One dye, one transition: how DCDHF-Me2 earns the two-level picture

1. The promissory note in the two-level model

The previous note on blinking and absorption modeled a fluorescent molecule as two electronic states, and was careful to say so twice: a real dye “does have higher electronic states and can show excited-state absorption,” and the two-level approximation “hides the higher excited states.” Those sentences are promissory notes, and this note pays them down for one specific, real molecule. We expected to find the hidden states crowding the band. What we found instead is the more interesting answer: for this dye the two-level idealization is not a convenient fiction that ignores nearby states — it is earned, and the reasons trace back to what the molecule was designed for.

The route: introduce the molecule and why it is the right example (§2), get its geometry onto defensible footing (§3), compute the singlet excitation manifold with time-dependent density functional theory (§4), set benzene — the dye’s own parent ring — beside it as the symmetry contrast (§5), build the three currencies a transition is quoted in — dipole moment, oscillator strength, molar absorptivity — from first principles (§6), and then read both manifolds against the two-level picture (§7).

2. The molecule: a push-pull dye built to be watched one at a time

DCDHF-Me2 is a donor–acceptor chromophore from the dicyanomethylenedihydrofuran (DCDHF) family developed for single-molecule fluorescence imaging: a dimethylamino donor conjugated through a phenyl ring to the DCDHF acceptor head, whose three nitrile groups and ring oxygen make it a strong electron sink.¹ The family was designed precisely for the experiment the previous note started from — single molecules blinking in a microscope — which makes it the natural molecule to ask the manifold question about. This site has computed on it before: the 2025 tooling notes used DCDHF-Me2 as their demonstration chromophore and shipped its Avogadro-built starting structure in their supporting information, which is the geometry this experiment inherits.

A push-pull dye is also where the two-level picture should work best: the lowest excitation is an intense charge-transfer transition from donor to acceptor — that is the design goal of the molecule. The push-pull chromophores note worked through that charge-transfer state on smaller analogs. The question here is how many neighbors that transition really has, and how much of the molecule’s absorption strength they carry.

3. Getting the geometry right: the force-field twist and the planar minimum

The inherited starting structure is a force-field object, not a quantum mechanical one: built in Avogadro and pre-optimized with the UFF force field, then exported explicitly for subsequent QM optimization — which is how the 2025 workflow used it.^2,3 How far that starting point sits from the DFT minimum turns out to be worth a section: UFF places the dimethylaniline ring 54.6° from coplanar with the acceptor plane. At the B3LYP/def2-SVP minimum the twist is 0.0° — a change of 54.6° — and the amine nitrogen, seeded slightly pyramidal, flattens into the ring plane entirely.

The chemistry of the planarization is the same push-pull story as the spectrum: the amine lone pair is being drawn into the π system by the acceptor on the far side of the ring, planar nitrogen maximizes that conjugation, and two methyl groups are not enough steric hindrance to resist it. The optimization step is not a formality for a dye like this — it is what restores the donor–acceptor conjugation the spectrum depends on.

Because an optimizer’s “converged” cannot distinguish a planar minimum from a planar saddle point — a trap this repository’s own tooling has hit before on aniline — we interrogated the stationary point along the two coordinates that could plausibly be unstable: rigid single-point displacements of the inter-ring twist (±10°, ±20°) and of the amine nitrogen out of its substituent plane (±0.15 Å), at the optimization level of theory. The energy rises for every displacement — by 0.44 and 1.52 kcal/mol along the twist and by 2.84 kcal/mol for the amine — and, because the optimized structure is exactly planar, each ± pair is mirror-equivalent and must agree by symmetry: the largest observed pair split is 0.51 microhartree, which validates the displacement construction itself. This rules out the two specific instabilities that motivated the check. It is not a frequency calculation, and no true minimum is claimed.

One caveat belongs beside the planarity result rather than buried in methods: B3LYP is known to favor planarizing amine donors, and the excitation spectra below — including CAM-B3LYP’s — are computed at the B3LYP geometry.

4. The computed manifold: one state owns the band

At the planar geometry we computed the 12 lowest singlet excitations with full-response TD-DFT (no Tamm–Dancoff approximation)⁴ at def2-TZVP, using CAM-B3LYP as the primary functional — range-separated hybrids are the standard guard against the charge-transfer failures of global hybrids^5,6 — with B3LYP⁷ alongside as the sensitivity check, in Psi4.⁸

The answer to the title question is: one. The lowest excitation sits at 374 nm (3.32 eV) and is a nearly pure HOMO→LUMO promotion carrying an oscillator strength of 1.124 — 74% of all the absorption strength in the computed window (Table 1). Of 12 computed states, 7 clear the f ≥ 0.01 brightness convention, but within ±0.35 eV of the lowest bright transition — the empirical band width this site’s earlier TD-DFT notes applied to their stick spectra — sits exactly 1 state: S₁ itself. The next state of any kind is 0.89 eV away and far weaker, and the manifold’s remaining bright states begin nearly another electron-volt beyond it: real, several of them genuinely bright, and all of them in the deep UV between 5.1 and 6.2 eV, where no visible-band measurement will conflate them with S₁ (Figure 1).

B3LYP tells the same story shifted red: lowest bright state at 413 nm, an S₁–S₂ gap of 0.51 eV, and still no second state within the band window. The 0.315 eV blue shift from B3LYP to CAM-B3LYP is itself diagnostic: it is the signature of substantial charge-transfer character in S₁ — the same signature the push-pull note measured at 0.42 eV for para-nitroaniline.

Table 1. The twelve computed singlet states of DCDHF-Me2 at CAM-B3LYP/def2-TZVP: one intense HOMO→LUMO transition, then nothing of comparable strength anywhere in the window. The generated per-state record in the experiment directory additionally carries hole–particle distances and an automated character label, omitted here for a reason given in §8.

5. The contrast: benzene’s band is a degenerate pair

Benzene earns its place in this note twice over: it is the simplest molecule whose strongly allowed band is not one transition, and it is literally the parent ring of DCDHF-Me2’s donor half — the same ring the push-pull note walked through the aniline → nitrobenzene → para-nitroaniline series. We recomputed it with the identical harness, functional, basis, band window, and software environment as the dye, because numbers imported from a different environment would not be a comparison.

The symmetry underneath the contrast is measured, not assumed: at the B3LYP/def2-SVP minimum all six carbon–carbon bonds agree to within 0.00001 Å, so the ring’s sixfold symmetry survives an optimization that ran in C1 and was free to break it. On that frame the manifold arranges itself in the opposite way to the dye’s (Table 2). The two lowest singlets, at 5.46 and 6.15 eV, are symmetry-forbidden and carry zero oscillator strength — the selection-rules note is about exactly this kind of extinction. The strongly allowed band at 7.12 eV is an exactly degenerate pair: two states separated by 0.000 eV, splitting the band’s strength down the middle — the lower partner carries 50% of the pair’s total. Under the same ±0.35 eV window that found one state for the dye, benzene holds 2 states, both bright (Figure 1). And the pattern repeats one rung up: the next pair, at 7.70 eV, is also exactly degenerate — dark, but degenerate — because degeneracy here is not numerical coincidence; it is what a two-dimensional irreducible representation of a sixfold-symmetric ring enforces.

Figure 1 shows the two manifolds in two measures, because a transition’s oscillator strength prices its dipole strength in units of its energy,

\[f = \tfrac{2}{3}\,\Delta E\,|\boldsymbol{\mu}|^{2},\]

with \(\Delta E\) in hartree and \(|\boldsymbol{\mu}|^{2}\) in atomic units — so the two panels rank the same sticks differently, and the difference is itself the physics; §6 builds it from first principles.

Figure 1. Both molecules’ computed states in two currencies on one energy axis at CAM-B3LYP/def2-TZVP, deliberately un-normalized in each panel. Top: oscillator strength — in dimensionless form, the integrated molar absorptivity. Bottom: dipole strength \(|\boldsymbol{\mu}|^{2}\). All twelve computed states of each molecule appear in both panels; dark states are markers on the axis, and no envelope is drawn because line widths are not computed in this experiment. Benzene’s exactly degenerate bright pair is drawn stacked at 7.12 eV, the lighter segment being the second member; A (top) and B (bottom) mark the same physical division seen in the two measures. The ranking inverts between panels: in oscillator strength benzene’s band total, 1.201, edges past the dye’s single transition at 1.124, while in dipole strength the dye’s 13.83 a.u. stands at twice benzene’s whole band (6.89 a.u.) — because oscillator strength prices dipole strength in units of transition energy, and benzene’s band sits at twice the energy.

The mechanism of the contrast fits in one sentence: benzene’s symmetry forces its bright transitions into a degenerate pair, and substituting a donor on one side of the ring and an acceptor on the other does two things at once — it destroys the symmetry that enforced the degeneracy, and it creates the low-lying charge-transfer state benzene does not have, which is where the funneled strength goes.

One thing this comparison is not: benzene’s allowed band sits at 7.1 eV — 174 nm, deep vacuum-ultraviolet — so nothing here compares visible colors. The comparison is about the composition of one apparent band. For benzene it is two transitions sharing one line; for the dye it is one transition wearing the whole band.

Table 2. The twelve computed singlet states of benzene under the identical protocol: forbidden states with exactly zero strength, and bright strength arriving only as a degenerate pair. Every dominant excitation is an almost exactly equal two-configuration mixture — the orbital-level fingerprint of a degenerate frame.

6. Three currencies for one transition: dipole moment, oscillator strength, absorptivity

Figure 1’s inversion confuses exactly the readers who know the most, so it is worth building the three quantities involved from the ground up and watching where the energy factor enters.

Start with what absorption is. Light is an oscillating electric field, and it can carry a molecule from \(|g\rangle\) to \(|e\rangle\) only to the extent that changing state moves charge. The transition dipole moment,

\[\boldsymbol{\mu}_{ge} = \langle\psi_e|\,e\,\hat{\mathbf{r}}\,|\psi_g\rangle,\]

is that motion’s measure: charge times displacement, evaluated between the two states rather than within either of them. It is not a dipole the molecule has; it is the dipole the molecule exercises in the act of switching states. The field couples to it the way a hand couples to a swing, and the rate at which resonant light drives the transition — the absorption cross section per molecule — goes as \(|\boldsymbol{\mu}_{ge}|^2\).⁹ Every note in this series has been reading this one quantity from a different side. For DCDHF-Me2 the S₁ excitation slides density from the amine end to the nitrile end of a long conjugated backbone — a lot of charge moved a long way — which is why its dipole strength, \(|\boldsymbol{\mu}|^{2}\) = 13.83 a.u., is about four times each benzene partner’s 3.44 a.u.: benzene’s excitation only rearranges π density within a ring a few bond lengths across.

Oscillator strength is an older yardstick with a deliberately chosen normalization. Before quantum mechanics, Lorentz modeled an absorbing electron as a charge on a spring, and one such classical electron has a definite total absorbing power. The oscillator strength of a transition asks: what fraction of one classical electron’s absorbing power does this transition deliver? In atomic units,

\[f = \tfrac{2}{3}\,\Delta E\,|\boldsymbol{\mu}|^{2},\]

and the normalization has a famous consequence: summed over every transition a molecule possesses, the f values add up to its number of electrons — the Thomas–Reiche–Kuhn sum rule.⁹ So f counts electrons’ worth of absorbing power, and the energy factor is what makes the count come out right: absorbing power is energy removed from the beam, each absorbed photon removes \(\Delta E\), and a transition operating at twice the photon energy extracts twice the energy per event from the same underlying charge motion. That is the entire resolution of Figure 1’s inversion. The dye has by far the larger charge displacement; benzene’s degenerate pair, working at 7.12 eV instead of 3.32 eV, is paid roughly twice per photon for about half the dipole strength — and the two products land within ten percent of each other (1.201 against 1.124).

Molar absorptivity is the bench quantity, defined by what a spectrophotometer measures through Beer–Lambert’s law, \(A = \varepsilon(\nu)\,c\,\ell\).¹⁰ It adds one ingredient the other two currencies lack: the line shape. The area under an absorption band, \(\int\varepsilon\,d\tilde{\nu}\), is fixed by the oscillator strength — f is that integral in dimensionless form, the same quantity the molar absorptivity note read as a rate constant. But the peak height a chemist quotes as \(\varepsilon_{\text{max}}\) depends on how vibronic structure and solvent spread that fixed area over a band width: the same area over a narrow band gives a tall peak, over a broad band a low one. This experiment computes no widths, so it can say nothing about peak heights — which is why \(\varepsilon\) appears in this section and on neither of Figure 1’s axes.

One reconciliation, for readers holding the entirely reasonable intuition that benzene is a feeble absorber: that intuition is about the visible and near-UV, where every low-lying benzene transition is symmetry-forbidden — the weak band a UV lamp shows near 254 nm is the forbidden S₁ borrowing intensity it does not own. The allowed E1u pair computed here is genuinely intense, but it sits at 174 nm, in the vacuum ultraviolet, where no cuvette measurement ever meets it. Benzene is a strong absorber; it is just not a strong absorber anywhere a chemist usually looks.

So the three currencies answer three different questions. \(|\boldsymbol{\mu}|^{2}\) answers how strongly does light couple to this transition — the molecule’s own property, the one the dye was engineered to maximize. f answers how much total absorbing power does that amount to, with the photon energy priced in. And \(\varepsilon(\nu)\) answers what will the instrument read at this wavelength, which further depends on a band width nothing in this note computes. Figure 1’s two panels are the first two currencies, and the exchange rate between them is the photon energy — that is all the inversion is.

7. Why the idealization is earned, not lucky

Set side by side under identical conditions, the two molecules answer the title question in opposite ways. Of the strength shared by its two lowest bright states, benzene’s lower partner carries 50%; DCDHF-Me2’s carries 98%. That number is the post in one line.

The currencies of §6 sharpen it further. In oscillator strength the dye’s S₁ carries 74% of its molecule’s computed total; in dipole strength — the energy weighting divided out — it carries 82%. The f-share understates how concentrated this molecule’s absorbing power actually is.

It is also not luck. Oscillator strength is a budgeted quantity, and an f of 1.124 in a single transition is a large fraction of what a chromophore this size can carry — a molecule engineered for single-molecule detection is a molecule engineered to be bright, and bright means concentrating the available transition strength into the one state the laser will drive. The DCDHF designers optimized brightness and photostability;¹ an isolated, dominant S₁ is a consequence of those goals, not their stated aim, but the consequence is what the manifold shows. Benzene, with no donor, no acceptor, and a symmetry that forbids favoritism among equivalent directions, spreads the same budget across a degenerate pair — and one substitution pattern later, on the same ring, the degeneracy is gone and the funnel exists.

What the two-level model hides is therefore real but relocated. The dye’s higher states — several genuinely bright — sit between 4.5 and 6.2 eV, which is exactly where the previous note’s excluded channels operate: excited-state absorption out of S₁ terminates in this manifold, and the states its nonlinear section reached with two photons live here as well. Nothing in the visible band needs them; everything beyond the two-level picture starts with them.

So the blinking note’s idealization, applied to the dye class it was written about, is not an approximation forced on an unwilling molecule. A fluorophore engineered for single-molecule brightness behaves like a two-level system because it was selected to; the approximation and the design are the same fact read twice.

8. Reproducibility

Both molecules ran through the same two-stage pipeline (research/dcdhf-me2-transitions/, driven by run_all.sh): geometry optimization at B3LYP/def2-SVP (C1, gau_tight), then full-response TD-DFT for the 12 lowest singlets at def2-TZVP with CAM-B3LYP and B3LYP, in Psi4 1.9.1 (conda env psi4_19, Python 3.10.17, NumPy 2.2.5), on 6 threads with a 6 GB memory target, on x86_64 Linux. The DCDHF-Me2 starting geometry is the Avogadro/UFF structure published in the 2025 supporting-information note; benzene was built and optimized under the same protocol and, unlike the dye, was recomputed rather than inherited from this site’s earlier push-pull work, because those numbers came from a software environment that no longer exists on this machine. One solver fact is worth recording: Psi4 1.9.1’s TD-DFT eigensolver runs at a fixed ceiling of 60 iterations — both the driver keyword and the global option that appear to control it are silently ignored — and an unconverged root raises an exception rather than passing silently; every state reported here converged at the requested residual tolerance of 10⁻⁵, and the experiment’s results record the requested and effective solver settings side by side. The rigid-displacement stationary check of §3, its symmetry-pair self-validation, and the full per-state records, spectra, metrics, and environment files are in the experiment directory, and every quoted number above resolves from its generated metrics.json. One reproduction datum: rerunning the stationary check after the canonical run, to capture its environment record, reproduced every reported value to the displayed precision.

9. Where the model stops

These are vertical electronic excitations of isolated molecules in vacuum, twelve states deep. Nothing here has vibrational structure: no line widths are computed anywhere in this experiment, which is why Figure 1 draws bare sticks, and a real absorption band’s width is dominated by vibronic progressions and solvent broadening this calculation does not attempt. Solvent shifts of charge-transfer states are large and absent here — both functionals sit blue of the dye’s experimental solution band, as vertical gas-phase numbers should be expected to. We can say nothing about states above the computed window, benzene’s Rydberg states are poorly served by a basis without diffuse functions, and the geometry underneath the dye’s spectrum is a stationary point interrogated along two suspect coordinates, not a frequency-confirmed minimum. Finally, one diagnostic deserves its own disclaimer: the automated charge-transfer classifier in our harness labels states by the distance between hole and particle centroids, a metric that understates charge transfer when both frontier orbitals delocalize over the same conjugated backbone — for S₁ of the dye we trust the functional-shift signature instead, and the classifier column stays in the experiment directory rather than in Table 1.

References