Every spec line is either a measurement or a promise, and the difference is

A buyer at a mid-size environmental lab sent me a column datasheet last month. The sheet listed a plate count of 118,000 for a 150 mm C18 column. What it did not list was the test solute, the mobile phase, or the temperature. The column never came close to that number on their LC-MS method, and the vendor refused a return because the sheet was technically accurate under its own undisclosed conditions. Reading a lab chromatography datasheet is not about trusting the printed number. It is about knowing which numbers were earned in a defined test and which were typed into a template.

Every spec line is either a measurement or a promise, and the difference is the whole game

I have spent 15 years buying, selling, and occasionally rejecting chromatography consumables. The first thing I teach new procurement staff is that a datasheet is a legal document with marketing formatting. Each line item falls into one of three categories: a tested value (measured on every unit or every lot, with a defined method), a typical value (measured on a reference batch, not your batch), or a nominal value (a design target, often never measured at all). The problem is that most sheets do not tell you which category a number belongs to. You have to infer it from the presence or absence of a test method reference, a tolerance band, and a certificate of analysis.

For example, a HPLC column datasheet from a reputable manufacturer will list particle size as 5 µm with a tolerance of ±0.5 µm and a method reference like ISO 13320 for laser diffraction. That is a tested value. The same sheet may list carbon load 12% with no tolerance and no method. That is often a typical value measured once during development. The difference matters when you are troubleshooting a retention time shift or defending a method to an auditor.

Particle size: the number that is always tested, but not always the way you think

Particle size is the single most important number on a column datasheet because it drives backpressure, efficiency, and the van Deemter curve. But the number printed as 5 µm or 3 µm is almost always a mean or median value from a distribution. The full specification includes D10, D50, and D90 values, which describe the particle size distribution. A column with a D50 of 5 µm and a D90 of 8 µm will behave very differently from one with a D50 of 5 µm and a D90 of 5.8 µm. The first will have a wider particle size distribution, lower efficiency, and higher backpressure than the second.

Ask the vendor for the span value, which is calculated as (D90 minus D10) divided by D50. A span below 0.8 is tight. A span above 1.5 is broad and will show up as tailing peaks on late-eluting compounds. I have seen a lot of cheap columns where the mean particle size is exactly 5 µm but the span is 1.7, and the plate count is half of what a tight-distribution column delivers. The datasheet will not volunteer the span. You have to ask.

For GC columns, the equivalent issue is the film thickness. A 0.25 µm film is not a single atomic layer. It is an average coating thickness along the capillary wall. The tolerance is typically ±0.05 µm for a 0.25 µm film, but that tolerance is measured on a reference column, not your column. In practice, film thickness varies along the length of the column and from lot to lot. A retention time shift of 0.1 minutes on a 30 m column can come from a film thickness difference of less than 0.02 µm. When you see a retention time spec on a GC column datasheet, look for the test solute and the temperature program. If those are missing, the number is a typical value, not a guarantee.

Plate count and efficiency: the most abused number on any datasheet

Plate count is the number that sells columns and the number that causes the most disputes. The formula is simple: N = 16 × (tR / W)², where tR is retention time and W is peak width at baseline. But the result depends entirely on the test conditions. A plate count measured with naphthalene on a C18 column at 25 °C with 70:30 acetonitrile:water will be much higher than the same column tested with a polar compound at 40 °C with a buffered mobile phase.

Here is the rule I use: a plate count spec without a named test solute, a mobile phase composition, a flow rate, and a temperature is a marketing number. It is not false, but it is not useful for your method. A proper plate count spec reads something like: N ≥ 80,000 per 150 mm column, tested with acenaphthene, 70:30 ACN:H2O, 1.0 mL/min, 25 °C. That is a tested value with a defined method. If the sheet says high efficiency or excellent resolution without numbers, walk away.

I once received a column where the datasheet claimed N = 95,000 for a 100 mm column. When we tested it with the vendor’s own method, we got N = 61,000. The vendor’s response was that our HPLC system had too much extra-column volume. They were right. Our autosampler and detector tubing added about 40 µL of extra-column dispersion, which is enough to cut the observed plate count by a third on a short column. The lesson is that plate count is a system property, not just a column property. When you compare datasheets, compare the test system too.

USP tailing factor and asymmetry: two numbers that look the same but are not

Peak shape is specified as either USP tailing factor (T) or asymmetry factor (As). They are calculated differently. USP tailing factor uses the peak width at 5% of peak height, while asymmetry factor uses the width at 10% of peak height. For a perfectly symmetrical peak, both equal 1.0. For a tailing peak, the asymmetry factor is always larger than the USP tailing factor. A column with T = 1.2 will have As ≈ 1.4. If a datasheet says tailing factor < 1.5 without specifying which calculation, assume it is the more forgiving one.

The acceptance limit depends on the method. USP monograph methods typically require T ≤ 2.0 for the analyte peak. EPA methods like SW-846 8270 often require T ≤ 2.0 for the internal standard peaks. A new column should deliver T < 1.3 for a well-behaved test solute. If the datasheet claims T < 1.1, check the test solute. Some vendors test with uracil, which is unretained and always gives a perfect peak. That tells you nothing about how the column will handle a basic pharmaceutical compound.

I keep a small library of columns that failed tailing specs on real samples but passed the vendor’s test. The common thread is that the vendor tested with a neutral, non-polar solute under ideal conditions. The customer tested with a basic amine under acidic conditions. The difference is not the column quality. It is the test method. When you read a tailing spec, ask: what solute, what pH, what temperature? If the vendor will not tell you, the number is decoration.

Pressure rating and temperature limits: the numbers that keep you out of court

Pressure ratings on HPLC columns are straightforward but often misunderstood. A column rated for 600 bar means the hardware can withstand that pressure without leaking or failing. It does not mean the column will perform well at 600 bar. The packing bed can collapse or form channels at high pressure, especially with sub-2 µm particles. The pressure rating is a safety limit, not an operating recommendation.

Temperature limits are similar. A C18 column rated for 80 °C means the bonded phase will not degrade catastrophically below that temperature. But silica-based phases begin to hydrolyze at pH above 7 and temperatures above 40 °C. The column will not fail immediately, but the retention time will drift as the phase is stripped. I have seen a column lose 30% of its retention in two weeks at pH 8 and 60 °C, even though the datasheet said it was stable to 80 °C. The datasheet was technically true for neutral pH. The pH-temperature interaction is the part that is not printed.

For GC columns, the temperature limit is a function of the stationary phase. A 5% phenyl-methylpolysiloxane phase is typically rated to 325 °C isothermal and 350 °C programmed. The isothermal limit is the one that matters for long runs. Exceeding it causes phase bleed, which shows up as baseline rise and ghost peaks. The programmed limit is for short excursions, not for holding at that temperature. I have replaced columns that were run at 340 °C for 40 minutes because the user read the programmed limit as the safe operating temperature. It is not.

Pore size and surface area: the hidden numbers that control retention and exclusion

Pore size is printed on most HPLC column datasheets, but it is rarely understood. The number, typically 100 Å or 120 Å for small molecules and 300 Å for proteins, is the average pore diameter. The full pore size distribution matters more than the average. A column with a mean pore size of 100 Å and a wide distribution will have some pores too small for your analyte to enter, reducing accessible surface area. A narrow distribution gives more consistent retention.

Surface area is directly related to pore size. A 100 Å silica has a surface area of about 300 to 350 m²/g. A 300 Å silica has about 100 to 150 m²/g. More surface area means more retention for small molecules, but it also means more silanol activity, which can cause tailing for basic compounds. The datasheet will list surface area as a typical value, not a tested value, because it is measured by nitrogen adsorption (BET method) on the raw silica, not on the finished bonded phase. The bonded phase reduces the accessible surface area by 20 to 40%, depending on the ligand density.

When I see a column datasheet that lists surface area but not pore size distribution, I assume the manufacturer is hiding something. A column with a mean pore size of 100 Å and a tail of pores below 40 Å will exclude larger analytes and show size-exclusion effects that are not in the datasheet. Ask for the pore size distribution curve if you are working with anything larger than a small molecule.

Carbon load and ligand density: the two numbers that are always nominal and always different

Carbon load is printed on nearly every reversed-phase column datasheet, and it is almost never a tested value on the finished column. It is calculated from the elemental analysis of the bonded silica, usually on a reference batch. The number, typically 10 to 20% for C18 phases, tells you how much organic material is attached to the silica. Higher carbon load generally means more retention for non-polar analytes, but the relationship is not linear and depends on the ligand type, the bonding chemistry, and the endcapping.

Ligand density, expressed in µmol/m², is a better predictor of retention and selectivity than carbon load. A C18 phase with a ligand density of 3.0 µmol/m² will have different selectivity from one with 2.5 µmol/m², even if the carbon load is the same. The difference comes from the residual silanol activity and the spacing of the bonded chains. Most datasheets do not list ligand density. You have to ask for it, and many vendors will not provide it because it is proprietary.

Here is a practical test. Take two C18 columns with the same particle size, pore size, and carbon load. Run a mixture of toluene, naphthalene, and biphenyl under identical conditions. If the retention times differ by more than 5%, the ligand density or the bonding chemistry is different. The datasheet did not tell you this. The chromatogram did.

How to spot a sheet that hides its test method

I have developed a simple checklist for evaluating a chromatography datasheet. It takes five minutes and catches most of the problems before you spend money.

• Does the sheet list a test method for each quantitative spec? Look for references like USP <621>, ISO 13320, or a vendor-specific method number. If the method is missing, the number is a typical or nominal value.
• Does the sheet list a tolerance band? A tested value always has a tolerance, like 5 µm ± 0.5 µm or N ≥ 80,000. A number without a tolerance is a target, not a guarantee.
• Does the sheet name the test solute and conditions? For plate count, tailing factor, and retention factor, the test solute, mobile phase, flow rate, and temperature must be specified. If they are not, the number is not reproducible in your lab.
• Does the sheet distinguish between tested, typical, and nominal values? Reputable manufacturers mark typical values with a footnote or an asterisk. If every number looks equally authoritative, the sheet is designed to mislead.
• Is there a certificate of analysis for the specific lot? A datasheet describes a product line. A certificate of analysis describes the lot you are buying. Always ask for the COA before you commit.

I once rejected a batch of PTFE syringe filters because the datasheet claimed a pore size of 0.45 µm with no tolerance and no test method. The COA, when I finally received it, showed a bubble point test result that corresponded to an actual pore size of 0.58 µm. The datasheet was technically true for the reference lot, but the lot I received was out of spec. The vendor’s argument was that the datasheet was a product description, not a lot-specific guarantee. My argument was that a product description without a tolerance is not a specification. I won that dispute because the purchase order referenced the datasheet as the specification. If you do not reference the datasheet in your PO, you lose that argument.

Comparison table: what a tested spec looks like versus a nominal one

Field case one: the disappearing plate count

A pharmaceutical QC lab bought a 150 mm, 3 µm C18 column for a USP monograph method. The datasheet claimed N ≥ 100,000. The lab’s system delivered N = 72,000 on the first injection. The lab manager called me, convinced the column was defective. I asked three questions: what is the extra-column volume, what is the injection volume, and what is the data acquisition rate.

The answers were 35 µL extra-column volume, 20 µL injection volume, and 20 Hz acquisition rate. The extra-column volume alone was enough to cut the observed plate count by 25% on a 3 µm column. The injection volume was too high for a 3 mm ID column, causing additional band broadening. The acquisition rate was too slow to capture the narrow peaks, adding another 10% loss. The fix was to reduce the injection volume to 5 µL, replace the connecting tubing with 0.005 inch ID tubing, and increase the acquisition rate to 80 Hz. The plate count came up to 96,000. The column was fine. The system was the problem.

Field case two: the septum that lied about its puncture count

An environmental lab running EPA 8260 volatile organic analysis kept seeing air leaks after about 50 injections on a 20 mL headspace vial. The septum datasheet claimed 100 punctures before failure. The lab was using a 23-gauge needle with a beveled tip, which is standard for autosamplers. The failure mode was a torn septum that allowed air into the vial, causing analyte loss.

The measurement that identified the problem was a leak test using a pressure decay method. The septum failed the leak test after 48 punctures, not 100. The vendor’s puncture count was measured with a blunt-tipped needle at 20 punctures per minute on a test rig, not with a beveled needle at autosampler speed. The fix was to switch to a pre-slit septum with a PTFE-faced silicone construction, rated for 150 punctures with a beveled needle. The difference was in the test method, not the material. The original septum was fine for its stated test conditions. It just was not tested under the conditions the lab actually used.

This is the same logic that applies to every line on a chromatography datasheet. The number is only as good as the test method behind it. If the method is not stated, the number is a claim, not a measurement. For more on the mechanical side of septum puncture, our HPLC filter team has documented the failure modes that occur when a septum is reused past its rated count.

What the COA tells you that the datasheet cannot

The certificate of analysis is the only document that describes the specific lot you are buying. A datasheet describes a product line, which may include hundreds of lots with varying properties. The COA should list the actual measured values for your lot: particle size distribution, plate count, tailing factor, carbon load, and pressure test results. If the COA lists the same values as the datasheet, the vendor is not testing each lot. They are copying the datasheet into the COA template.

I have seen COAs where the plate count, tailing factor, and retention factor were identical across three different lots over six months. That is statistically impossible if each lot is actually tested. The vendor was testing one lot per year and applying the results to all lots. When I confronted them, they admitted that the COA was generated from the datasheet, not from lot-specific testing. The fix was to require a lot-specific COA with a unique lot number and a test date. If the test date is more than six months old, ask for a new test.

For GC columns, the COA should include a retention index for a standard mixture, a bleed specification at a defined temperature, and a coating efficiency value. Coating efficiency is expressed as a percentage, with 100% being a perfectly uniform film. A value below 80% indicates a non-uniform film that will cause peak broadening. Most datasheets do not list coating efficiency. The COA does, if the vendor tests it. Ask for it.

For sample preparation products like PTFE membrane filters, the COA should include a bubble point test result and a water breakthrough pressure for hydrophobic membranes. These are the tested values that confirm the pore size rating. A datasheet that lists a pore size without a bubble point reference is using a nominal value.

How to write the purchase order so the datasheet actually binds

The single most effective thing you can do as a buyer is to reference the datasheet and the COA in your purchase order. A PO that says HPLC column, 150 mm × 4.6 mm, 5 µm C18 buys you a product description. A PO that says HPLC column, 150 mm × 4.6 mm, 5 µm C18, N ≥ 85,000 per vendor datasheet rev 3, lot-specific COA required, USP T ≤ 1.5 for naphthalene buys you a specification.

I learned this the hard way with a COD vial order for a wastewater lab. The datasheet claimed the vials were pre-cleaned to EPA standards. The lot we received had a blank COD of 12 mg/L, which is above the 5 mg/L limit for low-range COD analysis. The vendor refused the return because the datasheet said pre-cleaned, not blank COD ≤ 5 mg/L. The word pre-cleaned is not a specification. It is a claim. If I had written the PO to require a blank COD ≤ 5 mg/L per lot, the vendor would have been obligated to test and meet that number. For more on how COD vial specs are actually tested, see codvial.com.

The same principle applies to every chromatography consumable. A syringe filter datasheet that says low extractables is not a specification. A PO that requires extractables ≤ 0.1 mg per filter, tested per USP <661> is a specification. A guard column datasheet that says high efficiency is not a specification. A PO that requires N ≥ 10,000 for the guard plus analytical column combination is a specification.

The decision rule: buy the test method, not the number

Here is the rule I apply to every chromatography datasheet I read, and it has saved my labs more money than any other single practice: a number without a test method is a rumor, a number with a method but no tolerance is a target, and a number with a method and a tolerance is a specification. Buy only the third kind. When a vendor will not tell you how a number was measured, assume the number was not measured at all. When a vendor will not give you a lot-specific COA, assume the lot was not tested. When a vendor says a spec is typical, multiply the claimed performance by 0.7 and ask yourself if the product still works for your method. If it does, buy it. If it does not, find a vendor who will put the test method in writing.

For more on how chromatography consumables are specified and tested across our product lines, see Lab Chromatography. For vendor-specific technical documentation, the Agilent and Thermo Fisher resources are useful, but apply the same scrutiny to their sheets as to any other. Restek publishes some of the most complete test method documentation in the industry, and their chromatograms are a good reference for what a properly specified column should look like.

References & further reading

Related resources from our lab consumables network:

• Lab Chromatography at labchromatography.com
• Hplc Filter at hplcfilters.com
• Ptfe Membrane Filter at ptfemembranefilter.com
• Cod Vial at codvial.com

External references and standards:

• Agilent chromatography consumables and columns
• Thermo Fisher chromatography and sample handling
• U.S. EPA SW-846 test methods for waste and materials
• Restek technical resources for chromatography