Search

This study evaluated the safety, tolerability, pharmacodynamics, and pharmacokinetics of recombinant neorudin (EPR‐hirudin [EH]) in patients with acute coronary syndrome (ACS), providing a basis for further therapeutic research. This open‐label, single‐center, nonrandomized, nonblinded, and noncontrolled trial categorized 24 patients with nonprogressive ACS who met the screening criteria into 3 groups. They received an intravenous injection of neorudin (0.4 mg/kg), followed by an intravenous drip at doses of 0.15, 0.30, and 0.45 mg/kg/h for 3 days in the low‐, medium‐, and high‐dose groups, respectively. The safety, tolerability, pharmacodynamics, and pharmacokinetics of EH were assessed after treatment, indicating that neorudin was safe and well tolerated in nonprogressive ACS. No serious adverse events or clinical composite end points were observed. The activated partial thromboplastin time and thrombin time increased significantly and dose dependently following EH administration across all groups compared to pretreatment values. Conversely, thrombin activity significantly decreased after drug administration but returned to baseline levels shortly after drug withdrawal. Within the administered dose range, neorudin exposure increased with the dose, and its half‐life was approximately 2 hours. Neorudin was found to be safe and tolerable for treating patients with nonprogressive ACS, demonstrating therapeutic efficacy at doses up to 0.45 mg/kg/h over a 3‐day period.

This study addresses the issues of strong data dependency and limited prediction accuracy in temperature prediction using optical fiber spectroscopy technology by proposing a small-sample optical fiber spectroscopy-based temperature prediction method leveraging deep learning. This method aims to achieve high-precision temperature prediction with limited data samples through the powerful feature extraction and generalization capabilities of deep learning models. To achieve this goal, we first designed a precise experimental protocol to collect optical fiber spectroscopy data covering a temperature range from 30°C to 130°C, resulting in 84 high-quality data samples under controlled temperature variations. In the data processing stage, advanced signal processing techniques were employed to remove noise and outliers, and the data were normalized to ensure reliability and consistency. Subsequently, various models from the deep learning domain were utilized to train and learn from the processed spectroscopy data. By optimizing the model structures and parameters, we successfully established a nonlinear mapping relationship between spectroscopy and temperature. Experimental results demonstrate that compared to traditional methods, the deep learning model exhibits higher prediction accuracy and stronger robustness in spectroscopic temperature prediction, particularly under small-sample conditions. This study not only provides a novel and effective approach for spectroscopic temperature prediction with limited samples but also expands the application scope of deep learning in spectral analysis. Furthermore, this method holds broad application prospects in various fields such as industrial production, environmental monitoring, and biomedicine, promising to offer more precise and efficient solutions for temperature monitoring and control in related areas.

In actual physiological environments, bacteria can activate the immune system and release lactic acid. However, the detailed contribution of lactic acid to the passivation behavior of titanium (Ti) alloys is still unclear. The current work investigated the in vitro passivation behavior of Ti–6Al–4V (TC4) alloys fabricated by laser powder bed fusion in Hank's solution with and without adding lactic acid. Electrochemical methods, inductively coupled plasma atomic emission spectrometer, and X-ray photoelectron spectroscopy were jointly used. Adding lactic acid decreases the corrosion resistance of samples by degrading the formed passive film. The film formed in the (lactic acid)-containing solution exhibits a higher level of oxygen vacancies and a lower thickness, attributed to the suppressed formation of Ti4+transformed from Ti3+and Ti2+. Moreover, the presence of lactic acid would increase the open circuit potential, relieve the ions release, and hinder the deposition of calcium phosphates within 24 h immersion.

Photomemristors have been regarded as one of the most promising candidates for next-generation hardware-based neuromorphic computing due to their potentials of fast data transmission and low power consumption. However, intriguingly, so far, photomemristors seldom display truly nonvolatile memory characteristics with high light sensitivity. Herein, we demonstrate ultrasensitive photomemristors utilizing two-dimensional (2D) Ruddlesden–Popper (RP) perovskites with a highly polar donor–acceptor-type push–pull organic cation, 4-(5-(2-aminoethyl)thiophen-2-yl)benzonitrile+(EATPCN+), as charge-trapping layers. High linearity and almost zero-decay retention are observed in (EATPCN)2PbI4devices, which are very distinct from that of the traditional 2D RP perovskite devices consisting of nonpolar organic cations, such as phenethylamine+(PEA+) and octylamine+(OA+), and traditional 3D perovskite devices consisting of methylamine+(MA+). The 2-fold advantages, including desirable spatial crystal arrangement and engineered energetic band alignment, clarify the mechanism of superior performance in (EATPCN)2PbI4devices. The optimized (EATPCN)2PbI4photomemristor also shows a memory window of 87.9 V and an on/off ratio of 106with a retention time of at least 2.4 × 105s and remains unchanged after >105writing–reading–erasing–reading endurance cycles. Very low energy consumptions of 1.12 and 6 fJ for both light stimulation and the reading process of each status update are also demonstrated. The extremely low power consumption and high photoresponsivity were simultaneously achieved. The high photosensitivity surpasses that of a state-of-the-art commercial pulse energy meter by several orders of magnitude. With their outstanding linearity and retention, rabbit images have been rebuilt by (EATPCN)2PbI4photomemristors, which truthfully render the image without fading over time. Finally, by utilizing the powerful ∼8 bits of nonvolatile potentiation and depression levels of (EATPCN)2PbI4photomemristors, the accuracies of the recognition tasks of CIFAR-10 image classification and MNIST handwritten digit classification have reached 89% and 94.8%, respectively. This study represents the first report of utilizing a functional donor–acceptor type of organic cation in 2D RP perovskites for high-performance photomemristors with characteristics that are not found in current halide perovskites.

The power unit on board the ship generates periodic low-frequency vibration that affects the normal operation of the equipment on board, and the adaptive feedforward control algorithm can effectively suppress such harmful vibration noise. But the adaptive feedforward control algorithm needs to obtain the identification model of the secondary channels, and the frequency domain least squares method based on the linear Extended auto-regressive model (ARX) is difficult to obtain the identification model with nonlinear characteristics. The nonlinear auto-regressive model (NARX) adds nonlinear mapping layers to the topology of the ARX model to enhance the identification capability of the NARX model for complex systems. In this paper, a block diagram of the Fx-LMS feedforward control algorithm based on the NARX model is proposed, then the initial parameters of the NARX neural network are optimized using the Quantum Particle Swarm Optimization (QPSO) algorithm and the secondary channel is identified, and the identification results show that the accuracy of identifying the secondary channel using the NARX neural network is higher than that of the ARX model. The simulation and experimental results show that the vibration damping effect of the proposed method is better than the traditional Fx-LMS method for both single-line spectrum and multi-line spectrum periodic low-frequency disturbances, which provides a new method for the suppression of periodic low-frequency disturbances.

A silicon (Si) Mach-Zehnder modulator (MZM) is demonstrated for 10-km SMF Link at C-band (1550 nm), which allows the pre-emphasized 4-level pulse amplitude modulation (PAM-4) data transmission at 50 GBaud (100 Gbit/s) per channel. By employing the pre-emphasis to the PAM-4 data over different links, the channel dispersion-induced power fading can be compensated. The PAM-4 data with improved spectral usage efficiency in limited bandwidth is employed to encode the Si MZM, and the maximal transmission capacities of 50 GBaud for back-to-back (BtB), 50 GBaud for 2-km SMF, and 26 GBaud for 10-km DSF under KP4-FEC criterion are respectively obtained for providing 100/100/52 Gbit/s data transmission with qualified bit error ratios of 1.5×10−4/2×10−4/1.4×10−4. The penalties on the sensitivity of the receiving power are 0.32 dB for 50-GBaud PAM-4 data after 2-km-SMF link and 0.64 dB for 26-GBaud PAM-4 data over 10-km DSF as compared to the BtB case. Such a scheme takes advantages including the lower power consumption with less circuit complexity than traditional push-pull driven operation, which certifies the 50-GBaud PAM-4 data capability per channel for inter-data-center applications. In addition, this Si MZM device can also modulate the pre-emphasized 16-QAM OFDM data at 140 Gbit/s and the bit-loaded DMT data at 153 Gbit/s.

The COVID-19 pandemic resulted in seismic changes to healthcare service delivery. The use of telemedicine was widely adopted during the pandemic, although its value in the safe care of vascular patients is unknown.

We present a novel full hardware implementation of Streamlined NTRU Prime, with two variants: a high-speed, high-area implementation and a slower, low-area implementation. We introduce several new techniques that improve performance, including a batch inversion for key generation, a high-speed schoolbook polynomial multiplier, an NTT polynomial multiplier combined with a CRT map, a new DSP-free modular reduction method, a high-speed radix sorting module, and new encoders and decoders. With the high-speed design, we achieve the to-date fastest speeds for Streamlined NTRU Prime, with speeds of 5007, 10,989, and 64,026 cycles for encapsulation, decapsulation, and key generation, respectively, while running at 285 MHz on a Xilinx Zynq Ultrascale+. The entire design uses 40,060 LUT, 26,384 flip-flops, 36.5 Bram, and 31 DSP.

In this study, a multiphase pyrolysis model has been proposed under the large eddy simulation (LES) framework incorporating moving boundary surface tracking, char formation, and detailed chemical kinetics combustion modelling. The proposed numerical model was applied to simulate the cone calorimeter test of two kinds of materials: (i) pinewood (charring) and (ii) low-density polyethylene (non-charring). Using a cone calorimeter setup, good agreement has been achieved between the computational and the experimental results. The model is capable of predicting the formation of the char layer and thus replicating the flame suppressing thermal and barrier effects. Furthermore, with the application of detailed chemical kinetics, the fire model was able to aptly predict the generation of asphyxiant gas such as CO/CO2during the burning process. However, the pinewood experiments showed significant CO/CO2emissions post flame extinguishment attributed to char oxidation effects, which were not considered by the fire model. Despite the limitation, the fully coupled LES model proposed in this study was capable of predicting the fluid mechanics and heat transfer for the turbulent reacting flow, solid-phase decomposition, and gaseous products under flaming conditions. In the future, it can be further extended to include char oxidation mechanisms to improve predictions for charring materials.

In this paper, two arc modes of gas shielded metal arc welding (GMAW), a new pulsed mode (microMIG) and a non-pulsed mode (microMIG-CC), were used to melt ER5356 aluminum alloy wire for thin-wall fabrication. The influence of both arc modes on the melt–drop transition pattern, thin–wall properties, porosity, microstructure and mechanical properties was investigated at the same heat input. The microMIG mode adds SKS-based current time based on conventional pulsed GMAW for short-circuiting the transition, which leads to a faster spread of melt pool temperature. The rapid cooling of the melt pool contributed to the increase of grain nucleation rate and inhibited grain growth, resulting in smaller grain size in the microMIG mode specimens. In addition, the low porosity of microMIG specimens is attributed to the high voltage of the micro-MIG mode resulting in an unfocused heat source, which leads to a shallow penetration depth of the specimens and a reduced hydrogen escape distance. Compared with the microMIG-CC mode, the microMIG mode specimens have smaller grain size and lower porosity, therefore higher tensile strength is obtained. The isotropic property of thin walls is due to the existence of equiaxed grains and the random distribution of grain orientation. However, the aggregation of dislocations and the increase of precipitates make microMIG-CC have higher hardness. Overall, the mechanical properties of the thin wall specimens of WAAM 5356 aluminum alloy are higher than those of the as-cast specimens.

CoNi and FeCoNi hydroxide with narrow voltage distance between oxygen reduction reaction and oxygen evolution reaction was synthetized by electro-deposition in low solvent concentration. 5cm2 Membrane electrode assemble (MEA) electrolyzers composed with anion exchange membrane, homogenerated catalyst on both cathode and anode gas diffusion layer (GDL) was fabricated for oxygen electrochemical production from air. The current and yield of binary CoNi device reached up to 466.7mA and 4.4mmol/h (94.7% conversion rate) at 1.2V. The ternary FeCoNi device showed only 0.5% degradation from 394.0mA during 12h. The applicability of oxygen production from air by high performance electrochemical devices was demonstrated.

Modern buildings and structures are commonly equipped with fire safety detection and protection systems. Owing to the complexity in building architectures, performance-based fire engineering designs are often applied to achieve safety compliance criteria in stipulated fire events. With the uprising popularity of computer simulation fire predictive models benefited by the rapid improvement in computing speed and modelling techniques, the use of computational fluid dynamics (CFD) based fire field models has become an integrated component in fire tenability and assessment studies. This article delivers a comprehensive review on the history, past developments, and current state-of-the-art of CFD models for enclosure fires, as well as providing an in-depth review on the advancement in other sub-modelling components including turbulence, combustion, radiation, and soot models. Additionally, two types of multiphase modelling approaches involving solid-gas and liquid-gas phase models are reviewed. As for the preceding, the consideration of the solid phase combustibles is generally achieved via pyrolysis modelling under the context of CFD. Recent advancements in CFD-based pyrolysis studies are extensively discussed, including the consideration of porous media, charring layer formation, and kinetics search algorithms to describe the solid decomposition and charring processes. Meanwhile, fire suppression models involving the discrete phase model (DPM) approach are reviewed. This includes previous developments in simulation methods of water droplets, coupling approaches with the fire dynamics in the large eddy simulation (LES) framework. Finally, a future perspective regarding the need to develop a melting/dripping sub-model for building materials is discussed, whose reaction kinetics can be supported by molecular dynamics (MD).

ObjectivePatients with increased PD-L1+host cells in tumours are more potent to benefit from antiprogrammed death-1/programmed death ligand-1 (PD-L1) treatment, but the underlying mechanism is still unclear. We aim to elucidate the nature, regulation and functional relevance of PD-L1+host cells in hepatocellular carcinoma (HCC).DesignA total of untreated 184 HCC patients was enrolled randomly. C57BL/6 mice are given injection of Hepa1-6 cells to form autologous hepatoma. ELISpot, flow cytometry and real-time PCR are applied to analyse the phenotypic characteristics of PD-L1+cells isolated directly from HCC specimens paired with blood samples or generated from ex vivo and in vitro culture systems. Immunofluorescence and immunohistochemistry are performed to detect the presence of immune cells on paraffin-embedded and formalin-fixed samples. The underlying regulatory mechanisms of metabolic switching are assessed by both in vitro and in vivo studies.ResultsWe demonstrate that PD-L1+host macrophages, which constructively represent the major cellular source of PD-L1 in HCC tumours, display an HLA-DRhighCD86highglycolytic phenotype, significantly produce antitumourigenic IL-12p70 and are polarised by intrinsic glycolytic metabolism. Mechanistically, a key glycolytic enzyme PKM2 triggered by hepatoma cell derived fibronectin 1, via a HIF-1α-dependent manner, concurrently controls the antitumourigenic properties and inflammation-mediated PD-L1 expression in glycolytic macrophages. Importantly, although increased PKM2+glycolytic macrophages predict poor prognosis of patients, blocking PD-L1 on these cells eliminates PD-L1-dominant immunosuppression and liberates intrinsic antitumourigenic properties.ConclusionsSelectively modulating the ‘context’ of glycolytic macrophages in HCC tumours might restore their antitumourigenic properties and provide a precise strategy for anticancer therapy.

Graphical abstract:

ABSTRACTIn recent research, a method for manufacturing superhydrophobic metal surfaces using pulse laser texturing has been developed. After laser texturing, a sample can be converted into a superhydrophobic surface by aging the sample for several weeks in ambient air. Research on this method is currently focused on metal surfaces, such as aluminium, copper and so on. There are few studies of iron and steel materials, and the process and principle of wetting conversion is not sufficiently clear, and thus, further research is needed. In this study, a stainless steel surface with a large number of Gaussian pits was prepared using a nanosecond fibre laser, which was then naturally converted into a superhydrophobic surface in air. The effect of the roughness, lattice arrangement and chemical composition of a surface on its hydrophobic properties was studied to realize controllable hydrophobic properties. A hexagonal lattice surface prepared using 10 processing cycles showed the best hydrophobic performance, with contact and rolling angles of 161.3 and 3.2°, respectively.

In recent research, a method for manufacturing superhydrophobic metal surfaces using pulse laser texturing has been developed. After laser texturing, a sample can be converted into a superhydrophobic surface by aging the sample for several weeks in ambient air. Research on this method is currently focused on metal surfaces, such as aluminium, copper and so on. There are few studies of iron and steel materials, and the process and principle of wetting conversion is not sufficiently clear, and thus, further research is needed. In this study, a stainless steel surface with a large number of Gaussian pits was prepared using a nanosecond fibre laser, which was then naturally converted into a superhydrophobic surface in air. The effect of the roughness, lattice arrangement and chemical composition of a surface on its hydrophobic properties was studied to realize controllable hydrophobic properties. A hexagonal lattice surface prepared using 10 processing cycles showed the best hydrophobic performance, with contact and rolling angles of 161.3 and 3.2°, respectively.

At the beginning of the coronavirus outbreak in late 2019 (COVID-19), the procedure of epidemic prevention for each student in most campuses in Taiwan included taking the temperature of the forehead and disinfecting hands by use of alcohol or hand sanitizer; there was an obvious need for many operators to help in the epidemic-prevention process. In our campus, a scan of the QR code is also required to register on the campus cloud system for a footprint record. Therefore, before class, many students are waiting in line, and each student needs to spend about 10 minutes for the entire process. Hence, we designed a new campus epidemic-prevention system against the threat of COVID-19. The system can automatically perform the forehead temperature measurement, hand sanitizing, and ID registration on each user. The manually operated issues, such as infection risk of operators, manpower requirement of epidemic-prevention stations, and the entire process times, can all be reduced.

Objective: To observe the effects of acupoints, cone numbers and durations of moxibustion with different moxibustion methods on skin surface and inside temperature, and to provide references for the clinical standardization of moxibustion amount. Methods: The 42 big-ear white rabbits were divided into 6 groups according to the random number table method, a 1-cone direct moxibustion group, a 2-cone direct moxibustion group, a 3-cone direct moxibustion group, a 1-cone herbal cake-partitioned moxibustion group, a 2-cone herbal cake-partitioned moxibustion group, and a 3-cone herbal cake-partitioned moxibustion group, with 7 rabbits in each group. Shenque (CV 8), Shenshu (BL 23) and Zusanli (ST 36) were used in each group, but the moxibustion methods, cone numbers and durations of moxibustion were different. Rabbits in each group received moxibustion once every other day for 5 times in total. During the intervention, a thermoelectricity coupled probe and a temperature recorder were used to record the real-time acupoint skin temperature and the temperature at different time points, so as to observe, analyze and process the real-time changes in the temperature difference between the surface and inside of acupoint skin. Results: For herbal cake-partitioned moxibustion, the best temperature for cone changing was (46.38±0.51) °C when the highest surface temperature was (49.20±0.52) °C; the multi-factor comparison of acupoint × cone number × time and acupoint × moxibustion method × time showed that time × acupoint, time × moxibustion method and cone number × acupoint had interactive effects (all P<0.05). Comparing skin temperature differences between different cone numbers at the same acupoint, Shenque (CV 8) on the 1st and the 5th days, Shenshu (BL 23) on the 3rd and the 7th days, Zusanli (ST 36) on the 1st and the 9th days of experiment showed statistically significant differences (all P<0.05). The skin temperature comparison of different moxibustion methods at the same acupoint all had statistical differences (all P<0.05), except for Shenque (CV 8) before moxibustion, Shenshu (BL 23) before moxibustion and on the 5th day; Zusanli (ST 36) only showed statistical differences on the 5th and 7th days (both P<0.05). The skin temperature differences of different acupoints after moxibustion in the 1-cone, 2-cone and 3-cone groups were statistically different (all P<0.05); direct moxibustion and herbal cake-partitioned moxibustion at different acupoints were all statistically different (all P<0.05). Conclusion: Cone changing temperature under the same specifications of herbal cake-partitioned moxibustion was confirmed. Temperature difference between surface and inside of different acupoint skin at the same maximum temperature was significantly different due to the cone numbers and moxibustion methods, which showed the highest at Shenshu (BL 23), the second at Shenque (CV 8), and the lowest at Zusanli (ST 36). The influence of acupoint factor should be considered to determine the quantitative indicators of moxibustion.

Long-chain molecules can self-assemble into ordered and functional structures through chain folding. In crystalline polymers, chain folding leads to lamellar crystals that are extremely important to the ultimate properties. Understanding chain folding at the molecular level remains a great challenge. Here, we report that a dynamic helical cis-poly(phenylacetylene) bearing a bulky side group (P1), which is a side-chain liquid crystalline polymer, can also form folded chain lamellae in the hexagonal columnar phase. Our atomic force microscopy (AFM) experiments on P1thin films directly visualized the adjacent and nonadjacent folds at the liquid crystal–amorphous interface. While the helical segments of P1with a diameter of 2.2 nm closely pack in the hexagonal lattice, the disorder strands that connect the helices could compose the folds. Furthermore, we monitored the lamellar growth from the amorphous state using AFM. We demonstrate experimentally for the first time that the lamellar growth may obey the kinetics of Ostwald ripening. It is also found that with the initial lamellar thickness close to the persistence length of P1, the lamellae show simultaneous lateral extension and thickening.

Owing to the well-established Eulerian-Lagrangian framework on mixture fluids, computational fluid dynamics coupled with discrete element model (CFD-DEM) is an effective while appropriate tool to predict the complex interactive fire behaviours associate with suppression effects. Although suppression behaviours between hydrocarbon-fuelled fire and water-based suppression agents were extensively studied both numerically and experimentally, lack of numerical studies was conducted on fires involving water-reactive chemicals (i.e., Na, Li, and LiH), where extinguishment is barely performed by water-based active suppression system, as violent and explosive decomposition occurred between water and reactive fuel. In this research, a numerical investigation has been conducted on expandable graphite (EG) application for water-reactive fire suppression. Based on the discrete phase model (DPM) framework, a novel EG particle model is proposed to characterise the particle expansion that couples with superior thermal properties and chemical stability. A numerical assessment on large eddy simulation (LES) has been performed to study the temporal fire behaviours and the suppression effect of EG against the flame plume in various subgrid-scale (SGS) models. Four SGS models were adopted, which were namely Smagorinsky-Lilly, WALE, dynamic kinetic energy, and dynamic Smagorinsky-Lilly. As a result, the WALE SGS model was observed to be in a better agreement compared with the experimental data owing to its significant enhancement in flow diffusivity modelling. The WALE SGS model has achieved a more accurate temperature prediction and finer resolved turbulence compared with other SGS models.

The one of major impurities in quartz is iron, which has a great impact on the properties of the material and should be be strictly removed in many applications. In this study, a low-energy consumption, simple, fast processing is introduced by combination microwave treatment with different acid leaching. Selective heating of microwaves is used to process quartz, causing local phase changes to improve the purification effect. Acid leaching was used to remove the iron in the quartz matrix. Under optimized conditions, the iron content can be reduced to below 0.167 ppmw with one single purification pass.

The energetic bremsstrahlung photons up to 100 MeV produced in heavy ion collisions can be used as a sensitive probe for short-range correlation in atomic nuclei. The energy of the γ-rays can be measured by collecting the Čerenkov light in the medium induced by the fast electrons generated in the Compton scattering or electromagnetic shower of the incident γray. Two types of detectors based on pure water and lead glass as sensitive materials were designed for this purpose. The γresponse and optical photon propagation in the detectors were simulated based on electromagnetic and optical processes in Geant4. The inherent energy resolutions of 0.022(4)+0.51(2)/Eγ1/2for water and 0.0026(3)+0.446(3)/Eγ1/2for lead glass were obtained. The geometry sizes of the lead glass and water were optimized to 30 cm ×30 cm ×30 cm and 60 cm ×60 cm ×120 cm, respectively, to detect high-energy γ-rays at 160 MeV. The Hough transform method was applied to reconstruct the direction of the incident γ-rays, providing the ability to experimentally distinguish the high-energy γ-rays produced in the reactions on the target from random background cosmic-ray muons.

Exchange bias, a shift in the hysteresis loop of a ferromagnet arising from interfacial exchange coupling between adjacent ferromagnetic and antiferromagnetic layers, is an integral part of spintronic devices. Here, we show that spin–orbit torque generated from spin current, a promising approach to switch the ferromagnetic magnetization of next-generation magnetic random access memory, can also be used to manipulate the exchange bias. Applying current pulses to a Pt/Co/IrMn trilayer causes concurrent switching of ferromagnetic magnetization and exchange bias, but with different underlying mechanisms. This implies that the ferromagnetic magnetization and exchange bias can be manipulated independently. Our work demonstrates that spin–orbit torque in ferromagnet/antiferromagnet heterostructures facilitates independent manipulations of distinct magnetic properties, motivating innovative designs for future spintronics devices.

Modern Systems-on-Chip (SoC) designs are increasingly heterogeneous and contain specialized semi-programmable accelerators in addition to programmable processors. In contrast to the pre-accelerator era, when the ISA played an important role in verification by enabling a clean separation of concerns between software and hardware, verification of these “accelerator-rich” SoCs presents new challenges. From the perspective of hardware designers, there is a lack of a common framework for formal functional specification of accelerator behavior. From the perspective of software developers, there exists no unified framework for reasoning about software/hardware interactions of programs that interact with accelerators.This article addresses these challenges by providing a formal specification and high-level abstraction for accelerator functional behavior. It formalizes the concept of an Instruction Level Abstraction (ILA), developed informally in our previous work, and shows its application in modeling and verification of accelerators. This formal ILA extends the familiar notion of instructions to accelerators and provides a uniform, modular, and hierarchical abstraction for modeling software-visible behavior of both accelerators and programmable processors. We demonstrate the applicability of the ILA through several case studies of accelerators (for image processing, machine learning, and cryptography), and a general-purpose processor (RISC-V). We show how the ILA model facilitates equivalence checking between two ILAs, and between an ILA and its hardware finite-state machine (FSM) implementation. Further, this equivalence checking supports accelerator upgrades using the notion of ILA compatibility, similar to processor upgrades using ISA compatibility.