Reading molecules by the light they absorb
A chemist hands you a vial of colourless liquid and asks what it is. You cannot see molecules, and a liquid's appearance says almost nothing. But every molecule has a ladder of quantised energy levels — electronic, vibrational, rotational, nuclear-spin — and it absorbs or emits light only at the energy differences between them. Measure which energies it absorbs and you are reading its structure. This is spectroscopy, and together with mass spectrometry (which weighs molecules rather than shining light on them) it is how almost every new compound is identified, every drug batch is checked, and every pollutant is measured.
The previous two lessons supplied the energy ladders: Quantum mechanics for chemists gave quantised levels and the harmonic oscillator, and Molecular orbitals gave HOMOs and LUMOs. This lesson turns them into measurements, one technique at a time, and ends by identifying an unknown liquid from its spectra.
Light and energy levels
A photon of frequency , wavelength and wavenumber carries energy
Spectroscopists use whichever unit is convenient, so it pays to know the conversions: 1 eV corresponds to a wavelength of 1240 nm and a wavenumber of 8066 cm⁻¹. A molecule absorbs a photon only if matches the gap between two of its levels and a selection rule allows the jump, which in practice means the light's oscillating electric (or, for NMR, magnetic) field can couple the two states.
The different ladders have very different spacings, so each region of the spectrum probes a different motion:
UV–visible absorption and the Beer–Lambert law
Ultraviolet and visible photons lift electrons between molecular orbitals, usually from the HOMO or nearby filled orbitals to the LUMO. Groups that absorb in this range are called chromophores: conjugated π systems like the dyes and polyenes of the last two lessons, aromatic rings, carbonyl groups, and transition-metal ions. A substance's colour is the light it does not absorb: a dye that absorbs orange-red light near 630 nm looks blue.
A UV–visible spectrophotometer measures how much light of each wavelength passes through a sample.
Definition— Absorbance and the Beer–Lambert law
The transmittance is and the absorbance is . For a dilute solution of one absorbing species at concentration in a cell of path length ,
where , the molar absorption coefficient (units L mol⁻¹ cm⁻¹), is a property of the substance at that wavelength.
Why a logarithm? Each thin slice of solution, thickness , absorbs the same fraction of the light reaching it, proportional to the number of molecules in the slice: . Integrating gives , an exponential fall, and taking logarithms makes it linear in . So absorbance, not transmittance, is proportional to concentration, which is why it is the quantity analysts use. In practice one does not trust a literature ; one measures a calibration line from standards of known concentration, then reads the unknown off it.
import numpy as np
# Calibration standards of a blue dye, measured at 630 nm in a 1.00 cm cuvette (illustrative data)
conc_uM = np.array([0.0, 2.0, 4.0, 6.0, 8.0, 10.0]) # micromol per litre
absorb = np.array([0.003, 0.262, 0.519, 0.786, 1.041, 1.298])
slope, intercept = np.polyfit(conc_uM, absorb, 1) # least-squares straight line
path_cm = 1.00
epsilon = slope / path_cm * 1e6 # per (mol/L) per cm
fit = slope * conc_uM + intercept
r2 = 1 - np.sum((absorb - fit)**2) / np.sum((absorb - absorb.mean())**2)
print(f"slope = {slope:.4f} per uM, intercept = {intercept:.4f}, R^2 = {r2:.5f}")
print(f"molar absorption coefficient = {epsilon:.3g} L mol^-1 cm^-1")
# An unknown, diluted 1 in 20 before measuring, reads A = 0.655
A_unknown, dilution = 0.655, 20
c = (A_unknown - intercept) / slope
print(f"diluted sample: {c:.2f} uM; original: {c * dilution:.0f} uM = {c * dilution * 1e-6:.2e} mol/L")
# Absorbance is logarithmic: what fraction of the light gets through?
for A in [0.1, 0.5, 1.0, 2.0, 3.0]:
print(f"A = {A:3.1f}: {100 * 10**-A:6.2f} % transmitted")
# Why very high absorbances can't be trusted: 0.5 % stray light reaching the detector
stray = 0.005
print("\n true A measured A with 0.5% stray light")
for A_true in [0.5, 1.0, 2.0, 3.0, 4.0]:
T = 10**-A_true
print(f" {A_true:4.1f} {-np.log10((T + stray) / (1 + stray)):6.2f}")The line through the standards has slope 0.1297 per µmol/L, so L mol⁻¹ cm⁻¹: a strongly absorbing dye, as conjugated chromophores often are. The diluted unknown is 5.03 µmol/L, so the original was about mol/L. Note why the unknown was diluted: absorbance 2 means only 1% of the light gets through, and the stray-light table shows that a mere 0.5% of light leaking round the sample caps the measured absorbance near 2.3. Good practice keeps absorbances roughly between 0.1 and 1.
Note
Try it at home, with an adult. Put a few drops of blue or red food colouring in a tall clear glass of water. Pour half into a second identical glass and top it up with water; repeat a few times. Looking down through the glasses from above (a long path) and through their sides (a short path), you can see that the colour depends on concentration and path length together, just as says. The solutions are only for looking at: don't drink them.
Infrared: molecules as vibrating springs
A chemical bond behaves, near its equilibrium length, like a spring. In Quantum mechanics for chemists you met the quantum harmonic oscillator: a particle on a spring of force constant has equally spaced levels.
Definition— Harmonic vibration of a diatomic molecule
Two atoms of masses and joined by a bond of force constant vibrate like one particle of reduced mass . The vibrational levels are , , with
The selection rules for infrared absorption are and that the vibration must change the molecule's electric dipole moment.
So a stiff bond between light atoms vibrates fast, and a weak bond between heavy atoms slowly. The second selection rule means N₂ and O₂, whose vibration leaves them symmetric and non-polar, do not absorb infrared light at all, while CO, HCl, H₂O, CO₂ (in its bending and asymmetric stretching modes) and CH₄ do. That is why the minor gases carbon dioxide, methane and water vapour, not nitrogen and oxygen, determine how the atmosphere absorbs the heat radiated by the Earth (Environmental chemistry).
import numpy as np
u = 1.66053907e-27 # kg, atomic mass constant
c_cm = 2.99792458e10 # speed of light in cm/s
h = 6.62607015e-34
mass = {"H": 1.007825, "D": 2.014102, "C": 12.000000, "O": 15.994915, "35Cl": 34.968853, "37Cl": 36.965903}
def wavenumber(k, a, b):
"""Harmonic vibrational wavenumber (cm^-1) of a diatomic with force constant k (N/m)."""
mu = mass[a] * mass[b] / (mass[a] + mass[b]) * u # reduced mass, kg
return np.sqrt(k / mu) / (2 * np.pi * c_cm)
print("molecule k (N/m) reduced mass (u) harmonic wavenumber (cm^-1)")
for a, b, k in [("H", "35Cl", 516), ("H", "37Cl", 516), ("D", "35Cl", 516), ("C", "O", 1902)]:
mu_u = mass[a] * mass[b] / (mass[a] + mass[b])
print(f"{a + b:8s} {k:8d} {mu_u:14.4f} {wavenumber(k, a, b):14.0f}")
# Equally spaced harmonic levels E_v = (v + 1/2) h c nu~ ; the fundamental is v = 0 -> 1
nu = wavenumber(516, "H", "35Cl")
print(f"\nH35Cl zero-point energy: {0.5 * h * c_cm * nu / 1.602176634e-19:.3f} eV")
print(f"H-35Cl / H-37Cl splitting: {nu - wavenumber(516, 'H', '37Cl'):.1f} cm^-1")
print(f"isotope ratio nu(DCl)/nu(HCl) = {wavenumber(516, 'D', '35Cl') / nu:.4f}")
# Anharmonicity: the measured fundamental of H35Cl is 2886 cm^-1, below the harmonic value
observed = 2886
print(f"measured fundamental {observed} cm^-1 lies {100 * (nu - observed) / nu:.1f}% below the harmonic value")With force constants derived from spectroscopic data (516 N/m for HCl, 1902 N/m for CO; see the NIST Chemistry WebBook), the model puts HCl at 2990 cm⁻¹ and CO at 2170 cm⁻¹. Replacing hydrogen by deuterium leaves the bond, and so , unchanged but nearly doubles , lowering the frequency by a factor of 0.717 — close to . That isotope shift is a tool: chemists swap H for D to find out which band belongs to which bond. Even ³⁵Cl and ³⁷Cl are distinguishable; high-resolution spectra of HCl show every line doubled, about 2 cm⁻¹ apart (the harmonic estimate gives 2.3 cm⁻¹), in the 3:1 ratio of the isotopes' abundances.
The measured HCl fundamental is 2886 cm⁻¹, 3.5% below the harmonic value: real bonds are anharmonic, softer when stretched than a spring, as the figure shows. In polyatomic molecules each bond type has a characteristic range, which is what makes IR useful for spotting functional groups: O–H near 3200–3600 cm⁻¹ (broad when hydrogen bonded), N–H near 3300–3500, C–H near 2850–3100, C≡N near 2250, and the intense C=O stretch near 1650–1750 cm⁻¹ (esters near 1740, ketones near 1715). Below about 1500 cm⁻¹ lies the fingerprint region, a tangle of coupled bending and stretching modes unique to each molecule.
NMR: listening to hydrogen nuclei
Protons, like electrons, have spin ½. In a magnetic field a proton's two spin states differ in energy by , where is the proton's gyromagnetic ratio; radio waves of frequency flip them. A "400 MHz" spectrometer has a superconducting magnet of 9.4 T, about 200 000 times the Earth's field, in which protons resonate at 400 MHz. The same physics, with the body as the sample, is magnetic resonance imaging.
What makes NMR chemically useful is that protons in different environments resonate at slightly different frequencies.
Definition— Chemical shift
The electrons round a nucleus partly shield it from the applied field, so each chemically distinct proton resonates at its own frequency. The chemical shift is
measured from the protons of tetramethylsilane, Si(CH₃)₄ (TMS), whose is defined as 0. Being a ratio, is the same on every spectrometer.
Electronegative neighbours pull electron density away, deshield the proton and raise . Typical ¹H shifts: C–CH₃ in an alkane about 0.9 ppm; CH₃ next to C=O about 2.0–2.5; O–CH₂ or O–CH₃ about 3.3–4.2; aromatic C–H about 7–8; aldehyde CHO about 9–10; carboxylic acid O–H about 10–12. Three more pieces of information sit in every spectrum:
- Integration. The area under each signal is proportional to the number of protons producing it.
- Splitting. The spins of protons on neighbouring carbon atoms (usually three bonds away, H–C–C–H) slightly add to or subtract from the field, splitting each signal. With equivalent neighbours, a signal splits into lines (the n + 1 rule) with intensities from Pascal's triangle — 1:1, 1:2:1, 1:3:3:1 — spaced by the coupling constant , typically about 7 Hz between CH₂ and CH₃ in a chain.
- Equivalence. Protons related by symmetry or fast rotation (the three of a CH₃ group) give one signal and do not split each other.
The block simulates first-order stick spectra of two isomers with the formula C₄H₈O₂, ethyl ethanoate and methyl propanoate, from typical shifts.
from math import comb
FIELD_MHZ = 400.0 # proton frequency of the spectrometer
def multiplet(delta, n_h, n_neighbours, J):
"""First-order (n+1 rule) lines: list of (ppm, intensity). The group's total intensity is n_h."""
n = n_neighbours
lines = []
for k in range(n + 1):
offset_hz = (k - n / 2) * J
lines.append((delta + offset_hz / FIELD_MHZ, n_h * comb(n, k) / 2**n))
return lines
NAMES = {1: "singlet", 2: "doublet", 3: "triplet", 4: "quartet", 5: "quintet"}
def spectrum(name, groups):
print(f"{name} at {FIELD_MHZ:.0f} MHz")
for label, delta, n_h, n_nb, J in groups:
lines = multiplet(delta, n_h, n_nb, J)
print(f" {label:22s} {delta:5.2f} ppm, {n_h}H, {NAMES[n_nb + 1]}" + (f", J = {J} Hz" if n_nb else ""))
for ppm, inten in sorted(lines, reverse=True):
print(f" {ppm:7.4f} ppm {'#' * round(8 * inten):24s} {inten:.3f}")
# Two isomers of C4H8O2. Shifts (ppm) are typical values in CDCl3.
spectrum("ethyl ethanoate, CH3-CO-O-CH2-CH3", [
("O-CH2 (next to CH3)", 4.12, 2, 3, 7.1),
("CO-CH3 (no H nbrs)", 2.05, 3, 0, 0.0),
("CH2-CH3", 1.26, 3, 2, 7.1)])
print()
spectrum("methyl propanoate, CH3-CH2-CO-O-CH3", [
("O-CH3 (no H nbrs)", 3.67, 3, 0, 0.0),
("CO-CH2 (next to CH3)", 2.32, 2, 3, 7.5),
("CH2-CH3", 1.14, 3, 2, 7.5)])
# The n + 1 rule needs the shift difference (in Hz) to be much larger than J
dnu = (4.12 - 1.26) * FIELD_MHZ
print(f"\nethyl group: shift difference {dnu:.0f} Hz vs J = 7.1 Hz, ratio {dnu / 7.1:.0f}")Both molecules show the same pattern — a 2H quartet, a 3H triplet and a 3H singlet — because both contain an ethyl group and an isolated methyl group. What tells them apart is where the signals fall: the CH₂ quartet at 4.12 ppm in ethyl ethanoate is attached to oxygen, while in methyl propanoate it is next to C=O, at 2.32 ppm, and the deshielded singlet at 3.67 is the O–CH₃.
Mass spectrometry and isotope patterns
A mass spectrometer turns molecules into ions, separates them by their mass-to-charge ratio and counts them. In electron ionisation a 70 eV electron knocks an electron out of the molecule, giving the molecular ion M⁺• (whose is the molecular mass), which often breaks into fragment ions that reveal pieces of the structure. The tallest peak is called the base peak; it is often a fragment, not the molecular ion.
Elements are mixtures of isotopes, so even a single compound gives a cluster of molecular-ion peaks. Carbon is 1.07% ¹³C; chlorine is about 76% ³⁵Cl and 24% ³⁷Cl; bromine is almost half ⁷⁹Br and half ⁸¹Br (CIAAW gives the official ranges). The pattern for a molecule is the product of the patterns of its atoms — mathematically, a convolution — which the block computes, together with the exact masses that high-resolution instruments measure.
import numpy as np
# Isotope abundances (amount fractions; representative values within the ranges given by IUPAC's CIAAW)
# as arrays indexed by extra nominal mass: index 0 = lightest isotope, 1 = +1 u, 2 = +2 u
ISOTOPES = {"Cl": np.array([0.758, 0.0, 0.242]), # 35Cl, 37Cl
"Br": np.array([0.507, 0.0, 0.493]), # 79Br, 81Br
"C": np.array([0.9893, 0.0107])} # 12C, 13C
def pattern(counts):
"""Relative intensities (largest = 100) of the M, M+1, M+2, ... peaks, by repeated convolution."""
p = np.array([1.0])
for element, n in counts.items():
for _ in range(n):
p = np.convolve(p, ISOTOPES[element]) # adding one atom = convolving its isotope pattern
return 100 * p / p.max()
for name, counts in [("one Cl", {"Cl": 1}), ("two Cl", {"Cl": 2}), ("three Cl", {"Cl": 3}),
("one Br", {"Br": 1}), ("two Br", {"Br": 2}), ("one Br + one Cl", {"Br": 1, "Cl": 1})]:
p = pattern(counts)
peaks = " ".join(f"{'M' if i == 0 else f'M+{i}'}: {v:5.1f}" for i, v in enumerate(p) if v > 0.05)
print(f"{name:16s} {peaks}")
# Carbon-13 makes an M+1 peak of about 1.1% per carbon atom
for n_c in [1, 6, 20, 60]:
p = pattern({"C": n_c})
print(f"C{n_c:<3d} M+1 / M = {p[1] / p[0] * 100:5.1f} %")
# High-resolution MS: four formulas with nominal mass 88 have different exact (monoisotopic) masses
mono = {"C": 12.0, "H": 1.0078250, "N": 14.0030740, "O": 15.9949146}
for formula in [{"C": 4, "H": 8, "O": 2}, {"C": 5, "H": 12, "O": 1}, {"C": 3, "H": 4, "O": 3}, {"C": 4, "H": 12, "N": 2}]:
m = sum(mono[e] * n for e, n in formula.items())
dbe = formula.get("C", 0) - formula["H"] / 2 + formula.get("N", 0) / 2 + 1 # double-bond equivalents
name = "".join(e + (str(n) if n > 1 else "") for e, n in formula.items())
print(f"{name:10s} monoisotopic mass {m:9.4f} double-bond equivalents {dbe:.0f}")These patterns are the fingerprints every organic chemist learns to spot: one chlorine gives M and M+2 in about 3:1; one bromine gives two nearly equal peaks; two bromines give 1:2:1. A compound with twenty carbons has an M+1 peak about a fifth the height of M, so the M+1/M ratio estimates the number of carbon atoms. And four different formulas all have nominal mass 88, but their exact masses differ in the second or third decimal place — easily resolved by modern instruments that measure masses to a few parts per million. The double-bond equivalents (rings plus π bonds) follow from the formula: , with halogens counted like hydrogen.
Identifying an unknown
Example— A colourless liquid, C₄H₈O₂
A sample gives these data:
- Mass spectrum. A molecular ion at 88, with high-resolution mass 88.0524. No M+2 peak beyond the small one expected from oxygen and carbon isotopes, so no Cl or Br.
- IR. A strong band at 1740 cm⁻¹; nothing broad between 2500 and 3600 cm⁻¹.
- ¹H NMR (400 MHz, CDCl₃). δ 4.12 (quartet, 2H, = 7.1 Hz), 2.05 (singlet, 3H), 1.26 (triplet, 3H, = 7.1 Hz).
Step 1: the formula. Of the formulas with nominal mass 88, only C₄H₈O₂ has exact mass 88.0524. Its DBE is : one ring or one double bond.
Step 2: the functional group. A strong band at 1740 cm⁻¹ is a C=O stretch at the frequency typical of esters, and it uses up the one DBE. The absence of a broad O–H band rules out the isomeric carboxylic acids (butanoic acid would show a very broad O–H from about 2500 to 3300 cm⁻¹ and a 1H signal near 11–12 ppm).
Step 3: the pieces. The NMR integrals add to 8, matching the formula's hydrogens. A 2H quartet with a 3H triplet sharing the same is an ethyl group, CH₂CH₃. The 3H singlet is a methyl group with no hydrogen neighbours.
Step 4: put them together. Two esters of this formula contain one ethyl and one isolated methyl: ethyl ethanoate (CH₃CO–O–CH₂CH₃) and methyl propanoate (CH₃CH₂CO–O–CH₃). The quartet at 4.12 ppm means the CH₂ is bonded to oxygen, and the singlet at 2.05 ppm is a methyl on a carbonyl. The compound is ethyl ethanoate (ethyl acetate), exactly as the simulation above predicted; methyl propanoate would have put its quartet near 2.3 ppm and its singlet near 3.7.
Real structure elucidation follows the same logic with more evidence: ¹³C NMR (one signal per kind of carbon), two-dimensional NMR experiments that show which protons couple to which and which carbons they sit on, fragmentation patterns, and comparison with reference spectra in databases such as SDBS and the NIST Chemistry WebBook.
The professional view
Modern instruments are faster and far more sensitive than the principles above suggest, mostly thanks to two ideas. Fourier-transform methods record all frequencies at once and separate them mathematically: FT-IR spectrometers use an interferometer, and pulsed FT-NMR, introduced by Richard Ernst and Weston Anderson in 1966 and extended to two dimensions by Ernst in Zürich (Nobel Prize 1991), excites all the protons with one radio-frequency pulse and Fourier-transforms the ringing signal (Fourier series are the mathematics). Soft ionisation — electrospray (John Fenn) and laser desorption (Koichi Tanaka), recognised in the 2002 Nobel Prize — lets mass spectrometers weigh intact proteins, and Orbitrap and Fourier-transform ion-cyclotron-resonance analysers measure masses to within a few parts per million. In the same prize, Kurt Wüthrich was honoured for solving protein structures by NMR.
In industry these techniques run all day: NMR and LC–MS confirm every intermediate in a pharmaceutical synthesis, GC–MS screens food, water and forensic samples, and inline IR and Raman probes watch reactions in real time. Superconducting NMR magnets operating above 1 GHz for protons now exist for structural biology, while compact benchtop instruments at 60–100 MHz have brought NMR into teaching laboratories and production plants. Machine-learning models that propose structures directly from spectra are an active research area, introduced in AI for chemistry; the analytical side — calibration, uncertainty and validated methods — is in Analytical chemistry, and spectroscopy of solids in spectroscopy and thermal analysis in Lumenary Materials.
Optional Christian reflection · The heavens declare
A reflection from the Christian tradition that inspires Lumenary Academy. It is not part of the mathematics and you are welcome to skip it.
In 1868 astronomers studying the Sun's spectrum during an eclipse found a yellow line that matched no known element. Norman Lockyer and Edward Frankland named the unknown element helium, after the Greek word for the Sun; William Ramsay found it on Earth only in 1895. That light from a star can tell us what the star is made of, because hydrogen and helium obey the same quantum rules there as here, is a remarkable fact about the universe. "The heavens declare the glory of God; and the firmament sheweth his handywork" (Psalm 19:1, King James Version). Many spectroscopists, of every belief and none, have felt that sense of order and wonder.
Common misconceptions
Misconception
"A blue solution absorbs blue light." It absorbs the complementary colour, orange-red, and transmits the blue that reaches your eye. Colour is what is left.
Misconception
"Absorbance is the percentage of light absorbed." Absorbance is logarithmic: means 90% absorbed, means 99% absorbed. Only absorbance, not percentage absorbed, is proportional to concentration.
Misconception
"The n + 1 rule counts all the hydrogens in the molecule." It counts equivalent hydrogens on neighbouring atoms (usually three bonds away). Protons on the same carbon, if equivalent, do not split each other, and protons on oxygen often exchange too fast to split anything.
Misconception
"The biggest peak in a mass spectrum is the molecule." The biggest peak is the base peak, often a stable fragment. The molecular ion can be weak or even absent with electron ionisation, which is one reason softer ionisation methods are used.
Exercises
1. Spectroscopy
Warm-upHumanA solution of a dye has absorbance at 520 nm in a 1.00 cm cuvette. The dye's molar absorption coefficient at that wavelength is L mol⁻¹ cm⁻¹. What is its concentration, in micromoles per litre?
2. Spectroscopy
CoreHumanThe C≡O bond in carbon monoxide has a force constant of N/m. Treating ¹²C¹⁶O as a harmonic oscillator with atomic masses 12.000 u and 15.995 u ( kg, cm/s), what is its vibrational wavenumber in cm⁻¹, rounded to the nearest 10?
3. Spectroscopy
Warm-upHumanNitrogen, N₂, makes up 78% of the air but has no infrared absorption spectrum. Why?
4. Spectroscopy
CoreHumanDichloromethane, CH₂Cl₂, is analysed by mass spectrometry. Using nominal masses (C = 12, H = 1, ³⁵Cl = 35, ³⁷Cl = 37) and ignoring ¹³C, list the values of all the peaks in its molecular-ion cluster.
5. Spectroscopy
CoreHumanA liquid with formula C₄H₈O₂ shows a strong IR band at 1740 cm⁻¹ and no broad O–H band. Its ¹H NMR spectrum has δ 3.67 (singlet, 3H), 2.32 (quartet, 2H) and 1.14 (triplet, 3H). What is it?
6. Spectroscopy
Warm-upHumanHow many double-bond equivalents (rings plus π bonds) does a compound with formula C₇H₆O have?
7. Spectroscopy
CoreComputationalWrite isotope_pattern(n_cl, n_br) returning a NumPy array of the relative intensities of the M, M+2, M+4, … peaks of a molecule containing n_cl chlorine and n_br bromine atoms (ignore all other isotopes), scaled so that the largest peak is 100. Use the abundances provided: ³⁵Cl 0.758, ³⁷Cl 0.242, ⁷⁹Br 0.507, ⁸¹Br 0.493.
Since only steps of 2 u occur, index of the array is the M+2 peak. For a molecule with no Cl or Br, return array([100.]).
8. Spectroscopy
StretchComputationalWrite multiplet(delta_ppm, n_neighbours, J_hz, field_mhz) that returns the first-order (n + 1 rule) lines of a proton signal as a list of (ppm, intensity) tuples, sorted by increasing ppm, with intensities from Pascal's triangle normalised to sum to 1. The lines are centred on delta_ppm and spaced by J_hz hertz; on a spectrometer whose proton frequency is field_mhz MHz, 1 ppm is field_mhz Hz.
The tests include a quartet at 400 MHz and the same coupling on an old 60 MHz instrument, where the multiplet spreads over a much wider range of ppm.
Further reading and resources
- Spectral Database for Organic Compounds (SDBS) · National Institute of Advanced Industrial Science and Technology (AIST), JapanFree IR, ¹H and ¹³C NMR, mass and Raman spectra of tens of thousands of compounds.
- NIST Chemistry WebBookIR, UV–visible and mass spectra, and vibrational constants of diatomic molecules.
- nmrshiftdb2An open database of NMR spectra and shift predictions.
- IUPAC Commission on Isotopic Abundances and Atomic Weights (CIAAW)The official isotopic abundances and atomic masses used for isotope patterns.
- The Nobel Prize in Chemistry 1991Richard Ernst, for Fourier-transform and multidimensional NMR.
- The Nobel Prize in Chemistry 2002John Fenn and Koichi Tanaka for mass spectrometry of large biomolecules, and Kurt Wüthrich for protein NMR.
- MIT OpenCourseWare 5.61: Physical Chemistry