Bond Length Alternation
in Near-Infrared Cyanine Dyes

Cyanine dyes are a family of polymethine chromophores whose absorption maxima shift by approximately 100 nm per additional vinyl unit. Two nitrogen-bearing heterocyclic rings are bridged by an odd-numbered chain of alternating single and double bonds; in the canonical cationic form the positive charge delocalises symmetrically across both nitrogens. Cy7, Cy9, and Cy11 sit squarely in the biologically important near-infrared window of 650-1100 nm used for fluorescence imaging and photoacoustic spectroscopy.

The key microscopic quantity is the Bond Length Alternation (BLA); the difference between the mean single-bond length and the mean double-bond length in the polymethine bridge:

In the cyanine ideal (perfect π-delocalisation, coined by Dähne & Hoffmann 1977): BLA = 0, all C-C bonds ≈ 1.39-1.40 Å. In the polyene limit (carotenoids, conjugated polymers): BLA ≈ 0.08-0.12 Å. Real near-IR cyanines occupy an intermediate regime. Larger BLA → smaller electronic coupling → redder absorption via the two-state valence-bond model.

Why CAM-B3LYP and not B3LYP?

Standard GGAs (PBE, BLYP) over-delocalise the π density → artificially small BLA. B3LYP improves this but shows residual over-delocalisation. CAM-B3LYP incorporates Coulomb-attenuating long-range exact exchange with parameters μ = 0.33 bohr⁻¹, α = 0.19, β = 0.46; substantially improved BLA for polymethines (Jacquemin et al. 2007). For Cy11⁰ (neutral doublet, 243 electrons) we used B3LYP/UKS instead: CAM-B3LYP combined with UKS can introduce significant spin contamination for extended conjugated radicals, and B3LYP/UKS is the standard choice for π-radical neutrals in the cyanine/polyene literature.

All calculations ran on a Windows 10 workstation: serial execution, 1 CPU core, 4 GB RAM per job, total wall time 14 h 29 min. ORCA 6.1.1 removed the LOOSOPT keyword; we set convergence thresholds manually via the %geom block. Grid4 and NoFinalGrid keywords are also invalid in 6.1.1. The RI-J warning ("RI is on but no J-basis assigned; assigning Def2/J") is benign.

Input; Cations (Cy7⁺, Cy9⁺, Cy11⁺)

Input; Neutral Radical (Cy11⁰)

Molecular Structures

These are the exact |ΔE| (mH) values from every geometry step, directly from the ORCA output files. The dashed threshold at 0.050 mH = 5×10⁻⁵ Eₕ is TolE. Click any molecule button to isolate it. Note the anomalous spike in Cy9⁺ at cycle 8: the trust-radius algorithm temporarily accepted an uphill step (+10.33 mH), then self-corrected in cycle 9. Cy11⁰ required 31 cycles; the UKS wavefunction showed HOMO-LUMO gap <0.10 Eₕ in 16/31 cycles (near-degeneracy of SOMO and virtual orbitals).

Figure 3 (publication). |ΔE from final| (mH) on a log scale vs. geometry cycle. Dashed line: TolE threshold (5×10⁻⁵ Eₕ = 0.050 mH). Note the transient energy increase in Cy9⁺ at cycle 8 (trust-radius, self-corrected) and the slow final convergence of Cy11⁰ (31 cycles, UKS doublet).

Final Energies and Convergence Summary

All 122 C-C bond lengths from the optimised geometries, sorted ascending. Colour-coding: molecule-colour bars = short conjugated bonds (d < 1.420 Å, formal C=C); red bars = long conjugated bonds (1.420-1.500 Å, formal C-C); grey bars = sp³ C-C bonds (>1.500 Å, excluded from BLA). The bimodal gap near 1.420 Å validates the cutoff choice. BLA cutoff sensitivity: ±0.010 Å change in cutoff → <2% change in BLA.

Figure 4. Complete C-C bond length distributions (sorted). Bimodal gap near 1.420 Å validates the BLA cutoff. All 4 molecules shown.

Figure 7. All 122 C-C bonds across four molecules displayed sequentially. Horizontal lines at 1.420 and 1.500 Å.

BLA Analysis; Principal Results

Figure 1 (publication). BLA overview. (a) BLA values for all four molecules; dashed horizontal lines = integer multiples of δ_PC = 0.013643; thick dashed = mean BLA = 0.05019 Å. (b) Mean short C=C (hatched) and mean long C-C (solid) bond lengths with BLA arrows. (c) BLA in units of δ_PC; shaded band = 3.5δ-4.0δ.

The orbital energies reveal the microscopic origin of the 100 nm/vinyl rule. LUMO energies across the cation series: −4.184, −4.183, −4.156 eV (σ = 0.013 eV; essentially constant). HOMO energies change by σ = 0.198 eV. The ratio σ(LUMO)/σ(HOMO) = 0.065. The chain-length red-shift arises almost entirely from HOMO destabilisation: the LUMO has a node at the meso position and adding vinyl units at the ends does not change its energy.

ᵃ Cy11⁰ UKS α-spin orbitals: HOMO is the SOMO. Gap is α-SOMO to α-LUMO splitting; not directly comparable to closed-shell gaps.

Figure 2 (publication). Kohn-Sham orbital energies. (a) HOMO/LUMO levels for all four molecules with gap annotated. (b) HOMO, LUMO, and gap as a function of chain length n for the cation series, with linear regression overlay. Note the near-constant LUMO (σ = 0.013 eV) versus the rising HOMO.

Dipole Moments and C-N Bond Lengths

Figure 6. C-N bonds by type and Cy11 neutral vs cation structural comparison. Note 29 mÅ iminium elongation. BLA values scaled ×10 for display.

Figure 8. Computational statistics: dipole moments, wall-clock times (Cy11⁰ dominates at 7.4 h), geometry cycles vs. number of atoms.

The Pythagorean Comma arises from the fundamental incommensurability of powers of 3 and powers of 2 (they have distinct prime factorisations). Twelve perfect fifths overshoot seven octaves by exactly δ. The comma is irreducible; it cannot close. BLA in real cyanines is also a residual gap: the competition between Peierls distortion and π-delocalisation can approach zero but never reaches it. Both measure how far a nearly-ideal system is from its ideal.

Figure 5. Pythagorean Comma analysis. (a) BLA vs. chain length for cation series and Cy11⁰ with Jacquemin 2007 literature estimates. (b) BLA/δ for all four molecules; mean = 3.678δ; shaded ±1σ band.

Figure 9. BLA in structural context (cyanine → polyene axis). Background shading: blue = cyanine, yellow = intermediate, red = polyene. Gold dashed grid lines at nδ intervals. All four measured values fall between 3δ and 4δ.

From the three cation measurements (n=7,9,11) we fit the linear model: BLA(n) = 8.47×10⁻⁴·n + 0.04127 Å with dBLA/dn = 8.47×10⁻⁴ Å/carbon. This is used to predict Cy1, Cy3, Cy5, Cy13, and Cy15 at the same level of theory. Predicted KS gap: from Cy7⁺ (3.550 eV) to Cy11⁺ (3.094 eV), slope = −0.114 eV/carbon. Predicted λ_max = 1240/E_gap; the KS gap systematically underestimates the optical gap by ~20-30% (TD-DFT needed for quantitative values). Cy11⁰ (neutral radical) is shown separately as it uses B3LYP/UKS and is not part of the linear fit.

Full Predicted Series

Interactive Polymethine Chain; Animate BLA across the Series

Select any molecule to see the animated bond-alternation pattern. Bond thickness encodes bond order (thicker = shorter). The BLA progress bar shows position relative to cyanine ideal (0) and polyene limit (0.12 Å). Green buttons = ORCA-measured; white = model prediction.

The BLA observation in cyanine dyes is not isolated. It is one projection of a broader geometric object: the Spiral of Fifths. The same irrational number δ = (3/2)¹²/2⁷ − 1 = 0.013643 appears as a universal commensurability deficit in four independent domains. These are the figures from the companion paper: The Pythagorean Comma as Universal Commensurability Deficit: Spiral of Fifths, Kepler's Harmonice Mundi, Cyanine Dye Electronic Structure, and the Comma as a Quantum Berry Phase.

Domain 1; Music: The Perfect Fifth Landscape

The error landscape of fifth ratios (1.45-1.56) shows the deep structure of near-closure. The pure 3/2 = 1.500 (Pythagorean) gives n=12 closure with error δ. The special ratio 2^(3/5) = 1.5157 gives 5-step closure. The convergent ladder shows how larger n→ smaller closure error: n=12 (−1.955¢), n=41 (+0.48¢), n=53 (−0.069¢), n=306 (+0.006¢), n=665 (−0.0001¢). Sara's special fifth 1.5201749468 appears in the comparison plot; its identity and significance remain an open question.

Perfect Fifth Landscape; Pythagorean, Mercator & Kepler. Top: error landscape (min closure error over n=1..700) across fifth ratios 1.45-1.56. Middle: zoom near 3/2 = 1.5 (Pythagorean territory, blue) and near 2^(3/5) = 1.5157 (5-step closure, orange). Bottom left: convergent ladder showing EDO fifth accuracy for n=5,12,41,53,306,665. Bottom right: heat map of closure error for each (n, fifth) pair; the cyan dots mark exceptional near-closures.

Domain 2; Orbital Mechanics: Octave Closure Patterns

The same closure patterns appear in orbital resonances. The ultra-zoom around 2^(3/5) = 1.5157165665 reveals near-closures at n=303→2¹⁸², n=421→2²⁵³, n=718→2⁴³¹ etc. The circle of fifths plot shows that the 5-EDO (yellow) traces a perfect pentagram; exactly 5 steps closing; while the 12-EDO (cyan) traces the familiar twelve-note circle. The closure error comparison at bottom right shows Sara's fifth (1.5201749468) achieves exceptionally low closure error at specific n values; it appears as a white line with distinctive null patterns distinct from all standard systems.

Deep Dive: Octave Closure Patterns Around Key Fifths. Top left: ultra-zoom around 2^(3/5) showing near-closure pairs labeled with their (n, 2^k) identities. Top right: orbit of (2^3/5)^n reduced to [1,2]; stars mark near-closures, showing the pseudo-periodic structure. Bottom left: circle of fifths; pure 3/2 (pink), 12-EDO (cyan), 5-EDO (yellow). Bottom right: closure error vs n for five key fifths including Sara's special fifth 1.5201749468 (white); note the distinct deep nulls at n≈720.

Domain 3; Molecular: Vibrational Ladder Model

The vibrational ladder model extends the cyanine analysis beyond ground-state BLA to the full bond-pair sum y = 2.67ⁿ + 2.535 Å. The floor at 2.535 Å = 2 × C=N (two iminium C=N bonds) sets the geometric minimum; the exponential growth tracks the bond pair length as chain extends. The end correction solves where 2.67ⁿ + 2.535 = b_opt = 6.882 Å: at n = 1.496. The predicted λ_max vs experiment shows 146× improvement from the comma correction. The recurrence 2.67ⁿ connects directly to the Peierls gap equation E_gap = 4t·δ.

Sara's Vibrational Ladder Model: C=C-C Resonance + PIB. Top left: recurrence y = 2.67ⁿ + 2.535, floor = 2×C=N = 2.535 Å, n₁ = 5.205 Å. Top right: positive (▲) and negative (▽) transitions by group I-IV and substate i-iv, with 2.535 floor line. Bottom left: predicted λ_max vs experiment; standard model (blue, 250k: 560 nm offset), comma-corrected (green), experimental (☆). Bottom right: end correction; where does 2.67ⁿ + 2.535 = b_opt? At n = 1.496, y = 6.882 Å.

Domain 4; Quantum Mechanics: The Comma Berry Phase

A spin-1/2 wavefunction completing a comma-revolution of φ·360° = 364.912° accumulates a residual Berry phase of exp(iπ(φ−1)) per revolution that never cancels. The interference signal I(n) = cos²(nπ(φ−1)/2) predicts a 6.4% reduction at n=12 (measurable with modern neutron interferometers) and 82% reduction at n=53 (Mercator cycle). Standard QM predicts I = 1 for all n. The cyanine BLA is the physical manifestation of this accumulated Berry phase in a molecular system.

PIB Comma Correction; RMSE: 107 nm → 0.73 nm

The particle-in-a-box model for cyanine absorption uses box length L = a_opt·k + b_opt, where k is the chain length index. Standard PIB: a = 2.4872 Å (Simpson 1948), b = 5.6 Å → RMSE = 107 nm. Comma-corrected: a_opt = 2.5146 ≈ 2.5·φ^(1/2) Å, b_opt = 6.882 ≈ 5.6·φ^15 Å → RMSE = 0.73 nm. The comma correction improves prediction by 146×. The experimental points are at k=0 (Cy3, ~300 nm), k=1 (Cy5, ~420 nm), k=2 (Cy7, ~630 nm).

Guido of Arezzo (c. 1030 AD) mapped the hexachord syllables (Ut, Re, Mi, Fa, Sol, La) onto the joints of the left hand, giving singers a physical mnemonic for the spiral of fifths. Each joint corresponds to a pitch; the hand traces the circle of fifths, but when you reach the 12th fifth, you overshoot the starting octave by exactly φ−1 = 0.013643. The mutation at Fa (the fourth degree) is the hand's reset: a topological necessity, not a pedagogical convenience. Below: the hand animated, each joint lighting up as the fifths stack; watch the overshoot appear at step 12.

The π-electrons of a cyanine dye traverse the conjugated chain and accumulate a small Berry phase per pass. The bond alternation δ is the lattice's geometric response: the chain dimerizes to partially compensate for the phase mismatch. Below: select any of the four measured molecules (or the predictive series Cy1-Cy15). Toggle between representations: skeletal (bond-stick), electron density (wavefunction overlay), BLA map (colour-coded by bond length), and PIB box (particle-in-a-box comparison). The Kairos countdown shows how far the molecule is from crossing the absorption wavelength threshold.

CAM-B3LYP/6-31G* TD-DFT single-point vertical excitation calculations completed for Cy7⁺, Cy9⁺, and Cy11⁺ on 22 March 2026 using ORCA 6.1.1. All three terminated normally. Wall time for Cy7⁺: 18 min 26 sec (SCF 12.5 min, CIS module 5.8 min).

Each vinyl unit added to the polymethine bridge is one step around the spiral of fifths. BLA increases by ≈ 8.5 × 10⁻⁴ Å per carbon, never reaching zero - for the same reason the spiral never closes. The comma is the gap preventing closure in both systems. At the cyanine limit (Cy11⁺: gap = 0.068 eV), the bridge becomes electronically vulnerable. The molecule modulates rather than decomposes: it transforms into a new stable species via a predictable chemical pathway.