A chemical bond pulls the two atoms back when they move apart or together, like a spring with force constant k. The atoms swing about their shared centre of mass, which stays still, so the lighter atom does most of the moving. An ideal spring vibrates at ν̃ = (1/2πc)√(k/μ), where μ = m₁m₂/(m₁+m₂) is the reduced mass. A stiffer bond (larger k) vibrates faster; heavier atoms (larger μ) vibrate slower. Go from HF to HI to see both at once: the bond gets weaker and the halogen heavier. Swap an isotope (tap an atom or use the buttons): the bond, and so k, stay the same, and only μ changes. A real bond is not a perfect spring: its energy levels crowd together as they rise, so the fundamental (0 → 1) is a little lower than the ideal-spring value and the overtone (0 → 2) is a little less than twice the fundamental. The real-bond numbers come from published spectroscopic constants (NIST), and every isotope value is predicted from them, not fitted: DCl comes out at 2091.0 cm⁻¹ (measured 2091.0), ¹³CO at 2096.0 (measured 2096.1). Motion is slowed about 7.5 × 10¹³ times and magnified 3×; every bond is drawn with the same vibrational energy, so a stiff bond swings less far.
Infrared light is an oscillating electric field. The atoms of a molecule carry small partial charges (δ+ and δ−, shown here as computed effective charges), so the field pushes the positive atoms one way and the negative atoms the other, and the arrows on the atoms show these pushes. If the pushes change the molecule’s dipole moment, they can drive a vibration. Like pushing a swing, this only builds up a large motion when the light’s frequency matches the vibration’s own frequency. Then the motion grows over a second or so, energy flows from the light into the molecule (see the meter), and less light reaches the detector. That is absorption. Off resonance the bonds only jiggle a little in time with the field. In the photon picture the same rule reads: a photon is absorbed when its energy hcν̃ equals the gap between two vibrational levels. Turn the source on, sweep the wavenumber (or press play to scan from 4000 to 400 cm⁻¹) and watch the spectrum build. Drag in the 3D view to turn the molecule: the light stays put, and only the part of the dipole change that lies along the field gets pushed. A bond at right angles to the field is not driven at all; “Line it up with the field” turns the molecule back. The spectrum below is for a gas, where molecules point every way, so it does not depend on how you turn this one. Try N₂ or the symmetric stretch of CO₂: the frequency can match perfectly, yet nothing is absorbed, because the dipole does not change. The light is drawn as a soft violet wave that passes behind the molecule; its wavelength is shrunk to fit the screen but stays proportional to 1/ν̃, so it visibly shortens as you raise the wavenumber, and it dims after the molecule when energy is absorbed. The arrow beside the molecule is the field at the molecule right now. “Field arrow only” hides the wave. “True scale” shows what really happens: the wavelength is thousands of times longer than the molecule, so the whole molecule feels one uniform field that flips back and forth. Simplifications: a classical driven, damped oscillator with a band width of 16 cm⁻¹; motion slowed about 7.5 × 10¹³ times and magnified; the detector reading is scaled for display; molecules really vibrate all the time (zero-point motion), which is not drawn.
A molecule with N atoms has 3N − 6 vibrations (3N − 5 if it is linear). In each normal mode all the atoms move at the same frequency, in step. The modes here come from a quantum-chemistry force field (B3LYP/6-31G*, computed for this lab), with each frequency corrected to the measured gas-phase fundamental of the normal molecule (NIST/Shimanouchi tables). Tap any atom to make it heavier (H ↔ D, C ↔ ¹³C, O ↔ ¹⁸O, N ↔ ¹⁵N): the force field stays the same and the modes are solved again for the new masses, with a small correction for anharmonicity. Checked against measured isotopologues (D₂O, ¹³CO₂, D₂CO, CH₃D, CH₂D₂, CHD₃, CD₄, C₂D₂), the predictions are mostly within 1% and at worst 1.5%; for deuterated methanol, where the CD₃ motions mix with the C–O stretch, errors reach about 3%. Only atoms that move in a mode change its frequency: ¹³C does not shift the symmetric stretch of CO₂. The intensity of a band depends on how much the dipole changes (|∂μ/∂Q|²). Band heights are computed, not measured: for formaldehyde, where measured values exist, the strong bands are within a factor of about 2 and the weakest is about 5× too weak, so read the heights as a guide. Bands are drawn 16 cm⁻¹ wide. Ethyl acetate has no complete published gas-phase assignment, so its band positions are computed and scaled with acetone’s factors for the same kind of motion, and its vibrations are named from their motion. Its three ester bands come out at 1738 (C=O), 1257 and 1057 cm⁻¹ (C–O), close to the measured liquid values (1735–1752, 1241 and 1047 cm⁻¹); gas-phase bands lie somewhat higher.
An FTIR does not scan the wavenumbers one by one. Light from a hot source (a globar near 1500 K) is split by a beamsplitter into two arms; one mirror is fixed and the other moves. When the beams recombine, each wavenumber interferes with itself according to the optical path difference δ (twice the mirror travel), so the detector records the interferogram I(δ), the sum of a cosine for every wavenumber. A Fourier transform turns it back into a spectrum. Here the interferogram is computed from the true spectrum and transformed exactly as an instrument does. The mirror travel sets the resolution: two lines Δν̃ apart are resolved once δmax ≥ 1/Δν̃. Press play and watch the spectrum sharpen as the mirror moves. Gas-phase HCl shows its rotational lines (P and R branches), each split by the ³⁵Cl/³⁷Cl isotopes about 2 cm⁻¹ apart: visible at 1 cm⁻¹ resolution, blurred at 4. Stopping the interferogram abruptly gives ripples next to sharp lines; apodization (Happ–Genzel) removes them at the cost of slightly wider lines. A HeNe laser (632.8 nm) times the sampling, one point per fringe, which fixes the wavenumber scale and sets the upper limit to 7901 cm⁻¹. Averaging more scans lowers the noise as √N; finer resolution makes the spectrum noisier for the same time. The sample spectrum is divided by a background (empty cell) to give transmittance. Simplifications: an ideal interferometer with no phase error (one side of the interferogram is used), lines broadened as in a cell filled to about 1 atm, a 10 cm cell at 296 K, and polyatomic bands drawn without their rotational structure.
Four kinds of question rotate: the band of a bond from k and the masses, the shift when H becomes D, whether a vibration absorbs IR, and naming a gas from its spectrum. Every number you need is in the Conditions box. A wrong answer is crossed out and you can try again; ★ counts questions right on the first try. Hint gives the method without the answer. Show the solution asks twice whether you are sure (and offers the hint first), then shows the method, the working with numbers, and the answer.
Diatomic constants: NIST Computational Chemistry Comparison and Benchmark Database (Irikura, J. Phys. Chem. Ref. Data 2007) and NBS J. Res. 64A (1960) for HBr. Fundamentals: Shimanouchi, Tables of Molecular Vibrational Frequencies (NSRDS-NBS 6 and 39), NIST Chemistry WebBook, CDMS; formaldehyde intensities: Nakanaga et al., J. Chem. Phys. 1982. FTIR principles: Skoog, Holler and Crouch, Principles of Instrumental Analysis.