UV synchrotron radiation linear dichroism spectroscopy of the anti-psoriatic drug anthralin

Sample preparation

A sample of A (CAS No. 1143-38-0) was purchased from Sigma-Aldrich and purified by column chromatography as previously described (Andersen & Spanget-Larsen, 1997). The near-UV absorbance spectrum of the purified substance in n-heptane solution (Merck Uvasol) is shown in Fig. S1. Low-density polyethylene (PE) 100 µm sheet material was obtained from Hinnum Plast A/S. A 2.5 ×6 cm PE piece cut from the sheet was purified by extraction with chloroform (Merck Uvasol) at 50 °C for one day. A was introduced by submersion of the dried PE sample into a saturated chloroform solution of the substance in a sealed container at room temperature for three days. After evaporation of the chloroform from the doped sample, the surface was cleaned with ethanol (Merck Uvasol) to remove crystalline deposits. The PE sample was finally uniaxially stretched by 500%. A sample without solute was produced in the same manner for use as a reference. Further details on stretched polyethylene samples can be found in the literature (Michl & Thulstrup, 1986; Thulstrup & Michl, 1989).

Linear dichroism (LD) spectroscopy

LD spectra in the range 31,300–20,000 cm⁻¹ (320–500 nm) were recorded on a UV-2101PC Shimadzu spectrophotometer at Roskilde University. SRLD spectra were measured in the range 58,000–31,300 cm⁻¹ (170–320 nm) on the CD1 beamline (Miles et al., 2007; Miles et al., 2008) at the storage ring ASTRID at the Centre for Storage Ring Facilities (ISA). As previously described (Andersen & Spanget-Larsen, 1997; Nguyen et al., 2018; Nguyen et al., 2019), two absorbance curves were recorded at room temperature with the electric vector of the sample beam parallel (U) and perpendicular (V) to the stretching direction of the PE sample. The observed baseline-corrected LD absorbance curves $E_U\left(\tilde{\nu }\right)$ and $E_V\left(\tilde{\nu }\right)$ are shown in Fig. 2A. The isotropic absorbance curve $A_{\mathrm{ISO}}\left(\tilde{\nu }\right)=E_U\left(\tilde{\nu }\right)+2E_V\left(\tilde{\nu }\right)$ is shown in Fig. S2.

Calculations

Quantum chemical calculations considering isolated molecules in the gas phase were performed by using the Gaussian16 software package (Frisch et al., 2016). Optimizations of molecular geometry were performed with B3LYP (Becke, 1993; Lee, Yang & Parr, 1988). This DFT was found to be very successful in the prediction of molecular and vibrational structures of compounds like A with intramolecular C=O⋅⋅⋅HO hydrogen bonding (Spanget-Larsen, Hansen & Hansen, 2011; Hansen & Spanget-Larsen, 2012). Geometry optimizations for the ground state (S₀) and the lowest excited singlet state (S₁) of A were carried out with B3LYP and TD-B3LYP, respectively, using the 6-31+G(d,p) basis set (Foresman & Frisch, 2015; Frisch et al., 2016). Vertical electronic transitions from the ground state to the 60 lowest excited singlet states were computed with TD–CAM-B3LYP (Yanai, Tew & Handy, 2004) and the basis set aug-cc-pVTZ (Dunning Jr, 1989; Kendall, Dunning Jr & Harrison, 1992). We recently found that this long-range corrected procedure is adequate in the prediction of electronic transitions of aromatic chromophores in the near and vacuum UV regions (Nguyen et al., 2018; Nguyen et al., 2019). A constant term was subtracted from the excitation wavenumbers calculated with TD–CAM-B3LYP in order to facilitate comparison with the observed spectra (Grimme, 2004; Nguyen et al., 2018; Nguyen et al., 2019); an empirical correction of 3,500 cm⁻¹ was found to be adequate in the present case. The main predicted transitions are listed in Table 1 together with the observed ones. A complete listing of the calculated transitions is provided as Data S1. Graphical representations of transitions to ¹A₁ and ¹B₂ states are shown in Fig. 3. Gaussian convolutions considering all allowed transitions are provided as Fig. S3, using previously described procedures (Serr & O’Boyle, 2009; Nguyen et al., 2019). Detailed results for the ESIPT photoproduct 8,9-dihydroxy-1(10H)-anthracenone are given in Data S2, results for additional excited state configurations are outlined in Fig. S4.

Linear dichroism: orientation factors and partial absorbance curves

The observed SRLD absorption curves $E_U\left(\tilde{\nu }\right)$ and $E_V\left(\tilde{\nu }\right)$ for A partially aligned in stretched PE are shown in Fig. 2A. The directional information that can be extracted from these curves is represented by the orientation factors K_i for the molecular transition moment vectors ${\overrightarrow{M}}_i$ of the observed spectroscopic features i (Michl & Thulstrup, 1986; Thulstrup & Michl, 1989): (1)${\displaystyle K_i=〈{cos}^2\left({\overrightarrow{M}}_i,U\right)〉.}$

The orientation factor is defined as an average over all solute molecules in the light path, indicated by the pointed brackets in Eq. (1), where $\left({\overrightarrow{M}}_i,U\right)$ is the angle of the moment vector of transition i with the polymer stretching direction, U. The K_i values may be determined by the graphical TEM procedure (Michl & Thulstrup, 1986; Thulstrup & Michl, 1989) which involves construction of linear combinations of $E_U\left(\tilde{\nu }\right)$ and $E_V\left(\tilde{\nu }\right)$. In the present application we consider the reduced absorbance curves $r_K\left(\tilde{\nu }\right)$ (Madsen et al., 1992): (2)${\displaystyle r_K\left(\tilde{\nu }\right)=\left(1-K\right)E_U\left(\tilde{\nu }\right)-2K{E}_V\left(\tilde{\nu }\right).}$

A family of curves $r_K\left(\tilde{\nu }\right)$ for A is shown in Fig. 2B. A spectral peak or shoulder due to transition i disappears from the linear combination $r_K\left(\tilde{\nu }\right)$ for K = K_i and the K_i value can thus be determined by visual inspection. The molecular point group of A is C_2v (Ahmed, 1980) and allowed vertical transitions must be polarized along the three symmetry axes x, y, and z, corresponding to excited states of B₁, B₂, and A₁ symmetry, respectively. Hence, only three different K_i values corresponding to K_x, K_y, and K_z are expected. From the $r_K\left(\tilde{\nu }\right)$ curves in Fig. 2B K values close to 0.60 can be estimated for the features at 34,600, 39,200, 44,000 and 46,100 cm⁻¹, and values close to 0.25 for those at 37,600, 42,900, 51,000 and 53,400 cm⁻¹. As previously discussed (Andersen & Spanget-Larsen, 1997), the relatively broad band with maximum close to 27,500 cm⁻¹ (∼365 nm) is due to two near-degenerate, differently polarized transitions; their K values cannot be determined directly.

The K values 0.60 and 0.25 must be assigned to K_y and K_z, corresponding to transitions polarized along the in-plane long and short molecular axes y and z. Under the assumption that absorbance polarized along the out-of-plane x axis is negligible in the present spectra, it is possible to construct the partial absorbance curves $A_y\left(\tilde{\nu }\right)$ and $A_z\left(\tilde{\nu }\right)$ indicating y- and z- polarized intensity (Madsen et al., 1992):

(3)${\displaystyle A_y\left(\tilde{\nu }\right)={\left(K_y-K_z\right)}^{-1}\cdot r_{K_z}\left(\tilde{\nu }\right)A_z\left(\tilde{\nu }\right)={\left(K_z-K_y\right)}^{-1}\cdot r_{K_y}\left(\tilde{\nu }\right).}$

The curves $A_y\left(\tilde{\nu }\right)$ and $A_z\left(\tilde{\nu }\right)$ produced with (K_y, K_z) = (0.60, 0.25) are shown in Fig. 2C and in Fig. 3. They are shown on an expanded absorbance scale in Fig. S2, together with the isotropic absorbance curve $A_{\mathrm{ISO}}\left(\tilde{\nu }\right)=A_y\left(\tilde{\nu }\right)+A_z\left(\tilde{\nu }\right)=E_U\left(\tilde{\nu }\right)+2E_V\left(\tilde{\nu }\right)$. The latter is equal to three times the absorbance that would have been measured in an isotropic experiment on the same sample. Six long-axis (y) polarized features B, C, E, G, H, and L and six short-axis (z) polarized features A, D, F, I, J, and K are indicated. Observed peak wavenumbers, absorbances, and polarization directions are listed in Table 1.

Electronic transitions

The spectrum starts with a relatively broad band system with maximum close to 27,500 cm⁻¹ (364 nm). The absorbance is predominantly long-axis (y) polarized, overlapping weaker short-axis (z) polarized intensity. The short-axis polarized absorbance A with maximum at 27,300 cm⁻¹ (366 nm) can be assigned to the 2¹A₁ state computed at 27,900 cm⁻¹ (Table 1, Fig. 3). The computed transition is well described by promotion of an electron from the 6b ₁(π) HOMO to the 7b ₁(π*) LUMO. The intense long-axis polarized band B peaking at 27,700 cm⁻¹ (361 nm) is assigned to the 1¹B₂ state predicted at 27,700 cm⁻¹ (Table 1, Fig. 3). This transition is primarily due to the promotion from the 4a ₂ (π) SHOMO to the 7b ₁(π*) LUMO.

Band C with maximum at 34,600 cm⁻¹ (289 nm) is purely long-axis (y) polarized and is easily assigned to the 2¹B₂ state computed at 34,600 cm⁻¹ (Table 1, Fig. 3). The following absorbance band is resolved into two differently polarized components, D and E at 37,600 and 39,200 cm⁻¹ (266 and 255 nm). They can be assigned to the 3¹A₁ and 3¹B₂ states computed at 38,600 and 40,900 cm⁻¹ (Table 1, Fig. 3). Band F at 42,900 cm⁻¹ (233 nm) is predominantly short-axis (z) polarized and is assigned to the 5¹A₁ state predicted at 43,200 cm⁻¹ (Table 1, Fig. 3). The band overlaps the onset of long-axis polarized (y) intensity G around 44,000 cm⁻¹ (∼227 nm). This feature is possibly due to the 4¹B₂ state computed at 43,500 cm⁻¹ (Table 1, Fig. 3), but it may also involve b ₂ symmetric vibrational components of band F, gaining long-axis polarized intensity by vibronic coupling with the strong band H. In the isotropic spectrum recorded in n-heptane solution, transition F apparently corresponds to the shoulder close to 43,500 cm⁻¹ and transition G to the peak at 44,300 cm⁻¹ (Fig. S1).

Assignment of individual transitions in the high-wavenumber region is complicated by the presence of broad, overlapping bands and by a high density of electronic states. Our TD–CAM-B3LYP calculation predicts about 50 electronic transitions between 60,000 and 45,000 cm⁻¹ (Data S1). In addition, the applied theoretical model may be less accurate in this region and the suggested assignments of observed features to computed transitions are necessarily tentative.

The observed long-axis (y) polarized absorbance in this region displays a strong band H with maximum at 46,100 cm⁻¹ (217 nm). According to the calculated transitions, this band can be assigned to the 5¹B₂ state predicted at 47,400 cm⁻¹ (Table 1, Fig. 3). Band H has a long tail into the vacuum UV with an additional peak L at 54,100 cm⁻¹ (185 nm). Several states are predicted to contribute to this absorption, primarily 8¹B₂ and 9¹B₂ computed at 52,500 and 56,300 cm⁻¹ (Table 1, Fig. 3).

The short-axis (z) polarized intensity in this region is dominated by a very strong band K with maximum at 53,400 cm⁻¹ (187 nm) in the vacuum UV. This is the strongest absorbance observed in the investigated spectral range. The peak K at 53,400 cm⁻¹ is preceded by two diffuse shoulders I and J at 47,000 and 51,000 cm⁻¹ (213 and 196 nm). Several states are predicted to contribute to this band system, such as 7¹A₁, 8¹A₁, 9¹A₁, and 10¹A₁, with transition to the 9¹A₁ state computed at 52,800 cm⁻¹ as the most intense (Table 1, Fig. 3).

Excited state intramolecular proton transfer (ESIPT)

The observed fluorescence emission of A in n-hexane solution has a maximum at 17,500 cm⁻¹ (570 nm) with an unusually large Stokes shift of 10,500 cm⁻¹ relative to the excitation observed at 28,000 cm⁻¹ (360 nm) (Møller et al., 1998). The fluorescence quantum yield is nearly independent of temperature which suggests that the decay from the initially formed excited state to the ground state does not involve barrier crossing (Møller et al., 1998).

According to the orbital amplitudes in Fig. 4, the promotions 6b ₁(π) → 7b ₁ (π*) and 4a ₂(π) → 7b ₁(π*) involved in the transitions to the two lowest, near-degenerate singlet states 2¹A₁ and 1¹B₂ are associated with substantial transfer of electron density from the phenolic moieties to the carbonyl group. This is expected to affect the balance of forces involved in the intramolecular hydrogen bonding, facilitating excited state intramolecular proton transfer (ESIPT) (Andersen & Spanget-Larsen, 1997). Geometry optimization of the excited S₁ state with TD-B3LYP predicts barrier-less transition to the ESIPT product 8,9-dihydroxy-1(10H)-anthracenone, see Fig. 5, leading to good agreement with the observed fluorescence and excitation wavenumbers (the predicted excitation wavenumber indicated in Fig. 5 is slightly different from the one listed in Table 1 because of different TD-DFT procedures). The present TD-DFT results thus support the assignment previously suggested on the basis of semi-empirical model calculations (Andersen & Spanget-Larsen, 1997; Møller et al., 1998). Excited state equilibrium nuclear coordinates and other computational data for the ESIPT product are provided as Data S2; relative energies of this and additional excited state configurations are outlined in Fig. S4.

A similar excited state rearrangement is observed for the closely related compound 1,8-dihydroxy-9,10-anthraquinone (chrysazin, an oxidation product of A) as reported by Smulevich and coworkers (Smulevich et al., 1987; Marzocchi et al., 1998) and recently studied theoretically by Mohammed et al. (2014) and by Zheng, Zhang & Zhao (2017).