Download PDF
Synthesis 2023; 55(18): 2817-2832
DOI: 10.1055/a-2072-2617
short review
Special Issue Electrochemical Organic Synthesis

Accelerated Electrosynthesis Development Enabled by High-Throughput Experimentation

Huijie Chen
a   College of Chemical and Biological Engineering, Zhejiang University, 866 Yuhangtang Road, Xihu District, Hangzhou 310027, Zhejiang, P. R. of China
,
Yiming Mo  ∗
a   College of Chemical and Biological Engineering, Zhejiang University, 866 Yuhangtang Road, Xihu District, Hangzhou 310027, Zhejiang, P. R. of China
b   ZJU-Hangzhou Global Scientific and Technological Innovation Center, Zhejiang University, Hangzhou 311215, P. R. of China
› Author Affiliations

This work was supported by the National Key R & D Program of China (No. 2021YFA1502700) and National Natural Science Foundation of China (No. 22108242).
› Further Information
 
 PDF Download Permissions and Reprints All articles of this category


Abstract

Electrochemical synthesis has recently emerged as an environmentally benign method for synthesizing value-added fine chemicals. Its unique reactivity has attracted significant interests of synthetic chemists to develop new redox chemistries. However, compared to conventional chemistry, the increased complexity caused by electrode materials, supporting electrolytes, and setup configurations create obstacles for efficient reaction discovery and optimization. The recent increasing adoption of high-throughput experimentation (HTE) in synthetic chemistry significantly expedites the synthesis development. Considering the potential of implementing HTE in electrosynthesis to tackle the challenges of increased parameter space, this short review aims at providing recent advances in the HTE technology for electrosynthesis, including electrocatalysts screening, device miniaturization, electroanalytical methods, artificial intelligence, and system integration. The discussed contents also cover some topics in HTE electrochemistry for areas other than synthetic chemistry, hoping to spark some inspirations for readers to use interdisciplinary techniques to solve challenges in synthetic electrochemistry.

1 Introduction

2 Parallelized Reaction Screening

3 High-Throughput Screening for Electrocatalysts

4 Miniaturization of Screening Devices

5 Analytical Methods for Electrosynthesis Screening

6 Artificial Intelligence for High-Throughput Screening

7 Integrated Screening Systems

8 Conclusion and Outlook


#

Key words

high-throughput screening - electrochemistry - artificial intelligence - automation - device miniaturization
1

Introduction

Electrosynthesis is an old technology with a history over 200 years.[1] Davy and Moissan used electrochemical techniques to facilitate the discovery of the chemical elements, including lithium, sodium, and fluorine.[2] [3] Kolbe developed electrolytic decarboxylative dimerization of carboxylates to synthesize alkanes in 1848,[4] regarded as the first organic electrosynthesis method. Despite that electrosynthesis has always been considered as a complementary method to conventional chemistry due to its relatively complex setup and operation, it has already been implemented in large-scale production of specialty chemicals because it can provide some unique reactivity that, otherwise, would be hard to realize with conventional chemistry. For example, the world’s largest organic electrosynthesis process is the electrochemical hydrodimerization of acrylonitrile to produce adiponitrile developed by Monsanto.[5]

Zoom Image
Huijie Chenis a Ph.D. candidate mentored by Professor Yiming Mo in College of Chemical and Biological Engineering at Zhejiang University. She received her B.S. from the School of Chemistry and Chemical Engineering Nanchang University. Her current scientific interests focus on integration of micro-devices for electrochemical high-throughput screening. Yiming Mo is an assistant professor in College of Chemical and Biological Engineering at Zhejiang University. He received his B.S. in chemical engineering from Tsinghua University and his Ph.D. in chemical engineering from Massachusetts Institute of Technology. His current research focuses on utilizing microreactors, automation, and AI algorithms to advance the development of sustainable electrosynthesis methods.

The use of electrons as the traceless reagents in electrochemistry eliminates the needs for harmful or risky oxidants and reductants employed in conventional processes, satisfying the requirements of ‘Green Chemistry’.[6] With accurate control over electrolysis current or voltage, it is possible to initiate selective electron transfer between the electrode surface and substrates. In the past decade, there is a growing interest within the synthetic community on taking the advantage of electrosynthesis to furnish sustainable and unique transformations with practical value, such as oxidations,[7] [8] reductions, [9,10] and cross-couplings.[6,11] The advances in electrosynthesis methods have been well documented in recent reviews.[12] [13] [14] In addition, with the development of continuous flow chemistry and microreactors, employing microreaction engineering techniques to enable continuous electrosynthesis improves the process efficiency, scalability, and safety.[15]

Despite its advantages, electrosynthesis is still not a routine tool used by synthetic practitioners. The possible challenges in the development and application of electrosynthesis include special reaction apparatus,[16] unique electrode design,[17] heterogenous electron and species transfer,[18] and greatly expanded reaction parameter space compared to conventional chemistry.[19] Among these challenges, the extra parameters unique to electrosynthesis, such as supporting electrolyte, electrode material, electrolysis current/voltage, and setup configuration, have significantly limited the experimental throughput of exploring and optimizing new electrochemical reaction.[19] [20]

To expedite the reaction discovery in chemistry field including electrosynthesis, high-throughput experimentation strategy has recently attracted broad attention. With the help of robotics,[21] parallel reaction and analysis,[22] [23] control software,[24] and artificial intelligence,[25–27] systems with various levels of automation and intelligence have been designed, constructed, and validated. These systems aim at increasing the throughput of reaction screening to obtain enormous amount of high-quality experimental data, thus achieving superior efficiency in tasks like reaction condition optimization,[28] compound diversification,[29] multistep synthesis,[30] and discovery of unexpected reactivities.[31] One recent notable example illustrated the utilization of data-guided matrix down-selection, uncertainty-minimizing machine learning, and robotic experimentation to discover general reaction conditions for heteroaryl Suzuki–Miyaura coupling.[32] The emergence of high-throughput experimentation (HTE) along with artificial intelligence technology provides an opportunity for electrochemists to discover and optimize new chemistries at an accelerated speed.

There have been several seminal reviews available covering various aspects of HTE for organic electrosynthesis.[17] , [33] [34] [35] The aim of this short review is to describe recent advances in the enabling screening technology for electrosynthesis, including electrocatalysts screening, screening device miniaturization, electroanalytical methods, artificial intelligence, and screening system integration. The discussed contents cover some topics in HTE electrochemistry for areas other than synthetic chemistry, hoping to spark some inspirations for readers to use interdisciplinary techniques to solve challenges in synthetic electrochemistry.


# 2

Parallelized Reaction Screening

In 1986, Pfizer used a 96-well plate approach for high-throughput screening of natural products, which can screen 800 compounds each week. This marked the beginning of HTE technology in industry.[36] In 1999, Dow Chemical Company built a HTE facility for research and development of coatings, polymers, catalysts, and other products.[37] Since then, high-throughput systems have been continually improving their screening throughput as automation and miniaturization technologies advance.[38] Fast and accurate processing of liquids and powders is made possible by the use of automated liquid/solid handlers and robotics, and these automated tools not only minimize manual intervention in the experimental process but also improve the reproducibility.[39] In 2013, Eli Lilly reported a remotely controlled, fully automated robotic synthesis facility that can screen compounds at a scale of 100 mg for 120 samples/day.[40] [41] The use of microreactors and microplates has led to a more precise and accelerated operations, a decrease in reagent usage, and an improvement in heat and mass transfer.[41] In addition, the bottleneck of high-throughput analysis is being progressively tackled by the integration of high-throughput fast-response analytical methods (e.g., ultra-performance liquid chromatograph, in-line Raman spectroscopy, in-line infrared spectroscopy, in-line mass spectroscopy). With microplates together with LC-MS analysis, Pfizer researchers have been able to screen up to 1536 Suzuki–Miyaura coupling reaction conditions per day at nanomolar levels.[42]

For electrochemical high-throughput screening, the need for electrodes leads to significantly increased complexity of the screening device compared to conventional synthesis. As a result, the throughput is limited by the number of electrodes and potentiostat channels available. Similar to microplate concept, several kinds of parallelized electrolyzers have been developed to facilitate the reaction optimization, which are introduced as follows.

Siu et al. developed a spatially addressable electrolysis platform (SAEP) (Figure [1a]), in which the reactor consists of a 16-hole Teflon block and 4 × 4 arrays of glass vials.[43] With stainless steel as cathodes and graphite as anodes, it realized parallelized Shono oxidation of carbamates. However, this system is only capable of applying the same electrolysis potential for all parallelized channels. Waldvogel et al.[44] has reported an electrochemistry screening system for divided cells (six in parallel) and undivided cells (eight in parallel) with reaction temperature control (Figure [1b]). An eight-channel DC power supply was used to regulate individual cells. A number of electrosynthesis reactions, including anodic cross-coupling reactions, dehalogenations, and deoxygenations were successfully optimized using this system.

Zoom Image
Figure 1 Selected multi-channel setups for electrosynthesis. (a) Spatially addressable electrolysis platform (SAEP) for carrying out solution-phase electrosynthesis. (b) Electrochemical screening setup developed by Waldvogel group for optimization of various electrosynthesis. Reproduced from ref. [44] Copyright 2016 American Chemical Society. (c) Fully assembled HTe –Chem device invented by Lin group for reaction optimization and discovery. Reproduced from ref.[20] Copyright 2021 CC BY-NC-ND 4.0. (d) IKA Carousel with six parallelized electrochemical reactors. Reproduced from ref.[17] Copyright 2021 American Chemical Society. (e) IKA e-Hive designed by IKA for parallel constant voltage electrolysis. Reproduced from ref.[17] Copyright 2021 American Chemical Society.

This parallelized batch electrochemical screening concept has been further advanced by Lin et al. They developed HTe –Chem (Figure [1c]), a high-throughput screening system for electrochemistry with straightforward electrode changes and small reaction volumes (200–600 μL).[20] The shape and dimensions were designed to be suitable for integration with automated reagent pipetting systems. Various electrode materials, including graphite, nickel, stainless steel, copper, titanium, magnesium, zinc, platinum, tin, and aluminum can be used with an interelectrode distance of 1.54 mm. This screening reactor is capable of performing alternating current electrolysis, constant voltage electrolysis, and constant current electrolysis. They demonstrated the system with a variety of representative electrosynthetic reactions, such as oxidative azidooxygenation, reductive silylation, and chlorination of arylboronic acids.

On the commercial side, Baran Group in partnership with IKA created ElectraSyn 2.0,[17] which greatly simplifies the required apparatus to perform electrochemical reaction screening allowing chemists with no expertise in electrochemical devices to pick up electrosynthesis. The IKA carousel converts ElectraSyn 2.0 from a single reactor to a six-reactor screening system, enabling constant current operation through six vials in series with voltage monitoring. IKA e-Hive[17] (Figure [1d]) came out later enabling the simultaneous screening of up to 24 parallel reactions, increasing the throughput of synthetic reactions. To be noted, in e-Hive, the cathode surface is 20 times bigger than the anode surface, which necessitates the re-optimization of the reaction parameters during the following scale-up.


# 3

High-Throughput Screening for Electrocatalysts

In general, electrochemical synthesis relies on the selective electron transfer between target substrates and electrodes to achieve controlled transformation. However, this can be difficult due to the lack of electron recognition between electrode and target functional groups or substrate, especially for complex molecules often encountered in synthesis applications.[45] Nevertheless, such a challenge has been greatly alleviated by the recent development of electrocatalysts.[46] Redox-active electrocatalysts can enable the generation of reaction intermediates at reduced potentials with improved chemo-,[47] [48] regio-,[49,50] and stereoselectivity.[51,52]

Homogeneous electrocatalysts (sometimes referred as mediators) are often the first option for a variety of electroorganic reactions because of their high adaptability and selectivity.[53] [54] Representative examples include TEMPO and its derivatives for selective alcohol oxidation to aldehydes,[55–57] hydrogen atom transfer (HAT) catalysts for direct functionalization of aliphatic C–H bonds,[56,58,59] nickel complexes for C–C and C–N formation reaction,[31] , [60] [61] [62] [63] and biological enzymes for CO2 reduction.[64] Finding the suitable homogeneous electrocatalysts for the targeted transformations is often time-consuming and tedious, and high-throughput screening and optimization of homogeneous electrocatalysts are conducted in a manner like traditional electrochemical synthesis processes.

An early parallel electrocatalyst screening system was demonstrated by Siu et al. It consists of four four-well grids, each of which has a counter and parallel-connected working electrodes.[65] They repeatedly scanned the applied potentials of 0.5–1.4 V (vs Ag/AgCl) to initiate electrochemical copolymerization of 2,2′-dithiophene with TEMPO catalyst precursors containing pyrrole side chains for obtaining a library of catalysts with a wide range of dithiophene/pyrrole ratios. They screened the electrocatalysts against electrochemical oxidation of primary alcohols to aldehydes (Figure [2a]). This approach enables rapid synthesis and screening of catalytic materials on electrode surfaces.

Zoom Image
Figure 2 High-throughput screening for homogeneous electrocatalysts. (a) Electrochemical co-polymerization for immobilized catalyst preparation and screening. Reproduced from ref.[65] Copyright 2000 American Chemical Society. (b) A microfluidic device for automated kinetic measurements of allylic C–H oxidation electrocatalysts. Reproduced from ref.[19] Copyright 2020 Wiley-VCH GmbH. (c) A screening setup for photoelectrochemical catalyst applied for C–H amination. Reproduced from ref.[20] Copyright 2021 CC BY-NC-ND 4.0. (d) Application of HTe –Chem for redox mediator screening. Reproduced from ref.[20] Copyright 2021 CC BY-NC-ND 4.0.

Mo et al. developed a microfluidic device using interdigitated electrodes (IDEs) providing automated kinetics measurement of electrocatalysts (Figure [2b]).[19] Ingeniously, the authors configured a cyclic voltammetry (CV) flow cell with a 0.75 mm deep pocket in front of the working electrode in order to use the classical CV analytical equation[66] for automated electroanalysis. The measured kinetic constants were compared to those obtained from a traditional cyclic voltammetry cell to ensure validity and repeatability. Fast kinetic measurements of TEMPO-catalyzed alcohol oxidation and anodic NHPI-mediated allylic C–H oxidation kinetics were accomplished by using this microscale cyclic voltammetry flow cell. This droplet-based electrochemical screening methodology only requires 15 μL analyte solution for single measurement, which greatly reduces the reagent cost during the kinetic studies.

In addition, high-throughput screening of electrophotocatalysts has been accomplished with the help of HTe –Chem (Figure [2c]).[20] The C–H bonded amination reaction catalyzed by trisaminocyclopropenium radical dication was optimized on the HTe –Chem platform.[67] Three supporting electrolytes, three voltages, and two reaction times were screened on 24-well plates. After screening, an improved reaction condition was found which increased the conversion rate from 71% (72 h) to 78% (24 h). To further examine the reproducibility of the HTe –Chem platform, the identical reactions were repeated in a 24-well plate with an average yield of 75% and a standard deviation of only 5.2%.

Ruccolo et al. developed a method to use electricity to regenerate adenosine triphosphate (ATP) for complex enzymatic synthesis processes (Figure [2d]).[68] To find the optimal mediator for ATP regeneration, they first used cyclic voltammetry to determine the mechanism of the catalytic process before employing the HTe –Chem platform to screen 15 ferredoxene derivatives as the redox mediator. By monitoring ATP production, the greatest conversion rate was achieved when ferrocenemethanol (FcMeOH) is used. This optimized condition facilitated ADP to ATP conversion in the scaled-up tests at an ideal yield of 87%. The method is simple, robust, and scalable, avoiding usage of high-energy stoichiometric reagents.

Although homogeneous electrocatalysts have the advantages of remarkable selectivity and easy implementation, one notable challenge in practical large-scale synthesis is the separation and recycling of homogeneous catalysts after the reaction.[69] Heterogeneous catalysis accounts for around 3/4 of commercial catalytic chemical and petrochemical processes.[70] Similarly, it is desired to develop heterogenous electrocatalysts for electroorganic synthesis applications. However, current focus of electroorganic synthesis development is still on the solution chemistry, and electrodes are mainly considered as an interface for electron transfer. Despite screening electrode materials is a routine task when optimizing electrosynthesis conditions, designing catalytic electrode materials is still at its infancy currently. A seminal work from Baran group discovered the unique catalytic effect of in situ prepared silver-nickel nanoparticles on cathode for decarboxylative arylation.[71] Due to the early stage of heterogeneous electrocatalysts for organic synthesis, there are few works available on the topic of high-throughput screening for heterogeneous synthetic electrocatalysts. The following introduces some efforts on electrocatalysts HTE for CO2 reduction and water splitting in order to spark inspirations for future development in organic synthesis applications.

To find the optimal electrocatalysts for converting CO2 into value-added products, traditional manual approach only allows a screening throughput of 10 catalysts per day. Lai et al. employed a sputter system to deposit Cu and Pd-Zn thin film electrocatalysts (Figure [3a]).[72] The multi-source configuration in the sputter system was able to generate a continuous composition gradient across the electrode substrate, thus obtaining a variety of electrocatalyst with different compositions. To achieve rapid evaluation of performance of the synthesized electrocatalysts, they constructed a custom in-line mass spectroscopy for quantification of products. These two techniques combined significantly sped up the evaluation of electrocatalyst arrays for CO2 reduction by enabling the quasi-real-time measurement of the generated hydrogen, methane, and ethylene.

Zoom Image
Figure 3 Heterogeneous electrochemical catalyst screening device. (a) High-throughput screening of electrocatalyst arrays for CO2 reduction. Reproduced from ref.[72] Copyright 2019 American Chemical Society. (b) Bubble imaging method for high-throughput characterization of catalysts for OER and HER reactions. Reproduced from ref.[74] Copyright 2022 American Chemical Society.

The electrocatalyst characterization is usually the rate-limiting step in HTE. The recently developed bubble imaging method for characterizing oxygen or hydrogen evolution electrocatalysts greatly expedite the screening process.[73] For example, Liu et al. implemented multi-target magnetron sputtering technique to create ternary alloy films with various compositions, and characterized the catalyst activity by imaging the formation rate of evolved hydrogen gas bubbles (Figure [3b]).[74] A camera was utilized to capture the progression of the bubbles as the catalyst was kept under a specific electrochemical potential. The procedure is also applicable for screening oxygen evolution reaction (OER) catalysts.

Additionally, in silico screening of heterogeneous electrocatalysts can provide valuable information for catalyst design and reduce number of experiments required to find the optimal catalyst. For instance, Pathak et al. proposed an automated electrocatalyst screening workflow for CO2 reduction to methanol using machine learning (ML) and density function theory (DFT) calculations.[75] They selected Cu, Co, Ni, Zn, and Mg to create diverse alloy-based compositions. By using the microstructure model and machine learning algorithm, they have found 7 active catalysts from 495 alloy combinations.

In contrast to conventional one-factor-at-a-time techniques,[76] HTE has offered a revolutionary paradigm for rapid exploration of electrocatalysts.[39] Due to the complexity of electroorganic synthesis, current screening throughput of electrocatalysts is relatively low compared to what is regarded as high throughput in the pharmaceutical or fine chemical applications.[77] Current examples are still limited to homogeneous electrocatalysts, and heterogeneous electrocatalysts are a promising future direction, which requires extensive development to achieve selective transformations.


# 4

Miniaturization of Screening Devices

The target of HTE is rapid screening of a large number of compounds or conditions.[78] Miniaturization of the reaction and analysis modules is one of the key technologies for advancing modern HTE systems. Miniaturization typically comes with three significant benefits: lower reagent cost, faster turnaround, and reduced space requirement.[79] Thus, the core technology determining HTE system’s throughput is the accuracy to operate with miniaturized screening modules and the ability to expedite the screening using microscale process intensification mechanisms.[80 Electroorganic synthesis is a multiphase chemistry requiring species to transport across electrolyte between two electrodes. Minimizing the electrodes and their inter-electrode distances can enhance the inter-electrode material exchange.[81] In addition, the number of electrons required to complete an electrochemical reaction is proportional to moles of substrate, and, thus, reducing the amount of substrate used during the screening can reduce needed electrons, which can shorten the reaction time.[82] Thus, developing the miniaturization tools for electrochemistry screening is crucial to achieve a throughput that, otherwise, is hard to realize using conventional devices.[83]

The electrochemical microreactor is a representative technology for miniaturization, and it has been extensively used for continuous flow electrochemical synthesis. Many electrochemical reactions can benefit from the unique properties of microreactors, such as high electrode surface-to-volume ratio, low resistance, excellent heat and mass transfer, and accurate residence time control.[84] Yoshida et al. used an electrochemical microreactor to generate unstable carbon cation intermediates under extremely low temperature, and subsequentially mixed them with nucleophiles to form corresponding C(sp3)–C(sp3) bonds. This ‘cation flow’ strategy avoided the undesired oxidation of nucleophiles when generating highly energetic cation intermediates.[85] Mo et al. pioneered a microfluidic redox-neutral electrochemistry (μRN-eChem) system that relied on rapid interelectrode diffusion to avoid the decomposition of reactive radical intermediates generated from electrode surface. These radical intermediates then cross coupled in an efficient manner to form C(sp2)–C(sp3) bonds.[86] Electrochemical microreactors can be scaled-up to produce bulk chemicals. A notable example is that BASF uses thin-gap flow cell with a 250 μm interelectrode distance to produce adiponitrile in an annual sub-megaton scale.[87] [88] The recent development of continuous flow electrochemical microreactors have been reviewed and discussed in detail in several reviews.[89–91]

In addition to the benefits for continuous electrosynthesis, electrochemical device miniaturization can potentially improve the efficiency for electrochemistry HTE. Considering the relatively recent emergence of electrosynthesis, the examples of using miniaturized electrodes are still limited. Moeller et al. developed a strategy to use electrochemically generated organometallic reagents for catalyzing the surface functionalization reactions. With potential-controlled generation of confinement reagent, this strategy can selectively functionalize certain electrodes in the microelectrode array.[92] [93] [94] Mo et al. microfabricated an interdigitated electrode (IDEs) with interelectrode distance of 10 μm, enabling sub-second molecular diffusion between cathode and anode (Figure [4a]).[95] This intensified mass transfer allowed a high current density even when electrolyzing a 15 μL reagent solution. The implementation of IDEs significantly expedites the screening process of finding the optimal condition for α-amino C–H arylation reactions. Despite the limited examples available that are directly related to electrosynthesis HTE, there are extensive efforts that employed miniaturized electrodes for electrochemical sensing and electroanalysis, which should provide inspirations for future electrosynthesis HTE miniature device design.

Microelectrodes or ultramicroelectrodes are often used as working electrode in electroanalysis.[96] They have a diameter between 10 and 200 μm, and are made out of extremely thin wires or fibers that are enclosed in a glass tube. The small size results in greatly diminished double-layer capacitance and a relatively large diffusion, which enables measurement of extremely fast reaction kinetics during reaction mechanism studies.[97]

In electroanalytical applications, microelectrodes with minimal or negligible ohmic voltage losses are growing in popularity.[98] However, large size reference electrodes are insufficient for the manufacturing of tiny size cells.[99] Schuett et al. offered a straightforward approach for fabricating 3D-printed micro-reference electrodes (Figure [4b]).[100] The chemically resistant polyvinylidene fluoride was used to 3D print the micro-reference electrodes (REs), which have an outer diameter of 4 mm and a length of 40 mm. The electrodes are sealed with a magnesia stick to ensure sealing. Additionally, the compact size permits their usage in electrochemical micro cells [e.g., in situ scanning tunneling microscopy (STM) characterization]. This miniature reference electrode will be important in electrochemical miniature cells or other applications.

Zoom Image
Figure 4 Electrochemical device miniaturization and their applications. (a) The IDE chips combined with flow cells for electrocatalytic reactions. Reproduced from ref.[19] Copyright 2020 Wiley-VCH GmbH. (b) The miniature cell with the micro-REs used for the small-size applications. Reproduced from ref.[100] Copyright 2021 CC BY-NC-ND 4.0. (c) Interdigitated nano-gap electrodes for volatile organic compounds detection. Reproduced from ref.[101] Copyright 2017 The Royal Society of Chemistry. (d) Multiplex electrochemical biosensor for protein detection. Reproduced from ref.[104] Copyright 2015 with permission from American Chemical Society. And 3D-printed array. Reproduced from ref.[105] Copyright 2019 PubMed Central.

Detection of volatile organic compounds (VOCs) is essential for environmental applications. Interdigitated nanogap electrodes (IDEs) can be used as one of the approaches to detect VOCs. Nguyen Minh et al. developed an interdigitated electrodes with gaps from 50–250 nm on a silicon wafer (Figure [4c]).[101] These nanogap IDEs were then coated with poly(4-vinylphenol) as the sensing molecular layer for acetone detection. These sensors showed excellent sensing performance with a dynamic range from 10 ppm to 1000 ppm of acetone at room temperature.

Integrated microelectrode arrays have important applications in electrochemical biosensors.[102] Multiplexed protein assays used in point-of-care diagnostics need be operated in a simple, automated, and low-cost manner with high throughput.[103] Kadimisetty et al. described an inexpensive automated multiplexed protein immunoarray (Figure [4d]).[104] The microelectrode array used for detection is made of pyrolytic graphite chip working electrode, Ag/AgCl reference electrode, and stainless steel counter electrode. Four prostate cancer biomarker proteins can be detected simultaneously by the array in a 36-minute test with extremely low detection limits. In 2019, they combined 3D printing with optimization of this gadget (Figure [4d]). In comparison to its predecessor, the immunoassay array allows for the assessment of 8 biomarker proteins in 25 minutes with a low limits of detection (LOD) of 85–110 fg/mL.[105]


# 5

Analytical Methods for Electrosynthesis Screening

The analysis of reaction outcome is often the bottleneck that limits the throughput of HTE. Therefore, it is essential to employ the appropriate analysis and characterization method according to the screening target.[106] Since, for most of the time, the screening target for electrosynthesis is reaction yield, selectivity, or catalyst turnover, the characterization methods used for conventional organic synthesis can also be used for electrosynthesis HTE, including gas/liquid chromatography,[107] nuclear magnetic resonance,[107] and mass spectrometry.[23]

Due to the unique requirement of applying external current/voltage during electrosynthesis, many electrochemical analysis methods can be used to provide thermodynamic and kinetic information of the heterogeneous electron transfer. Common electroanalysis techniques are listed in Table [1]. These methods are relatively straightforward to implement and can characterize transient intermediates to understand the reaction mechanism.[108] In the following, cyclic voltammetry (CV) and its application is introduced, and the descriptions of additional electroanalysis methods can be found in Table [1] and their corresponding references. CV can provide information about the feasibility of redox process and determine the potential and current range of synthesis reaction.[109] Mo et al. used CV in an electrochemical HTE platform to measure the kinetics of TEMPO-catalyzed alcohol oxidations and NHPI-mediated allylic C–H bond oxidations.[19] It was achieved by varying the substrate concentration to obtain the corresponding catalytic plateau current. However, the voltammograms typically require manual analysis, which is incompatible with automated high-throughput screening. Hoar et al. applied a residual neural network (ResNet) to automatically analyze cyclic voltammograms and assign the most probable homogeneous molecular electrochemistry to the measured voltammogram.[110] Automatic voltammogram analysis will help to analyze complex electrochemical systems, and it is expected to conduct independent high-throughput electrochemical research with minimal human interference.

In the area of high-throughput analysis, additional types of voltammetry, including fast scanning cyclic voltammetry,[111] hydrodynamic voltammetry,[111] pulse voltammetry,[112] and anodic stripping voltammetry (ASV)[113] have also progressively been used in HTE. In addition, traditional electrochemical analytical methods like chronoamperometry,[114] and chronopotentiometry,[115] electrochemical impedance spectroscopy[116] have drawn researchers’ interest over time.

Table 1 Examples of Electrochemical Analytical Techniques

Category

Methods

Characteristics

References

Voltammetry

Cyclic voltammetry (CV)

CV is a method that records the current by scanning the potential back and forth between selected limits. It can gather quantitative and qualitative data on electrochemical processes, including redox potential, kinetics, reversibility, and reaction mechanisms.

117,118

Fast scan cyclic voltammetry (FSCV)

Compared to normal CV, FSCV uses (ultra)microelectrode and extremely large voltage scan rate (up to 106 V/s) to study short-lived intermediates (lifetime <milliseconds). Additionally, FSCV requires extremely small electrolyte volume, thus gaining popularity in the micro-bioelectrochemical analysis field.

119,120

Hydrodynamic voltammetry

Hydrodynamic voltammetry uses convection to enhance the rate of mass transport to the electrode, resulting in increased current and sensitivity in comparison to voltammetric measurements performed in stagnant solutions. Analysis of the current response can be performed under well-defined mass transport conditions.

121

Differential pulse voltammetry (DPV)

In DPV technique, the current is sampled periodically as a series of potential pulses is applied. DPV can measure Faradaic current free of background capacitive current. DPV is commonly employed to study the redox properties of analytes with low concentrations.

119

Stripping voltammetry (SV)

SV technique pre-concentrates analytes onto the working electrode, followed by a potential scan during which the analyte is stripped back to the solution. SV is highly sensitive (detection limits on the order of 10–12–10–10 mol/L) and is commonly for quantitative elemental detection.

122

Amperometry

Chronoamperometry (CA)

CA applies stepped electrode potential and the resulting current from faradaic processes is monitored as a function of time. CA can be used to measure the current-time dependence of the diffusion-controlled processes occurring at the electrodes.

123,124

Coulometry

Chronocoulometry (CC)

CC is a controlled-current technique that measures potential variation with time following programmed current steps. CC is used for measuring electrode surface area, diffusion coefficient, the time window of an electrochemical cell, adsorption of electroactive species, kinetics of chemical reactions coupled to electron transfer reactions.

125

Impedance

Electrochemical impedance spectroscopy (EIS)

EIS applies small sinusoidal voltage perturbation of different frequencies to an electrochemical system to measure the corresponding current response. The equivalent circuit can be constructed with EIS results, helping evaluation of electrochemical mechanisms at the electrode interface and within the electrolytic solution. EIS is used to investigate mass-transfer, charge-transfer, and diffusion processes.

126

# 6

Artificial Intelligence for High-Throughput Screening

The primary goal of HTE is to identify potential candidates through large-size experimental library screening. This is usually achieved using robotics and multichannel reaction/analysis apparatus. However, the parameter space of an electrochemical transformation is enormous, which makes it impractical, even with HTE, to enumerate through all parameter combinations.[127] Artificial intelligence has recently emerged as a tool to help HTE with chemistry knowledge extraction from HTE data and guiding HTE system to search across parameter space.[128]

Three major tasks in artificial intelligence-assisted high-throughput reaction screening are reaction prediction, catalyst design, and self-driven optimization of reaction conditions.

An accurate and qualitative prediction of the synthetic reaction is highly desired when designing and evaluating synthetic routes.[129] Ahneman et al. developed a random forest model achieving a high Pearson correlation for predicting the yields of C–N cross-coupling reactions.[130] Granda et al. used the AI algorithm for the exploration of Suzuki–Miyaura reaction space with the liquid-handling robot.[31] By utilizing only 10% of the total condition combinations, the outcomes of the remaining 90% can predicted with the accuracy over 80%.

The design of catalysts in organic synthesis still heavily relies on empirical understandings from chemists.[131] Modern machine learning methods can find patterns in large sets of data that are sometimes challenging to recognize by humans.[132] Discovering these structure-activity relationships may facilitate catalyst discovery and enabling the rapid optimization of catalytic transformations.[131] Zahrt et al. reported a computationally guided workflow for chiral catalyst selection.[131] They designed a set of robust molecular descriptors allowing to obtain a highly accurate AI model for predicting stereoselectivity of phosphoric acid-catalyzed thiol addition to N-acylimines.

Self-driven automated platforms have been demonstrated in several studies to significantly speed up the discovery of compounds and materials.[27] [133] For instance, MacLeod­ et al. reported a modular robotic platform driven by a model-based optimization algorithm capable of autonomously optimizing the optical and electronic properties of thin-film materials by modifying the film composition and processing conditions.[134] Shields et al. applied Bayesian optimization for palladium-catalyzed direct arylation reaction, their findings demonstrate that Bayesian optimization outperforms human decision making in both average optimization efficiency and consistency.[28]

Considering the exciting progress of utilizing AI for synthetic chemistry and HTE, it is believed that extensive implementation of AI to facilitate the development of electrosynthesis will emerge soon.[35] The following introduces a few existing data-driven approaches for electrochemical screening.

Traditionally, electrosynthesis conditions are optimized using the one variable at a time (OVAT) approach, where only one variable is modified while others are held constant. This approach not only finds the local optimal solution most of the time, but it is also inefficient. The application of design of experiments (DoE) method has solved this problem to some extent. By utilizing the DoE approach, Dörr et al. were able to accelerate the condition optimization for electrochemical (hetero)arenes C–H functionalization to form 1,1,1,3,3,3-hexafluoroisopropanol (HFIP) aryl ethers (Figure [5a]).[135] The parameter space of the phenol-arene cross-coupling process was studied by Hielscher et al. using DoE techniques.[138] High yields of up to 85% are possible with selective synthesis under the ideal ortho- and para-coupling conditions. And high current densities of 40 mA/cm2 can significantly reduce the response time. Seidler et al. utilized the DoE approach to not only improve the reaction parameters for the electrochemical reduction of l-cystine to l-cysteine, but also to obtain high repeatability of the screening findings when the flow cell area was enlarged to 10 times the original size.[139]

Zoom Image
Figure 5 Data-driven optimization strategies for electrochemical reactions. (a) DoE approach for accelerating the condition optimization of electrochemistry esterification. Reproduced from ref.[135] Copyright 2021 Wiley-VCH GmbH. (b) The artificial neural network for exploring the pulsed electrolysis parameters for electrosynthesis of adiponitrile. Reproduced from ref.[136] Copyright 2019 PNAS license. (c) Gaussian process with reaction fingerprints for redox potential prediction of carbonyl reductions. Reproduced from ref.[137] Copyright 2019 American Chemical Society.

Controlling the selectivity and obtaining high energy conversion efficiency at high current densities is the key target of electrosynthesis. Blancothe et al. developed a pulsed electrolysis method to balance the mass transfer rate and reaction rate for improved control over reaction performance (Figure [5b]).[136] They explored the pulsed electrolysis parameters for electrosynthesis of adiponitrile and fed the data into an artificial neural network. The neural model was able to identify the global optimal operation parameters, which was subsequently validated by experiment to achieve 30% and 325% improvement in adiponitrile production rate and selectivity compared to traditional direct current electrolysis, respectively.

However, scaling to higher dimensions can slow down the DoE process, and new methods are required to improve the efficiency.[140] The combination of automation and machine learning has shown considerable promise for speeding up screening by an order of magnitude.[141] A machine learning-based electrolyte metering/mixing/evaluation test stand was designed and assembled by Whitacre et al.[142] The test stand was used to query several 2-dimensional electrolyte search spaces discovering novel binary electrolyte solution blends such as aqueous LiNO3 and NaNO3 blends with voltage stabilities exceeding 2.8 V. With the help of Bayesian optimization, it took less than a day conduct these searches while conventional manual methods would likely have taken much more time. Implementation of a feedback loop enables autonomous optimization procedures using active learning algorithms.[143] Naito et al. aimed at exploring and determining the appropriate conditions for amino acid synthesis by electrochemical carboxylation.[144] They implemented a Bayesian optimization algorithm to find the suitable conditions for the electrochemical carboxylation of an imine to α-amino acids in a flow microreactor with a small number of experiments, demonstrating its potential as a tool for multiparameter screening.

The recent advances in machine learning for organic synthesis has resulted in a range of application-oriented neural network model and molecule encoders which has greatly improved the prediction performance. Jinich et al. used semi-empirical quantum chemistry calculations calibrated with Gaussian process for predicting the redox potentials of carbonyl functional group reductions to alcohols and amines (Figure [5c]).[137] This model is able to perform high-throughput virtual screening by predicting the standard potentials of more than 315k redox reactions involving approximately 70k compounds. Chen et al. constructed electro-descriptors including onset potential, Tafel slope, and effective voltage as inputs of machine learning algorithms.[145] This electro-descriptor approach successfully applies to three electro-organic reactions with different pathways, enabling accurate prediction of the reaction yield of unknown substrate and conditions.


# 7

Integrated Screening Systems

Integrating reagent preparation station, reactors, analysis tools, control software, and artificial intelligence algorithms to form a complete system is the final step to execute large-scale screening experiments in a fully automated and self-driven manner (Figure [6]).[146] The modular system construction strategy allows the HTE platform to be adopted in different screening scenarios, thus improving its applicability and generality.[147] According to the difference in how sample is handled, integrated systems typically fall into two categories: batch and flow systems. Batch system can execute parallelized reaction or analysis with the help of automated robotics, while, in comparison, flow system has the ability to run safely under extreme temperature or pressure condition to expedite the reaction process.[148] [149] Integrated modular reaction screening systems have been employed in a variety of fields, including synthetic biology,[150,151] organic synthesis,[39] , [152–154] and material discovery.[134] , [155] [156] [157] Due to the complexity introduced by the multiphase process of electrosynthesis, there are only few fully integrated HTE systems for electrosynthesis that have been demonstrated.[21]

Zoom Image
Figure 6 Schematics of a self-driven modular electrochemistry screening system

Discovering new electrochemical reactions is the most important goal for synthetic chemists who work on synthetic electrochemistry. However, finding unexpected redox reactivities relies on accidental discovery of unexpected products. To expedite the process of new electrosynthesis reaction discovery, Zahrt et al. constructed a machine-learning-guided reaction discovery workflow that integrates a reactivity prediction model and a microfluidic electrochemical screening platform (Figure [7a]).[95] [158] With this workflow, 38 865 molecules were pre-screened using the reactivity prediction model before the experiment, 824 of them were identified as potentially reactive. When experimentally testing the selected 20 molecules, 80% showed redox reactivity, demonstrating the prediction accuracy of the reactivity prediction model.

Most of the existing automated synthesis platforms execute experimental plans according to predefined procedures.[159] [160] Indeed, only with ‘AI brain’ can the automated system truly be self-driven to conduct all-round scientific research.[160] Along this line, Zhu et al. have built an all-round AI-chemistry laboratory with three core modules: a scientific text data-mining system using natural language processing (NLP), a mobile robot experimental module, and theoretical prediction models (Figure [7b]).[161] This robot can complete the full process of synthesis, characterization, and testing as well as independently generate new hypotheses and experimental designs and carry out the subsequent round of chemical experiments. The AI chemist’s closed-loop iterative design has demonstrated to search for efficient high-entropy metal electrocatalysts. It first used NLP algorithm to sort metal element recommendations obtained from ~16 000 papers. Five non-nobal metal elements were selected to construct 207 trial and error experiments, which were fully executed by the mobile robot. And two catalyst compositions were found showing the lowest overpotential for oxygen evolution reaction (OER).

For energy related electrochemistry, 2,5-di-tert-butyl-1,4-bis(2-methoxyethoxy)benzene (DBBB) is a promising active species for non-aqueous redox flow batteries, but its practical use requires extensive optimization of electrolytes. Su et al. employed an automated synthesis and characterization platform to optimize the performance for DBBB flow battery.[162] The characterization module contains an optical camera, a temperature controller, and conductivity meter. They explored various combinations of alkali ion salts and carbonate solvents with the goal of improving DBBB solubility and ionic conductivity. In addition, systematic electrochemical studies were performed on optimized electrolytes to determine trends in key parameters that impact the performance of redox flow batteries.

Zoom Image
Figure 7 (a) Modular workflow of machine-learning-guided reaction discovery system. Reproduced from ref.[158] Copyright 2022 The Authors. Published by American Chemical Society. (b) The key modules of an all-round AI-chemistry laboratory for electrocatalyst discovery and optimization. Reproduced from ref.[161] Copyright 2022 Oxford University Press.

# 8

Conclusion and Outlook

This short review summarizes the recent development of high-throughput experimentation for electrosynthesis. Electrosynthesis is still niche method, but it is becoming increasingly popular among synthetic chemists. Implementing the HTE method to facilitate the development of electrosynthesis can mitigate the challenges caused by additional reaction parameters compared to conventional chemistry. Despite that HTE for electrosynthesis is still an emerging area and there are only a few examples available, the technologies in other relevant electrochemistry areas can be adopted and used to increase the screening throughput. It is believed that there will be increasing amount of interdisciplinary work that combines several aspects among device engineering, electrocatalyst synthesis, electroanalytical methods, control software, and machine learning to illustrate the potential of HTE for accelerating the adoption, development, and application of electrosynthesis. We envision that the following aspects will receive extensive attention in the future:

(1) Efficient and easy-to-use apparatus. The existing HTE devices for electrosynthesis have already greatly streamlined the procedure of conducting parallelized electrochemical reactions. However, a large portion of organic reactions require oxygen- and/or moisture-free conditions, which creates obstacles to conduct parallelized reaction screening in existing devices if without a glove box. Thus, we would foresee that simplifying the device configuration and electric connections is one of the major directions for electrosynthesis device development. For example, using integrated circuit board and wireless electricity transfer technology to eliminate the needs for electrode wire connections can significantly avoid the tedious and error-prone electrode wire connection through vial caps.

(2) Heterogeneous electrocatalyst screening. Despite that majority of the electrocatalysts used in existing electrosynthetic methods are homogeneous electrocatalysts, functionalization of electrodes with heterogeneous electrocatalysts offers a new dimension for reaction selectivity control and facile catalyst recycling. However, the screening of heterogeneous electrocatalysts involves catalyst synthesis, immobilization, characterization, and testing, the procedure of which is much more complicated than that of homogeneous electrocatalysts. Designing corresponding HTE devices will be of importance to facilitate the development of heterogeneous electrocatalysts.

(3) Database for electrosynthesis. Despite the rapid emergence of new electrosynthesis reactions, there lacks a centralized database built for electrosynthesis. Thus, it is challenging for using artificial intelligence algorithms to learn the underlying rules of how to design and implement electrochemical transformations, thus, making it hard to achieve autonomous electrosynthesis reaction discovery using automated systems. The recent leap of artificial general intelligence (e.g., ChatGPT) opens the possibility to use AI algorithms to summarize chemistry literature and extract reaction data, making it easy to build a specialized database for electrosynthesis.


#
#

Conflict of Interest

The authors declare no conflict of interest.


Corresponding Author

Yiming Mo
ZJU-Hangzhou Global Scientific and Technological Innovation Center, Zhejiang University
Hangzhou 311215
P. R. of China   

Publication History

Received: 16 December 2022

Accepted after revision: 12 April 2023

Accepted Manuscript online:
12 April 2023

Article published online:
16 May 2023

© 2023. Thieme. All rights reserved

Georg Thieme Verlag KG
Rüdigerstraße 14, 70469 Stuttgart, Germany


   Permissions and Reprints All articles of this category
Zoom Image
Huijie Chenis a Ph.D. candidate mentored by Professor Yiming Mo in College of Chemical and Biological Engineering at Zhejiang University. She received her B.S. from the School of Chemistry and Chemical Engineering Nanchang University. Her current scientific interests focus on integration of micro-devices for electrochemical high-throughput screening. Yiming Mo is an assistant professor in College of Chemical and Biological Engineering at Zhejiang University. He received his B.S. in chemical engineering from Tsinghua University and his Ph.D. in chemical engineering from Massachusetts Institute of Technology. His current research focuses on utilizing microreactors, automation, and AI algorithms to advance the development of sustainable electrosynthesis methods.
Zoom Image
Figure 1 Selected multi-channel setups for electrosynthesis. (a) Spatially addressable electrolysis platform (SAEP) for carrying out solution-phase electrosynthesis. (b) Electrochemical screening setup developed by Waldvogel group for optimization of various electrosynthesis. Reproduced from ref. [44] Copyright 2016 American Chemical Society. (c) Fully assembled HTe –Chem device invented by Lin group for reaction optimization and discovery. Reproduced from ref.[20] Copyright 2021 CC BY-NC-ND 4.0. (d) IKA Carousel with six parallelized electrochemical reactors. Reproduced from ref.[17] Copyright 2021 American Chemical Society. (e) IKA e-Hive designed by IKA for parallel constant voltage electrolysis. Reproduced from ref.[17] Copyright 2021 American Chemical Society.
Zoom Image
Figure 2 High-throughput screening for homogeneous electrocatalysts. (a) Electrochemical co-polymerization for immobilized catalyst preparation and screening. Reproduced from ref.[65] Copyright 2000 American Chemical Society. (b) A microfluidic device for automated kinetic measurements of allylic C–H oxidation electrocatalysts. Reproduced from ref.[19] Copyright 2020 Wiley-VCH GmbH. (c) A screening setup for photoelectrochemical catalyst applied for C–H amination. Reproduced from ref.[20] Copyright 2021 CC BY-NC-ND 4.0. (d) Application of HTe –Chem for redox mediator screening. Reproduced from ref.[20] Copyright 2021 CC BY-NC-ND 4.0.
Zoom Image
Figure 3 Heterogeneous electrochemical catalyst screening device. (a) High-throughput screening of electrocatalyst arrays for CO2 reduction. Reproduced from ref.[72] Copyright 2019 American Chemical Society. (b) Bubble imaging method for high-throughput characterization of catalysts for OER and HER reactions. Reproduced from ref.[74] Copyright 2022 American Chemical Society.
Zoom Image
Figure 4 Electrochemical device miniaturization and their applications. (a) The IDE chips combined with flow cells for electrocatalytic reactions. Reproduced from ref.[19] Copyright 2020 Wiley-VCH GmbH. (b) The miniature cell with the micro-REs used for the small-size applications. Reproduced from ref.[100] Copyright 2021 CC BY-NC-ND 4.0. (c) Interdigitated nano-gap electrodes for volatile organic compounds detection. Reproduced from ref.[101] Copyright 2017 The Royal Society of Chemistry. (d) Multiplex electrochemical biosensor for protein detection. Reproduced from ref.[104] Copyright 2015 with permission from American Chemical Society. And 3D-printed array. Reproduced from ref.[105] Copyright 2019 PubMed Central.
Zoom Image
Figure 5 Data-driven optimization strategies for electrochemical reactions. (a) DoE approach for accelerating the condition optimization of electrochemistry esterification. Reproduced from ref.[135] Copyright 2021 Wiley-VCH GmbH. (b) The artificial neural network for exploring the pulsed electrolysis parameters for electrosynthesis of adiponitrile. Reproduced from ref.[136] Copyright 2019 PNAS license. (c) Gaussian process with reaction fingerprints for redox potential prediction of carbonyl reductions. Reproduced from ref.[137] Copyright 2019 American Chemical Society.
Zoom Image
Figure 6 Schematics of a self-driven modular electrochemistry screening system
Zoom Image
Figure 7 (a) Modular workflow of machine-learning-guided reaction discovery system. Reproduced from ref.[158] Copyright 2022 The Authors. Published by American Chemical Society. (b) The key modules of an all-round AI-chemistry laboratory for electrocatalyst discovery and optimization. Reproduced from ref.[161] Copyright 2022 Oxford University Press.