Discovery of temperature-induced stability reversal in perovskites using high-throughput robotic learning

Zhao, Yicheng; Zhang, Jiyun; Xu, Zhengwei; Sun, Shijing; Langner, Stefan; Hartono, Noor Titan Putri; Heumueller, Thomas; Hou, Yi; Elia, Jack; Li, Ning; Matt, Gebhard J.; Du, Xiaoyan; Meng, Wei; Osvet, Andres; Zhang, Kaicheng; Stubhan, Tobias; Feng, Yexin; Hauch, Jens; Sargent, Edward H.; Buonassisi, Tonio; Brabec, Christoph J.

doi:10.1038/s41467-021-22472-x

Cited by 113 publications

(103 citation statements)

References 36 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Recent advances in accelerated and combinatorial synthesis have allowed rapid fabrication of immense number of materials' compositions in the form of combinatorial spread libraries in pulsed laser deposition, [1][2][3][4][5][6][7] high-throughput microfluidic, [8][9][10][11][12] and pipetting robot sets. [13][14][15][16][17][18] In parallel, conventional sample fabrication has been drastically accelerated via laboratory robotization and autonomous and combined human-automation workflows. [19][20][21][22][23] Complementary to the large-throughput synthesis are rapid characterization methods that allow establishing structure-property relations in the specific systems and in certain cases target functionalities of interest.…”

Section: Introductionmentioning

confidence: 99%

Hypothesis Learning in Automated Experiment: Application to Combinatorial Materials Libraries

et al. 2022

View full text Add to dashboard Cite

Machine learning is rapidly becoming an integral part of experimental physical discovery via automated and high‐throughput synthesis, and active experiments in scattering and electron/probe microscopy. This, in turn, necessitates the development of active learning methods capable of exploring relevant parameter spaces with the smallest number of steps. Here, an active learning approach based on conavigation of the hypothesis and experimental spaces is introduced. This is realized by combining the structured Gaussian processes containing probabilistic models of the possible system's behaviors (hypotheses) with reinforcement learning policy refinement (discovery). This approach closely resembles classical human‐driven physical discovery, when several alternative hypotheses realized via models with adjustable parameters are tested during an experiment. This approach is demonstrated for exploring concentration‐induced phase transitions in combinatorial libraries of Sm‐doped BiFeO3 using piezoresponse force microscopy, but it is straightforward to extend it to higher‐dimensional parameter spaces and more complex physical problems once the experimental workflow and hypothesis generation are available.

show abstract

Section: Introductionmentioning

confidence: 99%

Hypothesis Learning in Automated Experiment: Application to Combinatorial Materials Libraries

et al. 2022

View full text Add to dashboard Cite

show abstract

“…[ 16 , 17 , 24 ] Furthermore, self‐driven laboratories based on machine learning (ML) or Bayesian optimization allow to drastically reduce the number of experiments in a given parameter space necessary to improve the performance of solar cells. [ 25 , 26 ]…”

Section: Introductionmentioning

confidence: 99%

Understanding the Microstructure Formation of Polymer Films by Spontaneous Solution Spreading Coating with a High‐Throughput Engineering Platform

et al. 2021

Self Cite

View full text Add to dashboard Cite

An important step of the great achievement of organic solar cells in power conversion efficiency is the development of lowband gap polymer donors, PBDBÀ T derivatives, which present interesting aggregation effects dominating the device performance. The aggregation of polymers can be manipulated by a series of variables from a materials design and processing conditions perspective; however, optimization of film quality is a time-and energy-consuming work. Here, we introduce a robot-based high-throughput platform (HTP) that is offering automated film preparation and optical spectroscopy thin-film characterization in combination with an analysis algorithm. PM6 films are prepared by the so-called spontaneous film spreading (SFS) process, where a polymer solution is coated on a water surface. Automated acquisition of UV/Vis and photoluminescence (PL) spectra and automated extraction of morphological features is coupled to Gaussian Process Regression to exploit available experimental evidence for morphology optimization but also for hypothesis formulation and testing with respect to the underlying physical principles. The integrated spectral modeling workflow yields quantitative microstructure information by distinguishing amorphous from ordered phases and assesses the extension of amorphous versus the ordered domains. This research provides an easy to use methodology to analyze the exciton coherence length in conjugated semiconductors and will allow to optimize exciton splitting in thin film organic semiconductor layers as a function of processing.

show abstract

“…41 Increases in transmittance are therefore interpreted as conversion of MAPbI3 to degradation products. Similar approaches were recently taken to identify thermal degradation rates in various perovskites, 30,42 where the reciprocal of the time to 80% of initial absorbance T80 was taken as a simple "rate constant" with units of s -1 . Here, we use transmittance measurements to rigorously calculate the initial MAPbI3 consumption rate (𝑟 𝑀𝐴𝑃𝐼 ) in mol⋅m -2 ⋅s -1 .…”

Section: Introductionmentioning

confidence: 99%

Water-Accelerated Photo-oxidation of CH3NH3PbI3 Perovskite: Mechanism, rate orders, and rate constants

Siegler

Dunlap-Shohl

Meng

et al. 2021

Preprint

View full text Add to dashboard Cite

Understanding the chemical reactions that hybrid organic-inorganic halide perovskite (HP) semiconductors undergo in the presence of moisture, oxygen, and light are essential to the commercial development of HP solar cells and optoelectronics. Here we use optical absorbance to study the kinetics of methylammonium lead iodide (MAPbI3) degradation in response to combinations of moisture, oxygen, and illumination over a range of temperatures. We identify two primary reaction pathways that dominate MAPbI3 material degradation in these mixed environmental conditions: (1) dry photooxidation (DPO) due to the combined role of oxygen and photoexcited electrons (with a rate of 2 x 10-9 mol/m2∙s in dry air at 25 C and an effective activation energy of 0.62 eV), and (2) a water-accelerated photooxidation (WPO) process due to the combined role of water, oxygen, and photoexcited electrons (with a rate of 1 x 10-7 mol/m2s in 50% RH air at 25 C and observed effective activation energy of 0.07 eV). Commonly reported humidity-only, blue light, and thermal degradation pathways are demonstrated to have rates that are respectively 100, 1000, and >1000 times slower than predominant photooxidation processes in ambient conditions. Extracting kinetic rate constants from the dynamics of the initial degradation, we calculate that in dry air, photooxidation rate of MAPbI3 follows a f(x)∝x/(1+Kx) relationship with respect to oxygen in the vapor phase (PO2) and excess concentration of photoexcited electrons (n). In humid air, photooxidation of MAPbI3 exhibits first order kinetics with respect to the partial pressure of water in the vapor phase (PH2O). However, with respect to PO2 and n, kinetics follow a f(x)∝x/(1+kx)^2 relationship with respect to rate. We then identify a plausible reaction mechanism for degradation of MAPbI3 material that is consistent with these rate functionalities. The rate determining step for both DPO and WPO is proton abstraction by photogenerated superoxide radicals. However, proton donation by adsorbed water proceeds much more rapidly than donation by methylammonium, resulting in faster degradation rates for WPO at typical ambient conditions (~50% RH). Rate laws derived from this mechanism were fit to the entire dataset to extract rate constants for DPO and WPO processes. Accurate predictions of material degradation rates, with narrow confidence intervals of fit parameters as identified by the Bootstrap algorithm, provide the first experimental estimates of the equilibrium constants of oxygen adsorption on MAPbI3 (Keq ≈ 3 x 10-3 kPa-1) and superoxide generation from adsorbed oxygen and photoexcited electrons in MAPbI3 (Keq ≈ 5 x 10-15 (photons/m2∙s)-0.7). Given that water has been reported as a degradation product of DPO, the results reported here highlight the need for the development of encapsulation schemes that rigorously block oxygen, as over longer time periods, product water (if generated) may accumulate inside the packaging and initiate the much faster WPO process.

show abstract

Discovery of temperature-induced stability reversal in perovskites using high-throughput robotic learning

Cited by 113 publications

References 36 publications

Hypothesis Learning in Automated Experiment: Application to Combinatorial Materials Libraries

Hypothesis Learning in Automated Experiment: Application to Combinatorial Materials Libraries

Understanding the Microstructure Formation of Polymer Films by Spontaneous Solution Spreading Coating with a High‐Throughput Engineering Platform

Water-Accelerated Photo-oxidation of CH3NH3PbI3 Perovskite: Mechanism, rate orders, and rate constants

Contact Info

Product

Resources

About