Colloidal quantum dots (QDs) refer to the uniform small semiconductor particles, in which the charge carrier transportation is confined by three dimensions^[1]. As a result, of the small size, the continuous energy band becomes discrete energy levels together with series of unique photoelectric properties, known as the quantum confinement effect. QDs have played a central role in the development of nanoscience with several related milestone works, such as the discovery of quantum size effect^[2-4], size controllable synthesis^{[5, 6]}, biology^[7-9] and device applications^{[10, 11]}. Nearly 40 years later, it continues research to be a hot topic in nanoscience, while the connotation of the QDs family is constantly expanding, showing amazing development vitality. Traditionally, QDs field mainly focused on II–VI and IV–VI semiconductors containing heavy metals. Now, environmentally friendly I–III–VI (I = Cu, Ag; III = Ga, In; VI = S, Se) QDs have attracted more and more attention as an important subset of the QD family^{[12, 13]}. In a certain sense, I–III–VI QDs can be considered as a substitute structure by replacing the +2 cations (such as Cd) in II–VI chalcogenides with +1 and +3 cations. However, they are different in several aspects from traditional II–VI QDs due to the complex nature of nonstoichiometric composition and crystal structure, which brings a series of new laws, new phenomena and new challenges to the fulfillment of the whole QDs field, especially on synthetic chemistry and photoluminescence (PL) properties. More importantly, QDs play an irreplaceable role as the structural and catalytic model systems in understanding and demonstrating the unique laws on the nanoscale between lay atomic and bulk materials. As Prof. Feynman pointed out half a century ago, ‘there is plenty of room at the bottom’, which was first demonstrated with II–VI semiconductor system. Now, when it is expanded to more and more emerging new materials, such as carbon and other 2D materials, the large gap on composition and structure imposes series of great challenges on the synthetic manipulation, optical properties and mechanism understanding. From this aspect, I–III–VI QDs, as a close relative to II–VI QDs but with much more complex composition and structure variation, provides a great opportunity as a gradual bridge to make up the big gap between traditional QDs and emerging new QDs, such as carbon dots.

I–III–VI chalcogenides and their solid solutions have the advantages of low toxicity and continuously adjustable band gap covering the UV–vis to near infrared region which has attracted huge amount of attention as promising alternatives to the traditional cadmium- or lead-containing QDs^{[14, 15]}. These I–III–VI chalcogenides are usually direct bandgap semiconductors with strong light absorption^{[16, 17]}. Early research of I–III–VI compounds were focusing on thin films^{[18, 19]} and nanoparticles^[20] for solar cell applications^[21]. Recently, the excellent optical properties have made these non-toxic I–III–VI QDs useful in bioimaging, light-emitting diodes (LEDs), and photocatalysis etc. The composition of these multinary chalcogenides can be adjusted in a very large range while the anion skeleton remains stable, which brings excellent control ability of the optical and electrical properties^[22]. The important features include widely adjustable bandgap and long PL lifetime that are the key to related applications in devices and biological systems. However, this also brings the difficulties on the synthesis of uniform particles with well-controlled composition and size. Moreover, the coexistence of two or more cations bring abundant inner defects, together with the surface defects, resulting in high-density trap states and significantly different optoelectronic properties from II–VI QDs and imposing new challenges on related applications.

QDs have played a key role in many fields such as biology, devices and catalysis. Among these application, photocatalysis, as potentially the key solution to both environmental and energy problems, has been the most prominent and widely studied field not only for traditional QDs, but also for I–III–VI QDs and carbon dots^[23]. As a class of widely used semiconductor materials, I–III–VI QDs are more like a close relative of the traditional II–VI QDs and are naturally similar in terms of roots and research methods^[24]. In the development process, it also obtained a lot of experience from II–VI QDs. For the recently emerged QDs, such as carbon dots, there is a great gap in structure and properties and subsequently more research challenges from that of II–VI QDs. Considering this, I–III–VI QDs provide an important gradient bridge for these emerging new QDs, to better understand their new photoelectric properties, challenges and development process. As we summarized earlier, it lacks transition and bridge from the traditional inorganic II–VI semiconductors to the complicated pseudo-organic/biological carbon dots system, which imposes great challenges on comparison and correlation of their structure, optoelectrical properties and applications. Relatively speaking, there are many reviews of traditional QDs and emerging carbon dots, but fewer reviews of I–III–VI QDs. More importantly, there is no review article has summarized this field in the perspective of bridging between traditional QDs and emerging new carbon dots. So, here we try to give a brief review of I–III–VI QDs focusing on the discussion of their photoelectric and photocatalytic properties in systematic comparison with traditional QDs, and also give a short perspective of their key bridging role in the whole QDs family (Fig. 1). More efforts were contributed to resulted from challenges and issues the multinary composition in I–III–VI QDs, such as the complicated structural characteristics, balancing reactivity during synthetic process, optical properties dominated by abundant defect states, the crucial role of the long-lived charge carriers, and engineering of the bandgap and trap states in photocatalysis.

I–III–VI chalcogenides constitute a large class of semiconductor materials, whose band gap can be tuned not only by size and shape, but also by continuous composition manipulation^{[15, 25]}. The classic composition of bulk I–III–VI semiconductors is I–III–VI₂, but there are also some I–III₅–VI₈ compounds. Most importantly, their composition can be continuously adjusted in a large range without affecting the stability of the anion framework^[14]. Conceptually, the ternary I–III–VI chalcogenides can be made from II–VI materials by replacing two divalent Cd²⁺ with one monovalent (such as Cu⁺) and one trivalent (such as In³⁺) cation^[26]. Crystal structure of the ternary I−III−VI QDs includes chalcopyrite, zinc blend, and hexagonal wurtzite phases^[26]. However, the zinc blend and wurtzite phases usually can only be observed at high temperatures. In the chalcopyrite structure, the M⁺₂M³⁺₂ tetrahedron usually centered on the chalcogenide atoms^[27]. Due to the coexistence of multi cations with different radius and deviation from the stoichiometry, I−III−VI QDs tend to form lots of internal defects, which act as deep donor-acceptor-pair (DAP) trap states for charge carriers and thus form the characteristic PL from DAP recombination^[28]. This is the biggest difference between I–III–VI and traditional II–VI QDs that mainly containing surface defects. As an example, the defect sites in chalcopyrite CuInS₂ QDs include sulfur vacancies (V_S), copper vacancies (V_Cu) and indium on copper sites (In_Cu), where the Cu-related defects are more easy to produce due to the relatively weaker Cu–S bonds than In–S^[29].

The energy band structure of the multinary I−III−VI chalcogenide compounds can be adjusted by the alloy composition, thereby affecting the light absorption and reducing capability^[28]. Through theoretical calculation and analysis of energy band structure, it was found that Cu 3d and Ag 4d orbits, as part of the valence band (VB) composition, play a key role in increasing the VB level^{[30, 31]}. The introduction of Zn 4s4p orbital improves the conduction band (CB) level and improves the reducing capability^[32]. A large number of studies on I–III–VI QDs have shown adjusted band gap along with CB/VB positions to optimize the photocatalytic activity^[14]. It is worth noting that in the current research, the quaternary alloy QDs of traditional cadmium chalcogenides have not been well developed. In contrast, QDs of I–III–VI-derived quaternary alloys have attracted great interest due to their promising properties as non-toxic substitutes for cadmium chalcogenides in various applications. Two kinds of quaternary alloys are derived from I–III–VI QDs, including the self-alloy within I–III–VI semiconductors (such as CuIn_xGa_1–xSe₂S₈ and CuIn_xGa_1–x(S_1–ySe_y))^{[33, 34]} and the alloy of I–III–VI semiconductor with zinc or cadmium chalcogenides (such as Cu–In–Zn–S^[35] and Ag–In–Zn–S^[36]), whose composition can all be manipulated flexibly.

The preparation method and growth mechanism of ternary and multinary I–III–VI QDs are similar to traditional binary II–VI QDs, since the basic synthetic principles are similar. Currently, the hot injection method is still the main synthesis strategy of the complex I–III–VI QDs containing three or more elements. The metal precursors and chalcogen precursors can be formed in the same way capped with corresponding ligands. The main synthesis challenge of I–III–VI QDs is the reactivity balancing of multiple metal cations with strong capping ligands, which is a prerequisite for the formation of particles with uniform composition, phase and controllable structure (Fig. 2)^[37]. An example of the hot injection method of Zn_xCu_yInS_1.5+x+0.5y QDs is the reaction of metal acetate and sulfur in octadecene with oleic acid and dodecanethiol as capping reagents^[38]. Strong coordination ligands are usually used to control the relative reactivity of different metal cations to ensure the growth of pure phase alloyed QDs^[39]. It has also been reported that capping reagents, such as dodecane thiol, play a key role in forming pure phase QDs (CuInS₂) instead of composite QDs (Cu₂S and In₂S₃)^[34]. Insufficient capping of metal precursors may lead to the preparation of phase or composition separation, and uniform QDs cannot be obtained. Cation or anion exchange is also used for I–III–VI QDs (Fig. 2), such as the preparation of uniform CuInS₂ QDs or Cu₂S/CuInS₂ heterojunctions starting from Cu₂S QDs^{[40, 41]}. The conversion of the Cu₂S–CuInS₂ composite to phase-pure CuInS₂QDs was also observed at an elevated temperature of 250 °C^[42]. As an important member of I–III–VI QDs, AgInS₂ is usually reported as an analogue of CuInS₂ in the early studies. In 2007, ZnS–AgInS₂ was the first reported solid solution QDs with PL systematically adjusted by the amount of ZnS, which started to attract lots of attention to alloyed or solid solution QDs^[43]. Recently, much work has also focused on the synthesis of AgInS₂and AgInS₂–ZnS alloy QDs. AgInS₂ is slightly different because of the relatively high reactivity of Ag⁺ with S, which requires strong protection of the precursors as well as low temperatures to ensure the controllable particle nucleation and growth^[44].

Aqueous synthesis methods are receiving more and more attention because of the promising biological and catalytic applications of I–III–VI QDs. With mercaptoacetic acid, mercaptopropionic acid and other thiol-containing amines/alcohols as ligands, different types of I–III–VI QDs can be synthesized^[45-47]. In addition, the microwave- and ultrasonic-assisted methods can further optimize the aqueous phase synthesis method that decompose reactants under high-frequency electromagnetic radiation and high temperature^[48]. However, the QDs synthesized by the aqueous method usually have lower PL efficiency, larger size distribution, and subsequently PL tailing^[49]. In addition, the simple thiol ligands have poor stability and tend to fall off the QDs surface, resulting in QDs aggregation and PL decrease. Of course, with more and more reliable methods developed to improve the quality of aqueous QDs, the green synthesis method using water as the solvent is still a very promising route^[50-52]. Scale-up preparation of the I–III–VI QDs up to 4 L was also reported using a low-cost electric cooker^[53]. To further improve the application potential, size-selective separation may be a key to improving the quality of multinary QDs in aqueous phase^[54].

Semiconductor heterojunctions are critical in the bandgap engineering and wave function engineering of QDs. In the binary II–VI QD system, surface passivation by growing a larger band gap semiconductor shell, are used to decrease surface defects and improve PL quantum yield (QY)^[55]. In the I–III–VI core/shell QDs, the weight of the long-lived component increases with extended lifetime, while the fast decay component decreases^[56]. By comparing CdS and ZnS as shell materials, it was found that CdS has a better passivation effect with a higher QY and an almost single exponential decay, while the ZnS passivation still shows double exponential decay^[57]. Due to its low toxicity, ZnS is still a commonly used coating shell material for improving the optical properties of QDs. However, it worth to mention that cation exchange is also often reported in the preparation of CuInS₂/ZnS or AgInS₂/ZnS core/shell QDs with blue-shifted PL, which is caused by zinc alloying or doping^[58]. During the preparation of core-shell structure, cation exchange or diffusion often compete with formation of the ZnS or CdS shell. The low contrast of high-resolution transmission electron microscopy (HRTEM) between CuInS₂ and ZnS makes it difficult to distinguish the atomic structure of the interface^[59]. These challenges have brought a lot of difficulties to the synthesis of QDs with specific composition, structure and size, but further in-depth research has also brought fruitful novel results on the regulation of photoelectric properties. For example, a series of AgInS₂–ZnS alloyed QDs were prepared with a programmed-temperature-increasing method (Fig. 3)^[60]. The diffusion of wide-bandgap ZnS in AgInS₂–ZnS and the selective cation exchange of Zn for Ag were observed at increased temperatures. Spectroscopic studies show significantly improved PL QYs and ultra-long lifetime during alloying, and a new mechanism of structural evolution due to selective cation exchange was proposed. By systematically regulating the photoelectric properties of related doping and solid solution QDs, strategies such as selective cation exchange and controlled cation diffusion have resulted in the nearly ten times increase of the PL lifetime^[60].

The declaration of particle growth mechanism played a crucial on the delicate control of QDs as well as the development of nanoscience and nanotechnology. In general, the related work of I–III–VI QDs preparation still cannot summarize out a clear nucleation and growth mechanism, especially the different aspects from that of II–VI QDs. It should be noted that although there have been a lot of reports on related research, the adjustment of the size distribution, surface structure and optical properties of I–III–VI QDs are still not comparable to that of II–VI QDs, and in-depth mechanism research is still lacking. This has brought a lot of challenges for the regulation of optoelectronic properties. Without compromising quality, mass production of ternary QDs remains a challenge, so optimized reactions that exceed laboratory scale is an urgent requirement. In order to achieve this goal, an in-depth understanding of chemical synthesis mechanism is necessary. From a basic scientific point of view, the precise structure and composition control is still challenging, which subsequently limits the manipulation and understanding of its optical properties. Comprehensive single-dot characterizations and spectroscopic measurements might be the key to clarify the correlation of optical performance and other physical parameters with structure. In addition, the synthesis and growth control research is carried out focusing on the ground state QDs, while the excited state mechanisms of these I–III–VI QDs are of great significance, especially their size, composition and surface/interface effects that are more critical in the design and preparation of photoelectric materials and clean energy applications (such as photovoltaics and photocatalysis).

There is no doubt that the I–III–VI QDs show the widest tunable optical properties due to changes in both size and composition. These materials also show special features, including longer excited state lifetimes, wider full width at half maximum (fwhm), larger Stokes shift, high quantum efficiency, along with simple and economical synthesis^{[39, 54, 56, 61-63]}. Omata et al. calculated the size-dependent optical band gaps of I–III–VI chalcopyrite QDs with various compositions (Fig. 4(a)). The structure can be considered as metal cations embedded in the S/Se scaffold, similar to the structure of Cu_2–xS. The results agree well with the experimental values (Fig. 4(b)) and show that I–III–VI QDs can cover almost the entire spectral range from UV to near-infrared (NIR) by adjusting their composition and size. This shows the great potential of I–III–VI semiconductors as non-toxic QDs without heavy metals^[25].

The I–III–VI materials have photoelectric properties that are significantly different from traditional II–VI QDs, where the clarification of the excited state photophysical properties is crucial. The most important feature of I–III–VI QDs is their deep donor–acceptor pairs (DAPs, Fig. 5), which results in wide PL peaks and relatively large Stokes shifts. DAPs consist of different types of defects such as vacancies (V_Cu and V_S), interstitial atoms (Cu_i) and substituting (Cu_In), which are abundant in these ternary semiconductors due to the coexistence of multiple cations in the crystal lattice^[64]. The deep donor–acceptor state has advantages such as long PL lifetimes, along with the low toxicity of these materials is crucial for related applications. Mao et al. used femtosecond transient absorption spectroscopy to identify the ultra-long lifetime of their excited states in ternary AgInS₂ QDs, which are of great significance for photoelectric conversion applications. They also systematically studied the PL lifetime of QDs of different sizes (1.9–3.1 nm) in the full spectral range, which clarified the source of its long-lived excited states and the different contribution of surface defects and internal defects (e.g. DAPs). They found that high-energy surface trap states are short-lived, and low-energy intrinsic trap states (e.g. DAPs) are long-lived. The ratio of surface trap states decreases with increasing particle size^[65].

For chalcopyrite CuInS₂ QDs, the main contributions of optical transitions are assigned to the electrons at V_S and In_Cu, and holes at V_Cu^[68]. Since the relatively weaker Cu–S bonds than In–S, Cu defects are more prone to the inversion defects^[29]. The early PL studies show that the peak energy of different spectral components shifts to lower energy as the delay time increases^{[69, 70]}, a characteristic of DAP-related emission. In recent years, researches on the transient absorption spectroscopy, excitation power dependence, and time-resolved PL of CuInS₂ QDs have gradually developed toward the direction of "free-constraint" mechanism. The electrons are delocalized in CB, and the holes are located on Cu⁺ or point defects that are generally considered to be V_Cu. For band gap engineering^[71], thanks to the strong foundation of the traditional I–III–VI semiconductor industry, basic physical parameters such as CB, VB, bulk band gap, mobility, n-/p-doping and other information are relatively clear. However, due to the influence of the points and the size together, the mechanism research of QDs and the control of photoelectric properties are still far behind those of II–VI group QDs. For some core/shell structured CuInS₂/ZnS with strong PL at short wavelengths, it is difficult to tell whether the spectral blue shift is the quantum size effect or caused by the introduction of wide-bandgap ZnS in alloy formation. With the increase of the ratio of zinc to copper, the absorbance and PL have a clear blue shift^[72]. In optical studies, the band gap or E_LUMO–E_HOMO are often reported, but without the absolute position information of these levels. However, this information is essential for designing an efficient energy conversion system, including photovoltaics and photocatalysis. X-ray photoelectron spectroscopy (XPS) an be used to determine the VB position (or molecular HOMO orbital), while ultraviolet photoelectron spectroscopy (UPS) c for the CB position (or molecular LUMO orbital). Electrochemical method is another convenient approach for measuring redox potentials, which provides supplementary information beyond spectroscopic measurements, especially the surface states that affect the electron transportation on the electrode. However, careful considerations are still required regarding the electrochemical measurement of I–III–VI QDs.

Band gap of I–III–VI QDs can be adjusted within a very wide range due to their composition manipulation capability, which represent the most promising visible-light-active photocatalysts for hydrogen production due to their excellent optoelectrical properties. Bulk I–III–VI sulfides were actually the first reported visible light photocatalysts with adjustable band gap used for photocatalytic water splitting^{[25, 73]}. The early work of I–III–VI sulfide photocatalysts mainly focused on the effect of composition adjustment on photocatalytic performance, such as (AgIn)_xZn_2(1–x)S₂, (CuIn)_(x)Zn_2(1–x) S₂ and other solid solution materials. In 5s5p and Cu 3d/Ag 4d orbitals in I–III–VI sulfides can participate in the formation of CB and VB, respectively, giving continuously adjustable band gap over a wide range^[74]. However, due to the complex composition and structure, the preparation usually requires high temperature annealing to increase crystallinity and reduce defects, which also results in its size increase and low specific surface area. For further improvement, nanostructuring can reduce the diffusion distance of the photo-generated charge carriers to the catalyst surface, and effectively improve the efficiency of surface hydrogen evolution reaction^{[75, 76]}. Recently, researchers have developed simple hydrothermal and solvothermal methods to prepare high-specific surface area I–III–VI solid solution nanostructures, such as porous microspheres and hieratical nanostructures, which can provide a rich surface active sites^[77]. Above all, high-crystallinity and small-size nanoparticles have always been the ideal structures pursued in photocatalytic applications. In addition to traditional composition and structure control, cocatalysts loading, and heterostructure construction strategies, the I–III–VI QDs also provide opportunities for size and surface ligand control, which provides unique advantages^{[36, 78-80]}. In-depth systematic research on the composition and particle size of different systems is expected to screen out I–III–VI-based solid solution QDs photocatalysts with excellent performance, which is expected to greatly boost the visible light catalytic efficiency^[14].

For multinary I–III–VI QDs, the small size of QDs and the narrow band gap of sulfides have important effects on the photocatalytic water splitting technology. As the size decreases, the band structure of QDs changes from continuous to discrete along with upshift of VB, resulting in the enhanced photoreduction capability, so reducing the size will usually facilitate photocatalytic H₂ production. The reduction in particle size results in larger surface area and more active sites on the surface^[81-84]. However, it has also been reported that the number of surface traps and the nonradiative recombination decrease with increasing size to promote the photocatalytic water splitting. Kameyama et al. systematically explored how the size of (AgIn)_xZn_2(1−x)S₂ QDs affects the band structure and optical properties (Fig. 6), and found that the band gap narrows with increasing size, where the photocatalytic activity is optimal when the average size is 4.2–5.5 nm^[85]. Shape control is another important factor for the I–III–VI QDs photocatalysts. One-dimensional (1D) nanostructures, such as elongated nanorods, are a promising type of photocatalyst that facilitates the charge separation and improves activity^{[84, 86-88]}. Han et al. successfully prepared a series of elongated morphological nanostructures, such as CuGaS₂ tadpoles, CuInS₂ bullets, CuInZnS rods and Cu₂ZnGeS₄ worms^[89].

Composition manipulation plays a critical role in I–III–VI QDs photocatalysts, not only for bang gap engineering, but also for catalytic activity. The band structure of the multinary compounds can be adjusted by composition, thereby affecting the light absorption ability and reduction ability to improve the photocatalytic performance (Fig. 6). Through theoretical calculation and analysis of energy band structure, it was found that Cu 3d or Ag 4d orbits play a critical role in the VB, while In 5s5p orbital constructs the CB of I–III–VI QDs^[30]. For I–III–VI-based solid solutions, the introduction of Zn 4s4p orbital improves the CB level and improves the reducing capability of the photogenerated electrons^[31]. A large number of studies on I–III–VI-based alloys have adjusted the band gap and the CB/VB positions via the ratio of metal cations to optimize the photocatalytic activity^[14]. For example, Chen et al. pointed out that the AgIn₅S₈ photocatalyst with visible light response can perform photocatalytic hydrogen production at wavelengths greater than 640 nm^[90]. I–III–VI-based solid solution QDs, such as ZnS–AgInS₂^[91], CuIn_1−xGa_xS₂^[92] and CuGa₂In₃S₈^[93], have shown excellent photocatalytic hydrogen production performance in recent studies. Among them, Yu et al. tested the influence of Ga content on catalytic activity of CuIn_1−xGa_xS₂ QDs (optimized composition of CuIn_0.3Ga_0.7S₂)^[94]. The hydrogen production rate of CuGa₂In₃S₈ QDs loading with Ru even exceeded the corresponding bulk material prepared by high temperature annealing method^[95]. The introduction of Group I elements (Cu or Ag) is an effective means to reduce the band gap in the composition control of I–III–VI photocatalysts, but it also brings severe charge carrier recombination and thus reduced photocatalytic efficiency, resulting in the dilemma of unable to cooperate band gap narrowing and charge recombination. For this, Mao et al. reported that a proper amount of Ag doping (i.e. Ag : In < 1.5 : 10) can lead to continuous increase of photocatalytic hydrogen production, which achieved synergetic band gap narrowing and lifetime increase ^[74]. PL mechanism study shows that the ratio of intrinsic donor–acceptor state increases with moderate silver doping, beneficial to the lifetime elongation. This discovery provides a new perspective for narrow-band-gap photocatalysts^[74].

The surface ligands not only determine the aqueous/organic dispersibility of the QDs, but may also serve as the surface catalytic center^{[96, 97]}. Generally, the surface of QDs synthesized under the conditions of high temperature and high boiling point or organic solvent are often covered with a layer of organic ligands with long hydrocarbon tails (such as oleylamine, oleic acid and trioctylphosphine oxide)^[98] to obtain high crystallinity and high dispersibility. They are easily dispersed in non-polar solvents or polymer matrix, but are difficult to disperse in aqueous phase. The pristine long-chain ligands are usually exchanged with short-chain bipolar ligands (such as 3-mercaptopropionic acid (MPA) or simple inorganic ions), which provide increased aqueous colloidal stability and convenient charge transport channels during photocatalysis beneficial for the photocatalytic activity. Lots of I–III–VI QDs photocatalysts were also directly prepared by aqueous methods with short-chain ligands. These short chain ligand-coated colloidal QDs exhibit better charge separation, high conductivity, and enhanced photocatalytic activity^[99]. Moreover, Mao et al. found that Ag–In–Zn–S QDs modified with mixed ligands -shows an unexpected synergetic effect on the photocatalytic properties, where the introduction of MPA led to an increase in PL lifetime and a decrease in non-radiative recombination that can be attributed to the stronger caping effect of MPA compared to that of L-cysteine^[100]. Rao et al. found that simple S^2– ion coated (ZnS)_0.4(AgInS₂)_0.6 QDs have more active sites than the original QDs, thus making the photocatalytic activity increase^[101].

In heterojunctions formed by combining semiconductor materials with different band gaps and energy levels, especially type-II heterojunctions, photogenerated electrons can transfer from high CB materials to relatively low CB materials, while photogenerated holes can migrate reversibly. This band arrangement thereby achieves efficient separation of electrons and holes. CuIn₅S₈/Ag₂S, CuInSe₂/TiO₂^[102], AgIn₅S₈/TiO₂^[103], AgInS₂/ZnS^[104] and other heterojunctions have shown the capability to efficiently transfer the photo-generated electrons and holes, extend the carrier lifetime, and promote the photocatalytic performance of the composite. In-situ growth method is also reported for constructing 0D/2D heterojunctions, such as Zn–AgIn₅S₈QDs/g-C₃N₄ nanosheets^[105].

Various cocatalysts, especially noble metals, are widely used as high-efficiency promoters in water splitting systems to enhance charge separation and catalytic activity. The Fermi level of the noble metals are often lower than that of the semiconductors and the electrons generated by the semiconductor can smoothly migrate to the active site of the precious metal to reduce the surface adsorbed H⁺ to H₂^[106]. The cocatalysts promote the separation of photo-generated carriers of semiconductor materials to a certain extent, which reduces the internal and surface electron/hole recombination. The reverse combination reaction is also suppressed to enhance the surface chemical reaction, and thus the photocatalytic performance is significantly improved. Various high-efficiency metal cocatalysts have loaded onto I–III–VI QDs, including Pt^[107], Ru^[108], Pd^[109], and Rh^[110]. In addition, adjusting the Cu content makes it easier to cover the entire solar spectrum, but it is often observed that even a relatively low Cu content will cause the photocatalyst to deactivate. Mao et al. recently found that loading Pt cocatalyst could cause a shift in the photocatalytic activity trend in a series of Cu–In–Zn–S QDs, where the optimized sample was changed from Cu/In/Zn of 0.1/10/5 without Pt to 1/10/5 with Pt, e.g. the tolerance to Cu doping was increased by 10 times^[15]. This phenomenon is ascribed to the competitive effect of band gap narrowing by Cu doping, charge recombination by Cu defect sites and charge separation induced by Pt cocatalysts. In addition to noble metals, transition metal sulfides are a class of cocatalysts with outstanding performance in photocatalytic hydrogen production. Mao et al. pointed out that the introduction of MoS₂ nanosheets increased the photogenerated electron transfer of Zn–AgInS QDs and optimized the photocatalytic activity^[111]. Ning et al. combined 0D AgInZnS QDs and 2D MoS₂ nanosheets to construct a heterojunction, and systesystematically explored the effects of the amount of Ag doping, sacrificial reagents, and MoS₂ co-catalyst loading on the photocatalytic activity. It was found that the introduction of the inexpensive MoS₂ nanosheets made the AQE up to 41% (at 400 nm) with a high hydrogen production rate of 40.1 mmol g^–1 h^–1[112].

Photocorrosion caused by the oxidative holes is a problem faced by most narrow-band-gap sulfide photocatalysts^[113]. Stability is still one of the biggest challenges for these narrow-bandgap sulfide QDs, similar to CdS and CdSe. It’s worth to note that sulfides play a critical role for visible-light-driven photocatalysis, so the stability and photocorrosion issues have attracted lots of attention in traditional QDs photocatalysts. For example, in addition to the above-mentioned modification methods, other experimental conditions for photocatalytic water splitting (such as pH, temperature, sacrificial agents, etc.) also affect the hydrogen production activity of the photocatalysts. Lots of efforts have been contributed to the separation of the photogenerated charge carriers, such as heterojunction construction and use of hole sacrificial agents, such as S^2–/SO₃^2– and ascorbic acid. It has also been realized that hole trapping in the defect states is probably the main cause of photocorrosion, and corresponding strategies have started to be explored, such as surface defect engineering and hole accepting ligand modification, which are also learned from traditional QDs. For metal sulfide photocatalysts, S^2– and SO₃^2– mixtures are usually used as sacrificial agents to capture the photo-generated holes and thus minimize recombination between electrons and holes^[114]. In the photocatalysis process, the holes oxidize S^2- to S as a competitive reaction to the photocorrosion of the catalyst themselves. The resulted dark-color S and S₂^2– will react with the SO₃^2– added to the system to form light-color S₂O₃^2– and get eliminated to avoid light shielding effect. In addition to the Na₂S/Na₂SO₃ system, triethanolamine, L-cysteine, and ascorbic acid have also been widely explored as sacrificial agents for QDs-based hydrogen production^[115].

As mentioned above, the QDs family has shown amazing developing vitality, and the connotation is also constantly expanding, bringing a series of new phenomena, new principles and new challenges to the QDs field. Now, as it expands from traditional II–VI semiconductors to carbon and other emerging materials, there is a huge gap in the composition and structure, leading to completely different synthetic chemistry, PL mechanisms and applications^[23]. Thus, it is difficult for carbon dots to learn directly from traditional II–VI QDs. As a close relative of II–VI QDs, I–III–VI QDs still belong to traditional semiconductors, for which the principles of synthesis and structure are similar, but with more complicated composition and structural variations that brings new optoelectronic properties. From this point of view, their corresponding manipulating strategies and development history that are closely related with II–VI QDs but with new challenges, provide a opportunity to bridge the huge gap between traditional QDs and emerging new carbon dots. For photocatalysis, traditional QDs play an important role in the clarification of photocatalysis mechanisms and high-efficiency photocatalyst design, while carbon dots show unique charm in all the three aspects of photocatalysis, e.g. enhancing light absorption, charge transfer and co-catalysis. Carbon dots reflect the dual merits of small-size QDs and carbon materials, show high catalytic activity for many reactions, and can be used as multi-functional compositions of many high-performance photocatalysts and electrocatalysts^[116-119]. The unique structural advantages make carbon dots not only provide photo-generation e^–/h⁺, but also provide photo-generated protons and photo-controlled electron transfer, which is useful in many organic catalytic reactions beyond photocatalysis^[120]. Of course, the carbon dot-based catalysts are still in its infancy in terms of fine structure manipulation, ultrafast spectral mechanism and in-situ catalytic mechanism characterizations, which still has a large gap compared with traditional QDs. I–III–VI QDs photocatalysts have made great progress by learning from II–VI QDs, especially the concise manipulation of size, shape and surface, but also on their own advantages including continuous band gap tuning via composition, utilization of long-lived charge carriers and delicate manipulation of the intrinsic (e.g. DAPs) and surface defect states. This actually provides an opportunity for carbon dots to learn and consult, especially the composition tuning, defect manipulation and the implication of ultrafast spectroscopy on the complicated photocatalytic mechanisms. More detailed comparison is listed below in Table 1.

In summary, I–III–VI QDs own lots of unique structural and optical properties and play a key role in photoelectric fields, including photocatalysis. It has similar advantages and characteristics of traditional QDs, such as the size-dependent quantum confinement effect and high specific surface area. On one hand, traditional QDs have developed lots of useful strategies to improve light harvesting and charge separation, such as the delicate control over size, shape, surface exposure and heterostructures. On the other hand, these ternary or multinary I–III–VI QDs provide several characteristic advantages that traditional binary QDs do not have, especially the wide range regulation of composition and band gap, as well as the rich long-lived trap states, which greatly expands the design of solid solution QDs and composite photocatalysts with complex compositions and structures. However, the complicated composition also brought a series of challenges on the structure and synthesis of these QDs, such as the precise control of the composition, the balancing of cation reactivity, the unwanted cation exchange and diffusion for heterojunction construction. This may rely on the in-situ monitoring of the growing process and the deep understanding of the synthetic chemistry. Subsequently, more efforts are needed on the clarification of size- vs. composition-dependent band gap, the PL origin from the abundant trap states, the engineering and utilization of the long-lived charge carriers, which requires investigation and manipulation of the excited states by ultrafast spectroscopy. As mentioned above, all structure controled methods of traditional QDs photocatalysts have been applied to the I–III–VI QDs system, including particle size, composition, surface ligands, cocatalysts, hole sacrificial agents, etc., which provide systematic research on related strategies as very good model system. In addition, the performance shortcomings of the traditional chalcogenide photocatalysts usually can also be found in the I–III–VI QDs system, such as poor stability, slow hole extraction, low charge carrier utilization efficiency, etc. These problems rely on ultrafast spectroscopic research, which also provides a good inspiration for other photocatalytic systems. In principle, photocatalysis is a special form of electrocatalysis, where the electrons and holes are provided by light irradiation. The advantages of I–III–VI QDs in mechanism research can also be extended to the entire photo/electrocatalysis fields Furthermore, as a natural multi-composition and multi-interface catalyst, I–III–VI QDs can be useful in the distribution process of active species, the adsorption of active species, and electron transfer in other catalytic systems in a much broader way. The identification and change of the active sites and other issues may provide useful help as a suitable model system for understanding the catalyst from the perspective of physical and chemical interface engineering, for which more profound ultrafast and in situ spectroscopy studies are the key. In terms of the overall significance of photocatalysis and other catalytic studies, I–III–VI QDs are a class of materials that far surpasses traditional binary QDs. With the joint efforts of researchers in related fields, we hope that they will play an increasingly important role in catalysis research.

This work is supported by the National Natural Science Foundation of China (21908081, 21501072, 51972216, 51725204, 21771132 and 52041202), the National MCF Energy R&D Program (2018YFE0306105), Innovative Research Group Project of the National Natural Science Foundation of China (51821002), the Jiangsu Specially-Appointed Professors Program, and the Natural Science Foundation of Jiangsu Province (BK20190041, BK20190828 and BK20150489).