An overview on emerging photoelectrochemical self-powered ultraviolet photodetectors

Abstract

In recent years, as a new member of ultraviolet photodetectors (UV-PDs), photoelectrochemical UV-PDs (PEC UV-PDs) have received great attention. Compared to conventional photoconductors, PEC UV-PDs exhibit a number of merits, including low cost, environmentally friendly nature, being self-powered, and fast response. This tutorial review provides a comprehensive introduction to this research field, covering from the basics of performance evaluation of PEC UV-PDs, the state-of-the-art advances in structural design, electrolyte matching, and electrode fabrication of PEC UV-PDs, to the integration of multiple functions into a PEC UV-PD. In the end, we present our perspectives on the future development of PEC UV-PDs and highlight the key technical challenges in aiming to stimulate further developments in this research field.

---

Jinyuan Zhou

Jinyuan Zhou received his B.S. and Ph.D. degrees in microelectronics and condensed matter physics from Lanzhou University of China in 2002 and 2010, respectively, and then joined Lanzhou University as a lecturer. From 2010 to 2012, he worked at Nanyang Technological University of Singapore as a postdoctoral fellow. In 2013, he was promoted to associate professor. His research interests include semiconductor and carbon nanotube based functional nanomaterials & their applications in energy and optoelectronic devices.

Lulu Chen

Lulu Chen received her B.S. degree in physics from Shangqiu Normal University, Henan, China, in 2011. At present, she is a Ph.D. candidate in condensed matter physics under the supervision of Prof. Erqing Xie at Lanzhou University. Her research focused on the synthesis of ZnO and SnO2 nanomaterials and their applications in electrochemical devices, such as dye-sensitized solar cells and self-powered UV-photodetectors.

Youqing Wang

Youqing Wang received his B.S. degree in physics from Lanzhou University of China in 2011, and then joined Prof. Erqing Xie's group. At present, he is a Ph.D. candidate in condensed matter physics at Lanzhou University. His research is focused on dye-sensitized solar cells and ultraviolet detectors.

Yongmin He

Yongmin He received his Ph.D. degree in condensed matter physics from Lanzhou University of China in 2015. From 2013 to 2015, He was supported by China Scholarship Council to study at Rice University as a visiting student. From 2015 to the present, he worked at Nanyang Technological University of Singapore as a postdoctoral fellow. His current research interests mainly focus on “smart” flexible ECs, and two-dimensional (2D) materials such as graphene, h-BN, transition metal dichalcogenides and their heterostructures.

Xiaojun Pan

Xiaojun Pan received his B.S. and Ph.D. degrees in Physics and condensed matter physics from Lanzhou University of China in 2003 and 2008, respectively, and then joined Lanzhou University as a lecturer. From 2008 to 2010, he worked at City University of Hong Kong as a senior research assistant. In 2012, he was promoted to associate professor of Lanzhou University. His main research interests focus on the fields of optical materials, nanomaterials and nano-optoelectronic devices.

Erqing Xie

Erqing Xie received B.S. and Ph.D. degrees in physics and condensed matter physics from Lanzhou University of China in 1984 and 1995, respectively, and then joined Lanzhou University as a lecturer. In 2001, he was promoted to full professor of Lanzhou University. His research interests are in the fields of superhard films (such as DLC, SiC(N), BN, and C3N4), thin-film technology for solar applications, and semiconductor functional nanomaterials; his current research mainly involves nanomaterials for environmental and green energy applications, such as dye-sensitized solar cells, supercapacitors, semiconductor oxide-based gas sensors, UV photodetectors, and photocatalytic technology.

---

1 Introduction

As is well known, ultraviolet (UV) light is a “double-edged sword”, and can bring not only benefits to our daily lives, but also harms. It is very important for us to precisely measure and control UV irradiation. Due to their simple device structure and high ON/OFF ratios, UV photodetectors (UV-PDs) based on photoconductance were considered to be very promising for practical applications.^1–4 Their working principle is based on the conductance change of the semiconductors under illumination, namely photoconductivity (Fig. 1a).^5,6 It can be understood that, in a semiconductor, electron–hole pairs are generated when illuminated by photons that possess energy larger than their band gap. The generated holes are usually trapped at the surface, and the electrons transit to the conduction band (CB) and become quasi-free, resulting in an increase in the conductivity. So far, a number of wide-band-gap semiconductors (such as ZnO,⁷ TiO₂,⁸ SnO₂,⁹ ZnS,¹⁰etc.) have been used for making photoconductive UV-PDs, and various nanostructures (nanoparticles (NPs), nanowires (NWs), nanotubes (NTs), and nanotube/nanowire arrays (NTAs/NWAs)) and/or nano-composites have also been adopted to improve their performances.¹¹

Fig. 1 (a) Schematic of a conventional photoconductor based on TiO₂ nanotube arrays (TNTAs). (b and c) Trapping and photoconduction mechanism in TiO₂ NWs. (d) Schematic of a ZnO–GaN p–n junction. (e) CdSe/graphene Schottky junction UV-PDs. (f) Schematic of self-powered UV-PDs based on PEC cells, and (g) working mechanism of the PEC UV-PD. (a) and (b and c) Reproduced from ref. 6 and 7, respectively. Copyright © 2010 and 2007 American Chemical Society. (d) Reproduced from ref. 20. Copyright © 2011 Wiley-VCH. (e) Reproduced from ref. 21. Copyright © The Royal Society of Chemistry 2012. (f) Reproduced from ref. 22. Copyright © 2012 Elsevier Ltd.

Generally, the performances of a photoconductive PD involve two aspects, namely the absorption of the incident light and the separation/transport of photogenerated electrons.^12–14 The magnitude of the conductivity change depends on the quantum yield of carrier generation and the mobility of photogenerated carriers. As schematically shown in Fig. 1b and c, the photo detecting process is normally determined by the absorption and desorption of oxygen molecules in air, both of which are slow processes that lead to a low photoresponse.^15–19 Moreover, this type of UV-PD usually needs an external power source, which will limit its potential applications in a harsh environment.

To solve the aforementioned problems, Schottky and p–n junctions are introduced into self-powered UV-PDs, which in nature are optoelectronic conversion devices similar to solar cells. The self-powered UV-PDs based on p–n junctions are constructed using a pair of p- and n-type semiconductors with a wide band gap, forming a p–n junction photodiode (Fig. 1d).^23,24 This type of photodiode can convert the incident light into an electrical signal without the necessity of an external power source. Furthermore, the photodiode usually possesses a good spectral selection (visible-blind) and a fast photoresponse (rise time (t_r) of several tens of μs, decay time (t_d) of several hundreds of μs). As illuminated in Fig. 1e, a UV-PD based on the Schottky junction is constructed using a metal/semiconductor interface, in which the semiconductor possesses a wide band gap. When illuminated by UV light (photon energy is larger than the E_g of the used semiconductor), the electron–hole pairs are generated and quickly separated due to the built-in electric field induced by Schottky junctions, showing a photovoltaic (PV) effect. Schottky-junction PDs often show extremely high photosensitivity (ratio of (I_photo − I_dark)/I_dark) and fast photoresponse (less than 100 μs).^25,26 Although many achievements have been obtained, especially ultra-fast photoresponse and extremely high photosensitivity, the reported self-powered PDs based on p–n or Schottky junctions usually show low spectral responsivity and high cost (a complex fabrication process), limiting their real applications.²⁷

Recently, photoelectrochemical (PEC) self-powered UV-PDs have attracted great attention.²² As shown in Fig. 1f, this type of UV-PD possesses a similar device configuration to other PEC devices, which, in principle, is a PEC PV device. Fig. 1g shows the working mechanism of the PEC UV-PDs using a wide-band-gap semiconductor as the photoanode. When the nanocrystalline TiO₂ (nc-TiO₂) film is illuminated by UV light, the absorbed photons will promote electrons from the valence band (VB) to CB, e_CB, leaving behind a hole, T⁺:

T + hν → T⁺ + e_CB.

The photogenerated holes then migrate to the semiconductor/electrolyte interface and trap an electron donor in the electrolyte (D), leaving the oxidized redox species, D⁺,

T⁺ + D → T + D⁺,

while the e_CB will diffuse through the nc-TiO₂ film and reach the transparent conductive oxide (TCO) electrode, becoming external circuit electrons (e). On the other hand, the formed D⁺ will diffuse through the electrolyte and reach the counter electrode. Then, the D⁺ will be reduced by the e at the counter electrode, completing the circuit.

D⁺ + e → D.

Therefore, in this type of PEC UV-PD, light harvesting and photogenerated carrier transport are achieved simultaneously.

Compared to the reported self-powered photoconductors based on p–n junctions and/or Schottky junctions, PEC UV-PDs usually exhibit two advantages:^28–34 (1) much higher photo responsivity. Most of the reported p–n junctions and/or Schottky junctions are made up of semiconductor nanoscale units, which usually lead to current output signals of nA order and require high-precision galvanometers; while PEC UV-PDs can output a large current of μA order. (2) Simple fabrication processes and low cost: the fabrication of p–n junctions and/or Schottky junctions usually requires not only complex processes, but also high-purity semiconductor raw materials, both of which determine their high cost; while PEC UV-PDs are usually fabricated via a physicochemical route, which shows a low requirement in fabrication and cost. Nevertheless, this proposal of PEC UV-PDs might open a new insight into the development of cheap, easily-fabricated, self-powered UV-PDs with high spectral responsivity, high photo sensitivity and fast photoresponse.

The aim of this review is to sum up recent advances in PEC self-powered UV-PDs, including (1) the basics of performance evaluation of PEC UV-PDs for an easy understanding of different research approaches and diverse performance metrics reported in the literature; (2) the latest developments in electrolyte and photoanode designs for PEC UV-PDs, including various electrolytes, various wide-band-gap semiconductors, and various micro-/nano-scale morphologies/modifications/hybrids; and we further analyze their effects on the properties of photoanodes (such as power conversion efficiency (η), photocurrent density (J), spectral responsivity, and photosensitivity). In this context, we provide examples from recent experimental results to elucidate these concepts in this section. And we will finally give an outlook on future directions in the research on PEC UV-PDs aiming to excite more research interest and stimulate further research progress in this research field.

2 Performance evaluations of PEC UV-PDs

The basic performance metrics of a PEC self-powered UV-PD usually include spectral responsivity, photosensitivity, response time, incident photo-to-current conversion efficiency (IPCE), etc.^28,31–36 These parameters indicate how a detector responds and are described as follows:

Spectral responsivity (R_λ, A W⁻¹) is the ratio of photocurrent (I_p) to incident light power. It mainly depends on the wavelength of incident UV light. The plot of responsivity as a function of wavelength gives the spectral response of a PD. The R_λ can be evaluated using

where P_inc is the incident optical power in watts, h is the Planck constant, ν is the light frequency, q is the electron charge, and η is the ratio of photogenerated electrons to incident photon numbers.

Photosensitivity (S), named as the ON/OFF ratio, is the ratio between I_p and dark current (I_d), i.e.,

This ratio depends on both I_p and I_d, reflecting not only the optoelectronic conversion performances of the devices, but also the transport characteristics of electrons in the devices’ photoanodes.

Response time (t) is used to characterize the speed of response to a sudden change in the input signals, i.e., to measure the time required for a PD's response to a light pulse. The response time is specified as the time taken by the photocurrent to rise or decay to a specific value under the “ON” or “OFF” state of light illumination. The t_r is often defined as the time taken by the photocurrent to increase from 0 to 63% (i.e., 1 − 1/e) of the maximum value, and the t_d is the time taken by the photocurrent to decrease from the maximum value to 37% (i.e., 1/e).

In most cases of PEC UV-PDs, the response speed is limited by the t_d, rather than the t_r, which is quite different from other types of UV-PDs.³⁷ In fact, the t_r of a PEC UV-PD is mainly dependent on two fundamental aspects: (1) the time taken by the e_CB generated in the solid/liquid interface to diffuse to the TCO electrode via the wide-band-gap semiconductor films; (2) the time taken by the D⁺ to drift to the counter electrode via the electrolyte. The drift process of D⁺ through the electrolyte is usually quite rapid due to free ion transport in the electrolyte, while the e_CB diffusion process is greatly influenced by the recombination time and is relatively slow. The diffusion time can be improved by ensuring that all electron–hole pairs are separated and transported through the electrode films under the built-in electric field of the semiconductor junctions.

IPCE is defined as the ratio of output I_p (electrons per second) to incident photon fluence (photons per second) on the device. It mainly depends on the light harvesting efficiency (LHE (λ)), electron injection efficiency (ϕ_inj), and electron collection efficiency (η_c). Then IPCE can be roughly expressed by³⁸

IPCE(λ) = LHE(λ) × ϕ_inj × η_c.

Thus, what we can do to increase the IPCE value is to improve LHE(λ), ϕ_inj and η_c, which generally guide the potential research directions in the PEC field. On the other hand, using R_λ, IPCE can also be estimated using

where E_ph is the photon energy measured in eV. As for certain incident UV light, the IPCE value is in direct proportion to the output I_p. This is to say that IPCE and I_p are equal for quantitative UV detecting.

Some important issues should be noted here for graduate students or researchers who are new in the research on PEC UV-PDs. First, the methodologies to reliably choose and prepare a material for use as an PEC UV-PD electrode are not well standardized due to the reported different techniques and parameters (e.g. different device structures (three-electrode or two-electrode), different calculations of response time (0 to 1 − 1/e, 0 to 90%, or 0 to a stable value), different electrode kinds/materials/morphologies, different electrolytes/ions, and different UV light wavelengths), thus yielding widely varying results. Second, the requirements in electrode materials are not the same as those in dye sensitized solar cells (DSSCs) or photocatalytic cells; e.g., the IPCE value is not the most important parameter for PEC UV-PDs. Third, for a high-performance UV-PD, it should be “visible-blind” for high UV selection, and “linear output” for quantitative UV measurement. This is to say, these two aspects are also important performance metrics for UV-PDs.

Overall, the various performance metrics discussed above indicate that the UV detection performances of the PEC devices are mainly determined by the photoanodes. Current research efforts are focused on improving the optoelectronic performance of the frequently-used photoanode materials (such as ZnO, TiO₂ and SnO₂), including spectral responsivity, photosensitivity, response time, visible-blind and linear-output characteristics. In addition, the used electrolytes and multi-functional PEC UV-PDs are also briefly reviewed, as well as the preparation of the micro-/nano-structure of photoanodes. The recent results are summarized in Table 1.

Table 1 Photoresponse performances of PEC UV-PDs based on various nanostructures of ZnO, TiO₂, and SnO₂

Nanostructures	Conditions	Electrolyte type	Electrode system	J p (μA cm^−2)	S ((IP − Id)/Id)	t r (s)	t d (s)	Estimate method of t	R λ (mA W^−1)	Visible blind	Linear output	Ref.
Hydrothermal ZnO NRs	250–600 nm, 100 mW cm^−2, 0 V	0.1 M Na2SO4	3-Electrode	1.5	7	4.9	134.3	Biexponential function fitting	—	No	—	39
ZnO(N) NR films	390 nm, 3 mW cm^−2	Spiro-MeOTAD	2-Electrode	50	300	4 × 10^−3	1 × 10^−2	10% to 90%	17 at 390 nm	Yes	—	40
TiO2 NRAs	365 nm, 0.7 mW cm^−2	Poly(9,9-dihexylfluorene)	2-Electrode	1.78 × 10^−2	1000	<0.2	<0.2	Estimated by the test limit	33 at 395 nm	Yes	—	41
ZnO nanoneedles	365 nm, 1.25 mW cm^−2	DI water	2-Electrode	4.1	—	0.1	0.1	10% to 90%	22 at 385 nm	Yes	No	30
ZnO@TiO2 nanostrawberries	365 nm, 20 mW cm^−2	I^−/I3^−	2-Electrode	379	37 899	0.022	0.009	0–1, 1 − e^−1	17.85 at 365 nm	Yes	Yes	29
ZnS/ZnO NRAs	Xe lamp, 100 mW cm^−2, 0 V	0.5 M Na2SO4	3-Electrode	50	—	—	—	—	—	Yes	—	42
TiO2 films by ALD	365 nm, 2.44 mW cm^−2	Water	2-Electrode	100	10^6	<0.5	<0.5	0 to 1	69 at 350 nm	Yes	Yes	43
nc-TiO2 films	365 nm, 33 mW cm^−2	I^−/I3^−	2-Electrode	550	2698	0.08	0.03	0–1, 1 − e^−1	16.7 at 365 nm	Yes	Yes	22
TiO2 NRAs	365 nm, 3 mW cm^−2, 0 V	0.5 M Na2SO4	3-Electrode	15	6	<0.1	<0.1	—	5 at 365 nm	—	No	44
TiO2 NRAs	365 nm, 10 mW cm^−2	DI water	2-Electrode	40	—	0.15	0.05	0–90%, 1 − e^−1	25 at 350 nm	Yes	—	36
TiO2 NRAs	365 nm, 2.5 mW cm^−2	Liquid crystal- I^−/I3^− gel	2-Electrode	175	—	<0.03	<0.03	0–90%, 1 − e^−1	90 at 383 nm	Yes	Yes	33
TiO2 nano-branched arrays	365 nm, 2 mW cm^−2	I^−/I3^−	2-Electrode	373	—	0.15	0.05	0–90%, 1 − e^−1	220 at 352 nm	Yes	—	34
Dendriform TiO2 NWs	365 nm, 25 mW cm^−2	I^−/I3^−	2-Electrode	570	1902	0.0053	0.0059	0–1 − e^−1, 1 − e^−1	22.8 at 365 nm	Yes	Yes	45
TNRCs	254/365 nm, 1.55 mW cm^−2, 0 V	I^−/I3^−	2-Electrode	17	—	1.7	2.1	0 to 1	19 at 345 nm	Yes	Yes	35
TNRs				235	—	0.3	0.1		89 at 355 nm	Yes	Yes
Multilayer TNRC/TNRs				278	—	0.3	0.2		152 at 360 nm	Yes	Yes
TNTAs	250–385 nm, 0 V	0.1 M NaOH	3-Electrode	210	—	<0.2	<0.2	0–1	11.2 at 350 nm	Yes	—	46
SrTiO3–TiO2 heterojunction				650	—	<0.2	<0.2		18.6 at 350 nm	Yes	—
B-SnO2 NFs	365 nm, 40 mW cm^−2	I^−/I3^−	2-Electrode	900	4549	0.03	0.01	0–1, 1 − e^−1	22.5 at 365 nm	Yes	Yes	28
SnO2@TiO2 microtubes	365 nm, 1.55 mW cm^−2	I^−/I3^−	2-Electrode	100	—	<0.1	<0.2	0 to 1	∼52 at 340 nm	Yes	—	32
SnO2 microtubes				78	—	<0.1	<0.2		∼38 at 340 nm	Yes	—
SnO2@TiO2 hollow nanospheres	365 nm, 35 mW cm^−2	I^−/I3^−	2-Electrode	1277	4020	0.03	0.01	0–90%, 1 − e^−1	36.48 at 365 nm	Yes	Yes	31
B-TiO2 nanoneedles–SnO2 nanosheets	365 nm, 40 mW cm^−2	I^−/I3^−	2-Electrode	709.5	4404	0.02	0.004	0–1, 1 − e^−1	17.7 at 365 nm	Yes	Yes	47

---

3 Recent developments in PEC UV-PDs

3.1 Prototype for PEC self-powered UV-PDs

PEC UV-PDs possess a similar device configuration to other PEC devices, evolved from a photoanode/electrolyte/counter electrode sandwich structure. Thus their working mechanisms are to some extent the same as that of other PEC devices. Generally, the improvement in detecting the performances of PEC UV-PDs hinges on the development of the combination of electrolytes (redox couples), counter electrodes, and photoanodes.⁴⁸ In 2011, Lee et al. first reported a type of UV-PD based on a TiO₂/water solid–liquid heterojunction,⁴³ containing two electrodes: a 50 nm TiO₂/Fluorine-doped Tin Oxide (FTO) electrode and 50 nm Pt/indium tin oxide (ITO). Both electrodes were assembled together using a sealant and some water used as an electrolyte was poured into the space between them. Under UV illumination (365 nm, 0.244 μW cm⁻² to 0.113 mW cm⁻²), the I_p of the devices linearly increases with increasing incident light power, exhibiting an excellent reproducible photoresponse with a fast response speed (both decay and rise times are less than 0.5 s). Here, it is noted that this linear variation of I_p will make this type of UV-PD easily usable for the quantitative measurement of UV light. Moreover, the S can reach an order of magnitude of 6 under 479 μW UV illumination, and the devices show a visible-blind characteristic, exhibiting their maximum responsivity of ∼69.2 mA W⁻¹ at ∼350 nm. Thus, it can be seen that this type of UV-PD possesses an exceptional competence for UV light detection and presents a promising direction for future development of commercially low-cost UV-PDs.

Alternatively, the three-electrode configuration is also adopted to test the electrochemical performances of electrode materials. For instance, Majumder et al.³⁹ tested the PV response of hydrothermal ZnO nanorod arrays (ZnO NRAs) grown on a FTO substrate using a three-electrode system, in which Ag/AgCl served as the reference electrode, a Pt wire as the counter electrode, and 0.1 M Na₂SO₄ aqueous solution as the electrolyte. The ZnO NRAs show a stable and repeatable PV response, showing a quick increase in PV but a slow recovery. The ZnO NRAs show the maximum PV at 350 nm and good visible-blind characteristics. In a strict sense, this three-electrode configuration is not a real device. However, it is technically more challenging to precisely analyze and discuss limited results from the reported two-electrode PEC cells. So, to give a far and wide discussion on PEC UV-PDs, the results from the three-electrode system have also been put into the whole frame for comparison.

3.2 PEC UV-PDs based on various electrolytes

The electrolyte is one of the most crucial components that provide pure ionic conductivity between the photoanode and counter electrode in PEC UV-PDs.⁴⁹ Redox electrolytes in UV-PDs can not only function as the medium for the inner carrier transport between electrodes, but also continuously reactivate the semiconductor and itself during the operation. Thus, several aspects are essential for the electrolytes used in PEC UV-PDs: (1) the electrolytes must be able to transport the carriers between the photoanode and counter electrode. Once the electrons are promoted from the VB to CB, the remaining holes will be rapidly oxidized by electron donors in the electrolyte. (2) The electrolytes must guarantee a fast drift of carriers (higher conductivity) and show large interfacial contact with the mesoporous semiconductor photoanode and the counter electrode. (3) The electrolyte should be chemically stable and inert to photoanode materials, which guarantees the efficiency and long-term stability of the PEC devices. According to the physical state, compositions, and formation mechanisms, the electrolytes used in PEC UV-PDs can be broadly classified into: aqueous electrolytes, organic liquid electrolytes, or quasi-solid gel electrolytes. Xie and Shen's groups reported PEC self-powered UV-PDs assembled using the traditional liquid I⁻/I₃⁻ redox couple electrolyte.^{22,28,29,31,32,35,45,47} And impressive performances were observed in these UV-PDs. However, the liquid I⁻/I₃⁻ redox couple electrolyte is not ideal for long-term operation, because it is highly corrosive, volatile, and photo-reactive, which often makes it interact with common metallic components and sealing materials.

Aqueous electrolytes are more suitable for PEC self-powered UV PDs due to the fact that water is one of the safest, most stable, and most environment-friendly electrolytes. And a type of UV-PD based on TiO₂/water solid–liquid heterojunction has been proposed recently.^30,36,43 Interestingly, this TiO₂/water-based UV-PD can exhibit high photosensitivity, excellent spectral selectivity, linear variations in I_p, and fast response.⁴³ Moreover, Chen's group reported that TiO₂ NR/water-based self-powered UV-PDs showed a spectral responsivity of 25 mA W⁻¹ and a fast photoresponse.³⁶ Furthermore, many aqueous electrolytes have been used during PEC tests, such as Na₂SO₄^{39,42,44,50–57} and KOH.^58–60 Cao et al. reported the I_p response of TiO₂ nanorod arrays (NRAs) under UV illumination using a 0.5 M Na₂SO₄ aqueous electrolyte,⁴⁴ in which TiO₂ nanostructures can harvest more incident light photons compared to a flat thin-film active layer because of the markedly enlarged TiO₂/electrolyte contact area. The photocurrent density (J_p) can reach 12.87 μA cm⁻² under UV illumination (365 nm, 3 mW cm⁻²). Though aqueous electrolytes are safe, stable, and environment-friendly, the performances of these UV-PDs are poorer compared to those with the I⁻/I₃⁻ electrolyte. So, further research is required for the development of the aqueous electrolyte.

Using aqueous electrolytes as the medium for the carrier transport, great developments have been achieved in various PEC devices. However, liquid electrolytes often cause some practical problems, such as leakage, volatilization of solvent, corrosion of counter electrode, and ineffective sealing of the cells for long-term applications. Recently, a type of quasi-solid-state electrolyte has been used in PEC cells. Han et al. demonstrate a high spectrum selectivity from TiO₂-NRA-based UV-PDs by using poly(9,9-dihexylfluorene) as the electrolyte.⁴¹ The maximum response of 33.2 mA W⁻¹ at 395 nm is obtained under zero bias. However, the J_p is only about 1.78 × 10⁻⁵ μA cm⁻² under 0.7 mW cm⁻², 365 nm UV illumination. More recently, Chen's group reported nanostructured quasi-solid-state UV-PDs fabricated using a liquid crystal (LC)-embedded I⁻/I₃⁻ electrolyte with a light-trapping scheme.³³ The maximum responsivity is about 90 mA W⁻¹ at 383 nm; the maximum IPCE is about 29%. Under 2.5 mW cm⁻² UV illumination (365 nm, 0 V), the J_p reaches 175 μA cm⁻² with t_r less than 0.03 s. Moreover, Game et al. report a facile solution-processed fabrication of a self-powered organic–inorganic hybrid PD using n-type oriented ZnO NRs and a p-type Spiro-MeOTAD semiconductor.⁴⁰ The assembled UV-PDs show a high S (10²), a high UV/visible ratio (300) and a fast photoresponse (t_r of 200 μs and t_d of 950 μs). Under UV illumination (390 nm, 3 mW cm⁻²), a J_p of 50 μA cm⁻² can be obtained. Although the efficiencies of these UV-PDs with quasi-solid-state electrolytes are often lower than those of the UV-PDs with liquid electrolytes, the quasi-solid electrolytes may become viable alternatives to the liquid electrolytes owing to the greatly improved stability and sealing ability, which will be beneficial to their practical applications.

3.3 PEC self-powered UV-PDs based on various photoanodes

So far, most work to enhance the performances of PEC UV-PDs has focused on developing photoanodes with high photoresponse, including spectral responsivity, photosensitivity, and response time. The key findings can be seen in Table 1. Researchers have used different materials, micro-/nano-structures, electrolytes, counter electrodes, and even diverse device structures to report the UV detecting performances of the assembled PDs, which makes it difficult to directly compare some results in the literature. In general, short-circuit photocurrent (I_sc) or density (J_sc) of electrode materials is obtained using either the assembled two-electrode cell or the three-electrode system. Moreover, many studies have reported the “visible-blind” and “linear output” characteristics, even though the estimate methods of photosensitivity, ON/OFF ratio, and rise and decay time are not always consistent. Thus, we have listed the different methods used in Table 1 for easy comparison. And it can be seen that various nanostructures were exploited for constructing photoelectrodes in PEC cells, including NPs, NRs, NWs, NTs, NRAs, nanosheets (NSs), and nano-forests.⁶¹ It is suggested that the electron transport rates of these nanostructures mentioned above are in the sequence of NP films < NR films < nano-forest films < NRA films, while the sequence for specific surface area (SSA) turns out to be NP films > NR/NW films > nano-forest films > NTA films > NRA/NWA films. Thus, there is a need for a strategy to balance the choice in constructing PEC electrodes.

On the other hand, the band-gap energy of the photoanode for UV-detection should be larger than 3.1 eV, corresponding to the wavelength of 400 nm. The recently reported wide-band gap oxide nanostructures (including binary metal oxides like ZnO (ZnS),⁴² TiO₂,¹¹ SnO₂,⁶² NiO,⁶³ Nb₂O₅,⁶⁴ Ga₂O₃,⁶⁵ ZrO₂,⁶⁶ and ternary metal oxides like SrTiO₃ ⁶⁴) can be potentially applied as photoanode materials for PEC UV-PDs (Fig. 2). So far, ZnO, TiO₂ and SnO₂ are the most frequently studied materials for PEC cells, and in this context, we present them in three sub-sections, i.e. ZnO, TiO₂, and SnO₂ nanostructured photoanodes.

Fig. 2 Schematic diagrams of the energy levels of the CB and VB of various wide-band-gap semiconductors.

3.3.1 TiO₂ nanostructures for PEC UV-PDs. TiO₂ is an n-type semiconductor with a wide band gap (anatase 3.2 eV and rutile 3.0 eV) and has been frequently studied for UV detection due to its distinctive UV absorption characteristics.^67,68 TiO₂ NP, NR, NT, and nano-forest films have been employed in the photoanodes for PEC UV-PDs.^{22,34,36,44,45,59} In 2012, Xie's group first formally proposed a prototype for PEC UV-PDs, as shown in Fig. 1e and f.²² The nc-TiO₂ films were used as photoanodes, which were prepared by the traditional sol–gel method and spin-coated on FTO glass. The resultant anatase TiO₂ nanoparticles possess an average diameter of ∼17 nm (Fig. 3a), and the optimized film thickness is ∼450 nm for 365 nm UV light in this case. Then, a dried TiO₂ photoanode and a platinized counter electrode were assembled into a sandwich cell, and the traditional I⁻/I₃⁻ ionic liquid was used as the electrolyte. The assembled devices show the highest I_p response at ∼330 nm (Fig. 3b), along with a good UV selectivity and a visible-blind property, which is very important for a UV-PD. Under UV illumination (365 nm, 33 mW cm⁻²), the J_sc signal is reproducible under the “ON” or “OFF” states of UV light, and reaches 550 μA cm⁻² with the R_λ of 16.7 mA W⁻¹ (Fig. 3c). Moreover, the S of the PDs is found to be about 269.8, and the t_r and t_d are ∼0.08 and 0.03 s for J_sc signals, respectively (Fig. 3d and e). These results suggest that these assembled cells possess a high photosensitivity to UV light (365 nm). And the J_sc increases linearly with increasing UV light intensity in a wide variation range from 25 μW cm⁻² to 33 mW cm⁻² (Fig. 3f), which might benefit their real application in quantitative measurement for UV light. However, the values of R_λ and S are not as high as the published results.⁴³ Nevertheless, this work successfully started a new research area of PEC UV-PDs.

Fig. 3 (a) SEM image of the spin-coated TiO₂ films; the inset is the corresponding XRD spectrum; (b) spectral responsivity of the self-powered UV-PDs, indicating a visible-blind characteristic; (c) time responses of J_sc under UV illumination (365 nm, 33 mW cm⁻²) measured for light-on/off states; (d and e) enlarged rise and recovery edges of the current response at a light intensity of 33 mW cm⁻², respectively; (f) J_sc and V_oc as a function of the incident UV (365 nm) intensity. Reproduced from ref. 22. Copyright © 2010 Elsevier Ltd.

Another suitable nanostructure of TiO₂ for UV-PDs is vertical NRAs due to their relatively high transport rates of electrons and enhanced transfer of electrolyte ions because of more interfaces between the NRs and the solution compared to a flat thin-film active layer. It is demonstrated that the basic parameters of NRAs (such as diameter, length and density) greatly affect the performances of PEC cells. Cao et al.⁴⁴ found that under UV illumination (365 nm, 3 mW cm⁻², 0.5 M Na₂SO₄, three-electrode system), 3 μm long TiO₂ NRAs show a J_p of about 15 μA cm⁻², while 6 μm long TiO₂ NRAs only 6 μA cm⁻², suggesting that a suitable length of NRAs is required for a high J_p. In view of this point, Chen's group applied vertical rutile TiO₂ NRAs (∼1 μm in height, 100–150 nm in diameter, 20 NRs per μm² in density) to PEC UV-PDs (Fig. 4a and b).³⁶ Under UV illumination (365 nm, 1.25 mW cm⁻², DI water), the J_p is 4.67 μA cm⁻² with the R_λ of 4 mA W⁻¹. This low responsivity should be mainly attributed to the high resistance of the used water electrolyte. Meanwhile, this device shows a fast photoresponse, i.e., a t_r of 0.15 s and a t_d of 0.05 s, indicating a rapid photoresponse characteristic. Moreover, the maximum responsivity of the photoanodes reaches about 25 mA W⁻¹ at 350 nm.

Fig. 4 (a) Schematic device structure of the PEC UV-PDs using water as the electrolyte, and (b) top-view SEM image of TiO₂ NRAs. (c) Schematic formation process of B-TiO₂ NRAs grown on FTO glass, (d) scheme of the possible electron transport mechanism in a B-TiO₂ NRA photoanode, (e) top-view SEM image of B-TiO₂-18 h NRAs, and (f) time responses of J_p values for different PDs under UV light pulsed illumination (365 nm, 2 mW cm⁻²). (a and b) Reproduced from ref. 36. Copyright © 2013 Springer. (c–f) Reproduced from ref. 34. Copyright © 2014 IOP Publishing.

Compared to NRAs, TiO₂ NTAs (TNTAs) are also attractive in the field of PEC applications, such as DSSCs, photocatalysts, and UV-PDs, due to their enlarged TiO₂/electrolyte interface and well-defined carrier transport path, and most importantly, ease of fabrication and low cost.^6,69 Li et al.⁵⁹ reported that under UV illumination (365 ± 15 nm, 70 mW cm⁻², 0 V bias), the 3 μm long TNTAs exhibit a J_p of ∼100 μA cm⁻² with the R_λ of 1.4 mA W⁻¹. And the J_p can also quickly drop to zero when switching off the light source, indicating fast charge transport in NTAs. However, compared to the nc-TiO₂, both TiO₂ NRA and NTA films show a much lower R_λ, which may be due to their lower SSA and thus lower light harvesting efficiency.

It is reported that branched TiO₂ (B-TiO₂) nanostructures (i.e., nano-forest) usually possess better charge transport and light absorption properties than NP films, and larger SSAs than NRAs/NTAs.^70–72 Chen's group prepared B-TiO₂ NRAs by a facile two-step chemical synthesis process and assembled them into photoanodes (Fig. 4c and d).³⁴ Under UV illumination (365 nm, 2 mW cm⁻²), the assembled PEC UV-PDs (18 h, Fig. 4e and f) also exhibit a high J_p of 373 μA cm⁻² with the R_λ of 186.5 mA W⁻¹, which is greatly enhanced compared to those of the above single-phase TiO₂ NRA/NTA and nc-TiO₂ films. Moreover, the UV-PDs also show a fast photoresponse (t_r of 0.15 s and t_d of 0.05 s). These improved performances are attributed to a markedly enlarged TiO₂/electrolyte contact area and a high electron conductivity in the TiO₂ NR trunk. However, it can be seen that the enhanced light scattering by B-TiO₂ NRAs leads to lower transmission in the visible light region, which is harmful to spectral selectivity of a UV-PD.

3.3.2 ZnO nanostructures for PEC UV-PDs. Due to its easy fabrication, low cost, and unique physical/chemical properties, ZnO is applied as building units in many optoelectronic devices, such as solar cells, nanogenerators, and field effect transistors.^61,73 Recently, the application of ZnO in UV light detection has attracted much interest because of its wide band gap (3.37 eV) and high electron mobility (200 cm² V⁻¹ s⁻¹).^67,74 It has been demonstrated that rationally designed ZnO nanostructures with a complex shape and compositions can become one of the most promising building blocks as detectors for the UVA range.⁴ NP films were first used in various optoelectronic devices. Although NP films possess the highest SSA among various types of nanostructures, their high recombination losses and long transport distance of photogenerated carriers greatly reduce their optoelectronic performances.⁷⁵ It was found that under UV illumination (365 nm, 5 mW cm⁻²) using a 0.5 M Na₂SO₄ solution as the electrolyte and the three-electrode system, the highly-crystalline ZnO films only exhibit a J_sc of ∼60 μA cm⁻² with the R_λ of 12 mA W⁻¹,⁵⁴ which is even lower than that of nc-TiO₂ films.²²

ZnO NRAs/NWAs have also been frequently used in PEC cells, such as water splitting and PV systems, due to their good conduction of electrons and holes, nontoxicity, environment friendliness, and controllable low-temperature synthesis.^76,77 Compared to NP films, NRAs usually possess higher light absorbance efficiency and lower electron recombination. Recently, Chen's group reported a type of PEC UV-PD using hydrothermal-grown ZnO nanoneedles as the active photoanode and DI water as the electrolyte (Fig. 5a).³⁰ The resultant ZnO nanoneedles possess an average length of 2–3 μm and a diameter of 80–100 nm at the base (Fig. 5b). The maximum R_λ of ZnO nanoneedle arrays is ∼22 mA W⁻¹ at ∼385 nm, showing good spectral selectivity in the UVA range. Under UV illumination (365 nm, 1.25 mW cm⁻²), the J_P is 4.08 μA cm⁻² with the R_λ of 3.26 mA W⁻¹. Both t_r and t_d of the device are ∼0.1 s for J_sc, which will make ZnO nanoneedle arrays show potential application in UV-PDs. However, the photoresponse region is obviously broadened to over 450 nm because of the light scattering between nanoneedles. This is unfavorable to UV detection. It is found that there exists a gradual saturation in J_p under higher UV illumination, which might be due to the poor carrier transportability of water.

Fig. 5 (a) Schematic device structure of UV-PDs based on ZnO nanoneedle arrays and (b) SEM image of ZnO nanoneedle arrays; the inset shows their cross-sectional view. (c) Schematic growth route for ZnO NWAs with different branches, (d and e) SEM images of ZnO NWAs after one time branched growth and ZnO NW nano-forest, respectively. (a and b) Reproduced from ref. 30. Copyright © 2013 Springer. (c–e) Reproduced from ref. 78. Copyright © 2011 American Chemical Society.

Hierarchical branched nanostructures are other promising structures that provide a high light absorption for the optoelectronic devices.⁷⁹ Ko et al.⁷⁸ studied the effect of branched growth on the efficiency of hierarchical ZnO NWA photoanodes (Fig. 4c–e), and found that after one-time branched growth, the J_sc of 7 and 13 μm ZnO NWAs are significantly enhanced 389% and 256% compared to the pristine ones, respectively; while with one more branched growth, the J_sc of 13 μm ZnO NWAs can be further slightly enhanced ∼14%. Thus, this J_sc increase can be explained by considering a combination of several effects:⁸⁰ (i) enhanced light absorption from the enlarged SSA results in a large J_sc increase. (ii) The dense network of crystalline ZnO NWs can increase the electron diffusion length and electron collection because the NWs provide direct conduction paths for electrons’ transport from the injection points to the collection electrode. (iii) Randomly branched NWs promote enhanced light-harvesting without sacrificing efficient electron transport. Furthermore, branched NWs also increase light-harvesting efficiency by scattering enhancement and trapping, which might be unfavorable to spectral selectivity of the assembled UV-PDs.

3.3.3 SnO₂ nanostructures for PEC UV-PDs. Tin oxide (SnO₂) is an important n-type semiconductor with good sensitivities, including chemical sensing, bio-sensing and UV sensing properties.⁹ Thus, SnO₂ has been considered to be another most promising alternative of TiO₂ for PEC UV-PDs due to its higher electronic mobility (∼100–200 cm² V⁻¹ s⁻¹) and larger band gap (3.6–3.8 eV) than TiO₂ (∼0.1–1.0 cm² V⁻¹ s⁻¹, 3.2 eV), fast response and “visible blind” characteristics together with excellent long-term stability. However, SnO₂-based PEC cells, including UV-PDs, were developed with less success, and the performances of SnO₂ photoelectrodes reported so far are much lower than those of TiO₂. In view of this situation, Xie's group has conducted a series of studies on enhancing the PEC performances of SnO₂ photoelectrodes using the traditional I⁻/I₃⁻ electrolyte.^28,31,47,81 First, SnO₂ nanofibers (NFs) were prepared using a mixture solution of polyvinylpyrrolidone (PVP) and SnCl₂via an electrospinning technique.^28,81 Under UV illumination (330 nm), the PEC UV-PDs based on SnO₂ NFs exhibit a J_sc of 15.8 μA cm⁻², which is far lower than that of nc-TiO₂ films (34.7 μA cm⁻²). However, the SnO₂ NFs show a high J_d of 10.7 μA cm⁻², indicating a poor photodetecting performance. To further increase the SSA of electrospun SnO₂ NFs, SnO₂ NTs were designed combining a nanoscale Kirkendall effect (Fig. 6a).⁸³ Compared to SnO₂ NFs, the J_sc and η values of SnO₂-NT photoanodes are improved by ∼30% and 39%, respectively, which is still far lower than those of P25.⁸¹ On the other hand, the SnO₂ NSs directly grown on FTO glass were also applied to PEC UV-PDs.⁴⁷ Under UV light illumination (365 nm, 40 mW cm⁻²), the UV-PDs show a J_sc of 30 μA cm⁻² with the R_λ of 0.75 mA W⁻¹, indicating a strong electron recombination at the interface between SnO₂ NSs and the electrolyte. In view of this situation, Xie's group further assembled SnO₂ NSs into hollow nanospheres (HNSs) via a low-temperature synthesis (Fig. 6b–d),³¹ and found that under UV irradiance at 330 nm, the SnO₂ HNS photoanodes show a much improved J_sc of 31.2 μA cm⁻² with the η of 3.3%, which is still far lower than that of nc-TiO₂ (40.8 μA cm⁻² with an η of 7.8%).

Fig. 6 (a) Schematic formation mechanism of SnO₂ NFs and NTs. (b) Schematic illustration of the formation of SnO₂ HNS, (c and d) SEM images of the prepared SnO₂ HNSs; the inset in (d) is the TEM image. (e) Typical SEM images of SnO₂ cloth woven by microtubes; the insets show amplified SEM images of a single SnO₂ microtube assembled by numerous NPs. (f) Structure scheme of the PEC UV-PDs; the inset shows an optical image of SnO₂ cloth. (a) and (e) Reproduced from ref. 81 and 82, respectively. Copyright © 2012 and 2013 The Royal Society of Chemistry. (b–d) Reproduced from ref. 31. Copyright © 2014 Elsevier Ltd. (f) Reproduced from ref. 32. Copyright © 2014 Wiley-VCH.

More recently, Shen's group has prepared a type of SnO₂ microtubes using carbon cloth as sacrificial templates,^32,82 which can be directly used as a binder-free photoanode (Fig. 6f). These microtubes consist of rutile SnO₂ NPs with the size of several tens of nanometers (Fig. 6e). The assembled PEC UV-PDs using the I⁻/I₃⁻ electrolyte show a high J_p response of 78 μA cm⁻² with the R_λ of 50.3 mA W⁻¹ under UV illumination (365 nm, 1.55 mW cm⁻²). Compared to TiO₂-based photoanodes, the poor performances observed from SnO₂ ones are mainly attributed to the fast interfacial electron recombination and lower trapping density.⁸⁴ Thus in a later section, we will give a review on recent studies of the PEC performances of SnO₂-based hybrid photoanodes, including SnO₂ NF/NP, and SnO–TiO₂ hybrids.

3.4 Modification of photoanodes for high-performance UV-PDs

Similar to an optoelectronic cell, the strategies for enhancing the performance of PEC PDs usually include three aspects, i.e., enhancing light harvesting in photoanodes (for high R_λ and S), enhancing charge separation or reducing charge recombination (for a repeatable and fast J_sc response), and enhancing charge transport (for short t_r and t_d). In this section, we will give a systematic introduction to these three strategies and detailed discussions on the design of photoanodes for high-performance PEC UV-PDs.

3.4.1 Enhancing the light harvesting efficiency. PEC UV-PDs are usually composed of an inorganic semiconducting photoanode, an electrolyte, and a platinized counter electrode in a sandwich configuration. Improving the light-harvesting efficiency and suppressing the carrier recombination process at the semiconductor/electrolyte interface are two key factors to improve the performance of the UV-PDs. And various approaches to enhance the light-harvesting efficiency have been reviewed in this sub-section, including porous films, length-wise NRAs, density-wise NWAs, and 3D micro-/nano-architectures.
3.4.1.1 Porous films. On entering a porous NP layer, the incident light can generally be divided into two parts: (i) one part is adsorbed by NPs, and (ii) the other part is penetrating the NP film after multiple refractions, which can increase the light harvesting. Then, based on this point, Mridha et al.⁸⁵ prepared a type of sponge-like porous ZnO with the assistance of activated carbon (AC). This AC-assisted ZnO (AC-ZnO) shows a SSA of 3.17 m² g⁻¹, which is increased by 33% compared to that of the bare ZnO (B-ZnO). The UV absorption of AC-ZnO in the range of 300–400 nm is enhanced 10% compared to that of B-ZnO. Moreover, the S of AC-ZnO is 162, which is four times greater than that of B-ZnO. And both t_r and t_d of AC-ZnO are 5–6 times shorter than those of B-ZnO, indicating a fast photoresponse. Moreover, Hu et al.⁸⁶ fabricated a type of mesoporous TiO₂ photoanode via a gel hydrothermal method using PVP as a pore-forming agent. It is found that the absorbance of TiO₂-1.5 wt% films is enhanced 23% in the UV region compared to that of dense TiO₂ films, implying that the multiple refractions can greatly enhance the light absorption efficiency in TiO₂ photoanodes. And the counterpart PEC cells yield an η of 9.86%, which is enhanced 53.8% compared to that of the dense TiO₂ ones. This enhancement in η suggests that the porosity and pore size of TiO₂ photoanodes not only enhance the light harvesting efficiency, but also affect the SSA of photoanodes for greatly enhanced electrolyte ion adsorption and interfacial charge transfer. Furthermore, Lei et al.⁵³ studied the effect of pore size on the optoelectronic performances of the porous films prepared via the sol–gel technique using polystyrene (PS) sphere templates. Under UV illumination (369 nm, 32 mW cm⁻²), the J_p of the arrayed 190 nm pore films is 10 μA cm⁻², which is ∼5 and 10 times as high as those of the 850 nm pore and nonporous films. The sample with a smaller pore size has larger SSA and is more effective in the improvement of the photocurrent.
3.4.1.2 Lengthwise NRAs. Generally, the insufficient SSA of NWAs constrains the J_sc to relatively low levels due to their low light absorbance. Ko et al.⁷⁸ performed a parametric study of the effect of NWA length on the efficiency of the photoanodes (Fig. 7a), and found that as the length of ZnO NWAs increases from 7 to 18 μm, the J_sc and η values are enhanced by 90% and 89%, respectively. This increase in J_sc and η is mainly due to the effective SSA increase. Moreover, Xie's group also studied the effect of NWA length on J_sc and η (Fig. 7b),⁸⁷ and found that on increasing the length of the NWAs from 6 to 23 μm, the J_sc and η values are enhanced by 285% and 283% respectively. It is noted here that the η increases by 85% when the NWA length increases from 6 to 19.4 μm, while it increases by 105% when the NWA length increases slightly from 19.4 to 23 μm. This increasing rate of enhancement is found to be due to the porous structure of ZnO NWs formed by etching during the long-time hydrothermal growth process. To avoid this situation, Xu et al.⁸⁸ demonstrate a convenient approach for synthesizing multilayer assemblies of ZnO NWAs (Fig. 7c–e), and found that the assembled four-layer ZnO NWAs possess an internal SSA more than 5 times larger than that of single-layer NRAs, and yield a high η of 7%, which is ∼2.3 times higher than that of single-layer NRAs.

Fig. 7 (a) Schematic growth route for ZnO NWAs with different NW lengths; their SEM images are shown in the bottom. (b) J_sc and η of ZnO NWAs as a function of NW length; the inset SEM images show the morphology changes of ZnO NWs. (c) Schematic process for synthesizing the multiple-layer assembly of ZnO NRAs and (d and e) SEM images of two- and four-layer assembly of ZnO NRAs, respectively. (a) and (c–e) Reproduced from ref. 78 and 88, respectively. Copyright © 2011 American Chemical Society. (b) Reproduced from ref. 87. Copyright © 2014 Elsevier Ltd.

3.4.1.3 Densitywise NWAs. Apart from length, the density of NW in an array can also affect the optoelectronic performance of the electrodes, especially light absorption in the visible region due to light scattering, which is unfavorable to UV detection. Hu et al.⁹⁰ designed various densities of ZnO NRAs using square patterns, and found that the patterned ZnO NRAs show a slight increase in UV absorption, and a large increase in visible absorption. Thus, a suitable density of NWA is required for low absorption in the visible region. In view of this point, Xie's group has prepared a type of dense dendriform TiO₂ (D-TiO₂) NWAs to serve as the photoanode of PEC self-powered UV-PDs.⁴⁵ The aspect ratio of the NWs is up to 150 and the surface density is about 90–110 NW μm⁻². The grown branches possess a cone shape with an average length of 80 nm and a base diameter of about 15 nm, densely filling the interspace between TiO₂ NWs. The UV light absorption of D-TiO₂ NWA films is enhanced ∼45% compared to that of TiO₂ NWA, and with a very low light absorption in the visible region. And under UV light illumination (330 nm), the J_sc of D-TiO₂ NWA films is respectively enhanced 176% and 46% compared to those of TiO₂ NWA and nc-TiO₂ films. Especially, the responsivity in the visible region drops 1/396 for J_sc as compared with the peak value, indicating a visible-blind property. Under UV illumination (365 nm, 25 mW cm⁻²), the J_sc value exceeds 570 μA cm⁻² with the R_λ of 22.8 mA W⁻¹ and the S of 1902. The t_r and t_d of the D-TiO₂ NWA based PEC cell are ∼5.3 and 5.9 ms for J_sc, which is much faster than the photoconductivity-based UV detectors made by 1D TiO₂ nanostructures.^2,74 In addition, the J_sc increases linearly on increasing the UV light intensity in a wide variation range from 0.05 to 27 mW cm⁻², showing a potential application for measuring UV intensity.
3.4.1.4 3D hierarchical architectures. Recently, Shen's group has reported a type of flexible and transferable TiO₂ NR cloth (TNRC) synthesized using carbon cloths as sacrificial templates, as illustrated in Fig. 8a.⁸⁹ The as-synthesized TNRCs are assembled by numerous aligned TiO₂ NRs with diameters of about 100 nm (Fig. 8b), showing good transferability and flexibility (Fig. 8c). Illuminated by UV light (365 nm, 1.25 mW cm⁻², −1 V), the assembled sandwich device without an electrolyte show an R of only ∼10 mA W⁻¹, a t_r of ∼1.4 s, and a t_d of ∼6.1 s. These low optoelectronic performances are due to the relatively lower film thickness and NR density compared to the traditional NP films, and a large proportion of the photons will become transmission loss for TNRCs. Then, Shen's group further tried to combine multilayers of TNRCs with vertical TiO₂ NRs (TNRs), and obtained a type of 3D hierarchical TNRC structures.³⁵ It is found that the use of the hierarchical 3D TNRC structures can greatly reduce the “transmission loss” problem in TNRC films (Fig. 8d and e). And the IPCE of the TNRCs combined with TNRs increases from ∼32.5% to ∼45% at ∼360 nm. Although the spectral response range is slightly widened, the photoresponse is still narrow enough to meet the “visible blind” property (Fig. 8f). Under UV illumination (365 nm, 1.55 mW cm⁻², inset of Fig. 8f), the PEC UV-PDs based on TNRC@TNRs using the I⁻/I₃⁻ electrolyte show a much higher photoresponse of 278 μA cm⁻² with an R of 179 mA W⁻¹, which is enhanced 18% compared to that of TNRs. Moreover, the t_r and t_d of TNRC@TNRs are 0.3 and 0.2 s, respectively. This indicates that introducing TNRC into UV-PDs increases the photoresponse by reducing light transmission, while retaining the carrier transport efficiency. Furthermore, all devices can not only respond to very weak optical signals, but also show highly-linear properties, which can make them ideal candidates for UV radiation-level measurement. However, the 30% increase in light absorption only leads to a 12% increase in IPCE, which is mainly due to the poor attachment existing between the TNRCs and the TNRs. Nevertheless, this type of freestanding multilayer nanostructured photoanode opens a new route for fabricating high quality UV-PDs with a fast and high response.

Fig. 8 (a) Scheme for fabricating the TNRCs, (b) top-view SEM image of the TNRCs; the insets show enlarged top and cross-sectional view SEM images, respectively, and (c) optical images of TNRCs under a ruler, and the inset (right) shows their optical image under bending. (d) Design and principles for the TNRC/TNR structure to achieve better performance, (e) scheme for fabricating the TNRC/TNR structure, and (f) IPCE testing for the PEC PD; the inset shows photoresponse of different PEC detectors under UV light rectangular pulses (365 nm, 1.55 mW cm⁻²). (a–c) Reproduced from ref. 89. Copyright © 2011 American Chemical Society. (d–f) Reproduced from ref. 35. Copyright © 2012 The Royal Society of Chemistry.

3.4.2 Enhancing charge separation or reducing charge recombination. Interfacial charge recombination is a problem that exists in PEC cells and causes the loss of photogenerated electrons.⁷⁵ Recently, semiconductor core–shell structures with type-II band alignment have attracted significant attention as the elementary building blocks for the fabrication of next generation electronics, because of their rapid carrier separation/injection to the photoanodes.⁹¹ In core–shell structures, it is required that the CB potential of the shell should be more negative than that of the core, which will establish an energy barrier at the semiconductor/electrolyte interface. Then, this established energy barrier can help in the photogenerated carriers’ separation and thus suppress interfacial charge recombination. According to the energy levels shown in Fig. 2, the core–shell structures of ZnO@TiO₂, ZnO@NiO, ZnO@ZnS, TiO₂@SrTiO₃ and SnO@TiO₂ possess great potential in high-performance PEC devices.
3.4.2.1 ZnO-based core–shell structures. The high I_d in ZnO nanostructured photoanodes caused by photo-corrosion in aqueous solution will greatly limit the photosensitivity of the devices. Thus, coating a layer of TiO₂ on ZnO nanostructures can effectively protect the ZnO NWs from degradation and corrosion in basic media under light illumination.⁹² Moreover, it is also demonstrated that the I_d was largely blocked and the J_sc was increased because of the CB offset of 0.1 eV and the VB offset of 0.2 eV between the ZnO core and the TiO₂ shell (Fig. 9b).^93–95 In 2013, Xie's group reported a type of PEC UV-PD based on ZnO@TiO₂ core–shell nanostrawberries (ZnO NS-TiO₂) using the traditional I⁻/ I₃⁻ electrolyte (Fig. 9a).²⁹ The resultant TiO₂ shell is 3 nm in thickness (inset of Fig. 9d), and the ZnO-NS core is ∼250 nm in length, ∼155 nm at the big side, ∼91 nm at the small side, consisting of ∼22 nm rod-like NPs (SEM image in Fig. 9c). Under UV illumination (365 nm, 20 mW cm⁻², Fig. 9d), the ZnO NS-TiO₂ based device gives a high J_sc of 357 μA cm⁻² with an R_λ of 17.85 mA W⁻¹, which is enhanced 52.5% compared to that for the ZnO-NS based one, indicating an improved R_λ. The S of the J_sc signal from ZnO NS-TiO₂ based devices is ∼37 900, which is 3.24 times as high as that from ZnO-NS based ones. The t_r and t_d are estimated to be 0.022 and 0.009 s for J_sc, respectively. Moreover, J_sc increases linearly with increasing the 365 nm UV light intensity in a wide variation range from 0.002 to 40 mW cm⁻².

Fig. 9 (a) Schematic of self-powered UV-PDs based on ZnO NS-TiO₂ films, (b) energetics of operation of the UV-PDs, (c) schematic diagram of the formation mechanism of strawberry-like and sphere-like ZnO aggregates, the right insets are their corresponding SEM images, and (d) J–V curves of the self-powered UV-PDs based on ZnO NS-TiO₂ and ZnO NS films. The inset is the HRTEM image of ZnO NS-TiO₂ taken from the ZnO/TiO₂ interface. Reproduced from ref. 29. Copyright © 2013 Elsevier Ltd.

Apart from TiO₂, other wide-band-gap semiconductors, such as NiO and ZnS, can also be applied to enhance the optoelectronic performances of ZnO-based nanostructures. NiO (p-type, 3.7 eV) can easily form a p–n heterojunction with n-type ZnO and generate an inner electric field at their interface of heterostructures. Dai et al.⁶³ constructed a type of honeycomb-like ZnO@NiO core–shell NRA via a photochemical deposition process, and found that the S of ZnO@NiO-NRA based PDs is as high as 167, which is nearly enhanced by 49 fold compared with the bare ZnO NRA under 365 nm UV illumination. Both t_r and t_d from ZnO@NiO NRAs are also greatly decreased to 1/7 of those from the ZnO NRAs. As for ZnS, it is also a wide-band-gap material with an especially high CB position, which can produce excitons with a strong driving force for reactions. Guo et al.⁴² found that ZnO@ZnS NRAs show similar UV light absorption to ZnO NRAs, but with higher transmittance in the visible light region, which is favorable to UV detection. Under UV illumination (0 V bias), the J_sc of ZnO@ZnS NRAs is enhanced over 100% compared to that of ZnO NRAs, which is mainly due to the high separation efficiency of photogenerated carriers.

3.4.2.2 TiO₂–SrTiO₃ core–shell structures. TiO₂ is one of the most commonly used materials in the photoanodes for PEC cells, but it shows low electron mobility, leading to slow electron diffusion transport and high recombination. It is expected that the heterojunction of photoanodes could facilitate the separation or suppress the recombination of photogenerated carriers at the semiconductor interface, resulting in the final high efficiency of the photoanodes. SrTiO₃ is a wide-band-gap (3.4 eV) optoelectronic ceramic material, which possesses a high PEC activity. Kim et al.⁴⁶ fabricated a type of TiO₂@SrTiO₃ core–shell heterojunction NTA by anodization of Ti foils and sequential hydrothermal reaction. The maximum η of the TiO₂@SrTiO₃-1 h NTAs is 6.6%, which is enhanced 65% compared to that of the bare TiO₂ ones. Under UV illumination (250–385 nm), they exhibit a quick and repeatable photoresponse with a J_P of ∼650 μA cm⁻², which is 3-fold that of the bare TNTAs. This enhancement is mainly because of their charge separation and suppression of charge recombination at the SrTiO₃/TiO₂ interface.
3.4.2.3 SnO₂–TiO₂ core–shell structures. Thus far, it has been demonstrated that the application of SnO₂ to PEC cells has been greatly limited due to its poor PEC behavior, which primarily arises from a larger interfacial electron recombination and lower trapping density. To solve this problem, many researchers have tried to coat TiO₂ nanostructures onto SnO₂ cores. This combination can simultaneously solve the large interfacial electron recombination of photogenerated carriers in SnO₂ and the slow diffusion transport of electrons in TiO₂ (Fig. 10c and h), which are mainly due to two aspects: (1) SnO₂ possesses higher electron mobility (∼100 to 200 cm² V⁻¹ s⁻¹) than TiO₂ (∼0.1 to 1.0 cm² V⁻¹ s⁻¹), which can work as an oriented charge transport path,⁶⁴ and (2) TiO₂@SnO₂ core–shell structure can yield a surface dipole layer toward SnO₂, which can benefit suppressing the interfacial charge recombination.^81,96–98 Xie's group reported a type of branched nanostructures of TiO₂ nanoneedles on the SnO₂ NF network (B-SnO₂), and applied this type of B-SnO₂ to the photoanodes for PEC UV-PDs (Fig. 10a and b).²⁸ Under UV irradiation (330 nm), the η of PDs reaches 14.7%, which is over twice as large as that of nc-TiO₂ films (6.4%), indicating that these branched heterojunction nanostructures can simultaneously exhibit a low degree of charge recombination and work as a direct pathway for electron transport. And under UV illumination (365 nm, 40 mA cm⁻²), the PDs exhibit a J_sc of 900 μA cm⁻² with an R_λ of 22.5 mA W⁻¹, an S of ∼4550, a t_r of 0.03 s, and a t_d of 0.01 s for J_sc signals. Moreover, a type of branched TiO₂ nanoneedle on SnO₂ NS (SnO₂ NS-TiO₂) was further applied to PEC UV-PDs (Fig. 10d).⁴⁷ These SnO₂ NS-TiO₂ electrodes are highly transparent and show a “visible-blind” characteristic (Fig. 10e). Under UV irradiation (365 nm, 40 mW cm⁻²), the UV-PD shows a J_sc of 709.5 μA cm⁻² with an R_λ of 17.73 mA cm⁻², an S of 440.5 for J_sc signals, and a fast photoresponse (0.02 s for t_r and 0.004 s for t_d). These results indicate that this ultra-thin SnO₂ NS core can serve as the fast electron transport network for a fast photoresponse, but at the same time also cause high electron recombination at their grain boundaries that leads to a low R_λ. Then, Xie's group further assembled SnO₂ NSs into a type of SnO₂@TiO₂ core–shell hierarchical hollow nanospheres (HNSs) (Fig. 10f),³¹ in which the TiO₂ shell consists of a large number of tapered TNRs with lengths of over 10 nm (Fig. 10g). The assembled UV-PDs show the maximum value of R_λ at 330 nm, which is 12.4 times as high as that of SnO₂ HNS films, 1.1 times as high as that nc-TiO₂ films, and even 39.5% higher than the above B-SnO₂ NFs.²⁸ Under UV illumination (365 nm, 35 mW cm⁻²), the J_sc from SnO₂@TiO₂ HNS based PDs exceeds 1277 μA cm⁻² with an R_λ of 35.05 mA W⁻¹, an S of ∼4020, a t_r of 0.03 s, and a t_d of 0.01 s. This great enhancement is mainly due to multiple reflection and scattering of light in the hollow spheres and the decrease of electron recombination at grain boundaries, which may lead to high light trapping and carrier generation efficiencies.^99,100 Besides, Shen's group has prepared a type of freestanding SnO₂@TiO₂ core–shell heterostructure microtube using carbon cloth as templates (Fig. 10i and j).³² The TiO₂ shell consists of rutile TiO₂ NRs and shows a superior synergistic effect on electrochemical performances. Under UV illumination (365 nm, 1.55 mW cm⁻²), the assembled PEC UV-PDs show a high J_p of 100 μA cm⁻² with an R_λ of 64.5 mA W⁻¹, which is enhanced 22% compared to that of the pure SnO₂ one. This enhanced photoresponse performance is due to effective charge separation and good transport that was achieved in the SnO₂@TiO₂ cloth.

Fig. 10 (a) Diagram of self-powered UV-PDs based on nc-TiO₂ (left) and B-SnO₂ NF films (right), (b) SEM image of B-SnO₂-NF films, the inset is their TEM image, (c) energetics of operation of B-SnO₂-NF UV-PD. (d) Top view image of SnO₂ NS-TiO₂ films, the inset is their TEM image, and (e) spectral response of the self-powered UV-PDs, the inset is the digital photograph of transparent SnO₂ NS-TiO₂ films. (f) Structure scheme of UV-PDs based on SnO₂@TiO₂ HNSs, (g) SEM images of SnO₂@TiO₂ hierarchical HNSs, the insets show the TEM image. (h) Detailed carrier transfer process of the PEC cell under illumination, (i) SEM images of SnO₂ NP-TiO₂ NR composites, (j) cross-sectional view SEM image of the SnO₂@TiO₂ microtube. (a–c) and (h–j) Reproduced from ref. 28 and 32, respectively. Copyright © 2012 and 2014 Wiley-VCH. (d and e) and (f and g) Reproduced from ref. 47 and 31, respectively. Copyright © 2014 Elsevier Ltd.

3.4.3 Enhancing the charge transport efficiency in photoanodes. It has been recognized that diffusion is the major mechanism for electron transport in the oxide films, while multiple trapping/de-trapping at the grain boundaries usually results in a decreased electron transport rate in the oxide films.^62,84 Since the electrons in the photoanode would be easily recombined with oxidized ions (such as I₃⁻) in the electrolyte, the transit time of electrons within the oxide films should be much shorter than the time needed for their recombination, i.e., lifetime. Thus, to enhance the charge transport efficiency is another key factor for increasing the PEC performances of a PD.
3.4.3.1 NP/1D nanostructure hybrids. Compared to NPs, 1D NWs or NRs usually possess the advantages of a fast electron transport rate and excellent light scattering. However, at the same time, these two nanostructures suffer from their low SSA which usually results in a low light absorption efficiency and hence leads to low J_sc and R_λ.^101,102 Zhang et al.⁵² assembled anodic TiO₂ NTA/NP composites into photoanodes. Compared to the bare TNTAs, the TNTA/NP composites exhibit a visible-blind characteristic but with a slight red shift in the absorption edge. Moreover, this type of NTA/NP photoanode shows a fast photoresponse to UV light (200–400 nm), and the J_sc is increased by 29% compared to that of the bare TNTAs. These improvements in their PEC performances could be mainly due to the higher separation efficiency of the photogenerated electron–hole pairs and effective charge transport between the TiO₂ NTs and NPs.
3.4.3.2 Reduced NRAs. Surficial reduction can greatly increase the donor density of oxide nanostructures, which can benefit the photogenerated carrier transport. Yao et al.¹⁰³ prepared a type of hydrogenated ZnO hierarchical NRA (H-ZnO HNRA) by a two-step electrochemical process. After hydrogenation annealing, the color of the white ZnO HNRA film changes to black H-ZnO HNRAs. The carrier density of H-ZnO HNRAs is enhanced over 14 fold compared to that of the pristine ones, and H-ZnO HNRAs show a broad UV photoresponse from 300 to 370 nm with an IPCE of more than 90%. Similar results can also be seen in the study by Xu et al.⁶⁰

Moreover, it is suggested that the existence of Ti³⁺ defects (oxygen vacancies) in bulky TiO₂ nanostructures can enhance their electrical conductivity.¹⁰⁴ In this case, Ti³⁺-induced mid-gap states, as shallow donor sites, could contribute to the charge density enhancement and narrowed band gap and could lead to further improvement in PEC efficiency.¹⁰⁵ Zhang et al.¹⁰⁶ reported the preparation of reduced TNTs (r-TNTs) using electrochemical doping via cathodic reduction. The r-TNTs show a visual change of color from grey to dark blue, and their charge transfer resistance decreases from 1246 to 90.79 Ω cm⁻². And the J_p of r-TNTs increases by 33.3% compared to that of the pristine TNTs under UV irradiation (375 nm, 300 mW cm⁻²).

3.4.3.3 Element doping. Doping suitable elements into TiO₂ is found to be important for the reduction of the photogenerated hole–electron recombination rate.¹⁰⁷ Wang et al. fabricated nitrogen-doped TiO₂ mesosponge (N-TMSE and N-TMSW) by transformation of TNTAs using a solvothermal method (Fig. 11a).⁵⁷ The color of TNTAs changes from the original grey to sky-blue after doping with the N element, which corresponds to higher light absorption in the UVA region. Under UV irradiation (254 nm, 500 W) in a 0.1 M Na₂SO₄ electrolyte, the saturated J_p values of N-TMSE and N-TMSW are respectively ∼427 and 370 μA cm⁻², which are much higher than those of the pristine TNTAs (318 μA cm⁻²). The improved J_p could be mainly because N doping facilitates the separation and transfer of photogenerated electron/hole pairs. Moreover, Xu et al.⁶⁰ doped the Sn element into TiO₂ NWs by a one-pot hydrothermal synthesis. The obtained Sn-doped TiO₂ (Sn/TiO₂) NWs with Sn/Ti ratios of 1–2% are single crystalline with a rutile structure. The Sn/TiO₂-12 h NW photoanode shows a high J_p of 2.0 mA cm⁻², which is enhanced 100% compared to the best pristine TiO₂ NW photoanodes. This increase in photocurrent density is mainly ascribed to the enhancement of photoactivity in the UV region and the significant improvement in density of n-type carriers by Sn doping.

Fig. 11 (a) Transformation processes of NTs into mesosponge of TiO₂ by solvothermal treatment in an ethanol solution of ammonia, and their corresponding optical and SEM images (top views). (b) Schematic illustration of the preparation of TiO₂:(W, C) NWs by ex situ doping method; the cross-sectional SEM image is shown in the bottom, the inset shows TEM and EDX mapping images. (a) Reproduced from ref. 57. Copyright © 2012 Elsevier Ltd. (b) Reproduced from Cho et al.¹⁰⁸ Copyright © 2013 Nature Publishing Group.

Although a suitable mono-doping (single element doping) can enhance the PEC performance of TiO₂ to some extent, excessive mono-doping will increase carrier recombination centers and reversely decreases the overall PEC performance. Then, a new type of donor–acceptor co-doping concept was proposed to improve the PEC performances of TiO₂ photoanodes.¹⁰⁹ This type of co-doping of TiO₂ with donor–acceptor pairs can not only enhance the charge transport, but also reduce charged defects and their associated recombination, and accordingly improve their PEC performances. Dong's group reported a type of bimetallic Au–Pt NP/NH₂-group functionalized TiO₂ hybrid nanoparticles (denoted as f-TiO₂–Au/Pt NPs) via a simple self-assembly approach.¹¹⁰ These f-TiO₂–Au/Pt NPs consist of evenly dispersed Au/Pt NPs (6–7 nm) on f-TiO₂ colloid spheres with a nanoporous surface. Under UV illumination (365 nm), the f-TiO₂–Au/Pt NPs show a stable I_p of 0.30 ± 0.02 μA, which is about 4-fold that of the f-TiO₂ NPs (0.08 ± 0.03 μA). Moreover, Cho et al. reported a novel ex situ method to co-dope TiO₂ with W and C elements by sequentially annealing W-precursor-coated TiO₂ NWs in carbon monoxide gas (Fig. 11b).¹⁰⁸ The integrated UV light absorption (plus scattering) values of (W, C) co-doped TiO₂ NWs is enhanced only 7.56% compared to that of the undoped ones, while the IPCE value of the co-doped TiO₂:(W, C) NWs is ∼80% at 380 nm, which is enhanced 150% compared to that of the higher undoped TiO₂ (∼32% at 395 nm). And the saturation J_p of the co-doped TiO₂:(W, C) NWs is 2.4 times as high as that of the undoped TiO₂ NWs. These enhancements suggest that co-doping with donor–acceptor pairs can greatly enhance charge separation and transport efficiency. Interestingly, the co-doped TiO₂:(W, C) NWs also show very fast photoresponse to UV light.

3.4.3.4 Optics–electrics highways. Recently, conductive nanostructure modified technology has been employed as a promising way to boost the performances of PEC cells via optics–electrics highways.^111–114 It has been applied to significantly accelerate the carriers’ transport and separation, and hence considered to be one of the most attractive technologies to improve the η. Dong et al.¹¹² first proposed a concept of “optics–electrics highways” using a type of bristled silver NW@TiO₂ core–shell nanostructures (Fig. 12a). It is demonstrated that these constructed composites significantly accelerate electron transport as well as reduce recombination, and then greatly improve the charge collection efficiency of the cells, and accordingly enhance the IPCE by 43%. However, this type of metal nanostructure can often generate localized surface plasmon resonance, which can not only enhance light harvesting in the UV region, but also greatly increase the photoresponse in the visible light region.

Fig. 12 (a) Schematic diagram of photoelectrons transport in photoanodes of (I) pure and (II) Ag NW-modified TiO₂. (b) Schematic illustrations of electron transport across SWCNT-doped TiO₂ films, and (c) SEM images of SWCNT/TiO₂ films. (d) Schematic diagram of the graphene/ZnO HSN composite photoanode and the possible electron transport path within the structure; and (e) top-view SEM images of graphene/ZnO HSN composites (∼1.2 wt% graphene loading), the inset is the magnified image of an individual ZnO HSN. (f) Schematic illustrations of electron transport in pristine TiO₂ (upper) and r-GO–TiO₂ (lower) nanocomposites, (g) photographs of TiO₂ films with different r-GO loadings, and (h) SEM image of r-GO–TiO₂ nanocomposites. (a) Reproduced from ref. 112. Copyright © 2014, Elsevier. (b and c), (d and e) and (g and h) Reproduced from ref. 58, 115 and 51, respectively. Copyright © 2007, 2013 and 2010 American Chemical Society.

Carbon nanostructures, such as single-walled carbon nanotubes (SWCNTs), were also used as conducting scaffolds in TiO₂-based PEC cells, in which TiO₂ nanoparticles were dispersed on SWCNTs (Fig. 12b and c).⁵⁸ Compared to the undoped TiO₂ cells, the SWCNT/TiO₂-based cells show a great increase (over 200%) in J_p and/or η values, indicating the beneficial role of the SWCNT as a conducting scaffold to facilitate charge transport in TiO₂ films. Meanwhile, multi-walled carbon nanotubes (MWCNTs) were also incorporated into rutile TiO₂ NWAs.⁵⁵ It is found that on increasing the MWCNT content, the absorption edges of the TiO₂@MWCNT NWA films are slightly blue-shifted with wavelength of ∼10 nm. And the maximum η was observed from the TiO₂@MWCNT-0.1 NWA films, and it is enhanced 194% compared to that of the undoped ones. Further EIS results suggest that the small amount of MWCNT would promote the photogenerated electron transfer, and thus enhance the PEC performances of the devices. However, like mono-element doping, excessive MWCNT contents can also increase the trap state formation, and thus photogenerated electrons would spend more time to reach the FTO substrate.

Graphene is another type of carbon nanostructures that can also be used to enhance electron transport properties of oxide photoanodes. Compared to CNTs, reduced graphene oxide (r-GO) can provide a 2-D conductive support path for charge transport and collection at the electrode surface (Fig. 12f). In view of this point, Kamat's group synthesized a type of r-GO–TiO₂ nanocomposite via a solution-based method involving photocatalytic reduction of GO.⁵¹ The color of TiO₂ thin films gradually changes from white to black with increasing the addition of r-GO (Fig. 12g and h). The r-GO–TiO₂ composites exhibit the maximum η value of 13.9% at ∼350 nm, which is enhanced 87.8% compared to that of the pure TiO₂ films (7.4%). Upon UV irradiation (320–400 nm, 100 mW cm⁻²) in the 1 M KOH electrolyte, the r-GO–TiO₂ films (with an r-GO of 0.05 mg cm⁻²) output a J_p of ∼38 μA cm⁻², which is enhanced 90% compared to that of the TiO₂ films (∼20 μA cm⁻²). These observed enhancements in the photocurrent and IPCE represent improved charge transport from r-GO–TiO₂ nanocomposites to the collecting electrode surface. Similar results can also be seen in the study by Xu et al.,¹¹⁵ in which graphene was incorporated into ZnO nanoparticle photoanodes (Fig. 12d). After incorporating 1.2 wt% graphene into the 3 μm thick ZnO hierarchical structured nanoparticle (HSN) photoanode (Fig. 12e), the assembled devices exhibit a high J_sc of 10.89 mA cm⁻² and an η of 3.19%, which are increased by 43.48% and 38.09%, respectively, compared to those of the devices without graphene. Moreover, it was also found that the incorporation of graphene into electrodes could not only effectively decrease the internal resistance within the photoanodes, but also markedly prolong electron lifetime and effective diffusion length. This allows the utilization of thicker photoanodes that could afford enhanced SSA for light harvesting, and an impressively high η of 5.86% was achieved for the 9 μm thick ZnO HSN photoanode. However, as in the case of MWCNTs, excessive r-GO loadings will adversely affect the electron transport in the composite films. And the incorporated graphene can cause more oxygen-related defects in the ZnO HSN photoanode, which often shows an enhanced visible light absorption.

3.5 Recent multi-functional PEC UV-PDs

3.5.1 Wearable ZnO NRA/metal wire microcells. Recently, Yu's group reported a prototype of double-sided, transparent, flexible, ITO-free PEC cells,¹¹⁶ in which hierarchical ZnO NWAs (200 nm in diameter, 2 μm in length, and 5000 mm⁻² in density)/metal microwire (stainless steel, Au, Ag, and Cu) serve as the working electrode, and a Pt wire serves as the counter electrode (Fig. 13a). This device shows good transparency and can increase the light harvesting efficiency via two-sided illumination (Fig. 13b). Moreover, the double-wire cells remain stable for a long period of time and can be reversibly bent under large angles, even to 107°, but without any loss of performance. However, the untreated metal wires can usually be electrochemically corroded by the electrolyte ions, leading to a high J_d of 2.03 mA cm⁻². Thus, an oxidation treatment of the metal wires is developed effectively to avoid the electrochemical corrosion in the experiment and to fasten the light response, which is crucial for real applications of double-metal-wire PEC devices. This double-wire-PET, planar solar-cell configuration can be used in various flexible PEC devices, such as PEC UV-PD, and can be readily realized for large-area-weave roll-to-roll processes.

Fig. 13 (a) Schematic prototype of double-sided transparent PEC devices. The left inset shows hierarchical ZnO-NWA/Fe microwires, and the right one shows the high flexibility of the device. (b) J–V curves of double-sided transparent cells under illumination from the top, bottom, and both sides. (c) General working principles of a self-powered, visual and self-recovered PD, the inset photographs show the color changes of the PB/ITO ON/OFF UV illumination. (a and b) Reproduced from ref. 116. Copyright © 2012 Wiley-VCH. (c) Reproduced from ref. 117. Copyright © The Royal Society of Chemistry 2014.

3.5.2 Visual UV-PDs. More recently, Dong's group reported a novel self-powered visual UV-PD based on electrochromic display from Prussian blue (PB).¹¹⁷ This UV-PD is constructed by Pt-modified TNT and PB-modified ITO electrodes. Fig. 13c shows the general working principle of this visual UV-PD. Upon UV illumination, a number of electron–hole pairs are generated in TNTs. And the holes will oxidize water to oxygen, while the electrons will flow through an external circuit to the cathode, where PB will be reduced to Prussian white (PW) with color changing from blue to transparent. Thus we can judge the existence of UV light through the color change of PB. Pt/TNT could also be used as the cathode in the dark. When connected with PW-modified ITO, the PW can be oxidized to PB whereas the electrons flow to the Pt/TNT cathode, which catalyzes oxygen reduction. Based on the discussions above, a self-powered, visual and self-recovered PD is achieved. This self-powered, visual and self-recovered UV-PD makes it possible to judge the existence of UV light easily by the naked eye. More interestingly, it is also possible to fabricate a type of UV test cell based on this principle, similar to pH test paper.

3.5.3 Multi-functional UV-PDs with bi-polar electrodes. Intelligent devices are attracting more and more attention in recent years. Generally, these so-called “intelligent devices” are the equipment that is integrated with multiple functional components. On the other hand, introducing a bi-polar electrode (BPE) into those multiple functional optoelectronic platforms will make them light-weight, low-cost, small-sized and portable. A BPE in optoelectronic devices can act as both the anode and the cathode when two different cells are integrated in series. Chen et al.¹¹⁸ reported a type of optoelectronic three-electrode system, in which a SC and a PEC cell were integrated and separated by sharing a common Pt electrode (Fig. 14a). This is the so-called “photocapacitor”.¹¹⁹ The photovoltage can rapidly increase and reach 90% of the maximum value (0.69 V) by illumination for 28 s. Moreover, Peng's group replaces the Pt electrode with aligned CNT materials (Fig. 14b),¹²⁰ and found that the photovoltage can reach 90% of the maximum value (0.72 V) by illumination for less than 5 s, indicating more efficient electron transport. Furthermore, Peng's group integrated CNT fibers and a Ti wire modified with aligned TiO₂ nanotubes on the surface into an optoelectronic device.¹²¹ As shown in Fig. 14c, the Ti wire is used as BPE for two cells and aligned CNT fibers were finally twisted with the modified Ti wire to produce the desired device. Upon exposure to the light, the voltage can be charged to 0.6 V in a few seconds, and the overall photoelectric conversion/storage efficiency is 1.5%. It is noted that this novel wire structure can also enable unique and promising applications in wearable and portable PEC UV-PDs.

Fig. 14 (a) Configuration of the optoelectronic device containing a DSSC and a polymer-based SC using a Pt wire as BPE. (b) Schematic illustration of a PEC cell/SC optoelectronic device using aligned MWCNT films as BPEs. (c) Schematic illustration of the wire-shaped device integrated with a PEC cell/SC optoelectronic device using a Ti wire as BPE. The upper insets show SEM images of a Ti wire coated with aligned TNTAs, and the lower insets show SEM images of a CNT fiber. (d) schematic illustration and work mechanism of a PEC cell/SC optoelectronic device using TNTAs as BPEs. (e) Fabrication process and structure of the bi-functional UV-PD. (a) Reproduced from ref. 118. Copyright © 2012 Elsevier Ltd. (b) Reproduced from ref. 120. Copyright © 2013 The Royal Society of Chemistry. (c and d) Reproduced from ref. 121 and 122, respectively. Copyright © 2012 and 2014 Wiley-VCH. (e) Reproduced from ref. 123. Copyright © 2015 Elsevier Ltd.

More recently, Shen's group employed a BPE of TNAs into an optoelectronic integration.¹²² As shown in Fig. 14d, 1D anodic titanium oxide (ATO) NTAs were used as the photoanode materials for PEC cells and hydrogenated ATO (ATO-H) NTA as the cathode materials for electrochemical SCs. On visible light illumination, a pulsed anodic photocurrent of ∼2.6 mA cm⁻² occurred promptly and then gradually decayed to 0.025 mA cm⁻² in 1 s. Consequently, a maximum, stable and repeatable charge voltage of 0.61 V is achieved on SC after being photo-charged for 1 s. These results indicate that this type of optoelectronic device can also show a rapid response to light, even to UV light. In view of this, Zhang's group tried to integrate a UV-PD with a supercapacitor (SC) on one vertical-aligned silicon NW (SiNW) array electrode,¹²³ in which carbon coated SiNWs were used as the electrode for the SC and the pristine SiNWs were used as the electrode for UV-PDs (Fig. 14e). The assembled SC shows a capacitance density of 13.6 F cm⁻³, which can be fully charged in 80–100 s and then support the UV-PDs for 10 h. Upon UV illumination (365 nm), the photoresponse can be observed from the external circuit potential increase (0.03–0.12 V) between the two electrodes of the SC. Although the reported UV detecting performances are not comparable to those from the above PEC UV-PDs, this type of self-powered multi-functional system that integrates power conversion/storage with UV light detection will open a new insight into the design of multi-functional PEC UV-PDs.

4 Summary and outlook

In this review, we have discussed the recent advances in self-powered PEC UV-PDs. Various nanostructures, including nc-films, NWs/NRs, NTs, NSs, NRAs, NTAs, nano-forest, etc., have been synthesized for photoanodes of PEC UV-PDs benefiting from their high UV light absorption, fast carrier separation/transport efficiencies, high spectral responsivity, high photosensitivity, and fast photoresponse. On the other hand, regarding the electrolyte, I⁻/I₃⁻, water, Na₂SO₄ aqueous solution, etc., have been adopted, and the recently reported multi-functional PEC optoelectronic devices have been simply presented. However, current research work is still at the initial stage focusing on the design of photoanode structures. And these reviewed research studies are continuously developed according to the traditional optoelectronic devices (such as DSSC), involving device structure (sandwich structure), light harvesting, separation and transport mechanism of carriers, electrolytes, and counter electrodes. In nature, the mechanism for PEC UV-PDs is highly different from those for other optoelectronic devices. However, several issues, such as the light absorption mechanism under UV illumination, interfacial charge dynamics, and recombination mechanism of carriers, have not been investigated in detail. Thus, future endeavors are needed to achieve further improvement in the material and device designs, and photo detecting performances for practical applications. The related outlook points are listed below:

4.1 Photoanode structures

Photoanode materials and structures are among the most important components of the whole PD, and greatly determine the light harvesting of a PEC cell. We can optimize the structures of photoanodes for PEC UV-PDs from the following aspects:

(i) UV light absorption. Usually, one optoelectronic device requires a photoanode film with sufficient thickness to absorb light, which is dependent on the penetration depth of UV light. Excessive thickness of the film will result in the increase of diffusion length of photogenerated electrons. Thus, the optimized thickness of photoanode films is an important issue. On the other hand, light scattering can increase the light propagation length, which is an effective approach to improve the light harvesting efficiency. It is noted here that light scattering can also enhance the visible light absorption, which is unfavorable to the UV selectivity of a PD. Thus, we should modify the size of nanostructures according to the wavelength of the target UV light, in order to achieve an ideal scattering effect.

(ii) Separation of photogenerated carriers. Under excitation by UV light, the electrons in the semiconductor will transit from the VB to CB, leaving a hole in the VB, which is subsequently reduced by ions in the electrolyte. Thus, first, it is necessary to guarantee an adequate semiconductor/electrolyte interface area, i.e., for the efficient separation of carriers, thus requiring a large SSA for the nanostructures in photoanodes. Second, apart from a high SSA, photoanodes should also possess suitable pore size and porosity, which will benefit the diffusion of the electrolyte to the whole film. It is also worth noting here that the penetration depth of UV light in the semiconductor is about 2 μm, which is much less than that of visible light. Then the film thickness required by PEC UV-PD (1–2 μm) is far less than that of a DSSC (10 μm for P25). This is to say, compared to a DSSC, the PEC UV-PDs need a smaller pore size in a photoanode film to meet effective diffusion of the electrolyte. Third, different crystal planes of semiconductors and size distribution of nanostructures will also affect this PEC interfacial charge dynamics.⁷⁰

(iii) Transport of electrons. Photogenerated carriers will be separated at the semiconductor/electrolyte interface, and the electrons that are injected into the CB of the semiconductor will diffuse through the photoanode to the collector. Then, the transport rate is the key factor to determine the efficiency of the electron collection and the response speed of a PD. It is demonstrated that 1D nanostructures with high crystallinity and high aspect ratio can provide a direct transmission channel for electrons, which can greatly reduce the electron recombination at the grain boundary during the transport.^78,87,88 And it is reported that the PEC UV-PDs assembled using TiO₂ NWAs with an aspect ratio of 150 show greatly improved performances in photo detection. However, so far, few reports can be seen on the PEC UV-PDs based on high-electron-mobility and high-aspect-ratio semiconductor nanostructure such as TiO₂ or SnO₂. Besides, it is widely reported that conductive nanomaterials, such as CNTs, graphene, or metal nanostructures, can work as optics–electrics highways to boost the performance of PEC cells. Then, we believe that these strategies can also be used to enhance the performances of PEC UV-PDs.

(iv) Reducing the recombination of photogenerated electrons. In a PEC UV-PD, the surface of the semiconductor nanomaterial can come into direct contact with the electrolyte, which usually makes the electrons that transit to the CB easily recombine with the electrolyte ions. The optics–electrics highways mentioned above can not only facilitate the transport of electrons, but also reduce the recombination of electrons. On the other hand, one more effective way is to make the semiconductor heterojunction (such as p–n junctions and Schottky junctions) and to form a type-II band structure, which can block electron recombination by the barrier. According to Fig. 2, many semiconductor heterojunctions, such as ZnO/ZnS, ZnO/NiO, SrTiO₃/TiO₂, etc., have a high potential for application in PEC UV-PDs.

In brief, for an ideal photoanode, it is important to balance these four aspects mentioned above into an integrated system, not just to prepare high-SSA nanostructures with amazing morphologies and structures.

4.2 Device configuration

At present, most of the reported PEC UV-PDs possess a sandwich device configuration similar to a DSSC, as shown in Fig. 15a. In this configuration, TCO glass (such as FTO and ITO) is used as a transparent conductive substrate, which supports the photoanode materials. Once the devices start working, the target light should enter from the side of the photoanode through the TCO glass, and then excite the semiconductor materials. However, not all UV light can transmit through TCO glass easily. According to our tests, as shown in Fig. 15b (black solid line), UV light with wavelength less than 300 nm cannot transmit through TCO glass, and the one with wavelength larger than 330 nm just remains only 38% in intensity after transmission. These facts will inevitably lead to a serious issue, i.e., most of the reported PEC UV-PDs exhibit a peak response with a full width at half maximum of about 40 nm in the range of 300–400 nm and a very low peak value in spectral responsivity at 330 nm, as shown in Fig. 15b (red solid line). If the filtering effect on the UV light source from the substrate material can be removed, the performances of PEC UV-PDs will be greatly improved. More interestingly, the response spectrum will broaden, showing an ideal response curve similar to the red dotted line shown in Fig. 15b. Thus, it is very urgent to solve this “UV-filtering” problem. Some potential approaches are summarized below:

(i) To design a novel UV transparent conductive substrate to replace the traditional TCO glass. For instance, Bao et al. have prepared a type of transparent conductive substrate based on a metal NW mesh by in situ etching (Fig. 15c),¹²⁴ which shows ∼90% transmittance within a ultra-wide wave range of 200–2000 nm and a very low resistance (2.2 Ω sq⁻¹). Moreover, Cui's group prepared a metal NW mesh via the sputtering or evaporation method using electrospun NWs as templates (Fig. 15d), which shows a sheet resistance of ∼2 Ω sq⁻¹ at 90% transmission.¹²⁵ However, when applied to PEC UV-PDs, these new transparent conductive substrates will meet some troubles, i.e., how to well contact the semiconductor photoanode films with the conductive layer, and how to conduct a low-temperature sintering of semiconductor films because the melting point of metal NWs is usually very low.

(ii) Another effective approach is to design a type of planar device structure, i.e., put the semiconductor photoanode and counter electrode on the same substrate, which are separated via some special method such as interdigital electrode or fiber devices. And then the electrolyte will be perfused between them, and the light source can enter from the surface UV transparent window without any loss.

Fig. 15 (a) Schematic of UV light entering a PEC PD using FTO as electrode connectors. (b) Typical transmittance spectrum of FTO glass (black solid line), experimental spectral response of PEC UV-PDs (red solid line), and the ideal spectral response of a PEC UV-PD (red dotted line). (c) (left) Schematic in situ fabrication of a metal NW network via the ion beam etching process using a polymer NF network as the shadow mask, and (right) optical transmission spectra of flexible metal networks. (d) Schematic of the polymer-NF templating process for fabricating nanotroughs. (c) Reproduced from ref. 124. Copyright © 2015 American Chemical Society. (d) Reproduced from ref. 125. Copyright © 2013, Nature Publishing Group.

4.3 Material/band-gap adjustable and deep-UV detectors

The newly-reported PEC UV-PDs are mainly based on the ZnO, TiO₂ and SnO₂, which have been widely used in PEC devices. However, there exist some drawbacks in those semiconductors, such as low electron mobility, poor stability and so on. Thus, it is of great importance to seek more alternative materials to promote the practical applications of PEC UV-PDs.

(i) In fact, compared with other PEC applications such as DSSC, photocatalysis, photolysis of water, and so on, PEC UV-PDs usually show a low requirement in the crystal structure, absorptivity, and energy band position of the semiconductor nanomaterials. And many wide-band-gap semiconductor oxides, nitrides and sulfides can be selected for their photoanode materials. For instance, GaN possesses a wide band gap of 3.4 eV, a high electron mobility of 900 cm² V⁻¹ s⁻¹, and a high UV radiation stability, which will make it show high potential for applications in PEC UV-PDs.¹²⁶ Besides, many ternary inorganic compounds with a perovskite structure, such as SrTiO₃ (3.15 eV), CaTiO₃ (3.6 eV), and BaTiO₃ (3.3 eV), shows excellent behavior in the field of DSSC, and it is believed that these materials might also play a good role in PEC UV-PDs.

(ii) For practical applications, it is often required that UV-PDs possess a tunable detecting cutoff wavelength, which can be realized by adjusting the semiconductors’ band gap. And this idea has been widely reported and realized in the photoconductors. For instance, some ternary alloy semiconductors such as ZnMgO,¹²⁷ AlGaN,¹²⁸ and so on show a good ability for band-gap modulation, and have successfully covered the wave range from 200 to 365 nm via the component ratios. However, few similar studies have focused on PEC UV-PDs up to now. And it is worth noting that the energy band structure of the semiconductor used in PEC UV-PDs should well match with the redox potential of the chosen electrolyte. In addition, the wetting property between the semiconductor and the used electrolyte is another important issue that can greatly influence the electrolyte ion diffusion, and thus determine the charge collecting efficiency.

(iii) Deep-UV detectors have many important applications in military, space, and other fields. At present, photoconductors/Schottky-type deep-UV detectors based on ultra-wide-band-gap semiconductors often exhibit a low performance. Thus, it is instructive and interesting for researchers to explore PEC deep-UV detectors.

4.4 Stabilities of PEC UV-PDs

Running stability is one important indicator of the practical applications of a device. Also the stability of PEC UV-PDs is an essential issue that we should pay attention to. However, few papers have reported a systematical study of the running stability of PEC UV-PDs. As is well known, compared to the solid UV-PDs based on photoconductance (p–n or Schottky junctions), PEC ones usually show a worse stability, which is mainly due to the volatilization of the electrolyte (depending on the packaging process), the corrosion and surface passivation of the semiconductor, the gas generation during the work process, the decreased catalytic ability of the counter electrode (such as Pt), and so on. Thus, in view of the above points, some future work should be conducted to improve the stability of PEC UV-PDs. For instance, Chen's group attempted a nanostructured quasi-solid-state UV-PD fabricated using a liquid LC-embedded I⁻/I₃⁻ electrolyte, and obtained improved light harvesting and stability.³³ Further more research on improving the stabilities of UV-PDs is still on the way.

4.5 Multi-functional PEC UV-PDs

Recent reports on flexible, visual, and multi-functional UV-PDs with BPEs have just given an initial attempt or proposed a prototype, and further development of multi-functional detectors will significantly broaden the application fields of PEC UV-PDs. On the other hand, compared to the PDs based on the metal–semiconductor–metal structure of bulky semiconductor materials, the PEC UV-PD is based on nanomaterials and an electrolyte system, thereby showing a great potential in flexible applications. As demonstrated, this new type of UV-PD can work in the electrolyte of “non-toxic, environmentally-friendly and cheap” water, which makes the devices safer to a large extent. Moreover, it is speculative that interval replacement and filling of the water electrolyte in the devices can easily realize their circular utilization. Therefore, in brief, those advantages of UV-PDs, including fast reponsibility, high reliability and environment protection properties, low-cost and self-powered system, and possible integration with other functional devices, will make them have great potential in various civilian application fields, such as medical and health protection, and wearable devices in the future.

Acknowledgements

The authors acknowledge the financial support from the National Natural Science Foundation of China (Grant No. 51202100 and 61404065). The authors also thank the great contribution from Drs Xiaodong Li and Caitian Gao who have graduated from our group.

References

1. E. Monroy, F. Omnès and F. Calle, Semicond. Sci. Technol., 2003, 18, R33.
2. G. Shen and D. Chen, Recent Pat. Nanotechnol., 2010, 4, 20–31.
3. C. L. Hsu and S. J. Chang, Small, 2014, 10, 4562–4585.
4. Z. Alaie, S. Mohammad Nejad and M. H. Yousefi, Mater. Sci. Semicond. Process., 2015, 29, 16–55.
5. G. Konstantatos and E. H. Sargent, Nat. Nanotechnol., 2010, 5, 391–400.
6. J. Zou, Q. Zhang, K. Huang and N. Marzari, J. Phys. Chem. C, 2010, 114, 10725–10729.
7. C. Soci, A. Zhang, B. Xiang, S. A. Dayeh, D. P. R. Aplin, J. Park, X. Y. Bao, Y. H. Lo and D. Wang, Nano Lett., 2007, 7, 1003–1009.
8. Y. Han, G. Wu, M. Wang and H. Chen, Polymer, 2010, 51, 3736–3743.
9. W. Tian, C. Zhang, T. Zhai, S. L. Li, X. Wang, M. Liao, K. Tsukagoshi, D. Golberg and Y. Bando, Chem. Commun., 2013, 49, 3739–3741.
10. X. Wang, Z. Xie, H. Huang, Z. Liu, D. Chen and G. Shen, J. Mater. Chem., 2012, 22, 6845–6850.
11. M. Gratzel, Nature, 2001, 414, 338–344.
12. C. Basant, P. Leela Srinivas, K. Salaru Baba and C. N. R. Rao, Jpn. J. Appl. Phys., 2011, 50, 100206.
13. C. W. Litton, D. C. Reynolds, T. C. Collins, J. Zhong and Y. Lu, in Zinc Oxide Materials for Electronic and Optoelectronic Device Applications, ed. P. Capper, S. Kasap and A. Willoughby, Wiley, 2011, vol. 11, p. 363.
14. S. Sawyer and D. Shao, in Handbook of Nanomaterials Properties, ed. B. Bhushan, D. Luo, S. R. Schricker, W. Sigmund and S. Zauscher, Springer, Berlin, Heidelberg, 2014, pp. 1177–1198.
15. Y. Li, F. D. Valle, M. Simonnet, I. Yamada and J. J. Delaunay, Appl. Phys. Lett., 2009, 94, 023110.
16. S. Liu, J. Ye, Y. Cao, Q. Shen, Z. Liu, L. Qi and X. Guo, Small, 2009, 5, 2371–2376.
17. F. Fang, J. Futter, A. Markwitz and J. Kennedy, Nanotechnology, 2009, 20, 245502.
18. W. W. J. Liu, S. Bai and Y. Qin, ACS Appl. Mater. Interfaces, 2011, 3, 4197–4200.
19. B. Liu, Z. Wang, Y. Dong, Y. Zhu, Y. Gong, S. Ran, Z. Liu, J. Xu, Z. Xie and D. Chen, et al. , J. Mater. Chem., 2012, 22, 9379–9384.
20. Y. Q. Bie, Z. M. Liao, H. Z. Zhang, G. R. Li, Y. Ye, Y. B. Zhou, J. Xu, Z. X. Qin, L. Dai and D. P. Yu, Adv. Mater., 2011, 23, 649–653.
21. W. Jin, Y. Ye, L. Gan, B. Yu, P. Wu, Y. Dai, H. Meng, X. Guo and L. Dai, J. Mater. Chem., 2012, 22, 2863–2867.
22. X. Li, C. Gao, H. Duan, B. Lu, X. Pan and E. Xie, Nano Energy, 2012, 1, 640–645.
23. Y. Q. Bie, Z. M. Liao, H. Z. Zhang, G. R. Li, Y. Ye, Y. B. Zhou, J. Xu, Z. X. Qin, L. Dai and D. P. Yu, Adv. Mater., 2011, 23, 649–653.
24. S. M. Hatch, J. Briscoe and S. Dunn, Adv. Mater., 2013, 25, 867–871.
25. Q. Yang, Y. Liu, Z. Li, Z. Yang, X. Wang and Z. L. Wang, Angew. Chem., Int. Ed., 2012, 124, 6549–6552.
26. Z. Zhan, L. Zheng, Y. Pan, G. Sun and L. Li, J. Mater. Chem., 2012, 22, 2589–2595.
27. X. Hu, B. Heng, X. Chen, B. Wang, D. Sun, Y. Sun, W. Zhou and Y. Tang, J. Power Sources, 2012, 217, 120–127.
28. X. Li, C. Gao, H. Duan, B. Lu, Y. Wang, L. Chen, Z. Zhang, X. Pan and E. Xie, Small, 2013, 9, 2005–2011.
29. C. Gao, X. Li, Y. Wang, L. Chen, X. Pan, Z. Zhang and E. Xie, J. Power Sources, 2013, 239, 458–465.
30. Q. Li, L. Wei, Y. Xie, K. Zhang, L. Liu, D. Zhu, J. Jiao, Y. Chen, S. Yan and G. Liu, et al. , Nanoscale Res. Lett., 2013, 8, 1–7.
31. L. Chen, X. Li, Y. Wang, C. Gao, H. Zhang, B. Zhao, F. Teng, J. Zhou, Z. Zhang and X. Pan, et al. , J. Power Sources, 2014, 272, 886–894.
32. X. Hou, X. Wang, B. Liu, Q. Wang, Z. Wang, D. Chen and G. Shen, ChemElectroChem, 2014, 1, 108–115.
33. Y. Xie, L. Wei, Q. Li, Y. Chen, H. Liu, S. Yan, J. Jiao, G. Liu and L. Mei, Nanoscale, 2014, 6, 9116–9121.
34. Y. Xie, L. Wei, Q. Li, Y. Chen, S. Yan, J. Jiao, G. Liu and L. Mei, Nanotechnology, 2014, 25, 075202.
35. Z. Wang, S. Ran, B. Liu, D. Chen and G. Shen, Nanoscale, 2012, 4, 3350–3358.
36. Y. Xie, L. Wei, G. Wei, Q. Li, D. Wang, Y. Chen, S. Yan, G. Liu, L. Mei and J. Jiao, Nanoscale Res. Lett., 2013, 8, 188.
37. J. Xing, H. Wei, E. J. Guo and F. Yang, J. Phys. D: Appl. Phys., 2011, 44, 375104.
38. J.-K. Lee and M. Yang, Mater. Sci. Eng., B, 2011, 176, 1142–1160.
39. T. Majumder, J. J. L. Hmar, K. Debnath, N. Gogurla, J. N. Roy, S. K. Ray and S. P. Mondal, J. Appl. Phys., 2014, 116, 034311.
40. O. Game, U. Singh, T. Kumari, A. Banpurkar and S. Ogale, Nanoscale, 2013, 6, 503–513.
41. Y. Han, C. Fan, G. Wu, H. Z. Chen and M. Wang, J. Phys. Chem. C, 2011, 115, 13438–13445.
42. P. Guo, J. Jiang, S. Shen and L. Guo, Int. J. Hydrogen Energy, 2013, 38, 13097–13103.
43. W. J. Lee and M. H. Hon, Appl. Phys. Lett., 2011, 99, 251102.
44. C. Cao, C. Hu, X. Wang, S. Wang, Y. Tian and H. Zhang, Sens. Actuators, A, 2011, 156, 114–119.
45. Y. Wang, W. Han, B. Zhao, L. Chen, F. Teng, X. Li, C. Gao, J. Zhou and E. Xie, Sol. Energy Mater. Sol. Cells, 2015, 140, 376–381.
46. C. W. Kim, S. P. Suh, M. J. Choi, Y. S. Kang and Y. S. Kang, J. Mater. Chem. A, 2013, 1, 11820–11827.
47. C. Gao, X. Li, X. Zhu, L. Chen, Y. Wang, F. Teng, Z. Zhang, H. Duan and E. Xie, J. Alloys Compd., 2014, 616, 510–515.
48. J. Halme, P. Vahermaa, K. Miettunen and P. Lund, Adv. Mater., 2010, 22, E210–E234.
49. J. Wu, Z. Lan, J. Lin, M. Huang, Y. Huang, L. Fan and G. Luo, Chem. Rev., 2015, 115, 2136–2173.
50. Z. Sun, L. Xu, W. Guo, B. Xu, S. Liu and F. Li, J. Phys. Chem. C, 2010, 114, 5211–5216.
51. Y. H. Ng, I. V. Lightcap, K. Goodwin, M. Matsumura and P. V. Kamat, J. Phys. Chem. Lett., 2010, 1, 2222–2227.
52. Y. Zhang, X. Li, M. Feng, F. Zhou and J. Chen, Surf. Coat. Technol., 2010, 205, 2572–2577.
53. J. Lei, X. Li, W. Li, F. Sun, D. Lu and J. Yi, Int. J. Hydrogen Energy, 2011, 36, 8167–8172.
54. N. L. Tarwal, V. V. Shinde, A. S. Kamble, P. R. Jadhav, D. S. Patil, V. B. Patil and P. S. Patil, Appl. Surf. Sci., 2011, 257, 10789–10794.
55. M. Chang, L. Wu, X. Li and W. Xu, J. Mater. Sci. Technol., 2012, 28, 594–598.
56. Y. Z. Su, K. Xiao, Z. J. Liao, Y. H. Zhong, N. Li, Y. B. Chen and Z. Q. Liu, Int. J. Hydrogen Energy, 2013, 38, 15019–15026.
57. Q. Wang, X. Yang, X. Wang, M. Huang and J. Hou, Electrochim. Acta, 2012, 62, 158–162.
58. A. Kongkanand, R. Martínez Domínguez and P. V. Kamat, Nano Lett., 2007, 7, 676–680.
59. Y. Li, H. Yu, W. Song, G. Li, B. Yi and Z. Shao, Int. J. Hydrogen Energy, 2011, 36, 14374–14380.
60. M. Xu, P. Da, H. Wu, D. Zhao and G. Zheng, Nano Lett., 2012, 12, 1503–1508.
61. W. Chen, Y. Qiu and S. Yang, Phys. Chem. Chem. Phys., 2012, 14, 10872–10881.
62. P. Tiwana, P. Docampo, M. B. Johnston, H. J. Snaith and L. M. Herz, ACS Nano, 2011, 5, 5158–5166.
63. W. Dai, X. Pan, S. Chen, C. Chen, Z. Wen, H. Zhang and Z. Ye, J. Mater. Chem. C, 2014, 2, 4606–4614.
64. S. Yang, Y. Hou, J. Xing, B. Zhang, F. Tian, X. H. Yang and H. G. Yang, Chem. – Eur. J., 2013, 19, 9366–9370.
65. J. Zhao, W. Zhang, E. Xie, Z. Liu, J. Feng and Z. Liu, Mater. Sci. Eng., B, 2011, 176, 932–936.
66. B. Liu, X. Zhao, C. Terashima, A. Fujishima and K. Nakata, Phys. Chem. Chem. Phys., 2014, 16, 8751–8760.
67. T. Zhai, L. Li, X. Wang, X. Fang, Y. Bando and D. Golberg, Adv. Funct. Mater., 2010, 20, 4233–4248.
68. H. Xue, X. Kong, Z. Liu, C. Liu, J. Zhou, W. Chen, S. Ruan and Q. Xu, Appl. Phys. Lett., 2007, 90, 201118.
69. J. M. Macak, H. Tsuchiya, L. Taveira, S. Aldabergerova and P. Schmuki, Angew. Chem., Int. Ed., 2005, 44, 7463–7465.
70. I. S. Cho, Z. Chen, A. J. Forman, D. R. Kim, P. M. Rao, T. F. Jaramillo and X. Zheng, Nano Lett., 2011, 11, 4978–4984.
71. W. Q. Wu, B. X. Lei, H. S. Rao, Y. F. Xu, Y. F. Wang, C. Y. Su and D. B. Kuang, Sci. Rep., 2013, 3, 1352.
72. W. Q. Wu, H. S. Rao, H. L. Feng, X. D. Guo, C. Y. Su and D. B. Kuang, J. Power Sources, 2014, 260, 6–11.
73. J. Xu, Z. Chen, J. A. Zapien, C. S. Lee and W. Zhang, Adv. Mater., 2014, 26, 5337–5367.
74. T. Zhai, L. Li, Y. Ma, M. Liao, X. Wang, X. Fang, J. Yao, Y. Bando and D. Golberg, Chem. Soc. Rev., 2011, 40, 2986–3004.
75. Q. Zhang and G. Cao, Nano Today, 2011, 6, 91–109.
76. M. Law, L. E. Greene, J. C. Johnson, R. Saykally and P. Yang, Nat. Mater., 2005, 4, 455–459.
77. L. Li, T. Zhai, Y. Bando and D. Golberg, Nano Energy, 2012, 1, 91–106.
78. S. H. Ko, D. Lee, H. W. Kang, K. H. Nam, J. Y. Yeo, S. J. Hong, C. P. Grigoropoulos and H. J. Sung, Nano Lett., 2011, 11, 666–671.
79. M. J. Bierman and S. Jin, Energy Environ. Sci., 2009, 2, 1050.
80. X. Sun, Q. Li, J. Jiang and Y. Mao, Nanoscale, 2014, 6, 8769–8780.
81. C. Gao, X. Li, B. Lu, L. Chen, Y. Wang, F. Teng, J. Wang, Z. Zhang, X. Pan and E. Xie, Nanoscale, 2012, 4, 3475–3481.
82. X. Hou, B. Liu, X. Wang, Z. Wang, Q. Wang, D. Chen and G. Shen, Nanoscale, 2013, 5, 7831–7837.
83. J. C. Fu, J. L. Zhang, Y. Peng, C. H. Zhao, Y. M. He, Z. X. Zhang, X. J. Pan, N. J. Mellors and E. Q. Xie, Nanoscale, 2013, 5, 12551–12557.
84. N. K. Elumalai, R. Jose, P. S. Archana, V. Chellappan and S. Ramakrishna, J. Phys. Chem. C, 2012, 116, 22112–22120.
85. S. Mridha, M. Nandi, A. Bhaumik and D. Basak, Nanotechnology, 2008, 19, 275705.
86. B. Hu, Q. Tang, B. He, L. Lin and H. Chen, J. Power Sources, 2014, 267, 445–451.
87. L. Chen, X. Li, L. Qu, C. Gao, Y. Wang, F. Teng, Z. Zhang, X. Pan and E. Xie, J. Alloys Compd., 2014, 586, 766–772.
88. C. Xu, J. Wu, U. V. Desai and D. Gao, J. Am. Chem. Soc., 2011, 133, 8122–8125.
89. Z. Wang, H. Wang, B. Liu, W. Qiu, J. Zhang, S. Ran, H. Huang, J. Xu, H. Han and D. Chen, et al. , ACS Nano, 2011, 5, 8412–8419.
90. Y. Hu, X. Yan, X. Chen, Z. Bai, Z. Kang, F. Long and Y. Zhang, Appl. Surf. Sci., 2015, 339, 122–127.
91. R. S. Selinsky, Q. Ding, M. S. Faber, J. C. Wright and S. Jin, Chem. Soc. Rev., 2013, 42, 2963–2985.
92. S. Hernández, V. Cauda, D. Hidalgo, V. Farías Rivera, D. Manfredi, A. Chiodoni and F. C. Pirri, J. Alloys Compd., 2014, 615, S530–S537.
93. M. Wang, C. Huang, Y. Cao, Q. Yu, Z. Deng, Y. Liu, Z. Huang, J. Huang, Q. Huang and W. Guo, et al. , J. Phys. D: Appl. Phys., 2009, 42, 155104.
94. M. Wang, Y. Wang and J. Li, Chem. Commun., 2011, 47, 11246.
95. S. Panigrahi and D. Basak, Nanoscale, 2011, 3, 2336–2341.
96. S. H. Ahn, D. J. Kim, W. S. Chi and J. H. Kim, Adv. Mater., 2013, 25, 4893–4897.
97. U. V. Desai, C. Xu, J. Wu and D. Gao, J. Phys. Chem. C, 2013, 117, 3232–3239.
98. C. Gao, X. Li, X. Zhu, L. Chen, Z. Zhang, Y. Wang, Z. Zhang, H. Duan and E. Xie, J. Power Sources, 2014, 264, 15–21.
99. J. Qian, P. Liu, Y. Xiao, Y. Jiang, Y. Cao, X. Ai and H. Yang, Adv. Mater., 2009, 21, 3663–3667.
100. A. Thapa, J. Zai, H. Elbohy, P. Poudel, N. Adhikari, X. Qian and Q. Qiao, Nano Res., 2014, 7, 1154–1163.
101. W. P. Liao and J. J. Wu, J. Mater. Chem., 2011, 21, 9255.
102. Y. F. Wang, K. N. Li, W. Q. Wu, Y. F. Xu, H. Y. Chen, C. Y. Su and D. B. Kuang, RSC Adv., 2013, 3, 13804–13810.
103. C. Yao, B. Wei, H. Ma, H. li, L. Meng, X. Zhang and Q. Gong, J. Power Sources, 2013, 237, 295–299.
104. F. Zuo, L. Wang, T. Wu, Z. Zhang, D. Borchardt and P. Feng, J. Am. Chem. Soc., 2010, 132, 11856–11857.
105. X. Zhang, H. Tian, X. Wang, G. Xue, Z. Tian, J. Zhang, S. Yuan, T. Yu and Z. Zou, Mater. Lett., 2013, 100, 51–53.
106. Q. Zhang, L. Wang, J. Feng, H. Xu and W. Yan, Phys. Chem. Chem. Phys., 2014, 16, 23431–23439.
107. I. Justicia, P. Ordejón, G. Canto, J. L. Mozos, J. Fraxedas, G. A. Battiston, R. Gerbasi and A. Figueras, Adv. Mater., 2002, 14, 1399–1402.
108. I. S. Cho, C. H. Lee, Y. Feng, M. Logar, P. M. Rao, L. Cai, D. R. Kim, R. Sinclair and X. Zheng, Nat. Commun., 2013, 4, 1723.
109. Y. Gai, J. Li, S. S. Li, J. B. Xia and S. H. Wei, Phys. Rev. Lett., 2009, 102, 036402.
110. D. Wen, S. Guo, Y. Wang and S. Dong, Langmuir, 2010, 26, 11401–11406.
111. Y. C. Pu, G. Wang, K. D. Chang, Y. Ling, Y. K. Lin, B. C. Fitzmorris, C. M. Liu, X. Lu, Y. Tong and J. Z. Zhang, et al. , Nano Lett., 2013, 13, 3817–3823.
112. H. Dong, Z. Wu, F. Lu, Y. Gao, A. El-Shafei, B. Jiao, S. Ning and X. Hou, Nano Energy, 2014, 10, 181–191.
113. F. Teng, G. Zhang, Y. Wang, C. Gao, L. Chen, P. Zhang, Z. Zhang and E. Xie, Appl. Surf. Sci., 2014, 320, 703–709.
114. X. An, F. Teng, Z. Zhang, X. Pan, J. Zhou and E. Xie, Electron. Mater. Lett., 2014, 10, 95–99.
115. F. Xu, J. Chen, X. Wu, Y. Zhang, Y. Wang, J. Sun, H. Bi, W. Lei, Y. Ni and L. Sun, J. Phys. Chem. C, 2013, 117, 8619–8627.
116. W. Wang, Q. Zhao, H. Li, H. Wu, D. Zou and D. Yu, Adv. Funct. Mater., 2012, 22, 2775–2782.
117. L. Han, L. Bai and S. Dong, Chem. Commun., 2014, 50, 802–804.
118. H. W. Chen, C. Y. Hsu, J. G. Chen, K. M. Lee, C. C. Wang, K. C. Huang and K. C. Ho, J. Power Sources, 2010, 195, 6225–6231.
119. T. Miyasaka and T. N. Murakami, Appl. Phys. Lett., 2004, 85, 3932–3934.
120. Z. Yang, L. Li, Y. Luo, R. He, L. Qiu, H. Lin and H. Peng, J. Mater. Chem. A, 2013, 1, 954–958.
121. T. Chen, L. Qiu, Z. Yang, Z. Cai, J. Ren, H. Li, H. Lin, X. Sun and H. Peng, Angew. Chem., Int. Ed., 2012, 51, 11977–11980.
122. J. Xu, H. Wu, L. Lu, S.-F. Leung, D. Chen, X. Chen, Z. Fan, G. Shen and D. Li, Adv. Funct. Mater., 2014, 24, 1840–1846.
123. L. Sun, X. Wang, K. Zhang, J. Zou, Z. Yan, X. Hu and Q. Zhang, Nano Energy, 2015, 15, 445–452.
124. C. Bao, J. Yang, H. Gao, F. Li, Y. Yao, B. Yang, G. Fu, X. Zhou, T. Yu and Y. Qin, et al. , ACS Nano, 2015, 9, 2502–2509.
125. H. Wu, D. Kong, Z. Ruan, P.-C. Hsu, S. Wang, Z. Yu, T. J. Carney, L. Hu, S. Fan and Y. Cui, Nat. Nanotechnol., 2013, 8, 421–425.
126. M. Zhang, Y. Wang, F. Teng, L. Chen, J. Li, J. Zhou, X. Pan and E. Xie, Mater. Lett., 2016, 162, 117–120.
127. K. Yudai, K. Toshiyuki, N. Hiroyuki and F. Shizuo, Jpn. J. Appl. Phys., 2006, 45, L857.
128. A. Osinsky, S. Gangopadhyay, B. W. Lim, M. Z. Anwar, M. A. Khan, D. V. Kuksenkov and H. Temkin, Appl. Phys. Lett., 1998, 72, 742–744.

---

Footnote

† These authors contributed equally to this work.

---