NMR yield and isomer ratios from a crude spectrum: what relaxation is, and why a short delay gives wrong integrals
Share
NMR yield and isomer ratios from a crude spectrum: what relaxation is, and why a short delay gives wrong integrals
The integral of a signal is proportional to the number of nuclei only if the magnetisation has fully recovered before each pulse. This article explains relaxation at the molecular level and its consequences for yields, conversions and isomer ratios
A 1H NMR spectrum of a crude reaction mixture is the fastest way to assess the outcome of a reaction. With a weighed internal standard, the spectrum gives a yield. Without a standard, it gives a conversion or an isomer ratio.
All three values are ratios of integrals, and they are correct only if each integral is proportional to the number of nuclei that produce the signal. In a routine spectrum this condition is met only approximately. In the example of this article, one signal is recorded at 93% of its full intensity and another at 99%, and the yield calculated from them is overestimated by 6.5%.
The cause is incomplete relaxation. In a routine experiment the scan is repeated every few seconds, and some nuclei are not fully relaxed when the next scan starts. Their signals are then weaker than they should be, and a ratio of integrals is no longer correct (Figure 1).
Figure 1. Two signals with equal numbers of nuclei, recorded in four scans. The magnetisation of the signal that relaxes slowly has not recovered when the next scan starts, so from the second scan onwards this signal is recorded at a lower intensity. Schematic, drawn for a 90° pulse.
The first half of this article explains relaxation: the origin of the NMR signal, the effect of a pulse, and the molecular process by which the nuclei return to equilibrium. The second half gives measured relaxation times, the size of the error, a check of the acquisition parameters, and the conditions for reliable yields, conversions and isomer ratios.
Eight terms used in this article. The magnetisation is the vector sum of the magnetic moments of all the nuclei that give one signal. At equilibrium it points along the external magnetic field. The pulse angle is the angle by which a radiofrequency pulse rotates the magnetisation away from the external field. T1, the longitudinal relaxation time, is the time constant of the exponential recovery of the magnetisation along the external field after a pulse. The acquisition time is the period after each pulse during which the signal is recorded. The relaxation delay is the time between the end of the acquisition and the next pulse; on Bruker spectrometers this parameter is called d1. The time between pulses is the sum of the acquisition time and the relaxation delay. An internal standard is a compound that is added to the sample in a known amount. Its signal is used as the reference to calculate the amounts of the other compounds in the solution. The NMR yield is the amount of product in the crude mixture, determined by comparison with the internal standard and expressed as a percentage of the theoretical amount.
At a glance: what can be determined from one crude spectrum
| Quantity | Signals that are compared | Sources of error |
|---|---|---|
| NMR yield | A signal of the product and the signal of a weighed internal standard | A time between pulses that is too short, when the product and the standard have different T1 values. Loss, exchange or reaction of the standard. Overlap of signals |
| Conversion | A signal of the product and a signal of the starting material | A time between pulses that is too short, when the two signals have different T1 values. Material that is not in solution and therefore gives no signal |
| Isomer ratio | The signals of the same proton in the two isomers | Overlap of the two signals. 13C satellites of the major isomer. Noise on the signal of the minor isomer. Rotamers mistaken for isomers |
| Ratio of two nuclei, for example 19F to 1H | Signals in two spectra, related through a reference compound that contains both nuclei | A time between pulses that is too short, in either of the two spectra |
The origin of the NMR signal: the magnetisation
Each 1H nucleus has a magnetic moment: it behaves as a very small bar magnet. An NMR sample contains about 1020 such nuclei.[2]
In the external magnetic field of the spectrometer, each magnetic moment precesses: it rotates around the direction of the field at a constant angle to the field.[1] The frequency of precession is the resonance frequency, also called the Larmor frequency. For 1H in a 400 MHz spectrometer it is 400 million rotations per second.
Many diagrams show the magnetic moments in two groups, one aligned with the field and one aligned against it. This picture is a simplification.[1] The energy of a magnetic moment in the field is far smaller than the thermal energy of the molecules. The magnetic moments therefore point in all directions, with only a very small excess along the field.[1, 2]
The spectrometer detects the vector sum of the magnetic moments of all the nuclei that give one signal. This sum is the magnetisation. The magnetic moments that are distributed evenly over all directions cancel in the sum. Only the small excess along the field remains, so at equilibrium the magnetisation is a single vector along the field (Figure 2).[2, 3]
For 1H at 25 °C and 400 MHz, the magnetisation is equal to that of 32 nuclei per million aligned exactly with the field. At 600 MHz the value is 48 per million. One proton site in 0.6 mL of a 10 mM solution contains 3.6 × 1018 nuclei, which at 400 MHz give the magnetisation of 1.2 × 1014 fully aligned nuclei.
Figure 2. The magnetisation is the vector sum of the nuclear magnetic moments. The excess along the field is too small to be drawn, so the first two panels look identical. The value of 32 per million is our calculation for 1H at 400 MHz and 25 °C.
The pulse and the recorded signal
The spectrometer detects the magnetisation with a receiver coil. Only a magnetic field that changes with time induces a voltage in a coil, so the constant magnetisation of a sample at equilibrium gives no signal.[2]
A radiofrequency pulse is a weak magnetic field that is perpendicular to the external field and oscillates at the resonance frequency. Because this frequency is also the precession frequency, the effect of the pulse on each magnetic moment accumulates, and every magnetic moment is rotated by the same angle around the same axis.[1, 2] This angle is the pulse angle, and it increases with the strength and the duration of the pulse.[2]
The pulse therefore rotates the whole distribution of magnetic moments, and not only the small excess along the field:[1]
- The magnetic moments that are distributed evenly over all directions remain evenly distributed. Their vector sum is zero before and after the pulse, so they induce no net voltage in the coil.
- The excess is rotated with them, so the magnetisation makes an angle with the field that is equal to the pulse angle.
After the pulse, the magnetic moments continue to precess, and the magnetisation precesses with them. Its longitudinal component is parallel to the field and does not rotate, so it induces no voltage. Its transverse component is perpendicular to the field and rotates at the resonance frequency, so it induces an oscillating voltage in the coil. This voltage is recorded as the free induction decay (FID), and the Fourier transform of the FID is the spectrum (Figure 3).[2]
The pulse angle determines the size of the two components:
- A 90° pulse converts all of the magnetisation into transverse magnetisation.
- A 30° pulse gives a transverse component of 50% of the magnetisation (sin 30° = 0.50) and leaves a longitudinal component of 87% (cos 30° = 0.87). The two components are perpendicular, so the percentages do not add up to 100%.
- A 180° pulse inverts the magnetisation and gives no transverse component, so there is no signal.
The intensity of a signal is therefore proportional to the longitudinal magnetisation immediately before the pulse and to the sine of the pulse angle.[2] For a single pulse on a sample at equilibrium, this intensity is proportional to the number of nuclei that produce the signal.
Figure 3. From the pulse to the line in the spectrum. Schematic.
A spectrum is recorded with many pulses
A spectrum is the sum of many scans, and each scan consists of a pulse, the acquisition of the FID and the relaxation delay.[4, 5]
The first pulse acts on the equilibrium magnetisation. Each later pulse acts on the longitudinal magnetisation that is present at that time, which is zero immediately after a 90° pulse and then recovers gradually.
The integral of a signal is therefore proportional to the number of nuclei only if the longitudinal magnetisation has fully recovered before each pulse. This recovery is longitudinal relaxation.
T1: how the magnetisation recovers after a pulse
Immediately after a 90° pulse, the excess of magnetic moments is perpendicular to the field, and there is no excess along the field. The longitudinal magnetisation recovers as the individual magnetic moments change their angles to the field, until the equilibrium excess is restored.[1, 3]
The external field cannot restore this excess, because it only makes each magnetic moment precess at a constant angle. The angle changes only under a second magnetic field, one that is perpendicular to the external field and oscillates at the resonance frequency. During a pulse, this second field comes from the spectrometer. After the pulse, it can come only from within the sample.[3]
The cause of relaxation: local magnetic fields
Each nucleus is exposed to weak local fields, which are produced by the magnetic moments close to it. For a proton there are two main sources:[3]
- Neighbouring nuclei. The contribution of a neighbouring nucleus to the relaxation rate decreases with the sixth power of the distance. At the same distance, a proton contributes 16 times more than a 13C nucleus.
- Unpaired electrons. Dissolved oxygen and some transition-metal compounds are paramagnetic. The magnetic moment of an unpaired electron is much larger than that of a nucleus, so low concentrations of these species are sufficient to increase the relaxation rate.
Molecules in solution rotate and collide continuously, so the local field at each nucleus fluctuates randomly. The component of these fluctuations at the resonance frequency acts on the magnetic moment as a pulse does, but it is very weak, random in direction and different for every nucleus.[3] Each magnetic moment is therefore reoriented in small, random steps.
Through many such steps, the nuclei transfer the energy absorbed from the pulse to the thermal motion of the molecules, which is called the lattice, and the distribution of magnetic moments returns to equilibrium.[1, 3] Longitudinal relaxation is therefore also called spin–lattice relaxation.[3] The relaxation rate is not a property of the nucleus alone: it depends on the magnetic moments close to the nucleus and on the motion of the molecule.
The time constant T1
The longitudinal magnetisation recovers exponentially with the time constant T1, the longitudinal relaxation time.[3] After a 90° pulse, the recovered fraction of the equilibrium magnetisation is:
- 63.2% after T1;
- 95.0% after three times T1;
- 99.3% after five times T1.
Figure 4 shows two signals of one sample. After 5 s, the signal with T1 = 1.5 s has recovered to 96.4% and the signal with T1 = 4.7 s to 65.5%. A second 90° pulse at this time records the two signals at these fractions of their full intensity.
Figure 4. Recovery of the magnetisation along the field after a 90° pulse, for two signals with T1 = 1.5 s and 4.7 s. Calculated.
The effect of molecular tumbling
The local fields fluctuate because the molecule tumbles, and the rate of tumbling determines the fraction of the fluctuations that occurs at the resonance frequency. This rate is described by the correlation time, the time over which a molecule keeps its orientation. The correlation time is tens of picoseconds for small molecules and a few nanoseconds for proteins.[3]
Relaxation is fastest when the rate of tumbling matches the resonance frequency, which at 400 MHz corresponds to a correlation time of 0.4 ns. With faster or slower tumbling, a smaller fraction of the fluctuations occurs at the resonance frequency, and relaxation is slower (Figure 5).[3]
The correlation times of small molecules are shorter than 0.4 ns. For small molecules, slower tumbling therefore gives faster relaxation: a larger molecule or a more viscous solution has a shorter T1.[3, 9]
Figure 5. Rate of T1 relaxation as a function of the correlation time at 400 MHz, relative to its maximum value. T1 is the inverse of the rate, so T1 is long at both ends of the curve. Calculated from the spectral density given in source 3.
Consequences for a spectrum
The mechanism has three consequences:
- Each signal has its own T1, which depends on the protons close to the observed proton and on the motion of the molecule.
- A proton without neighbouring protons relaxes slowly, because its main source of local fields is absent.
- A sample without dissolved oxygen relaxes more slowly than the same sample in air.
T1 is therefore not a constant of a compound: it changes with the solvent, with the dissolved oxygen and with the other components of the mixture.
T2 and the width of the line
A second relaxation process acts on the transverse magnetisation and causes the decay of the FID. The local fields are different for every nucleus, so the magnetic moments do not precess at exactly the same frequency and gradually lose their common direction in the plane perpendicular to the field. Their vector sum therefore decays. The time constant of this decay is T2, the transverse relaxation time.[3]
T2 determines the width of the line: the width at half height is 1/(πT2). Part 7 uses this relation for quadrupolar nuclei. For small molecules in solution, relaxation caused by molecular tumbling gives equal values of T1 and T2.[3]
T2 affects the shape of a line but not its area, because the area is determined by the transverse magnetisation at the start of the acquisition. For integration, the only consequence is that a broad line requires wider integration limits, as described in Part 4.
Measured T1 values
The error in a ratio of integrals depends on the length of the T1 values and on their difference between the signals of one spectrum. For organic molecules under ordinary conditions, the T1 of a proton is usually between 0.5 and 4 s.[4] Within one molecule the values can differ widely: for the protons of strychnine in CDCl3, the fitted values range from 0.53 s to 3.7 s.[7]
For the compounds used as internal standards, T1 values have been measured in several solvents.[6]
T1 of reference compounds, in seconds
| Compound | Signal | CDCl3 | DMSO-d6 | CD3OD |
|---|---|---|---|---|
| 1,3,5-Trimethoxybenzene | Aromatic H, 6.1 ppm | 4.7 | 3.2 | 4.8 |
| 1,3,5-Trimethoxybenzene | OCH3, 3.8 ppm | 2.2 | 1.4 | 2.7 |
| Dimethyl sulfone | CH3, 3.0 ppm | 2.7 | 2.4 | 3.3 |
| Dimethyl terephthalate | Aromatic H, 8.1 ppm | 3.6 | 2.9 | 4.4 |
| Dimethyl terephthalate | OCH3, 4.0 ppm | 1.8 | 1.1 | 2.4 |
| 1,2,4,5-Tetramethylbenzene | Aromatic H, 7.0 ppm | 6.1 | 4.7 | 5.9 |
| 1,2,4,5-Tetramethylbenzene | CH3, 2.2 ppm | 4.0 | 2.6 | 4.3 |
| 1,2,4,5-Tetrachloro-3-nitrobenzene | Aromatic H, 7.8 ppm | 10.7 | 12.6 | 6.4 |
| Pentachlorobenzene | Aromatic H, 7.6 ppm | 5.8 | 13.8 | 8.2 |
Each compound was measured on its own, at about 5 to 20 mg/mL and 25 °C. The chemical shifts are those in CDCl3.[6]
The table shows three effects:
- Protons without neighbouring protons relax most slowly. The only proton of 1,2,4,5-tetrachloro-3-nitrobenzene has a T1 of 10.7 s in CDCl3.
- Aromatic protons relax more slowly than methyl or methoxy protons of the same molecule. In 1,3,5-trimethoxybenzene, T1 is 4.7 s for the aromatic protons and 2.2 s for the methoxy protons. Each methyl proton has two other protons close to it, which shortens its T1.
- T1 depends on the solvent. The aromatic signal of 1,3,5-trimethoxybenzene has a T1 of 4.7 s in CDCl3 and 3.2 s in DMSO-d6.
The source states that the values can differ in a mixture and recommends that they are checked before a quantitative experiment.[6]
In these examples, the T1 values within one molecule differ by a factor of 1.5 to 7. This difference is the origin of the error in a ratio of integrals.
Figure 6. Four reference compounds and their T1 values in CDCl3. Values from source 6.
The effect of dissolved oxygen
Paramagnetic ions shorten T1, and the removal of dissolved oxygen usually lengthens it.[7] For pyridine in methanol-d4 at 60 MHz, the T1 values of the protons were 4.8 to 5.3 s in a sample prepared in air. In a degassed sample, which also contained an iridium catalyst, they were 19.6 to 25.4 s. The source attributes the increase to the absence of oxygen.[8]
This effect is relevant for air-sensitive compounds. A sample prepared under inert gas in an NMR tube with a PTFE valve contains less oxygen than a routine sample, so its T1 values are longer and it requires a longer time between pulses.
Between two pulses: acquisition time, relaxation delay and pulse angle
When T1 is known, the fraction of each signal that is recorded can be calculated from two acquisition parameters: the time between pulses and the pulse angle.
The longitudinal magnetisation recovers during the acquisition time as well as during the relaxation delay, so the time between pulses, and not the relaxation delay alone, determines this fraction.[4, 5]
Acquisition with a 90° pulse
After a 90° pulse no longitudinal magnetisation remains, so each scan records the magnetisation that has recovered since the previous pulse. A recovery of 99.3% requires a time between pulses of five times T1, which is the rule given in Part 4.
One facility guide gives an example: for a longest T1 of 4 s, the time between pulses is 20 s, divided into an acquisition time of 5 s and a relaxation delay of 15 s. When T1 is not known, the same guide suggests a relaxation delay of 30 s for 1H, 19F and 31P, and 60 s for 13C.[5]
Acquisition with a smaller pulse angle
Routine spectra are not recorded in this way, because 20 to 30 s per scan makes the experiment slow. A smaller pulse angle gives less signal per scan but leaves most of the longitudinal magnetisation (87% for a 30° pulse), so the scans can be repeated more quickly.
The pulse angle that gives the highest signal per unit of time is the Ernst angle, cos θ = exp(−t / T1), where θ is the pulse angle and t is the time between pulses.[11] The Ernst angle is 53° for t = 0.5 × T1 and 82° for t = 2 × T1. One university facility lists a 30° pulse and a relaxation delay of 1 s as its standard 1H experiment.[10] Routine acquisition parameters are chosen for the highest signal in a given time, and not for correct ratios of integrals.
With a small pulse angle and a short time between pulses, the longitudinal magnetisation reaches a steady state after a few scans. Its value before each pulse is then a constant fraction of the equilibrium value:[11]
fraction = (1 − E) / (1 − E cos θ), with E = exp(−t / T1)
The equation is valid when no transverse magnetisation remains at the next pulse.[11] Each signal is recorded at this fraction of its full intensity.
Fraction of the full signal intensity that is recorded
| Pulse angle | t = 1 × T1 | 2 × T1 | 3 × T1 | 5 × T1 |
|---|---|---|---|---|
| 30° | 92.8% | 97.9% | 99.3% | 99.9% |
| 45° | 85.4% | 95.6% | 98.5% | 99.8% |
| 90° | 63.2% | 86.5% | 95.0% | 99.3% |
Our calculation with the equation above. t is the time between pulses.
A 30° pulse with a time between pulses of three times T1 records 99.3% of the signal, the same fraction as a 90° pulse with five times T1. The facility of source 10 gives the same two rules.[10] At a time between pulses equal to T1, the fractions are 92.8% for a 30° pulse and 63.2% for a 90° pulse.
The effect on a ratio of integrals
The recorded fraction depends on T1, so it is different for each signal. The measured ratio of two signals therefore differs from the true ratio by a factor equal to the ratio of their recorded fractions.
The example is an NMR yield with 1,3,5-trimethoxybenzene as the internal standard in CDCl3. The T1 of the aromatic signal of the standard is 4.7 s,[6] and a T1 of 1.5 s is assumed for the signal of the product. With a 30° pulse and a relaxation delay of 1 s, the time between pulses is 3 s or 5 s, for an acquisition time of 2 s or 4 s.
NMR yield with a standard that relaxes more slowly than the product
| Pulse angle | Time between pulses | Standard recorded at | Product recorded at | Yield overestimated by |
|---|---|---|---|---|
| 30° | 3 s | 87.0% | 97.9% | 12.6% |
| 30° | 5 s | 93.4% | 99.5% | 6.5% |
| 90° | 5 s | 65.5% | 96.4% | 47.3% |
| 30° | 14.1 s (3 × 4.7 s) | 99.3% | 100.0% | 0.7% |
| 90° | 23.5 s (5 × 4.7 s) | 99.3% | 100.0% | 0.7% |
Our calculation. The values in the last column are relative errors: with an error of 6.5%, a true yield of 80% is reported as about 85%.
In the second row, the signal of the standard is recorded at 93.4% of its full intensity and the signal of the product at 99.5%. The yield is therefore overestimated by a factor of 99.5 / 93.4 = 1.065, that is, by 6.5%. These are the values quoted in the introduction.
The direction of the error depends on which signal relaxes more slowly:
- If the signal of the standard relaxes more slowly, the yield is overestimated.
- If the signal of the product relaxes more slowly, for example the signal of an isolated aromatic proton, the yield is underestimated.
In this example, the error is below 1% when the time between pulses is longer than about 13 s with a 30° pulse, or about 22 s with a 90° pulse (Figure 7).
Figure 7. Error in the NMR yield of the example as a function of the time between pulses. Calculated.
13C and other nuclei
The same calculation applies to every nucleus, with different T1 values. For 13C they range from 0.1 s to tens of seconds, and the longest values belong to quaternary carbons.[5] A quaternary carbon has no directly attached proton, so its main source of local fields is absent.
A routine 13C spectrum is not quantitative for two reasons. The signals of quaternary carbons are recorded at much less than their full intensity, and proton decoupling changes the signal intensities through the nuclear Overhauser effect. A quantitative 13C or 31P spectrum is therefore recorded with inverse-gated decoupling and a long relaxation delay, and a relaxation agent can be added when the T1 values are very long.[5]
Measuring T1
The rules of five and three times T1 require the longest T1 among the signals that are compared. T1 can be measured, or the time between pulses can be tested directly.
T1 is measured by inversion recovery. A 180° pulse inverts the magnetisation, and after a variable delay a 90° pulse records how much of it has recovered.[2, 3, 4] With a short delay, the magnetisation is still inverted and the line is negative. With longer delays, the intensity of the line passes through zero and becomes positive (Figure 8).
- Quick estimate. T1 is approximately 1.4 times the delay at which the intensity of the line is zero.[3] A few spectra with different delays are sufficient to find this delay for the signal that relaxes most slowly.
- Full measurement. Spectra are recorded for a series of delays, and the intensities are fitted to an exponential function.[4] The time between scans must be four to five times the expected T1, and one facility guide gives 30 s as the minimum for organic compounds.[4]
Figure 8. Inversion recovery of two signals with T1 = 1.5 s and 4.7 s. The intensity of each line is zero at a delay of 0.69 times its T1. Calculated.
A check that does not require T1
The spectrum is recorded a second time with a much longer relaxation delay. If the ratio of the integrals is unchanged, the first relaxation delay was long enough. If the ratio changes, at least one signal was recorded at less than its full intensity in the first spectrum, which is called partial saturation.
The check follows from the equation for the recorded fraction: the ratio of two integrals changes with the time between pulses only while the two recorded fractions are different.
A measured example: fluoride per cation in TMAF
The NorrChemica Lab Journal reports the fluoride content of the TMAF–MeOH adduct, determined by quantitative 1H and 19F NMR as the number of fluoride ions per tetramethylammonium cation. The complete description is in the article TMAF–MeOH adduct: 91 mol% active fluoride by quantitative 1H/19F NMR.
The conditions were:
- The solvent was CD3OD, and the spectra were recorded at 400 MHz for 1H and at 376 MHz for 19F.
- 1,3,5-Trifluorobenzene was the reference compound for both nuclei. It contains three protons and three fluorine atoms, so it relates the 19F spectrum to the 1H spectrum, and its amount cancels in the ratio.
- A 90° pulse and a relaxation delay of 30 s were used for both nuclei.
The adduct contained 0.91 fluoride ions per cation, and two commercial samples of anhydrous TMAF contained 0.92 and 0.86 (Figure 9).

Figure 9. Quantitative 1H and 19F spectra of the TMAF–MeOH adduct and of commercial anhydrous TMAF, with 1,3,5-trifluorobenzene as the reference. Measured spectra from the NorrChemica Lab Journal.
The result is calculated from the integrals of four signals: fluoride and the reference compound in the 19F spectrum, and the cation and the reference compound in the 1H spectrum. Each signal has its own T1, so with a relaxation delay that is too short the size of the error could not be predicted without the four T1 values.
The Lab Journal article does not report T1 values. The measurement used the relaxation delay recommended when T1 is not known, and the delay was checked as described above: spectra recorded with shorter relaxation delays gave materially different integrals, which indicates partial saturation.
Fraction of the signal recorded with a 90° pulse, for two different values of the time between pulses
| T1 of the signal | 5 s between pulses | 34 s between pulses |
|---|---|---|
| 2 s | 91.8% | 100.0% |
| 5 s | 63.2% | 99.9% |
| 7 s | 51.0% | 99.2% |
Our calculation. The time of 34 s is the relaxation delay of 30 s plus an assumed acquisition time of 4 s.
With 34 s between pulses, every signal with a T1 of up to 7 s is recorded at 99% or more of its full intensity. With 5 s between pulses, a signal with a T1 of 5 s is recorded at 63%.
NMR yield with an internal standard
A known amount of an internal standard is added to the crude mixture, and the amount of product is calculated from two integrals:[5]
n(product) = n(standard) × [ I(product) / N(product) ] ÷ [ I(standard) / N(standard) ]
In this equation, n is an amount in moles, I is the integral of a signal, and N is the number of nuclei that produce this signal. If the purity of the standard is below 100%, the amount of the standard is multiplied by its purity.[5]
The following example uses invented numbers:
- A reaction is performed on a 0.100 mmol scale, and 0.100 mmol of 1,3,5-trimethoxybenzene (16.8 mg) is added to the crude mixture.
- The integral of the aromatic signal of the standard (3 H) is set to 3.00.
- A signal of the product that corresponds to 2 H has an integral of 1.42.
- The amount of product is 0.100 × (1.42 / 2) ÷ (3.00 / 3) = 0.071 mmol, so the NMR yield is 71%.
Choice of the internal standard
One facility guide lists the properties of a suitable standard:[5]
- It contains the nucleus that is observed, and its signals do not overlap with the signals of the sample.
- It does not react with the sample or with the solvent.
- It dissolves completely and is not volatile. If part of the standard is lost when the sample is concentrated, the yield is overestimated.
- Its signals are simple, preferably singlets, and its T1 is not very long.
- It is available in pure form and is easy to weigh.
The same guide names maleic acid, dimethyl sulfone, 1,3,5-trimethoxybenzene, 1,4-dinitrobenzene and 3,4,5-trichloropyridine among the standards for 1H.[5] A table of certified standards is given in Part 4.
1,3,5-Trimethoxybenzene as the standard
1,3,5-Trimethoxybenzene gives two singlets: one at 6.1 ppm for the three aromatic protons, and one near 3.8 ppm for the nine methoxy protons.[5, 6] Three of its properties are relevant for an NMR yield:
- The aromatic signal relaxes slowly. Its T1 in CDCl3 is 4.7 s, so the time between pulses must be 23.5 s with a 90° pulse, or 14 s with a 30° pulse.[6]
- The methoxy signal relaxes faster. Its T1 in CDCl3 is 2.2 s, so the time between pulses must be 11 s with a 90° pulse.[6] This signal can be used as the reference if no other signal overlaps with it.
- 1,3,5-Trimethoxybenzene is sensitive to acid. One industrial source reports that it can exchange hydrogen for deuterium under acidic conditions, and that the assay value then increases with time. The remedy given by this source is to use a different standard.[12] An increase is the expected result of this exchange, because the exchange lowers the 1H integral of the standard.
The residual solvent signal as the standard
One published method uses the residual CHCl3 signal of the solvent as the reference, after the concentration of CHCl3 in the batch of solvent has been determined. According to the publication, the accuracy is comparable to that obtained with 1,3,5-trimethoxybenzene when the guidelines of the method are followed, and the sample can be recovered by evaporation of the solvent.[13]
Other requirements for the spectrum
A quantitative spectrum has further requirements:[5]
- The signal-to-noise ratio is at least 250:1, which is required for an integration error below 1%.
- The phase and the baseline are corrected, and the result of the correction is checked visually.
- The integration limits extend at least 20 line widths on each side of the signal, and the same limits are used for every signal.
- The 13C satellites are either included for every signal or excluded for every signal.
- The signal that is integrated is assigned with certainty and does not overlap with other signals. Signals of exchangeable protons (NH, OH) are not used.
Part 4 describes these requirements and the weighing in detail. For measurements of this kind, NorrChemica supplies NMR tubes made for quantitative NMR.
Conversion and isomer ratios without a standard
A conversion and an isomer ratio require no internal standard, because both signals that are compared belong to compounds of the reaction mixture.
Conversion
A conversion is calculated from a signal of the product and a signal of the starting material. It has two limitations:
- The two signals belong to different molecules with different T1 values, so the error described for the NMR yield also applies to a conversion.
- A spectrum shows only the compounds that are in solution. Material that has left the solution as a gas or as a solid gives no signal. A conversion is therefore not a yield.
Isomer ratios
An isomer ratio is calculated from the signals of the same proton in the two isomers. Its determination is one of the typical applications of quantitative NMR.[5]
Relaxation causes a smaller error in an isomer ratio than in a yield. Two signals with the same T1 are recorded at the same fraction of their full intensity, so their ratio is correct for any time between pulses. The same proton in two diastereomers has the same neighbouring protons, so the two T1 values are usually similar. This similarity is an expectation and not a rule, and it can be checked with a second spectrum recorded with a longer relaxation delay.
Overlap of signals
The most common difficulty is overlap of the signals: the spectra of larger molecules are crowded, and the difference in chemical shift between two isomers can be small.[14]
In one published example, three methods gave three different ratios for the same crude mixture of two diastereomers:[14]
- 7:93 by integration of the conventional 1H spectrum;
- 8:92 by integration of a pure shift spectrum, in which each multiplet is reduced to a singlet;
- 12:88 by line-shape fitting of the pure shift spectrum.
The minor isomer and the 13C satellites
At ratios such as 95:5 or 99:1, the low intensity of the signal of the minor isomer causes two additional limitations:
- Noise. An integration error below 1% requires a signal-to-noise ratio of 250:1.[5] For the minor isomer of a 95:5 mixture, this value corresponds to about 4,750:1 for the major isomer (250 × 95 / 5).
- 13C satellites. Each 13C satellite has 0.55% of the intensity of its parent signal, as described in Part 4, and routine 1H spectra are recorded without 13C decoupling.[5] At a ratio of 99:1, one satellite of the major isomer has about half the intensity of the signal of the minor isomer (Figure 10). A satellite that overlaps with this signal is added to its integral.
Figure 10. The signal of a minor isomer and the 13C satellites of the major isomer, at ratios of 95:5 and 99:1. Calculated for a line width of 1.2 Hz, with the satellites at 70 Hz on each side of the signal of the major isomer (an assumed one-bond coupling constant of 140 Hz).
Rotamers
Two sets of signals with a constant ratio do not always belong to two isomers. Rotamers of a single compound give the same appearance, and Part 8 explains how to distinguish the two cases.
The limitations of an NMR yield
- An NMR yield is not an isolated yield. It is the amount of product in the crude mixture, calculated from one signal. An isolated yield is calculated from the mass of product obtained after purification.
- An NMR yield depends on the assignment of the signal. If a signal of a by-product overlaps with the signal of the product, the by-product is counted as product.
- With routine acquisition parameters, the error caused by relaxation alone can be several per cent. In the example, it was 6.5% or 12.6%, depending on the acquisition time. A difference of two or three per cent between two experiments is then smaller than the error of the method.
- With quantitative acquisition parameters, the integration error can be kept below 1%.[5] The standard must then be weighed with the same accuracy.
A published study compared the known composition of mixtures with the ratios determined for these mixtures by HPLC, GC and NMR, and it also examined the accuracy of weighing at different scales.[15]
A reader can assess an NMR yield when the report states:
- the internal standard and the two signals that were integrated;
- the pulse angle, the acquisition time and the relaxation delay;
- whether T1 was measured, or whether the relaxation delay was checked with a second spectrum.
Summary
- The NMR signal is produced by the magnetisation, which is the vector sum of the magnetic moments of the nuclei. A pulse rotates all the magnetic moments, but only the small excess along the field gives a signal.
- A spectrum is the sum of many scans. From the second scan onwards, the intensity of each signal depends on the longitudinal magnetisation that has recovered since the previous pulse.
- The recovery is caused by fluctuating local magnetic fields, mainly those of neighbouring protons and of dissolved oxygen. Its time constant is T1, and each signal has its own T1. Protons without neighbouring protons relax most slowly, with measured T1 values of 10 s and more.
- For a recovery of 99.3%, the time between pulses must be five times T1 with a 90° pulse, or three times T1 with a 30° pulse.
- With a shorter time between pulses, two signals with different T1 values are recorded at different fractions of their full intensity. In the example, the yield is overestimated by 6.5% with a 30° pulse and 5 s between pulses.
- When T1 is not known, a second spectrum recorded with a longer relaxation delay shows whether the first relaxation delay was long enough.
- Isomer ratios are limited by overlap of signals, by noise on the signal of the minor isomer and by the 13C satellites of the major isomer.
Quantitative NMR from NorrChemica
Purity by quantitative NMR against a calibrated standard is Package V of the NMR service. A 1H spectrum is Package I, and 1H with 19F, 31P or 11B is Package II. The service is available worldwide for chemical and pharmaceutical R&D. Each report is signed and is supplied with the native data. Research use only.
- 1,3,5-Trimethoxybenzene internal standard for 1H
- TMAF–MeOH adduct the measured example
- NMR tubes for quantitative NMR yields, ratios and purity
- NMR tubes with PTFE valve degassed samples, samples under gas
- Screw-cap NMR tubes closed or septum top
NMR tubes made in Germany by Deutero and shipped from Finland. See the NMR tube guide →
References
- Hanson LG. Is quantum mechanics necessary for understanding magnetic resonance? Concepts Magn. Reson. Part A 2008, 32A, 329–340. DOI
- Keeler J. Understanding NMR spectroscopy, chapter 3: The vector model. Lecture notes, University of Cambridge, 2002. Link
- Keeler J. Understanding NMR spectroscopy, chapter 8: Relaxation. Lecture notes, University of Cambridge, 2002. Link
- University of Oxford, Department of Chemistry, NMR facility. Measuring T1 relaxation times. Guide, 2025. Link
- University of Oxford, Department of Chemistry, NMR facility. Quantitative NMR spectroscopy. Guide, 2017. Link
- Merck. Standards for quantitative NMR. Flyer FL2747EN, version 1.0. Link
- University of California, Santa Barbara, NMR facility. T1 relaxation measurement. Protocol. Link
- Ashworth D, Clayton L, Mewis R. Using Pulsar to measure spin-lattice relaxation data. Technical note, Oxford Instruments, 2015. Link
- Columbia University, NMR facility. NOESY and ROESY experiments. Facility guide. Link
- KU Leuven, NMR core facility. qNMR: the do’s and don’ts. Seminar slides, 2022. Link
- Spencer RGS, Fishbein KW. Measurement of spin–lattice relaxation times and concentrations in systems with chemical exchange using the one-pulse sequence: breakdown of the Ernst model for partial saturation in nuclear magnetic resonance spectroscopy. J. Magn. Reson. 2000, 142, 120–135.
- Heggarty C, Graham S. 1D quantitative experiments and key areas to consider during analysis. Poster, Almac Sciences. Link
- Muhamadejev R, Melngaile R, Paegle P, Zibarte I, Petrova M, Jaudzems K, Veliks J. Residual solvent signal of CDCl3 as a qNMR internal standard for application in organic chemistry laboratory. J. Org. Chem. 2021, 86, 3890–3896. DOI
- Adams RW, Byrne L, Király P, Foroozandeh M, Paudel L, Nilsson M, Clayden J, Morris GA. Diastereomeric ratio determination by high sensitivity band-selective pure shift NMR spectroscopy. Chem. Commun. 2014, 50, 2512–2514. DOI
- Wernerova M, Hudlicky T. On the practical limits of determining isolated product yields and ratios of stereoisomers: reflections, analysis, and redemption. Synlett 2010, 2701–2707. DOI
About this article. Sources 1 to 9, 11, 12 and 14 were read in their freely available copies. Source 10 was read in an excerpt only. Sources 13 and 15 were read in their abstracts only, and they are cited only for what the abstracts state. Sources 4, 5, 7, 9 and 10 are guides of university NMR facilities, and sources 6, 8 and 12 are technical documents of companies. They were used for measured values and practical parameters that papers seldom state. The population figures, all percentages of recorded signal, the Ernst angles, the yield examples and Figures 3, 4, 6, 7 and 9 are our own calculations with the equations given in the text. Figures 1, 2 and 5 are schematics. Figure 8 is a measured spectrum from the NorrChemica Lab Journal. The sources were read on 13 November 2026, and sources 1, 3, 8 and 12 were read again on 14 November 2026.
NMR Series
- Part 1 — Reading 1H and 13C NMR
- Part 2 — Reading 19F, 31P and 11B NMR
- Part 3 — Reading 2D NMR: COSY, HSQC and HMBC
- Part 4 — Quantitative NMR: how a purity figure is made
- Part 5 — 15N NMR: which nitrogen is alkylated, protonated or coordinated?
- Part 6 — NMR tubes by experiment: which tube for which measurement
- Part 7 — Quadrupolar nuclei in NMR: broad lines and missing carbons
- Part 8 — Two sets of NMR signals: rotamers, diastereomers or an impurity
- Part 9 — NMR yield and isomer ratios: relaxation and the delay between pulses (this article)
- Part 10 — Silicon NMR: the glass signal, the negative NOE and M, D, T, Q units
- Part 11 — Solid-state NMR: magic-angle spinning, cross-polarisation and what a CP-MAS spectrum counts
- Part 12 — Enantiomeric excess by NMR: when NMR is the best method, the chiral auxiliaries and the detection limit for the minor enantiomer
- Part 13 — Predicted NMR shifts: how accurate ChemDraw, MestReNova and other predictors are
- Part 14 — Phosphorus NMR in biology: phospholipids, phosphorylated proteins, oligonucleotide drugs and phosphates in food
- Part 15 — Multiplets in ¹H NMR: coupling trees, second-order patterns and two signals from one CH₂
- Part 16 — NMR for cosmetics: adulterated essential oils, free formaldehyde, aloe vera, rancid lipsticks and mineral oil in lip products
- Part 17 — Polymers by NMR: composition, sequence, chain ends and branches, and what a spectrum of many chains can and cannot measure
- Part 18 — Benchtop NMR: what a permanent-magnet spectrometer does well, and when a sample needs high field
- Part 19 — NOESY, ROESY and TOCSY: regioisomers, E and Z isomers, anomers, and the signals of one compound in a mixture
- Part 20 — Rapid-injection NMR: starting a reaction inside the spectrometer, how fast the spectra can follow it, and the intermediates it has revealed
- Part 21 — Fluorine NMR for PFAS: total organofluorine, what targeted LC-MS/MS does not detect, and the detection limit
- Part 22 — Polysiloxanes by proton NMR: silicone oil in medicines, food and cosmetics, the reactive groups before curing, and the rubber network after it