Written and maintained by CASRAI Editorial Board
Last updated
A ghost peak is any signal on an HPLC chromatogram that does not correspond to a real component of the sample being run. It matters because it does not announce itself as an artefact — it looks exactly like a peak, and can be misread as an unknown impurity, a degradation product, or worse, integrated into a calibration curve. Ghost peaks fall into four mechanistically distinct classes — injector carryover, mobile-phase contamination, late-eluter breakthrough, and gradient artefacts — and each one has a different fix. Treating all of them as “system noise” and re-running the method is how the same ghost peak gets diagnosed three different times by three different people. The four classes are separable with two simple diagnostic runs: a blank injection and a blank gradient.
The Two Diagnostic Tests
Before assigning a cause, run both of these. Together they localise almost every ghost peak without needing to touch the sample at all.
- Blank injection. Inject clean solvent (ideally the injection diluent, not the mobile phase) using the normal method, immediately after the run that showed the ghost peak. This tests the injector and flow path for anything left behind by the previous injection.
- Blank gradient. Run the full gradient program — same time, same %B ramp, same flow rate — with nothing injected at all, ideally at the start of a sequence before any sample has touched the system. This tests the mobile phase and the gradient itself, independent of any injection.
A peak’s behaviour across these two tests, plus how its retention time relates to the method’s normal elution window, is what assigns it to one of the four causes below.
Injector (Autosampler) Carryover
Mechanism. A small amount of analyte from a previous injection is retained in the injection needle, needle seat, sample loop, or transfer lines and is re-introduced into the next injection. It is most likely after a high-concentration sample, a viscous or “sticky” matrix, or a compound with poor solubility in the needle-wash solvent.
How it behaves across the tests. Carryover shows up on a blank injection run immediately after the suspect sample, at the same retention time as the analyte in that sample. Its peak area is small relative to the original peak and typically shrinks on a second consecutive blank as the residual clears. It does not appear in a blank gradient run at the start of a sequence, before any sample has been injected — there is nothing yet to carry over.
Fixes. Strengthen or extend the autosampler needle-wash program (a wash solvent that actually dissolves the analyte, not just the mobile phase); check needle-seat and rotor-seal wear, which creates dead volume that traps sample; insert a genuine blank or weak-needle-wash injection between a high-concentration sample and the next quantitative injection if the method is otherwise carryover-prone; for persistent carryover on a specific analyte class, consider a stronger wash solvent segment (e.g. an organic-rich or acidified wash) rather than relying on the mobile phase alone to clean the flow path.
Mobile-Phase Contamination
Mechanism. The mobile phase itself carries something that elutes as a peak: trace organic impurities in the solvent lot, plasticizers or biocide leaching from bottle caps and tubing, buffer-salt contamination, or biological growth in an aqueous buffer left standing too long.
How it behaves across the tests. This is the one class that appears on a blank gradient run before any sample has ever been injected — the system does not need a sample to produce it. It also appears on a blank injection, but crucially at a similar magnitude whether or not a sample was run beforehand; it does not diminish the way carryover does, because the source is continuously present in the mobile phase, not left behind by one injection. Retention time is usually consistent run to run, since it reflects the mobile-phase composition rather than anything analyte-specific.
Fixes. Prepare fresh mobile phase and confirm solvent lot/grade (HPLC or gradient grade, not reagent grade, for the organic modifier); switch to glass reservoirs with PTFE-lined caps if plastic components are suspected; filter buffers through the correct pore size and do not hold aqueous buffers unrefrigerated for more than the manufacturer’s stated window; replace inline solvent filters/frits; if a specific reagent (e.g. a buffer salt or ion-pairing agent) is the suspect, test a fresh bottle from a different lot before concluding the method itself is at fault.
Late-Eluter Breakthrough
Mechanism. A compound in the sample matrix is real, but it is more retentive than anything the method’s run time is designed to elute. It does not come off the column within the programmed run window; instead it continues migrating slowly during the next injection’s equilibration and run, and elutes there — in a chromatogram that has nothing to do with it.
How it behaves across the tests. This is the pattern most often mistaken for carryover, and the two are told apart by retention time. Carryover reproduces the analyte’s own retention time; a late eluter appears at a retention time that does not correspond to any expected analyte in the current run, often later than the method’s normal elution window would predict for a real component, and its appearance correlates with which sample ran two or three injections earlier, not the immediately preceding one. It typically disappears if the sequence includes an extended isocratic hold or an extra high-organic wash segment after a “dirty” matrix sample — because that segment is what finally elutes it, before it has the chance to bleed into a later run.
Fixes. Extend the run time or add a high-organic wash-and-re-equilibration segment after known dirty-matrix samples, even if it is not needed for the analytes of interest; consider a guard column to trap the most retentive matrix material before it reaches the analytical column; if the same late eluter recurs on a given sample type, characterise it once (even approximately) so it can be recognised on sight rather than re-diagnosed from scratch each time it appears.
Gradient-Artefact Ghosts
Mechanism. The gradient itself produces the peak, independent of the sample or any prior injection. Two common sources: a UV/baseline mismatch between mobile phases A and B that produces a step or hump as the gradient ramps through their absorbance difference, and trace impurities in the organic modifier (mobile phase B) that pre-concentrate at the head of the column during the initial low-%B loading period and then elute together as a single sharp peak once the gradient reaches the %B where they are no longer retained — effectively a mini trace-enrichment of whatever was dissolved in the solvent.
How it behaves across the tests. This is the class the blank gradient test exists for: run the full program with nothing injected, and the peak still appears, at a retention time (or %B) that is highly reproducible run to run because it tracks the gradient profile itself, not anything introduced by an injection. It is unaffected by whether a blank or a real sample was run beforehand, which separates it from both carryover and late-eluter breakthrough.
Fixes. Use gradient-grade (not isocratic-grade) solvent for mobile phase B specifically, since it carries the trace-enrichment load; run a solvent-matched blank-gradient subtraction if the software supports it, so the artefact is subtracted from every sample chromatogram automatically; if the mismatch is baseline/UV-driven rather than a true impurity peak, adjusting detection wavelength or accepting a documented baseline step (rather than misreading it as an analyte) may be the correct response instead of chasing a fix that does not exist.
Telling the Four Apart: A Quick-Reference Table
| Ghost peak type | Appears on blank gradient (nothing injected)? | Appears on blank injection? | Retention-time signature | Diminishes on repeat blanks? |
|---|---|---|---|---|
| Injector carryover | No | Yes, right after the suspect sample | Same RT as the source analyte | Yes, shrinks on successive blanks |
| Mobile-phase contamination | Yes | Yes, regardless of prior sample | Consistent RT run to run | No, stays constant |
| Late-eluter breakthrough | No (unless a prior late-eluting sample ran first) | Only after the specific dirty-matrix sample, may lag by 2–3 injections | Outside the normal elution window for any expected analyte | Clears after an extended wash/hold, not a simple blank |
| Gradient artefact | Yes, reproducibly | Yes, unaffected by sample history | Tracks the gradient (%B) profile, not the sample | No, tied to the gradient program itself |
A Note on Vials and Septa
One source of apparent “carryover” is not the injector at all: autosampler vial and septa chemistry. Vials that are not adequately deactivated for polar or basic analytes can adsorb and slowly release analyte across injections, and septa without a PTFE or silicone laminate can bleed siloxanes that appear as extra peaks unrelated to any real carryover mechanism in the autosampler hardware itself. If the fixes above do not resolve an apparent carryover pattern, the vial/septa combination is worth checking before assuming the autosampler needle or wash program is at fault — see the CASRAI guide on HPLC vials and septa selection for compatibility guidance by analyte class.
Frequently Asked Questions
What is a ghost peak in HPLC?
A ghost peak is a signal on the chromatogram that does not come from a genuine component of the sample injected for that run. It can originate from the injector, the mobile phase, a carried-over compound from an earlier sample, or the gradient program itself — not from the sample under test.
How do I tell carryover from mobile-phase contamination?
Run a blank gradient with nothing injected at all, before any sample has touched the system. If the peak still appears, it is coming from the mobile phase, not the injector — carryover has no source to draw on until at least one sample has been run.
What is a blank gradient run, specifically?
It is the method’s full gradient program — same timing, same flow rate, same %B ramp — executed with no injection, or with only the injection diluent injected. It isolates the mobile phase and gradient mechanics from anything the sample or injector could contribute.
Why does a ghost peak show up only every few injections instead of every run?
That intermittent pattern is the signature of late-eluter breakthrough: a highly retentive matrix compound that does not clear the column within one run’s time window and instead surfaces two or three injections later, whenever it finally elutes.
Can a ghost peak be mistaken for a real impurity or degradation product?
Yes, and this is the practical risk that makes the four-way diagnosis worth doing properly — a gradient artefact or contamination peak that gets misread as a genuine impurity can trigger unnecessary investigation, or worse, get folded into a reportable result if it happens to co-elute near a specification limit.
Related CASRAI reading: HPLC: columns, mobile phases, and a peak-problem troubleshooting table for the instrument fundamentals this page assumes; HPLC column selection and HPLC vials and septa selection for the hardware choices that prevent several of the causes above; HPLC vs UPLC vs UHPLC for how dwell volume and system design change gradient-artefact behaviour on newer instruments.
