Gas Chromatography

How a GC separates molecules

You inject 1 µL of a liquid mixture into a heated injection port. The liquid flashes to vapor in milliseconds. An inert carrier gas (helium, hydrogen, or nitrogen) sweeps the vapor onto the head of a long, thin capillary column whose inside wall is coated with a thin film of viscous liquid called the stationary phase. The column sits inside an oven that ramps from 40 °C to 300 °C during the run.

Each component partitions between the moving carrier gas and the stationary liquid film. Molecules with higher boiling points and higher affinity for the film spend more time dissolved in it and lag behind. Molecules with lower boiling points spend most of their time in the gas phase and arrive first. After 5–30 minutes, separated peaks emerge from the end of the column and hit the detector, which generates a chromatogram of signal vs. time.

The instrument layout

  carrier-gas tank (He / H₂ / N₂)
         │
         ▼
   ┌──────────────┐    ┌─────────────────────┐    ┌────────────┐
   │   inlet      │───▶│  column oven 40–350°C │───▶│  detector  │──▶ data
   │  injector    │    │  ╔═══════════════╗  │    │ FID/MS/ECD │
   │  250–280 °C  │    │  ║  capillary    ║  │    └─────┬──────┘
   │  split/      │    │  ║   30 m × 0.25 ║  │          │
   │  splitless   │    │  ║   mm × 0.25 µm║  │          ▼
   │              │    │  ╚═══════════════╝  │     chromatogram
   │  syringe ↓   │    └─────────────────────┘
   │   1 µL       │
   └──────────────┘

The capillary itself is a 0.25 mm inner diameter fused-silica tube, 30 m long, coiled into a 20 cm hoop that fits inside the oven. The interior is coated with a 0.25 µm film — the famous "30 × 0.25 × 0.25" geometry that's sold as nearly every general-purpose GC column.

Worked example — calculating resolution

You injected a hydrocarbon mixture and got these adjacent peaks on a 30 m DB-5 column:

• n-decane: retention time t₁ = 7.42 min, baseline width w₁ = 0.18 min
• n-undecane: retention time t₂ = 8.15 min, baseline width w₂ = 0.20 min

Resolution: R = 2 × (8.15 − 7.42) / (0.18 + 0.20) = 2 × 0.73 / 0.38 = 3.84. This is severely over-resolved — anything above 1.5 is already baseline-separated. You're wasting time. Possible speedups: ramp the oven faster (15 °C/min instead of 5 °C/min), use a shorter 15 m column, or raise the carrier flow rate from 1 mL/min to 1.5 mL/min. Each cuts run time roughly in half while keeping R well above 1.5.

Now suppose two coeluting isomers gave t₁ = 6.80, w₁ = 0.10, t₂ = 6.92, w₂ = 0.10. Then R = 2 × 0.12 / 0.20 = 1.20 — only partially resolved. Fixes: longer column (R scales as √length), lower oven temperature for that section (better partitioning), or switch to a more polar stationary phase that interacts differently with the two isomers.

GC vs GC-MS vs HPLC vs HPLC-MS

	GC-FID	GC-MS	HPLC-UV	HPLC-MS	UPLC-MS	Headspace-GC
Mobile phase	He / H₂ gas	He gas	Liquid solvent	Liquid solvent	Liquid solvent	He / H₂ gas
Sample requirement	Volatile, <500 Da	Volatile, <500 Da	Soluble, <2000 Da	Soluble, <2000 Da	Soluble, <2000 Da	Volatile from matrix
Typical run time	5–30 min	10–40 min	10–40 min	10–30 min	2–8 min	15–45 min
Detection limit	~1 pg	~10 fg (SIM)	~10 ng	~10 pg	~1 pg	~10 ppt in headspace
Identifies unknowns?	No (only retention time)	Yes (NIST library)	No	Yes (with standards)	Yes	Depends on detector
Hardware cost (USD)	$15–40k	$60–120k	$25–60k	$120–250k	$200–400k	$30–60k
Best for	Petroleum, solvents	Forensics, drug screen	Pharma QC, dyes	Metabolomics, peptides	High-throughput pharma	BAC, residual solvents

Headspace-GC is the technique that legally measures blood alcohol — it samples only the vapor above the liquid, which contains a known fraction of any volatile present, and avoids injecting the messy biological matrix onto the column.

Detector zoo

• FID (flame ionization) — universal for organics, picogram sensitivity, six-decade linear range. Cannot detect water, CO, CO₂, formaldehyde well, and is destructive.
• TCD (thermal conductivity) — universal but ~1000× less sensitive than FID. Non-destructive, the only common detector that sees H₂, He, and noble gases.
• ECD (electron capture) — selective for halogenated and nitro compounds. Detects pesticides like DDT at femtogram levels.
• NPD (nitrogen-phosphorus) — selective for N- and P-containing compounds. Used for drug screening.
• MS (mass spectrometer) — fragments analytes by electron impact (70 eV) and counts ions at each m/z. The gold standard for identification.

Stationary phases

Modern capillary columns use cross-linked silicone polymers. Polarity ranges from non-polar (100% dimethyl siloxane, "DB-1") through 5% phenyl ("DB-5", the most universal choice) to highly polar polyethylene glycol ("Wax", "DB-WAX") for fatty acids and aromatic alcohols. A typical synthesis lab keeps three columns in rotation: DB-5 for general work, DB-WAX for polar oxygen-containing compounds, and a chiral phase like Chirasil-Dex for enantiomer separations.

Variants and adjacent techniques

• GC-MS/MS (triple quadrupole) — adds a collision cell and second mass filter. Eliminates matrix interference; routine in pesticide regulation at parts-per-trillion.
• GCxGC (two-dimensional GC) — separates the effluent of one column on a second, orthogonal column. Resolves 5,000+ peaks in petroleum samples.
• Pyrolysis-GC — flash-heats a polymer to 600 °C, separates the fragments. Identifies plastics in microplastic studies.
• Solid-phase microextraction (SPME) — a coated fiber concentrates analytes from aqueous or vapor samples without solvent. Used for environmental and food-flavor analysis.

Common pitfalls

• Overloading the column. A 0.25 mm column saturates around 50–100 ng per peak. Above that, peaks fronting and retention times shift.
• Cold spots in the transfer line. Any unheated section condenses analytes, giving ghost peaks on the next run.
• Septum bleed. Old septa shed siloxanes; you'll see TIC peaks at m/z 207, 281 in the MS.
• Column bleed at high temperature. Past the column's max isothermal limit, the stationary phase decomposes and raises baseline. Stay 30 °C below the rated maximum for long life.
• Active sites in the inlet liner. A dirty liner adsorbs polar analytes and gives non-reproducible peak areas. Replace liners every few hundred injections.

The van Deemter curve in one paragraph

Plate height H (a measure of band spreading per unit column length) varies with linear gas velocity u as H = A + B/u + C·u. The A term (eddy diffusion) is zero for capillaries. The B term (longitudinal diffusion) dominates at low velocity. The C term (mass-transfer kinetics) dominates at high velocity. The minimum gives the optimal flow. For helium, optimum u ≈ 25 cm/s; for hydrogen, u ≈ 40 cm/s and the curve is flatter — that's why hydrogen runs faster without losing resolution.