HPLC autosampler troubleshooting: common problems and fixes

Most autosampler problems trace back to a short list of parts: the needle and its seat or injection port, the metering syringe and its seal, the injection valve rotor seal, and the vial with its septum. Start by finding out whether the fault follows the sample or the instrument: inject several times from one vial of a clean standard, compare a zero-volume injection with a real one, and switch the valve between load and inject while watching the pressure. Those three checks take a few minutes and point you to the right row of the table below.

Key points

• A healthy autosampler gives about 0.3-0.5% area RSD for six 10 µL injections of a well-retained standard. Above 1% under those conditions it needs service (Dolan, LCGC).
• Small injections magnify any fixed volume error: an extra 0.6 µL is a 30% bias at 2 µL but 6% at 10 µL (Louie and Dong, LCGC).
• Repeated injections from one vial separate autosampler faults from sample preparation faults.
• If system pressure falls back to normal when the valve is in bypass, the blockage is inside the autosampler flow path.
• Split or broad early peaks are usually an injection effect: sample solvent stronger than the mobile phase, or a volume above about 15% of the first peak's volume.

Before you start: three habits that make troubleshooting fast

Know your injector architecture. In a flow-through-needle design (the most common type, as Stoll notes), the needle, seat and metering device sit in the high-pressure path, so a blocked seat raises system pressure. In a fixed-loop design with a low-pressure needle, only the loop is switched into the flow, so a blocked needle or injection port shows up as a failed draw instead. How an HPLC autosampler works explains both.

Record what normal looks like: the pressure with the valve in each position, the area RSD of your check standard and the blip on the pressure trace at injection. You cannot spot an abnormal value without a normal one.

Divide and conquer. Dolan's LC troubleshooting strategies column recommends tests that split the possible causes into two groups, then repeating. Change one thing at a time, and swap a suspect part for one known to be good.

Autosampler problems at a glance

Why are my peak areas not reproducible?

Likely causes, most probable first:

1. The vials differ, not the injections. Louie and Dong chased erratic potency results on a UHPLC assay where duplicate calibrators differed by about 20%. Peak area repeatability from a single vial was below 0.3% RSD, which cleared the autosampler. The cause was a film of aqueous standard trapped in the vial neck after the analyst inverted the vial, which added about 0.6 µL to injections from the affected vials (the method injected 3 µL) (LCGC).
2. Air or vacuum during the draw. A draw that is too fast for a viscous sample forms bubbles in the syringe (Dolan). Overfilled vials and non-slit septa that seal around the needle can leave a partial vacuum (see the air section below).
3. A leak between the needle and its seat, or at a fitting. Agilent lists this as a frequent hardware cause (KP96). A worn rotor seal can do the same (KP818).
4. An injection volume near the bottom of the metering range (see the worked example).
5. A worn syringe seal or metering device, usually seen as precision drifting worse over weeks.

Quick test. Make six injections from one vial of a stable, well-retained standard. Dolan's check uses 2 µg/mL anthracene on a C18 column with 80/20 methanol/water, UV at 250 nm and 10 µL injections, and expects no more than 1% RSD. Then inject once from each of six vials of the same solution. Good same-vial RSD with poor vial-to-vial RSD points at sample preparation. Poor RSD in both points at the instrument: repeat at half the draw speed with half-filled vials, and if RSD improves, the draw is the problem.

Fix. Agilent's note suggests filling vials to 1/2-2/3 of their volume, keeping the room temperature constant, and dropping the draw speed for viscous samples (it gives 50 µL/min instead of a 200 µL/min default). Do not invert filled vials; vortex instead. Inject a larger volume of a more dilute sample where sensitivity allows. If none of that helps, replace the needle seat (or injection port seal), the rotor seal and the syringe seal, in that order.

Worked example: why small injections are less precise

Two kinds of volume error behave differently. A random error with a standard deviation σ_V gives an area RSD that shrinks with volume. A fixed bias, such as the 0.6 µL film in the Louie and Dong case, gives a percentage error that also shrinks with volume.

RSD_area (%) ≈ 100 × σ_V / V
bias (%)     ≈ 100 × ΔV / V
RSD_total    ≈ √(RSD_injection² + RSD_detector² + RSD_integration²)

With an illustrative random error of σ_V = 0.02 µL and the 0.6 µL bias from the case study:

The 2 µL and 10 µL bias figures match those Louie and Dong report. Independent errors add as squares, so a 0.4% injection RSD plus 0.6% from integrating a small, noisy peak gives √(0.4² + 0.6²) ≈ 0.72%: check integration before blaming the autosampler. Where you can, inject 10 µL of a dilute solution rather than 1 µL of a concentrated one. For how far you can push the volume, see how much you can inject.

Why is there no peak, or why are injections missing?

Likely causes, most probable first:

1. The needle could not reach any sample: an empty vial, a sample level below the needle tip (common with inserts and low fill volumes), or a sequence position that does not match the tray.
2. A blocked needle or seat: the syringe moves but nothing is drawn.
3. The valve did not switch. Agilent lists "the valve does not switch from bypass to main pass" among rotor seal symptoms (KP818).
4. A large air bubble in the metering line, so the syringe draws air instead of sample.
5. Inconsistent vial bottoms. Samplers that learn the vial bottom can skip injections or draw from different depths when internal bottom heights vary between vials (MicroSolv).
6. Not the autosampler: data acquisition did not start, or the detector lamp is off.

Quick test. Watch one complete cycle, through the door window or with the door open if the instrument allows it. Mark the liquid level on a vial and make 10 injections of 10 µL: the level should fall by at least 100 µL plus any flush volume your sampler draws. Then inject a known standard from a fresh vial at a known position. A small blip on the pressure trace at the injection time confirms that the valve switched.

Fix. Correct the sequence or tray definition, raise the fill volume or adjust the needle depth for inserts, replace a blocked needle, and purge the metering line. If the valve does not switch, the drive or the rotor seal needs service.

Where do extra or ghost peaks come from?

Likely causes, most probable first:

1. Late eluters from the previous injection in isocratic methods. Waters suggests collecting data for twice the method end time to catch them (WKB246120).
2. Mobile phase contaminants in a gradient. The weak starting eluent traps mildly hydrophobic impurities at the column head, and the gradient focuses them into sharp peaks (Taylor, LCGC).
3. Carryover from the needle, seat or valve (next section).
4. Vial and septum extractables, such as plasticisers. Taylor lists septa, gloves and pipettes as possible sources.
5. The sample itself: degradation in the vial or matrix components.

Quick tests (from the Waters and Taylor articles):

For gradient methods, Waters also recommends a blank or zero-volume injection with a double gradient (the gradient programmed twice in one run) to narrow down whether the injector, mobile phase, column or sample is responsible.

Fix. Use fresh mobile phase from high-grade solvents. Rinse glassware well and keep plastic out of eluent preparation. Taylor shows that cleaning the injection valve reduced contaminant peaks, so clean or replace needle seats and rotor seals. Extend the run or add a strong-solvent wash step for late eluters, and try septa from another supplier.

Is it carryover?

Carryover is a peak from the previous sample appearing in the next injection. The usual sources are the outside and inside of the needle, wash port and injection port seals, scratches in the rotor seal grooves, unswept volumes in fittings, and adsorption to metal surfaces (LCGC / CHROMacademy).

Quick test. Inject a blank right after your highest standard, then a second blank. A peak that shrinks sharply from the first blank to the second is carryover. A peak of the same size in both is contamination of the blank, the mobile phase or the system.

Fix. Use a wash solvent at least as strong as the composition that elutes your most retained analyte, and a wash volume of at least 10× the injection volume (the CHROMacademy guidance). If carryover persists, Thermo Fisher's guidance is to replace the needle seat (Thermo Fisher), and the rotor seal is the next candidate. Our autosampler carryover guide covers measurement, acceptance limits and wash design in detail.

Why does pressure spike when the sample is injected?

First, separate normal from abnormal. When a loop that has been filled at atmospheric pressure is switched into the flow, the pump has to compress its contents, so a small, repeatable dip at injection is normal. Agilent's pump documentation lists the compressibility of water as 46 × 10⁻⁶ per bar, with methanol at 120 × 10⁻⁶ and acetonitrile at 115 × 10⁻⁶ (Agilent). For a 100 µL loop of water switched into a 300 bar stream:

ΔV ≈ V_loop × compressibility × pressure
   ≈ 100 µL × 46 × 10⁻⁶ bar⁻¹ × 300 bar ≈ 1.4 µL
At 1 mL/min (16.7 µL/s) the pump makes that up in about 0.08 s.

A brief dip of the same size at every injection is expected. A rise that lasts, grows from run to run, or appears only with some samples is a fault.

Likely causes of a real spike, most probable first:

1. A partial blockage in the part of the flow path that only carries flow in the inject position: the needle seat capillary, the loop or a rotor seal groove. Particles from the sample or the vial cap get drawn in and are pushed into the narrow seat capillary when the valve switches (Agilent KP673). Debris from a wearing rotor seal lodges there too (Stoll).
2. A viscous sample plug. Pressure scales with viscosity, so a large plug of a more viscous solvent (DMSO, for example) raises the pressure while it passes through the column.
3. Precipitation, where the sample or its buffer is not soluble in the mobile phase.

Quick test. With flow on, switch the valve between bypass and inject (mainpass) and compare the readings. Agilent's procedure: if pressure returns to normal in bypass, the blockage is in the autosampler path; if it stays high, look at other modules (KP100). Then inject mobile phase from a vial. A spike only with real samples points to viscosity or precipitation.

Fix. A needle seat capillary can be backflushed, but Stoll found that the relief is usually temporary, with the capillary often clogging again within days, so his lab simply replaces it. Replace a scratched or blocked rotor seal and inspect the stator. Filter or centrifuge samples, and dilute viscous ones in a solvent closer to the mobile phase.

How do I get air out of the syringe or metering device?

Likely causes, most probable first:

1. The draw is too fast for the sample's viscosity. An Agilent Vialsampler help page warns that "if the draw speed is too high, air bubbles may form in the sample plug, affecting precision" (Agilent).
2. A partial vacuum in the vial. Non-slit septa can seal tightly around the needle, so air cannot enter as sample is withdrawn, which can cause over-aspiration and inaccurate volumes (MicroSolv).
3. Wash or flush solvent that is not degassed, which outgasses when the plunger retracts (general practice).
4. A loose fitting at the syringe or metering head, or a worn plunger seal, which pulls in air on every draw stroke.
5. Too little sample, especially in low-volume or tapered vials. Shimadzu's troubleshooting course advises inspecting such vials for bubbles (Shimadzu).

Quick test. Purge the metering line and watch it. A bubble that appears at the same joint on every stroke is a leak at that fitting or seal. Bubbles that appear only after drawing from capped vials, and not from an uncapped vial, point to vacuum. Bubbles only with viscous samples point to draw speed.

Fix. Degas the wash solvent, purge the needle and syringe lines (Dolan), tighten or replace the fitting, replace the syringe seal, switch to pre-slit septa, and slow the draw.

Leaks at the valve, needle seat or injection port

Likely causes, most probable first:

1. A worn needle and needle seat. Agilent notes that the needle wears against the seat in normal operation until both need replacing, and recommends changing them together to prevent premature leakage (KP141).
2. A worn rotor seal. Stoll describes two kinds of leak: an external one, where liquid or a white salt crust appears on the valve body, and an internal cross-port leak that lets mobile phase into the needle path, seen as small puddles on top of vials after a draw.
3. Loose or over-tightened fittings at the valve ports.
4. The syringe or metering seal.

Quick test. With flow on, hold a dry tissue against each joint for a few seconds. Look for salt crystals around the seat and valve. If you suspect a leak you cannot see, measure the flow: collect the column outlet into a 10 mL volumetric flask at 1.00 mL/min and time it. A fill time of 10.4 min instead of 10.0 min means about 4% of the flow is going somewhere else; repeat at the pump outlet to see whether the loss is before or after the pump. In one LCGC case the pump delivered the right flow but a badly worn rotor seal was leaking mobile phase out through the valve's waste port (LCGC).

Fix. Replace the needle and seat as a pair. Replace the rotor seal and check the stator face, replacing it too if it is marked. Tighten stator screws evenly to the maker's torque; for its 1100-1290 series samplers Agilent gives 2.25 Nm and warns that over-tightening can stop the valve turning (KP818). Because rotor wear cannot be prevented, Stoll recommends scheduled replacement based on injections per day, pH, sample cleanliness and pressure. More on valve wear in the 6-port injection valve explained.

What blocks an autosampler needle?

Likely causes, most probable first:

1. Septum fragments cored out by the needle (Dolan).
2. Particles in the sample, or debris from the vial cap (Agilent KP673).
3. Dried buffer or precipitated sample in the needle tip, often after an idle weekend.
4. A bent or crimped needle after a crash.

Quick test. Draw from an uncapped vial of clean solvent. If the syringe moves and the level does not drop, the needle path is blocked. On flow-through-needle samplers, follow the bypass and mainpass pressure test above, then flush the needle and loop as Agilent describes.

Fix. Replace the needle, or backflush it if the maker gives a procedure. Filter or centrifuge samples; Dolan also suggests a 0.5 µm in-line filter upstream of the autosampler to trap particles from the pump or mobile phase. Flush buffers out with water before shutting down.

Why is the septum coring?

Likely causes, most probable first:

1. A damaged needle tip that cuts the septum instead of piercing it (Dolan).
2. A septum too hard or thick for the needle. Thermo Fisher advises lower-hardness materials to reduce needle problems, soft silicone laminated to PTFE for fine-gauge needles, and PTFE/silicone/PTFE for large, blunt-tip needles, which it rates as the most coring-resistant (Thermo Fisher).
3. Many punctures in the same spot from repeated injections.

Quick test. Look at the underside of a used septum and at the vial bottom for fragments, and inspect the needle tip with a magnifier. If vials lift with the needle, the septum is gripping it: an Agilent support answer suggests testing an injection without the septum, and pre-slit septa if that cures it (Agilent Community).

Fix. Replace a burred needle, then switch to softer or pre-slit septa. Pre-slit septa reduce penetration force and coring, but a slit is a permanent opening, so watch for evaporation of volatile solvents on long sequences. See HPLC vials, caps and septa for the trade-offs.

Needle crashes and vial-not-found errors

Likely causes, most probable first:

1. The method or tray setup does not match the hardware: wrong vial or rack type, vials in other positions than the sequence expects.
2. Vials not fully seated, tilted in the rack, or caps not fully screwed or evenly crimped.
3. Vials outside the sampler's specification. Inconsistent internal bottoms can cause needle-positioning errors and, on older systems, needle damage (MicroSolv).
4. A bent or damaged needle. Waters traced a persistent "Missing plate or rack" error on a 2707 autosampler to a damaged needle; the fix was a new needle and recalibration (WKB7624).
5. Mechanical faults in the positioning drive or its sensor. For a 717plus "could not position to vial" error that survives a power cycle and a tray swap, Waters points to the carriage reader or carriage drive (WKB3808).

Quick test. Inspect the needle tip and shaft. Run a short sequence of capped vials filled with solvent at several positions, then repeat with vials from another lot. If the error follows the vial lot, it is the vials; if it follows a tray position, it is the tray or its calibration.

Fix. Correct the vial and rack definitions, reseat the rack, buy vials that meet the instrument maker's dimensions, and recalibrate after every needle change. Newer designs add sensors to stop a crash early: the Elutrolab AS-48 prototype, for example, includes needle crash detection and a door sensor in its design (specifications).

Why does the sequence stop part-way?

Likely causes, most probable first:

1. A pressure limit tripped: the upper limit from a blockage, the lower limit from a leak or an empty solvent bottle. Agilent's high-pressure procedure starts by checking that the limits suit the method (KP100).
2. A leak sensor tripped, often by a small drip at the needle seat or valve.
3. A missing vial or position error with the sequence set to abort on missing vials.
4. An empty wash bottle or a full waste container, on samplers that monitor them.
5. The door opened during a run, or a communication time-out between the module and the software.

Quick test. Read the instrument logbook at the exact stop time and note the injection number. If stops always come at the same sequence line, look at that vial and position; if they come after a similar number of injections, look at consumables (wash, waste, solvent); if the pressure trace climbs before each stop, go back to the blockage section.

Fix. Treat the root cause, and set pressure limits with some margin around normal: a lower limit just below the lowest pressure of your gradient catches a dry solvent line without tripping on normal variation.

Why are my early peaks broad or split after injection?

Likely causes, most probable first:

1. Sample solvent stronger than the mobile phase. Stronger solvents flush the sample rapidly through the first part of the column and "can cause broad, split, doubled, or distorted peaks" (Dolan, LCGC).
2. Injection volume too large for the first peak. A small volume of a strong solvent does less harm than a large volume of the same solvent (Dolan and Alsehli).
3. The first peak elutes too close to the void. Dolan aims for the first peak of interest at k ≥ 1, better k ≥ 2.
4. Not the injection at all: a void or blocked frit at the column inlet, which usually distorts every peak.

Quick test. Halve the injection volume, or dilute the sample 1:1 with water or the weak mobile phase component (Dolan and Alsehli suggest a twofold change). If the early peaks sharpen, the injection is the cause. If all peaks stay distorted, look at the column and fittings.

Worked example: how much can you inject before early peaks suffer?

Dolan's rule of thumb is that an injection in mobile phase should be under 15% of the first peak's volume for less than 1% resolution loss (up to about 40% is acceptable when resolution is well above 2). The peak volume at baseline is about four standard deviations:

V_peak ≈ 4 × V_m × (1 + k) / √N
V_max  ≈ 0.15 × V_peak

In Dolan's 2015 case, V_m ≈ 0.68 mL, k ≈ 0.02 and N ≈ 15,000:

V_peak ≈ 4 × 680 µL × 1.02 / 122 ≈ 22.7 µL
V_max  ≈ 0.15 × 22.7 ≈ 3.4 µL

That matches the article. Now take a 150 × 4.6 mm column with V_m ≈ π × (0.23 cm)² × 15 cm × 0.68 ≈ 1.7 mL and assume N = 10,000:

The same column tolerates almost three times more volume once the first peak moves from k = 0.5 to k = 3. A sample in a solvent stronger than the mobile phase tolerates less than these figures.

Fix. Dissolve the sample in the starting mobile phase or something weaker, inject less, or increase retention of the first peak.

A short preventive routine

Most of the faults above build up slowly, so a routine check catches them before they cost a sequence:

1. Each sequence: a blank after the highest standard, and bracketing check standards.
2. Weekly, or after any service: six injections of a check standard (area RSD), and the system pressure in bypass and inject recorded against your baseline.
3. Monthly: inspect septa and the needle tip, top up and degas wash solvents, clean salt from the seat and valve.
4. On a schedule set by injection count: needle and seat together, rotor seal, syringe seal.

The tests used to accept a new instrument are a good template for this routine; see autosampler qualification.