Managing photoinduced electron transfer in AgInS₂–CdS heterostructures

Attaining clean renewable energy as an alternative to unsustainable fossil fuels continues to gain traction across multidisciplinary fields. Of particular interest are ternary I–III–VI₂ semiconductors, such as AgInS₂ and CuInS₂, which are finding applications in light-emitting diodes (LEDs),^1–5 photocatalysis,^6–8 batteries,⁹ and photovoltaic devices.^10–15 The photophysical properties of I–III–VI₂ semiconductors are complex and differ from their analog, cadmium-based binary chalcogenides. Binary semiconductor systems, such as CdTe and CdSe, exhibit distinct excitonic peaks, narrow photoluminescence (PL) bandwidths, and short emission lifetimes.^16–18 These excitonic peaks and sharp absorption onsets are indicative of the dominant band edge transitions. These semiconductor nanocrystals also exhibit relatively small Stokes shift, which is indicative of dominance of band–band charge carrier recombination channels with minimal contribution from intraband states.¹⁶ 

In contrast to II–VI semiconductors, the optical properties of I–III–VI₂ systems are characterized by a lack of distinct excitonic peaks and emergence of a long tail in their absorption spectra, large Stokes shifts, broad PL bandwidths,^19,20 and long PL lifetimes.^20–25 The spectral features arise from surface and interstitial defects (trap states) and donor–acceptor pairs (DAPs), which act as radiative and non-radiative recombination sites.^24,26–28 The surface trap states originate from the non-bonding orbitals of the undercoordinated surface atoms. Proper choice of ligands can passivate the surface of quantum dots (QDs) if they bond to the undercoordinated surface atoms. These uncoordinated atoms yield electronic states that lie within the bandgap of the semiconductor and, thus, act as a trap for electrons and holes.²⁹ The use of a higher bandgap material as a shell to the AgInS₂ core has been seen to significantly improve its photophysical properties. ZnS shell on AgInS₂ is reported to have reduced surface trap centers and, thus, enhance the PL quantum yield (PLQY).^19,30 Therefore, it is intriguing to study how capping AgInS₂ with CdS affects the photophysical properties, particularly interfacial charge transfer processes that are relevant to photocatalysis.

To achieve efficient conversion of light energy into chemical energy, it is important to maximize charge separation by suppressing the charge recombination within the semiconductor nanocrystal as well as back electron transfer from the electron transfer product. Careful design strategy is required to suppress these undesired processes. Viologen molecules are useful probes to evaluate the interfacial electron transfer properties as they can capture electrons and yield a stable reduced viologen radical in an inert atmosphere. By tracking the formation of viologen radicals, one can elucidate kinetic factors limiting interfacial electron transfer processes.^31,32 We have now employed ethyl viologen, EV²⁺ having one-electron reduction potential (E⁰ = −0.449 V vs NHE) less negative than the conduction band (CB) of the AgInS₂ and AgInS₂–CdS QDs (E_CB ∼ −1.0 V vs NHE) can readily capture photogenerated electrons.³¹ Using EV²⁺ as a probe, we have now succeeded in elucidating the involvement of conduction band and sub-bandgap states in promoting electron transfer. The role of AgInS₂–CdS heterostructure in enhancing the efficiency of charge transfer is also discussed.

AgInS₂ and AgInS₂–CdS QDs were synthesized by using modified hot-injection methods. Experimental details of the synthesis of AgInS₂ and AgInS₂–CdS QDs are presented in the supplementary material (Fig. S1). The transmission electron microscopy (TEM) images of the two QDs are shown in Figs. 1(a), 1(b), 1(d), and 1(e). The size distribution of the QDs was determined by measuring the edge length of ∼150 QDs [see Figs. 1(c) and 1(f)]. AgInS₂ and AgInS₂–CdS are pyramidal shaped with average edge lengths of 4.8 ± 0.6 and 5.5 ± 0.9 nm, respectively. The slight increase in size for the AgInS₂–CdS can be attributed to the introduction of the desired thin CdS shell. This thin shell is able to remediate the surface defects without blocking the charge transfer at the heterostructure interface. In addition, these nearly monodispersed particles show a tail feature in absorption and broad emission, as shown in Figs. 2(a), 2(c), and 2(d), hence confirming that polydispersity is not the key source of such features in these materials.³³ 

The absorption and emission properties of AgInS₂ and AgInS₂–CdS are presented in Figs. 2(a)–2(d). The absorption spectra [Fig. 2(a)] show absorption below 700 nm. The intraband transitions that induce tail absorption^24,26–28 slightly shift to blue upon capping with CdS. Both of these nanocrystals continue to show broad emission in the visible-near IR region. These broad emissions imply that the PL originates from the recombination of carriers trapped by intraband levels and not exciton recombination.^6,24,34

In accordance with the blue shift in absorbance, we also see a shift in the emission spectra from ∼750 nm (AgInS₂) to ∼650 nm (AgInS₂–CdS). The excitonic transition that appears near the band edge (∼500 nm) is not visible in the absorption spectrum. It is buried in the tail absorption and, hence, cannot be resolved from absorption spectra. The excitation spectra recorded at different monitoring wavelengths [Figs. 2(c) and 2(d)], however, show the origin of different optical transitions contributing to the emission. When monitored at longer wavelength (e.g., 800 nm), we see excitation spectra that closely follow the absorption spectra indicating contributions from bandgap and intrabandgap transitions. When the monitoring wavelengths are set close to the band edge (e.g., 570 nm), we see the appearance of a shoulder in the range of 450–500 nm region. This shoulder that corresponds to excitonic transition indicates a bandgap of ∼2.5 eV, similar to the one reported elsewhere.^20,35 Thus, the excitation spectra presented in Figs. 2(c) and 2(d) allow us to identify the origin of emission from different transitions. Interestingly, the CdS capping seems to have little effect on these transitions that contribute to overall emission.

Ternary semiconductors such as AgInS₂ exhibit both direct bandgap excitation and intraband transitions arising from the vacancies and interstitial (antisite) defects (Scheme 1).^35–39 The lack of the bandgap (excitonic) emission dominance in the steady state emission spectra has been attributed to the sub picosecond decay of the bandgap states to the intraband states.^35,38 Earlier reports show that the emissions of ternary QDs are dominated by donor acceptor pair (DAP) recombination. Sulfur vacancy (V_S) and silver interstitial (Ag_int.) act as plausible defects for donor state, whereas silver vacancy (V_Ag) and sulfur interstitial (S_int.) act as the acceptor states.^{25,36,40–42}

The emission decay of AgInS₂ and AgInS₂–CdS nanocrystal suspensions was monitored in a nanosecond laser flash photolysis set up using 355 nm laser pulse excitation [Fig. 2(b)]. Long-lived emission lifetimes of AgInS₂ (avg. lifetime of 493 ns) and AgInS₂–CdS (avg. lifetime of 300 ns) recorded for these samples indicated the dominance of DAP radiative pathway for the deactivation of the excited state.^24,25 Electrons from shallow donors (close to VB) recombining with shallow acceptors (close to CB) give rise to the emission in the higher energy part of the emission band, while deep seated donor and acceptor states are responsible for the emission seen in the lower energy part (red region) of the emission band. The broad emission and tail absorption are independent of shape of the nanocrystals but dependent on the ratio of Ag:In.⁴³ Related works on single particle of the CuInS₂ counterpart have also ruled out the possibility of polydispersity³³ and size⁴⁴ as the only major sources of broadening in emission linewidth. Therefore, trapping of charge carriers in the randomly positioned local states within the bandgap (recombination centers) is the plausible reason for the broad emission and long lifetime seen in ternary I–III–VI₂ QDs.^33,40,43,44

Other models depart from this defect mediated model (DAP) by invoking self-trapped exciton (STE) as another plausible mechanism in CuInS₂^45,46 and AgInS₂.⁴⁴ Knowles et al.⁴⁵ argue that excitation of CuInS₂ nanocrystal from its ground state to a delocalized excitonic state is followed by rapid relaxation to a self-trapped excitonic state. This trapping of exciton is modulated by hole localization (a direct consequence of strong vibronic coupling and nuclear distortion at the copper state). As a result, emission from this self-trapped excitonic state was characterized by a large Stokes shift, longer lifetimes, and a broadened band shape. Similarly, a long tail in the absorption spectrum has been associated with direct excitation of this self-trapped excitonic state.^44–46 However, the presence of self-trapped exciton cannot be the sole contributor to the observed emission at longer wavelengths. The absorption tail extending beyond the band edge contributing to the longer wavelength emission [excitation spectra in Fig. 2(d)] and wavelength dependent emission lifetimes suggest that intraband transitions contribute to the observed emission at longer wavelengths.

The adsorbed species play an important role in dictating the excited state interactions of semiconductor nanocrystals.⁶ For example, the excited semiconductor nanocrystals can undergo energy transfer or electron transfer with the adsorbed molecule depending upon the energy match and/or redox potentials.^47–52 In a previous study,^53,54 we probed the surface interactions of perovskite nanocrystals with viologen molecules and how these interactions dictate the interfacial electron transfer.

The interaction of AgInS₂ and AgInS₂–CdS nanocrystals with ethyl viologen was probed through the quenching of emission [Figs. 3(a) and 3(c)]. With increasing concentration of ethyl viologen, we see a decreased emission of nanocrystals indicating an electron transfer to adsorbed viologen species. In the absence of redox species, the QDs under bandgap excitation undergo charge carrier recombination, both radiatively and non-radiatively [reaction (1)]. When a redox molecule (viz., ethyl viologen) is present near the surface, an additional deactivation pathway of electron transfer [reaction (2)] is encountered,

while EV²⁺ refers to ethyl viologen and QD represents AgInS₂ or AgInS₂–CdS.

The quenching of the QD emission observed at relatively low concentrations of EV²⁺ indicates strong interaction between the QDs and the EV²⁺. This interaction is represented by the existence of an equilibrium between adsorbed and unadsorbed EV²⁺ with an apparent association constant, K_app,

where the observed quantum yield [ϕ_f(obs)] refers to the resultant emission arising from viologen-bound (ϕ_f′) and unbound QD (ϕ_f⁰).⁵⁵ The linearity of the double reciprocal plot of observed emission difference and EV²⁺ concentration [Figs. 3(b) and 3(d)] shows the validity of expression (4). From the slope and intercept of this plot, we obtain the apparent association constant (K_app) for equilibrium (3). The K_app values were 6.2 × 10⁶ and 2.0 × 10⁶ M⁻¹ for AgInS₂ and AgInS₂–CdS, respectively. The high K_app values indicate that EV²⁺ molecules are strongly bound to the surface of the QDs. Slightly lower K_app observed for AgInS₂–CdS (as compared to AgInS₂) indicates that the influence of CdS in dictating the attachment to the QD surface as opposed to direct attachment to AgInS₂ surface.

Elucidation of mechanistic and kinetic details of interfacial electron transfer at semiconductor interface is important to establish the photocatalytic properties of semiconductor nanocrystals.^56,57 The photoluminescence quenching results discussed in the previous section point to the electron transfer between excited AgInS₂ and AgInS₂–CdS nanocrystals and EV²⁺. Since the product of electron transfer, EV^+•, has the spectroscopic fingerprint with maxima at 605 nm (ε = 1.4 × 10⁴ M⁻¹ cm⁻¹),^58,59 we can directly probe the production of electron transfer product under steady state irradiation conditions. Figures 4(a) and 4(b) show the accumulation of EV^+• when AgInS₂ and AgInS₂–CdS suspensions containing 100 μM EV²⁺ were irradiated with visible light (>420 nm). By employing >420 nm excitation wavelengths, both the AgInS₂ core are excited without directly exciting the EV²⁺. Note that the CdS absorption is too small to make any direct contribution to the EV^+· yield. Figure 4(d) shows the increase in EV^+· concentration with a time of irradiation. The EV^+· attains a steady state concentration after about 800 s of irradiation. The attainment of steady state concentration at longer times is reflective of forward and back electron transfer rates becoming equal,^31,53,60,61

It is evident that the AgInS₂–CdS heterostructure enables higher steady state concentration of EV^+· reflective of better charge separation. The observed yield of electron transfer product in AgInS₂–CdS is almost double that observed in AgInS₂.

We also conducted irradiation using a long pass filter of 515 nm and tracked EV^+· growth, as shown in Figs. S2(a) and S2(b). Under these irradiation conditions, only AgInS₂ can be excited, thus avoiding any direct contributions from CdS excitation. We see the formation of EV^+· but attaining a lower steady state concentration [Fig. 4(e)] because of filtering of excitations below 515 nm. Again, we observed a higher yield of EV^+· with AgInS₂–CdS heterostructure. The presence of a CdS shell around AgInS₂ facilitates better charge separation and suppresses the back electron transfer.

It should be noted that the experiments in Figs. 4(d) and 4(e), the selected excitation wavelength allowed both band edge and sub-bandgap transitions. To check the participation of the sub-bandgap states, we irradiated the samples using excitation wavelengths >550 nm. At these excitation wavelengths, direct bandgap excitation of CdS or AgInS₂ is not feasible. Figure 4(f) shows EV^+· formation as we subject AgInS₂ and AgInS₂–CdS to sub-bandgap excitations [corresponding spectra are presented in Figs. S2(d) and S2(e)]. Interestingly, the difference in EV^+· yields with AgInS₂ and AgInS₂–CdS was relatively small. These results indicate that the electrons from the trap states are accessible and can as well be leveraged for photocatalysis [see scheme in Fig. 4(c)]. Under these different excitation wavelengths, we see a higher yield of electron transfer product with AgInS₂–CdS NCs. We also carried out control experiments by exposing only EV²⁺ in deaerated toluene/ethanol 87:13 v/v% (i.e., without QDs) to similar irradiation conditions described above. For the >420 nm excitation wavelengths, we see negligible contribution as a result of direct excitation [see trace (c) Fig. 4(d)]. By utilizing >515 and >550 nm filters, we do not see any generation of EV^+· [Figs. S2(c) and S2(f)].

We also independently measured quantum yields of EV^+· formation using ferrioxalate actinometry.^62,63 The details of actinometry measurements are presented in the supplementary material. Under steady state irradiation conditions (>420 nm, 30 min), we obtain a quantum yield (QY) of EV^+· of 14.5% and 29.3% for AgInS₂ and AgInS₂–CdS nanocrystals, respectively. Although the initial electron transfer from excited AgInS₂ could be very high (>90%), a significant fraction of transferred electrons is lost in back electron transfer (vide infra). This increase in electron transfer yield is consistent with the results obtained with selective bandgap excitation of AgInS₂ core in Figs. 4(d) and 4(e). The higher EV^+· QY observed for AgInS₂–CdS further supports the argument that the heterostructure configuration of AgInS₂–CdS facilitates improved charge separation.

Transient absorption spectroscopy is convenient to probe the ultrafast electron transfer events following the laser pulse excitation of AgInS₂ and AgInS₂–CdS nanocrystals. When excited with 400 nm laser pulse excitation, AgInS₂ shows a broad bleach in the 500–750 nm region [Fig. 5(a)]. The excited state is long-lived as evident from the slow bleach recovery. Corresponding time-resolved transient absorption spectra of AgInS₂–CdS NCs are shown in the supplementary material (Fig. S3). The recovery becomes faster in the presence of EV²⁺ [Fig. 5(b)]. This trend further confirms that the ultrafast electron transfer to EV²⁺ is the preferred pathway for excited state deactivation of both AgInS₂ and AgInS₂–CdS QDs. The fast-bleaching recovery recorded in nanosecond time scale at different EV²⁺concentrations for these two suspensions represents kinetics of conduction band electrons [Figs. 5(c) and 5(d)]. The residual bleach at the end of these traces was further analyzed on a longer microsecond time scale [Figs. 5(e) and 5(f)]. The bleach recovery traces were analyzed with biexponential kinetic fit and kinetic parameters are summarized in Table I (see supplementary material Tables S2 and S3 for short timescale and Tables S4 and S5 for long timescale).

In the absence of EV²⁺, we observe a relatively long lifetime for the recovery of the bleach for both AgInS₂ and AgInS₂–CdS nanocrystals. The average lifetime estimated for AgInS₂ and AgInS₂–CdS is 780 and 502 ns, respectively. (These lifetimes are slightly greater than the emission lifetime as the transient absorption represents various deactivation pathways.) As discussed earlier,^20–25 the surface and the interstitial trap states extend the lifetime of charge carriers.^24,25 The CdS shell remediates some of the surface defects that are responsible for long-lived recovery as evident from the decreased lifetime of recovery.

The traces recorded in Figs. 5(c) and 5(d) show faster bleach recovery of AgInS₂ and AgInS₂–CdS with increasing concentration of EV²⁺. More than 70% of the recovery occurs within 50 ps at the highest EV²⁺ concentrations. The short lifetime of bleach recovery decreases from 64.7 to 6.3 ns for AgInS₂ suspension containing 8 μM EV²⁺ and 110.0 to 7.3 ns for AgInS₂–CdS suspension containing 8 μM EV²⁺ suggesting rapid electron transfer in both of these systems. If we consider the observed decrease is entirely due to the electron transfer process [reaction (2)], we can obtain rate constant for electron transfer (k_et) using the following expression:

where τ₀ and τ′ are the fast processes lifetimes of bleaching recovery in the absence and presence of an electron acceptor.

The rate constants determined at 8 μM EV²⁺ were 2.0× 10¹¹ and 1.3 × 10¹¹ s⁻¹ for AgInS₂ and AgInS₂–CdS, respectively, and represent electron transfer from conduction band electrons. The CdS shell around AgInS₂ appears to slow down the electron transfer process. As noted in our earlier work,^53,54 the surface interaction plays an important role in dictating the rate and efficiency of interfacial electron transfer. As seen in the case of emission quenching measurements, the lower association constant observed for EV²⁺ with AgInS₂–CdS would also influence the rate of electron transfer. Strong interaction between viologen (MV²⁺) and CsPbBr₃ nanocrystals has also been shown to exhibit extended charge separation with electrons residing in the viologen moiety and holes residing in the semiconductor nanocrystals.⁶⁴ Given the long lifetime of the excited state of AgInS₂, we would expect a long-lived charge separation with the stabilization of electron transfer product.

The residual bleach at the end of the initial fast recovery was tracked further at longer times [Figs. 5(e) and 5(f)]. The long-term bleach exhibited faster recovery in the presence of EV²⁺. This faster recovery over a long time scale arises from the transfer of trapped electrons to EV²⁺. It should be noted that the long-term bleach that arises from trapped charge carriers is only a small fraction, since nearly 90% of the recovery is seen within 200 ps (see Figs. S4 and S5). The rate constants from the longer time kinetics at 8 μM EV²⁺ were 3.6 × 10⁸ and 1.6 × 10⁸ s⁻¹ for AgInS₂ and AgInS₂–CdS, respectively. Although the contribution of trapped electrons to EV²⁺ is small, it highlights the ability to participate in the interfacial electron transfer process. This observation is also in agreement with the steady state photolysis experiments in which we observe EV^+· formation under sub-bandgap excitations [Fig. 4(f)]. The photophysical and electron transfer properties of AgInS₂ and AgInS₂–CdS are compared in Table I.

Despite the faster bleach recovery seen in the presence of EV²⁺, the residual bleach signal in the 500–700 nm region in the previous measurements prevented us from characterizing the electron transfer product (EV^+·). We employed a nanosecond laser flash photolysis to probe the formation of EV^+· by using 355 nm laser pulse excitation (pulse width 10 ns). The transient absorption spectra recorded with AgInS₂ and AgInS₂–CdS are shown in Figs. 6(a) and 6(c), respectively. The formation of the electron transfer product with its characteristic absorption of around 600 nm could be seen immediately after the laser flash excitation. A fraction of this transient absorption decays quickly (within ∼15 µs) indicating contribution from the back electron transfer. The transient absorption decay profiles [Figs. 6(b) and 6(d)] show the initial decay followed by the stabilization of the absorbance signal.

It is interesting to note that the initial decay component is significantly smaller in AgInS₂–CdS heterostructures, indicating smaller loss of the electron transfer product in the back electron transfer process. The initial decay observed at 600 nm can be attributed to back electron transfer between trapped holes and EV^+· species,

In the case of AgInS₂–CdS, the CdS shell acts as a barrier to suppress the back electron transfer between EV²⁺ and trapped holes (Scheme 2). The fraction of residual bleach [ΔA(30 µs)/ΔA(0 µs)] in the kinetic traces in Figs. 6(c) and 6(d) corresponds to stabilization of electron transfer product. It is interesting to note that the residual bleach fraction observed for AgInS₂–CdS is nearly twice that observed for AgInS₂. This observation parallels the increased quantum yield of EV^+· observed in steady state irradiation experiments with AgInS₂–CdS nanocrystals (29.3%) as compared to pristine AgInS₂ nanocrystals (14.5%).

The results presented here shed light on two important issues. The first one is related to new mechanistic insights into the electron transfer processes of ternary semiconductors. Both bandgap and sub-bandgap excitations induce electron transfer as probed through EV^+· formation under wavelength selective irradiation. The quantum yields obtained with sub-bandgap excitation (λ > 550 nm) are relatively lower compared to broadband white light (λ > 420 nm) irradiation. The transfer of conduction band electrons to EV²⁺ is fast, which occurs with a rate constant of 0.6–2 × 10¹¹ s⁻¹ and constitutes a major part of interfacial electron transfer. The contribution of trapped electrons from the interstitial defects occurs at a slower rate (0.5–2.8 × 10⁸ s⁻¹). The second aspect is the stabilization of electron transfer product. Both steady state irradiation and transient absorption measurements indicate suppression of back electron transfer in AgInS₂–CdS heterostructures. The beneficial role that the CdS layer plays results in enhancement in the electron transfer quantum yield from 14.5% to 29.3%. As evident from the analysis of electron transfer kinetics, the interfacial electron transfer in ternary semiconductors is a kinetically driven process. Careful control of surface interaction of the acceptor molecule and suppression of back electron transfer is needed to maximize the efficiency of photocatalytic reduction. The heterostructures such as AgInS₂–CdS can provide new design strategies to amplify the use of ternary semiconductors in photocatalysis.

The supplementary material includes experimental methods and procedures, actinometry, data of control experiments, and kinetic analysis.

The research described herein is supported by the Division of Chemical Sciences, Geosciences, and Biosciences, Office of Basic Energy Sciences of the U.S. Department of Energy, through Award No. DE-FC02-04ER15533. This contribution is NDRL No. 5354 from the Notre Dame Radiation Laboratory. We also acknowledge the University of Notre Dame Equipment Restoration and Renewal (ERR) program for the purchase of the Spectra Physics laser used for the transient absorption measurements.

The authors have no conflicts to disclose.

A.K. contributed to conceptualization, synthesis, measurements, data analysis, validation, visualization, and writing of the first draft and revision. P.V.K. contributed to conceptualization, method development, funding acquisition, resources, supervision, discussions, and writing.

The data that support the findings of this study are available within the article and its supplementary material.