Optimization of fluorinated phenyl azides as universal photocrosslinkers for semiconducting polymers

Abstract

Fluorinated phenyl azides (FPA) enable photo-structuring of π-conjugated polymer films for electronic device applications. Despite their potential, FPAs have faced limitations regarding their crosslinking efficiency, and more importantly, their impact on critical semiconductor properties, such as charge-carrier mobility. Here, we report that azide photolysis and photocrosslinking can achieve unity quantum efficiencies for specific FPAs. This suggests preferential nitrene insertion into unactivated C‒H bonds over benzazirine and ketenimine reactions, which we attribute to rapid interconversion between the initially formed hot states. Furthermore, we establish a structure‒activity relationship for carrier mobility quenching. The binding affinity of FPA crosslinker to polymer π-stacks governs its propensity for mobility quenching in both PM6 and PBDB-T used as model conjugated polymers. This binding affinity can be suppressed by FPA ring substitution, but varies in a non-trivial way with π-stack order. Utilizing the optimal FPA, photocrosslinking enables the fabrication of morphology-stabilized, acceptor-infiltrated donor polymer networks (that is, PBDB-T: ITIC and PM6: Y6) for solar cells. Our findings demonstrate the exceptional potential of the FPA photochemistry and offer a promising approach to address the challenges of modelling realistic molecular interactions in complex polymer morphologies, moving beyond the limitations of Flory‒Huggins mean field theory.

Fig. 1. Basic characterization of the FPA photocrosslinkers.

a Chemical structure. b Thermogravimetry characteristics. c Differential scanning calorimetry characteristics, color coded same as (b). d Solution-state absorption spectra; e Computed absorption spectra for vertical excitations at the optimized ground-state geometry (S_o). f Molecular model: blue, nitrogen; yellow, fluorine; red, oxygen; gray, carbon; white, hydrogen. g Molecular orbital wavefunctions for dominant one-electron excitation contributions to ES4 of FPA0, and to both ES3 and ES4 of FPA1, with corresponding coefficients given. h Dominant one-electron excitation contributions to ES1 of FPA0 and FPA1. The wavefunctions for FPA1 are also similar to those of other substituted FPAs in this work. PA1 is the non-fluorinated analogue of FPA1.

Fig. 2. Photolysis characteristics in polystyrene: FPA films.

a Plot of surviving azide fraction against radiant exposure at 254-nm wavelength, measured by FTIR spectrometry of the asymmetric azide stretching mode (ν_as N₃, ca. 2125 cm^–1), for selected FPAs in PS: PFA films with thickness d given (in nm) after sample name. b Plot of reflectance, transmittance, and absorptance, together with absorptance enhancement factor given by: A/(1 – exp(–αd)), where α is inverse absorption length (4π κ/λ = 2.0 × 10⁴ cm^–1), computed for parallel-beam illumination. The relevant optical constants at 254-nm-wavelength are: Si, 1.60 – 3.75i; and PS: FPA, 1.79 – 0.04i. Interference oscillations are damped by the elongated tube lamp illumination and film thickness variation, where the damped function (fitted black curve) follows: A + B *exp(–βd), where A is (1 – R) at large film thicknesses (0.9196), B is 0.695, and β is 2.4 × 10⁴ cm^–1. c Film retention characteristics for 200-kD polystyrene. Lines are fit to ζ = 1 – (C_c/C)ⁿ, where ζ is the ratio of film thickness after solvent wash to original film thickness, C is crosslinker concentration, C_c is the experimental gel point, and n is the characteristic exponent. The theoretical gel points for this PS are marked by C_c,th and C’_c,th at 3.2 and 1.6 × 10¹⁸ cm⁻³ for bisazide and tetrakisazide, respectively. A number concentration of 1.0 × 10¹⁹ cm^–3 corresponds to a weight concentration of about 0.77 w/w% for the FPAs here. Table gives ξ_XL,PS and n values, with estimated uncertainties (1 sd) of 0.05 and 0.15, respectively.

Fig. 3. Photoreaction product analysis in thin films.

FTIR spectroscopy for PS: FPA films before, and after partial or complete photolysis, together with selected authentic amino derivatives as reference. a FPA0, b FPA1, c PA1, d FPA8a, e FPA8b, f FPA6a, and g FPA6b. Vertical division, typically 25 milli-absorbance units (mAU); baseline correction error typically less than 1 mAU. The broad “bump” feature at 1120 cm^–1 is contributed by substrate SiO₂, while the sharp features at 2349 cm^–1 are due to imperfect correction for atmospheric CO₂, and those at 1400–1900 cm^–1 are due to atmospheric H₂O. The difference spectra were obtained for the same film, thus giving changes in the crosslinker and PS matrix, independent of the substrate. PA1 is the non-fluorinated analogue of FPA1.

Fig. 4. Computed internal energies.

a Reaction profiles for FPAs¹.N₃ is azide ground state¹, N₃* is lowest excited singlet azide (dark)¹, N₃^# is lowest optically-coupled excited singlet azide¹, N is open-shell singlet nitrene ground state¹, N* is lowest closed-shell singlet nitrene³, N is ground-state triplet nitrene, Bza is benzazirine, Kti is cyclic ketenimine. Transition state energies for ¹N → Bza → Kti are shown. For FPA8a and FPA6a, transition states are shown for the lower-energy product where the N atom bridges and then inserts at the C atom away from Me and i-Pr, respectively. The other product is a few tens of meV higher in energy. Energies are computed with unrestricted Kohn-Sham DFT/CAM-B3LYP/6-31 G(d). Levels represent zero-point internal energies at 298 K relative to ¹N₃, combining the internal energies of the products, i.e., nitrene + N₂, after photolysis.¹N₃* and ¹N₃^# are vertical energies from TD-DFT at the same level of theory in acetonitrile. b Proposed hot interconversions: ¹N ⇄ Bza ⇄ Kti, for slow cooling in solid matrices; and freeze-in for fast cooling in liquid media or in the presence of a high density of C‒H bonds on the FPA moiety. IC is internal conversion, and VR is vibrational relaxation.

Fig. 5. Photocrosslinking efficiency of π-stacked polymer semiconductor models.

a Chemical structures of PM6 and PBDB-T. Film retention characteristics of b PM6 (M_n 43 kD, Đ 2.3), and c PBDB-T (M_n 34 kD, Đ 2.1), with selected FPAs. Data are fitted to ζ = 1 – (C_c/C)ⁿ, following Fig. 2. The theoretical gel point (C_c,th) for these PM6 and PBDB-T samples are 1.7 and 2.1 × 10¹⁹ cm^–3, respectively, based on a polymer density of 1.2 g cm^–3.

Fig. 6. Plots of hole-carrier mobility of polymer against crosslinker concentration, evaluated in hole-only diodes: glass/ ITO/ PEDT: PSSH/ polymer: FPA/ Ag.

The polymer is a PM6 and b PBDB-T. PEDT: PSSH provides ohmic hole injection, while Ag provides ohmic hole exit. Carrier mobility is extracted in the high-field region where the space-charge-limited conduction regime holds, using the Mott–Gurney equation: $J=\frac{9}{8}\epsilon \mu \frac{{\left(V-V_{*}\right)}^2}{d^3}$, where ε is permittivity, d is thickness, V_* is the apparent built-in potential, and μ is mobility, averaged over three representative diodes. Insets: Mott–Gurney plots for selected FPA crosslinkers. Legend: open squares, 0.5; diamonds, 1; and triangles, 2 w/w% of crosslinker. The yellow region gives a span of mobility quenching.

Fig. 7. Binding affinity model to interpret FPA molecular segregation tendencies.

a Schematic cartoon for (top) tight sandwich binding, and (bottom) loose sandwich binding. Green ribbons represent the polymer backbone; linked orange disks represent the crosslinker. The dashed circle encloses a perturbed region. b Schematic cartoon for intercalation on transport density-of-states. Illustrative OPLS4 molecular models for c FPA0 in amorphous alkyl side-chain nanophase modeled as alkane cluster and d polymer backbone/FPA complexes. FPA is rendered in space-filling representation for clarity. For FPA0, the entire crosslinker is computed as a unit; for the other FPAs, the hemi-crosslinker is computed, since each moiety can quasi-independently bind to the polymer backbone or side-chain phase.

Fig. 8. Validation of structure‒activity relationship.

Plots of the experimental logarithm of mobility quenching against computed differential binding energy for tight sandwich (blue) and loose sandwich (red) binding models for a PM6 and b PBDB-T. Binding energy for tight sandwich in well-ordered π-stacks (ΔΔU_PP–A); and loose sandwich in less-ordered π-stacks (ΔΔU_P–A), relative to amorphous alkyl side-chain nanophase (definition and data in Supplementary Table 3). μ is hole mobility in crosslinked polymer film with 1 w/w% FPA; μ_o is hole mobility in pristine film.

Fig. 9. Comparative performance of solar cells with infiltrated-polymer-network or demixed-biblend photoactive layer (PAL).

Device structure: glass/ ITO/ PEDT: PSSH/ PAL/ BT-F3NMe3: Ox, OAc/ Ag, where PAL is: a PBDB-T: ITIC (1:1 w/w), and b PM6: Y6 (1:1.2 w/w). Crosslinker concentration, 0.4 w/w%, FPA0 for PBDB-T and FPA6a for PM6. JV curves are shown for four typical devices. UV–Vis–NIR spectra for the PAL films: c PBDB-T: ITIC, and d PM6: Y6. e Statistics summary of devices where “±” denotes one standard error of the mean. J_sc, short-circuit current density (mA cm^‒2); V_oc, open-circuit voltage (V); FF, fill factor; PCE, power conversion efficiency (%). f Thermal stability of infiltrated-network vs demixed-biblend PM6: Y6 cells. Data averaged over four diodes. The standard error is 0.5% for normalized J_sc, 0.002 V for V_oc, and 0.005 for FF.

Fig. 10. Energy levels of FPAs and organic semiconductors.

(left) Theoretical HOMO and LUMO levels of crosslinkers: FPA moiety (dotted line) and ideal crosslink (solid line), computed in gas phase by DFT/CAM-B3LYP/6-31 G, and corrected to condensed phase by 1.5-eV inward polarization shift, i.e., HOMO(s) = HOMO(g) + 1.5 eV, and LUMO(s) = LUMO(g) – 1.5 eV, with estimated uncertainty of 0.1 eV. (right) Measured HOMO and LUMO band edges of organic semiconductors, obtained by ultraviolet photoelectron spectroscopy for HOMO band edge and estimated from bandgap and exciton binding energy correction for LUMO band edge.