Hydroxyaromatic Fluorophores.

Many fluorophores that are widely used in analytical biochemistry and in biological microscopy contain a hydroxyaromatic component. One could also find fascinating chemistries of hydroxyaromatic dyes, especially those capable of excited state proton transfer (ESPT) to produce dual emission, in the literature of materials and physical chemistry. The ESPT-capable compounds have attracted interest based on their fundamental intellectual values in molecular photophysics and their potential utilities as light emitters in organic light-emitting diodes (LEDs) or fluorescent sensors. The hydroxyaromatic dyes could undergo either intra- or intermolecular proton transfer in either electronic ground or excited states. Although having long been applied for various purposes, some of their absorption and emission properties have not always been clearly described because of the insufficient attention given to proton transfer equilibria in either the ground or excited state and the challenges in computationally modeling the true emitters of these dyes under any given conditions. In this article, an attempt is made to summarize the spectroscopic properties of a few common hydroxyaromatic dyes that have been studied for both fundamental and practical purposes, with the help from quantum chemical calculations of the absorption and emission energies of these dyes in neutral and anion forms. The goal of this article is to provide readers some clarity in the optical properties of these compounds and the tools to understand and to predict the photon-initiated behaviors of hydroxyaromatic fluorophores.

Introduction

Many fluorescent molecules contain a hydroxyaromatic moiety. Prominent examples include fluorescein, which is a reliable choice of fluorescent labels for staining biological specimens;[1] 1- and 2-naphthol, which are archetypal photoacids[2] that have inspired the development of many more for achieving temporal and spatial control of proton-driven processes;[3] and the light-emitting core of green fluorescent protein (GFP), which is a genetically encoded illuminator of subcellular targets and processes.[4] The neutral and anion forms of these dyes differ in their emission wavelengths and quantum yields, with the anion often as the dominant emitter that characterizes the fluorescence. In this article, the differences and commonalities of the proton-transfer-dependent spectroscopic properties of a few hydroxyaromatic dyes are summarized. Quantum chemical calculations are used to understand how proton transfer alters the electronic structures in both ground and emissive excited states. The objective of this article is to help readers understand how the emission color and brightness of these compounds change upon deprotonation, which we hope would prompt the discovery of more utilities of these photophysically intriguing compounds with beautiful emission colors.

Fluorescein and Green Fluorescent Protein (GFP): The Preamble

The pKa of the monoanion form of n class="Chemical">fluorescein (Scheme a) is ∼6.4.[1] Therefore, in biological fluids, blood serum for example, with pH = 7.4, ∼90% of fluorescein is in the phenolate dianion form that has an emission quantum yield of 0.93.[1] The dianion affords a relatively narrow absorption band centered at 490 nm, while the monoanion exhibits two bands of similar molar absorptivities at shorter wavelengths (430–480 nm).[5]

Scheme 1

Two Prototropic Forms of Fluorescein (a) and the Chromophore of GFP (b)

The emission of fluorescein (∼515 nm) is attributed to those of the dianion and, to a lesser extent, the monoanion forms. Other prototropic species (i.e., neutral/lactone, cation) of fluorescein have been reported,[5a] the descriptions of which are not included in this article because they do not involve the deprotonation of the hydroxyaromatic moiety. The overlaps between the emission spectra of the monoanion and the dianion, as well as the overlaps of their absorption spectra,[6] have led to the challenges in the characterization of the spectroscopic properties of either species. The absorption and emission spectra of the neutral and anion forms have been shown separately,[6] from which the difference in profile of the spectra is evident. However, that information is lost in λmax listings, such as the data in Table .

Table 1

Reported Absorption and Emission Maxima (λabs and λem), Emission Quantum Yield (φ), and Acidity Constants of Ground and Excited States (pKa and pKa*) of the 10 Dyes (N: Neutral; A: Anion) Described in This Articlea

entry	name	λabs (N)b	λabs (A)b	λem (N)c	λem (A)c	φ (N)	φ (A)	pKa	pKa*
1	fluorescein	453/472	490	515	515	0.37	0.93	6.4	6.3
1’	xanthene	508 (EtOH)	508 (EtOH)	517 (EtOH)	517 (EtOH)	0.76 (EtOH)	0.95 (EtOH)	6.0	-
2	wtGFP	395	475	460d	508	-	0.77	4.5/8.1	<1.0
3	7HC	324	366	397 (MeOH)	453	0.27 (MeOH)	0.76e	7.8	0.4
4	2-naphthol	329	350	360	425	0.16	0.36	9.5	2.8
5	pyranine	400	450	418 (EtOH)	510	-	0.82	7.2	0.4
6	HBO	333 (DMSO)	400 (DMSO)	370/480f (DMSO/DCM)	450 (DMSO)	0.02 (DCM)	0.55 (DMSO)	-	-
7	bipyVHBO	350 (DMSO)	457 (DMSO)	445/550f (DMSO/DCM)	585 (DMSO)	0.32 (DCM)	0.78 (DMSO)	-	-
8	HP-TZ1	299 (ACN)	335 (ACN)	449 (DCM)	434 (ACN)	0.16 (DCM)	0.31 (ACN)	-	-
9	HP-TZ2	318 (DMSO)	353 (DMSO)	409 (DCM)	524 (DMSO)	0.30 (DCM)	0.14 (DMSO)	-	-
10	HP-TZ3	293 (DMSO)	361 (DMSO)	-	435 (DMSO)	∼0	0.16 (DMSO)	-	-

Organic solvents in which these measurements were made are noted in parentheses and are listed in Tables S2–S6, in which the calculated excited energies were modeled in the same solvents. If not noted in parentheses, the data were collected in water or aqueous buffers.

The wavelength of maximal absorption of the band with lowest energy and a substantial molar absorptivity (not necessarily the largest).

The wavelength of the maximal intensity of the emission band.

Emission only observed in ultrafast time-resolved spectroscopy.

Implied.

The emission maxima of normal/tautomer bands.

Because the emission band maxima via exciting either a mono- or dianion coincide at ∼515 nm, one would consider the possibility that the monoanion may have dissociated to the dianion in the excited state. In other words, given the documented propensity of hydroxyarenes to eject a proton to a willing acceptor in the excited state, fluorescein could be a “photoacid”,[7] of which the acidity is higher in the excited state than in the ground state. Although experiments have so far discounted this hypothesis,[5a] it has been shown that the excited state proton transfer (ESPT) equilibrium between the monoanion and the dianion could be catalyzed by applying a buffer (e.g., phosphate over a threshold concentration) with proton-shuffling capacities.[8] Therefore, the ESPT reaction of fluorescein could materialize in a supportive medium.[6,8] However, fluorescein cannot be considered as a photoacid as conventionally understood because the acidity values of both ground (pKa) and excited (pKa*) states are close.[8]

Green fluorescent protein (GFP) is grouped with fluorescein in this opening section not only because GFP is arguably the most well-known naturally formed dye with a hydroxyarylated emitter but also due to the fact that the discovery was driven in part by a desire to find a more convenient and more versatile fluorescent label for biological imaging that carries the spectroscopic characteristics of the already widely used fluorescein.[4,9] The fluorophore of GFP (Scheme b) is formed via the oxidative cyclization of three amino acid residues, Ser65, Tyr66, and Gly67, and is protected inside a β-barrel (or β-can) fold. The wildtype (wt) GFP shows two excitation bands that were characterized to be the neutral (395 nm) and anion (475 nm) forms. The absorption maxima (λmax) of the neutral and the anion species change to different values, 384 and 448 nm, upon denaturation,[4] thus providing a context of the “accuracy” of the λmax values of the chromophore and revealing the dependence of absorption or emission on the microenvironment in which the chromophore is confined. The pKa of the denatured wtGFP is 8.1. The fully folded and therefore fluorescent wtGFP can be “quenched by acidic pH values with an apparent pKa near 4.5”.[4] The discrepancy between these two numbers might be attributed to the difference in acid–base equilibria that were recorded in the two separate measurements—the latter appears to be impacted by the acid–base equilibrium in the excited state where the acidity of GFP is increased upon photoexcitation (infra vide).

The emission (508 nm) of the GFP anion is observed when either of the absorption bands of GFP is excited. The emission from the excited neutral species at a shorter wavelength (450–460 nm) is only observed before ESPT to water occurs within a few picoseconds of excitation to produce the excited anion, as revealed in ultrafast time-resolved emission studies.[10] Therefore, the neutral GFP chromophore is a bona fide photoacid (pKa* < 1.0). It is understood that many hydroxyaromatic dyes that are excited under physiological (or similar) conditions undergo excited state deprotonation to produce emissive anionic fluorophores. Although most of the practitioners of these dyes are aware of this extra chemical event prior to emission, it has not been made clear in many of the publications. Therefore, it is worthwhile to clarify this chemistry in the present article.

Based on the observations of fluorescein and GFP, the following generalizations are drawn (Figure ): the neutral forms of the fluorophores are usually found to absorb (#1) and to emit (#2) at shorter wavelengths than the anions. The anions are brighter emitters than the neutral precursors (#3), and the neutral form could ionize either in the ground state or as a photoacid in the excited state (#4, #5).

Figure 1

Five conclusions of hydroxyaromatic fluorophores summarized from the photophysical properties of fluorescein and GFP. N: neutral; A: anion; GS: ground state; XS: excited state.

Five conclusions of hydroxyaromatic fluorophores summarized from the photophysical properties of n class="Chemical">fluorescein and GFP. N: neutral; A: anion; GS: ground state; XS: excited state.

In the following sections, the absorption and emission properties of ten hydroxyaromatic dyes (Table ) are described, with the help of quantum chemical calculations to understand the observations summarized in Figure , as well as the exceptions (#6 in Figure ) to these conclusions. Some challenges that we faced in the calculations while trying to balance the numerical accuracy and the computing currency are described in the Supporting Information. The calculated excitation energies of neutral and anion forms of the dyes at either ground state (S0) geometries (comparable with absorption) or excited state (S1) geometries (comparable with emission) are tabulated in Tables S2/3 and S4/5. The calculated excitation energy values are plotted against the experimental absorption and emission maxima in Figure .

Figure 2

Calculated excitation energies of (a) neutral forms at optimized S0 geometries (i.e., absorption); (b) anion forms at optimized S0 geometries (i.e., absorption); (c) neutral forms at optimized S1 geometries (i.e., emission); and (d) anion forms at optimized S1 geometries (i.e., emission). The range of y-axes is fixed at 1.5–4.5 eV for easy comparison of relative magnitudes of excitation energies. Gray bars represent experimental absorption or emission maxima, while gold and garnet bars are excitation energies calculated using def2-SVPD and aug-cc-pVDZ basis sets, respectively. Red arrows point to three instances where calculations deviate from experimental data by large margins.

The accuracies of the calculations, as judged by the deviations in eV from the experimentally observed absorption or emission maxima, are “reasonable” (quotation marks are used to emphasize the lack of precision of this word) in most cases and will be more specifically described in the Discussion section. The data in Figure show (1) the calculations correctly reflected the relative magnitude of excitation energies among these dyes, and (2) the use of the aug-cc-pVDZ basis set over the def2-SVPD basis set produced only slightly better mean absolute deviations of the excitation energy calculations (see the values in Tables S2–5). Regardless of the deviations from the experimental data, these calculations captured the changes (decrease or increase) of absorption or emission energies upon deprotonation and offered some explanations from the lens of frontier molecular orbitals (e.g., highest occupied molecular orbital (HOMO) and lowest unoccupied molecular orbital (LUMO)). In short, in most cases, the HOMO energy level is elevated to a larger extent than the LUMO energy upon deprotonation, which leads to a smaller HOMO–LUMO gap and therefore a lower excitation energy. Such changes of the first seven compounds calculated based on their ground state optimized geometries, which we have the most confidence in because we were able to use the same method and other computational parameters for the calculations of the neutral and anion pair in question, are tabulated in Table S6. The limited space of this Mini-Review is therefore reserved for describing results that are specific to each compound in the individual sections and how they could help explain experimental observations of absorption and emission of these dyes, as well as the limitations of these methods of computation in the Discussion section.

Fluorescein and Hydroxyxanthene

The hydroxyxanthene core[11] and the C9-carboxyphenyl substituent of fluorescein (Scheme ) are not conjugated with each other. The charge on the carboxyphenyl moiety, and the phenyl group itself, created complications in the calculations (e.g., difficulty in reaching convergence; unreasonable frontier molecular orbital (FMO) arrangements; etc.) of the excitation energies of the ground (absorption) and excited (emission) states. In fact, any fluorophore that allows the rotation of an internal single bond may create additional challenges in the optimization of excited state structures and the calculations of excitation energies, especially those of the anions.

Scheme 2

Structure of Hydroxyxanthene Which Is the Emitting Component of Fluorescein

The emitter of fluorescein is hydroxyxanthene (Scheme ), for which the emission and absorption spectra and those of its conjugate base are considered “virtually identical to those of fluorescein”.[11] It was also reported that deprotonation did not alter the absorption and emission maxima in the mixed solvent ethanol/dichloromethane (1:1) but only increased both the molar absorptivity at the λmax and the emission quantum yield from 0.75 to 0.95.[12]

The excitation energy calculated at the relaxed S0 geometry of the conjugate base of hydroxyxanthene matches well with the experimental absorption peak (Table S2), while the calculated excitation energy of the neutral form (422 nm) is much higher (+0.5 eV, deviations listed in Table S2) than its reported absorption maximum (508 nm), which coincides with the absorption maximum of the anion. The coincidence of neutral and anion absorption maxima of hydroxyxanthene was reported by a single paper[12] and is an outlier of the trend that deprotonation decreases the absorption energy of a hydroxyaromatic dye, including fluorescein, for which the neutral form has a shorter absorption maximum than the anion form. Therefore, for the purpose of reaching a higher accuracy in the calculation of the excitation energy of the ground state of neutral hydroxyxanthene, in addition to the refinement of the theoretical method, a reexamination of the absorption properties of hydroxyxanthene (Scheme ) perhaps is justified.

The lowest absorption bands of both neutral and anion forms are principally contributed from the HOMO → LUMO (H to L) transitions. Deprotonation raises the HOMO energy more so than the LUMO level (Table S6). Therefore, the increase of wavelength maximum of the lowest absorption band upon deprotonation is attributed to the preferential elevation of the HOMO relative to the LUMO. This interpretation of the absorption energy decrease upon deprotonation can be applied to other hydroxyaromatic dyes if (1) both HOMO and LUMO include the hydroxyaromatic component and (2) the lowest absorption band is dominated by the H to L transition.

The H to L transitions constitute also the majorities of the excitations at the S1 geometries (Figure ) of both neutral and anion forms of hydroxyxanthene. Between them, the percentage of the H to L contribution to the anion excited state (92%) is higher than that of the neutral (86%) form. This higher level of dominance of the H to L transition of the anion emissive state than its neutral counterpart holds true for most examples in this article and is correlated to the higher emission quantum yield of the anion than the neutral. The FMOs of the excited neutral form do not offer a hint of charge transfer from the hydroxy to the carbonyl end, consistent with the experimental conclusion that fluorescein is not a photoacid,[5b] in contrast to other hydroxyarenes that are discussed in later sections.

Figure 3

HOMO and LUMO of the neutral (n, left) and anion (a, right) forms of xanthene at their relaxed S1 geometries after single-point calculation. H to L transitions constitute the majority of the S1 excitations (i.e., emission). The contributing percentages (squared configuration interaction coefficients) are shown after the comma.

HOMO and LUMO of the neutral (n, left) and n class="Chemical">anion (a, right) forms of xanthene at their relaxed S1 geometries after single-point calculation. H to L transitions constitute the majority of the S1 excitations (i.e., emission). The contributing percentages (squared configuration interaction coefficients) are shown after the comma.

p-Hydroxybenzylidene Imidazolidinone (p-HBDI), the Chromophore of wtGFP

The chromophore of wtGFP is the post-translationally synthesized p-HBDI (Scheme ), which only exhibits the emission properties of wtGFP when constricted in the correctly folded protein environment. Due to the tremendous interests in various aspects of wtGFP, computation has been conducted to understand the photophysical properties of the chromophore. A couple of papers are listed here for the interested readers,[13] in which extensive computing powers were invested to accurately model the excitation energies of the chromophore with specific solvent effects.

Scheme 3

Chromophore in a cis-Planar Conformation As Seen in the Crystal Structures of wtGFP

The current work aims to find a compromise between the level of accuracy of excitation energy prediction (judged by the deviation from the absorption or emission band maximum) and the computational currency. The initial structure of p-HBDI for optimization has a cis-planar conformation as observed in the crystal structures of wtGFP. The lowest excitation energy of neutral and anion forms at the ground state geometries was found at 365 and 474 nm, respectively (Table S2). Both excitations were primarily contributed from the H to L transitions of the allowed π → π* nature, indicated by the high oscillator strengths (Table S2). These results are consistent with the experimentally observed large molar absorptivity values of both species and therefore are reassuring that the calculation reflects the physical reality in that aspect. As is the case with hydroxyxanthene, the larger increase of the HOMO energy than that of the LUMO accounts for the decrease of excitation energy upon deprotonation (Table S6). This conclusion can be drawn for the next few compounds as listed in Table S6.

The H to L transitions account for most of the excitation to the S1 of both forms at the relaxed excited state (S1) geometries (Figure ). Based on the difference in occupancies between HOMO and LUMO of p-HBDI (Figure ), a charge transfer transition is expected to result in a higher acidity of the phenolic OH upon exciting to the S1 state. Experimentally, wtGFP has been characterized as a photoacid.

Figure 4

HOMO and LUMO of the neutral (n, left) and anion (a, right) forms of p-HBDI, the chromophore of GFP, based on single-point calculations at the relaxed S1 geometries. H to L transitions constitute the majority of the S1 excitations (i.e., emission). The contributing percentages (squared configuration interaction coefficients) are shown after the comma.

HOMO and LUMO of the neutral (n, left) and n class="Chemical">anion (a, right) forms of p-HBDI, the chromophore of GFP, based on single-point calculations at the relaxed S1 geometries. H to L transitions constitute the majority of the S1 excitations (i.e., emission). The contributing percentages (squared configuration interaction coefficients) are shown after the comma.

7-Hydroxycoumarin

7-Hydroxycoumarin (7HC, aka umbelliferone) is a naturally occurring compound that is accumulated by certain fungal-infected plants to fight fungal growth.[14] It has been used as an indicator in assays to report the activities of enzymes[15] or the presence of small molecular targets.[16] The neutral and anion forms (Scheme ) absorb at 330 and 370 nm, respectively.[17] Although not pointed out specifically each time, the strong blue emission (λmax ∼ 460 nm) of 7HC in aqueous solutions (pH > 2) described in published works is from the deprotonated form, despite the fact that the ground state pKa value of 7HC is 7.8.[18] Therefore, the deprotonation occurs in the excited state,[14] making 7HC a photoacid in water (pKa* ∼ 0.4).[14] The antifungal property of 7HC has been attributed to its photoacidity.[14] The neutral dye emits at 397 nm in methanol where the photoinitiated deprotonation does not occur.[14]

Scheme 4

Neutral and Anion Forms of 7-Hydroxycoumarin (7HC)

Calculations show a decrease of both absorption and emission excitation energies of 7HC upon deprotonation (Tables S2 and S4), consistent with experimental observations. The H to L transitions account for the majority of the emissive S1 states of both the neutral and anion forms (Figure ). Both MOs are π orbitals that extend over the entire fluorophore including the hydroxy (or the oxide when deprotonated). Charge transfer from the phenol (or phenolate) moiety to the lactone component is conspicuous from the FMO plots (Figure ). For this reason, deprotonation would affect (i.e., raise) the HOMO level more than the LUMO level because the amplitude of HOMO on the hydroxyphenyl side is higher than that of the LUMO. The calculated change of HOMO and LUMO energies upon deprotonation is consistent with this interpretation (see data calculated at S0 geometry in Table S6).

Figure 5

HOMO and LUMO of the neutral (n, left) and anion (a, right) forms of 7-hydroxycoumarin (7HC) based on single-point calculations at the relaxed S1 geometries. H to L transitions constitute the majority of the S1 excitations (i.e., emission). The contributing percentages (squared configuration interaction coefficients) are shown after the comma.

HOMO and LUMO of the neutral (n, left) and n class="Chemical">anion (a, right) forms of 7-hydroxycoumarin (7HC) based on single-point calculations at the relaxed S1 geometries. H to L transitions constitute the majority of the S1 excitations (i.e., emission). The contributing percentages (squared configuration interaction coefficients) are shown after the comma.

2-Naphthol

2-Naphthol is one of the earliest fluorescent photoacids ever discovered[2] and has been thoroughly studied both experimentally and theoretically.[3a,19] The pKa value of 2-naphthol drops by ∼7 units (from 9.5 to 2.8) upon photoexcitation. It was later found that the isomer 1-naphthol (Scheme ) was an even stronger photoacid with a pKa* of about 0.[20] Because, unlike 1-naphthol, both the neutral and anion forms of 2-naphthol are fluorescent, the latter isomer has been used to illustrate the utilities of the Förster cycle and Förster equation (Figure )[21]—the simple but effective tools for preliminary assessment of the photoacidity of hydroxyaromatic fluorophores. The limitations of applying the Förster cycle to estimate photoacidity and causes of discrepancy from experimentally measured data (e.g., from ultrafast excited state dynamics experiments) were described in the review article by Ireland and Wyatt.[7] The excitation energy of 2-naphtholate is less than that of the neutral 2-naphthol, which is the driving force for the increased thermodynamic acidity in the excited state and is illustrated in the Förster cycle.

Scheme 5

2-Naphthol, Its Conjugate Base, and 1-Naphthol in the Box

Figure 6

Förster cycle (a) and Förster equation (b). NA: Avogadro number; h: Planck’s constant; ν: 0–0 transition frequency, which could be estimated using the average of absorption and emission maxima.

Similar to 7HC, the electron density of 2-naphthol redistributes after excitation away from the hydroxy moiety, as shown in the FMO plots of the neutral form at the relaxed S1 geometry (Figure , left). The acidity of 2-naphthol is therefore transiently enhanced to the extent that it is able to pass the proton to a capable base within reach during the lifetime of the excited state to produce the excited anion. The decrease of absorption and emission energies upon excitation can be explained similarly as with 7HC via the uneven changes of HOMO and LUMO effected by deprotonation.

Figure 7

HOMO and LUMO of the neutral (n, left) and anion (a, right) forms of 2-naphthol based on single-point calculations at the relaxed S1 geometries. H to L transitions constitute the majority of the S1 excitations (i.e., emission). The contributing percentages (squared configuration interaction coefficients) are shown after the comma.

HOMO and LUMO of the neutral (n, left) and n class="Chemical">anion (a, right) forms of 2-naphthol based on single-point calculations at the relaxed S1 geometries. H to L transitions constitute the majority of the S1 excitations (i.e., emission). The contributing percentages (squared configuration interaction coefficients) are shown after the comma.

Pyranine (a.k.a. 8-Hydroxypyrene-1,3,6-trisulfonic Acid, Trisodium Salt, HPTS)

HPTS is a pH indicator. More people may have encountered pyranine as the emitter in yellow highlighters. Its pKa was reported at ∼7.2,[22] which is within the narrow pH range of physiological fluids. Pyranine has only one emission band centered at 510 nm in water in the pH range of 4–10, while in ethanol, which is a weaker proton acceptor than water, the emission of the neutral form was found to center at 418 nm (Scheme ).[23] The absorption spectra of the neutral and the anion are centered at 400 and 450 nm, respectively,[22] and cross at 415 nm (i.e., the isosbestic point during a pH titration). Therefore, a ratiometric emission measurement via the excitation at 415 nm (independent of pH) and 460 nm (anion absorption and dependent on pH) would produce the degree of dissociation in the ground state which is used to calculate the pH value of the solution. The neutral pyranine is a photoacid (pKa* = 0.4) that has been used in pH-jump experiments to study biochemical proton transfer reactions.[24] Pyranine produces an excited anion in water, and consequently the anion emission (λmax = 510 nm, φ = 0.82) within the applicable pH range. The ESPT of pyranine to water was characterized by ultrafast spectroscopies.[23]

Scheme 6

Structures of Pyranine and Its Conjugate Base

The optimization of the trisodium salt form of n class="Chemical">pyranine was challenging. The positions of sodium ions were elusive, which prevented convergence. Sodium ions were then replaced by protons, which led to the successful convergence of both the neutral and the anion forms. Out of the four calculated excitation energy values (at S0 and S1 geometries of neutral and anion, Tables S2 and S4), the absorption of the anion of pyranine yielded the largest deviation with an underestimation of 0.44 eV. All the rest were within 0.3 eV of the experimental values. The lowest energy excitations of pyranine are primarily represented by the H to L transitions (Figure ). Both MOs cover the hydroxy group and the pyrene core. Deprotonation modulates the MO energies the same way as described for 7HC, which leads to the decrease in both absorption and emission energies.

Figure 8

HOMO and LUMO of the neutral (n, left) and anion (a, right) forms of pyranine (a.k.a. HPTS) based on single-point calculations at the optimized S1 geometries. H to L transitions constitute the majority of the S1 excitations (i.e., emission). The contributing percentages (squared configuration interaction coefficients) are shown after the comma.

HOMO and LUMO of the neutral (n, left) and n class="Chemical">anion (a, right) forms of pyranine (a.k.a. HPTS) based on single-point calculations at the optimized S1 geometries. H to L transitions constitute the majority of the S1 excitations (i.e., emission). The contributing percentages (squared configuration interaction coefficients) are shown after the comma.

2-(2′-Hydroxyphenyl)benzoxazole (HBO)

HBO forms an intramolecular hydrogen bond (HB) between the photoacid hydroxyphenyl and the photobase benzoxazole moieties (see the “enol” structure in Scheme ). Electronic excitation would lead to the ESPT within the same molecule, which is conventionally referred to as “excited state intramolecular proton transfer” (ESIPT) to transform from an O–H isomer (“normal” or “enol”) to an excited N–H isomer (“tautomer” or “keto”).[25] HBO is one of the smallest ESIPT-capable compounds that is emissive in both normal and tautomer forms, and for that reason, HBO and its derivatives have been studied extensively. The enol S1 state emits at ∼370 nm, which is observed in solvents that would engage HBO via an intermolecular HB (Scheme ). The keto form emits at ∼480 nm and dominates in weakly hydrogen bonding solvents (e.g., hexanes, dichloromethane, or acetonitrile) in which the intramolecular HB is preserved. The emission of the anion form is centered at ∼450 nm.[26] Therefore, if the keto emission is considered as the fluorescence from the neutral form, HBO would be an exception from the previously stated observations that the emission of the anion has a longer wavelength than its neutral, conjugate acid form.

Scheme 7

Origins of Three Emission (Enol, Keto, and Enolate) Bands of HBO

The ground state (GS) of the n class="Chemical">HBO neutral form has two major conformers in HB-permitting solvents: syn-enol and anti-enol (Scheme ). Syn-enol is the one that is conducive to ESIPT, while anti-enol is not. At the DFT/B3LYP/def2-SVPD level of theory, anti-enol is predicted to be 6.1 kcal/mol higher in energy than syn-enol. This energetic disparity is consistent with the observation that, in non-hydrogen bonding solvents, only the keto tautomer emission after the ESIPT from the syn-enol is observed.

Scheme 8

Major Conformers of the Neutral Forms (Top) and of the Conjugate Base (Bottom) of HBO

The computational characterization of the HBO anion presents a new challenge (see the detailed description in the SI). The single bond rotation between benzoxazole and phenolate creates conformational isomers (conformers). Three major conformers of the HBO anion that were investigated are shown in Scheme . Of the two planar conformers, one is referred to as cis-planar, where the oxide is found on the same side of the C–C bond with the oxygen in oxazole, while the other is trans-planar in which the oxide is on the opposite side of the C–C bond with regard to the heterocyclic oxygen. The third conformer exhibits an almost right dihedral angle between benzoxazole and phenoxide, hereby referred to as the twisted conformer. Both planar conformers were identified as minima on the GS, between which the cis-planar is more stable than the trans-planar (Table S7). On the S1 surface (XS), with some challenges (see the SI) the cis-planar conformer was optimized to a minimum with excitation energy and oscillator strength consistent with experimental emission data.

The H to L transitions are major contributors to the emissions of all three species of HBO: enol, anion, and keto (Figure ). The HOMO of the enol form, as well as the LUMOs of all three species, are delocalized over the entire molecule that includes the deprotonatable O–H or N–H bond. The HOMOs of the anion and the keto forms on the other hand are heavily localized on the phenoxide moiety, which of the neutral tautomer form does not involve the N–H bond (Figure ). Therefore, deprotonation of the tautomer keto form would impact (i.e., increase) the LUMO level much more than the HOMO level, resulting in a larger HOMO–LUMO gap, as opposed to a smaller one as shown for the earlier examples, which would account for the shorter emission wavelength maximum (higher excitation energy) of the anion than that of the keto form.

Figure 9

HOMO and LUMO diagrams of the enol (left, E), anion (middle, A), and keto (right, K) forms of HBO at the relaxed S1 geometries. H to L transitions constitute the majority of the S1 excitations (i.e., emission). The contributing percentages (squared configuration interaction coefficients) are shown after the comma.

HOMO and LUMO diagrams of the n class="Chemical">enol (left, E), anion (middle, A), and keto (right, K) forms of HBO at the relaxed S1 geometries. H to L transitions constitute the majority of the S1 excitations (i.e., emission). The contributing percentages (squared configuration interaction coefficients) are shown after the comma.

BIPYVHBO

BIPYVHBO was constructed as a fusion of the ESIPT-capable HBO and an internal charge transfer (ICT)-type stilbenoid (Figure ).[27] Deprotonation of BIPYVHBO would amplify the ICT of the stilbenoid component. The presence of the bipyvinyl substituent on the 5′-position of HBO shifts both the enol (when solvated) and keto (when intramolecularly hydrogen bonded) emission of HBO to cyan and green regions, respectively, while deprotonation moves the emission maximum to almost 600 nm. Therefore, this compound, depending on solvent and the presence of a base additive (e.g., DBU), may show three emission colors that roughly coincide with the three primary colors (blue, green, and orange/red).[27]

Figure 10

BIPYVHBO is a fusion of an ESIPT fluorophore HBO and an ICT stilbenoid.

BIPYVHBO is a fusion of an n class="Chemical">ESIPT fluorophore HBO and an ICT stilbenoid.

Similar to HBO, the syn-enol conformer of BIPYVHBO is lower in energy than the anti-enol conformer in the GS. The GS syn-enol conformer exhibits the FMO arrangements (Figure a) that are uneven over the landscape of the molecule: both HOMO and LUMO reside primarily on the stilbenoid component, while LUMO+1, which is only 0.1 eV higher above LUMO, is primarily on the HBO. However, the lowest electronic excitation of S0 to the S1 state is attributed in a large portion to the H to L+1 transition, which is charge transfer in nature and therefore introduces a dipole stabilization factor to become the major contributing transition to the lowest excited state. This transition that involves the HBO component (on LUMO+1) is the one that drives the ESIPT. When the hydroxyl group is solvated by a HB-basic solvent (e.g., DMSO) or is deprotonated, the S1 state is then represented more by the H to L transition restricted on the stilbenoid component (unpublished). By this analysis, BIPYVHBO exhibits the properties of its component fluorophores depending on the nature of the solvent or the presence of a base.

Figure 11

FMOs of BIPYVHBO that contains an intramolecular N···O-H hydrogen bond at the (a) ground state and (b) excited state (E: enol; A: anion; and K: keto) geometries. H to L transitions constitute the majority of the S1 excitations (i.e., emission) in (b). The contributing percentages (squared configuration interaction coefficients) are shown after the comma.

At the relaxed S1 geometries with the intramolecular HB preserved, the H to L transitions are found to be the major contributors to the S1 excitations (i.e., emission) of enol, anion, and keto species (Figure b). Different from the GS of the enol, now the LUMOs of both enol and keto forms at the relaxed S1 geometries are found on the HBO component, while the HOMOs of both forms remain on the stilbenoid. The deprotonated BIPYVHBO at the relaxed S1 geometry has a further drastically altered FMO localization profile: the HOMO is primarily on the HBO component that now includes the phenoxide moiety, while the LUMO is found on the bipyvinyl section (Figure b). The fact that FMOs localize on different sections (HBO vs stilbenoid) of the dye in a deprotonation-dependent manner separates this compound apart from earlier examples.

Hydroxyphenyl-Substituted 1,2,3-Triazoles

These compounds are similar to HBO in possessing intramolecular HBs (Scheme ).[28] Unlike HBO, they upon excitation do not produce an emissive tautomer; i.e., either no ESIPT occurs, or there is no emission from a product resulting from the ESIPT. Unlike the hydroxyaromatic dyes described earlier other than HBO, the ESIPT-incapable HP-TZ1 upon deprotonation in DMSO produces an anion that emits at a higher energy than the neutral. In contrast to HP-TZ1, its regioisomer HP-TZ2 undergoes an emission red shift upon deprotonation (Table ).

Scheme 9

Computationally Studied Conformers of HP-TZ1 (a), HP-TZ2 (b), HP-TZ3 (c), and Their Conjugated Bases

In this series of compounds, the chromophore can be considered as the combination of the C4 substituent (see triazole numbering in Scheme a) and the triazolyl moiety, which are more coplanar (based on both experiment and calculation) than the combination of the N1 substituent and triazolyl.[28] This putative chromophore is charge transfer in nature where the triazole heterocycle is the e-withdrawing charge acceptor.[28] When HP-TZ1 is deprotonated at the N1-substituted hydroxyphenyl, a negative charge is placed at the negative end of the excited state dipole of the chromophore, which would explain the blue shift of emission upon deprotonation. When HP-TZ2 is deprotonated (Scheme b), the C4-substituted phenolate becomes a stronger e-donor in the charge transfer chromophore to instead result in a red shift of emission. Computation has revealed a more nuanced picture that is not contradictory to this descriptive explanation.

HP-TZ3 contains hydroxyphenyl (HP) groups at both N1 and C4 positions, with two intramolecular HBs. Its neutral form is not emissive. The change of emission properties upon deprotonation would have to depend on which HP dissociates first. The N1 position of a 1,2,3-triazole exerts a larger e-withdrawing effect than the C4 position.[28] Therefore, similar to HP-TZ1, the N1-HP is deprotonated to result in a blue-light-emitting monoanion (Scheme c).

The HOMO and LUMO of neutral and anionic HP-TZ1 in the S1 states are shown in Figure . There are two key differences from all other fluorophores described thus far that may help explain the peculiarity of the fluorescence properties of this compound. First, the hydroxyphenyl moiety of HP-TZ1 is not involved in the HOMO of the neutral form, which is consistent with the lack of ESIPT of HP-TZ1. Second, deprotonation raises the LUMO level much more than it does to the HOMO level (Figure ), which would lead to the shift of emission to a shorter wavelength upon deprotonation, and that was indeed experimentally observed. Deprotonation puts the phenoxide in position to dominate the HOMO. This observation is replicated in HP-TZ2 and HP-TZ3.

Figure 12

HOMO and LUMO of the neutral (n, left) and anion (a, right) forms of HP-TZ1 at the relaxed S1 geometries. Changes of orbital energies upon deprotonation are listed on the right side. H to L transitions constitute the majority of the S1 excitations (i.e., emission). The contributing percentages (squared configuration interaction coefficients) are shown after the comma.

HP-TZ2 is a regioisomer of HP-TZ1. The hydroxyphenyl group of HP-TZ2 is substituted on the C4 position of triazole, unlike in HP-TZ1 where the hydroxyphenyl is on the N1 position of the triazole. Deprotonation of HP-TZ2 moves the emission to a longer wavelength, an opposite effect from that witnessed for HP-TZ1. The HOMO and LUMO plotted at the relaxed S1 geometry are shown in Figure . As is the case with HP-TZ1, the HOMO of the neutral form of HP-TZ2 does not involve the hydroxyphenyl, which is consistent with the absence of ESIPT. Different from HP-TZ1, deprotonation raises the HOMO level more than the LUMO level, which explains the observed emission shift to a lower energy upon deprotonation.

Figure 13

HOMO and LUMO of the neutral (n, left) and anionic (a, right) forms of HP-TZ2 at the relaxed S1 geometries. The changes of orbital energies upon deprotonation are listed on the right side. H to L transitions constitute the majority of the S1 excitations (i.e., emission). The contributing percentages (squared configuration interaction coefficients) are shown after the comma.

HP-TZ3 has two hydroxyphenyl (HP) groups on the N1 and C4 positions of the triazole. Evidence has been presented that N1 exerts more e-withdrawing power of a 1,2,3-triazole than C4, therefore rendering the hydroxyphenyl group on N1 more acidic than the one on C4.[28] Deprotonation is depicted to occur at the N1 substituent (Scheme c), and calculation showed that the C4-deprotonated isomer was higher in energy in the GS.

The neutral form of HP-TZ3 is not emissive. Therefore, the calculation focused on its monoanion form. Both the calculated excitation energies and oscillator strengths of the anions of HP-TZ1 and HP-TZ3 are similar. The similarity can also be seen in the profiles of the FMOs (Figures and 14) that characterize the emissive states of both anions. The calculated outcomes are consistent with the closeness of the experimentally observed emission band positions and brightness of the (mono)anions of HP-TZ1 and HP-TZ3 (Table ).

Figure 14

HOMO and LUMO of the anion (a) form of HP-TZ3 at the relaxed S1 geometry. H to L transitions constitute the majority of the S1 excitations (i.e., emission). The contributing percentages (squared configuration interaction coefficients) are shown after the comma.

HOMO and LUMO of the n class="Chemical">anion (a) form of HP-TZ3 at the relaxed S1 geometry. H to L transitions constitute the majority of the S1 excitations (i.e., emission). The contributing percentages (squared configuration interaction coefficients) are shown after the comma.

Discussion

Ten hydroxyaromatic fluorescent dyes are described in this article. The purpose is to help interpret the photophysical properties of these compounds, as summarized in Figure , and to understand the exceptions to the generalizations. These compounds are mostly well-known and should be of general interest to individuals who study the properties and applications of fluorescent dyes. The selections include widely applied fluorescent labels (fluorescein, GFP, 7HC), archetypal photoacids (2-naphthol, pyranine), one of the most known dyes capable of ESIPT (HBO), and intramolecular HB-containing dyes that our group has invested much effort in to understand (BIPYVHBO, HP-TZs).

Deprotonation in most cases decreases both the lowest absorption and emission energies of these dyes (#1 and #2 in Figure ). This effect on absorption would lead to photoacidity of the dye, as described by the Förster equation. In reality, whether the photoacidity manifests itself depends on the proton-accepting ability of the environment. The lowering of absorption and/or emission energy upon deprotonation can be attributed to the differential effect of deprotonation on the involved FMO levels (i.e., HOMO (H) and LUMO (L) of all the examples in this article), assuming the H to L transition is the major contributor to the S1 states that are spectroscopically characterized by the lowest energy absorption or emission bands. Deprotonation generally raises the HOMO level more than the LUMO level if both MOs delocalize on the entire chromophore including the hydroxy group. Consequently, the HOMO–LUMO gap is reduced, which is materialized with the drop of absorption or emission energies upon deprotonation. When the same functional, basis sets, and solvation parameters were applied in calculations, these FMO energy changes were reliably obtained at the relaxed S0 geometries (i.e., related to absorption, see Table S6). The HOMO and LUMO energy changes of the relaxed emissive S1 states upon deprotonation follow the same trend based on our preliminary calculations. However, they are not tabulated because in some cases different functionals, basis sets, or solvent parameters were used to optimize the S1 geometries and/or determine the excitation energies of the neutral and anion forms of the same dye, which lowers our confidence of their quantitative comparability.

The emission of the neutral form of HBO, an ESIPT-capable dye, in nonpolar solvents is found at a longer wavelength than its anion, which appears to be an exception to the rule (#6 in Figure ). However, the emission of HBO in nonpolar solvents almost exclusively comes from the excited state proton transferred tautomer form. The HOMO of the excited tautomer only occupies part of the structure that does not involve the N–H bond, while the LUMO covers the entire molecule, including the N–H bond. Deprotonation of the N–H bond would then raise the LUMO level more than the HOMO that does not cover the site of deprotonation. This explains the apparent “emission blue shift” caused by deprotonation of HBO in its neutral, albeit a photoexcited, tautomer form. This observation and the interpretation expose a major difference between HBO (and several compounds described afterward), where FMOs may only occupy part of the chromophore that excludes the deprotonatable bond, and the first five compounds, where HOMO and LUMO diffuse over the entire hydroxyaromatic moiety.

BIPYVHBO is also an n class="Chemical">ESIPT-capable compound whose anion emission is however found at a longer wavelength than both the neutral enol and keto (i.e., tautomer) forms. Therefore, it is an example that appears to immediately contradict the preceding conclusion. BIPYVHBO is a fusion of the ESIPT-capable HBO and the push–pull stilbenoid dye. Consequently, it possesses the photophysical potentials of both and expresses the emission of either as the conditions dictate. The neutral form adopts the properties of HBO, which produces the normal and tautomer emissions in blue and green regions of the spectrum. Upon deprotonation, the push–pull stilbenoid becomes the dominant fluorophore and emits at a lower energy that is red-shifted from both emission bands of the neutral form.

The non-ESIPT, intramolecular HB-containing HP-TZ1 is the clearest exception to conclusion #2 listed in Figure . The hydroxyphenyl group is not a part of the HOMO of HP-TZ1 in its excited neutral form. Upon deprotonation, the HOMO translocates from the dimethylanilinyl moiety to the phenoxide that is stabilized by the N1 position of 1,2,3-triazole. The overall effect of deprotonation only slightly raises the energy of the HOMO level while increasing the LUMO level that includes the phenol (or phenoxide) moiety in both neutral and anion forms by a larger margin. For this reason, the HOMO–LUMO gap widens upon deprotonation of HP-TZ1 at the relaxed S1 geometry, and consequently the emission shifts to a shorter wavelength. HP-TZ2, the regioisomer of HP-TZ1, contains a hydroxyphenyl group on the C4 position which upon deprotonation produces a phenoxide that is not stabilized effectively at the C4 position of 1,2,3-triazole. As such, the HOMO level is raised more than the LUMO. The latter translocates to the dimethylanilinyl portion upon deprotonation, to result in a red shift of emission. The comparison of HP-TZ1 and HP-TZ2 exposes the different electron-withdrawing abilities of N1 and C4 positions of 1,2,3-triazole on a substituent (N1 is more effective than C4), which is manifested as the drastically different effects of deprotonation on their emission properties.

Photoacidity (#4 in Figure ) is a consequence of the reduction of the S0–S1 energy gap upon deprotonation.[7] This is a thermodynamic conclusion that may or may not lead to the fulfillment of a proton transfer event considering the brevity of the lifetime of a singlet excited state (#5 in Figure ). The enhanced acidity upon excitation may be masked by the pairing of a relatively weak proton acceptor (e.g., MeOH instead of water) so that a kinetic barrier is enacted to prevent proton transfer from occurring in the allotted short period of time. This is why, in some cases, ESPT has to be catalyzed by buffers,[6,29] which would aid the delivery of a proton from the photoacid to a reluctant acceptor.

The difference of emission quantum yields between the neutral and anion forms of the same dye (#3 in Figure ) should not be principally attributed to the difference in rates of radiative decays because the calculated oscillator strengths of the emissive states of neutrals and anions are all of respectable values. Therefore, the difference lies in how the compound interacts with solvents, in particular water, to open up nonradiative relaxation pathways. The strengthened hydrogen bonds with solvent molecules of the excited states of neutral species could very well be the conduit of excitation energy dissipation, therefore providing an efficient path of relaxation without emission.[30]

The last part of the Discussion is reserved for the summary of the challenges in the computational inquisition of the structural and electronic properties of hydroxyaromatic dyes in ground and excited states. First, more often than not, only one conformer out of many is selected for the calculation. The chosen conformer may or may not be representative of the overall emission properties of the subject under investigation (e.g., it may be dominant in the ground state but would not afford an emissive excited state, or vice versa). Second, hydroxyaromatics tend to engage in specific solvent interactions, e.g., hydrogen bonding. The anions would interact with the counter cations, a process that is also highly medium-dependent. These properties lead to high sensitivities of the calculated excitation energies to the choices of basis sets, exchange-correlation functionals, and the methods of modeling solvent effects. Most calculations conducted in this work require diffuse functions in the basis sets (aug-cc-pVDZ or def2-SVPD). There is evidence that range-separated hybrid (RSH) functionals (such as LC-BLYP, which provided a success story of modeling the excited states of the HP-TZ series of dyes), rather than, for example, the popular B3LYP, are required for modeling highly polarizable or anionic molecular dyes. Third, in this work as well as others, the accuracy of excitation energy calculations is defined by the mean absolute deviation (MAD) of the calculated excitation energies from the experimental absorption or emission maxima, which may slightly vary between different reports. Therefore, the accuracy depends on not only the modeling of the electronic structure of the dye and the interaction of the excited dye with solvent molecules but also the systematic errors that are specific to each experimental measurement.

Surveying of the data in Tables S2 and S4 gives (1) a glimpse of the accuracies that could be achieved by the methods applied in this work and (2) the patterns of deviations that depend on the charge and polarizability of the structure. The excitation energy calculated at the relaxed S0 geometry is used to compare with the maximum of the absorption band, while the excitation energy calculated at the relaxed S1 geometry would correspond to the maximum of the emission band. The mean absolute deviations (MADs) of calculated lowest absorption energies of both neutral and anions, as well as the calculated emission energies of the neutral, are about 0.2 eV or less. The emission energies of the anions are the most challenging to model (MAD = 0.25 eV), and several of them required an RSH functional in the optimization of the S1 states. The MAD values reported in this work are comparable to ones considered acceptable from several works on benchmarking the calculations of excitation energies of organic fluorophores.[31]

The deviations of excitation energies are mostly positive (i.e., overestimation as marked in blue in Tables S2–5) for the absorption of the neutral species (8 out of 10, see Table S2), while they are mostly negative (i.e., underestimation as marked in red in Tables S2–5) for the emission of the anionic species (9 out of 10, see Table S4). The deviations from the experimental values accrued for the absorption of the anions and the emission of the neutral species are more random. Therefore, it appears that increasing charge (e.g., anion rather than neutral) and/or polarizability (e.g., excited rather than ground state) would skew the calculated energies to lower values using the methods in this work, while the excitation energy of a neutral molecule with a low polarizability tends to be overestimated.

For the purpose stated in this work, which is to provide preliminary predictions of excitation energies without an extravagant computational cost, what we could strive for is to make informed choices of functional and basis set that are critical in a reasonable description of the electronic structure of the subject. The solvent effect would have to be simulated with an implicit continuum model without considering specific interactions from explicit solvent molecules or cations within the first solvation shell because factoring in these specific solvent or counterion effects would raise the sophistication of the method to a level that is not affordable or required for our limited purpose. The errors originating from modeling electronic structure and solvation would be difficult to separate, which would collectively reach values on the order of a few tenths of an eV. Understanding these limitations is important for one to determine what compromises could be made in order to extract useful information via computation at an affordable cost.

Conclusion

This article emphasizes the following facts of hydroxyaromatic fluorophores: (1) the observed emission of many of these fluorophores belongs to the excited conjugate bases, rather than the neutral structures; (2) the absorption and emission wavelengths in most cases, but not all, of the neutral dyes are shorter than their conjugate bases. The lowering of excitation energy upon deprotonation is attributed to the more effective elevation of the HOMO level than the LUMO level; (3) exceptions to #2 are known, especially the emissions of the dyes of which the chromophore contains an intramolecular hydrogen bond; (4) the anions are usually brighter than the neutral because (a) the lowest energy transition of the excited anion is more allowed than the excited neutral form based on the calculated oscillator strengths and percentages of the contributing allowed π → π* transitions and (b) the OH hydrogen bonding with solvent provides a major quenching pathway of the excited states of the neutrals. This article also presents an overview on how to use quantum chemical calculations to understand, and hopefully to predict, the excitation energies (i.e., absorption and emission) of hydroxyaromatic fluorophores, while laying bare the challenges in the calculation of anions.

Theoretical Methods

In this article, the criteria of a successful calculation is (1) the geometry optimization needs to converge to a minimum; (2) the excitation energy values and oscillator strengths need to be consistent with the experimental absorption or emission wavelength maxima and intensity; and (3) it needs to be done without an extravagant amount of computing time. The following procedure was used to calculate the excitation energies of hydroxyaromatic fluorophores and their conjugate bases at their relaxed ground (S0, comparable to absorption λmax) and excited (S1, comparable to emission λmax) state geometries with a compromise between efficiency and accuracy.

Geometry Optimization

Unless otherwise noted, the S0 and S1 structures of fluorophores were optimized using Kohn–Sham density functional theory (KS-DFT)[32] and time-dependent (TD) DFT,[33] respectively, with the Becke, 3-parameter, Lee–Yang–Parr (B3LYP) exchange-correlation (XC) functional[34] and aug-cc-pVDZ[35] or def2-SVPD[36] basis set (see Table S1). Both basis sets contain diffuse functions which are important for describing anions.[37] For all the optimizations, the solvent effect was not considered because (1) a solvation model such as COSMO has not been implemented in excited-state gradient calculations in Turbomole 7.4,[38] the quantum chemistry package in the possession of the corresponding author (solvent effects were considered in single-point excitation energy calculations, see below) and (2) numerical, rather than analytical, frequency calculations have to be conducted for structures optimized under COSMO, which to us is too computationally costly. Hessian matrices and the associated eigenvalues at optimized geometries were computed to make sure that the optimization had resulted in a true minimum. For the ground states, the neutral and anion forms of each pair were optimized using the same method and basis set (in most cases B3LYP/aug-cc-pVDZ) so that the comparison of the calculated properties (e.g., excitation energies and FMO levels) could be confidently related to the difference in observed properties. For the excited states, the relatively small basis set def2-SVPD was selected as the first option for optimizations to save computing time. Excited states are optimized using time-dependent density functional theory (TDDFT). If an optimization using TDDFT/B3LYP/def2-SVPD failed to converge or converged but failed to result in a true minimum, basis sets aug-cc-pTDZ or def2-TZVP[36] (or others in rarer instances) were used with the convergence thresholds tightened, which cost more computing time but usually located a true minimum.

Several compounds (anions of HBO and HP-TZ3 and both neutral and anion forms of HP-TZ1 and HP-TZ2) failed to be optimized to emissive excited state minima using TDDFT/B3LYP. They were optimized using the long-range-corrected BLYP (LC-BLYP) functional,[39] which is a range-separated hybrid (RSH) functional[40] in which the long-range exchange is treated by 100% HF for producing correct exchange potential outside the molecular subject. TDDFT/LC-BLYP calculations were performed using the ORCA program.[41] The neutral forms of the HP-TZ compounds were calculated using 6-311G* basis sets,[42] since we were not able to obtain their true minima using def2-SVPD basis sets. The excited anions of HP-TZs and HBO were calculated using def2-SVPD basis sets. For the anion calculations, the resolution of identity (RI) approximation is used to calculate the Coulomb integrals, and the numerical chain-of-sphere integration is used for calculating the HF exchange (COSX) integrals.[43] The CAM-B3LYP[44] in place of LC-BLYP was also tried for the optimizations of the excited states of the six structures in question. However, the excitation energy values based on CAM-B3LYP-optimized structures yielded larger deviations from experimental emission maxima than those optimized using TDDFT/LC-BLYP (Table S8).

A summary of the XC functionals and basis sets used for geometry optimizations is given in Table S1. More systematic comparison between different XC functionals and efforts to achieve higher accuracy in the excitation energy calculations of small organic charged or polarized fluorophores will be made in the future as an independent study.

Excitation Energy Calculation

The excitation energy and oscillator strength were calculated at the ground state structures for UV/vis absorption and excited state structures for emission, using the algebraic diagrammatic construction (ADC)(2) scheme[45] with either aug-cc-pVDZ or def2-SVPD basis sets. The emission spectra of many of the dyes are sensitive to solvent. Therefore, the effect of solvent must be considered in the calculations of the excitation energies. In this work solvation was treated using the implicit continuum solvation model (COSMO)[46] with the approach of perturbation theory on energy and density (PTED).[47] The absorption energies calculated using PTED-COSMO-ADC(2)/aug-cc-pVDZ are listed in Table S2 with the deviations from experimental data, while the calculated emission energies using the same method are included in Table S4. The results using the def2-SVPD basis set are listed in Tables S3 and S5. The use of basis set aug-cc-pVDZ achieved slightly better, but not substantial, mean absolute deviations from experimental data than def2-SVPD. The changes of HOMO and LUMO energies upon deprotonation (ΔHOMO and ΔLUMO) of the first 7 compounds after single-point excitation energy calculation, which takes into consideration of solvation, are listed in Table S6. The ΔHOMO and ΔLUMO upon deprotonation of the same set of compounds calculated at the relaxed excited state (S1) geometries, though tracking the trend of the data derived from ground state geometries, were not listed because for each pair of excited neutral and anion forms the functional, basis set, and/or solvation parameters could not be kept identical to provide full confidence in their comparability. The single-point excitation energy calculations were done using the quantum chemistry package Turbomole (V7.4).[38]