Bond Lengths from Rotational Spectra

Point a microwave beam at a sample of gaseous carbon monoxide. You see absorption lines spaced with almost metronomic regularity. That regularity is not a curiosity. It is a direct readout of a bond length accurate to fractions of a picometer. Rotational spectroscopy turns a measured frequency spacing into $I$ , the moment of inertia. From there it yields $r$ , the internuclear distance. This happens without ever needing a structure model built from guesswork. This matters because bond lengths anchor almost everything else in molecular science. Force constants, thermochemical cycles, dipole moment interpretations all depend on them. So do the starting geometries for larger polyatomic structure determinations. All these numbers trace back to values first pinned down this way. Diatomics are the cleanest laboratory for the method. A single number, $r$ , fixes the entire geometry. This article builds the rigid-rotor result you already know into a working measurement tool. You will see how spectral line spacing gives $B$ . Then you will see how $B$ gives $I$ . Finally you will see how isotopic substitution separates $r$ from any lingering ambiguity. You will also see where the rigid-rotor approximation starts to break down.

Key takeaways

A rigid-rotor rotational spectrum consists of nearly equally spaced lines with spacing $2B$ in wavenumber units.
The rotational constant $B$ is inversely proportional to the moment of inertia $I = \mu r^2$ .
Measuring $B$ to spectroscopic precision (megahertz-level) gives $r$ to sub-picometer precision.
Isotopic substitution changes $\mu$ but not $r$ , so comparing isotopologues checks the rigid-rotor picture and isolates $r$ cleanly.
Centrifugal distortion and vibrational averaging are small but real corrections that limit ultimate accuracy.
Key idea: the spacing between adjacent rotational absorption lines is a direct, model-independent measurement of a molecule's moment of inertia, and hence its bond length.

The core insight

Because a rigid diatomic's rotational energy levels are quantized as $E_J = BJ(J+1)$ with $B \propto 1/I$ , the uniform $2B$ spacing seen in a rotational absorption spectrum is a fingerprint of $I = \mu r^2$ ; inverting that one relation, with the reduced mass known from atomic masses, delivers the bond length directly from frequency measurements alone.

The physical picture

Recall the rigid rotor: two point masses $m_1$ and $m_2$ held at a fixed separation $r$ by a rigid bond, free to tumble in space. Its allowed energies depend on a single geometric quantity, the moment of inertia $I = \mu r^2$ , where $\mu = m_1 m_2/(m_1+m_2)$ is the reduced mass. Everything in this article follows from asking: given that a real molecule absorbs microwave photons at a specific set of frequencies, what does that tell us about $I$ ? A polar diatomic (a permanent dipole moment is required for the transition to be allowed) absorbs radiation that drives it from rotational level $J$ to $J+1$ . The selection rule $\Delta J = +1$ combined with the quadratic level spacing $E_J = BJ(J+1)$ produces a spectrum of lines at $2B, 4B, 6B, \ldots$ — a ladder with constant rung spacing $2B$ . That constancy is the experimental signature you look for first: if the spacing is uniform, the rigid-rotor model applies, and the single fitted parameter $B$ contains the entire structural information. The logic of the method runs in one direction only: measure frequencies, extract $B$ , compute $I$ , then compute $r$ . No structural assumption enters except that the molecule behaves, to good approximation, as two masses on a rigid stick. Because microwave transition frequencies can be measured to parts in $10^8$ or better, this crude-sounding model yields bond lengths that rival diffraction methods for precision, while requiring only a gas sample and a tunable microwave source.

Figure

Rotational energy ladder and the resulting absorption spectrum

The core relations

E_J = h c B\, J(J+1), \qquad J = 0, 1, 2, \ldots

where: $B$ is the rotational constant in wavenumbers (cm⁻¹), $J$ the rotational quantum number.

B = \frac{h}{8\pi^2 c I}

where: $I$ is the moment of inertia in kg·m², $c$ the speed of light, $h$ Planck's constant.

\tilde\nu_{J\to J+1} = 2B(J+1)

where: successive absorption lines occur at $2B, 4B, 6B, \ldots$ , so the constant spacing is $2B$ .

I = \mu r^2, \qquad \mu = \frac{m_1 m_2}{m_1+m_2}

where: $\mu$ is the reduced mass, $r$ the equilibrium (or effective) bond length.

r = \sqrt{\frac{h}{8\pi^2 c \mu B}}

where: this is the working formula — measure $B$ spectroscopically, know $\mu$ from atomic masses, solve for $r$ .

Derivation AFrom line spacing to bond length

Goal. Starting from the rigid-rotor energy levels and the rotational selection rule, derive the formula that converts a measured line spacing into an internuclear distance.

Step 1 — Energy levels of the rigid rotor. Quantum mechanics of a rigid rotor (two masses $m_1, m_2$ separated by fixed $r$ ) gives quantized angular momentum with eigenvalues $J(J+1)\hbar^2$ , so the rotational energy is

E_J = \frac{\hbar^2}{2I}J(J+1), \qquad I = \mu r^2 .

Converting to the spectroscopist's wavenumber unit ( $\tilde E = E/hc$ ) defines the rotational constant,

B \equiv \frac{h}{8\pi^2 c I}, \qquad E_J/hc = B J(J+1).

Step 2 — Apply the selection rule. A polar rotor absorbing electric-dipole radiation obeys $\Delta J = +1$ . The photon wavenumber for the transition $J \to J+1$ is the energy difference divided by $hc$ :

\tilde\nu_{J\to J+1} = B(J+1)(J+2) - BJ(J+1) = B\big[(J+1)(J+2) - J(J+1)\big].

Expanding the bracket,

(J+1)(J+2) - J(J+1) = (J+1)\big[(J+2)-J\big] = 2(J+1),

\tilde\nu_{J\to J+1} = 2B(J+1).

Step 3 — Read off the spacing. Adjacent lines correspond to $J$ and $J+1$ as lower states, so the difference between consecutive transition wavenumbers is

\tilde\nu_{J+1\to J+2} - \tilde\nu_{J\to J+1} = 2B(J+2) - 2B(J+1) = 2B.

Every rung in the ladder is separated by exactly $2B$ (in the rigid-rotor limit), which is why a real spectrum lets you read $B$ directly off a ruler laid across the line positions, then refine it with a least-squares fit across many lines.

Step 4 — Invert for $I$ , then for $r$ . From the definition of $B$ in Step 1,

I = \frac{h}{8\pi^2 c B}.

Using $I = \mu r^2$ and solving for $r$ ,

r^2 = \frac{I}{\mu} = \frac{h}{8\pi^2 c \mu B},

so that

\boxed{r = \sqrt{\dfrac{h}{8\pi^2 c \mu B}}}

with $\mu$ obtained independently from atomic masses (via the periodic table) and $B$ obtained from the measured line spacing. No assumption about $r$ itself enters anywhere in this chain — the bond length is a pure output.

Worked examples

Example 1 — Carbon monoxide, ¹²C¹⁶O (foundational). The rotational absorption spectrum of gaseous CO shows lines close to 3.845 cm⁻¹ apart. Because the spacing is $2B$ ,

B = \frac{3.845\ \text{cm}^{-1}}{2} = 1.9225\ \text{cm}^{-1}.

Convert $B$ to SI (moment-of-inertia) form. Using $I = h/(8\pi^2 cB)$ with $B$ expressed in m⁻¹ ( $1.9225\ \text{cm}^{-1} = 192.25\ \text{m}^{-1}$ ):

I = \frac{6.626\times10^{-34}\ \text{J s}}{8\pi^2 (2.998\times10^{8}\ \text{m s}^{-1})(192.25\ \text{m}^{-1})}

Numerator: $6.626\times10^{-34}$ . Denominator: $8\pi^2 = 78.96$ ; then $78.96 \times 2.998\times10^{8} = 2.368\times10^{10}$ ; then $\times 192.25 = 4.553\times10^{12}$ .

I = \frac{6.626\times10^{-34}}{4.553\times10^{12}} = 1.455\times10^{-46}\ \text{kg m}^2.

Now find the reduced mass. Atomic masses: $m_C = 12.000\ \text{u}$ , $m_O = 15.995\ \text{u}$ (¹⁶O). In kilograms, using $1\ \text{u} = 1.6605\times10^{-27}\ \text{kg}$ :

\mu = \frac{(12.000)(15.995)}{12.000+15.995}\ \text{u} = \frac{191.94}{27.995}\ \text{u} = 6.856\ \text{u}

\mu = 6.856 \times 1.6605\times10^{-27}\ \text{kg} = 1.1385\times10^{-26}\ \text{kg}.

Finally,

r = \sqrt{\frac{I}{\mu}} = \sqrt{\frac{1.455\times10^{-46}}{1.1385\times10^{-26}}} = \sqrt{1.278\times10^{-20}}\ \text{m} = 1.130\times10^{-10}\ \text{m}.

So $r_{CO} \approx 113.0$ pm. This matches the accepted CO bond length (112.8 pm) to within 0.2 pm — well inside the precision expected once vibrational averaging (Step-4 caveat below) is taken into account.

Example 2 — Using isotopic substitution to separate mass from geometry (applied). Suppose you measure $B(^{12}\text{C}^{16}\text{O}) = 1.9225\ \text{cm}^{-1}$ (as above) and, in a second sample enriched with the heavier isotope, $B(^{13}\text{C}^{16}\text{O}) = 1.8397\ \text{cm}^{-1}$ . Isotopic substitution changes the mass distribution but — under the Born–Oppenheimer approximation — leaves the equilibrium bond length essentially unchanged, since it doesn't alter the electronic potential energy surface. This gives an internal consistency check. First compute $\mu$ for ¹³C¹⁶O: $m_{^{13}C} = 13.003$ u, $m_{^{16}O} = 15.995$ u.

\mu' = \frac{(13.003)(15.995)}{13.003+15.995}\ \text{u} = \frac{207.98}{28.998}\ \text{u} = 7.172\ \text{u} = 1.1909\times10^{-26}\ \text{kg}.

From $B'=1.8397\ \text{cm}^{-1} = 183.97\ \text{m}^{-1}$ :

I' = \frac{6.626\times10^{-34}}{78.96 \times 2.998\times10^{8}\times 183.97} = \frac{6.626\times10^{-34}}{4.359\times10^{12}} = 1.520\times10^{-46}\ \text{kg m}^2.

Then

r' = \sqrt{\frac{1.520\times10^{-46}}{1.1909\times10^{-26}}} = \sqrt{1.277\times10^{-20}} = 1.130\times10^{-10}\ \text{m} = 113.0\ \text{pm}.

The two isotopologues give the same $r$ to three significant figures (113.0 pm both times), even though $B$ differs by about 4.3% and $\mu$ differs by about 4.6%. This agreement is exactly the physical check that validates the rigid-rotor extraction: a real change in $\mu$ with an essentially fixed $r$ predicts $B'/B \approx \mu/\mu'$ , and indeed $1.8397/1.9225 = 0.9569$ compares well with $\mu/\mu' = 1.1385/1.1909 = 0.9560$ . The small residual difference (about 0.1%) is the fingerprint of the zero-point vibrational amplitude differing slightly between isotopologues — the first hint of the vibration–rotation coupling treated in the next stage of the theory.

Limiting cases & common traps

Situation	What actually happens	Consequence for extracted $r$
Rigid-rotor limit ( $J$ small, low resolution)	Lines equally spaced by $2B$	Simple formula above is accurate to ~0.1–1%
High $J$ , real molecule	Centrifugal distortion stretches the bond; levels shift by $-DJ^2(J+1)^2$	Spacing shrinks slightly at high $J$ ; fit $\tilde\nu = 2B(J+1) - 4D(J+1)^3$ instead
Vibrational averaging	Spectroscopic $B$ is really $B_v$ for a specific vibrational state, dominated by $v=0$	Extracted $r$ is $r_0$ (vibrationally averaged), slightly longer than the equilibrium $r_e$
Nonpolar diatomics (H₂, N₂, O₂, Cl₂)	No permanent dipole moment	No pure rotational (microwave) absorption spectrum; must use Raman rotational spectroscopy instead
Isotopic substitution	$\mu$ changes, $r_e$ essentially unchanged (Born–Oppenheimer)	Used to cross-check assignments and separate mass effects from bond-length effects

The chemist behind it

Rotational spectroscopy as a structural tool matured through the work of physicists and chemists who built microwave sources during and after the Second World War, adapting radar technology to spectroscopy. Among the pioneers who systematically turned this into quantitative structural chemistry was Charles Townes, whose group demonstrated that microwave absorption lines of gases could be measured with extraordinary precision, laying groundwork that connected directly to maser and laser development.

Origins · 1 minCharles Hard Townes and how this was discovered

Charles Hard Townes

1915–2015 · Pioneered microwave spectroscopy of gases, showing rotational transitions could yield bond lengths and dipole moments with unprecedented precision, work that also led toward the maser

Ask the tutor: How does centrifugal distortion let us extract a force constant alongside the bond length, and how does this connect to vibration–rotation coupling in real spectra?

Why does a molecule need a permanent dipole moment to show a pure rotational spectrum?

Absorption or emission of a photon in a rotational transition requires the molecule's rotating charge distribution to couple to the oscillating electric field of the radiation. That coupling is governed by the transition dipole moment, which is nonzero only if the molecule has a permanent dipole moment that changes orientation as it rotates. Homonuclear diatomics like N₂ or O₂ have zero permanent dipole moment by symmetry, so pure rotational transitions are electric-dipole forbidden; their rotational structure is instead studied through Raman spectroscopy, which relies on polarizability changes rather than dipole moments.

How precise is a bond length obtained this way compared to X-ray or electron diffraction?

Microwave-derived bond lengths for small gas-phase molecules are typically accurate to a few tenths of a picometer or better, because frequency measurements can achieve parts-per-billion precision. X-ray diffraction on solids is excellent for heavier atoms and larger structures but is less sensitive to light atoms like hydrogen and is subject to crystal-packing effects. Gas-phase microwave spectroscopy, free of intermolecular perturbations, is often the reference method for the most accurate small-molecule equilibrium and ground-state structures.

What happens to the line spacing if the bond isn't perfectly rigid?

Real bonds stretch slightly under the centrifugal force of rotation, an effect called centrifugal distortion. This adds a small negative correction term, $-4D(J+1)^3$ , to the transition wavenumula formula, so the line spacing decreases slightly as $J$ increases rather than staying perfectly constant. The distortion constant $D$ is typically four to six orders of magnitude smaller than $B$ , so it matters only at high resolution or high $J$ , and fitting both $B$ and $D$ from the full spectrum gives a more accurate structure.

Can this method be extended beyond diatomics?

Yes, though it becomes more involved. Linear and symmetric-top polyatomic molecules still have simple rotational spectra characterized by one or two rotational constants, and isotopic substitution at different atomic positions (Kraitchman analysis) allows chemists to locate each atom's coordinates one at a time. Asymmetric tops require three rotational constants and more complex fitting, but the same underlying principle holds: line positions encode moments of inertia, and moments of inertia encode geometry.