## Introduction

Strain engineering is an important method to modulate the physical properties in conventional inorganic semiconductor material. The strained silicon, germanium, GaAs, and newly developed 2D materials have been widely used and studied for high-performance electronic circuits1,2,3,4. Over the past decade, due to the potential application in flexible electronics, such as electronic skin and low-cost flexible displays5,6,7, the strain effect on organic semiconductors is also extensively studied, where impressive material performance improvements are observed8,9,10. It’s worth noticing that, due to differences in material structure, the strain effect shows fascinating distinctions in inorganic and organic semiconductors, which deserves further investigation. On the other hand, for photoresponsive molecules, if the photoisomerization group is incorporated, molecules’ rearranged packing can induce intrinsic strain, which has been utilized in light-responsive actuators11,12,13. It is natural to the hypothesis that the light-induced strain in organic semiconductors should also demonstrate light-strain modulation. To our knowledge, such a light-responsive self-strained organic semiconductor has not been proposed, although the such effect may have been unintentionally used in many previous photo-responsive devices. Here, we developed a strategy-directly grafting the photochromic groups onto the high-performance semiconductor molecule motif and studied the light-induced self-straining effect. Specifically, benzo[b]benzo[4,5]thieno[2,3-d]thiophene (BTBT) is employed as the semiconductor backbone due to its high intrinsic mobility14. The BTBT backbone is alkylated on one end with octane to improve molecular flexibility and solubility15, while the opposite end is covalently bound to azobenzene (AZO) to endow with reversible photoisomerization property16,17. These molecules with asymmetric modification can also obtain good molecular packing through simple post-processing, such as thermal annealing, thus improving performance15,18,19. The AZO group in the as-synthesized AZO-BTBT-8 undergoes a switch from trans to cis conformation upon ultraviolet (UV) irradiation and back to the thermodynamically stable trans conformation with visible-light (Vis) excitation or high-temperature treatment20,21. Due to the steric hindrance16,22,23,24, most photo-inducing folding occurs on the top thin layer, in contrast to a fraction of the molecules inside the OSCs film. The vertical layer difference will induce uniform lattice strain to the bulk semiconductors, thus positively feedback to long-range ordered crystalline25,26 and increasing the mobility. On this basis, a large-scale flexible OFET device array is fabricated utilizing AZO-BTBT-8 as the active semiconductor layer, exhibiting reversible light response and good stability under complex deformation, further indicating the AZO-BTBT-8 molecule performs an experimental foundation for molecular engineering and strategy optimization with specialized functionality.

## Results and discussion

### Characterizations of AZO-BTBT-8

As designed, the molecular structure, DFT-calculated molecular conformation, and frontier orbitals of AZO-BTBT-8 are shown in Fig. 1a. AZO-BTBT-8 mainly presents as planar trans conformation in the ambient environment, while the UV light triggered the isomerization to cis conformation with benzene rings tilted to each other. This photoisomerization slightly increases the highest occupied molecular orbital (HOMO), benefiting the energy level alignment with electrode Au (−5.1 eV) in the top-contact/bottom-gate device architecture (Fig. 1b).

All the experiments are conducted with spinning-coated film samples annealed at 80°C unless otherwise stated. The thermodynamic properties of the bulk material were first investigated. Supplementary Fig. 2a shows the breakdown temperature of 340°C, demonstrating strong thermal stability, which is a critical requirement for OFET electronics. According to the differential scanning calorimetry plots (DSC, Supplementary Fig. 2b), liquid crystals and Maltese crosses were revealed by polarized optical microscope (POM) in the incubation at 80°C (Supplementary Figs. 3 and  4). Thus, to optimize molecular packing in thin-film devices, an annealing temperature of 80 °C (30 min, Supplementary Fig. 5) was selected to help with the molecular rearrangement27,28. AZO-BTBT-8 thin film was further examined with atomic force microscope (AFM). As shown in Fig. 2a, the film of AZO-BTBT-8 shows good continuity and flatness with molecular step edges, which is suitable for electronic device fabrication. After exposure to UV light (365 nm, 20 mW∙cm-2) for two hours, the average step height of the surface monolayer decreased from 3.11 to 2.85 nm (Fig. 2b), which is consistent with the DFT calculated value (27.2 Å to 24.9 Å, difference of 2.3 Å, Supplementary Fig. 16) and indicates the complete photoisomerization on the film surface. The time-dependent UV-Vis experiment is then conducted to investigate the photoisomerization kinetics, showing the reversible photoisomerization of AZO-BTBT-8 in both solid film and solution (Supplementary Figs. 9a and 10). However, compared to the solution (Supplementary Fig. 9b), the thin films sample (800 nm, Fig. 2d) shows a much weaker response, which can be attributed to the steric hindrance in the condensed phase29. Due to the steric hindrance, a fraction of the molecules is triggered for photoisomerization inside the film under UV irradiation, in contrast to almost 100% on the surface. Therefore, the heterogeneous conformational change in the film is predictable, with higher conversion ratio towards the top surface (Fig. 2c). As a result, the thicker film shows lower overall conversion ratio, slower conversion kinetics and longer half-life time (t1/2), as observed experimentally in Fig. 2e and Supplementary Table 4. It is well-known that strain engineering is a general strategy applied in semiconductor materials to enhance device performance30,31. Previously, the strain-enhanced mobility in organic semiconductors has been achieved by mechanically shearing the growth solution during the crystallization, where a substantial mobility improvement is achieved8. In our thin-film device, since the top surface contributes to the majority of the photoisomerization while the bottom surface contributes the majority to the electrical conductivity, the top surface isomerization can induce uniform lattice strain to the bulk semiconductors (Fig. 2c). Here, the molecule configuration deep inside the film is analyzed with grazing incidence X-ray diffraction (GIXD), conventional X-ray diffraction (XRD), and the electric property is monitored with conductive AFM. Contrary to the high crystallinity on the film surface, from the GIXD, the film shows low-order diffraction ring (left of Fig. 2f), which indicates more irregular packing inside the bulk phase of the film. Upon UV irradiation, GIXD diffraction shifts into distinct Bragg rods (110) and (020), indicating improved packing in xy plane (upper right of Fig. 2f). Interestingly, after visible light irradiation, this pseudo crystalline phase can be switched back to lower-order phase with weak diffraction and a d-space expansion is observed, where (020) peak moves to higher fields with d-value increasing from 2.704 to 2.729 Å (lower right of Fig. 2f). It is well known that the strain can induce ordering and crystalline in amorphous materials, and the rich dynamics of glass composed of photoisomerization molecules have been discussed recently32. Since this reversible crystallinity change can not be explained by direct photoisomerization of molecules, we attribute this effect to the strain-induced crystallization, while the details mechanism requires further investigation. From the thin film XRD, the d-space shrinking is also observed, where the diffraction peaks at 21.05° shift to higher fields under UV irradiation (Fig. 2g and Supplementary Fig. 13), with d-value of the main peak decreasing correspondingly from 4.093 to 4.073 Å. The photoisomerization-induced strain can change the carrier mobility and material conductivity, similarly to previous mechanical-induced strain33. As shown in Fig. 2h and Supplementary Fig. 8, the conductive AFM scan shows a significant and instant current increase upon UV irradiation and a gradual current decay (K = 0.01587 min1, ca. 120 min falling with the baseline) at room temperature in the dark, which can be attributed to the thermal isomerization of cis to trans in the film and the strain-induced ordering, resulting in high carrier mobility25,34,35.

### Theoretical calculations

Moreover, it is calculated that cis is an unstable conformation excited from trans (Fig. 3a). However, compared with those with single conformation, the binding energy between cis and trans is much higher (Fig. 3b), resulting in regular molecular arrangement and alignment, which also explains that despite the limited concentration of cis, especially inside the film, it is still able to optimize the molecular packing and further stress on the deeper molecules. In addition, the average distance between the trans-trans conformations (3.129 Å) in the DFT calculation is larger than that between trans-cis conformations (3.104 Å), which indicates light-induced isomerization also synergistically optimizes molecular interaction with dense stacking and ultimately improves device performance (Supplementary Fig. 17). With displacement boundary conditions, we simplified the film section and calculated the strain and stress distributions (Fig. 3c and Supplementary Fig. 18). It was observed that after UV irradiation both the stress and strain accumulated along the x-axis and eventually released at the film edge. The stress gradually decreases with film depth and ultimately reaches the interface between dielectric and semiconductor, affecting the device output through molecular packing optimization. This is in good agreement with the above-proposed mechanism.

### Sensing tests by the OFET array

The as-prepared single OFET device exhibits good performance as described above, however, to implement true high availability, how to make large-scale flexible array is inevitable36,37. To realize it, we design a strategy by depositing patterned electrodes directly on a desired flexible substrate, as shown in Fig. 5a. Specifically, patterned AZO-BTBT-8, source electrodes, hafnium oxide dielectric layer, and drain electrodes were deposited sequentially on an Al2O3 covered ITO-PET. Thus, the precise device array layout was obtained (Fig. 5b). Figures 5c and d shows the optical image of a 33 × 40 device array, the magnified view of the 3 × 3 device array, and the SEM image of the individual device, respectively. The average mobility increased from 0.016 ± 0.005 cm2 V−1 s−1 to 0.152 ± 0.009 cm2 V1 s−1 under UV illumination with the on/off ratio of 105 (L = 50 and W = 1080 μm), which is consistent with the individual OFET device and demonstrates the reproducibility of the device performance. The optical pattern is obtained after top UV illumination to the array through a smiling cartoon sun garland mask for 20 minutes (Fig. 5e), where the exposed area switched from OFF state (~0.2 μA) to ON state (~2.8 μA) (Fig. 5f). Since the photoisomerization is reversible, the image can be erased with visible light irradiation and repatterned. As shown in Fig. 5g, h the previous pattern is erased with 30 min exposure of visible light and can be repatterned under the UV illumination with the diaglyph pattern “Hi! Chem!” and a happy emoji. All evidence suggests that the OFET array devices are reversible, programable, and scalable, with promising potential in other fields, like flexible sensors, displays, wearable devices, health care, and tissue detection.

In this work, a functional molecule AZO-BTBT-8 was designed and synthesized by integrating photochromic azobenzene, high-performance semiconductor backbone BTBT, and flexible alkyl chains. The light-induced self-strain engineering is observed in this material and leads to reversible mobility switching in solid-state devices. Based on this mechanism, a large-scale flexible OFET device was fabricated on a flexible substrate using AZO-BTBT-8 molecules as photoisomerization OSCs, showing good stability and reproducibility. In conclusion, this work establishes a new strategy for designing and developing light-responsive OSCs and corresponding functionalized OFETs.

## Methods

### Materials

All reagents and chemicals were obtained from commercial sources and used without further purification unless otherwise noted. All reactions were performed under an inert atmosphere of argon in dry solvents using standard Schlenk techniques. The synthetic route used to obtain linker AZO-BTBT-8 is outlined in Supplementary Fig. 1.

### Characterization

To ensure the reliability of the experiments, unless otherwise stated, all film spinning-coated samples are annealed at 80 °C for 30 min before being characterized and tested. The morphology of thin films was investigated by a JPK atomic force microscope (AFM) under ambient conditions in QI mode. The conductive AFM was conducted in contact conductive module. Film and powder X–ray diffraction data were collected on PANalytical high resolution PXRD. GIXD data were obtained at beamline BL14B1 of the SSRF at a wavelength of 1.2398 Å.

### Device fabrication and measurement

Heavily doped n-type silicon wafers were cleaned in a Piranha solution (volume ratio of components H2SO4/H2O2 = 70:30) by heating at 110 °C for 2 h followed by rinsing thoroughly with de-ionized (DI) water, sonicated for 15 min in an RCA solution (volume ratio of components DI water/ammonium hydroxide/H2O2 = 5:1:1), rinsed and dried under nitrogen, and used immediately. 200 nm Al2O3 thin film was deposited on the clean silicon wafer by atomic layer deposition (ALD) at 200 °C for 1850 cycles. A cleaned ITO-coated PET substrate (Sigma-Aldrich, 1 ft × 1 ft × 5 mil) was used as the flexible substrate. 200 nm Al2O3 thin film was deposited on PET by ALD at 70 °C for 2000 cycles. The device was fabricated by the spin-coating of a solution of AZO-BTBT-8 in 10 mg/mL CHCl3 (annealed at 80 °C for 30 min) and thermal evaporation of Au through a designed mask. The device array on PET is prepared by continuous evaporation of AZO-BTBT-8, Au and HfO2 insulating layers by a set of aligned shadow masks. The transistor characteristics were obtained at the room temperature in air by a standard probe station and two semiconducting parameter analyzers (Keithley 2400). The mobilities of the devices were calculated in the saturation regime by the standard method:

$${I}_{{DS}}=(W/2L){C}_{i}u{({V}_{G}-{V}_{T})}^{2}$$
(1)

Where W/L is the channel width/length, and VG and VT are the gate voltage and threshold voltage, respectively. Ci is the insulator capacitance per unit area.

### DFT and mechanical calculations

#### DFT calculation

All geometric structure calculation has been carried out using Gaussian 09 package and Gauss view molecular visualizing program package which has provide itself to be extremely useful to get a clear knowledge of optimized parameters, electronic structure properties and other molecular properties. The geometry is fully optimized at Beck3-Lee-Yang-Parr (B3LYP) [1,2] level with standard 6-311 + G (d, p) basis set 38,39.

#### Mechanical analysis

The mechanical analysis model uses the displacement boundary condition, and it is assumed that all molecules in the film are stacked along the x-dimension, that is, all film deformations occur in the x- dimension. The aspect ratio of the film section is simplified to 8000:800. The Poisson’s ratio and elastic modulus are estimated to be 0.35 and 7 GPa, respectively, according to the literature40,41. The strong form in the proposed thermodynamically consistent phase field model can be written as follows42,43:

$$\nabla \cdot {{{{{\boldsymbol{\sigma }}}}}}({{{{{\boldsymbol{\varepsilon }}}}}},{d})+{{{{{\bf{b}}}}}}=0\;{{{{{\rm{in}}}}}}\;\Omega$$
(2)
$${{{{{\boldsymbol{\sigma }}}}}}({{{{{\boldsymbol{\varepsilon }}}}}},d)\cdot {{{{{\bf{n}}}}}}=\bar{{{{{{\bf{t}}}}}}}\;{{{{{\rm{on}}}}}}\;\partial {\Omega }_{t}$$
(3)
$$u=\bar{u}\;{{{{{\rm{on}}}}}}\;\partial {\Omega }_{{{{{{\rm{u}}}}}}}$$
(4)
$$\left(1-d\right)\left(\frac{{H}_{n}}{{G}_{{Ic}}}+\frac{{H}_{t}}{{G}_{{IIc}}}\right)\eta+\frac{{l}_{c}}{2}{\nabla }^{2}d-\frac{d}{{2}{l}_{c}}=0\;{{{{{\rm{in}}}}}}\;\Omega$$
(5)
$$\nabla d\cdot{{{{{\bf{n}}}}}}=0\;{{{{{\rm{on}}}}}}\;\partial \, \Omega \cup \Gamma$$
(6)

where b is the body force vector, $$\bar{{{{{{\bf{t}}}}}}}$$ is the traction vector, u is the displacement field, d is the phase field, σ is the Cauchy stress tensor, ε is the strain tensor, GIc and GIIc are the critical fracture energy release rates for mode-I and mode-II fracture modes, respectively.