Evolution of the electronic structure in open-shell donor-acceptor organic semiconductors

Most organic semiconductors have closed-shell electronic structures, however, studies have revealed open-shell character emanating from design paradigms such as narrowing the bandgap and controlling the quinoidal-aromatic resonance of the π-system. A fundamental challenge is understanding and identifying the molecular and electronic basis for the transition from a closed- to open-shell electronic structure and connecting the physicochemical properties with (opto)electronic functionality. Here, we report donor-acceptor organic semiconductors comprised of diketopyrrolopyrrole and naphthobisthiadiazole acceptors and various electron-rich donors commonly utilized in constructing high-performance organic semiconductors. Nuclear magnetic resonance, electron spin resonance, magnetic susceptibility measurements, single-crystal X-ray studies, and computational investigations connect the bandgap, π-extension, structural, and electronic features with the emergence of various degrees of diradical character. This work systematically demonstrates the widespread diradical character in the classical donor-acceptor organic semiconductors and provides distinctive insights into their ground state structure-property relationship.

Similarly the authors comment on the long absorption tail being characteristic of diradical character, but given that the measurements are not correctly for reflection and scattering, I don't think they can assert this. I strongly suggest they provide film measurements which have been corrected for these factors.
Why is all EPR data in the solid state only? What happens in solution measurements? Is diradical character observed? If not, why not. This should be addressed in the manuscript.
Given the important of purity in these sensitive measurements, I find it surprising that no information on molecular purity is given except for NMR. Many of the mass spectra show peaks at different mass to that of the molecular ion. Given that all of these materials were made by Pd coupling, can the presence of homo-coupled defects or varying Pd levels be ruled out? Some information on purity (for example HPLC trace, melting points, elemental analysis) should be provided.
No info is given on how 2N-TDPP and 2N-NTT were sublimed. observed, largely matching that from polymer examples, in which lower band gap materials show larger diradical character. I think this is an interesting finding, and the paper will certainly generate interest in the community. I believe it is suitable for publication in this journal, but there are some issues to address first.

Response:
We are encouraged by the reviewer's positive assessment and the meaningful comments to improve our paper. We have carefully revised the manuscript based on the suggestions from the reviewer. The detailed point-to-point response to each comment and the modifications of manuscript are listed as below.
Comment 1: Given the large absorption tails, how is the absorption onset defined in the measurements? I would suggest they use a more precise method than currently used, for example the intersection of absorption and emission, given the importance of the band gap to their discussion.
Response: Thank for the helpful comment. In our earlier version of the manuscript, we determined the absorption onset by the intersection of the linear fit trendline of the absorption spectrum and the tangent of absorption tail (Adv. Mater. 2007, 19, 173-191). This approach has been commonly used in numerous papers (citation). However, as the reviewer commented, this approach is not very precise, especially when there is no strict linear region in the absorption edge or when the light scattering is very significant for the absorption tail, which is the often case for the spin-coated organic films. Furthermore, open-shell radical molecules always show large absorption tail and lowest-energy absorption band originated from the presence of low-lying singlet excited state dominated by a doubly excited electronic configuration (H,H→L,L). (J. Phys. Chem. Lett. 2010, 1, 3334-3339) Therefore, the determination of absorbance onset is very subjective. To further give a declaration of the optical energy gap, we measured the absorption and the fluorescence emission spectra ( Figure R1 and Figure S3) of all the samples, and presented the normalized curves together on the same abscissa axis. (Adv. Energy Mater. 2018, 8, 1801352) The optical bandgap is experimentally estimated as the overlap of absorption and fluorescence (zero-phonon line). The corresponding vertical transition from the initial ground state (lowest excited state) to most probable excited state (ground state) forms the maximum absorption Emax,abs (luminescence Emax,pl). The relaxation occurs to excited state (ground state) form the relaxed energy λabs (luminescence λfl). The corresponding transition energy was defined as: Emax,abs=E0-0 + λabs Emax,fl=E0-0 -λfl in which E0-0 is defined as the optical gap of the material. When the absorption and emission curves in the overlapped range are nearly symmetric, E0-0 (or Eg) can be defined as the energy at the intersection of normalized absorbance and emission spectra. This method to determine Eg are reproducible and physically credible. The updated data was summarized in Table R1. (Figure 2, Figure 3c, Table S1 in revised manuscript and supplementary information) Figure R1. The photoluminescence spectra of (e) DPP-based small molecules with fused-phenyl groups, (f) TDPP, Th-TDPP, Flu-TDPP, TPAOMe-TDPP and TPAOMe-TTDPP, (g) NT-based small molecules, and (h) TPAOMe-based small molecules in film. Comment 2: Similarly the authors comment on the long absorption tail being characteristic of diradical character, but given that the measurements are not correctly for reflection and scattering, I don't think they can assert this. I strongly suggest they provide film measurements which have been corrected for these factors.

Response:
We have defined the optical gaps as the intersection of normalized absorbance and emission curves. It can be more precise and therefore can eliminate the influence of reflection and scattering. Please see Table R1. Response: Solution EPR spectra are very sensitive to molecular motion. Some conditions such as solvent viscosity, temperature, concentration can have a profound influence on the EPR spectra given their influence on molecular dynamics. (Chem. Soc. Rev., 2018, 47, 2534-2553 Therefore, it was complicated to give a comparation between the solution samples. We have also complemented the EPR spectra of solution samples. It can be observed that the EPR response of the materials greatly weakened from powders to solutions (see the EPR spectra of 2N-NTC, NTC, TPAOMe-TDPP, TPAOMe-TTDPP in powders and in toluene solutions, Figure R1). Some materials with weak solid-EPR response exhibited nearly invisible signal in solutions. We summarized the reasons of the greatly reduced solution-EPR intensity and are listed as followed.
(1) The solvent will absorb the emitted microwave energy, which leading to a significant reduction of EPR signal in solution state. This phenomenon is more severe in polar solvents such as chloroform, chlorobenzene, and tetrahydrofuran (Chem. Soc. Rev., 2005, 34, 164-178 Comment 4: Given the important of purity in these sensitive measurements, I find it surprising that no information on molecular purity is given except for NMR. Many of the mass spectra show peaks at different mass to that of the molecular ion. Given that all of these materials were made by Pd coupling, can the presence of homo-coupled defects or varying Pd levels be ruled out? Some information on purity (for example HPLC trace, melting points, elemental analysis) should be provided.

Response:
We thank the reviewer for raising this concern. We provided more methods and spectra information for purity analysis. We conducted HPLC analysis with Cosmosil 5C18-MS-II column. A single sharp peak can be observed for each molecule (as shown in Figure R2), which demonstrated the purity for these molecules. According to the DSC analysis, we detected a sharp and strong endothermic peak in the heating stage for some samples. It is a typical fingerprint characteristic assigned to the melting points. We mention that some samples such as Flu-NTC did not exhibit any endothermic peaks due to the weak crystallinity. Accordingly, we did not detect a phase transition from melting point determination. It may due to the increase of flexible components including the multiple long alkyl chains (-C12H25) and the dimethyls on fluorenyls. The melting points of the materials are summarized in Table R2. Furthermore, the element composition (C, H, O, S, N) of the carefully purified and dried samples obtained through elemental analysis are nearly consistent with the calculated data (Table R3), further giving an evidence of the purity information. We are sorry for the unsatisfied mass spectra in the manuscript. The updated spectra are shown in Figure R3-R18. The amount of metallic impurities of the samples are measured by inductively coupled plasma mass spectrometry (ICP-MS) and particle-induced X-ray emission (PIXE). We found that the metallic impurities are lowered than 20 ppm. This amount of impurities is far from resulting the observed phenomenon in this manuscript. Therefore, we can eliminate the influence of Pd impurities.   Table R3. Element components of the materials.

Response:
We are sorry for the missing of the detailed methods of sublimation. We have complemented the related descriptions in SI. The high vacuum sublimation was conducted at high vacuum of 0.01 Pa and temperatures above 300 ℃ for 72 hours. Nitrogen was injected into the tube after sublimation and subsequently, ESR tests were performed on sublimated solids. All the operations were performed in an inert environment and the isolation of metal iron (to eliminate the influence of external ferromagnetism).

The replacement of SQUID data in Figure 4c and 4d
In the resubmitted manuscript, we have replaced the SQUID data in Figure 4c and 4d. The data in the initial submitted manuscript was wrongly collected because of the operational problems in measurement. The centering of the SQUID appears off, which results in a weak signal and relatively low singlet-triplet energy gaps. We carefully rerun the measurement to ensure the accuracy and authenticity of data. dependent on the computational methods utilized (Phys. Chem. Chem. Phys. 2018, 20, 24227-38.). Using the spin-projected unrestricted Hartree-Fock (PUHF) method we find that all the materials possess variable open-shell character (Table S2), which correlates with their bandgap. As the PUHF method prone to a large spin contamination, we have tested the widely accepted broken symmetry (BS) approach with different density functionals. However, unlike the other open-shell small materials (Nat. Chem. 2016, 8, 753-9;Nat. Chem. 2018, 10, 1134 Ed. 2015, 54, 12308-13.). Fractional orbital density (FOD) is an extremely simple and cost-effective method based on smearing the electrons over the molecular orbitals using finite temperature DFT (FT-DFT). NFOD accurately quantifies the static electron correlation and molecules with a delocalized FOD and a large NFOD have multireference character. Table R5 shows the NFOD values of the molecules studied in this work. A large NFOD values reveal that the electrons are strongly correlated, and different materials provide different NFOD values. Interestingly, the trends are qualitatively consistent with diradical index computed from PUHF (see Table R4), showing good linear correlation between NFOD and y0.
To better understand the radical nature for these molecules, we explore their open-shell characters (yi) using fractional orbital occupancy. Table R6 shows the y0, y1, y2 and y3 values for all the molecules computed using the FT-DFT at B3LYP/6-31G(d,p) level of theory and basis set. Our result reveals polyradical character in all the molecules. The spatial distribution of unpaired electrons in these molecules are evaluated using FOD plots (see Fig. R4-R6). FOD plots show partially delocalized/localized electron density distribution along the molecular backbones, disclosing strongly correlated electrons.
We have also computed the vertical singlet-triplet energy gap (EST) using FT-DFT with B3LYP/6-31G(d,p) and the results are presented in Table R5. It was shown that the EST gap computed using FT-DFT is comparable to that of CASPT2 method (Phys. Chem. Chem. Phys. 2018, 20, 7112-24.).
The computed EST gap for 2N-NTC (11.03 kcal/mol) and TPAOMe-TTDPP (10.65 kcal/mol) is overestimated compared to the experimental gap, 4.76 and 5.52 kcal/mol, respectively. We believe this is due to the medium effects that we are not able to capture in the isolated molecule calculations.
However, we find a good correlation between y0 (and NFOD) and EST gap using FT-DFT method, a larger NFOD value indicates a smaller EST gap.      However, the results and discussion was not impressive. The design of donor-acceptor organic semiconductor was already carried out for the proper ICT to apply the proper application. The only interesting point is planar conjugated structure has ESR spectra. But the result was also already reported as described by author. In addition, this paper did not show any electronic application by using this phenomena. Therefore, I do not recommend this paper to publish in nature communications.
Reviewer #2 (Remarks to the Author): The authors have addressed many of the concerns raised, and I believe the manuscript is indeed suitable for this journal. In particular the additional data regarding band gap and purity helps to partially eliminate that an impurity may be the cause of the effect, but given the sensitivity of EPR this is still a concern. One important issue still not addressed is the spin density in the sample. Whilst I don't expect that this be measured for every sample, for those materials showing the strongest apparent EPR signals it is important that the number be quantified. Is it significantly less than 1% per molecule, or higher? The spin density is important to further exclude impurities (which might not show up on HPLC or NMR, where the limit of detection is higher than for PER). This is relatively simple to measure versus a spin standard, and I think it is important to include. I would support publication once included.

Minor points:
This sentence does not make sense to me: 'Occasionally, in 2018 we found that Professor 88 Fred Wudl has also reported the similar results in narrow bandgap D-A molecules and copolymers based 89 on benzo[1,2-c;4,5-c]bis[1,2,5]thiadiazole (BBT) in 2015 (Fig. 1c) and we apologized that we ignored 90 this work and didn't cite it.31,32' It reads like they apologised in reference 31 or 32; but I think they mean they apologize here? It shouldn't start with 'occasionally' either. Needs to be rephrased.

List of point-to-point response of reviewers' comments
Reviewer 1  However, the results and disicussion was not impressive. The design of donor-acceptor organic semiconductor was already carried out for the proper ICT to apply the proper application. The only interesting pont is planar conjugated structure has ESR spectra.
But the result was also already reported as described by author. In addition, this paper did not show any electronic application by using this phenomena. Therefore, I do not recommend this paper to publish in nature communications.
Response: Thank you very much for your careful review and comments on the manuscript.

1) For the novelty of this work:
As noted by the reviewer, the interesting point is the planar conjugated structure has ESR spectra and the result was also already reported as described by us in the reference 31 (Yuan Li, et al., J. Phys. Chem. C. 2017, 121, 8579-8588 Lett. 2008, 93, 093303;Chem. Commun. 2015, 51, 2239-2241Adv. Mater. 2018, 30, e1705052;J. Org. Chem. 2018, 83, 3651-3656). It is an extremely challenging job to disclose the origin of the ESR signal with various precise experimental technologies and theoretical methods.
The fundamental understanding of the open-shell diradical character of the D-A organic semiconductors will further promote the rational design of these materials.
In  Mater. Chem. C, 2016, 4, 11427-11435) In this work, we did not focus on applications as we did not have adequate evidence to clearly present underlying mechanism and structure-properties-performance relationship. We are trying the diradical compounds as charge transport layer in OFETs, emission layer in OLEDs, as singlet fission candidates in OPVs, photothermal and thermoelectric areas. These projects are currently under way in our laboratory.
Reviewer #2 (Remarks to the Author): The authors have addressed many of the concerns raised, and I believe the manuscript is indeed suitable for this journal. In particular the additional data regarding band gap and purity helps to partially eliminate that an impurity may be the cause of the effect, but given the sensitivity of EPR this is still a concern. One important issue still not addressed is the spin density in the sample. Whilst I don't expect that this be measured for every sample, for those materials showing the strongest apparent EPR signals it is important that the number be quantified. Is it significantly less than 1% per molecule, or higher? The spin density is important to further exclude impurities (which might not show up on HPLC or NMR, where the limit of detection is higher than for PER). This is relatively simple to measure versus a spin standard, and I think it is important to include. I would support publication once included.
Response: Thank you so much for your affirmation and the insightful interpretations.
We sincerely express our heartiest thanks to you because your key and important comments guided us to study our results more deeply and factually improve the manuscript quality. We will response to all the comments point by point carefully.
You raised the concern of spin density of the D-A compounds. The spin concentration correlated with the ESR intensity is an important concern for researchers in the field of organic radical materials. Using the monoradical DPPH (S=1/2) as the standard, the spin concentration (spin density) of TPAOMe-TTDPP and 2N-NTC were carefully tested and calculated as 0.15 % and 0.17 % NA, respectively (Fig. R1). This low triplet spin concentration is expected as the variable temperature and susceptibility measurements confirmed the singlet ground state electronic structure with large singlettriplet energy splitting of these compounds (EST of TPAOMe-TTDPP and 2N-NTC is -5.52 kcal/mol and -4.76 kcal/mol), indicating the weak contribution of paramagnetic triplet in ground state. It must be noted that the well-known diradicaloid indenoindenodibenzothiophene exhibited silent ESR response at room temperature, in consistent with its large EST of -8.8 kcal/mol (Michael M. Haley et al., Nat. Chem. 2018, 10, 1134-1140. The monoradical DPPH showed weak spin-spin interaction and strong paramagnetic property. When an external magnetic field (H) is applied in a direction, the electron's magnetic moment aligns itself either parallel (ms = −1/2, β spin) or antiparallel (ms = +1/2, α spin) to the field (Fig. R2), contributing to its high ESR intensity. Diradicals To study the spin density and confirm the intrinsic diradical character of the D-A-D compounds in our manuscript, we synthesized a well-known stable p-quinodimethane   TPAOMe-TTDPP is reasonable and acceptable.
To further demonstrate the rationality of low spin concentration, we complemented some discussions of the Chichibabin's diradical molecules. Chichibabin's hydrocarbon and its analogues were reported in the beginning of last century, and were widely recognized having open-shell diradical character in their ground state (Fig. R6).
However, the intensity of the ESR intensity of these typical diradical molecules are not always very strong depending on the interaction and coupling effect of the two unpaired radicals. Diradical materials with distorted conformation exhibited weak intramolecular spin-spin interaction because of the remarkable steric protection effect. In this case, the diradical is more likely to behave as two monoradicals, thus significantly enhancing the    Chem. Soc., 2012, 134, 14913−14922;Michael M. Haley et al., Nat. Chem. 2018, 10, 1134-1140 Besides, in the valence bond description, the spin-spin interaction is not only represented by the resonance within a molecule, but also by the intermolecular interaction, as shown schematically in Fig. R8. Intermolecular interaction between the spins will greatly affect the spin concentration. Materials with planar molecular conformation and small - distance feature a strong intermolecular covalent character, and thus exhibited weak ESR intensity (Fig. R8). in 2015 (Fig. 1c) and we apologized that we ignored 90 this work and didn't cite it.31,32' It reads like they apologised in reference 31 or 32; but I think they mean they apologize here? It shouldn't start with 'occasionally' either. Needs to be rephrased.

Response:
We are sorry for the mistake. Thank you very much for your comment. We have changed this sentence as follow.
After we published our work, we found that Prof. Wudl et al. has reported the open-shell character of the narrow bandgap D-A small molecules and polymers based on benzo[1,2-c;4,5-c]bis[1,2,5]thiadiazole (BBT) in 2015 (Fig. 1c), 32 and we regret this oversight on our part. 31