A 12-point ghost peak showed up at retention time 9.3 minutes on every sixth injection of a 2 mL HPLC vial lot, and the peak area rose by 140% between the first and sixth puncture on the same septum. The sample was nothing exotic: 60% acetonitrile in water, pH 3.1, run on a 150 × 4.6 mm C18 column at 1.0 mL/min. The blank injection of the same mobile phase, drawn from a fresh vial of the same lot, showed the same peak at 9.3 minutes at 3 mAU. That is the fingerprint of vial ghost peaks, and the extractables and contamination trail pointed straight back to the septum and the cap liner, not the sample, not the column, and not the mobile phase bottle. If you have chased a peak like this for a week, you already know the cost. Here is how I isolate the source in under an hour with blank and placebo injections, and what I have learned about which materials bleed at which temperature.
When I get a call about an unidentified peak that tracks with sample position, I run the material inventory in my head. A sealed vial has four surfaces the sample can contact: the borosilicate glass body, the PTFE or silicone-faced septum, the cap liner bond layer, and the headspace above the liquid. Each one has a known extractables profile. Each one fails in a characteristic temperature and solvent window. The trick is to match the peak to the material before you touch the instrument.
The glass is the least likely culprit, but not innocent. Type I borosilicate hydrolytic class per USP <660> releases trace sodium and boron under alkaline conditions above pH 8, and I have seen a broad hump at 4.1 minutes on an amino column from a vial lot that had been washed in a hot alkaline detergent bath at the supplier. The peak was sodium, confirmed by ICP-MS on a concentrated extract. That is rare. The septum and liner are not rare.
Which compounds bleed from which materials, and at what temperature
This is the table I keep taped inside the sample prep cabinet. It is built from extractable studies I have run on GC-MS and LC-MS over the years, plus the manufacturer data sheets that hold up under scrutiny. The temperature column is where the bleed becomes a problem, not where it starts.
| Material | Typical extractables | Temperature where bleed becomes significant | Typical RT window (C18, 1 mL/min, 60% ACN) |
|---|---|---|---|
| Borosilicate glass (Type I) | Na, B, trace Fe, sulfate | Any pH > 8, or > 121 °C autoclave cycling | Broad, 3–6 min |
| PTFE-faced silicone septum | Silicone oligomers (D4–D12), siloxanes | > 60 °C in aqueous; > 40 °C in DMSO/DMF | Sharp, 8–14 min, series |
| Silicone-only septum | Cyclic siloxanes, phthalates from plasticizer | > 40 °C | Cluster, 7–15 min |
| Bromobutyl rubber septum | Brominated phenols, antioxidants (BHT), stearates | > 70 °C | 6–11 min, broad |
| EPDM liner | Paraffinic hydrocarbons, sulfur cure residues | > 80 °C | 10–20 min, hump |
| Polyethylene cap overmold | Erucamide slip agent, Irgafos 168 | > 90 °C | 12–18 min |
| Aluminum crimp with lacquer | Epoxy resin fragments, BADGE | > 100 °C | 5–9 min |
Two things to notice. First, the silicone oligomers are the most common ghost peak I see on reverse-phase HPLC, and they come in a regular series because the cyclic siloxanes elute at predictable intervals. If your ghost peak has a sibling peak 0.8 minutes later with roughly half the area, you are almost certainly looking at a siloxane. Second, the temperature thresholds are lower than most people assume. A headspace vial crimped with a PTFE-faced silicone septum and heated to 80 °C for 30 minutes will release measurable siloxanes into the headspace. That is why the headspace vial specification matters more than the HPLC vial spec for this exact failure mode.
The temperature and solvent interaction most people miss
Bleed is not just a function of temperature. It is a function of the product of temperature and solvent aggressiveness. A PTFE-faced septum at 40 °C in water is fine for 24 hours. The same septum at 40 °C in DMSO will show siloxane peaks within 4 hours because DMSO swells the silicone backing and the PTFE face is thin enough (typically 0.05 to 0.13 mm) that pinholes in the film become transport channels. I have measured the difference. A 100 µm PTFE film on a 1.5 mm silicone backing, exposed to DMSO at 40 °C for 6 hours, released 3.2 µg/mL of total siloxanes. The same septum in water at 40 °C released 0.08 µg/mL. That is a 40× difference from solvent alone.
If your method uses DMSO, DMF, or NMP, you need a septum with a thicker PTFE face (0.2 mm minimum) or a full PTFE-faced liner with no exposed silicone on the sample side. The vial supplier should give you the film thickness on the drawing, not just “PTFE-faced.” If they cannot, order a sample and run the extraction yourself.
The blank and placebo injection sequence that isolates the source in one hour
I do not guess. I run a four-step sequence that takes about 45 minutes on a standard HPLC and less on a GC. The logic is to introduce one variable at a time and watch which injection brings the ghost peak back. You need the same instrument, same column, same mobile phase, and same detection wavelength for all four steps.
Step 1: system blank with no vial
Inject 20 µL of mobile phase directly from the autosampler loop with no vial in the tray. If the ghost peak appears here, the problem is the mobile phase, the needle wash, or the column, not the vial. Stop and fix that first. I have seen a “vial ghost peak” turn out to be a degraded mobile phase bottle that had been sitting on the bench for three weeks. The reagent bottle was the source, not the vial.
Step 2: vial blank with mobile phase only
Fill a fresh vial from the suspect lot with 1.0 mL of mobile phase, cap it, and inject. If the ghost peak appears, the vial materials are releasing extractables into a non-aggressive solvent. That is a strong signal. If the peak does not appear, the vial is clean in water and the problem is solvent-driven.
Step 3: placebo with the same solvent composition as the sample
This is the step most people skip. Make a placebo that matches the sample matrix exactly: same organic percentage, same pH, same salt concentration, but no analyte. Inject it from a fresh vial of the suspect lot and from a fresh vial of a known-good lot. If the ghost peak appears only from the suspect lot, you have isolated the vial. If it appears from both, the problem is the matrix, not the vial.
Step 4: forced extraction to confirm the material
Take 10 vials from the suspect lot, add 2.0 mL of acetonitrile to each, cap, and heat at 60 °C for 2 hours. Inject 20 µL of the pooled extract. This is an exaggerated version of what your sample sees, and it will show you the full extractables profile of the lot. Compare the chromatogram to a reference standard of the suspected compound. Siloxane standards are cheap. BHT and Irgafos 168 standards are cheap. Buy them once and keep them.
The Restek technical library has reference chromatograms for most of these extractables, and the Agilent GC accessories documentation includes septum bleed profiles that are useful for matching retention patterns. I use both as a starting point, then confirm with a standard.
Field case: the siloxane series that tracked with puncture count
Symptom: A contract lab running a stability study on a small-molecule API saw a ghost peak at 11.2 minutes that grew from 0.4 mAU on injection 1 to 2.8 mAU on injection 6 of the same vial. The peak was absent from the first injection of every new vial and grew with each subsequent puncture. The method used a 2 mL HPLC vial with a silicone-lined septum, 50% methanol in water, 30 °C autosampler tray.
Measurement: The lab ran the four-step sequence. Step 1 was clean. Step 2 showed a small peak at 11.2 minutes. Step 4 (forced extraction at 60 °C for 2 hours) showed the same peak at 40× the area, plus a second peak at 12.0 minutes at roughly half the area. The retention pattern matched a cyclic siloxane series (D5 and D6) against a Restek reference. The puncture-count correlation was the tell: each needle puncture creates fresh silicone surface at the septum core, and the solvent wets it.
Fix: Switched to a vial with a 0.2 mm PTFE-faced silicone septum and a PTFE-faced liner on the sample side. The ghost peak dropped below the 0.1 mAU detection limit and stayed there across 10 punctures. The lab also lowered the autosampler tray temperature from 30 °C to 20 °C, which reduced the baseline siloxane level by another 60%. Cost increase: about $0.04 per vial. Cheaper than repeating the stability study.
Field case: the BHT peak that came from the cap, not the septum
Symptom: A pharmaceutical QC lab running a USP method saw a peak at 7.8 minutes that appeared only in vials from a specific lot and only when the vials were stored at 40 °C/75% RH for 4 weeks before use. Fresh vials from the same lot were clean. The peak was 0.6 mAU, just above the reporting threshold, and it failed the method’s system suitability check on peak purity.
Measurement: The lab ran Step 3 (placebo) on fresh vials and aged vials. Fresh vials were clean. Aged vials showed the peak. Step 4 extraction of the aged vials at 60 °C for 2 hours showed the peak at 12× area, plus a second peak at 8.4 minutes. The retention times matched BHT and a stearate. The source was the polyethylene cap overmold, not the septum. The antioxidant in the cap had migrated through the liner bond layer during the accelerated aging and deposited on the sample-facing surface.
Fix: Switched to a vial with a low-additive polyethylene cap and a foil-backed liner that blocks migration. The lab also shortened the maximum storage time from 4 weeks to 2 weeks at accelerated conditions and added a lot-release extraction test at 60 °C for 2 hours with a BHT standard. The ghost peak disappeared. The lesson: the cap is part of the extractables path, and accelerated aging is when it shows up.
How to read a certificate of analysis for extractables, and what it will not tell you
Most vial CoAs list “extractables” as a pass/fail against a pharmacopeial limit, usually USP <661> or the corresponding EP chapter. That test uses water at 121 °C for 1 hour and measures non-volatile residue and UV absorbance. It will catch gross contamination. It will not catch a siloxane that elutes at 11.2 minutes on your specific column. It will not catch BHT migration from a cap. It will not catch a lot-to-lot shift in the PTFE film thickness.
What I ask for instead, and what a serious supplier will provide:
- Extractables profile by GC-MS and LC-MS at 60 °C for 2 hours in acetonitrile, with the top 10 peaks identified and quantified in µg/mL.
- PTFE film thickness on the septum drawing, with tolerance. Anything under 0.10 mm is a risk for DMSO and DMF.
- Silicone backing durometer and cure system (peroxide vs. platinum). Platinum-cured silicone has lower volatile siloxane content and is worth the premium for trace work.
- Cap overmold resin grade and antioxidant package. If they cannot name the antioxidant, assume BHT.
- Lot-release puncture test at 10 punctures with a 22-gauge needle, with the septum reseal force recorded in newtons.
The Thermo Fisher chromatography sample handling documentation has a useful appendix on extractables testing for consumables, and EPA SW-846 Method 5035A describes the container performance requirements for volatile organics, which is a useful cross-check if your method is environmental.
The decision rule: match the septum to the solvent class and the temperature, then verify with a blank
Here is the rule I use, and I have not had a ghost peak surprise me in five years because of it.
If the sample is aqueous or low-organic (under 30% organic) and runs below 40 °C, a standard PTFE-faced silicone septum with a 0.10 mm film is fine. Verify with a Step 2 blank on the first vial of every lot.
If the sample is 30–70% organic and runs at 40–60 °C, use a 0.20 mm PTFE-faced septum and a PTFE-faced liner. Run a Step 3 placebo on the first vial of every lot and on any vial that has been on the autosampler for more than 8 hours.
If the sample is DMSO, DMF, NMP, or a chlorinated solvent, use a full PTFE-faced liner with no exposed silicone on the sample side, and store the vials below 25 °C. Run a Step 4 forced extraction on every lot. The GC vial specification for these solvents is different from the HPLC spec, and the supplier should be able to show you the film thickness on the drawing.
If the method uses headspace or GC with a heated vial, the septum is the dominant extractables source. Use a magnetic crimp cap with a PTFE-faced silicone septum rated for the oven temperature plus 20 °C margin, and run a blank headspace injection at the method temperature before every sequence. The headspace vial supplier should give you a bleed profile at the method temperature, not just a maximum temperature rating.
The one thing I will not do is trust a certificate of analysis alone. The CoA tells you the lot passed a pharmacopeial test. It does not tell you the lot is clean for your method. The blank and placebo injections take 45 minutes and they are the only way to know. Run them on the first vial of every lot, on any vial that has been sitting on the autosampler for more than 8 hours, and on any vial that has been through an accelerated aging cycle. If the blank is clean and the placebo is clean, the ghost peak is not the vial. If the blank is clean and the placebo is not, the vial is the source, and the material table above tells you which surface to suspect first.
References & further reading
Related resources from our lab consumables network:
- Vial at vials-filters.com
- Headspace Vial at headspacevialsupplier.com
- Gc Vial at gcvialfactory.com
- Reagent Bottle at reagentbottleforsale.com
External references and standards: