A 20 µL blank injection on a 30 m × 0.25 mm × 0.25 µm DB-5 column came back with a peak at RT 11.42 min, area 84 mAU·s, and nobody had injected anything but isopropanol. The sequence had run clean the week before. The difference was a new case of septa, lot 22-C, and a GC inlet that had been sitting at 280 °C for four days. That is the shape of most analytical testing ghost peaks, extractables and contamination problems I get called about: a peak with a retention time that matches nothing on the method, appearing after a consumable change, and reproducible to within 2% RSD across three blanks. The instrument is fine. Something in the flow path is off-gassing into it.

This is a troubleshooting piece for people who already know how to run a sequence. It covers which materials bleed which compounds at which temperature, how to read the bleed signature, and how to use blank and placebo injections to corner the source without swapping parts at random.

Every polymer in the flow path is above its service temperature somewhere, or close to it. Silicone septa, PTFE frits, polypropylene vial bodies, PEEK tubing, EPDM pump seals, and the epoxy that holds a frit in a column all release volatile and semi-volatile species when heated. The question is not whether they release anything. It is whether the release lands on your method’s retention window at a concentration above the limit of detection.

Three variables set that: the material, the temperature it sees, and the residence time of the gas or liquid in contact with it. A septum at 250 °C for 40 minutes of a GC run is a different material from the same septum at 40 °C in an autosampler tray. I have seen the same lot of septa pass a headspace method at 80 °C and fail a split injection method at 300 °C, and both results were correct.

What bleeds from what, and at which temperature

The table below is the one I keep taped inside a cabinet door. It is built from our own lot testing and from the temperature ratings the manufacturers publish, cross-checked against Restek’s technical resources on septa and inlet consumables.

The siloxane pattern is the one people misread most often. D4, D5, and D6 elute as a ladder of peaks spaced about 2.2 minutes apart on a standard 5% phenyl column at 30 °C/min. If your unknown peak has two or three siblings up the chromatogram, suspect the septum before you suspect the sample.

Temperature is a multiplier, not a switch

Bleed rate roughly doubles for every 15 to 20 °C above the onset temperature. A septum rated for 250 °C maximum is not clean at 250 °C; it is at the edge where the bleed rate is high enough to show on a sensitive method. Run that same septum at 300 °C and you can get a peak area 10 to 20 times larger from the same lot. I have watched a lab chase a “new contaminant” for two weeks when the only change was that someone raised the inlet to 300 °C to sharpen a tailing peak.

This is why the first question in any ghost peak investigation is not “what is the sample” but “what temperature did the flow path see, and what is the rating of every polymer in contact with it.” If the inlet is hotter than the lowest-rated polymer in the flow path, you have a bleed source by design.

Reading the bleed signature before you touch anything

A ghost peak has a fingerprint. Two minutes of reading it correctly saves a day of swapping parts.

• Reproducible retention time, variable area: usually an inlet consumable (septum, liner, O-ring) with a temperature-dependent release. Area tracks how long the inlet sat hot.
• Reproducible area, drifting retention time: usually a column problem, not contamination. Check the column before blaming the consumables.
• Peaks that grow across a sequence: carryover or accumulation on the inlet liner. The peak gets bigger as the liner loads.
• Peaks that only appear in blanks: the sample matrix is masking or suppressing the contaminant. Run the blank at the same inlet temperature and split ratio as the method.
• A ladder of peaks spaced 2.2 minutes apart: siloxanes. Go straight to the septum.
• Broad humps in the 20 to 30 minute window on headspace: vial or cap oligomers. Check the headspace oven temperature against the vial rating.

I use a 30 m × 0.25 mm × 0.25 µm 5% phenyl column for this diagnostic because it separates the siloxane ladder cleanly and the retention times are familiar. If you run a different phase, build the equivalent table once and keep it.

The blank injection is a control, not a formality

Most labs run a blank at the start of a sequence and call it done. That blank tells you almost nothing about contamination because it runs on a clean inlet. The useful blank is the one that runs at the method’s real inlet temperature, with the method’s real split ratio, after the inlet has been hot for the same time as a sample run. Run three of them. If the peak area is stable within 5% across the three, you have a steady-state source. If it climbs, you have accumulation.

Run the same three blanks with a fresh septum in place and the old one in a labeled bag. If the peak disappears, you have your answer and you have a retained sample of the failed lot for the supplier conversation. That bag is worth more than the argument.

Isolating the source with blank and placebo injections

Placebo injections are the half of the investigation most labs skip, and they are the half that separates a sample-derived contaminant from a consumable-derived one. A placebo is the sample matrix without the analyte: the same diluent, the same vial, the same cap, the same autosampler tray, the same needle. If the ghost peak appears in the placebo, the source is in the matrix or the container. If it appears in the blank but not the placebo, the source is in the flow path.

The matrix below is the decision logic I use. It takes about 90 minutes to run and it does not require a single instrument swap.

Run the series in that order. Each row narrows the search space by one variable, and by row six you either have the source or you have eliminated every consumable in the path, which is also a result.

Field case: the 11.4-minute peak that was not in the sample

Symptom: a contract lab running a USP method on a 30 m × 0.25 mm × 0.25 µm column saw a peak at RT 11.42 min, area 84 mAU·s, in every blank and every sample, starting on a Monday after a Friday septum change. Three analysts had already re-prepped the standards.

Measurement: three blanks at the method inlet temperature (280 °C) gave areas of 81, 85, and 86 mAU·s. The same three blanks at 220 °C gave areas of 4, 4, and 5 mAU·s. The peak was temperature-dependent and reproducible. A second blank series with a septum from the prior lot gave no peak at 280 °C.

Fix: the new septum lot had a silicone core with a lower cure state, and the 280 °C inlet drove siloxane release well above the method’s LOD. The lab switched to a PTFE-faced septum rated for 300 °C, re-ran the blanks, and got a clean baseline. They also added a lot-acceptance blank to their incoming consumable check: one blank injection at method temperature before the lot goes into the cabinet. The peak never came back.

Field case: the headspace hump that was the cap, not the column

Symptom: a food lab running headspace GC on a 60 m × 0.32 mm × 1.8 µm column saw a broad hump at 24 to 27 min in every vial, growing over a 12-hour sequence. The column had been replaced twice.

Measurement: a placebo series using the same matrix and the same vial lot reproduced the hump. A second series using a different cap lot, same vials, did not. The hump tracked the cap lot, not the vial lot and not the column. The cap liner was a silicone-faced material rated for 180 °C; the headspace oven was running at 200 °C.

Fix: the lab moved to a PTFE-faced cap rated for 250 °C and dropped the oven setpoint to 190 °C. The hump disappeared and the method’s system suitability passed on the first run. The column was never the problem, and two replacements were wasted.

The consumable selection rules that prevent most of this

Most ghost peak investigations end with a consumable that was wrong for the temperature, wrong for the matrix, or wrong for the lot. A few selection rules prevent the majority of them.

• Match the septum rating to the inlet temperature plus 50 °C. If the method runs at 280 °C, use a septum rated for 330 °C or better. The rating is a maximum, not a comfortable operating point.
• Use PTFE-faced septa for any inlet above 250 °C. The PTFE barrier delays siloxane release and buys you 50 to 80 °C of headroom.
• Match the vial and cap liner to the headspace oven temperature. A 180 °C liner in a 200 °C oven is a bleed source by design. Check the liner datasheet against the oven setpoint before the run, not after.
• Use a fresh needle for every sequence. A needle that has seen 200 injections carries matrix residue into the next blank. I replace at 100 injections for dirty matrices and 200 for clean ones.
• Change the inlet liner at 100 to 150 injections for dirty matrices, and at the first sign of a growing peak. A liner is cheaper than a re-run.
• Condition new columns per the manufacturer’s ramp, not a shortened version. A column that has not been conditioned fully will bleed for its first 20 to 30 runs. The Agilent GC column and accessory documentation gives the conditioning ramp for each phase; follow it.

The vial and cap side of this is where I see the most avoidable failures. A HPLC vial with a cap liner rated for 180 °C in a 200 °C headspace oven is a contamination source you built into the method. The same logic applies to lab vials used for LC: a polypropylene vial at ambient temperature is fine, but the same vial in a heated autosampler tray at 60 °C can leach oligomers into a sensitive method. Check the tray temperature against the vial rating the same way you check the inlet against the septum.

Torque, puncture count, and the numbers that matter

Crimp torque on a 11 mm aluminum seal should land at 1.5 to 2.0 N·m. Under-crimped, the seal leaks and the headspace pressure drops, which changes the split and the peak area. Over-crimped, the septum is deformed and the puncture channel does not reseal, which is a leak problem rather than a contamination one, but it shows up as area variability that people misread as a ghost peak. Use a torque wrench and check one seal per 50.

Puncture count is the other number. A septum is good for roughly 50 to 100 punctures in a split/splitless inlet before the channel stops resealing, and fewer at high inlet temperatures. Track punctures per septum on the sequence log. A septum at 90 punctures at 280 °C is a bleed source even if it looks intact.

For LC, the equivalent numbers are simpler. Replace the inline filter frit at 500 injections or when backpressure climbs 15% above baseline. Replace the needle seat at 1,000 injections. Replace pump seals at the manufacturer’s interval, not when they leak. A seal that is starting to fail sheds EPDM fragments into the mobile phase for weeks before it leaks, and those fragments show up as late-eluting peaks that nobody can source.

A troubleshooting checklist for the next ghost peak

When a ghost peak appears, work this list in order. It takes about two hours and it does not require a single instrument swap until step six.

1. Record the retention time, area, and peak shape. Compare against the method’s known peaks and the siloxane ladder.
2. Run three blanks at method inlet temperature. Note whether the area is stable or climbing.
3. Run three blanks at 50 °C below method temperature. If the peak shrinks by more than 80%, the source is temperature-driven bleed.
4. Run a placebo series with the same vial lot and a different cap lot. Then the reverse. One of the two will separate.
5. Swap the septum for a fresh one from a different lot. Run the three blanks again.
6. Swap the inlet liner and re-run. If the peak persists, move to the column.
7. Run a column conditioning ramp per the manufacturer’s spec. If the peak persists after conditioning, the column is the source and the lot should be quarantined.
8. Retain the failed consumable in a labeled bag with the lot number and the date. The supplier conversation needs the sample, not the story.

The order matters because each step eliminates a variable. Swapping the column first is the most common mistake I see, and it is expensive. The column is the last thing to change, not the first.

One more thing: document the blank result in the run file. A ghost peak that is investigated and resolved without a written record will be investigated again in six months by a different analyst. The EPA SW-846 method structure and the Thermo Fisher chromatography documentation both expect a blank result in the data package; use that expectation to force the record.

The decision rule

When a ghost peak appears, the decision rule is this: treat it as a material problem until the blank series proves otherwise. Run three blanks at method temperature, run three at 50 °C below, and run a placebo series against two cap lots and two septum lots. If the peak tracks temperature, it is a bleed source in the flow path. If it tracks a lot, it is a consumable. If it tracks the matrix, it is the sample, and that is a method problem rather than a contamination problem. The instrument is the last thing to suspect and the last thing to touch. Most of the time, the fix is a septum rated 50 °C above the inlet, a cap liner rated above the oven, and a blank injection logged in the run file before the next sequence starts.

For the wider workflow context, the analytical testing resource and the chromatography consumables reference both carry the lot-acceptance templates I use. Build the blank series into the incoming consumable check, and most of these investigations never start.

References & further reading

Related resources from our lab consumables network:

• Analytical Testing at testinganalytical.com
• Hplc Vial at vialforhplc.com
• Chromatography Consumables at chromatographyconsumables.com
• Lab Vials at labvials.com

External references and standards:

• Agilent chromatography consumables and columns
• Thermo Fisher chromatography and sample handling
• U.S. EPA SW-846 test methods for waste and materials
• Restek technical resources for chromatography