One matrix element, two experiments: molar absorptivity and the Pockels effect
There is a habit, nearly universal in optical-materials talks, of plotting every absorption spectrum normalized to 1. It makes the slides tidy: a dozen compounds, all their bands scaled to the same peak height, the eye free to compare where they absorb without the clutter of how much. It is also, quietly, the deletion of the single most physically loaded number on the plot. The height of an absorption band — the molar absorptivity you scaled away — is a direct readout of the transition dipole moment, and that same transition dipole sets the refractive index, the polarizability, and, for a push–pull chromophore, the electro-optic coefficient that makes it useful in a modulator. Normalize to 1 and you have thrown away the axis that predicts whether the material does anything.
The previous post made half of this case: molar absorptivity \(\varepsilon\) is a rate constant in disguise, because the integrated band is proportional to \(|\boldsymbol{\mu}_{ge}|^2\) and that fixes the Einstein coefficients — a kinetic quantity. This post makes the other half. The very same \(|\boldsymbol{\mu}_{ge}|^2\) runs straight into the optical response — linear and nonlinear — so that Beer’s law and the Pockels effect turn out to be two measurements of one matrix element. It is the bridge from this series’ spectroscopy posts to its hyperpolarizability posts, and it is why a spectroscopist’s instinct — high molar absorptivity tends to go with a strong electro-optic response — is not a coincidence but a theorem.
1. The one number every optical spectrum is really measuring
The transition dipole moment between the ground state \(|g\rangle\) and an excited state \(|e\rangle\) is the matrix element
\[\boldsymbol{\mu}_{ge} = \langle e | \hat{\boldsymbol{\mu}} | g \rangle, \qquad \hat{\boldsymbol{\mu}} = -e\sum_i \mathbf{r}_i,\]
the same object that appears in Fermi’s golden rule and in the configuration- interaction and response calculations of the excited-state post. Every oscillator strength, every absorption band, is a measurement of its magnitude. Recall the packaging from last time: the integrated band gives the oscillator strength, and the oscillator strength gives \(|\boldsymbol{\mu}_{ge}|^2\),
\[f = 4.32\times10^{-9}\!\int_{\text{band}}\!\varepsilon(\tilde\nu)\,d\tilde\nu \;\propto\; \tilde\nu_{ge}\,|\boldsymbol{\mu}_{ge}|^2 .\]
That is the entire content of an absorption intensity: height (and width) of the band \(\to |\boldsymbol{\mu}_{ge}|^2\). When a talk normalizes the band to unit height, it is discarding exactly this — keeping \(\tilde\nu_{ge}\) (where the band sits) and erasing \(|\boldsymbol{\mu}_{ge}|^2\) (how much dipole the transition carries). We can put a number on it and carry that number all the way to a modulator. Take a representative donor–acceptor chromophore of the nitroaniline / azo-dye family: a charge-transfer band at \(\lambda_{\max}=480\) nm (\(\tilde\nu_{ge}\approx20{,}800\ \mathrm{cm^{-1}}\), \(E_{ge}=2.58\) eV), peak molar absorptivity \(\varepsilon_{\max}\approx4\times10^{4}\ \mathrm{L\,mol^{-1}\,cm^{-1}}\), band width \(\approx5000\ \mathrm{cm^{-1}}\). Integrating gives \(f\approx0.92\) and
\[|\boldsymbol{\mu}_{ge}| \approx 9.7\ \text{D}.\]
Hold onto that 9.7 D. It is the number the normalized plot would have hidden, and it is about to reappear in three different experiments.
2. The linear response is the absorption band, continued
Before the nonlinearity, the linear optical response already runs on \(|\boldsymbol{\mu}_{ge}|^2\). Second-order perturbation theory gives the frequency- dependent polarizability as a sum over excited states,
\[\alpha(\omega) = \frac{2}{\hbar}\sum_{e} \frac{\tilde\omega_{ge}\,|\boldsymbol{\mu}_{ge}|^2}{\tilde\omega_{ge}^2-\omega^2},\]
and near an isolated, dominant transition this is one term: \(\alpha(\omega)\propto|\boldsymbol{\mu}_{ge}|^2/(\tilde\omega_{ge}^2-\omega^2)\). The macroscopic refractive index \(n(\omega)\) is built from \(\alpha(\omega)\), so the dispersion of \(n\) — how the glass bends blue light more than red — is set by the same transition dipoles that set the absorption. This is not an analogy; it is Kramers–Kronig. Absorption is the imaginary part of the response and refraction is the real part, and the two are Hilbert transforms of one another: a single complex function \(\chi(\omega)\) whose strength is \(|\boldsymbol{\mu}_{ge}|^2\). The refractive-index spectrum is the absorption spectrum continued into the transparent region. You cannot scale one to 1 without scaling the other. Already at linear order, “how much” is not a cosmetic axis — it is the whole amplitude of the response.
3. The nonlinear response: the two-level model
Now push the field harder. Expand the induced molecular dipole in powers of the local field,
\[\mu_{\text{ind}} = \mu_0 + \alpha E + \tfrac{1}{2}\beta E^2 + \tfrac{1}{6}\gamma E^3 + \cdots,\]
and the first hyperpolarizability \(\beta\) is the leading nonlinear term — the one responsible for second-harmonic generation and the linear electro-optic (Pockels) effect. Like \(\alpha\), it has an exact sum-over-states form, a double sum over excited states weighted by products of transition dipoles. For the donor–acceptor chromophores that matter in electro-optics, one charge-transfer state dominates so completely that the sum collapses to a single term — the two-level model of Oudar and Chemla.1,2 Its static value is
\[\beta_0 \;=\; \frac{3\,\Delta\mu_{ge}\,|\boldsymbol{\mu}_{ge}|^2}{E_{ge}^{2}},\]
where \(\Delta\mu_{ge} = \mu_{ee}-\mu_{gg}\) is the change in permanent dipole between ground and excited states (Figure 1). Look at what governs it: the transition dipole squared \(|\boldsymbol{\mu}_{ge}|^2\) — the absorption intensity — in the numerator, and the transition energy squared \(E_{ge}^2\) — where the band sits — in the denominator. A hyperpolarizability is an absorption band read through a different lens: strong (\(|\boldsymbol{\mu}_{ge}|^2\) large) and red (\(E_{ge}\) small) is what makes \(\beta_0\) big.
Figure 1. The two-level model. A single charge-transfer excitation carries a transition dipole \(\boldsymbol{\mu}_{ge}\) (what absorption measures) and a change in permanent dipole \(\Delta\mu_{ge}\) (the donor–acceptor push–pull). The static first hyperpolarizability is built from both, with the same \(|\boldsymbol{\mu}_{ge}|^2\) that sets the molar absorptivity sitting in the numerator.
Put the numbers in. With \(|\boldsymbol{\mu}_{ge}|=9.7\) D from the absorption band, a typical push–pull dipole change \(\Delta\mu_{ge}\approx6\) D, and \(E_{ge}=2.58\) eV,
\[\beta_0 = \frac{3\,(6\,\text{D})(9.7\,\text{D})^2}{(2.58\,\text{eV})^2} \approx 1.0\times10^{-28}\ \text{esu} = 99\times10^{-30}\ \text{esu},\]
squarely in the range of a good organic chromophore. Every factor that went into that number except \(\Delta\mu_{ge}\) came straight off the absorption spectrum — and the dominant, squared factor is precisely the peak height a normalized plot erases.
4. From \(\beta\) to the Pockels effect
A single molecule’s \(\beta\) becomes a device property through two more steps, both of which preserve the proportionality. First, orientation: a centrosymmetric arrangement has no net \(\beta\) (the second-order response cancels by symmetry), so the chromophores are poled — aligned in a field while the polymer is warm, then frozen in place. The macroscopic second-order susceptibility is the molecular \(\beta\) times the number density and an orientational average,
\[\chi^{(2)} \propto N\,\langle\cos^3\theta\rangle\,\beta ,\]
and the linear electro-optic coefficient — the Pockels coefficient \(r\) that tells you how much the refractive index shifts per volt — is
\[r \;\propto\; \frac{\chi^{(2)}}{n^4} \;\propto\; \frac{N\,\langle\cos^3\theta\rangle\,\beta}{n^4}.\]
This is the number that matters for a modulator: apply a voltage, change \(n\) via \(r\), change the optical path length, and you have written an electrical signal onto a beam of light — the heart of the electro-optic modulators in every long-haul fiber link.3 And it descends, factor by factor, from \(\beta\), hence from \(|\boldsymbol{\mu}_{ge}|^2\), hence from the absorption intensity. The spectroscopist’s instinct is now a derivation: a molecule with a strong, low-lying charge-transfer band (\(\varepsilon_{\max}\) large, \(\lambda_{\max}\) red) has a large \(|\boldsymbol{\mu}_{ge}|^2\) and a small \(E_{ge}\), so it has a large \(\beta_0\), so — once poled — it makes a good electro-optic material. The normalized-to-1 plot in the seminar had, in every one of its curves, silently discarded the quantity that ranks the compounds for exactly the application they were presumably being screened for.
At an operating wavelength the static \(\beta_0\) is enhanced by dispersion as the photon energy climbs toward the transition. The two-level Pockels dispersion has a resonance at the one-photon energy — i.e. at \(\lambda_{\max}\) itself:
Figure 2. Two-level dispersion of the Pockels hyperpolarizability \(\beta(-\omega;\omega,0)\) for the \(\lambda_{\max}=480\) nm chromophore. In the transparent telecom window the enhancement over the static value is modest (about \(1.2\)–\(1.3\times\)), but it grows without bound as the operating wavelength approaches \(\lambda_{\max}\) — where the molecule also absorbs. The knob that raises \(\beta\) is the knob that destroys transparency.
5. The nonlinearity–transparency trade-off
Figure 2 is the whole design problem in one curve. Everything that makes \(\beta\) large pulls the material toward its own absorption band. You can raise \(\beta_0\) by lowering \(E_{ge}\) — red-shifting \(\lambda_{\max}\) — but the Pockels resonance sits at \(\lambda_{\max}\), so a redder chromophore brings its absorption (and optical loss) closer to your operating wavelength. You can raise \(\beta_0\) by increasing \(|\boldsymbol{\mu}_{ge}|^2\) — a stronger band — but that is, by definition, a larger \(\varepsilon\) and more absorption. And you can borrow resonant enhancement by operating nearer the band, at the cost of the transparency the device needs. This is the nonlinearity–transparency trade-off, and it is not an engineering nuisance layered on top of the physics; it is the physics. One matrix element, \(|\boldsymbol{\mu}_{ge}|^2\), and one energy, \(E_{ge}\), set both the thing you want (a big electro-optic coefficient) and the thing you must avoid (absorption at the operating wavelength). Chromophore design is the art of playing that single matrix element against itself.4,5
6. Back to the seminar
So the offhand question — why not show any of the kinetics of absorption for these materials? — was not a non-sequitur. It was pointing at the axis the normalization erased. The peak molar absorptivity of each of those bands was, simultaneously:
- a rate constant, fixing the Einstein coefficients and the radiative lifetime (the previous post);
- the amplitude of the linear response, and hence the refractive-index dispersion, through Kramers–Kronig;
- the dominant factor in the first hyperpolarizability, and hence in the poled-film electro-optic coefficient — the figure of merit for the modulator the compounds were likely being screened to become.
All three are \(|\boldsymbol{\mu}_{ge}|^2\), read out in different experiments. To normalize every spectrum to unit height is to keep the one thing that is easy to see — the color, \(\lambda_{\max}\) — and discard the one thing that is hard-won and decisive: how much dipole the transition actually carries. The spectra looked comparable because they had been made comparable, scaled to hide the very quantity that distinguished them. A transition dipole is not a detail of an absorption band. It is the band’s whole reason for mattering — in the kinetics, in the refraction, and in the volt-to-light conversion at the end of the line.