GLOBAL KNOWLEDGE ARK

Materials Science · Radiation, vacuum and reactive environments

Vacuum, outgassing and surface change

Follow volatile material from bulk and surface release to transport and retention on a sensitive component. Distinguish pressure, release rate, collected mass and functional change using bounded diffusion and surface-residence models, then interpret outgassing records with their actual denominators and limits.

Useful background: Durability in real environments; Electrical transport and electronic materials; Heat transport, expansion and thermal response. Each worked case explains the quantities it uses.

Back to the chapter

After reading, you can

  • Explain how vacuum changes thermal paths and released-species transport without imposing one specimen temperature.
  • Distinguish surface desorption, bulk diffusion, evaporation and continuing supply, and test a diffusion-limited release model against its boundary and inventory limits.
  • Separate release, arrival, initial sticking and later desorption in a sensitive-surface mass balance.
  • Calculate TML/CVCM-style quantities on the correct mass bases and connect deposited mass to an explicitly modelled or measured functional endpoint.
01

Removing gas changes several pathways

In vacuum, ordinary gas convection is suppressed, while radiation and solid conduction remain. Materials can release adsorbed or volatile species through outgassing. Those species can contaminate colder or sensitive surfaces elsewhere.

A warm polymer component and a colder optical window can share a low-pressure enclosure while having very different temperatures and surface histories. Heat can still move through their supports and wiring and by radiation. The word vacuum supplies neither a fixed temperature nor a guarantee of cooling: absorbed power, solid connections, surface optical properties and the surrounding radiation field still matter.

At finite pressure, residual gas can still carry heat. As gas density falls, ordinary bulk convection is suppressed; gas-mediated heat exchange also depends on molecular collisions and surface accommodation. A perfect-vacuum approximation removes gas heat transfer, not radiation or solid conduction. The source temperature governing release and the receiver temperature governing residence are therefore separate inputs.

When the molecular mean free path is large compared with the relevant dimensions, molecule–molecule collisions are rare between wall encounters. This molecular-flow limit makes surface interactions and geometry central. It is not enough to call a system high vacuum without checking its length scale and gas conditions. A molecule may reach a sensitive surface, rebound, remain temporarily or later be released again.

02

Surface release and bulk replenishment are different sources

Adsorption places molecules at a surface; absorption or dissolution places them within the material. Desorption releases adsorbed molecules. Bulk diffusion can replenish the surface from a dissolved inventory. Evaporation or sublimation removes a volatile constituent, while chemical decomposition can generate new gaseous products. Gas entering through permeation or a leak is another supply route and should not automatically be assigned to the specimen’s stored inventory.

The slowest relevant step controls a simple release model. A readily depleted surface population can behave differently from a thick material whose bulk continues feeding the surface. Heating can change desorption, diffusivity and chemical stability at once. Preparation, curing, storage atmosphere and earlier temperature exposure determine what is initially present; there is no universal outgassing curve associated with a material name alone.

Keep units attached to the reported quantity. Mass released per time, molecules released per time and gas throughput in Pa·m³/s are related only after the species and gas-temperature interpretation are supplied. A pressure is a different quantity again. Partial pressures describe different gaseous species; a total-pressure decrease can reflect altered removal or redistribution rather than disappearance of all surface contamination.

03

Derive release from a stated concentration field

Consider an original model of a wide slab with only one face exposed, area A and thickness L. Let C(x,t) be the mass of a single releasable volatile per bulk volume, in kg/m³, at depth x measured into the solid. This is not the solid’s total density. Initially C = C₀ for x > 0. For t > 0 impose C(0,t) = 0, representing immediate removal at the exposed surface. Take constant diffusivity D, no internal reaction, no readsorption, and negligible edge or back-face effects.

With Fickian transport, ∂C/∂t = D∂²C/∂x². The semi-infinite solution is C(x,t) = C₀ erf[x/(2√(Dt))], where erf(u) = (2/√π)∫ from 0 to u of exp(−v²) dv. The concentration rises from zero at the exposed face towards C₀ deeper inside. Its gradient at x = 0 is C₀/√(πDt). Because x points inward, outward mass flux is j_out = D(∂C/∂x)|₀ = C₀√[D/(πt)], in kg/(m²·s). The inward-coordinate Fick flux has the opposite sign.

Integrating the outward flux gives released mass m_rel(t) = 2AC₀√(Dt/π). Thus a fourfold time increase doubles cumulative release but halves the instantaneous release rate, under these assumptions. The diffusion length √(Dt) indicates how deeply the initial inventory is being depleted. For a finite slab the semi-infinite approximation requires that length to be small compared with L; it cannot supply unbounded mass from a finite inventory.

The ideal instantaneous boundary change also predicts a singular flux as t approaches zero. This is a boundary idealisation, not a real infinite initial outgassing rate. Finite surface kinetics and earlier concentration profiles matter at early times. At long times, finite thickness matters. A prior release or heating history changes C(x,0), so restarting the clock does not restore the original uniform inventory.

In the illustrative case A = 0.0100 m², L = 2.00 mm, C₀ = 1.00 kg/m³ and D = 10⁻¹² m²/s, the initial releasable inventory is AC₀L = 20.0 mg. At 1 h the model releases 0.677 mg, and at 4 h 1.35 mg. The diffusion lengths are 0.0600 and 0.120 mm, respectively, versus the 2.00 mm thickness. These are invented material inputs, not data for a named polymer or a recommended treatment.

Volatile depletion profile and declining release rateOriginal semi-infinite one-face diffusion model. C is volatile mass per bulk volume, not total solid density or surface coverage. The vertical axis C/C₀ is dimensionless. D = 10⁻¹² square metres per second, C₀ = 1 kilogram per cubic metre, exposed area 0.0100 square metre, slab thickness 2.00 millimetres. Solid blue is the concentration profile after 1 hour; dashed violet after 4 hours. Both have C = 0 at the exposed face and approach C₀ inside. The plotted depth range is only the first 0.400 millimetre; the model mass integral extends deeper. At 1 hour the diffusion length is 0.0600 millimetre, released mass 0.677028 milligram and instantaneous release rate 0.338514 milligram per hour. At 4 hours these are 0.120 millimetre, 1.35406 milligrams and 0.169257 milligram per hour. Deeper depletion accompanies more cumulative release even as the current rate falls. No source material data or measured profile is plotted.Bulk depletion grows while release rate fallsOriginal one-face model · C(0,t) = 0 · constant D1 h4 hVolatile concentration C/C₀0.00.000.10.250.20.500.30.750.41.00Depth into solid x (mm)1 h: released 0.677 mg; current rate 0.339 mg/h4 h: released 1.35 mg; current rate 0.169 mg/hOnly the first 0.4 mm of the 2.00 mm slab is shown.Finite thickness and early surface kinetics limit extrapolation.
Original semi-infinite one-face diffusion model. C is volatile mass per bulk volume, not total solid density or surface coverage. The vertical axis C/C₀ is dimensionless. D = 10⁻¹² square metres per second, C₀ = 1 kilogram per cubic metre, exposed area 0.0100 square metre, slab thickness 2.00 millimetres. Solid blue is the concentration profile after 1 hour; dashed violet after 4 hours. Both have C = 0 at the exposed face and approach C₀ inside. The plotted depth range is only the first 0.400 millimetre; the model mass integral extends deeper. At 1 hour the diffusion length is 0.0600 millimetre, released mass 0.677028 milligram and instantaneous release rate 0.338514 milligram per hour. At 4 hours these are 0.120 millimetre, 1.35406 milligrams and 0.169257 milligram per hour. Deeper depletion accompanies more cumulative release even as the current rate falls. No source material data or measured profile is plotted.
04

Released mass is not the same as mass remaining on a window

Follow three stages: release from the source, transport to the receiver, and retention there. Geometry changes the fraction that arrives. Initial sticking depends on the species, surface, temperature and coverage; it is not a universal fraction. A molecule that initially sticks can later desorb. Colder surfaces often have longer residence for a given binding process, but neither all molecules stick nor all retained material remains forever.

In a separate original receiver model, let q be a constant mass-release rate of one species and f its supplied fraction reaching a particular window. Let α be the probability of initial sticking, assumed constant over the chosen low-coverage range. For one species it is also the corresponding mass fraction. Define a = αfq, the initially retained arrival rate. Let M be resident mass on that window and k_des a constant first-order desorption coefficient. Assume no chemical conversion, no saturation and no return of desorbed molecules to this tracked surface. Then dM/dt = a − k_des M.

With M(0) = 0, the solution is M(t) = (a/k_des)[1 − exp(−k_des t)]. Initially M ≈ at; later desorption balances arrival at M_* = a/k_des. After the source contribution stops, M decays as M_end exp(−k_des Δt) in this particular model. Irreversible deposits, multilayer condensation, different binding sites or repeated return paths require another description. This constant-q example is not a continuation of the earlier diffusion source, whose release rate decreases.

For a single activated desorption process one often writes k_des = ν exp[−E_des/(k_B T_w)], with surface temperature T_w, prefactor ν and activation energy per molecule E_des. It explains why changing receiver temperature can change residence even if release and transport are unchanged. The equation assumes a defined kinetic regime; changing temperature may also alter sticking, incoming composition or the source itself. The parameters below are supplied teaching inputs.

Arrival, sticking and later residence determine contaminationOriginal separate receiver model with one species. Constant source mass release q = 1.00 microgram per hour acts from 0 to 10 hours, then stops. Supplied arrival fraction f = 0.200 gives 0.200 microgram per hour arriving. Initial sticking α = 0.500 gives a = 0.100 microgram per hour entering the resident population. Desorption removes resident mass at k_des M, with no returning flux, chemical change or saturation. The plotted mass M is what remains on the receiver, not cumulative released or initially stuck mass. Solid blue uses k_des = 0.0100 per hour and dashed violet 1.00 per hour. At 10 hours M is 0.951626 or 0.0999955 microgram, respectively, despite equal released and arriving amounts. At 20 hours after ten hours without source arrival, M is 0.861067 or 0.00000453979 microgram. The source/arrival/sticking arrows are a supplied flow accounting, not a device geometry or a claim of universal capture efficiency. Invented model coefficients, no measured data.Release, arrival and resident mass are distinctOriginal receiver model · one species · M(0) = 0Source releaseq = 1 μg/hArrivalsfq = 0.20 μg/hResident Ma = 0.10 μg/hf = 0.20α = 0.50k_des M leavesNo return, saturation or chemistry in this limited balance.k_des = 0.0100 h⁻¹k_des = 1.00 h⁻¹Resident mass M (μg)00.0050.25100.50150.75201.00Source stopsElapsed time (h)At 10 h: the same 10 μg was released in both cases.What remains differs because residence differs.Model curves; no universal temperature or capture rule.
Original separate receiver model with one species. Constant source mass release q = 1.00 microgram per hour acts from 0 to 10 hours, then stops. Supplied arrival fraction f = 0.200 gives 0.200 microgram per hour arriving. Initial sticking α = 0.500 gives a = 0.100 microgram per hour entering the resident population. Desorption removes resident mass at k_des M, with no returning flux, chemical change or saturation. The plotted mass M is what remains on the receiver, not cumulative released or initially stuck mass. Solid blue uses k_des = 0.0100 per hour and dashed violet 1.00 per hour. At 10 hours M is 0.951626 or 0.0999955 microgram, respectively, despite equal released and arriving amounts. At 20 hours after ten hours without source arrival, M is 0.861067 or 0.00000453979 microgram. The source/arrival/sticking arrows are a supplied flow accounting, not a device geometry or a claim of universal capture efficiency. Invented model coefficients, no measured data.
05

Assess the whole configuration

Temperature, preparation, time and pressure influence release. A low bulk mass loss does not necessarily establish acceptable contamination for an optical surface. Use test conditions and collected-condensable information appropriate to the intended system.

In the mass definitions used by the NASA outgassing guide, total mass loss is TML = 100(m₀ − m_after)/m₀ percent. Collected volatile condensable material is CVCM = 100Δm_col/m₀ percent, where m₀ is the specimen’s starting mass on the stated preparation basis and Δm_col is collector mass gain after appropriate background accounting. CVCM is not normalised to the collector mass or to the specimen’s lost mass.

The collector samples what reaches and remains on it under that test’s geometry, temperature and duration. It need not collect all released species. A separate water-vapour-regained measurement concerns uptake after the vacuum exposure; it is not the same as total lost mass or collected mass. The example table uses the mass definitions only and is not a claim that an ASTM procedure or a current acceptance standard has been followed.

A database result belongs to a tested formulation, preparation and history. Different lots, cure states or storage can change the inventory and kinetics. A screening percentage does not alone give a species-resolved release rate, the arrival fraction at a new window, residence there, or acceptable optical performance. A suitability conclusion needs the application’s endpoint and the transfer assumptions.

Invented mass records using TML/CVCM definitions. These are not database material values or a conforming standard test.
CandidateStarting specimen massSpecimen mass lossCollector gainTMLCVCM
A2.000 g10.0 mg0.400 mg0.500%0.0200%
B2.000 g4.00 mg0.800 mg0.200%0.0400%
06

Translate a deposit into the property being assessed

For an assumed uniform, compact deposited film, mean thickness is h_f = M/(ρ_f A_w), where ρ_f is film density and A_w the coated area. This is a geometrical conversion, not an optical model. Patches, porosity, droplets or a nonuniform temperature field can invalidate the uniform-film interpretation even when total mass is known.

To predict transmission, the optical response must also be specified. An absorption-only teaching model might supply T/T₀ = exp(−βh_f), with β an absorption coefficient at the chosen wavelength. It suppresses changes in reflection, interference and scattering. A real thin film can require those effects and wavelength-dependent optical constants; a mass or thickness alone does not determine transmission.

The original optical-window question can therefore be made quantitative only by adding transport, retained-mass and optical information. Measuring release, measuring collected mass and measuring optical change answer different parts of the argument. A reference surface, background measurements, defined wavelengths and before/after comparisons can help distinguish the sample’s effect from a changing instrument or chamber contamination.

07

State what the evidence can carry to a new configuration

A release record should identify the specimen and exposed area, preparation/history, temperature and time, measured gas or collected species, background and the conversion used for the reported rate. A collector or optical witness adds its location, temperature, surface and observation interval. A total-pressure trace by itself does not provide all of these quantities.

Distinguish a model check from a material approval. A t⁻¹ᐟ² trend over a bounded interval is consistent with the diffusion model, but is not unique proof of that mechanism. A declining chamber pressure can coexist with accumulation on a colder surface. A small collector gain under one set of conditions does not establish small gain on a different colder, larger or more exposed receiver.

For transfer, ask which assumption changes. Changing thickness challenges the semi-infinite diffusion window; changing preparation changes the initial inventory; changing orientation changes transport; changing receiver temperature changes retention; changing wavelength changes the optical endpoint. A reasoned conclusion follows those links and leaves the unmeasured steps explicit.

Worked examples

Check a declining rate against growing cumulative release

Use the one-face diffusion model and its supplied A, L, C₀ and D. Calculate released mass and instantaneous mass rate at 1 h and 4 h. Is the lower rate at 4 h evidence that no more material has been released?

  1. Convert times to 3600 s and 14400 s. At 1 h, √(Dt) = 6.00 × 10⁻⁵ m = 0.0600 mm; at 4 h it is twice as large. These are 0.0300L and 0.0600L, respectively.
  2. m_rel(1 h) = 2 × 0.0100 × 1.00 × √(10⁻¹² × 3600/π) = 6.77028 × 10⁻⁷ kg = 0.677028 mg. At 4 h, cumulative release doubles to 1.35406 mg.
  3. The instantaneous rate is A j_out. Converted to mg/h, it is 0.338514 at 1 h and 0.169257 at 4 h. Equivalently dm_rel/dt = m_rel/(2t), using consistent time units.
  4. Both released amounts remain below the initial 20.0 mg inventory. The declining rate measures current release, not the amount accumulated earlier. The model’s short diffusion lengths and stated boundary assumptions are part of the calculation.

Cumulative release grows from 0.677 to 1.35 mg while the instantaneous rate halves. A declining rate is not zero release or a return of the earlier released mass.

Compare two residence times with the same incoming supply

In the separate receiver model, q = 1.00 μg/h, f = 0.200 and α = 0.500. Compare k_des = 0.0100 h⁻¹ and 1.00 h⁻¹ at t = 10.0 h, starting with clean receivers. Find resident mass and account for initially stuck mass that has desorbed.

  1. a = αfq = 0.100 μg/h. Over 10.0 h the source releases 10.0 μg, 2.00 μg reaches the receiver and 1.00 μg initially sticks. These are three different quantities.
  2. For k_des = 0.0100 h⁻¹, M = (0.100/0.0100)(1 − exp(−0.100)) = 0.951626 μg. The difference 1.00 − 0.951626 = 0.0483742 μg has left the tracked resident population.
  3. For k_des = 1.00 h⁻¹, M = 0.100(1 − exp(−10)) = 0.0999955 μg. About 0.900005 μg of the initially stuck amount has desorbed. The same released mass does not imply the same retained mass.
  4. The steady masses would be 10.0 and 0.100 μg, but the longer-residence case is far from its steady value at 10 h. A future plateau or continued first-order behaviour is a model prediction, not a measured material capacity.

At 10 h the two resident masses are about 0.952 and 0.100 μg. Arrival, initial sticking, residence and any later removal must be distinguished.

Interpret a lower TML without inventing lower collection

Use the two invented mass records. Which candidate has lower TML? Which has lower CVCM? For candidate A, what fraction of lost mass is collected, and is that its CVCM percentage?

  1. The starting mass is 2.000 g = 2000 mg in both cases. TML_A = 100 × 10.0/2000 = 0.500%; TML_B = 100 × 4.00/2000 = 0.200%. B loses less of its starting mass.
  2. CVCM_A = 100 × 0.400/2000 = 0.0200%; CVCM_B = 100 × 0.800/2000 = 0.0400%. A has the smaller collector gain on the stated starting-mass basis.
  3. For A, the fraction of lost mass collected is 0.400/10.0 = 0.0400, or 4.00%. That denominator is lost mass, so it is not CVCM. The two percentages answer different questions.
  4. Neither candidate is thereby qualified for a particular optical system. The test collector and the application receiver can differ, and the relevant spectral or other functional endpoint remains unspecified.

B has lower TML, while A has lower CVCM. A’s collected/lost ratio is 4.00%, distinct from its 0.0200% CVCM.

Complete the optical-window argument with stated transfer assumptions

A material releases little total mass but a fraction deposits on an optical window. Can total mass loss alone establish optical suitability?

  1. No: total mass loss describes the source, not the application window’s deposited state or optical response. Now add a separate invented application model: 10.0 mg is released; 0.200% of that mass reaches and remains on a 10.0 cm² window over the observation, with no later loss. This supplied fraction already combines transport and retention.
  2. The retained deposit is 10.0 mg × 0.00200 = 0.0200 mg = 20.0 μg = 2.00 × 10⁻⁸ kg. If it forms a uniform compact film with ρ_f = 1000 kg/m³, A_w = 10.0 cm² = 0.00100 m² gives h_f = M/(ρ_f A_w) = 2.00 × 10⁻⁸ m = 20.0 nm.
  3. Only if the independently supplied absorption-only model applies, with β = 10⁷ m⁻¹ at the chosen wavelength, is T/T₀ = exp(−0.200) = 0.818731. Its conditional loss is about 18.1%, rather than the 0.500% source mass-loss fraction in candidate A’s record.
  4. The absorption-only calculation omits reflection, interference and scattering and is separate from the low-coverage residence example. A measured film distribution and optical response are needed if those assumptions fail. No universal 20 nm optical limit or test acceptance rule follows.

No. The system’s contamination endpoint matters.

Try the reasoning

  1. Does vacuum eliminate radiative cooling? Explain why a warm source and cold window can remain at different temperatures in the same low-pressure enclosure.

    Show the answer and reasoning

    No. Radiation and solid conduction remain, and the components can have different absorbed powers and thermal connections.

    Suppressed convection does not impose a common temperature. Residual-gas heat exchange also depends on conditions; the perfect-vacuum limit removes that route, not every thermal pathway.

  2. Why report preparation and bake history? If a diffusion-limited specimen has already lost near-surface inventory, may a later exposure simply restart the original uniform-C₀ solution at t = 0?

    Show the answer and reasoning

    No. Preparation and prior exposure change inventory and surface state; a new calculation must use the actual remaining concentration profile.

    The fresh-slab solution assumes uniform initial C₀ and an instantaneous surface boundary. Neither is automatically restored by pausing a test, changing temperature or resetting a clock.

  3. For a new, initially uniform slab in the same semi-infinite regime, suppose only D increases by a factor of nine. At the same time, how do j_out, cumulative release and diffusion length change? What validity condition must be rechecked?

    Show the answer and reasoning

    All three increase by a factor of three. The diffusion length must still be small compared with slab thickness.

    Each is proportional to √D at fixed C₀ and t in this model. This is not a general claim that every outgassing rate scales as √D after an arbitrary earlier history.

  4. Extrapolating the original √t mass law to 40000 h predicts about 135 mg released. The specimen initially held only 20.0 mg. Is reducing the predicted number by a rounding correction enough?

    Show the answer and reasoning

    No. The model has been used beyond its finite-thickness/inventory regime.

    √(Dt) is 12.0 mm, six times the 2.00 mm thickness, so the semi-infinite assumption failed long before this extrapolation. A finite-body model with the remaining inventory and boundary kinetics is needed.

  5. For a fictional single desorption process with E_des = 0.600 eV and unchanged ν, compare k_des at 330 K and 300 K. Use k_B = 8.61733 × 10⁻⁵ eV/K. If a alone stays fixed and both states reach the linear model’s steady state, what happens to M_*?

    Show the answer and reasoning

    k_des increases by about 8.25, so the conditional steady resident mass falls by the same factor.

    k(330)/k(300) = exp{(E_des/k_B)[1/300 − 1/330]} ≈ 8.25. M_* = a/k_des. This conclusion requires unchanged incoming supply, sticking and kinetic regime; heating a real system can change those too.

  6. A different invented specimen starts at 0.500 g, loses 1.50 mg and produces a background-corrected collector gain of 0.0500 mg. Calculate TML, CVCM and collected/lost percentage. Which one predicts optical transmission without more information?

    Show the answer and reasoning

    TML = 0.300%, CVCM = 0.0100%, and collected/lost = 3.33%. None alone predicts optical transmission.

    Use 500 mg for the TML/CVCM denominator: 100 × 1.50/500 and 100 × 0.0500/500. Collected/lost instead divides 0.0500 by 1.50. A new receiver still needs transport, residence, film and wavelength-specific response evidence.

  7. Two configurations have the same source release q and the same α and k_des. One has twice the supplied arrival fraction f and half the coated window area. Within the constant-source resident-mass model and uniform-film assumption, compare M_* and h_f. Does the same total chamber pressure prevent this difference?

    Show the answer and reasoning

    Steady resident mass doubles and mean thickness quadruples. Equal total pressure does not rule it out.

    M_* = αfq/k_des scales with f, and h_f = M/(ρ_f A_w) adds the inverse area factor. Total pressure does not specify the gas composition, local arrival fraction or receiver retention. The comparison assumes the changed surface remains in the model’s valid regime.

What to carry forward

  • Vacuum changes thermal and molecular pathways; it does not specify a temperature or remove all surface interactions.
  • Identify the release mechanism, initial inventory and history before extrapolating an outgassing curve.
  • Transport, initial sticking and residence separate released mass from resident contamination.
  • TML and CVCM use the specimen’s starting mass; collected/lost fraction and optical response are different quantities.
  • Transfer evidence through the actual geometry, temperatures, duration, surface state and functional endpoint.

References

Teaching resources and research supporting this lesson’s explanations and examples.

  1. Outgassing Database User GuideNASA GoddardDatabase guide · selected definitions and limitsetd.gsfc.nasa.govRead source
  2. Outgassing properties of vacuum materials for particle acceleratorsPaolo Chiggiato · CERNCERN Accelerator School, 2017 · 2020 author archiveOutgassing properties of vacuum materials for particle acceleratorsAuthor archive recordarxiv.org
  3. Vincent Baglin · CERN2 sources
    1. Cryopumping and Vacuum SystemsVincent Baglin · CERNCERN Accelerator School · 2020 author archiveCryopumping and Vacuum SystemsAuthor archive recordarxiv.org
    2. Vacuum in Accelerators for Mechanical & Materials EngineeringVincent Baglin · CERNCERN Accelerator School, 2024 · 2026 author archiveVacuum in Accelerators for Mechanical & Materials EngineeringAuthor archive recordarxiv.org
  4. 7.0 Thermal ControlNASA Small Spacecraft Systems Virtual InstituteThermal-control guide · selected heat pathsnasa.govRead source
  5. A Multi-Chamber System for Analyzing the Outgassing, Deposition, and Associated Optical Degradation Properties of Materials in a VacuumJack Singal, Rafe Schindler, Chihway Chang, Patrick Czodrowski and Peter KimAuthor paper · archive version 2, 2010A Multi-Chamber System for Analyzing the Outgassing, Deposition, and Associated Optical Degradation Properties of Materials in a VacuumAuthor archive recordarxiv.org