You stare at the mass spec readout. Still, the peak sits at 412. 3 m/z. Also, your notebook says the theoretical mass is 410. 2. Close — but not exact. Now you're wondering: is that difference real, or did something go sideways in the prep?
That gap between theory and reality? That's where experimental molecular weight lives. And figuring it out properly separates solid data from guesswork.
What Is Experimental Molecular Weight
Experimental molecular weight is the mass you measure*, not the one you calculate from a formula. It comes from an instrument — mass spectrometer, light scattering detector, osmometer, whatever — and it reflects what's actually in your vial. Impurities, adducts, degradation products, solvent clusters. All of it.
Theoretical molecular weight assumes a pure, perfect molecule. Experimental molecular weight deals with the messiness of reality.
The distinction matters more than people think
A peptide synthesized on a resin might have a theoretical mass of 2,847 Da. But the crude cleavage mixture? You'll see peaks for the target, the deletion sequences, the truncated versions, maybe a sodium adduct at +22, a potassium at +38. The "molecular weight" of that sample isn't a single number. It's a distribution.
And if you're characterizing a polymer? Which means forget a single value entirely. You're looking at Mₙ (number average), M_w (weight average), M_z (z-average) — each telling you something different about the chain length distribution.
Why It Matters / Why People Care
You'd be surprised how many papers report "molecular weight confirmed by mass spec" with zero details on how they got the number. No error estimate. Now, no calibration info. No charge state assignment. Just a value.
That's a problem.
Regulatory and quality consequences
In pharma, an incorrect molecular weight assignment can mean a failed IND filing. So if your experimental MW doesn't match theoretical within the expected tolerance, you have an impurity. The FDA wants to see that you know* what's in your drug substance — not just the main peak, but the related substances too. Practically speaking, or a degradation pathway. Or a structural misassignment.
In polymers, the difference between Mₙ and M_w dictates mechanical properties. A batch with M_w = 100k and Mₙ = 50k (Đ = 2.0) behaves differently than one with M_w = 100k and Mₙ = 90k (Đ = 1.1). Same "average" — totally different material.
Publication standards are tightening
Journals increasingly require raw data deposition. If your experimental MW claim doesn't hold up to scrutiny — wrong charge state, uncalibrated axis, ignored isotope pattern — the paper gets rejected. Reviewers check. Or worse, published and later retracted.
How It Works (or How to Do It)
The method depends entirely on what you're analyzing and what instrument you have access to. Here's the practical breakdown.
Mass spectrometry — the workhorse
Most labs reach for MS first. But "getting a mass spec" isn't a single procedure.
ESI vs. MALDI — pick the right ionization
Electrospray (ESI) gives you multiply charged ions for proteins, peptides, oligonucleotides. Think about it: a 15 kDa protein might show +12, +13, +14 charge states simultaneously. Software does the math — but you need to verify the charge state assignment manually at least once. You deconvolute* those to get the neutral mass. Automated algorithms hallucinate.
MALDI gives mostly singly charged ions. And the "sweet spot" on the target plate? Practically speaking, great for polymers, synthetic peptides under ~5 kDa, intact proteins if you optimize matrix and laser. It moves. But MALDI has mass bias — higher masses suppress lower ones. Run standards every session.
Calibration — don't skip it
Internal calibration (spiking a known standard into your sample) beats external every time. External calibration drifts with temperature, vacuum, phase of the moon. Internal corrects in real time.
For high-res instruments (Orbitrap, FT-ICR, Q-TOF), you want < 5 ppm error. 1–0.Know your instrument's spec. 5 Da is typical. And for low-res quadrupoles or ion traps, 0. Don't claim 2 ppm accuracy on a machine that does 20.
Adducts and isotopes — read the pattern
See a peak at M+22? But that's sodium. M+38? Potassium. M+18? Water adduct (common in ESI). Even so, the isotope pattern — spacing, relative intensities — tells you the elemental composition. Chlorine gives a distinctive M+2 at ~33% intensity. That said, bromine gives nearly 1:1 M and M+2. Sulfur, silicon, phosphorus all leave fingerprints.
If the isotope pattern doesn't match your formula, your formula is wrong. Or you're looking at a different compound.
Light scattering — for polymers and proteins in solution
Size-exclusion chromatography (SEC) with multi-angle light scattering (MALS) gives you absolute* molecular weight. So the detector measures scattered light intensity at multiple angles. Consider this: no column calibration needed. Zimm plot or Debye plot extrapolation gives M_w directly.
But — and this trips people up — you need dn/dc* (refractive index increment). Day to day, for proteins, 0. Think about it: you must* measure it. For polymers? Even so, 185 mL/g is the standard assumption. Guessing dn/dc introduces systematic error that no amount of signal averaging fixes.
And the concentration detector (RI or UV) must be calibrated. If your concentration is off by 10%, your MW is off by 10%. Linear.
Osmometry — old school, still gold for Mₙ
Membrane osmometry measures vapor pressure lowering. Below 1k Da, the signal gets noisy. Works for polymers up to ~100k Da. Gives number-average molecular weight (Mₙ) directly. Above 100k, the membrane doesn't equilibrate fast enough.
For more on this topic, read our article on how to read peptide elution time and intensity heatmap or check out an ion with a negative charge. formed by gaining electrons.
Vapor pressure osmometry (VPO) goes lower — down to ~200 Da. But it's solvent-sensitive. And both methods need careful temperature control. A 0.01°C drift ruins the run.
Analytical ultracentrifugation — the absolute standard
Sedimentation equilibrium AUC gives M_w without a matrix, without a column, without calibration. Here's the thing — you spin the sample at multiple speeds, measure concentration gradients, fit to the Lamm equation. It's slow (hours per run), sample-hungry (hundreds of µL at mg/mL), and the data analysis has a learning curve.
But when you need* to know — really know — AUC is the referee.
Common Mistakes / What Most People Get Wrong
Reporting "the molecular weight" as a single number for a polymer
There is no such thing. Report Mₙ, M_w, Đ (dispersity = M_w/Mₙ). At minimum
Reporting "the molecular weight" as a single number for a polymer is like reporting "the height" of a crowd — meaningless without context. Polymers are distributions. Always report:
- Mₙ (number-average molecular weight) — sensitive to chain count, dominated by low-mass species
- M_w (weight-average molecular weight) — dominated by high-mass chains
- Đ (dispersity = M_w/Mₙ) — measures breadth of the distribution
A sample with Đ = 1.05 is nearly monodisperse. Which means đ = 3. Also, 0 means significant heterogeneity. Ignoring Đ is like ignoring error bars.
Trusting MS peak intensity for quantitation without calibration
Mass spectrometers are not inherently quantitative. Which means ionization efficiency varies wildly between compounds, matrices, and instruments. A 10x difference in response factor between two analytes is routine.
Use internal standards. Validate linearity, precision, and recovery. Calibrate with at least 5 points. If you didn't calibrate, you're guessing.
Ignoring matrix effects in LC-MS
Co-eluting compounds suppress or enhance ionization. Here's the thing — your analyte peak might be 50% smaller (or larger) than you think. This isn't theoretical — it's why bioanalytical methods fail validation.
Post-column infusion, post-extraction spiking, matrix factor calculations — pick one and use it.
Using the wrong column chemistry
C18 for everything? Basic compounds on C18? Expect poor peak shape and low recovery. That's lazy. Switch to HILIC, SCX, or specialty phases.
Match your stationary phase to your analyte's chemistry. Consider this: acidic compounds → C18 or phenyl. Basic → SCX or HILIC. Neutral polar → HILIC or amide.
Forgetting detector compatibility
UV detection at 214 nm works great for peptides. But what about your mobile phase? High aqueous content + UV detection = baseline drift nightmares.
Check your detector's spectral cutoff. Verify compatibility with your mobile phase pH. Some detectors (ELSD, CAD) don't care about wavelength — use them when UV fails.
Misinterpreting retention time shifts
Retention time drift is real. So column aging, temperature fluctuations, mobile phase composition changes — all shift RTs. Practically speaking, a 0. 1-minute shift can cause misidentification.
Use relative retention times (vs. But internal standard). Or better, use accurate mass + MS/MS. RT alone is not identification.
Calibration and Standards — The Foundation You Can't Skip
External standards only work if your matrix matches
Calibrating in pure solvent and running samples in plasma, urine, or food extract? On the flip side, your calibration curve lies. Matrix effects distort every point.
Use matrix-matched standards. Or standard addition. Or both.
Stock solutions aren't stable forever
DMSO stocks degrade. Water stocks grow bacteria. Freeze-thaw cycles break down analytes. Label everything with prep date and expiration.
Store stocks at -80°C in small aliquots. Never leave them at room temperature "for just a few minutes."
Calibration curves need weighting
Linear regression assumes constant variance. In LC-MS, variance increases with concentration. Use 1/x or 1/x² weighting. Unweighted fits bias low concentrations.
Check your r² — but don't worship it. That's why look at residuals. Systematic deviation means your model is wrong.
Conclusion
Analytical chemistry isn't about chasing the latest instrument or the fanciest software. It's about understanding what your data actually represents and knowing where uncertainty creeps in.
Every measurement has assumptions baked in — about ionization efficiency, column behavior, detector response, sample stability. The difference between a good analyst and a great one is recognizing those assumptions, testing them, and reporting results with appropriate caveats.
Your data is only as good as your weakest step. Characterize your column. Which means validate your method. Because of that, calibrate your detectors. Question your results.
Because when someone asks "how sure are you?" — you need to be able to answer.
