!"
"
Current understanding and challenges of polymeric !"
photocatalysts for solar-driven hydrogen generation #"
$"
Yiou Wang
a
, Anastasia Vogel
b
, Michael Sachs
c
, Reiner Sebastian Sprick
b
, Liam Wilbraham
d
, Savio J. %"
A. Moniz
a
, Robert Godin
c,e
, Martijn A. Zwijnenburg
d*
, James R. Durrant
c*
, Andrew I. Cooper
b*
, &"
Junwang Tang
a*
'"
a. Department of Chemical Engineering, University College London, Torrington Place, London, WC1E 7JE, UK. Email: ("
junwang.tang@ucl.ac.uk )"
b. Department of Chemistry and Centre for Materials Discovery, University of Liverpool, Crown Street, Liverpool, L69 7ZD, UK. E-mail: *"
aicooper@liverpool.ac.uk !+"
c. Department of Chemistry and Centre for Plastic Electronics, Imperial College London, South Kensington Campus, London SW7 2AZ, !!"
U.K. E-mail: j.durrant@imperial.ac.uk !#"
d. Department of Chemistry, University College London, 20 Gordon Street, London WC1H 0AJ, U.K. E-mail: m.zwijnenburg@ucl.ac.uk !$"
e. Department of Chemistry, The University of British Columbia, Okanagan Campus, 3247 University Way, Kelowna, British Columbia !%"
V1V 1V7, Canada !&"
!'"
The use of hydrogen as a fuel, when generated from water using semiconductor photocatalysts and !("
driven by sunlight, is a sustainable alternative to fossil fuels. Polymeric photocatalysts are based on !)"
earth-abundant elements and have the advantage over their inorganic counterparts that their !*"
electronic properties are easily tuneable through molecular engineering. Polymeric photocatalysts #+"
have developed rapidly over the last decade, resulting in the discovery of many active materials. #!"
However, our understanding of the key properties underlying their photoinitiated redox processes ##"
has not kept pace, and this impedes further progress to generate cost-competitive technologies. Here, #$ "
we discuss state of the art polymeric photocatalysts and our microscopic understanding of their #%"
activities. We conclude with a discussion of five outstanding challenges in this field: non-#&"
standardized reporting of activities, limited photochemical stability, insufficient knowledge of #' "
reaction mechanisms, balancing charge carrier lifetimes with catalysis timescales, and the use of #("
unsustainable sacrificial reagents. #)"
#*"
1 Introduction $+"
With the global reliance on non-renewable fossil fuels and increasing concern over their impact on $! "
climate, there has never been greater urgency to secure alternative clean and renewable energy $#"
supplies. Solar hydrogen has attracted much interest because of the abundance of solar energy and the $$"
cleanliness and high gravimetric energy density of hydrogen fuel. To scale-up photocatalytic water $%"
splitting to produce renewable hydrogen, we require a low-cost, earth-abundant photocatalyst with a $&"
~10% solar-to-hydrogen (STH) energy conversion efficiency
1
. Considering that nearly half of the $'"
energy in the sunlight that reaches the earth surface comes from visible light photons (400–700 nm), $("
their efficient use is one of the biggest challenges
2
. The long-standing target here is to find efficient, $)"
reasonably-priced semiconductor photocatalysts that can thermodynamically drive both proton $*"
reduction to hydrogen and water oxidation to oxygen, while staying chemically and photolytically %+ "
stable over long periods
3
. Four decades of extensive exploration into inorganic semiconductor %!"
photocatalysts—mostly metal oxides—has demonstrated that tuning their properties is challenging
4
. %#"
By contrast, conjugated polymeric semiconductors have a potential advantage of easy-to-tune %$ "
properties through synthetic control. This tuneability and their proven performance in fields including %% "
solar cells
5
, photoelectrochemical devices
6
and light-emitting diodes
7
, make organic materials %& "
attractive alternatives to inorganic semiconductor photocatalysts. %'"
The first reports on conjugated polymer photocatalysts date back to the late 1980s when Japanese %("
researchers demonstrated that poly(p-phenylene)
8,9
could reduce protons to hydrogen under %)"
illumination in the presence of sacrificial electron donors. In 2009, polymeric carbon nitride (CN
x
H
y
) %*"
was shown to evolve hydrogen from a 10 vol% triethanolamine (TEOA) aqueous solution and oxygen &+ "
from an 0.01M silver nitrate aqueous solution under visible-light illumination
10
. This report triggered &!"
a massive interest in CN
x
H
y
for hydrogen production and the development of new polymeric &#"
#"
"
photocatalysts, including conjugated poly(azomethine) networks, pyrene-based conjugated !"
microporous polymers (CMPs), covalent triazine-based frameworks (CTFs), covalent organic # "
frameworks (COFs), and planarized-fluorene-based conjugated polymers, to name a few
11-15
. Recently, $"
overall water splitting (OWS) using polymeric photocatalysts has also been claimed
16-19
, albeit at %"
impractically low efficiencies. &"
Despite the recent interest in polymeric photocatalysts, and particularly reduction photocatalysts for '"
hydrogen generation, we have yet to reach an understanding of their photophysics comparable to that ( "
developed for organic photovoltaics (OPVs) and metal oxide photocatalysts. Even factors well-known )"
to influence the performance of OPVs and metal oxides, such as defects and charge trapping, have *"
only just started to receive attention. The intimate interaction of the sacrificial electron donor and/or !+"
co-catalysts on the surface of polymer photocatalysts most likely affects their charge transfer kinetics !!"
and possibly exciton dissociation, yet little is known to date. The structural tunability of polymers and !#"
the multitude of electron donor-polymer combinations suggests that significant efforts are needed to !$"
understand their association and its impact fully. Difficulties in producing samples with identical !%"
physical properties such as molecular weight, degree of branching, and terminations, will likely !&"
complicate the comparison of results between different laboratories. Furthermore, it is still debatable !'"
whether such materials can directly drive the multi-redox chemistry of water oxidation and reduction !("
or whether co-catalysts are always required
14,20
. !)"
Here we discuss state-of-the-art polymeric photocatalysts for the hydrogen and oxygen evolution half !*"
reactions. We also explore design rules for these systems and how they can be characterised through #+"
spectroscopy. Furthermore, we discuss challenges facing this field, such as uncertainties surrounding #!"
the functional characterisation of photocatalysts, where the mechanism for hydrogen production is ## "
usually unknown. #$"
#%"
2. Framework for understanding polymer photocatalytic activity #&"
A photocatalytic reaction is initiated by light absorption to generate excited electron-hole pairs #' "
(excitons), followed by their separation into free charges, which ultimately drive redox reactions. The #( "
amount of light that a photocatalyst absorbs, and thus the number of charges it can generate, is #)"
determined by the overlap of its absorption spectrum with the irradiance spectrum of the light source. #*"
In contrast to polymers in solution, absorption spectra of solid polymers generally have a block-like $+"
shape
21
,
and their light absorption is therefore often simplified to a consideration of the optical gap. $!"
The optical gap of a semiconductor is the minimum energy (longest wavelength) that a photon $#"
requires to generate excitons and can vary sharply from polymers to polymers. For example, poly(p-$$"
phenylene) has an optical gap in the violet (E
gap
~ 3 eV, λ
edge
~ 420 nm)
8
, just barely in the visible $%"
range of the spectrum, while poly(thiophene) already starts absorbing in the near-infrared (~ 1.5 eV, $& "
830 nm)
22
. Shifting the optical gap to the red generally results in more absorbed photons, thus $'"
generating more excitons and more free charge carriers if these excitons dissociate. $("
Light absorption by organic materials typically produces Frenkel excitons with binding energies that $)"
are more than an order of magnitude larger than kT at room temperature
23
. This strong interaction $*"
tends to prevent spontaneous exciton dissociation and opens loss pathways such as re-emission of %+ "
light (photoluminescence) or internal conversion into heat/phonons. Spontaneously separated charges %!"
may also relax to the ground state through electron-hole recombination or reassociate into excitons. %# "
The large binding energy means that excitons usually must diffuse to an interface—such as the %$ "
polymer-solution interface—to dissociate, where one of the formed charge carriers takes part in a %%"
solution reaction, and the other remains on the polymer, poised to undergo a subsequent reaction. %&"
These charge carriers must possess a sufficiently high driving force for a targeted reaction such as %'"
proton reduction or the oxidation of water/sacrificial electron donors. %("
Free electrons in a polymer can thermodynamically drive the reduction of protons to molecular %) "
hydrogen if the electron affinity (EA) of the polymer, when expressed as a redox potential, is more %*"
negative than the potential of the proton reduction reaction (H
+
(aq) + e
-
→ ½ H
2
(g), E = -0.41 V vs. &+"
SHE at pH 7). Similarly, free holes can drive the oxidation of water if the ionisation potential (IP) of &!"
$"
"
the polymer is more positive than the potential of the overall oxidation of water (O
2
(g) + 4H
+
(aq) + !"
4e
-
→ 2H
2
O (l), E = + 0.82 V vs. SHE at pH 7). Hence, the EA and the IP of the polymer should #"
straddle both reactions to work as an OWS photocatalyst (Fig. 1). In the case of excitons, the $"
corresponding IP and EA potentials are labelled as IP
*
for reduction and EA
*
for oxidation (the %"
asterisk denotes that an exciton provides the electron or hole). The oxidation of sacrificial donors like &"
triethylamine (TEA) is thermodynamically less demanding than for water (diethylamine (aq) + '"
acetaldehyde (aq) + 2H
+
(aq) + 2e
-
→ triethylamine (aq) + H
2
O (l), E = -0.72 V vs SHE at pH 11.5
21
, ("
the likely pH of a triethylamine solution). It is also kinetically faster because two holes rather than )"
four are required. As a result, the activity of polymers for hydrogen evolution is often tested in the *"
first instance using such sacrificial donors, rather than attempting OWS. !+"
!!"
It is known from the literature on OPVs
24
that the exciton diffusion length (the distance an exciton !#"
travels before decaying back to the electronic ground state) is typically much shorter than the optical !$"
absorption depth (the distance light penetrates a material). Rapid charge separation across the !%"
interface is promoted by a large interfacial area between donor and acceptor domains, but the decrease !& "
of pure domains that serve as long-range selective charge transport channels can also lead to more !'"
rapid geminate recombination of electron-hole formed from the same absorbed photon
25
. For !("
polymeric photocatalysts, reduction of the typical particle/domain size, or increased solution !)"
permeability, should help in minimising the loss of excitons before they can dissociate at the !* "
polymer/solution interface. The interaction between the polymer and water (i.e., the wettability) can #+"
be expected to influence the activity of these materials, particularly for OWS. For linear polymers, #!"
contact angle values (lower angles correspond to better wetting) measured for pure water are reported ##"
to range from ~90° for purely hydrocarbon polymers
26
, such as poly(p-phenylene), to ~60° or lower #$"
for polymers containing heteroatoms
27
, such as poly(2,5-pyridine) and undoped CN
x
H
y
, or even lower #%"
for suitably doped CN
x
H
y
.
28
#&"
#'"
2.1 Characterisation of the activity of polymers #("
The activity of a photocatalyst for a targeted reaction can be quantified by measuring the formation #)"
rate of the product; for example, the hydrogen evolution rate. Importantly, the rates strongly depend #* "
on experimental conditions such as the spectrum and intensity of the light source. To further $+"
complicate things, rates are reported with or without considering the mass of photocatalyst $!"
(µmol h
−1
g
−1
vs µmol h
−1
). We provide some clarification of the experimental conditions in our $#"
overview tables below. However, we note that spectrum and output intensity even vary for different $$"
models of nominally the same light source (e.g., different 300 W Xe light sources). In addition, the $% "
lack of standards for the different lamp-to-sample distances and use of focusing/collimating optics $&"
make direct comparisons between reports from different groups difficult. Complimentary apparent $'"
quantum yield (AQY) measurements directly relate the amount of formed product to the amount of $("
incoming monochromatic photons, which improves comparability of activities; this makes AQY a $) "
preferable and more reliable metric than hydrogen evolution rate. AQY for hydrogen evolution is $*"
calculated as the ratio of the number of reacted electrons (2 × number of hydrogen molecules %+"
produced) to the number of incident photons of defined energy: %!"
𝐴𝑄𝑌 =
2× 𝑛
!
!
𝑛
!!!"!#
Even with well-defined illumination conditions, AQY (and hydrogen evolution rate) measurements %#"
are still affected by other reaction parameters such as photocatalyst concentration, sacrificial donor %$"
used, mixing, the bandwidth of a band filter and reactor pressure (p
initial
). For instance, a reduced %% "
headspace pressure suppresses back reactions compared to ambient pressure, often improving the %&"
hydrogen evolution yield significantly. Other factors, such as the addition of phosphate salt, can %'"
enhance the hydrogen evolution yield by accelerating the proton reduction and TEOA oxidation
29,30
. %("
Therefore, all these factors should be carefully considered when comparing the performance of %) "
different photocatalysts. %*"
%"
"
!"
3 Performance of polymeric photocatalysts #"
3.1 Carbon nitrides
and their photocatalytic activity $"
Carbon nitrides are a family of triazine or heptazine-based polymers containing carbon and nitrogen, which %"
is often referred to in the literature as graphitic carbon nitride (g-C
3
N
4
), which would be a heptazine-based &"
layered crystalline structure (Fig. 2, 1). However, both characterisation reported and calculations of the '"
CN
x
H
y
phase diagram
31
show that ideal g-C
3
N
4
is unlikely to form under the synthetic conditions ("
employed so far for photocatalytically active samples. In practice, materials of relevance to photocatalysis, )"
even if referred to as g-C
3
N
4
, contain significant amounts of hydrogen
32
and appear to consist of melon ( 2 *"
in Fig. 2)
33,34
. Melon is a linear polymer formed of heptazine units linked through amine (–NH–) bridges !+"
with a solid-state structure that is stabilised by intermolecular hydrogen bonds involving the amine groups. !!"
Photocatalytically active carbon nitride materials are often poorly crystalline or X-ray amorphous and are !#"
more accurately represented by CN
x
H
y
based on both experimental characterisation and modelling. !$"
Alternatively, materials prepared using molten salt as the reaction medium can yield a crystalline, layered !%"
poly(triazine imide) structure
35
in which some of the salt ions are incorporated ( 3 in Fig. 2). !&"
Table 1 summarises representative examples of CN
x
H
y
photocatalysts for hydrogen or oxygen evolution. !' "
Systems with AQYs larger than a few percentage are mostly reported in the presence of a sacrificial !("
scavenger and a co-catalyst (e.g. Pt or Ru, usually by photodeposition). Among various pristine CN
x
H
y
!)"
materials synthesised from common precursors, urea-derived materials exhibit slightly wider optical gaps !*"
(2.9–3.0 eV) than their dicyandiamide/melamine-derived counterparts (2.7~2.8 eV), and represent the #+"
benchmark for efficiency, perhaps due to a higher degree of polymerisation
36,37
. Recently, several other #!"
organic precursors (e.g., semicarbazide hydrochloride, 5-aminotetrazole) were reported to produce CN
x
H
y
##"
with improved performance
28,30
. A range of synthetic modifications to promote, for example, charge #$ "
separation, have been proposed
38
and used to achieve higher AQYs (generally for light in the range of #%"
395~420 nm). Such engineering strategies include increasing the degree of polymerisation
36
, nanosheet #&"
fabrication
39
, use of templates
40
, fabrication in molten salts
41
, creating p-n homojunction
42
, and selective #' "
doping
43,44
. Another emerging approach to control the properties of CN
x
H
y
is the utilisation of self-#("
assembled supramolecular structures as reactants
45-48
, such as using of halogen-based assemblies
49-52
and #)"
supramolecular single crystals
53-55
. Interestingly, a few CN
x
H
y
materials with controlled terminal groups, #*"
including cyanamide, urea and hydroxyl species
28,56,57
, have boosted HERs, suggesting that terminal $+"
groups in CN
x
H
y
structures can play an important role. So far, materials prepared via the molten salt $!"
approach have exhibited benchmark AQYs of 57~65% at 420 nm measured under reduced pressure in a $# "
phosphate environment, although their stability is seldom reported
30,37,58,59
. $$"
Meanwhile, to better match light absorption and solar spectrum, strategies have been devised to narrow the $%"
optical gap of CN
x
H
y
60-68
. Recently, location-controlled doping of CN
x
H
y
(e.g., selective linker/terminal $&"
replacement, surface layer doping) was reported to not only enhance charge separation but also stepwise $'"
narrow the optical gap to below 2 eV, leading to enhanced HERs and a benchmark AQY of 2.1% at 500 $("
nm measured at ambient conditions
28,69
. Despite these impressive advances in hydrogen evolution rate $) "
performance, fundamental understanding of CN
x
H
y
photocatalytic activity is still relatively limited, as we $*"
discuss further below. %+ "
Compared to proton reduction, there have been far fewer reports on water oxidation
69-72
and OWS
3,16,18,19
%!"
using CN
x
H
y
, most likely due to the inherent kinetic and energetic challenges of water oxidation. CN
x
H
y
%#"
was the first polymer to perform OWS in Z-scheme systems
3,73
or with suitable co-catalysts (e.g., Pt/PtO
x
, %$ "
Pt/CoP)
18,19
. In a Z-scheme system, CN
x
H
y
and WO
3
(or BiVO
4
) worked as reduction and oxidation %%"
photocatalysts, with I
-
/IO
3
-
(or Fe
2+
/Fe
3+
)
as redox mediators, respectively. Later on, reduced graphene %&"
oxide was reported as a shuttle in CN
x
H
y
/WO
3
heterojunction for OWS. Remarkably, carbon-quantum-%'"
dots (QD) cocatalysts are suggested to facilitate charge separation and decompose kinetically favourable %("
H
2
O
2
74
to O
2
via a two-electron process, bypassing the slow four-hole kinetics of direct oxidation of H
2
O %)"
to O
2
. The reported STH efficiency of 2% on QD/CN
x
H
y
composite is notable
16
, although this has proved %*"
challenging to reproduce
16,75
. &+"
&!"
&#"
&"
"
!"
#"
3.2 Synthesis and photocatalytic activity of organic polymers $"
Despite the substantial progress made in tailoring the (photophysical) properties of CN
x
H
y
by tuning %"
the synthetic routes and/or post-synthetic modifications, the degree of control and synthetic diversity &"
are inherently limited by the reaction conditions required to prepare the materials. This is less of an '"
issue when preparing conjugated polymers via metal-catalysed coupling reactions. Reaction ("
conditions can be relatively mild, and many functional groups in the reactants can be tolerated. This )"
allows the study of families of materials using related building blocks and, hence, the study of *"
structure-performance relationships. While there are still limits to synthetic control in such !+"
polymers—for example, regarding molecular weight, architecture, and monomer sequence !!"
distribution in copolymers—there is, in general, a broader scope for molecular engineering of a !#"
specific function in polymers than in materials synthesised at a high temperature, such as CN
x
H
y
. !$"
Tables 2 and 3 give an overview of reported polymer-based materials (linear and CTF in Table 2; !%"
CMP and COF in Table 3) for hydrogen and oxygen evolution. The values provided in this table are !&"
intended to provide a summary of the variety of reported materials and their properties, rather than a !'"
numerical comparison of their “success” as photocatalysts. Note that there is no agreement upon !("
systematic nomenclature in this field nor a conclusive analytical portfolio for a full characterisation of !)"
these (mostly) insoluble materials. Materials are sorted by decreasing optical gap values. !*"
#+"
#!"
3.3 Linear Polymers ##"
Poly(p-phenylene) was shown to evolve hydrogen under illumination in the presence of sacrificial #$"
donors in 1985, the first example of a polymeric photocatalyst
8
. While the reported AQY of 0.03% #% "
was low, the effect of different sacrificial donors and additional doping with precious metals was #&"
already investigated at that time
8,9,76
. Shortly afterwards, a bipyridine-based linear polymer was shown #'"
to reduce protons to hydrogen under illumination using triethylamine as a sacrificial donor; the #("
hydrogen evolution rate increased by two orders of magnitude in the presence of RuCl
3
77
. More #)"
recently, the use of linear homo- and co-polymers (Fig. 3) prepared through coupling reactions have #*"
received renewed attention as photocatalysts. For example, a series of photocatalytically active $+"
phenyl-co-polymers with fluorene derivatives were reported, the most active of which, a co-polymer $!"
of phenylene and dibenzo[b,d]thiophene sulfone (P7), was significantly more active than $#"
poly(p-phenylene)
15
. Another photocatalytically active linear co-polymer (B-BT-1,4)
78
featured $$"
alternating electron-donor-acceptor units in the form of phenyl and 2,1,3-benzothiadiazole units. $%"
Cobalt-chelating PPDI-bpy (perylenediimide–bipyridine) and PPDT-bpy (benzo [1,2-b:4,5-$& "
b]dithiophene- bipyridine) co-polymers were developed, combining a light-harvesting polymeric $'"
backbone with molecular catalytic active sites (bpy-metal complex)
79
. $("
$)"
In contrast to the discovery of materials with high photocatalytic activity, the processability of $* "
polymeric photocatalysts remains poorly explored for the fabrication of multicomponent and scaled-%+ "
up devices. Some of us recently reported a solution-processable co-polymer made of carbazole units %!"
