Integration: Counting Protons

Integration remains the quantitative backbone of every routine structure-elucidation workflow in organic and medicinal chemistry — a synthetic chemist checking a reaction's product purity by ¹H NMR reads integral ratios to confirm both structure and to estimate the ratio of product to unreacted starting material or side product, effectively using integration as an internal quantitative assay without needing a separate calibration curve, provided all relevant protons are cleanly resolved. In pharmaceutical quality control, quantitative NMR (qNMR) uses a calibrated internal standard of known concentration alongside the analyte, and integration ratios against that standard give absolute concentration measurements traceable to a primary standard — a technique now used by pharmacopeias as an alternative to chromatographic assay for active pharmaceutical ingredient purity. In polymer chemistry, end-group analysis by integration (comparing the integral of a chain-end proton signal, present once per polymer chain, against the integral of a repeat-unit signal, present many times per chain) is a standard method for determining number-average molecular weight, $M_n$, directly from a single ¹H spectrum. In metabolomics and natural-products isolation, integration of characteristic signals in complex mixture spectra allows relative quantitation of multiple components without physical separation, and diffusion-ordered spectroscopy (DOSY) combines integration with diffusion coefficients to deconvolute overlapping spectra of mixtures. At the instrumentation level, modern spectrometers report digital integrals directly under each multiplet in processing software, and automated structure-verification software cross-checks integration ratios against a proposed molecular formula as a first-pass sanity check before a chemist even examines multiplet splitting patterns in detail.

Ask the tutor: If two multiplets partially overlap, how do you separate their individual integrals reliably before computing a ratio?

Picture a sample of ethyl acetate, CH₃–CO–O–CH₂–CH₃, sitting in the bore of a superconducting magnet with field $B_0$. Every proton in the molecule is a spin-½ nucleus, and in the external field each proton's spin can align either with $B_0$ (the lower-energy $\alpha$ state) or against it (the higher-energy $\beta$ state). The energy gap between these two states, $\Delta E = \gamma\hbar B_0$, is minuscule compared to $k_BT$ at room temperature — for a 400 MHz spectrometer this gap corresponds to a radiofrequency photon, roughly five orders of magnitude smaller than $k_BT$. The Boltzmann distribution therefore gives only a tiny population excess in the $\alpha$ state:

The population difference $N_\alpha - N_\beta$ is what generates the net bulk magnetization $M_0$ along $B_0$, and it is this net magnetization — tipped into the transverse plane by a radiofrequency pulse and then detected as it precesses and relaxes — that produces the NMR signal. Crucially, $\Delta E = \gamma\hbar B_0$ depends only on the nucleus (through its gyromagnetic ratio $\gamma$) and the applied field $B_0$, not on the tiny chemical-shift perturbation from a proton's electronic environment. All protons in the molecule, whether they are on the methyl group of the acetate or the methylene of the ethyl group, see essentially the identical $B_0$ and identical $\gamma$, so each one contributes an essentially identical increment $\Delta E$ to the population imbalance and hence an essentially identical increment of detectable signal. This is the whole secret. The chemical shift moves a proton's resonance left or right along the frequency axis (because the local electron density shields or deshields it, shifting the effective field it feels by parts per million), but it does not change how much signal that proton contributes when it does resonate. Ten chemically distinct methyl protons scattered across ten different molecules, each with a wildly different shift, would each still contribute the same unit of signal. So if a cluster of peaks integrates to an area proportional to $3$ and a neighboring cluster to an area proportional to $2$, the underlying physical fact is that three times as many protons are resonating in the first cluster as some common unit compares to two in the second — the ratio $3:2$ is a count, not a coincidence of amplitude. In practice the detected signal is the free induction decay (FID), a time-domain voltage induced in the receiver coil by the precessing transverse magnetization, and Fourier transformation converts this into the frequency-domain spectrum you actually look at. The height of a peak in that spectrum depends on the linewidth (itself set by $T_2$ relaxation and field homogeneity) as well as on how many spins contribute — a broad peak from many protons can be shorter than a sharp peak from few. But the area under a peak, being the integral of a Lorentzian lineshape, is independent of linewidth and scales purely with the number of contributing spins. This is precisely why the technique is called "integration" and why height is never trusted for quantitation, only area.

The central relationship of integration is a direct proportionality between area and spin count. For a set of $k$ chemically distinct proton environments in a molecule, each with $n_i$ equivalent protons, the areas $A_i$ measured under their resonances obey

where:

• $A_i$ is the integrated area (in arbitrary instrument units) under the $i$-th resonance,
• $n_i$ is the number of equivalent protons producing that resonance,
• $c$ is a proportionality constant, identical for every environment in the same spectrum, set by instrument gain, sample concentration, and receiver settings.

Because $c$ cancels between any two signals in a single spectrum, what you actually measure and use is a ratio:

This ratio equation is the entire content of quantitative integration: it lets you convert measured areas into a ratio of small integers, which you then match against candidate structures. If you additionally know the total proton count $N$ from a molecular formula (e.g., from a parent mass in mass spectrometry combined with a plausible formula), you can pin down absolute counts via

where $\sum_j A_j$ is the sum of all measured integral areas across the whole spectrum, and $N$ is the total number of protons in one molecule of the compound. This is the step that turns "the ratio is 3:2:3" into "there are exactly 3, 2, and 3 protons," rather than 6, 4, and 6 (which has the same ratio).

Goal. Show, from the physics of pulsed FT-NMR detection, that the area under a resonance in the frequency-domain spectrum is strictly proportional to the number of contributing spins, independent of linewidth, while peak height is not.

Step 1 — Equilibrium magnetization per spin. At thermal equilibrium, a collection of $n$ identical spin-½ nuclei of gyromagnetic ratio $\gamma$ in field $B_0$ produces a net longitudinal magnetization

This comes directly from the population difference $N_\alpha - N_\beta$ derived from the Boltzmann factor times the magnetic moment per spin; every factor in this expression ($\gamma, \hbar, B_0, T$) is common to all protons in the sample except the count $n$ of protons in that particular environment. So $M_0 \propto n$, exactly, for a fixed instrument and sample temperature.

Step 2 — RF pulse tips the magnetization into the detectable plane. A 90° radiofrequency pulse rotates $M_0$ from the $z$-axis (along $B_0$, undetectable by the receiver coil) into the transverse $xy$-plane, where it precesses at the Larmor frequency and induces an oscillating voltage in the receiver coil by Faraday induction. The initial transverse magnetization magnitude immediately after a perfect 90° pulse is simply $M_0$ itself (rotated, not diminished), so the initial amplitude of the induced signal for that environment's contribution is still directly proportional to $n$.

Step 3 — Free induction decay and Fourier transform. The transverse magnetization for environment $i$ decays with characteristic time $T_{2,i}^*$ (combining true spin–spin relaxation and field-inhomogeneity dephasing), giving a time-domain signal contribution

Fourier transforming an exponentially decaying cosine yields a Lorentzian lineshape in the frequency domain centered at $\omega_i$, with peak height inversely proportional to linewidth (i.e., inversely proportional to $1/T_{2,i}^*$) but with total area under the Lorentzian equal to $M_{0,i}$ times a constant that does not depend on $T_{2,i}^*$:

This is the standard result that the integral of a Lorentzian absorption line is independent of its width — broadening the line (poorer shimming, faster relaxation, more overlapping neighbors) redistributes area over a wider frequency range without destroying the total area.

Step 4 — Height depends on linewidth; area does not. Because peak height for a Lorentzian scales as (area)/(linewidth), two environments with the same proton count $n$ but different linewidths (say, a broad NH proton next to a sharp CH₃ singlet) would show different heights despite equal spin count. This is precisely why height must never be used quantitatively: only the area, which strips out the linewidth dependence entirely, faithfully tracks $n$.

Step 5 — Ratio of areas equals ratio of proton counts. Combining Steps 1–4, the measured area for environment $i$ is

where $k_{\text{instr}}$ absorbs receiver gain, coil geometry, number of scans, and other instrument-specific constants — but crucially not anything that depends on which proton environment $i$ refers to. Taking the ratio of any two areas in the same spectrum, every one of these common factors cancels:

This is the working equation of NMR integration, and it required no assumption about chemical shift, multiplicity, or molecular structure — only that all protons share the same $\gamma$, sit in the same $B_0$, and are recorded in the same experiment.

Goal. Take arbitrary raw integration numbers straight off an instrument printout and derive the small-integer ratio, then the absolute proton count, showing every arithmetic step explicitly (no rounding sleight of hand).

Step 1 — Record the raw areas. Suppose a spectrometer reports raw (arbitrary-unit) integrals for three resonances of an unknown compound:

These numbers are not integers and carry an arbitrary overall scale $k_{\text{instr}}$ that we do not know and do not need to know.

Step 2 — Divide every integral by the smallest one. Normalizing removes the unknown scale constant, since it cancels in every ratio:

This gives the ratio $3.00 : 2.00 : 1.00$, already close to small integers — good raw integration data often lands within a few percent of exact integers because of instrumental noise and baseline drift, and a chemist rounds sensibly rather than forcing exactness.

Step 3 — Check whether the integer ratio is chemically plausible on its own, or needs scaling up. A ratio of $3:2:1$ is a relative statement; the true proton counts could be $3,2,1$ or equally well $6,4,2$ or $9,6,3$, since all these have the same ratio. You cannot distinguish between these possibilities from the ratio alone — you need an external constraint.

Step 4 — Bring in the molecular formula as the external constraint. Suppose independent evidence (mass spectrometry giving $M^+ = 88$, elemental analysis) establishes a molecular formula of $\mathrm{C_4H_8O_2}$, which has $N = 8$ total protons. Sum the ratio units:

Since $8$ is not evenly divisible into $3:2:1$ scaled by an integer factor unless we scale by $8/6$, check first whether an integer multiple of $6$ equals $8$ — it does not. This signals that the naive $3:2:1$ ratio from Step 2 must be reconsidered: either the true underlying integers are $3:2:1$ scaled to match $N=8$ imperfectly (a red flag for measurement error or an overlooked exchangeable proton), or a fourth signal was missed.

Step 5 — Resolve the discrepancy using chemical reasoning. For $\mathrm{C_4H_8O_2}$ with degrees of unsaturation

one degree of unsaturation is consistent with an ester carbonyl. A plausible structure is ethyl acetate itself, $\mathrm{CH_3CO_2CH_2CH_3}$, with three environments: $\mathrm{OCOCH_3}$ (3H), $\mathrm{OCH_2}$ (2H), and $\mathrm{CH_2CH_3}$ (3H) — total $3+2+3 = 8$, matching $N=8$ exactly. Comparing against the measured ratio $3:2:1$ from Step 2, the third signal's true count should be $3$, not $1$; re-examining the raw data, the correct normalization should have used $A_3 = 4.71$ (the terminal CH₃ triplet, often integrating similarly to the acetate singlet) as a third independent value rather than assuming it was the smallest. This is exactly the kind of arithmetic trap integration problems are designed to test: never assume which signal is the reference unit — anchor the ratio to the molecular formula, not to whichever number happens to be smallest.

Step 6 — Corrected calculation. Redo the ratio with correctly assigned raw areas $A_1 = 4.71$ (CH₃CO, 3H), $A_2 = 3.14$ (OCH₂, 2H), $A_3 = 4.71$ (CH₃CH₂, 3H):

Sum of ratio units $= 3+2+3 = 8$, matching $N=8$ exactly, so absolute counts are read off directly:

consistent with ethyl acetate's actual proton distribution, $3:2:3$.

Example 1 (foundational) — Toluene, a clean two-environment case. Toluene, $\mathrm{C_6H_5CH_3}$, has a molecular formula $\mathrm{C_7H_8}$, giving $N=8$ total protons. Two ¹H environments exist by symmetry: the aromatic ring protons (formally 5 distinct positions, but closely overlapping into one broad multiplet near 7.1–7.3 ppm for a first approximation) and the methyl protons (singlet near 2.35 ppm). Suppose integration gives raw areas $A_{\text{Ar}} = 6.25$ and $A_{\text{CH}_3} = 3.75$. Normalize:

This matches the expected ratio of 5 aromatic protons to 3 methyl protons exactly. Using the molecular formula constraint $N=8$:

recovering the known structure directly from integration plus formula, with no ambiguity. As a limiting-case check: if instead the ring had one substituent removed entirely (benzene itself, $\mathrm{C_6H_6}$), all six protons become equivalent by symmetry, giving a single resonance whose integral, taken alone, carries no ratio information at all — a reminder that integration only discriminates between chemically distinct environments, never within a single equivalent set.

Example 2 (applied) — Distinguishing two isomeric esters by integration ratio. Consider two possible structures for a compound of formula $\mathrm{C_4H_8O_2}$ ($N=8$ protons, DoU $=1$): methyl propanoate, $\mathrm{CH_3CH_2CO_2CH_3}$, versus ethyl acetate, $\mathrm{CH_3CO_2CH_2CH_3}$. Both are esters with the same molecular formula, so mass spectrometry alone cannot distinguish them — but integration ratios can, because the proton distributions differ. For methyl propanoate the environments are: $\mathrm{OCH_3}$ (3H, singlet, ~3.67 ppm), $\mathrm{CH_2}$ (2H, quartet, ~2.32 ppm), $\mathrm{CH_3}$ of the ethyl-like propanoyl chain (3H, triplet, ~1.13 ppm) — predicted ratio $3:2:3$, summing to $8$. For ethyl acetate the environments are: $\mathrm{COCH_3}$ (3H, singlet, ~2.05 ppm), $\mathrm{OCH_2}$ (2H, quartet, ~4.12 ppm), $\mathrm{CH_3}$ of ethyl (3H, triplet, ~1.25 ppm) — also predicted ratio $3:2:3$. Integration alone gives the identical ratio $3:2:3$ for both isomers — a crucial limiting case showing that integration is necessary but not sufficient for full structure determination. The two compounds are distinguished not by integration but by the actual chemical-shift values (the $\mathrm{OCH_2}$ of ethyl acetate resonates near 4.12 ppm, strongly deshielded by the adjacent ester oxygen, whereas methyl propanoate's $\mathrm{OCH_3}$ resonates near 3.67 ppm) and, definitively, by the splitting patterns — this is exactly the handoff point where integration must be combined with chemical shift correlation and with spin–spin splitting to complete a structure determination.

Example 3 (exam/competition level) — Reconstructing an unknown formula from integration data with an exchangeable proton trap. An unknown compound shows a mass spectrum parent ion at $m/z = 74$, and its ¹H NMR spectrum (in CDCl₃) shows four signals with raw integrals:

Note only three signals are listed with integrals here — a fourth broad exchangeable proton signal was too weak/broad to integrate reliably (common for OH/NH/COOH in impure or wet samples). Assume the parent formula, from $m/z=74$ and typical fragmentation of a carboxylic acid, is $\mathrm{C_3H_6O_2}$ (propanoic acid, $\mathrm{CH_3CH_2CO_2H}$), giving $N = 6$ total protons. Sum the integrated (non-exchangeable) ratio units:

This matches $\mathrm{CH_2}$ (2H, quartet, from coupling to the adjacent methyl — foreshadowing spin–spin splitting) and $\mathrm{CH_3}$ (3H, triplet). Sum $= 2+3 = 5$ ratio units for the non-exchangeable protons, but the molecular formula demands $N=6$ total. The missing proton, $6 - 5 = 1$, must be the carboxylic acid $\mathrm{OH}$ — consistent with the broad downfield signal at $\delta$ 11.8 that was observed but whose integral was unreliable (broad exchangeable peaks routinely under-integrate due to line-broadening from chemical exchange, as anticipated by Mechanism 1's linewidth-independence argument, which assumes an ideal Lorentzian that real exchange-broadened peaks only approximate). Absolute proton counts:

confirming propanoic acid. As a limiting-case sanity check: if the compound had instead been the methyl ester of acetic acid's higher homolog with no exchangeable proton at all, every proton would integrate cleanly and the sum of integrals would equal $N$ exactly with no "missing" unit — the appearance of a formula-integral mismatch is itself the diagnostic signal that an exchangeable or otherwise non-integrating proton is present.

Quantitative NMR integration as a routine structural tool emerged from the broader development of nuclear magnetic resonance spectroscopy pioneered by Felix Bloch and Edward Purcell, who independently detected NMR in bulk matter in 1946 and shared the 1952 Nobel Prize in Physics. Bloch, working at Stanford, developed the phenomenological equations describing bulk magnetization (the Bloch equations), which directly underlie the derivation in Mechanism 1 connecting equilibrium magnetization, RF pulses, and detected signal. Purcell, at Harvard, approached the same phenomenon from a more quantum-statistical direction, emphasizing the Boltzmann population argument that gives the proportionality between magnetization and spin count used throughout this article. Neither man set out to build an organic-chemistry structure-elucidation tool — their early experiments were on paraffin wax and water — but within a decade chemists including James Arnold, Herbert Gutowsky, and later industrial NMR groups at Varian Associates realized that the chemical shift and, crucially, the quantitative integral of resonances could unlock organic structure determination on a routine basis. The transition from continuous-wave NMR (where integration was performed by mechanically or electronically tracing a step-function overlay on the spectrum, literally an analog integrating circuit) to Fourier-transform NMR in the 1970s (pioneered by Richard Ernst, who won the 1991 Nobel Prize in Chemistry) made integration far faster, more sensitive, and — with proper relaxation delays — more quantitatively reliable, because a full FID could be digitally integrated with high precision rather than read by eye off a chart recorder.