermometer with slow response can miss a faster local excursion.

Heat paths

Conduction through supports and wiring

For steady one-dimensional conduction along a uniform link of length $L$ and area $A$, with no heat generation along it and endpoint temperatures $T_c<T_h$,

$$\dot Q_{\mathrm{cond}} = \frac{A}{L} \int_{T_c}^{T_h} k(T)\,dT,$$

where $k(T)$ is the material’s thermal conductivity. Contact thermal resistances require additional temperature drops. At cryogenic temperatures, room-temperature material tables can be badly misleading: purity, alloying, work hardening, superconducting transitions, plating, and geometry matter.

Low-conductivity alloys reduce passive heat leaks but can add electrical loss. High-conductivity copper provides strong thermal anchoring but also transports heat efficiently if it bypasses an intermediate stage. Superconducting conductors suppress electrical resistance below their transition but do not eliminate phonon heat conduction, connector losses, or heat carried by other materials in a cable.

Thermal interception turns one large temperature span into several smaller spans. Each cable, braid, support, waveguide, fibre, capillary, and shaft should be included with its actual thermal contacts. Shields, dielectrics, connectors, DC wiring, and mechanical supports can conduct heat even when they do not carry the intended signal.

Radiation

For two surfaces represented by an effective emissivity $\epsilon_{\mathrm{eff}}$ and area $A$, a useful first estimate is

$$\dot Q_{\mathrm{rad}} = \epsilon_{\mathrm{eff}}\sigma A \left(T_h^4-T_c^4\right).$$

Radiation shields intercept room-temperature power at stages with much larger cooling capacity before it reaches millikelvin components. Geometry, view factors, apertures, surface finish, multilayer insulation, seams, cable ports, and optical windows determine the actual load. A direct line of sight can defeat an otherwise excellent shield.

The same principle applies spectrally. Microwave filtering does not guarantee infrared shielding. Infrared photons can break Cooper pairs in superconductors, while room-temperature blackbody radiation can drive atomic or molecular transitions. A credible system specifies the relevant spectral band and path.

Active dissipation

Control power is intentionally dissipated in attenuators, resistive wiring, filters, switches, terminations, amplifiers, heaters, and electronics. For a purely resistive electrical element,

$$P_{\mathrm{Joule}} = I_{\mathrm{rms}}^2R = \frac{V_{\mathrm{rms}}^2}{R}.$$

This elementary equation becomes a systems problem because the dissipation must be assigned to a thermal stage and duty cycle. A pulse attenuated by $20\,\mathrm{dB}$ deposits most of its incident power as heat at that matched attenuator, apart from negligible reflected power in this model. For fixed power $P_{\rm device}$ delivered immediately after an attenuator of transmission $\eta$, its absorbed signal power is

$$P_{\rm absorbed}=P_{\rm device}(\eta^{-1}-1).$$

Moving attenuation colder can reduce incoming thermal noise but consumes scarcer cold-stage cooling power. Account separately for cable conduction, signal absorption, and noise power over the relevant bandwidth.

Optical absorption deserves the same accounting. A fibre may conduct little heat yet deliver microwatts or milliwatts of optical power to a millikelvin absorber. Scattered trapping light can heat a vacuum enclosure or charge a dielectric. Optical readout, photonic links, and transducers must report absorbed power and recovery as well as delivered photons.

Signal Lines Are Thermal Reservoirs

Attenuation thermalizes only when the attenuator is thermalized

A matched, linear passive attenuator with power transmission $\eta$ and an independent thermal bath at $T_a$ transforms occupation at a given frequency as

$$\bar n_{\mathrm{out}} = \eta\bar n_{\mathrm{in}} + (1-\eta)\bar n_{\mathrm{th}}(T_a).$$

Cascading stages gives

$$\begin{aligned} \bar n_N &= \eta_{1:N}\bar n_0 + \sum_{i=1}^{N} (1-\eta_i)\bar n_{\mathrm{th}}(T_i) \eta_{i+1:N},\\ \eta_{a:b} &\equiv \prod_{j=a}^{b}\eta_j, \qquad \eta_{N+1:N}\equiv1. \end{aligned}$$

For fixed total attenuation, allocating more to a well-thermalized colder stage reduces the contribution from warmer reservoirs, but it also dissipates control power at the coldest stage. Distributed attenuation is therefore a compromise among thermal occupation, pulse amplitude, bandwidth, cooling power, and component reliability.

A metal attenuator bolted to a plate is not automatically at the plate’s effective noise temperature. Contact resistance, connector heating, dielectric loss, insufficient dwell time, and poorly thermalized internal resistive films can matter. In situ noise or qubit thermometry tests the complete chain.

Signal paths have different thermal requirements

Line type	Primary function	Infrastructure tension
microwave drive	coherent fast control	low thermal noise versus cold attenuation and pulse power
flux or gate-bias line	DC and low-frequency control	filtering versus rise time; conductor heat leak versus resistance
readout output	carry weak signals outward	reverse noise, gain, isolation, amplifier dissipation
trap RF and electrode lines	confinement and shuttling	high voltage or current, filtering, dissipation, vibration and pickup
optical fibre or free-space beam	control, readout, networking	optical absorption, blackbody path, charging, alignment and vibration
thermometer and heater wiring	measure and stabilize temperature	self-heating, calibration, thermal lag and added wiring

Input and output chains are asymmetric. Readout often needs a near-quantum- limited parametric amplifier, isolators or circulators, and a higher-power amplifier at a warmer stage. Each component has finite reverse isolation, insertion loss, saturation power, bandwidth, magnetic sensitivity, and heat load. Interconnects and Transduction explains how loss and added photons change the transmitted state. Krinner and colleagues analyze complete cryogenic line configurations; Yeh and Wang and their collaborators examine attenuator thermalization.

Multiplexing and cryogenic electronics move costs

Frequency, time, code, or switch multiplexing can reduce the number of lines crossing the cryostat. Cryogenic CMOS, superconducting digital logic, local microwave sources, and optical links can move control functions closer to the device. None removes cost; they exchange:

• passive cable conduction for local active power;
• room-temperature bandwidth for cryogenic switching and calibration;
• line count for crosstalk, insertion loss, and failure modes;
• latency for serialization or local processing;
• accessible serviceable electronics for less accessible cold electronics.

The right question is not “How many wires per qubit?” in isolation. It is whether the power, bandwidth, noise, and timing remain compatible with the intended operations when all lines are active.

Shielding and relative motion

Different frequencies enter through different paths. Conductive enclosures and filtered feedthroughs can suppress radio-frequency pickup while leaving infrared leakage through seams or windows. Infrared photons above a superconductor’s pair-breaking threshold can generate quasiparticles; Barends and colleagues study this mechanism. An absorber that is poorly thermalized can itself become a source of radiation.

Static magnetic field, gradients, drift, and field-noise spectra affect different operations. Superconducting shields may trap flux during cooldown; high-permeability shields can saturate. Spin devices that require a finite bias field and nearby superconducting circuits therefore need a compatible field geometry rather than a single nominal shielding factor.

Vibration matters through the relative position of a device and its phase reference. For an effective wave vector $\mathbf k_{\rm eff}$,

$$\delta\phi(t)=\mathbf k_{\rm eff}\cdot\delta\mathbf x(t).$$

Two objects moving together can have large laboratory-frame displacements and small relative phase noise. Conversely, a narrow vibration line near a trap frequency can cause heating even when the integrated displacement is small. Measurement bandwidth and the operation’s frequency response must therefore accompany a root-mean-square displacement. Flexible thermal links can reduce force transmission while increasing thermal resistance; Dubielzig and colleagues describe a trapped-ion implementation.

Ionizing radiation can generate high-energy phonons and nonequilibrium quasiparticles, causing correlated charge or relaxation events across a chip. Stress release can also deposit energy through a non-radiative mechanism. Vepsäläinen and colleagues, Wilen and colleagues, and McEwen and colleagues provide experimental studies of radiation-related errors. Shielding, phonon management, materials, and event-aware control address different parts of this problem; a reduced average local error rate does not by itself exclude rare correlated events.

Residual gas and collision-limited lifetime

For dilute equilibrium gas species $j$,

$$n_j=\frac{p_j}{k_BT_j}, \qquad \Gamma_{\rm coll}=\sum_j n_j\langle\sigma_j(v)v\rangle.$$

The average uses the relative-velocity distribution of the trapped particle and background gas. Total pressure alone does not fix composition, temperature, or cross sections. A directed hot atomic beam need not admit a single equilibrium pressure/temperature description at all.

For loss from a trap of depth $U$, use a loss rate coefficient $K_j(U)$: collisions that transfer too little energy to eject the particle still count as collisions but not as loss. Including other independent loss processes,

$$\Gamma_{\rm loss}=\Gamma_{\rm other}+\sum_jn_jK_j(U).$$

Constant independent rates give $P_{\rm survive}(t)=e^{-\Gamma_{\rm loss}t}$. A deep ion trap may retain an ion after an event that reorders the crystal, changes chemistry, or heats a motional mode. A shallow neutral-atom trap can instead lose the particle. A fitted lifetime is consequently not a universal pressure meter.

For one known species at temperature $T$, after subtracting other loss,

$$p=\frac{k_BT(\Gamma_{\rm loss}-\Gamma_{\rm other})}{K(U)}.$$

Cold-atom vacuum standards use independently calculated or measured loss coefficients and explicit trap-depth corrections (Barker and colleagues). For an unknown mixture, one lifetime generally cannot determine all partial pressures. A remote ionization gauge also has species-dependent sensitivity and can disturb the gas through heating, desorption, or pumping.

For $N$ independent identical particles of lifetime $\tau$, the probability of no losses in time $t$ is

$$P_0(t)=e^{-Nt/\tau}.$$

This can be small even when each particle’s loss probability is small. Correlated bursts, unequal lifetimes, and replacement require a different model. In particular, a single-particle lifetime does not determine the joint distribution of array losses.

Gas transport and effective pumping

Sources and sinks in one volume

In a well-mixed volume $V$ at constant temperature, let $Q_{\rm source}$ be the positive gas throughput from outgassing, leaks, permeation, loading, and desorption. A linear effective pumping speed $S_{\rm eff}$ gives

$$V\frac{dp}{dt}=Q_{\rm source}-S_{\rm eff}p.$$

For constant coefficients and negligible pump base pressure,

$$\begin{aligned} p_\infty&=\frac{Q_{\rm source}}{S_{\rm eff}}, \\ p(t)&=p_\infty+[p(0)-p_\infty]e^{-S_{\rm eff}t/V}. \end{aligned}$$

Throughput has units $\mathrm{Pa\,m^3\,s^{-1}}$; pumping speed has units $\mathrm{m^3\,s^{-1}}$. Also, $1\,\mathrm{mbar\,L\,s^{-1}}=0.1\,\mathrm{Pa\,m^3\,s^{-1}}$. For surface outgassing rate $q$ per unit area, $Q_{\rm out}=qA$. The assembled chamber’s materials, cleaning, bake history, trapped volumes, and temperature affect $q$; Fedchak and colleagues compare measured chamber outgassing. Real pumpdown often departs from one exponential because desorption and pump performance evolve.

Distributed getters can be included as additional species-dependent pumping speeds at their local pressures, or as explicit removal rates in a spatial model. Do not count the same getter both as a negative source and inside $S_{\rm eff}$. With finite gas capacity or nonlinear sorption, its speed also changes in time.

A connecting tube can dominate

In isothermal molecular flow, let a chamber at pressure $p$ connect through conductance $C$ to a pump inlet at pressure $p_p$. Equating throughputs gives

$$Q=C(p-p_p)=S_p p_p.$$

Thus

$$\frac{1}{S_{\rm eff}}=\frac1{S_p}+\frac1C.$$

This single-species formula assumes a negligible pump base pressure and no additional source in the connecting path. If $S_p\gg C$, increasing the pump speed has little effect because $S_{\rm eff}\simeq C$. Conductance depends on geometry, molecular mass, and temperature. Multiple sources or chambers require a gas-flow network, and nonisothermal molecular flow needs temperature corrections.

Cold surfaces do not pump every gas equally

Cryocondensation and cryosorption depend on species, surface temperature, coverage, and sorbent. Water and many heavier gases can be trapped where hydrogen and helium remain difficult to remove. Suitable cold sorbents can improve hydrogen and helium pumping; ordinary chemical getters do not pump helium or other noble gases. Getters also have species-specific capacity and activation requirements. Jousten’s handbook and Baglin’s CERN lectures explain these distinctions.

Cold surfaces can greatly reduce outgassing and local gas density, but they store adsorbed material. Warming changes both desorption and residual-gas composition. An apparent improvement during cooling and a release during warm-up are different measurements of the same surface reservoir.

Warm gauges and thermal transpiration

Even without net pumping, two molecular-flow regions at different temperatures need not have equal pressure. For a small aperture connecting thermalized gas of one species, the one-way molecular flux is $n\bar v/4$, with $\bar v\propto\sqrt T$. Equal opposing particle fluxes therefore imply

$$\begin{aligned} \frac{p_c}{\sqrt{T_c}}=\frac{p_h}{\sqrt{T_h}}, \\ \frac{n_c}{n_h}=\sqrt{\frac{T_h}{T_c}}. \end{aligned}$$

This thermal transpiration limit has lower cold-side pressure but higher cold-side number density. It is not evidence for removal of particles. Net pumping, adsorption, nonthermal directed fluxes, and incomplete thermal accommodation can change the relation. Applying it blindly to a strongly cryopumped chamber is therefore inappropriate; Baglin discusses the regime in which a warm-gauge correction is meaningful.

Optical access, loading, and local diagnostics

Large numerical aperture improves collection and resolution, but it does not follow from short working distance alone: aperture and optical design also matter. Windows and in-vacuum optics affect birefringence, wavefront quality, thermal radiation, charging, and allowable bake temperatures. A warm view through a window can admit blackbody radiation even while chamber walls are cold. Baffles and filters reduce that view while adding optical loss and alignment constraints.

Atom sources, photoionization, ablation, and loading light can add heat, gas, or deposits. Source-off pressure is therefore insufficient to predict source-on loss. A pressure recovery measurement and a local trapped-particle lifetime can distinguish a loading transient from a persistent leak. Likewise, device population measurements can reveal hot signal lines when the cold plate remains stable.

Device	Local quantity to measure	Why a remote specification can miss it
Microwave qubit or cavity	Excited population, mode occupation, relaxation bursts	Warm signal ports and quasiparticles need not equilibrate with the plate
Semiconductor spin device	Electron temperature, reservoir broadening, charge noise	Wiring, dissipation, and filtering can differ from lattice temperature
Trapped ion	Collisions, reordering, reactions, and motional heating	Retained ions can still suffer interruptions or heating
Neutral-atom array	Single-atom loss and correlated array loss during operation	A long single-atom lifetime does not fix whole-array survival
Optical or spin–photon node	Relative phase, linewidth, and collection stability	Cold-head vibration is not the same as device–optics relative motion

Thermometry and gauge readings remain essential, but their uncertainty, location, response time, and operating conditions must enter the inference. Comparing them with local quantum observables is often the most direct way to find a missing heat or gas path.

Exercises

1. Thermal occupation of a microwave mode

A $5.0\,\mathrm{GHz}$ resonator is alternately coupled to baths at $4.0\,\mathrm K$, $100\,\mathrm{mK}$, and $20\,\mathrm{mK}$. Estimate the equilibrium mean photon occupation at each temperature. Use $h\nu/k_B\approx0.240\,\mathrm K$.

Solution

Use

$$\bar n = \frac{1}{e^{0.240/T}-1}.$$

The three values are approximately

$$\begin{array}{c|c} T & \bar n\\ \hline 4.0\,\mathrm K & 16.2\\ 0.100\,\mathrm K & 0.100\\ 0.020\,\mathrm K & 6.2\times10^{-6} \end{array}$$

Cooling from $4\,\mathrm K$ to $100\,\mathrm{mK}$ changes the mode from many-photon thermal occupation to a small but non-negligible occupation. Reaching $20\,\mathrm{mK}$ suppresses equilibrium occupation by another four orders of magnitude. A measured excess would indicate a warmer effective reservoir or non-equilibrium excitation.

2. Conductive heat leak

A support between $4.0\,\mathrm K$ and $0.10\,\mathrm K$ has $A/L=1.0\times10^{-6}\,\mathrm m$ and $k(T)=aT$ with $a=2.0\times10^{-2}\,\mathrm{W\,m^{-1}\,K^{-2}}$. Estimate the conducted heat.

Solution

Integrate the temperature-dependent conductivity:

$$\begin{aligned} \dot Q &= \frac{A}{L} \int_{0.10}^{4.0}aT\,dT\\ &= \frac{A}{L}\frac{a}{2} \left(4.0^2-0.10^2\right). \end{aligned}$$

Numerically,

$$\dot Q \approx 1.60\times10^{-7}\,\mathrm W = 0.160\,\mu\mathrm W.$$

That can be small at a kelvin stage and significant at a millikelvin stage. The calculation also shows why the actual low-temperature conductivity and geometry are needed.

3. Cascaded cryogenic attenuation

A $5\,\mathrm{GHz}$ input mode begins at $300\,\mathrm K$, for which $\bar n_0\approx1250$. It passes through three ideal $20\,\mathrm{dB}$ attenuators at $4\,\mathrm K$, $100\,\mathrm{mK}$, and $20\,\mathrm{mK}$. Each has $\eta=0.01$. Use the occupations from Exercise 1 and estimate the final occupation.

Solution

Apply

$$\bar n_{\mathrm{out}} = 0.01\bar n_{\mathrm{in}} +0.99\bar n_{\mathrm{th}}$$

at each stage. After the $4\,\mathrm K$ attenuator,

$$\bar n_1 \approx 0.01(1250)+0.99(16.2) \approx 28.5.$$

After the $100\,\mathrm{mK}$ attenuator,

$$\bar n_2 \approx 0.01(28.5)+0.99(0.100) \approx 0.384.$$

After the $20\,\mathrm{mK}$ attenuator,

$$\begin{aligned} \bar n_3 &\approx 0.01(0.384)+0.99(6.2\times10^{-6}) \\ &\approx 3.85\times10^{-3}. \end{aligned}$$

Sixty decibels of total attenuation is not enough information by itself; its distribution among temperatures controls the result. Real components add frequency dependence and imperfect thermalization.

4. Radiation-shield staging

Take $\epsilon_{\mathrm{eff}}=0.05$ and $A=0.050\,\mathrm{m^2}$. Estimate the radiative load on a $4\,\mathrm K$ surface if it sees $300\,\mathrm K$ directly. Then estimate the load if a $50\,\mathrm K$ shield blocks that view. Use $\sigma=5.67\times10^{-8}\,\mathrm{W\,m^{-2}\,K^{-4}}$.

Solution

The direct estimate is

$$\begin{aligned} \dot Q_{300\to4} &\approx 0.05(0.050)\sigma \left(300^4-4^4\right) \\ &\approx 1.15\,\mathrm W. \end{aligned}$$

With a $50\,\mathrm K$ shield, the $4\,\mathrm K$ stage sees

$$\begin{aligned} \dot Q_{50\to4} &\approx 0.05(0.050)\sigma \left(50^4-4^4\right) \\ &\approx 0.89\,\mathrm{mW}. \end{aligned}$$

The shield moves most of the load to a warmer stage with greater cooling capacity. Real multilayer geometries require view factors, apertures, and temperature-dependent emissivity.

5. Collision-limited lifetime

Suppose one residual species has pressure $p=1.0\times10^{-10}\,\mathrm{Pa}$ at $300\,\mathrm K$, effective cross section $\sigma=1.0\times10^{-18}\,\mathrm{m^2}$, and mean relative speed $500\,\mathrm{m\,s^{-1}}$. Assume every collision causes loss. Estimate the number density, collision rate, and lifetime.

Solution

The density is

$$\begin{aligned} n &= \frac{p}{k_BT} \\ &\approx \frac{10^{-10}} {(1.38\times10^{-23})(300)} \\ &\approx 2.4\times10^{10}\,\mathrm{m^{-3}}. \end{aligned}$$

Then

$$\Gamma = n\sigma v \approx 1.2\times10^{-5}\,\mathrm{s^{-1}},$$

so

$$\tau = \Gamma^{-1} \approx 8.3\times10^4\,\mathrm s \approx 23\,\mathrm h.$$

The result is model dependent: real cross sections depend on species and speed, and not every collision necessarily has the same observable outcome.

6. Conductance-limited pumping

A pump has nameplate speed $S_p=100\,\mathrm{L\,s^{-1}}$ but connects to the science chamber through a path of molecular-flow conductance $C=10\,\mathrm{L\,s^{-1}}$. Find the effective pumping speed.

Solution

Use

$$\frac{1}{S_{\mathrm{eff}}} = \frac{1}{100} + \frac{1}{10}.$$

Thus

$$S_{\mathrm{eff}} \approx 9.09\,\mathrm{L\,s^{-1}}.$$

The effective speed is close to the conductance, not the pump nameplate speed. Replacing the pump by a much larger one would produce little improvement unless the connecting conductance also changed.

7. Outgassing-limited pressure

A chamber has internal area $A=2000\,\mathrm{cm^2}$ and specific outgassing rate $q=1.0\times10^{-10}\,\mathrm{mbar\,L\,s^{-1}\,cm^{-2}}$. If $S_{\mathrm{eff}}=100\,\mathrm{L\,s^{-1}}$ and all other gas loads are ignored, estimate the steady pressure.

Solution

The gas throughput is

$$Q_{\mathrm{out}} = qA = 2.0\times10^{-7}\,\mathrm{mbar\,L\,s^{-1}}.$$

Therefore

$$p \approx \frac{Q_{\mathrm{out}}}{S_{\mathrm{eff}}} = 2.0\times10^{-9}\,\mathrm{mbar}.$$

The calculation is only a steady-state estimate. Time-dependent water desorption, hydrogen diffusion, source operation, and species-dependent pump speed can change both pressure and composition.

8. Whole-array survival

An array contains $N=6000$ atoms, each with independent vacuum lifetime $\tau=1200\,\mathrm s$. What is the probability that no atom is lost during a $1.0\,\mathrm s$ interval?

Solution

The no-loss probability is

$$P_0 = \exp\!\left(-\frac{Nt}{\tau}\right) = e^{-5} \approx 6.7\times10^{-3}.$$

Although a 20-minute single-atom lifetime sounds long, a 6000-atom array is very unlikely to remain completely full for one second under the independent model. Useful architectures therefore need longer lifetime, shorter loss-sensitive intervals, replacement, erasure handling, or a task that does not require zero loss everywhere.

9. Cooling margin at a specified temperature

At the maximum permitted temperature $20\,\mathrm{mK}$, a stage can extract a total of $20\,\mu\mathrm W$, before subtracting the loads below. Static package and support loads use $7\,\mu\mathrm W$; 120 lines contribute $50\,\mathrm{nW}$ each; cold readout components dissipate $1.5\,\mu\mathrm W$; and control pulses add an average $2.0\,\mu\mathrm W$. Find the remaining margin and its fraction of available power.

Solution

The line load is

$$120(50\,\mathrm{nW}) = 6.0\,\mu\mathrm W.$$

The total load is

$$\dot Q_{\mathrm{load}} = 7+6+1.5+2 = 16.5\,\mu\mathrm W.$$

The margin is

$$20-16.5 = 3.5\,\mu\mathrm W,$$

or $17.5\%$ of the available power. Whether that is adequate depends on load uncertainty, transients, degradation, and the permitted temperature rise.

10. A warm gauge and a cold gas

Two regions containing one noncondensing gas are connected by a small aperture in the molecular-flow limit, with no net particle flow. They are thermalized at $T_h=300\,\mathrm K$ and $T_c=4\,\mathrm K$. Find the pressure and density ratios. Would using $p_h/(k_BT_c)$ infer the correct cold density?

Solution

Equal one-way fluxes give

$$\begin{aligned} \frac{p_c}{p_h}=\sqrt{\frac4{300}}\simeq0.115, \\ \frac{n_c}{n_h}=\sqrt{\frac{300}{4}}\simeq8.66. \end{aligned}$$

Using the warm pressure with the cold temperature would instead predict $n_c/n_h=300/4=75$, overestimating the density by $8.66$. This idealized example contains thermal transpiration without particle removal. Adsorbing cold walls or a finite net flow require a fuller model.

11. A heat pulse shorter than thermal recovery

An effective device has $C_T=2\,\mu\mathrm{J\,K^{-1}}$ and $G_T=0.5\,\mu\mathrm{W\,K^{-1}}$ near its operating point. Starting at equilibrium, it absorbs an extra $0.02\,\mu\mathrm W$ for $0.5\,\mathrm s$. Find the temperature rise at the end of the pulse and two seconds later. Assume constant $C_T,G_T$ over the excursion.

Solution

The thermal time constant is $\tau_T=C_T/G_T=4\,\mathrm s$. During the pulse,

$$\begin{aligned} \delta T_{\rm peak} &=0.04\,\mathrm K\,(1-e^{-0.5/4}) \\ &\simeq4.70\,\mathrm{mK}. \end{aligned}$$

After the pulse the excess decays exponentially, giving

$$\begin{aligned} \delta T(2\,\mathrm s\text{ after pulse}) &=\delta T_{\rm peak}e^{-2/4}\\ &\simeq2.85\,\mathrm{mK}. \end{aligned}$$

Replacing the pulse by its long-term average power would miss its peak and recovery. Large excursions would require temperature-dependent heat capacity and conductance rather than this linearized model.

12. Same single-particle lifetime, different array losses

Each atom independently suffers loss at rate $\gamma$. In addition, a Poisson event of rate $\lambda$ removes the entire array. Find a single atom’s survival and the probability that an initially full $N$-atom array has no losses. Compare with treating the measured single-atom lifetime as entirely independent.

Solution

A tagged atom survives only if neither process occurs:

$$P_1(t)=e^{-(\gamma+\lambda)t}.$$

No array losses requires no independent loss anywhere and no common event:

$$P_0(t)=e^{-(N\gamma+\lambda)t}.$$

Blindly multiplying single-atom survivals gives $e^{-N(\gamma+\lambda)t}$, which is different. For the same marginal lifetime, concentrating losses in common bursts makes loss-free intervals more likely but the affected intervals more destructive. Neither correlation nor the full distribution is determined by the single-atom lifetime alone.

References

• V. Baglin, “Cryopumping and Vacuum Systems,” CERN Accelerator School lectures (2020), arXiv:2006.01574.

• R. Barends et al., “Minimizing Quasiparticle Generation from Stray Infrared Light in Superconducting Quantum Circuits,” Applied Physics Letters 99, 113507 (2011), doi:10.1063/1.3638063.

• D. S. Barker et al., “Precise Quantum Measurement of Vacuum with Cold Atoms,” Review of Scientific Instruments 93, 121101 (2022), doi:10.1063/5.0120500.

• D. S. Barker et al., “Accurate Measurement of the Loss Rate of Cold Atoms Due to Background Gas Collisions for the Quantum-Based Cold Atom Vacuum Standard,” AVS Quantum Science 5, 035001 (2023), doi:10.1116/5.0147686.

• T. Dubielzig et al., “Ultra-Low-Vibration Closed-Cycle Cryogenic Surface-Electrode Ion Trap Apparatus,” Review of Scientific Instruments 92, 043201 (2021), doi:10.1063/5.0024423.

• C. Enss and S. Hunklinger, Low-Temperature Physics (Springer, 2005), doi:10.1007/b137878.

• J. A. Fedchak et al., “Outgassing Rate Comparison of Seven Geometrically Similar Vacuum Chambers of Different Materials and Heat Treatments,” Journal of Vacuum Science and Technology B 39, 024201 (2021), doi:10.1116/6.0000657.

• K. Jousten, ed., Handbook of Vacuum Technology, 2nd ed. (Wiley-VCH, 2016), doi:10.1002/9783527688265.

• S. Krinner et al., “Engineering Cryogenic Setups for 100-Qubit Scale Superconducting Circuit Systems,” EPJ Quantum Technology 6, 2 (2019), doi:10.1140/epjqt/s40507-019-0072-0.

• M. McEwen et al., “Resolving Catastrophic Error Bursts from Cosmic Rays in Large Arrays of Superconducting Qubits,” Nature Physics 18, 107–111 (2022), doi:10.1038/s41567-021-01432-8.

• J. F. O’Hanlon, A User’s Guide to Vacuum Technology, 3rd ed. (Wiley, 2003), doi:10.1002/0471467162.

• F. Pobell, Matter and Methods at Low Temperatures, 3rd ed. (Springer, 2007), doi:10.1007/978-3-540-46360-3.

• S. Simbierowicz et al., “Inherent Thermal-Noise Problem in Addressing Qubits,” PRX Quantum 5, 030302 (2024), doi:10.1103/PRXQuantum.5.030302.

• A. P. Vepsäläinen et al., “Impact of Ionizing Radiation on Superconducting Qubit Coherence,” Nature 584, 551–556 (2020), doi:10.1038/s41586-020-2619-8.

• Z. Wang et al., “Cavity Attenuators for Superconducting Qubits,” Physical Review Applied 11, 014031 (2019), doi:10.1103/PhysRevApplied.11.014031.

• C. D. Wilen et al., “Correlated Charge Noise and Relaxation Errors in Superconducting Qubits,” Nature 594, 369–373 (2021), doi:10.1038/s41586-021-03557-5.

• J.-H. Yeh et al., “Microwave Attenuators for Use with Quantum Devices below 100 mK,” Journal of Applied Physics 121, 224501 (2017), doi:10.1063/1.4984894.