1. Before changing anything
Write a measurable problem: “The same check standard gives less peak area than its recent history under this method.” Specify whether the quantity is area, height, count rate, mass position, retention time, or calculated concentration. Compare like acquisition modes, units, concentrations, and processing versions. A differently scaled plot is not a sensitivity measurement.
Preserve raw files, method versions, sequence order, readiness messages, and the last satisfactory control. Review recent changes in samples, consumables, software, and maintenance. Select compatible blanks and independently prepared controls, then vary one factor at a time. Repeatedly changing several settings can erase the evidence you need.
Safety comes first. If there is a vacuum or gas fault, solvent leak, overheating, arcing, unusual noise, or an interlock warning, pause analysis and follow the instrument's safety procedure. Do not open energized housings, defeat interlocks, adjust high-voltage hardware, or improvise vacuum work. This is a decision guide, not a service procedure.
“Check-tune” and “autotune” may do different things. Establish whether a routine only measures performance or changes settings before using it diagnostically. Keep required calibration and suitability checks in place; this guide does not replace a validated method or set universal acceptance limits.
2. Choose the failed measurement
| Observation | First comparison | Avoid assuming |
|---|---|---|
| Less or no response | Raw sample signal versus a suitable standard | The detector has failed |
| Wrong reported m/z | Several correctly assigned reference ions | One shifted feature proves axis drift |
| Unexpected blank signal | Blanks before and after a known exposure | Every blank peak is carryover |
| Results vary | Repeated measurement versus independent preparation | All variation is instrumental |
3. Low or missing signal: was it acquired?
Observe before optimizing
Inspect the raw acquisition and compare it with the reported result. Confirm the intended ion or elemental isotope, polarity where relevant, mass range or transition, and acquisition time segment. A signal outside the acquired window cannot be recovered by lowering a reporting threshold. Check sample identity, preparation records, and the actual method loaded.
Choose the controlled comparison
Measure an appropriate check standard and blank using the sample method. If both the standard and sample are weak, investigate the common acquisition and delivery path using approved checks. If the standard is satisfactory but the sample is weak, prioritize sample preparation, stability, and matrix effects. A passing clean standard only demonstrates performance for that material and level.
For LC-MS, a matched post-extraction spike versus a neat standard can test matrix-related response separately from extraction loss. The source lesson develops that experiment; its assumptions matter. Matuszewski and colleagues established practical comparisons for these distinct effects.
Platform boundary: GC-MS does not need an electrospray. The Agilent 5977 troubleshooting manual distinguishes a missing chromatographic response with a passing tune from a failed MS tune. In solution ICP-MS, review uptake and the prescribed performance solution; a passing solution check does not validate every sample matrix.
Case: the peak vanished after a retention-time change. What next?
Synthetic observation: an LC-MS/MS standard previously eluted at 4.6 minutes; an approved diagnostic acquisition now places it at 5.3 minutes. The routine scheduled transition ends at 5.0 minutes. Other instrument checks pass.
Inference: the routine method stopped monitoring that transition before the peak arrived. Mass-axis recalibration cannot add those missing measurements. Investigate the retention shift and restore the validated chromatographic conditions, or validate an authorized acquisition-method change. Verify the relevant standards and controls before resuming. This explains the absent recorded peak, not why retention changed, and does not clear the affected samples.
4. Mass drift: an axis problem or a feature problem?
Observe: distinguish a spectral peak moving along m/z from a chromatographic peak moving in time. Check ion composition, charge, isotope assignment, reference value, and processing. Compare the same feature under the same conditions; do not compare a neutral molecular mass with an ion's m/z.
Controlled test: examine several suitable reference ions across the relevant range and over time, within the method's useful signal range. Record signed errors, not just their absolute magnitudes. A consistent pattern across independent references supports an axis-level investigation. A change confined to one crowded or weak feature makes assignment, interference, or peak estimation more immediate questions.
If lock-mass correction is used, inspect whether the correct reference was actually found and used. Enabling the option is not proof of successful correction. Lommen and colleagues' Orbitrap study demonstrates how missing or weak internal references can compromise correction. That platform-specific result is not a universal lock-mass algorithm.
Case: two references move together. Calculate and decide
Synthetic data: references at m/z 200.0000 and 800.0000 are measured at 200.0012 and 800.0048. For each, signed error is 10⁶ × (measured − reference) / reference: both are +6 ppm.
The proportional offsets support checking mass-axis calibration and correction status, provided identity, signal quality, and operating state are sound. Two points do not establish the whole calibration function. Follow the prescribed calibration procedure and verify with suitable independent references. Whether +6 ppm fails depends on this method's criteria; it is not a universal cutoff. Review the affected interval rather than assuming a later passing calibration repairs earlier data.
Limit: accurate-mass ppm checks are not the sole readiness measure for every instrument. Nominal-mass GC-MS or ICP-MS methods may specify mass-position, resolution, response, and interference checks differently. Use the applicable acceptance criteria, rather than importing a high-resolution specification.
5. Background or carryover: does sequence history matter?
Observe: inspect the analyte-specific channel, nearby spectral features, and chromatographic context where applicable. Total signal can rise because unrelated ions increase. Carryover means a previous exposure contributes to a later measurement; background can exist before that exposure.
Controlled test: under an approved procedure, compare a fresh blank before a representative high sample or standard with consecutive fresh blanks afterward. Keep acquisition, containers, and processing comparable. A sequence-dependent increase followed by decline supports a memory contribution; it does not locate the responsible surface.
Use the blank that answers the question. A preparation blank includes reagents and preparation steps. A solvent blank tests a narrower path. In LC, a no-injection gradient can help separate some injection-related effects from persistent system contributions, but its exact behavior depends on the autosampler. Waters' carryover guidance describes further controlled comparisons for LC ghost peaks; apply only compatible procedures.
Case: the first blank is high, then the signal declines
Synthetic integrated responses: the initial blank is 8 arbitrary area units. A high standard gives 100,000. Three subsequent blanks give 208, 48, and 9 in the same analyte window.
The first post-standard blank is 0.208% of the high-standard area; subtracting the initial blank gives 0.200%. State which definition you use. The decaying excess supports carryover, but neither percentage establishes acceptability without the method's required denominator and limit. A small fraction of a high sample may seriously bias a low sample.
Do not name the needle or column as the cause from this pattern alone. Use the approved isolation tests, then verify both blanks and low-level controls after correction. Identify potentially affected samples in the original sequence; do not silently subtract an assumed carryover contribution.
Platform boundary: solution ICP-MS can show element-dependent washout memory, including documented boron retention in sample-introduction components. GC-MS background may involve inlet or column contributions. Neither diagnosis licenses an LC wash recipe on another platform: solvents, acids, materials, and cleaning requirements are instrument-specific.
6. Poor repeatability: where does variation enter?
Observe: plot results in sequence order. Random scatter, a steady trend, and a step after a sample change suggest different tests. Track raw response, internal-standard response where used, their ratio, retention time, and integration boundaries separately. A concentration report can hide which measurement changed.
Controlled test: compare repeated measurements of one stable prepared solution with independently prepared replicates, interleaved with a suitable control to reveal time effects. If only independent preparations vary, investigate preparation, homogeneity, and stability. If repeated measurements also vary, investigate delivery, acquisition, and processing. Repeatedly processing one raw file is not an independent instrument replicate.
Define what your precision study actually varies. NIST's variability guidance distinguishes short-term variation from changes over longer periods. A few consistent injections do not establish between-day or preparation precision.
Case: an excellent ratio hides a large response loss
Synthetic ICP-MS control data: analyte response falls from 10,000 to 5,000 counts per second while its internal standard falls from 20,000 to 10,000 counts per second. The ratio remains 0.50.
A shared response change is plausible, and ratio correction may compensate under the method's assumptions. It does not prove acceptable operation: reduced raw response can compromise low-level performance, and the internal standard may not track every analyte. Check internal-standard acceptance, controls, and the common delivery path before accepting results. Do not increase gain simply to restore the displayed count rate.
Agilent's ICP-MS internal-standard guidance explains that correction relies on suitably matched behavior. An internal standard is evidence to examine, not a guarantee that all matrix effects or interferences disappeared.
Limit: stable retention with variable area does not uniquely identify injection failure. Changing integration boundaries, interference, or source response can produce that combination. Compare raw traces using consistent justified processing; retain the audit trail rather than selecting the most favorable result.
7. Verify, document, or escalate
Close the investigation with the original failed measurement, the controlled comparison, the evidence-supported explanation, and what remains unresolved. Record any authorized change and demonstrate improvement using the relevant independent controls, blanks, and method criteria. A stronger signal alone is not successful repair.
Keep potentially affected data identifiable until the laboratory's review is complete. Escalate repeated failed checks, unexplained step changes, unavailable reference materials, safety faults, or work beyond operator training. Give the responsible specialist raw examples, method versions, the last satisfactory check, sequence history, and tests already performed.
The next test should distinguish explanations. Calibration addresses the mass axis; tuning can change operating behavior; sample, acquisition, and processing checks address different parts of the measurement. Choose from evidence, then verify the complete workflow.
Primary references and scope
The linked manufacturer documents and original studies support the specific examples beside them. Current model-specific documentation and the laboratory's approved procedure govern actual operation. All numerical cases on this page are original synthetic teaching examples, not reported instrument results or acceptance limits.