Showing papers by "Christopher C. Cummins published in 2022"

Photochemical Alkene Hydrophosphination with Bis(trichlorosilyl)phosphine.

Abstract: Bis(trichlorosilyl)phosphine (HP(SiCl3)2, 1) was prepared from [TBA][P(SiCl3)2] ([TBA]2, TBA = tetra-n-butylammonium) and triflic acid in 36% yield. Phosphine 1 is an efficient reagent for hydrophosphination of unactivated terminal olefins under UV irradiation (15-60 min) and gives rise to bis(trichlorosilyl)alkylphosphines (RP(SiCl3)2, R = (CH2)5CH3, 88%; (CH2)7CH3, 98%; (CH2)2C(CH3)3, 76%; CH2Cy, 93%; (CH2)2Cy, 95%; CH2CH(CH3)(CH2)2CH3, 82%; (CH2)3O(CH2)3CH3, 95%; (CH2)3Cl, 83%; (CH2)2SiMe3, 92%; (CH2)5C(H)CH2, 44%) in excellent yields. The products require no further purification beyond filtration and removal of volatile material under reduced pressure. The P-Si bonds of prototypical products RP(SiCl3)2 (R = -(CH2)5CH3, -(CH2)7CH3) are readily functionalized to give further phosphorus-containing products: H3C(CH2)7PCl2 (56%), [H3C(CH2)5P(CH2Ph)3]Br (84%), H3C(CH2)7PH2 (61%), H3C(CH2)5P(O)(H)(OH) (81%), and H3C(CH2)5P(O)(OH)2 (55%). Experimental mechanistic investigations, accompanied by quantum chemical calculations, point toward a radical-chain mechanism. Phosphine 1 enables the fast, high-yielding, and atom-efficient preparation of compounds that contain phosphorus-carbon bonds in procedures that bypass white phosphorus (P4), a toxic and high-energy intermediate of the phosphorus industry.

Automating reinforcement learning architecture design for code optimization

Abstract: Reinforcement learning (RL) is emerging as a powerful technique for solving complex code optimization tasks with an ample search space. While promising, existing solutions require a painstaking manual process to tune the right task-specific RL architecture, for which compiler developers need to determine the composition of the RL exploration algorithm, its supporting components like state, reward, and transition functions, and the hyperparameters of these models. This paper introduces SuperSonic, a new open-source framework to allow compiler developers to integrate RL into compilers easily, regardless of their RL expertise. SuperSonic supports customizable RL architecture compositions to target a wide range of optimization tasks. A key feature of SuperSonic is the use of deep RL and multi-task learning techniques to develop a meta-optimizer to automatically find and tune the right RL architecture from training benchmarks. The tuned RL can then be deployed to optimize new programs. We demonstrate the efficacy and generality of SuperSonic by applying it to four code optimization problems and comparing it against eight auto-tuning frameworks. Experimental results show that SuperSonic consistently improves hand-tuned methods by delivering better overall performance, accelerating the deployment-stage search by 1.75x on average (up to 100x).

Reactions of Tri-tert-Butylphosphatetrahedrane as a Spring-Loaded Phosphinidene Synthon Featuring Nickel-Catalyzed Transfer to Unactivated Alkenes.

Abstract: Cage-opening reactions of the highly strained tri-tert-butylphosphatetrahedrane (1), shown here to function as a synthon of (tri-tert-butylcyclopropenyl)phosphinidene, are described. Treatment of 1 with a base-stabilized silylene led to the corresponding phosphasilene, which was isolated in 72% yield as a red crystalline solid. Phosphinidene transfer was also observed when 1 (2 equiv) was combined with the Wittig reagent Ph3PCH2 to form a diphosphirane (50% isolated yield). The reaction is proposed to proceed through a generated phosphaalkene intermediate, which was characterized by NMR spectroscopy. In addition, we report on nickel-catalyzed phosphinidene transfer to styrene, ethylene, neohexene, and 1,3-cyclohexadiene; the corresponding phosphiranes were isolated in 51-64% yield. Computational studies suggest the intermediacy of a nickel phosphinidene species. Treatment of the ethylene-derived phosphirane product with triflic acid delivered elimination of [tBu3C3]OTf and formation of a P-H bond, illustrating the ability of the tri-tert-butyl cyclopropenyl group to serve as a protecting group that is removable following phosphinidene transfer.

Beyond Triphosphates: Reagents and Methods for Chemical Oligophosphorylation.

Abstract: Oligophosphates play essential roles in biochemistry, and considerable research has been directed toward the synthesis of both naturally occurring oligophosphates and their synthetic analogues. Greater attention has been given to mono-, di-, and triphosphates, as these are present in higher concentrations biologically and easier to synthesize. However, extended oligophosphates have potent biochemical roles, ranging from blood coagulation to HIV drug resistance. Sporadic reports have slowly built a niche body of literature related to the synthesis and study of extended oligophosphates, but newfound interests and developments have the potential to rapidly expand this field. Here we report on current methods to synthesize oligophosphates longer than triphosphates and comment on the most important future directions for this area of research. The state of the art has provided fairly robust methods for synthesizing nucleoside 5'-tetra- and pentaphosphates as well as dinucleoside 5',5'-oligophosphates. Future research should endeavor to push such syntheses to longer oligophosphates while developing synthetic methodologies for rarer morphologies such as 3'-nucleoside oligophosphates, polyphosphates, and phosphonate/thiophosphate analogues of these species.

Chemical Challenges that the Peroxide Dianion Presents to Rechargeable Lithium–Air Batteries

Abstract: 1 Department of Chemistry and Chemical Biology, Harvard University, 12 Oxford Street, Cambridge, Massachusetts 02138, United States 2 Centro de Investigaciones Químicas IICBA, Universidad Autónoma del Estado de Morelos, Av. Universidad 1001, Col. Chamilpa, Cuernavaca, Morelos 62209, Mexico 3 Department of Chemistry, Massachusetts Institute of Technology, 77 Massachusetts Avenue, Cambridge, Massachusetts 02139, United States

Staudinger Reactivity and Click Chemistry of Anthracene (A)-Based Azidophosphine N3PA.

Abstract: 11-Azido-9,10-dihydro-9,10-phosphanoanthracene (N3PA) has been demonstrated recently as a transfer reagent for molecular phosphorus mononitride (PN) because it easily dissociates at room temperature into dinitrogen (N2), PN, and anthracene (A). Here we report further reactivity studies of the N3PA molecule including strain-promoted 1,3-dipolar cycloaddition with cyclooctyne and Staudinger-type reactivity. Calculations at the DLPNO-CCSD(T)/cc-pVTZ//PBE0-D3(BJ)/cc-pVTZ level of theory indicate that the click reaction is faster than the dissociation of N3PA. The Staudinger-type reactivity enabled transfer of the NPA fragment to a base-stabilized silylene. The previously reported intermediate of vanadium trisanilide with an NPA ligand could be isolated in 61% yield and structurally characterized in a single-crystal X-ray diffraction experiment. In line with the previously reported phosphinidene reactivity of the transient vanadium phosphorus mononitride complex, thermolysis or irradiation of the complex leads to A elimination and formation of the corresponding vanadium PN dimer or trimer, respectively.

Q-gym: An Equality Saturation Framework for DNN Inference Exploiting Weight Repetition

Abstract: The high computation cost is one of the key bottlenecks for adopting deep neural networks (DNNs) in different hardware. When client data are sensitive, privacy-preserving DNN evaluation method, such as homomorphic encryptions (HE), shows even more computation cost. Prior works employed weight repetition in quantized neural networks to save the computation of convolutions by memorizing or arithmetic factorization. However, such methods fail to fully exploit the exponential search space from factorizing and reusing computation. We propose Q-gym, a DNN framework consisting of two components. First, we propose a compiler, which leverages equality saturation to generate computation expressions for convolutional layers with a significant reduction in the number of operations. Second, we integrate the computation expressions with various parallelization methods to accelerate DNN inference on different hardware. We also employ the efficient expressions to accelerate DNN inference under HE. Extensive experiments show that Q-gym achieves 19.1% / 68.9% more operation reductions compared to SumMerge and original DNNs. Also, computation expressions from Q-gym contribute to 2.56× / 1.78× inference speedup on CPU / GPU compared to OneDNN and PyTorch GPU on average. For DNN evaluation under HE, Q-gym reduces the homomorphic operations by 2.47× / 1.30× relative to CryptoNet and FastCryptoNet for HE tasks with only 0.06% accuracy loss due to quantization.

Introducing N-Heterocyclic Iminophosphoranes (NHIPs): Synthesis by [3 + 2] Cycloaddition of Azophosphines with Alkynes and Reactivity Studies.

Abstract: Azophosphines (Ar-N═N-PR2) were prepared from N-aryl-N'-(trimethylsilyl)diazenes (Ar-N═N-SiMe3) and R2PCl by Me3SiCl elimination or oxidation of phosphinohydrazines (Ar-NH-NH-PR2) by 2,5-dialkyl-1,4-benzoquinones. Azophosphines underwent 1,3-dipolar cycloaddition with cyclooctyne and dimethylacetylene dicarboxylate to give N-heterocyclic iminophosphoranes (NHIPs), which are structurally similar to cyclic (alkyl)(amino)carbenes. The cycloaddition reaction is compatible with various phosphorus atom substituents including phenyl (NHIP-1,4,6), isopropyl (NHIP-2), cyclohexyl (NHIP-3), and dimethylamino (NHIP-5) groups. The pKBH+ values of the NHIPs in acetonitrile range from 13.13 to 23.14. On the basis of the Huynh electronic parameter, NHIP-1 and NHIP-2 have σ-donor strengths comparable with that of 1,8-diazabicyclo[5.4.0]undec-7-ene. NHIP-1 underwent facile 1,2-addition with pentafluoropyridine to form a rare fluorophosphorane. The treatment of NHIP-1 with triphenylsilane resulted in P-N bond cleavage, accompanied by the reduction of phosphorus(V) to phosphorus(III). A homoleptic, cationic CuI-NHIP-1 complex was also prepared. The potential utility of π-donating NHIPs was demonstrated by the stabilization of a reactive iminoborane (Cl-B≡N-SiMe3). The facile scalable synthesis, tunability of steric demands, and basicity of NHIPs suggest that this new heterocycle class may find a wide range of applications in synthetic chemistry.

BenchPress: A Deep Active Benchmark Generator

Abstract: Finding the right heuristics to optimize code has always been a difficult and mostly manual task for compiler engineers. Today this task is near-impossible as hardware-software complexity has scaled up exponentially. Predictive models for compilers have recently emerged which require little human effort but are far better than humans in finding near optimal heuristics. As any machine learning technique, they are only as good as the data they are trained on but there is a severe shortage of code for training compilers. Researchers have tried to remedy this with code generation but their synthetic benchmarks, although thousands, are small, repetitive and poor in features, therefore ineffective. This indicates the shortage is of feature quality more than corpus size. It is more important than ever to develop a directed program generation approach that will produce benchmarks with valuable features for training compiler heuristics. We develop BenchPress, the first ML benchmark generator for compilers that is steerable within feature space representations of source code. BenchPress synthesizes compiling functions by adding new code in any part of an empty or existing sequence by jointly observing its left and right context, achieving excellent compilation rate. BenchPress steers benchmark generation towards desired target features that has been impossible for state of the art synthesizers (or indeed humans) to reach. It performs better in targeting the features of Rodinia benchmarks in 3 different feature spaces compared with (a) CLgen - a state of the art ML synthesizer, (b) CLSmith fuzzer, (c) SRCIROR mutator or even (d) human-written code from GitHub. BenchPress is the first generator to search the feature space with active learning in order to generate benchmarks that will improve a downstream task. We show how using BenchPress, Grewe's et al. CPU vs GPU heuristic model can obtain a higher speedup when trained on BenchPress's benchmarks compared to other techniques. BenchPress is a powerful code generator: Its generated samples compile at a rate of 86%, compared to CLgen's 2.33%. Starting from an empty fixed input, BenchPress produces 10× more unique, compiling OpenCL benchmarks than CLgen, which are significantly larger and more feature diverse. We release BenchPress's source code on a publicly available repository at https://github.com/fivosts/BenchPress. This work was supported by the Engineering and Physical Sciences Research Council (grant EP/L01503X/1), EPSRC Centre for Doctoral Training in Pervasive Parallelism at the University of Edinburgh, School of Informatics. This work was supported by the Royal Academy of Engineering under the Research Fellowship scheme.

Synthesis of phosphiranes via organoiron-catalyzed phosphinidene transfer to electron-deficient olefins

Abstract: Herein is reported the structural characterization and scalable preparation of the elusive iron–phosphido complex FpP(tBu)(F) (2-F, Fp = (Fe(η5-C5H5)(CO)2)) and its precursor FpP(tBu)(Cl) (2-Cl) in 51% and 71% yields, respectively. These phosphide complexes are proposed to be relevant to an organoiron catalytic cycle for phosphinidene transfer to electron-deficient alkenes. Examination of their properties led to the discovery of a more efficient catalytic system involving the simple, commercially available organoiron catalyst Fp2. This improved catalysis also enabled the preparation of new phosphiranes with high yields (tBuPCH2CHR; R = CO2Me, 41%; R = CN, 83%; R = 4-biphenyl, 73%; R = SO2Ph, 71%; R = POPh2, 70%; R = 4-pyridyl, 82%; R = 2-pyridyl, 67%; R = PPh3+, 64%) and good diastereoselectivity, demonstrating the feasibility of the phosphinidene group-transfer strategy in synthetic chemistry. Experimental and theoretical studies suggest that the original catalysis involves 2-X as the nucleophile, while for the new Fp2-catalyzed reaction they implicate a diiron–phosphido complex Fp2(PtBu), 4, as the nucleophile which attacks the electron-deficient olefin in the key first P–C bond-forming step. In both systems, the initial nucleophilic attack may be accompanied by favorable five-membered ring formation involving a carbonyl ligand, a (reversible) pathway competitive with formation of the three-membered ring found in the phosphirane product. A novel radical mechanism is suggested for the new Fp2-catalyzed system.

MIT Open Access Articles A Joint Experimental and Computational Study of the Negative Ion Photoelectron Spectroscopy of the 1-Phospha-2,3,4-Triazolate Anion, HCPN₃

Abstract: : We report here the results of a combined experimental and computational study of the negative ion photoelectron spectroscopy (NIPES) of the recently synthesized, planar, aromatic, HCPN 3 − ion. The adiabatic electron detachment energy of HCPN 3 − (electron a ffi nity of HCPN 3 • ) was measured to be 3.555 ± 0.010 eV, a value that is intermediate between the electron detachment energies of the closely related (CH) 2 N 3 − and P 2 N 3 − ions. High level electronic structure calculations and Franck − Condon factor (FCF) simulations reveal that transitions from the ground state of the anion to two nearly degenerate, low-lying, electronic states, of the neutral HCPN 3 • radical are responsible for the congested peaks at low binding energies in the NIPE spectrum. The best fi t of the simulated NIPE spectrum to the experimental spectrum indicates that the ground state of HCPN 3 • is a 5 π -electron 2 A ″ π radical state, with a 6 π -electron, 2 A ′ , σ radical state being at most 1.0 kcal/mol higher in energy.