Column Chromatography

How a column separates a mixture

Pour a slurry of silica gel into a glass column. Tap to settle, top with a layer of sand, then load your dissolved mixture as a thin band on top. Open the stopcock and let solvent (the eluent) drip through. Each compound in the mixture spends some fraction of its time stuck to the silica and some fraction dissolved in the moving solvent. Compounds that bind weakly travel almost as fast as the solvent. Compounds that bind tightly lag behind. After enough column volumes, the bands separate.

The principle is a partition equilibrium repeated thousands of times. At any instant, a molecule is either adsorbed on a silanol (Si-OH) site or dissolved in the mobile phase. The fraction of time spent moving determines the velocity:

v_compound = v_solvent × (1 / (1 + k))

where k = (mass on stationary phase) / (mass in mobile phase)

A compound with k = 4 (80% adsorbed) moves at one-fifth the solvent speed. A compound with k = 0.1 nearly keeps up. Different functional groups give different k values on silica — that's the whole game.

The apparatus

      ┌─────────────┐  ← solvent reservoir
      │    eluent   │
      └─────┬───────┘
            │ drip
      ╔═════╧═════╗
      ║ . . . . . ║   sand top layer
      ║░░░░░░░░░░░║   ← sample band (loaded as thin disc)
      ║▓▓▓▓▓▓▓▓▓▓▓║
      ║▓▓ silica ▓▓║   stationary phase
      ║▓▓▓▓▓▓▓▓▓▓▓║
      ║ . . . . . ║   sand bottom (supports bed)
      ║▓▓▓▓▓▓▓▓▓▓▓║
      ╚═════╤═════╝
            │ stopcock
            ▼
       fractions
       in test tubes

The sand caps protect the silica bed from disturbance when you add fresh solvent and from migration into the stopcock at the bottom. The sample is loaded either as a concentrated solution (wet-load, fast but risks streaking) or pre-adsorbed onto a small mass of silica that you dry-load on top (slower but gives the cleanest band).

Worked example — interpreting R_f

You ran a TLC plate with your reaction mixture before committing to a column. The solvent front travelled 8.0 cm. Three spots are visible:

• Starting material at 1.6 cm → R_f = 1.6 / 8.0 = 0.20
• Desired product at 3.2 cm → R_f = 3.2 / 8.0 = 0.40
• Side product at 5.6 cm → R_f = 5.6 / 8.0 = 0.70

The product sits in the sweet spot. Translating to the column: the eluent is well-chosen, but you can collect the side product first (it will elute fastest), wait until the product band exits, then change to a stronger solvent to flush off the starting material. A fraction collector with 5–10 mL test tubes lets you spot every other tube on TLC and pool only the pure-product fractions. A 20 g column run at 10 mL/min with this eluent typically takes 12–18 minutes total.

Five column modes compared

Mode	Stationary phase	Separates by	Typical eluent	Best for	Cost / column
Normal-phase silica	Si-OH on porous silica	Polarity (polar = retained)	Hexane → ethyl acetate gradient	Synthetic organic mixtures	$3–10 (10 g)
Alumina (basic/neutral)	Al₂O₃ activated	Polarity, but tolerates bases	Hexane / DCM / methanol	Amines, basic alkaloids	$8–25 (10 g)
Reverse-phase (C18)	C18-alkyl-bonded silica	Hydrophobicity (lipophilic = retained)	Water → methanol / acetonitrile	Peptides, polar drugs, natural products	$30–80 (10 g)
Ion-exchange	Sulfonate or quaternary-ammonium beads	Net charge at given pH	Buffer with rising salt gradient	Proteins, nucleotides, amino acids	$50–150 (10 mL resin)
Size-exclusion (gel filtration)	Crosslinked dextran or agarose beads	Hydrodynamic radius (large = first)	Aqueous buffer, isocratic	Protein desalting, polymer MW	$80–200 (10 mL resin)
Affinity	Immobilized ligand (His-tag Ni²⁺ etc.)	Specific binding	Wash buffer + elution buffer	Recombinant proteins	$120–400 (5 mL resin)

Normal-phase silica handles ~80% of synthetic organic chemistry separations. Reverse-phase dominates pharmaceutical and biomolecule work because water-compatible. Size-exclusion is unusual: nothing binds the stationary phase — large molecules thread past the bead pores and elute first; small molecules detour through the pores and elute last.

Solvent strength and gradients

The eluotropic series ranks solvents by their ability to displace adsorbed compounds from silica, from weakest to strongest:

hexane < toluene < DCM < diethyl ether < ethyl acetate
       < acetone < ethanol < methanol < water

A typical synthetic column starts with 90:10 hexane:ethyl acetate, ramps to 70:30, then 50:50, holding each stage for 2–3 column volumes. This is a step gradient — easier than a true linear gradient and almost as resolving. For automated flash systems, a 1–80% ethyl acetate linear ramp over 15 column volumes is the workhorse.

On reverse-phase, polarity inverts: water is the weak solvent, methanol and acetonitrile are strong. A 5% → 95% acetonitrile in water gradient covers nearly all organic small molecules.

Variants and adjacent techniques

• Flash chromatography — 40–63 µm silica with 5–15 psi gas pressure. Standard since Still's 1978 paper, now usually run on automated boxes (Biotage, Combiflash) with UV detection.
• Dry-column flash — short fat column, vacuum-pulled. Cheap and fast for crude prep.
• Medium-pressure liquid chromatography (MPLC) — 15 µm silica, 50–300 psi. Bridge between flash and HPLC.
• Preparative HPLC — 5 µm silica, >1000 psi. Higher resolution than flash but expensive; reserved for milligrams of high-value compound.
• Centrifugal partition chromatography — no solid support; two immiscible liquid phases. Useful for crude natural-product extracts that destroy silica columns.

Common pitfalls

• Letting the bed run dry. Air entering the silica creates channels that ruin resolution. Always keep solvent above the top.
• Overloading. More than ~5% sample-to-silica by mass causes bands to overlap. If you need to load more, scale the column or split the run.
• Cracks in the bed. Tap the column gently while packing; pour the slurry quickly to avoid layered settling.
• Running too fast. Equilibration takes time. A 20 g column at >15 mL/min often shows worse resolution than at 8 mL/min.
• Using too-volatile eluent. Diethyl ether evaporates from the reservoir and changes solvent strength mid-run; cap or chill the reservoir.
• Silica killing acid-sensitive compounds. Boc groups, silyl ethers, and ketals can hydrolyze on the bed. Switch to neutral alumina or buffer the eluent.

When to skip the column

If your product is volatile and the impurity is not, distillation is faster. If your product crystallizes preferentially, recrystallization gives higher purity in fewer steps. If your mixture has only two components and one is very polar (like a salt), an acid–base extraction in a separating funnel finishes in five minutes — no silica required. The decision tree: extraction → recrystallization → distillation → column. Reach for the column only when the cheaper methods can't get you to purity.