troubleshoot-separation

pjt222

更新于 2 days ago

4 次查看

文档word

关于

This Claude Skill systematically diagnoses and resolves GC and HPLC chromatographic separation issues. It helps document symptoms, identify root causes for peak shape and retention anomalies, and evaluate matrix effects. Developers should use it to implement targeted fixes via a structured, one-variable-at-a-time approach.

快速安装

Claude Code

技能文档

Troubleshoot Chromatographic Separation

GC/HPLC sep diagnosis → symptom doc → peak shape diag → retention anomaly → matrix effects → verified fix via one-var-at-time.

Use When

Peaks tail|front|split|broaden
Retention shifted|irreproducible
Resolution between critical pairs degraded
Baseline drift|ghost peaks|negative peaks
Sensitivity drop|S/N worse
Working method now fails sys suitability

In

Required

Problem chromatogram: Current data → issue
Reference chromatogram: Recent good, same method → compare
Method conditions: Column, mobile phase|carrier gas, temp|gradient, detector, flow
System log: Recent maint, col changes, mobile phase prep, instrument events

Optional

Blank chromatogram: Most recent blank|solvent inj
Sys suitability trends: Historical tailing, resolution, plates, RT
Col history: # injections, sample types, age
Instrument maint log: Pump seal, lamp hrs, detector service

Do

Step 1: Doc Problem

Symptom precise: which peaks, how differ from ref
When started: gradual|sudden
All peaks or specific?
Standards|samples|both?
Current sys suitability vs historical
Photo|export problem chromatogram side-by-side w/ ref

Got: Problem statement w/ timeline, scope (all|specific peaks, std|sample), ref comparison.

If err: No ref → inject fresh standard prep under documented method → establish current baseline before troubleshoot.

Step 2: Diag Peak Shape

Symptom	Possible Causes	Solutions
Tailing (T > 1.5)	Secondary interactions (silanol activity), dead volume in fittings, contaminated column frit, overloaded active sites	Add amine modifier (HPLC), deactivate liner (GC), replace frit, trim column inlet, reduce injection mass
Fronting (T < 0.8)	Column overload (mass or volume), mismatch between sample solvent and mobile phase strength	Reduce injection volume or concentration, dilute in weaker solvent, use larger-bore column
Split / double peaks	Partially blocked frit, void at column head, two polymorphic forms, isomeric interconversion	Replace frit, repack column head, verify sample stability, adjust pH to lock one form
Broad peaks (all)	Extra-column band broadening, wrong tubing ID, large detector cell, old column, low plate count	Minimize post-column tubing length and ID, check connections, replace column
Broad peaks (early eluters)	Poor focusing at column head, injection solvent too strong (HPLC), cold on-column (GC)	Use weaker injection solvent, reduce injection volume, increase initial oven temp
Broad peaks (late eluters)	On-column diffusion, temperature too low (GC), insufficient gradient steepness (HPLC)	Increase final oven temperature, steepen gradient, add organic wash
Negative peaks	Sample solvent refractive index/absorbance differs from mobile phase, vacancy peaks (IEX)	Match sample solvent to mobile phase, use different detection wavelength
Ghost peaks	Carryover from previous injection, contaminated mobile phase, column bleed, septum bleed (GC)	Run blank to confirm, clean or replace injection system, filter/degas mobile phase, replace septum
Baseline drift (upward)	Column bleed (GC at high temp), gradient elution baseline shift (HPLC), lamp instability (UV)	Reduce max temp, use low-bleed column (GC), run blank gradient to characterize (HPLC), replace lamp
Baseline noise (high-frequency)	Electrical interference, pump pulsation, air bubbles in detector, contaminated detector	Ground instrument, replace pump seals, degas mobile phase, clean detector cell

Match symptom(s) → table
Narrow causes: all peaks|specific, sudden|gradual
Prioritize → sys history (recent changes, col age, maint)

Got: 1-2 most-likely causes from sym-cause map, prioritized by history.

If err: No table match or multi-symptom → compound problem (col degradation + leak). Address obvious first → re-eval.

Step 3: Diag Retention

Symptom	Possible Causes	Solutions
All peaks shifted earlier	Increased flow rate, higher column temperature, stronger mobile phase, column void	Check flow rate setting and actual delivery, verify temperature, remake mobile phase, inspect column
All peaks shifted later	Decreased flow rate, lower column temperature, weaker mobile phase, partially blocked tubing	Check for leaks (pressure drop), verify temperature, remake mobile phase, check inline filter
Retention time drift (gradual)	Column degradation, mobile phase evaporation (open reservoir), temperature fluctuation	Replace column, seal reservoir, stabilize oven, use column thermostat
Retention time irreproducible	Leak at fitting, check valve malfunction, autosampler timing error, inadequate re-equilibration	Pressure-test fittings, replace check valves, verify autosampler, increase equilibration volume
Lost retention (k' near 0)	Phase collapse (RP at high aqueous), column dewetting, wrong mobile phase, reversed connections	Use polar-embedded or AQ-type column, re-wet column with organic, verify mobile phase, check plumbing
Co-elution (previously resolved)	Column selectivity lost (bonded phase stripped), mobile phase composition changed, temperature changed	Replace column, verify mobile phase preparation, check temperature setpoint vs. actual

Shifts uniform (all) or selective (specific)?
Uniform → systematic (flow, temp, mobile phase)
Selective → col chemistry|specific analyte
Check pressure trace: sudden change → leaks|blockage
Re-inject ref std → confirm sys vs sample

Got: Retention root cause categorized: systematic (instrument/MP) or col-related.

If err: Re-inject std on new col resolves → orig col = problem. Persists on new col → upstream (MP, instrument, method).

Step 4: Matrix Effects

Std vs sample chromatogram:
- Extra peaks in sample absent from std?
- Baseline elevated|noisy in retention windows?
- Analyte peaks differ in sample vs std (broader, more tailing)?
LC-MS → ion suppression/enhancement:
- Post-col infusion: infuse analyte continuous while inject blank matrix → dips = suppression regions
- Analyte RT coincides w/ suppression → shift elution
Col contamination:
- Solvent blanks after sample seq → persistent peaks = col contamination
- Flush col w/ strong solvent (100% organic for RP, or per mfr)
Sample prep:
- Dirty injector (autosampler needle, GC liner) → replace|clean
- Insufficient cleanup → add filter, SPE, protein precipitation
GC: non-volatile residue in inlet liner → causes tailing + ghost over time

Got: Matrix effects characterized (interferents, ion suppression zones for LC-MS, col contamination) w/ actions.

If err: Can't characterize → matrix-matched cal curve vs solvent cal curve. Slope diff > 15% → significant matrix effects → method mod.

Step 5: Fix + Verify

One var at a time. Doc what + why.
Per change → re-inject sys suitability std vs ref
Sequence (least → most disruptive):
- Fresh mobile phase | carrier gas tank
- Replace consumables (septum, liner, frit, inline filter, lamp)
- Tighten|replace fittings + tubing
- Flush|regen col
- Adjust method (temp, flow, gradient, pH)
- Replace col
- Service instrument (pump seals, check valves, detector)
Fix found → full sys suitability test (n >= 5 injections)
Compare all params (RT, area, resolution, tailing, plates) vs spec
Doc root cause, action, verify in instrument|col logbook
Recurs → preventive maint schedule

Got: Problem resolved, sys suitability params restored to spec. Root cause + action + verify documented.

If err: All single-var changes fail → multi simultaneous failures. Replace all consumables + col together → verify w/ fresh std → rebuild from new baseline. Persists after total replacement → escalate to instrument service.

Check

Problem documented w/ symptom, timeline, scope
Root cause via sym-cause maps
One var at a time
Fix verified by sys suitability (n >= 5 replicate inj)
All sys suitability params restored to spec
Root cause + action in logbook
Preventive measure ID'd

Traps

Multi vars at once: Can't ID root cause. One change → test → decide.
Col first: Expensive, masks real problem (leak, wrong MP, contaminated inlet). Exhaust simpler first.
Ignore logbook: Many problems → recent maint, MP batch, col swap. Check what changed.
Blame sample w/o evidence: Run ref std first. Std also shows problem → sys, not sample.
Incompatible solvents: Never flush RP w/ pure water (phase collapse) or silica HILIC w/ pure aqueous (irreversible). Per mfr protocol.
No documentation: Failed attempts valuable. Record every change + outcome → avoid repeat + build knowledge.

→

interpret-chromatogram — understanding chromatographic data revealing sep problems
develop-gc-method — GC method dev, relevant when troubleshooting requires redesign
develop-hplc-method — HPLC method dev, relevant when troubleshooting requires redesign
validate-analytical-method — re-validation may need after significant method changes

GitHub 仓库

pjt222/agent-almanac

路径: i18n/caveman-ultra/skills/troubleshoot-separation

agentsagentskillsai-assisted-developmentclaude-codeskillsteams