Report 175 · Lab Science
It is one of the most common lines on a certificate of analysis, and it sounds like it means 99% of the powder in the vial is the compound on the label. Usually it means something narrower: of everything the detector could see, the main peak was 99% of the signal.
HPLC (high-performance liquid chromatography) is the workhorse of chemical quality control. You dissolve a sample, push it through a packed column, and the components come out at different times. A detector at the end, most often a UV/visible absorbance detector, draws a trace with a peak for each thing it can see. It is an excellent separation tool, and I have no quarrel with it.
The trouble is the fastest way to turn that trace into a purity number. You add up the areas of all the peaks, then report the main peak as a percentage of the total. Chemists call this area normalization, or the "100% method." It needs no reference standard and no calibration, which is exactly why it is everywhere.
An area-percent purity quietly assumes two things, and the method itself can check neither.
That everything in the sample shows up as a peak. A UV detector sees molecules that absorb UV light. Water doesn't make a peak. Neither do most salts, or silica left over from a purification column. A 2014 review in the Journal of Medicinal Chemistry by Guido Pauli and colleagues, led from the University of Illinois at Chicago, put it plainly: "Materials such as silica gel, sorbents, and inorganic and polymeric matter frequently escape chromatographic analysis." If 4% of a vial is water and salt, a chromatogram can still be one clean peak.
That every compound gives the same signal per gram. It doesn't. A UV detector's response depends on how strongly each molecule absorbs at the chosen wavelength, so an impurity can be heavily underweighted or overweighted relative to the main compound. This is the same physics I wrote about in when Beer's law breaks down: absorbance is a property of the molecule, not of its mass. A Fujifilm Wako poster on purity testing lists both problems side by side: there can be a "non-detectable impurity depending on the kind of a detector," and the "main component differs from impurity in the sensitivity of detection."
That same poster, from Wako Pure Chemical, Japan's National Institute of Health Sciences and JEOL Resonance, includes a worked example on the plant compound (−)-epigallocatechin, one of the green tea catechins. The same material was assayed two ways:
By HPLC: 100.0%.
By quantitative NMR: 95.8%.
The NMR run also identified 0.6% residual ethanol, a solvent the HPLC number simply didn't register. The poster does not itemize the rest of the gap, and I won't guess at it. The point stands without it: the same material, measured honestly two ways, differed by more than four percentage points, and the chromatography said it was perfect.
Quantitative proton NMR (qNMR) works on a different principle. Signal area is proportional to the number of hydrogen nuclei producing it, more or less regardless of what molecule they sit in. Add a precisely weighed internal standard of known purity, and you can calculate how many milligrams of your compound are in the tube, not just what share of the visible signal it makes up. Pauli's review calls this "nearly universal detection" and describes qNMR as able to capture "analytes that frequently escape detection (water, sorbents)."
It is not magic. The same review notes that even qNMR, if you use it in the shortcut 100% mode without calibration, "frequently results in an overestimation of purity" because it misses "nonobservables" like salts and silica. The protection comes from calibrating against a weighed standard, not from the brand of instrument. That is a general rule in analytical chemistry, and it is the one I keep coming back to: a number is only as good as what it was calibrated against.
The most quotable line in the review is aimed at the very format on most certificates. Purity thresholds like "at least 99% pure," the authors write, "commonly found in journal, pharmacopoeial, and regulatory guidelines as well as certificates of analysis must be considered as arbitrary values," and they add that "without full disclosure of the HPLC method including detection parameters and quantitative calibration, the results likely will not be reproducible in other laboratories."
That's strong language from analytical chemists. It does not mean HPLC purity numbers are fake. It means the number answers a narrower question than the reader thinks it does, unless the method says otherwise.
1. Look for the word "area." "99.2% (HPLC, area %)" is a relative number. "Assay 98.5% w/w against a reference standard" is an absolute one. They are different measurements.
2. Check the detector and wavelength. A purity at one UV wavelength says nothing about compounds that don't absorb there.
3. Look for water, residual solvent and ash. Separate lines for these (Karl Fischer water, residual solvents, residue on ignition) are how a lab can account for what HPLC can't see. If they're missing, the purity figure is carrying more weight than it can bear.
4. Ask whether a second, independent method agrees. Pauli's review argues for exactly this ("orthogonal" methods). Two methods with different blind spots agreeing is worth far more than one method reported to a decimal place.
I read the review in full and the poster as published. The Wako poster is undated on its face and shows one compound's result; I have not seen the underlying HPLC conditions for that comparison. The 100.0% vs 95.8% example is one material, not a survey of how far area-percent figures typically drift.
I make no claim about any specific product or supplier. This is a method piece. Plenty of certificates do report absolute assays and full impurity accounting.
"99% by HPLC" usually means the main peak was 99% of the peak area a detector could see at one wavelength. Water, salts, silica and anything that doesn't absorb are outside the frame, and compounds that do absorb are weighted by how strongly, not by how much is there. A reference standard measured both ways read 100.0% by HPLC and 95.8% by qNMR. A purity number is only as meaningful as the method written next to it, so read the method.
Scope note: this is a methods explainer. No sample was analyzed. The four reading questions are the author's framework, built on the sources above.
U.S. Army combat veteran (Counter-IED / Electronic Warfare), peer-reviewed researcher in microwave spectroscopy, and founder & CEO of Shroombiosis. Consults on laboratory operations, AI, and supplement formulation.