Molecular engineering of tetraazapentacene with different numbers of fluorine and chlorine substituents fine-tunes the frontier molecular orbitals, molecular vibrations, and π-π stacking for n-type organic semiconductors. Among the six halogenated tetraazapentacenes studied herein, the tetrachloro derivative (4Cl-TAP) in solution-processed thin-film transistors exhibits electron mobility of 14.9 ± 4.9 cm V s with a maximum value of 27.8 cm V s , which sets a new record for n-channel organic field-effect transistors. Computational studies on the basis of crystal structures shed light on the structure-property relationships for organic semiconductors. First, chlorine substituents slightly decrease the reorganization energy of the tetraazapentacene whereas fluorine substituents increase the reorganization energy as a result of fine-tuning molecular vibrations. Second, the electron transfer integral is very sensitive to subtle changes in the 2D π-stacking with brickwork arrangement. The unprecedentedly high electron mobility of 4Cl-TAP is attributed to the reduced reorganization energy and enhanced electron transfer integral as a result of modification of tetraazapentacene with four chlorine substituents.

Herein we report the synthesis, crystal structures, and semiconductor properties of new derivatives of bisnaphtho­[2′,3′:3,4]­cyclobut­[1,2-b:1′,2′-i]­anthracene (BNCBA). It is found that the π–π stacking of BNCBA in single crystals can be largely modified by alkyl substituting groups of different lengths. In particular, the tetrahexyl derivative exhibits π–π stacking with an unusual zigzag arrangement. The variation of molecular packing also leads to a change in charge transport characteristics as found from the theoretical calculation of mobility on the basis of single-crystal structures. All of these BNCBA derivatives function as p-type semiconductors in solution-processed thin film transistors, and the tetrahexyl derivative exhibits a field effect mobility as high as 2.9 cm2/(V s).

A theoretical study was carried out to investigate the electronic structures and the charge transport properties of a series of naphthodithiophene diimide (NDTI) thiophene α-substituted derivatives NDTI-X using density functional theory and classical Marcus charge transfer theory. This study deeply revealed the structure-property relationships by analyzing the intermolecular interactions in crystal structures of C8-NDTI and C8-NDTI-Cl thoroughly by using the Hirshfeld surface, QTAIM theories and symmetry-adapted perturbation theory (SAPT). Our results suggested that a 2-D brick-like π-stacking structure makes C8-NDTI-Cl a more excellent n-type semiconducting material with μ of 2.554 cm V s than C8-NDTI with a herringbone-like slipped π-stacking motif. In addition, the calculated results showed that by modifying the thiophene α-positions of NDTI with electron-withdrawing substituents, -F, -Cl and -CN, low-lying LUMO energy levels and a high adiabatic electron affinity EA(a) can be obtained; while introducing electron-donating groups, benzene (-B), thiophene (-T), benzo[b]thiophene (-BT) and naphtha[2,3-b]thiophene (-NT), expanded the molecular π-conjugated backbone, and narrow band gaps, high EA(a) and small reorganization energies can be obtained. Theoretical simulations predict that NDTI-CN is an excellent air-stable n-type organic semiconducting material with an average electron mobility μ of up to 1.743 cm V s. Owing to their high EA(a), moderate adiabatic ionization potential IP(a) as well as small hole and electron reorganization energies, NDTI-BT and NDTI-NT are two well-balanced air-stable ambipolar semiconducting materials. The theoretical average hole/electron mobilities are as high as 2.708/3.739 cm V s for C8-NDTI-NT and 1.597/2.350 cm V s for C8-NDTI-BT, respectively.

No abstract

2,5-Difluoro-7,7,8,8-tetracyanoquinodimethane (F-TCNQ) was recently reported to display excellent electron transport properties in single crystal field-effect transistors (FETs). Its carrier mobility can reach 25 cm V s in devices. However, its counterparts TCNQ and F-TCNQ (tetrafluoro-7,7,8,8-tetracyanoquinodimethane) do not exhibit the same highly efficient behavior. To better understand this significant difference in charge carrier mobility, a multiscale approach combining semiclassical Marcus hopping theory, a quantum nuclear enabled hopping model and molecular dynamics simulations was performed to assess the electron mobilities of the F-TCNQ (n = 0, 2, 4) systems in this work. The results indicated that the outstanding electron transport behavior of F-TCNQ arises from its effective 3D charge carrier percolation network due to its special packing motif and the nuclear tunneling effect. Moreover, the poor transport properties of TCNQ and F-TCNQ stem from their invalid packing and strong thermal disorder. It was found that Marcus theory underestimated the mobilities for all the systems, while the quantum model with the nuclear tunneling effect provided reasonable results compared to experiments. Moreover, the band-like transport behavior of F-TCNQ was well described by the quantum nuclear enabled hopping model. In addition, quantum theory of atoms in molecules (QTAIM) analysis and symmetry-adapted perturbation theory (SAPT) were used to characterize the intermolecular interactions in TCNQ, F-TCNQ and F-TCNQ crystals. A primary understanding of various noncovalent interaction responses for crystal formation is crucial to understand the structure-property relationships in organic molecular materials.

The charge transport properties of a series of rubrene derivatives were systematically investigated by density functional theory and molecular dynamics (MD) simulations. It was found that functionalizing electron-withdrawing groups (−CN, −CF3, or fluorination) on the peripheral phenyls not only enhance the chemical stability of materials but also favor electron injection by lowering the energy levels of frontier molecular orbitals and increasing the electron affinities. Derivatives 2–5 and 9, exhibiting packing motifs similar to rubrene but closer π-stacking distances, possess large hole and electron-transfer integrals, significant bandwidths, and small effective masses, suggesting excellent ambipolar semiconductor behavior. The maximum hole­(electron) mobilities in the Marcus hopping mechanism based on kinetic Monte Carlo simulation can reach 14.0–16.5(1.6–3.5) cm2 V–1 s–1. Interestingly, the antiparallel 2-D brick stacking and twisted backbones of fluorinated derivatives 11 and 12 result in nearly 1-D percolation network but balanced hole and electron transport property. In contrast, the parallel 2-D brick stacking of 14 leads to 2-D percolation network. Their maximum hole and electron mobilities fall in the range of 0.5–3.6 and 2.0–4.8 cm2 V–1 s–1. Furthermore, MD simulations show that dynamic disorder is strongly detrimental to the hole transfer but has a little influence on the electron transfer for 1–5. Moreover, severe twist of backbones of 9 leads to almost 1 order of magnitude lowered mobility. In addition, the influences of different substituents on the molecular structure, packing motif, and intermolecular reorganization energy are discussed.

To obtain anthracene-based derivatives with electron transport behavior, two series of anthracene-based derivatives modified by trifluoromethyl groups (−CF3) and cyano groups (−CN) at the 9,10-positions of the anthracene core were studied. Their electronic structures and crystal packings were also analyzed and compared. The charge-carrier mobilities were evaluated by quantum nuclear tunneling theory based on the incoherent charge-hopping model. Our results suggest that introducing −CN groups at 9,10-positions of the anthracene core is more favorable than introducing −CF3 to maintain great planar rigidity of the anthracene skeleton, decreasing more lowest unoccupied molecular orbital energy levels (0.45–0.55 eV), reducing reorganization energies, and especially forming a tight packing motif. Eventually, the excellent electron transport materials could be obtained. The molecule 1-B in Series 1 containing −CF3 groups is an ambipolar organic semiconductor (OSC) material with a 2D transport network, and its value of μh‑max/μe‑max is 1.75/0.47 cm2 V–1 s–1 along different directions; 2-A and 2-C in Series 2 with −CN groups are excellent n-type OSC candidates with the maximum intrinsic mobilities of 3.74 and 2.69 cm2 V–1 s–1 along the π–π stacking direction, respectively. Besides, the Hirshfeld surface and quantum theory of atoms in molecules analyses were applied to reveal the relationship between noncovalent interactions and crystal stacking.

Deep learning methods have got fantastic performance on lots of large-scale datasets for machine learning tasks, such as visual recognition and neural language processing. Most of the progress on deep learning in recent years lied on supervised learning, for which the whole dataset with respect to a specific task should be well-prepared before training. However, in the real-world scenario, the labeled data associated with the assigned classes are always gathered incrementally over time, since it is cumbersome work to collect and annotate the training data manually. This suggests the manner of sequentially training on a series of datasets with gradually added training samples belonging to new classes, which is called incremental learning. In this paper, we proposed an effective incremental training method based on learning automata for deep neural networks. The main thought is to train a deep model with dynamic connections which can be either ''activated'' or ''deactivated'' on different datasets of the incremental training stages. Our proposed method can relieve the destruction of old features while learning new features for the newly added training samples, which can lead to better training performance on the incremental learning stage. The experiments on MNIST and CIFAR-100 demonstrated that our method can be implemented for deep neural models in a long sequence of incremental training stages and can achieve superior performance than training from scratch and the fine-tuning method.INDEX TERMS Supervised learning, incremental learning, learning automata.