Getting HPLC to actually work on the bench
High Pressure Liquid Chromatography is one of those techniques where the theory is clean and the reality is a series of small failures that compound until your chromatogram looks like static. I've spent enough years at this instrument to know that most problems don't come from the detector or the column itself. They come from things you stop noticing because they've been there for months. Degassed solvent sitting in an open beaker. A pump seal that's been wearing slowly since last spring. A frit that clogged one step at a time instead of all at once. The method starts the same way every time. You make your mobile phase, filter it, and degas it. That last part is where people cut corners. If you're using an in-line degasser, you still need to sonicate or sparge with helium if you're doing gradient elution and running low UV wavelengths. Bubbles in the line at 210 nm will ruin your baseline faster than anything else. I learned that the hard way during a five-day method transfer when my baseline noise spiked every time the gradient valve switched from 90% water to 10% organic. Turned out the degasser was cycling but the solvent reservoir wasn't covered. Helium sparging for thirty minutes before the run fixed it permanently.
High Pressure Liquid Chromatography method development from scratch
You pick a column first. C18 is the default for a reason. If your analyte is ionizable, you're going to want a column with end-capped silanols or consider a mixed-mode phase, but start simple. A 150 by 4.6 mm column with 5 micron particles and 100 angstrom pores will handle most small molecules. For peptides or larger molecules, bump to 300 angstroms. Don't overthink this on the first attempt. Set your flow rate at 1.0 ml per minute for that column size. Your injection volume should be no more than ten percent of your peak width at the base. If your peaks are running at forty seconds wide, a ten microliter injection is fine. If you're using UPLC-grade columns with sub-two-micron particles, you're looking at one or two microliters max before peak broadening becomes noticeable. For the initial mobile phase, try water with zero point one percent formic acid and acetonitrile. That covers a surprising number of compounds. If your analyte is basic, swap the formic acid for ammonium formate at five millimolar. If it's acidic, you can try phosphoric acid instead. The pH matters more than people admit. A difference of half a pH unit can shift retention by twenty to thirty percent on a C18 column with ionizable compounds. That's not theoretical. I had a method that looked stable for three weeks then completely collapsed when someone used a different batch of water with a different CO2 content. The pH drifted by 0.4 and our late eluters moved back by eight minutes.
Run a quick gradient from five percent to ninety-five percent organic over twenty minutes. Watch where your peaks come out. If everything elutes before three minutes, you need more organic or a shallower gradient. If nothing comes out before fifteen, you're running too much organic at the start or your column is too long. Adjust from there. This iterative process usually takes three to five runs before you land somewhere usable. Here's something nobody tells beginners about gradient elution: the gradient delay volume matters. Every instrument has a built-in delay between the mixing chamber and the column head. On older systems it can be four to six milliliters. On newer ones closer to one milliliter. If you transfer a gradient method from one instrument to another without accounting for this delay, your retention times will shift. I once spent two days troubleshooting what I thought was a column problem before measuring the actual delay volume and realizing the second instrument was just slower to deliver the gradient. The column was fine the whole time. Detector setup is straightforward but the wavelength choice is where people make mistakes. If you're running a UV detector, 254 nm is standard for aromatic compounds. For non-aromatics or carboxylic acids, drop down to 210 nm. But 210 nm means your mobile phase needs to be UV transparent. Acetate buffers absorb there. Use formate or phosphate instead. Also, don't run above 220 nm with organic modifiers that have cutoffs near that region. Methanol is fine at 210 nm. Acetonitrile is fine down to about 190 nm. THF becomes problematic below 220 nm. I once ran a method with THF at 200 nm and spent an hour trying to figure out why my baseline was rising exponentially through the run. The solvent itself was absorbing.
What actually goes wrong and how to fix it
Pressure spikes are the most common headache. If your backpressure jumps suddenly and stays elevated, you have a blockage. Work backwards from the column. Disconnect the column and run the method. If the pressure normalizes, the column is the problem. If it doesn't, check the frits, the guard column, and the tubing fittings before the column inlet. A clogged guard column frit can add fifty to one hundred bars of backpressure and you won't know until your main column starts showing tailing. I had a persistent issue where my pressures would drift up by ten percent over the course of a day-long batch run. No leaks, no visible contamination. It turned out to be temperature. The column oven was set to thirty degrees but the room around the instrument fluctuated between twenty and twenty-six degrees during the day. Mobile phase viscosity changes about two percent per degree Celsius. That ten-degree swing was enough to push my pressures up noticeably and shift retention times by almost a minute. Installing a proper column oven with ±0.1 degree stability eliminated both problems immediately. Tailing peaks are another thing people blame on column degradation when it's often something else. Silanol interactions with basic compounds cause tailing on standard C18 columns. Adding an ion-pairing agent or adjusting pH to suppress ionization usually fixes it. But if your sample solvent is stronger than your mobile phase, you'll get fronting peaks and distorted shapes. Keep the injection solvent weaker than the initial gradient composition. If you need to dissolve your sample in pure acetonitrile and your starting condition is five percent organic, you're going to have a bad time. Dilute the sample or switch to a solvent closer to your mobile phase composition.
Peak splitting happens more often than it should and it's almost always mechanical. Check your injector loop for bubbles. Make sure the needle is seated properly. Verify that your septum isn't shredded and letting air in. A loose connection between the injector and the column will also split peaks. I've seen it multiple times with PEEK fittings that were tightened once and never checked again. Vibration from the pump can work them loose over a week or two. There are real limitations to this technique. High Pressure Liquid Chromatography cannot separate enantiomers on a standard C18 column. You need a chiral stationary phase for that. It also struggles with very large molecules above twenty thousand daltons. Size exclusion chromatography is the right tool there. If your sample contains particulate matter larger than the frit pore size, you'll clog the column regardless of how many guard columns you put in front of it. Filtration at 0.22 microns is non-negotiable. I lost a four-hundred-dollar column in a single run because someone filtered through 0.45 micron paper and the particles passed right through. That column showed up as two merged peaks instead of the three I was expecting. The efficiency dropped by nearly half and the backpressure jumped two hundred bars. It was dead after that. Column lifetime is another thing people optimize for badly. A good C18 column under normal conditions will last three to six months with daily use. That assumes you're filtering your samples, running a guard column, and avoiding pH values outside the two to eight range. If you're running pH one methods regularly, even high-quality columns will degrade within weeks. The siloxane bonds hydrolyze and your selectivity shifts. I once compared a column I'd been running at pH 1.8 for two weeks against a fresh one side by side. The retention of my most basic analyte had shifted by four minutes and the peak shape was noticeably broader. The manufacturer's stated pH limit was 2.0. I'd been over it the entire time and hadn't noticed because I wasn't comparing against a reference.
For quantitation, external calibration is the standard approach. Prepare standards spanning your expected concentration range. A five-point calibration curve with replicates at each point is adequate for most work. The correlation coefficient should be above 0.998. If it's not, check your pipetting technique and whether your standards are degrading. I've seen formic acid decompose in aqueous solutions over the course of a week, causing calibration curves to drift. Keeping standards in acetonitrile instead of aqueous mixtures improved their stability significantly. Method validation sounds like a heavy lift but for routine lab work you really need to confirm precision, linearity, and detection limits. Precision should be under two percent relative standard deviation across six replicate injections of the same standard. Linearity is your calibration curve. Detection limits depend entirely on your detector and your compound's molar absorptivity at your chosen wavelength. A compound with a strong chromophore at 254 nm might give you nanogram-level detection. Something without one could be in the microgram range. Run a serial dilution and determine the signal-to-noise ratio. Three to one is your limit of detection. Ten to one is your limit of quantitation. Documentation matters more than most people think. Every method change, every column lot number, every mobile phase batch should be recorded. When a method fails unexpectedly six months later, having a traceable history is what separates someone who can identify the root cause from someone who just restarts from scratch. I once traced a persistent retention time shift to a different supplier for the same grade of HPLC solvent. The new supplier used a different water purification system and the trace ionic content was slightly higher. It didn't show up on the certificate of analysis but it changed the selectivity enough to matter for our method. Switching back resolved it. That kind of thing doesn't show up in any textbook.
