Accurate and efficient prediction of polymer properties is of great significance in polymer design.Conventionally,expensive and time-consuming experiments or simulations are required to evaluate polymer functions.Rece...Accurate and efficient prediction of polymer properties is of great significance in polymer design.Conventionally,expensive and time-consuming experiments or simulations are required to evaluate polymer functions.Recently,Transformer models,equipped with self-attention mechanisms,have exhibited superior performance in natural language processing.However,such methods have not been investigated in polymer sciences.Herein,we report TransPolymer,a Transformer-based language model for polymer property prediction.Our proposed polymer tokenizer with chemical awareness enables learning representations from polymer sequences.Rigorous experiments on ten polymer property prediction benchmarks demonstrate the superior performance of TransPolymer.Moreover,we show that TransPolymer benefits from pretraining on large unlabeled dataset via Masked Language Modeling.Experimental results further manifest the important role of self-attention in modeling polymer sequences.We highlight this model as a promising computational tool for promoting rational polymer design and understanding structure-property relationships from a data science view.展开更多
A power-scaled laser operation of Pr:YLi F4(YLF)crystal at 720.9 nm pumped by a 443.6 nm laser diode(LD)module was demonstrated.The 20 W module was used to pump the Pr:YLF crystal,and a maximum output power of 3.03 W ...A power-scaled laser operation of Pr:YLi F4(YLF)crystal at 720.9 nm pumped by a 443.6 nm laser diode(LD)module was demonstrated.The 20 W module was used to pump the Pr:YLF crystal,and a maximum output power of 3.03 W with slope efficiency of 30.04%was obtained.In addition,a 5 W blue LD was also used to pump the Pr:YLF laser,and a maximum output power of 0.72 W was obtained at room temperature.The output power was limited by the wavelength mismatch between the single-emitter LD and the absorption peak of the crystal.展开更多
Radially polarized beams characterized by an axially symmetric polarization distribution can be sharply focused to produce strong longitudinal fields in the vicinity.Future applications of these beams will be facilita...Radially polarized beams characterized by an axially symmetric polarization distribution can be sharply focused to produce strong longitudinal fields in the vicinity.Future applications of these beams will be facilitated by the availability of higher powers and shorter durations.Currently,the ultrafast radially polarized pulse is typically generated via wavefront reconstruction from conventional linearly polarized states.Achievable pulse duration and intensity limits are strictly dependent on extra-cavity optics.Herein,a chirp-assisted near-degenerate type-II parametric process is presented as a pulse-energy-scalable method of accessing ultrafast radially polarized pulses.In a proof-of-principle experiment,the broadband gain balance between the orthogonally polarized signal components was realized via controlling the chirp of the pump pulse.Through an analogous pulseduration transfer effect,the radially polarized signal inherited the temporal and spectral characteristics of the pump pulse and maintained the radial polarization state of each frequency component of the signal.With a shorter pump pulse,the generation of few-cycle radially polarized pulses should be achievable,which may facilitate a wide range of ultrafast applications such as vacuum electron acceleration and high-harmonic generation.展开更多
文摘Accurate and efficient prediction of polymer properties is of great significance in polymer design.Conventionally,expensive and time-consuming experiments or simulations are required to evaluate polymer functions.Recently,Transformer models,equipped with self-attention mechanisms,have exhibited superior performance in natural language processing.However,such methods have not been investigated in polymer sciences.Herein,we report TransPolymer,a Transformer-based language model for polymer property prediction.Our proposed polymer tokenizer with chemical awareness enables learning representations from polymer sequences.Rigorous experiments on ten polymer property prediction benchmarks demonstrate the superior performance of TransPolymer.Moreover,we show that TransPolymer benefits from pretraining on large unlabeled dataset via Masked Language Modeling.Experimental results further manifest the important role of self-attention in modeling polymer sequences.We highlight this model as a promising computational tool for promoting rational polymer design and understanding structure-property relationships from a data science view.
基金supported by the National Natural Science Foundation of China(Nos.61405126 and 61704112)the Science and Technology Planning Project of Shenzhen Municipality(Nos.ZDSYS201707271014468 and JCYJ20170817094438146)the Educational Commission of Guangdong Province(No.2016KCXTD006).
文摘A power-scaled laser operation of Pr:YLi F4(YLF)crystal at 720.9 nm pumped by a 443.6 nm laser diode(LD)module was demonstrated.The 20 W module was used to pump the Pr:YLF crystal,and a maximum output power of 3.03 W with slope efficiency of 30.04%was obtained.In addition,a 5 W blue LD was also used to pump the Pr:YLF laser,and a maximum output power of 0.72 W was obtained at room temperature.The output power was limited by the wavelength mismatch between the single-emitter LD and the absorption peak of the crystal.
基金supported by the National Natural Science Foundation of China(Grant No.92050203)the Natural Science Foundation of Guangdong Province(Grant No.2020A1515010541)the Science and Technology Project of Shenzhen(Grant Nos.JCYJ20200109105606426,JCYJ20190808143419622,and JCYJ20190808145016980)。
文摘Radially polarized beams characterized by an axially symmetric polarization distribution can be sharply focused to produce strong longitudinal fields in the vicinity.Future applications of these beams will be facilitated by the availability of higher powers and shorter durations.Currently,the ultrafast radially polarized pulse is typically generated via wavefront reconstruction from conventional linearly polarized states.Achievable pulse duration and intensity limits are strictly dependent on extra-cavity optics.Herein,a chirp-assisted near-degenerate type-II parametric process is presented as a pulse-energy-scalable method of accessing ultrafast radially polarized pulses.In a proof-of-principle experiment,the broadband gain balance between the orthogonally polarized signal components was realized via controlling the chirp of the pump pulse.Through an analogous pulseduration transfer effect,the radially polarized signal inherited the temporal and spectral characteristics of the pump pulse and maintained the radial polarization state of each frequency component of the signal.With a shorter pump pulse,the generation of few-cycle radially polarized pulses should be achievable,which may facilitate a wide range of ultrafast applications such as vacuum electron acceleration and high-harmonic generation.