Vacuum Leak and Outgassing Diagnosis

R01 decision: Separate leakage process load vapour release and measurement error.

Required asset: Vacuum load diagnostic decision tree.

Depth: Rate-of-rise behaviour; isolated volumes; temperature; vapour; seals; gauge response.

Worked comparison: Distinguish a leak from outgassing using controlled evidence.

Release gate: No single pressure-rise result proves the leak mechanism.

1 Direct answer and purpose

A rising vacuum pressure is an observation, not a diagnosis. Verify the gauge, temperature and isolated boundary before calculating a gas load. Then compare repeat tests, controlled changes and independent evidence to separate external leakage, surface gas release, trapped gas, permeation, process admission and instrument effects. A pressure-rise result alone cannot establish the mechanism or location.

It includes a Vacuum load diagnostic decision tree, an editable evidence record, five bounded numerical cases and an acceptance procedure. Every worked value is hypothetical. The document establishes no universal leak limit, pump rating, conditioning temperature, test duration or detector sensitivity.

The intended outcome is a defensible disposition: confirmed mechanism with supporting evidence, a bounded unresolved load requiring further investigation, or invalid data requiring a corrected test. Their boundaries remain distinct even when one investigation uses all three.

2 Define the question before starting a test

Write the actual symptom: failure to reach a chamber target, excessive pressure increase after isolation, slow recovery after loading, contamination during operation or an unexpected gauge signal. Record when the symptom began and what changed. A new seal, cleaning process, product, line connection or gauge replacement can be more informative than a general description of poor vacuum.

Define the decision the test must support. Locating an external leak, quantifying total gas load and proving process readiness are different tasks. A diagnostic trend may justify further investigation without proving a contractual leak-rate requirement. Conversely, a leak test can pass while the installed process still exceeds its pressure target because it releases vapour.

Agree the pressure range, operating state, allowed materials and test method with the responsible project team. Keep test criteria separate from equipment protection limits. Never substitute a detector's display resolution for an application acceptance criterion.

3 Establish the isolated boundary

Draw the complete test volume from the isolation valve to the chamber, including gauge heads, connecting pipes, instrument branches, fixtures and product displacement. Record which valves are open, closed or independently verified. A controller command does not demonstrate seat tightness. A valve can leak from a pressurised branch into the test volume even while an external leak test finds nothing.

Determine the actual volume inside that boundary. Include spaces communicating during the selected interval; exclude disconnected regions. Narrow connections may equalise slowly, so a geometrical volume sum is not always a valid single-volume dynamic model. Retain the drawing, method and uncertainty. If communicating volumes are not understood, interpret the trace as a network response until validated.

Record the pump's condition after isolation. A pump left connected through another route, a cold surface, a getter or a sorbing product can remove gas and mask an incoming load. The pressure-rise calculation measures net accumulation under its assumptions; it does not automatically measure total ingress.

4 Record the operating history

Outgassing depends on what happened before the test. Document exposure to ambient air, humidity, wet cleaning, solvents, lubricant, process materials and time under vacuum. Record the product and fixture condition, previous conditioning and the time since the last vent. A chamber tested after extended pumping cannot be compared directly with one tested just after opening without acknowledging the difference.

Keep matched initial conditions for comparison tests: pressure, gas, temperature, pump duration, valve sequence, volume and gauge setup. If changing a factor is the experiment, state it explicitly. Otherwise a difference in slope may reflect the changed preparation rather than the component being investigated.

Do not invent a standard bake temperature or drying interval. Conditioning must be compatible with the actual chamber, seals, instruments, process material and supplier restrictions. Record the approved procedure and the achieved condition rather than assuming the command setting describes every surface.

5 Separate the candidate mechanisms

These mechanisms can coexist. A declining load after conditioning does not prove that a smaller external leak is absent. A successful seal repair does not remove surface moisture. Keep a multi-cause explanation available until the evidence resolves the residual load.

6 Pressure and gas load definitions

Use absolute pressure p at an identified location. State units and measurement range. A vacuum gauge percentage requires a defined ambient reference and is not a direct substitute for absolute pressure in gas-law calculations. Temperature T is absolute temperature in kelvin.

For a fixed volume V, approximately uniform and constant gas temperature, and a sufficiently isolated system with negligible gas removal, apparent throughput is q_app = V dp/dt. Over an interval, q_app = V(p₂ − p₁)/(t₂ − t₁). If V is litres and pressure is mbar, q is mbar L/s. One mbar L/s equals 0.1 Pa m³/s. P1 supports the pressure-rise method and the need to separate gas liberation from leakage before assigning a leak rate.

Call this apparent or net gas load until the mechanism is established. The mathematical result does not identify whether gas crossed a seal, left a wet surface or entered through a process valve. Preserve this naming in spreadsheets, charts, website extracts and acceptance reports.

7 Temperature and changing volume

From the ideal-gas relation pV = nRT, the net molar accumulation expressed as local pressure-volume throughput is q_net = RT dn/dt = V dp/dt + p dV/dt − (pV/T)dT/dt. For fixed volume, the second term is zero. This separates apparent pressure change from thermal expansion under the stated model.

The correction requires a representative gas temperature. A single outer wall sensor may not capture nonuniform gas or product temperature. Thermal gradients, changing vapour composition and gauge temperature sensitivity can invalidate a simple correction. Stabilise the experiment where possible and state the remaining uncertainty.

Flexible walls, moving fixtures, diaphragms or thermal volume changes can also affect pressure. If these effects are material, use an appropriate volume model or redesign the test. Do not report an uncorrected slope as leakage merely because temperature data are unavailable.

8 Gauge response and measurement error

Select a gauge whose principle, range and gas response suit the test. Retain calibration status, zero checks, contamination history and correction factors. Some pressure principles depend strongly on gas composition; a changing mixture can change the indicated pressure differently from the true total pressure. P2 identifies gas dependence as an interpretation issue in pressure-rise measurements.

Record sensor location and connection geometry. A long or restricted gauge tube can delay response; a gauge at the pump inlet can differ from chamber pressure while flow continues. Compare simultaneous readings rather than readings taken in different operating states.

Check timestamp alignment, sampling, filtering, automatic range changes and data gaps. Wait for the declared valve and instrument settling behaviour before choosing a fit window. Retain the excluded interval and reason rather than silently deleting an inconvenient transient. Cross-check with an appropriate independent instrument when evidence points to a measurement problem.

9 Reading the rate of rise trace

Plot absolute pressure against elapsed time, with valve events and temperature on the same timeline. Compute slopes over declared intervals or fit a model with residual checks. A straight line over a short interval may represent a nearly constant net load from several sources. Curvature may arise from depletion, changing temperature, redistribution, sorption or a pressure-dependent leak.

A declining contribution after repeated evacuation is consistent with conditioning-sensitive gas release. It is not, by itself, proof of outgassing. P3 describes the distinction between outgassing and leakage contributions; the decision tree below requires corroboration.

Do not use curve shape as a universal fingerprint. The correct claim is often that the data are consistent with a hypothesis and identify the next controlled test. Report what remains unresolved, including a residual source below the test's discrimination capability.

10 Seals joints and trapped spaces

Identify seal material, installation orientation, gland condition, mating surfaces and exposure history. A mechanically similar seal does not establish equivalent vacuum behaviour. Permeation, outgassing and a damaged joint are different mechanisms and can require different corrective actions.

Record approved inspection and reassembly evidence. A seal disturbed during diagnosis changes the experiment; compare the retest with matched preparation where practicable. Do not assume that tightening, applying a coating or substituting a material is an acceptable repair without equipment-specific authority.

Blind holes, double seals, porous materials and trapped pockets can release gas through restricted paths. Investigate geometry when a long transient does not fit an external-leak explanation. Enlarging a communicating path or venting a pocket is a design change that needs review, not a generic field instruction.

11 Vapour and process gas

A wet surface or volatile product can dominate pressure without a hole in the vessel. Record material identity, temperature, recent cleaning and potential liquid reservoirs. Cooling may lower observed pressure through condensation while leaving the source present. A pressure plateau can reflect gas release and removal processes rather than a sealed and clean system.

Use appropriate composition or process evidence where available. Residual-gas analysis may help, but overlapping fragments, instrument backgrounds and gas-specific sensitivities require expert interpretation. A single mass peak does not establish a unique contaminant or leak mechanism.

Check deliberate gas paths, purge, analyser flows and pneumatic interfaces against the isolation drawing. Process control logs should establish actual admission timing. Repeat a controlled blank or empty-fixture test only when it represents a meaningful comparison. Product-dependent load must remain part of process acceptance even if an empty chamber performs well.

12 Selective isolation and independent leak evidence

Divide the system into testable sections where the installed design and approved procedure permit. Preserve comparable thermal and preparation conditions. A reduced slope after isolating a branch assigns a net contribution to that boundary; it does not automatically distinguish a seal leak from a wet component within it.

For localisation, use an appropriate validated leak-detection method with documented calibration, background, response time, tracer conditions and equipment compatibility. Observe whether a repeatable signal follows the targeted exposure and returns to background. Account for transport delays and contamination of nearby regions. P4 supports tracer-gas localisation as a separate method from an integral pressure-rise test.

Do not equate a helium response directly to operating-air ingress without the required flow-regime, gas and pressure basis. A non-detection must be reported with the applicable threshold and test coverage, not as proof of zero leakage. Avoid unapproved tracer gases, pressures or conditioning methods; use the actual instrument and project procedure.

13 Design the diagnostic experiment

Write the hypothesis, expected observation, controlled variables and alternative explanations before changing the system. Test one planned difference at a time where practical. Keep an evidence record that links each result to configuration, instruments, raw files and the inference it supports.

Include a baseline, a controlled intervention and a repeat. Where conditions cannot be matched, identify the difference and bound the conclusion. A before-and-after repair comparison is stronger when accompanied by localisation evidence and an unchanged test boundary than when it is only two screenshots.

Define stopping and escalation conditions through the process owner. Do not continue a test outside the equipment envelope simply to obtain a longer trace. Unresolved gas composition, incompatible media or unexplained pressure excursions require the responsible specialist's review rather than an invented acceptance limit.

14 Required asset Vacuum load diagnostic decision tree

Use the nodes in order. Each outcome records a next action or a supported disposition; no branch labels a mechanism solely from one curve. The node table is the accessible text decision tree and can be implemented as an interactive questionnaire without changing its evidence conditions.

The decision tree does not force one exclusive label. For example, D6 can support a surface-related component while D8 independently identifies a remaining external leak. D10 tests the result of intervention; D11 separates a repaired defect from acceptable operation.

15 Editable test identity and preparation record

Complete one record per test configuration. Use Measured, Confirmed, Calculated, Assumed or Unknown for evidence status. Preserve source revision, owner and disposition. These are project input fields; they are not recommended settings.

16 Editable hypothesis and disposition record

17 Worked case A Distinguish leakage and outgassing

All values in Sections 17–21 are hypothetical training data. They are neither OEM specifications nor customer results. Case A uses a verified 300 L isolated volume, stable temperature, appropriate calibrated absolute-pressure instrumentation and no deliberate process admission. Tests begin at 0.010 mbar and each declared fit interval is 600 s. Pump and gauge settling periods are excluded under the fictional test plan.

First inference: the decrease from 0.006 to 0.002 mbar L/s supports a conditioning-sensitive component. The difference is 0.004 mbar L/s under this comparison. It is not yet proof of outgassing because preparation might also change temperature, seal behaviour or sorption. Here, fictional repeated tests confirm matched temperature and valve state, while the preparation log identifies recent wet cleaning and a controlled drying response. Those observations support surface or vapour release as a contributing mechanism.

Independent evidence: a calibrated tracer test produces a repeatable local response at one joint, followed by background recovery. A blank-location check remains at background. The test record includes instrument calibration, transport delay and applicable sensitivity; it establishes a local leak pathway rather than assigning an operating-air rate directly from tracer response.

Intervention and result: an approved repair is followed by a matched retest, reducing apparent load from 0.002 to 0.0005 mbar L/s. The prior tracer response is no longer detected under the documented method. Together, localisation, repair and repeat evidence support an external-leak contribution. The 0.0015 mbar L/s difference is a comparative net-load change, not automatically a calibrated air leak rate.

Disposition: mixed causes are supported. Conditioning reduced one component; independent evidence identified another. The remaining 0.0005 mbar L/s must be compared with the actual process requirement and test uncertainty. Neither a lower slope nor a tracer non-detection proves zero leakage. This case deliberately cannot be diagnosed from its first pressure-rise result alone.

18 Worked case B A thermal rise without added gas

Assumptions: a rigid 200 L volume contains an ideal gas with no net addition or removal. It warms uniformly from T₁ = 293 K to T₂ = 296 K over 600 s. Initial pressure is 0.500 mbar absolute. Gauge temperature effects are assumed absent.

Method: conservation of gas gives p₂ = p₁T₂/T₁ = 0.505119 mbar. An uncorrected pressure-rise calculation gives q_app = 200(0.505119 − 0.500)/600 = 0.001706 mbar L/s. It would be wrong to call this a leak rate.

For a reference temperature T_ref = 293 K, the interval gas accumulation is q_ref = VT_ref[(p₂/T₂) − (p₁/T₁)]/Δt. Substitution gives zero within arithmetic precision. This is consistent with the stipulated no-addition model.

Interpretation: pressure increased because temperature increased. A real vessel may have thermal gradients, vapour release and instrument drift simultaneously. The case shows the correction's purpose; it does not demonstrate that one wall sensor can correct every thermal transient or that all warming tests contain no leaks.

19 Worked case C Incorrect connected volume

Assumptions: a hypothetical verified steady pressure slope is 2.0 × 10⁻⁵ mbar/s. Temperature is stable and gas removal is negligible. The chamber nameplate volume is 500 L, but open pipework, gauge branches and fixtures add a net 120 L within the actual boundary.

Method: using 500 L gives q_app = 500 × 2.0 × 10⁻⁵ = 0.0100 mbar L/s. Using the reconciled 620 L gives 0.0124 mbar L/s. The corrected value is 24 percent larger than the original calculation; the original understates the corrected result by about 19.4 percent.

Interpretation: both calculations use the same pressure trace. The error is the boundary definition, not arithmetic or a changed leak. The 620 L model still requires pressure equalisation across communicating sections within the fit interval. If a restricted pocket equalises slowly, use a suitable dynamic treatment instead of simply adding its full volume.

20 Worked case D Uncertainty near acceptance

Assumptions: V = 300 L with 3 percent relative standard uncertainty; p₁ = 0.010 mbar and p₂ = 0.014 mbar, each with independent standard uncertainty 0.0002 mbar; interval 600 s with 0.2 s standard uncertainty. Temperature and boundary validity are assumed established. The endpoint-error independence is a fictional simplification; correlated offsets require different propagation.

Method: q_app = 0.002 mbar L/s. The uncertainty in pressure difference is sqrt(0.0002² + 0.0002²) = 0.0002828 mbar. Combining independent relative terms gives u(q)/q = sqrt(0.03² + (0.0002828/0.004)² + (0.2/600)²) ≈ 0.0768, or 7.68 percent. Thus u(q) ≈ 0.0001536 mbar L/s.

An illustrative coverage factor k = 2 gives expanded uncertainty about 0.0003073 mbar L/s, subject to the actual uncertainty model and coverage assumptions. Suppose the fictional project limit is 0.0021 mbar L/s and its agreed rule requires measured value plus expanded uncertainty not to exceed the limit. The upper value is about 0.0023073, so the result cannot be accepted under that rule.

Disposition: hold or improve the evidence; do not declare a pass solely because 0.002 is numerically below 0.0021. The limit, coverage factor and decision rule are examples, not prescribed settings. Also, this test quantifies apparent load; uncertainty analysis alone does not prove an external-leak mechanism.

21 Worked case E Conflicting gauge readings

Assumptions: two hypothetical gauges measure a demonstrated common, well-connected 250 L volume over 300 s. Gauge A rises from 0.020 to 0.026 mbar; reference gauge B rises from 0.0200 to 0.0206 mbar. Temperature and timing are matched for comparison. The gases and gauge response are still under investigation.

Method: Gauge A implies 250 × 0.006/300 = 0.005 mbar L/s; Gauge B implies 250 × 0.0006/300 = 0.0005 mbar L/s. The tenfold difference is too consequential to select whichever result is preferred.

Interpretation: hold the mechanism diagnosis while checking calibration, gas sensitivity, contamination, range switching, logger scaling and response. Confirm that the supposedly common pressure is actually common at both ports. If corrected independent evidence validates B, record the reason A was invalid and repeat the baseline. This fictional discrepancy does not identify contamination, gas change or any other specific gauge fault by itself.

22 Commissioning and acceptance sequence

Before testing, approve the boundary drawing, operating method, product condition, gauge arrangement, preparation, evidence requirements and acceptance rule. Distinguish total-load acceptance from external-leak acceptance. A criterion expressed in one test gas under one pressure differential must not silently become a different operating-gas requirement.

Record baseline and repeat tests with raw data. Verify any relevant valve leakage and adjacent pressure conditions. Demonstrate that thermal, volume and instrument effects are bounded. Compare like configurations or state precisely why they differ. If the test is intended to prove process readiness, include representative product, fixtures and process load rather than relying exclusively on an empty vessel.

Close the diagnostic tree with evidence, not a selected label. An external-leak conclusion requires independent support appropriate to the method and scope. A surface-release conclusion requires a controlled response with competing explanations considered. An unresolved residual load can be reported quantitatively without assigning an unsupported mechanism.

After intervention, repeat the agreed tests and confirm the installed process requirement. Record the release owner, date, operating conditions, measurement limits and open actions.

23 Failure modes in diagnosis

24 Frequently asked questions

Does a linear pressure rise prove an external leak

No. A near-constant net load can arise from several sources over the selected interval. Verify instruments, isolation, temperature and gas accounting, then obtain independent evidence before identifying a leak mechanism.

Does a declining slope prove outgassing

No. A decline is consistent with depletion or conditioning-sensitive gas release, but redistribution, thermal effects and other changing conditions may contribute. Repeat controlled tests and record the alternatives considered.

Why is the result different after venting

Venting can change surface exposure, moisture, temperature and preparation time. Retain that history and match initial conditions before treating results as comparable. The test should represent the intended process condition.

Can the empty chamber pass while the process fails

Yes. Product, fixtures and deliberate gas admission can add loads or change thermal behaviour. Empty-system diagnosis helps isolate contributions, but process acceptance needs the relevant installed operating state.

Can a pressure rise locate the leak

An integral rise does not identify location. Selective isolation can narrow the boundary, and an appropriate independent localisation method can supply stronger evidence. Each method must retain its calibration, coverage and detection limits.

What should be sent for an engineering review

Provide the symptom, boundary drawing, connected volume, gas and product state, preparation history, raw pressure and temperature traces, gauge details, valve sequence and any independent leak-test record. Unknown inputs should remain visibly unknown.