The_Catalyst_Review_January_2024 - 16

Experimental Abstracts
Machine Learning Accelerated Exploration of Ternary Organic Heterojunction Photocatalysts
for Sacrificial Hydrogen Evolution
Organic photocatalysts are intriguing materials due to their tunable properties, but their large exciton binding energies and low
charge-carrier mobilities hinder efficient charge separation. Introducing donor-acceptor (D-A) bulk heterojunctions can enhance exciton
dissociation and charge separation, thereby improving photocatalytic activity. However, the wide choice of possible donors and
acceptors and the enormous diversity of organic molecules and materials that might be combined to form donor−acceptor-acceptor
(DAA) or donor−donor−acceptor (DDA) ternary heterojunctions poses a challenge for the rational design of organic heterojunction
photocatalysts, particularly for large ternary
phase spaces. Herein, the authors present a
method for accelerating the exploration of
ternary organic heterojunction photocatalysts
(TOHP) using machine learning and highthroughput
experimental screening.
The TOHPs used in this study included
both DDA and DAA combinations. These
ternary combinations were derived from 12
conjugated polymer donors (D1−D12) and 8
molecular acceptors (A1 to A8), illustrated in
Figure 1, yielding a combinatorial space of
528 DDA-type and 336 DAA type or a total
of 864 potential TOHPs. These workers then
down selected a diverse set of 104 TOHPs
using the Kennard-Stone (KS) algorithm, which
functions by selecting samples that exhibit
the greatest dissimilarity from one another
using a distance metric or a precomputed
distance matrix as the basis for comparison.
It was shown that the 104 TOHPs exhibited a
wide range of photocatalytic activities, with
hydrogen evolution rates (HERs) varying from
zero to 737.4 mmol g−1
h−1
; most samples
(65%) had HER values below 150 mmol g−1
under 1 sun illumination.
Of these 104 candidates a series of TOHPs
with exceptionally high mass-normalized
sacrificial photocatalytic hydrogen production
activities were identified using diversityfocused
chemical space exploration,
augmented by machine learning predictions.
Of the ten most active TOHPs, all with a HER
greater than 500 mmol g−1
h−1
; only three
of these were found in the diversity-driven
selection. The remaining seven of the top
ten TOHPs, including the two most active
compositions (D1D6A4 and D1D9A4), were
found by machine learning predictions. The
rates of photocatalytic hydrogen generation
associated with these candidates are among
the highest reported for organic photocatalysts in the literature.The three top TOHPs were then selected to investigate their time course
for sacrificial photocatalytic hydrogen evolution. These materials were also compared with their associated binary combinations (Figure 2).
The most active TOHPs shared several common characteristics, such as synergistic light absorption, fluorescence quenching, and reduced
average fluorescence lifetime. These features can provide valuable insights into designing future organic heterojunction photocatalysts.
Haofan Y, Yu C, Cooper AI, et al. (2023). J. Am. Chem. Soc., https://doi.org/10.1021/jacs.3c10586
h−1
Figure 1. Donors and acceptors used to define the ternary phase space. Chemical
structures of conjugated polymer donors (a) and molecular acceptors (b) for
generating ternary organic heterojunction photocatalysts (TOHPs).
Figure 2. Photocatalysis results for selected TOHPs. Time course for sacrificial photocatalytic
hydrogen production for three selected ternary combinations and their
associated binary combinations: (a, d) D1D6A8, (b, e) D9A1A8, and (c, f) D1D9A4;
0.1 M ascorbic acid, 3 wt % H2
PtCl6
, 300 W Xe lamp, λ > 420 nm.
16
The Catalyst Review
January 2024

The_Catalyst_Review_January_2024

Table of Contents for the Digital Edition of The_Catalyst_Review_January_2024