HPLC Method Transfer to a Narrow-Bore Column: Reading a Plate-Count Collapse
A contract lab is transferring an isocratic reversed-phase assay for a small-molecule API to a narrow-bore column to cut solvent use. Original system: 150 × 4.6 mm, 5 µm C18, 1.0 mL/min, 50:50 (v/v) acetonitrile–20 mM phosphate buffer pH 6.8, 30 °C, 214 nm; main peak k′ = 2.8, N = 7,000, USP tailing 1.05, resolution 2.5 from the nearest specified degradant (specification ≥ 2.0). New system: 100 × 2.1 mm, 2.7 µm core–shell C18 of the same chemistry, 0.4 mL/min, same mobile phase, 4 µL injection, same detector cell; main peak k′ = 2.6, retention 2.1 min, N = 3,400, tailing 1.85, resolution 1.7, backpressure 480 bar. The autosampler uses a 4 µL loop with 0.010-inch tubing and a standard 12 µL detector cell; the sample diluent is 90:10 water–acetonitrile. The analyst proposes raising acetonitrile to 55% 'to sharpen the peaks and get the resolution back'.
1. Is the analyst's 55% acetonitrile proposal sound? Support your answer with the resolution equation.
Read the worked answer
Back-calculate selectivity from the original data. Rs = (√N/4)(α − 1)[k/(1 + k)]. For the original: 2.5 = (83.67/4)(α − 1)(2.8/3.8) → 2.5 = 20.9 × 0.7368 × (α − 1) → (α − 1) = 0.162, α ≈ 1.16.
Predict the new column with the same chemistry. 1.16 is fixed by the stationary/mobile phase pair, so: Rs = (58.31/4)(0.162)(2.6/3.6) = 14.58 × 0.162 × 0.722 = 1.71, which matches the observed 1.7. The loss of resolution is entirely a loss of N; selectivity has not changed.
Effect of 55% acetonitrile. Adding 5 points of organic to a 50:50 method typically halves k′ for a small molecule, to about 1.3. k/(1 + k) falls from 0.722 to 0.565. Even granting a 30% gain in plates (√1.3 = 1.14), Rs ≈ 1.71 × (0.565/0.722) × 1.14 ≈ 1.5 - worse, not better. Selectivity will usually erode as well, because the pair is pushed toward the void.
Verdict. Reject the proposal. Peaks would look 'sharper' in height because they elute earlier, but the separation that matters degrades. The correct lever is to recover the lost plate count (band broadening), and secondarily to raise k′ by lowering acetonitrile toward 45%, which increases k/(1 + k) at essentially unchanged α.
2. Expected plate count for the new column is well above 7,000, yet 3,400 was measured. Rank the plausible root causes and give the experiments that discriminate between them.
Read the worked answer
Expected performance. Scaling N ≈ (L/dp): 7,000 × (100/150) × (5/2.7) ≈ 8,600, and core–shell particles usually exceed that. Observed 3,400 is a factor of 2.5–3 low.
Variance budget. σ_obs = tR/√N = 2.1/58.3 = 0.036 min = 2.16 s ≡ 14.4 µL at 0.4 mL/min. Column-only (N = 10,000): σ = 0.021 min ≡ 8.4 µL. Variances add: σ_ec² = 14.4² − 8.4² = 137 µL², σ_ec ≈ 11.7 µL. That is roughly 65% of the observed variance and several times what a correctly plumbed narrow-bore system gives (σ_ec < 4 µL). Tailing rising from 1.05 to 1.85 is the same signature.
Ranked causes.
- Extra-column volume - 0.010-inch tubing (long runs, poorly cut connections), a 12 µL cell on a 2.1 mm column, and the injection loop/needle seat. Most likely, given the variance number.
- Detector time constant / data rate set too slow for a 2 s-wide peak (a classic transfer oversight).
- Injection volume too large relative to the early eluting peak (4 µL is modest, but the peak volume is only ~100 µL).
- Column damage or a badly packed bed - least likely for a new column.
- The strong-diluent effect is ruled out: 90:10 water–acetonitrile is weaker than the 50:50 mobile phase, which focuses rather than broadens the band.
Discriminating experiments. (a) Replace the column with a zero-dead-volume union and inject a small volume of tracer - the measured peak width is the system variance. (b) Halve the injection volume; no change means the loop is not the problem. (c) Re-plumb with 0.005-inch tubing, shortest possible path, and re-measure N. (d) Raise the detector sampling rate to ≥ 20 points across the peak. (e) Run the same sample on the old 4.6 mm column at the same linear velocity to confirm the chemistry is unchanged.
3. State the acceptance criteria you would apply to the new system and the minimum change needed to get resolution back to ≥ 2.0.
Read the worked answer
Acceptance criteria for the transfer.
- Extra-column variance ≤ 10% of the column variance (equivalently σ_ec ≤ 0.3 σ_col); for this column that means σ_ec ≤ 3 µL, i.e. well under 10 µL² of added variance.
- Injection volume ≤ 15% of the volume of the first peak of interest (here ≈ 100 µL, so ≤ 15 µL - the 4 µL injection is comfortable).
- Detector cell ≤ 3 µL and tubing ≤ 0.005 inch for 2.1 mm columns; data rate ≥ 20 points per peak.
- Confirmed tailing ≤ 2.0 and N within 30% of the L/dp prediction.
Minimum change to restore Rs ≥ 2.0. Since α is unchanged at 1.16 and k′ is adequate, only N must move: Rs = (√N/4)(0.162)(0.722) requires √N ≥ 4 × 2.0/(0.162 × 0.722) = 68.4, i.e. N ≥ 4,700. Fixing the plumbing to recover a realistic 8,000–10,000 plates gives Rs ≈ 2.4–2.7 with no change to the mobile phase, which is the cleanest, most defensible outcome and requires no revalidation of selectivity.
If plumbing alone cannot reach 4,700 plates, the fallback is to lower acetonitrile to ~45%: k′ rises to ~4, k/(1 + k) → 0.80, and Rs ≈ 1.71 × (0.80/0.722) ≈ 1.9–2.0, marginally passing but leaving no margin for column ageing. The robust answer is to fix the band broadening first and treat the mobile-phase adjustment as a secondary robustness lever.