Exciton Emission and Light induced Charge Separation in colloidal ZnO Nanocrystals

Abstract

Adsorption of organic molecules at ZnO nanoparticle surfaces enables the transfer of energy or charge across resulting organic-inorganic interfaces and, consequently, determines the optoelectronic performance of ZnO based hybrids. We investigated on aqueous colloidal ZnO dispersions adsorption-induced changes with photoluminescence (PL) and electron paramagnetic resonance (EPR) spectroscopy. Citrate and acetate ion adsorption increases or decreases radiative exciton annihilation at hν = 3.3 eV and at room temperature, respectively. Searching for a correspondence between PL emission and the yield of trapped charge carriers originating from exciton separation - using photon energies of hν = 4.6 eV and fluxes of = 10¹⁴ cm^-2 s^-1 for excitation - we found that there is a negligible fraction of paramagnetic products that originate from exciton separation. Upon polychromatic excitation with significantly higher photon fluxes (Ṅ_ph= 10¹⁶ cm^-2·s^-1), ZnO specific shallow defects trap unpaired electrons in citrate and acetate functionalized samples. The adsorption dependent PL intensity changes and the excitation parameter dependent yield of separated charges (EPR) in colloidal ZnO nanoparticles underline that the distribution over the different exciton annihilation channels sensitively depends on interface composition and the intensity of the photoexcitation light.

1. Introduction

Powders, colloids and hybrids containing ZnO nanoparticles as active components are important materials for electronic and optical applications.^[1,2] As an n-type semiconductor, ZnO shows a rich defect chemistry. In part, defect related spectroscopic phenomena are controversially discussed for respective nanomaterials, as different synthesis and processing approaches lead to variations in nature and concentration of electronically relevant defects. Structure and composition of the interface, which are also linked to sample history, have an additional impact on electronic and spectroscopic properties.^[1,3–5]

Apart from defect-related electronic transitions, excitonic emissions play a key role for the optical properties of ZnO nanostructures. Characteristic UV emission bands exist in the near-band-edge energy range and at photon energies around hν = 3.3 eV. The exciton binding energy is as high as 60 meV and associated spectroscopic effects are accessible at room temperature and above. (At cryogenic temperatures, i.e. below 30 K, related PL emission spectra show several near-band-edge emission features corresponding to the free exciton emission at the high energy side and to bound excitons with emissions at slightly lower energies. Single or multiple photon replicas of bound excitons do also appear in this energy range.)^[6]

For semiconducting metal oxides, UV-Vis-NIR absorption spectroscopy, photoluminescence (PL) and electron paramagnetic resonance (EPR) are widely used for the characterization of defects such as conduction band electrons, which compensate for oxygen deficiency (Vis-NIR), point defects with energy levels in the bandgap (PL) or unpaired electrons in local defect structures (EPR).

An experimental challenge in the investigation of optoelectronic ZnO nanomaterials’ properties is related to the question whether specific defects and associated electronic transitions can be tracked with more than one spectroscopic technique. Progress in the complementary characterization of defects may lead to a more coherent and therefore deeper understanding of their electronic and geometric structure.^[7] Different research groups have performed such an integrated characterization approach.^[4,8,9]

The present study combines photoluminescence (PL) and continuous wave electron paramagnetic resonance (cw-EPR) spectroscopy to study the influence of citrate and acetate adsorption on the spectroscopic properties of vapor phase grown ZnO nanocrystals in aqueous colloidal dispersions. While steady-state PL spectroscopy tracks the time-averaged emission from a sample under continuous UV light excitation, its discontinuation also stops the emission process. With cw-EPR spectroscopy, on the other hand, one can monitor the concentration build-up of paramagnetic defects under exposure to excitation light as well as their stability after its discontinuation.^[10]

For nanoparticle powders of TiO₂, as another prominent metal oxide semiconductor, it was demonstrated that surface potential changes, which are either induced by UV light or by the adsorption of donor and acceptor molecules, affect the intensity of characteristic PL emission bands.^[11] Independent cw-EPR spectroscopy measurements provided complementary information about the time-dependent concentration build-up of spectroscopically accessible charging states^[12], which result from a complex reaction network with several electronic transitions occurring within a nanosecond time scale.^[13]

In this study, we performed a complementary photoluminescence and EPR spectroscopy study to identify key factors with regard to the optoelectronic performance of colloidal ZnO nanocrystals. The adsorption of acetates and citrates to the nanoparticle surface, as carried out in this study, affects the intensity of the free exciton emission and the zeta potential of the particle systems at the same time. Based on these adsorbate induced trends we searched for a potential anti-correlation between the two competing processes (Figure 1), namely the radiative exciton annihilation (PL emission), on the one hand, and exciton dissociation that is followed by charge trapping and the emergence of paramagnetic states, on the other hand.

Figure 1.

Different deactivation channels a photoexcited state can undergo inside ZnO nanostructures. Radiative exciton annihilation, which corresponds to photoluminescence emission and is illustrated by the emission curves of arbitrary intensity in the schematic particle drawings (left column), stands in competition i) with energy transfer processes leading to subsequent defect related PL emission at lower photon energies, ii) exciton dissociation and trapping of paramagnetic charges, or iii) alternative non radiative deactivation pathways such as charge transfer across the interface.

2. Results and Discussion

2.1. Exciton emission and the specific adsorption of acetate and citrate ions

The zeta potential of bare metal oxide particle surfaces in aqueous dispersions is governed by the protonation state of the surface hydroxyls. For this reason, it depends on both the surface (i.e. pK_A values of M-OH and M-OH₂ groups) and on the properties of the solution (pH). The here investigated colloidal dispersions are made of vapor phase grown and crystalline ZnO particles with wurtzite structure and an average particle diameter of 14 nm.^[14] The measured zeta potential value of +24 mV is consistent with previous reports on comparable materials.^[15] Ion adsorption modifies the charge distribution at the interface, the zeta potential (Figure 2, left panel) and consequently also the stability of particle dispersions.

Figure 2.

Zeta potential (left panel) and photoluminescence emission curves (right panel) measured for aqueous colloidal dispersions with and without organic salts. Citrate and acetate addition to the aqueous dispersion decreases or increases the zeta potential φ_ζ, respectively. The schematic in the left panel shows the hypothesized distance-dependent development φ_ζ for ZnO nanoparticles in contact with liquid water (blue curve) and after the specific adsorption of acetate (violet curve) or citrate ions (green curve) at the nanoparticle surfaces, respectively. At the same time, photoluminescence emission spectra (right panel) that were acquired on identical colloidal ZnO samples show that citrate and acetate salt addition increases or decreases the intensity of the excitonic emission at hν = 3.3 eV (λ = 380 nm), respectively.

The addition of 30 w/w-% acetate (relative to particle mass) produces a slight shift of the zeta potential toward more positive values. This is different to acetate functionalized ZnO nanoparticles which are synthesized in organic solvents such as ethanol and which typically exhibit negative values. To resolve this discrepancy one needs to recall that the surfaces of oxide nanoparticles are far from being homogeneous, as they include many synthesis related heterogeneities like defects, variable levels of hydroxylation and surface strain, acetate and solvent occupied surface sites, and many more factors the identification of which requires a systematic comparison of samples of different synthetic origin.

On the other hand, citrate addition to vapor phase grown nanoparticles in aqueous dispersion leads to an inversion of the zeta-potential (Figure 2 left). In case of complete adsorption of all the citrate molecules available in solution, a concentration of 3 w/w% (relative to the particle mass) corresponds to approximately 560 molecules per particle. At this concentration we measured a zeta potential value of 0 and, consistently observed an increase of the particle agglomerate size from 70 nm to 2000 nm (Table S3). Citrate addition shifts the zeta potential toward more negative values and thereby stabilizes the colloidal dispersion. As a result, the hydrodynamic agglomerate diameters (d_h = 60 nm) decrease. Preliminary experiments have shown that the zeta potential values show a strong dependence on citrate concentration in the range of < 8 w/w %. Assuming strong irreversible citrate adsorption on monodisperse ZnO nanoparticles this concentration corresponds to ~ 1500 citrate ions per primary particle. Further citrate addition changes the zeta potential only moderately (Table S3). From these preliminary results related to the concentration dependence of the zeta potential, we speculate that citrate ions adsorb onto the ZnO surface forming a monolayer at a citrate concentration of ~ 8 w/w %. This whould be consistent with an estimate, which takes into account that at the average a citrate ion requires a surface area of 0.76 nm2 for adsorption.^[16] In such as case, monolayer formation on the ZnO particle surface (d = 14 nm) is achieved with a citrate concentration of 6 w/w %, or to ~ 1100 citrate ions per primary ZnO particle.

Room temperature PL emission spectra of aqueous ZnO nanoparticle colloids (0.1 mg·ml^-1 or 1.2 ·10¹³ nanoparticles·ml^-1) are plotted in Figure 2. Consistent with literature about RT luminescence properties of ZnO nanowires^[1], colloids^[17], or nanoparticle powders^[7,14] we measured at room temperature a UV emission band at hν = 3.3 eV (λ = 380 nm) and a broad emission feature in the range between 500 and 800 nm. The UV emission is a result of a near-edge excitonic transition, while the broad emission feature is related to interstitial oxygen in this material as described in an earlier study.^[18]

Acetates and citrates adsorb via attachment of their carboxyl groups to surface Zn ions. Apparently, the localization of positive (acetate) or negative partial charges (citrate) at the particle interface alters the electronic structure in such a way that the free exciton emission is reduced by a factor of 0.6 (acetate) or enhanced by a factor of 3 (citrate). We did not observe any correspondence between the adsorbate induced increase/ decrease of the exciton emission, on the one hand, and intensity changes related to the broad emission feature which is centered at λ = 670 nm and keeps its original intensity, on the other. Potential energy transfer processes, where the direct generation of excitons ultimately leads to defect related emission in the visible light range (Figure 1)^[19], are responsible for the adsorbate induced changes in the exciton emission intensity.

2.2. Exciton separation and charge trapping

With Electron Paramagnetic Resonance (EPR) we searched for paramagnetic species that would originate from exciton separation (Figure 1). For this purpose we needed to freeze out the aqueous colloidal dispersions by a temperature quench to 10 K in order to eliminate the dielectric loss of microwave radiation in liquid water which would compromise the EPR measurements. Prior to UV exposure the ZnO samples with a nanoparticle concentration of 0.1 mg·ml^-1 (Figure 3a) shows an isotropic signal at g = 2.003 that is attributed to carbon radicals (details about its orgin are provided below). The related spin concentration is (1 ± 0.5) ·10¹² spins (2·10^-1 spins·particle^-1).

Figure 3.

EPR spectra of colloidal ZnO nanocrystals with different colloid concentrations: (a) 0.1 mg·ml^-1 (1.2·10¹³ particles·ml^-1) and (b) 6 mg·ml^-1 (7.2·10¹⁴ particles·ml^-1), before (black) and after UV exposure (red). The here shown spectra were obtained on ZnO colloidal dispersions in the presence of citrates, but uniformly describe the results obtained for all three types of colloids.

Apart from the gain in sensitivity that is associated with the EPR measurements at cryogenic temperatures, the employed set-up allowed us to perform in-situ UV excitation without changing the sample position inside the cavity.

EPR measurements during continuous light exposure (not shown) reveal the increase in the concentration of the carbon radical (g = 2.003)^[20] up to a factor of 10 at the maximum after 60 minutes of light exposure (Figure 3a, red curve). A second feature of very small intensity emerges at g = 1.961 and originates from electrons in ZnO specific shallow defect states (e_st).^[21]. At this particle concentration and for all three types of colloids measured, the UV induced concentration build up of e_st levels at a concentration which remains within the error margin of the overall method (x ± 50 %, see experimental section).

To improve the signal-to-noise ratio we increased the nanoparticle concentration by a factor of 60 to 6 mg·ml^-1 and obtained a stable aqueous colloid with 7.2·10¹⁴ particles·ml^-1 (black spectrum in Figure 3b). Before UV exposure, the EPR spectra of the three ZnO colloidal samples, i) ZnO nanoparticles in salt-free aqueous dispersion or surface functionalized with ii) acetate or iii) citrate ions showed the EPR signals much more clearly. The isotropic signal at g = 2.003 is attributed to paramagnetic carbon species^[9,20]which remain as pyrolysis products and minority species at the ZnO nanoparticles after gas phase synthesis and thermal processing. The g-value of the signal is indicative of unpaired electrons in dangling bonds with significant sp³ character and lacks noticeable g-value anisotropy or hyperfine structure (hfs).^[20] (Carbon radicals with a g-value below that of the free spin value are associated with electron containing orbitals of pronounced sp² character.) This type of radical is typically observed on metal oxide nanoparticles that were obtained from metal organic chemical vapor synthesis.^[22] Since oxygen treatment at 873 K is required to achieve their perfect elimination, they are still present on the here investigated ZnO nanoparticle systems. We have also previously shown with Auger Electron Spectroscopy^[18] that, after annealing and processing of vapor phase grown ZnO nanocrystals in oxygen, the concentration of residual carbon remains at a level of 5 %. Additional evidence from FT-IR spectroscopy revealed that the carbonaceous species do also include surface carbonates and carboxylic groups.^[14] It has to be noted that the addition of acetate or citrate ions does not show any effect on the EPR signal intensity of the paramagnetic carbon-centered signal in the dark.

The shape of the resonance feature at g = 1.96 (black curve in Figure 3b) points to more than one paramagnetic species in the magnetic field range that is specific to shallow donors in ZnO.^[21] We will describe in the text below, that two types of electron centers e_st contribute to this resonance explaining the signal shape observed. Increase of the colloid concentration from 0.1 to 6 mg·ml^-1 enhances the number of spins by an order of magnitude. Before UV excitation we determined for both colloid concentrations an average spin concentration of $\varphi =\frac{N_{spins,max}}{N_{particle}}$ = (0.5 ± 0.3) spins per particle. UV excitation increases the EPR signal intensities of the carbon center (g = 2.003) to ~ 7·10¹³ spins which corresponds to an enhancement factor of 2 to 3. The same applies for paramagnetic electron centers (e_st) in shallow trap states of ZnO.

Searching for a potential anti-correlation between radiative and non-radiative exciton deactivation channels we plotted the photoluminescence emission yields (Figure 4a) in comparison to the UV generated number of spins (Figures 4b and c) as measured for the different types of colloidal ZnO nanoparticle samples (top row in Figure 4).

Figure 4.

Photoluminescence emission yields (a) and numbers of UV generated spins in ZnO colloids (b and c) for ZnO nanocrystals in different colloidal dispersions (top row) and, therefore, with different surface species (water, acetate, citrate) present. The Figure compares results for the different samples using either monochromatic (a and b) or polychromatic UV excitation (c) with photon fluxes of 10¹⁴ and 10¹⁶ photons·s^-1·cm^-2, respectively.

The typical acquisition time for the steady-state PL measurement (Figure 2, with continuous UV excitation) was eight minutes (480”) and guarantees a reasonable signal-to-noise ratio. The photoluminescence spectra were perfectly reproducible after experiment repetition. Therefore, for the experimental conditions related to the PL measurements we can rule out significant UV-induced depletion effects that would occur on a time scale of minutes. This is different from our EPR observations. With i) the integral number of photons (480·10¹⁴ s^-1·cm^-2) having a photon energy of hν = 4.6 eV and impinging on the colloid inside the cavity of the EPR spectrometer and ii) the maximum number of spins measured after UV excitation (7·10¹³ spins), we determined a relative yield that is averaged over the integral measurement time of the PL experiment. For all three types of colloids described above, this value corresponds to (1.5 ± 0.75) · 10^-3 spins · photon^-1. Apparently, for colloids with a solids content of 0.1 mg·ml^-1 the UV induced formation of paramagnetic trap states is independent from whether acetate or citrate ions are adsorbed to the particle surfaces or not. Therefore, as another major conclusion of this study, there is no observable correspondence between PL emission intensity (Figure 2) and type and concentration of paramagnetic products. Hence, the relative decrease in the PL emission yield observed for ZnO nanoparticles in pure aqueous dispersion with and without acetates (but in the absence of citrates), must originate from non-radiative exciton deactivation pathways (Figure 1) that are different from exciton dissociation and consecutive charge trapping at ZnO specific defects.

2.3. Charge separation in ZnO colloids with polychromatic excitation and higher photon fluxes

In another set of UV excitation experiments we used the unfiltered polychromatic light of the Xe lamp that was solely equipped with a water filter (Figure 4c). For the energy range between 3.4 and 6.2 eV we measured a light power of 30 mW·cm^-2 which corresponds to an averaged flux of 10¹⁶ photons cm^-2·s^-1 with supra-bandgap energy (Supplementary Information). With regard to yield and nature of paramagnetic defects we determined spin concentrations that are higher only by a factor of 10 as compared to the experiments with monochromatic excitation. Moreover, the concentration of paramagnetic carbon species is significantly smaller than in the monochromatic experiment.

Using the polychromatic UV excitation profile we achieved only for the aqueous ZnO colloid an intensity increase for the carbon signal at g = 2.003 (Figure 5b). In addition, an axially symmetric signal related to electrons in ZnO specific shallow donor states (Figures 5b and c) gained in intensity.^[21] The shape of the overall resonance feature at g = 1.96 is clearly influenced by the nature of the adsorbate, i.e. surface acetates (c) or citrates (d). Analysis of the EPR powder spectra (see for example Figure 6) provided the g-tensor components of two shallow donor types, which are designated here as species A and B (Figures 5 and 6, Table 1).

Figure 5.

Paramagnetic species measured on colloidal ZnO nanocrystals before (a) and after (b-d) polychromatic UV irradiation at 10 K. Spectrum (a) was acquired in the dark on ZnO in pure H₂O, while the spectra b, c and d correspond to colloidal dispersions of ZnO in H₂O with acetate and ZnO in H₂O with citrate after 30 minutes of photoexcitation, respectively. The dashed curves plot the results from powder simulation and are discussed along Figure 6 and Table 1.^[23]

Figure 6.

EPR powder signal analysis for paramagnetic species measured in aqueous colloidal ZnO dispersions containing acetate ions (Figure 5c). The powder spectrum is made up from a weak axial contribution assigned to donor species A and an isotropic signal assigned to the donor species B (Table 1).^[23]

Table 1.

g-values of related carbon species C and shallow donor defects (e_st) measured in colloidal ZnO nanocrystals before and after UV excitation at 10 K.

Paramagnetic center	g┴	gII
carbon [Ref. 8,20]	giso = 2.0031
species A in ZnO	1.9596	1.9727
species B in ZnO	giso = 1.9651

The polychromatic excitation experiment is associated with fluxes of supra-bandgap photons that exceed those of the monochromatic PL and EPR experiments (Figure 4 a and b) by a factor of 100. However, the concentration of photogenerated paramagnetic centers is only by a factor of 10 higher. On the one hand, this underlines that the experiments using different emission profiles of excitation light are not comparable with each other, since the overall process starting with exciton formation and ending with the surface chemistry of photogenerated holes and electrons is complex. Most likely, it involves more than one photoinduced reaction step, and different paramagnetic and diamagnetic intermediates and products.^[24] On the other hand, the enhanced signal-to-noise ratio of the obtained EPR spectra (Figures 5 and 6) has enabled us to isolate the different paramagnetic species present (Table 1) and to learn that the relative abundance of the two trapped electron centers e_st depends on the composition of the interface, i.e. on the nature of the ions adsorbed (Figure 7).

Figure 7.

Comparison of integral yield of two types of paramagnetic electron centers (e_st) which results from exciton dissociation (top) with the PL emission yield for colloidal ZnO nanocrystals (bottom).

At the solid-liquid interface the photogenerated charge carriers encounter carboxyl groups that complexate surface Zn²⁺ ions as identical acceptor/ donor species. In principle, the carboxyl groups do participate in the interfacial charge transfer and may either become oxidized or reduced. It is important to note that for all the excitation profiles employed (Supplementary Information) only electrons in paramagnetic shallow trap states (e_st) were observed, while EPR fingerprints indicative of trapped hole centers^[9,25]are essentially absent. Thus, the hole component of the dissociated exciton undergoes oxidation reactions and ultimately leads to diamagnetic product states. In the absence of organic ions, water oxidation leads to an increase of the carbon radicals (Figure 5b), whereas the carboxyl bound surface adsorbates favor the localization of the electronic exciton component at shallow trap states inside the ZnO structure (Figure 5c and d). A quantitative analysis of the EPR powder signals (an example for the powder spectrum analysis is shown in Figure 6, Table 1) indicate the increase of the total concentration of paramagnetic electron centers e_st is in the order aqueous colloid < aqueous colloid with acetate < aqueous colloid with citrate (Figure 4c and top panel in Figure 7).

The citrate induced signal enhancement as observed both by PL and EPR spectroscopy is remarkable. Using PL spectroscopy we track the probability of radiative recombination as one of many different deactivation processes. The polychromatic excitation profile in conjunction with higher photon flux (Ṅ_ph = 10¹⁶ cm^-2·s^-1) shows for the EPR measurements the same trend and leads to two types of paramagnetic defects (species A and B) inside the ZnO lattice with a relative abundance that again does not show the expected anti-correlation with respect to the PL emission yields (Figure 7, bottom). Exciton separation and subsequent charge (electron and/or hole) trapping or interfacial charge transfer belong to the competing deactivation processes, which possibly may result in a quenching of radiative deactivation. However, the observation that citrate addition enhances the probabilities for both radiative recombination (PL) and exciton separation (EPR) points to a multiple impact of this molecule on the fate of the excited states. This impact may have physical origins (e.g. polarization of the interface upon dipole adsorption) as well as chemical origins (resulting e.g. from donor properties of the adsorbed molecule) giving rise to interfacial charge transfer processes.

In composite materials systems the hybrid functionality emerges from the distinct properties of the inorganic/ organic interface. This study underlines the sensitive dependence of non-radiative transitions on changes at the nanoparticle interface, which naturally occur during formation of organic-inorganic hybrids and further materials processing.^[3,24] While radiative exciton annihilation or energy transfer (Figure 1) is the key process in light emission, charge separation is the essential mechanism in solar cells and photocatalysis.^[24] The experiments performed in this study accounted for the comparability of colloid concentration and UV excitation parameters during the intrinsically different PL and EPR spectroscopic measurements. On this basis, we can state the absence of a direct connection between the probability for the radiative exciton annihilation and the excitation dependent emergence or depletion of a paramagnetic defect sites. On the other hand, the here reported findings cannot provide additional new insights into origin and properties of the broad PL feature in the visible light region (Figure 2). For vapor phase grown ZnO nanocrystals and with support from Auger Electron Spectroscopy measurements we previously related this feature to interstitial oxygen ions which become depleted by vacuum annealing.^[18] Different from the excitonic band at hν = 3.3 eV adsorption induced zeta potential changes of the aqueous colloid do not show any effect on the intensity and band shape of the visible light emission. Moreover, for room temperature measurements and in terms of energy transfer we can exclude a correspondence with the annihilation exciton process at 3.3 eV at room temperature.

Finally we want to emphasize that charge separation and localization of electrons in shallow trap states can be considered as a self-doping process. The integral yield of trapped electron centers observed in the course of the experiments of Figures 6 and 7 is roughly of the order of particles excited. Assuming a monodisperse particle size distribution and a particle diameter of 14 nm, which is consistent with the average particle/ crystallite size as determined with TEM and XRD, this number of spins corresponds to an electron doping concentration of n = 7·10¹⁷ cm^-3 or 1.7·10^-5/ O^2-.

3. Conclusions

Citrate adsorption on ZnO nanocrystals in aqueous colloidal dispersion not only enhances the PL emission associated with the free exciton annihilation, it also favors trapping of the exciton derived electrons at shallow trap states inside the ZnO structure. Using comparable experimental conditions for photoluminescence emission and electron paramagnetic resonance we did not observe any type of anti-correlation between the PL emission yield and the yield of paramagnetic products originating from exciton separation. In light of the multitude of pathways the photoexcited state can relax (Figure 1)^[26], the here presented observations together with discrepant results in literature illustrate that the following main sources are effective with regard to variations in the PL and/ or photoexcitation properties of colloidal ZnO nanoparticles.

• i)Nanoparticles from different synthetic provenience show variations in purity, crystallinity as well as nature and concentration of spectroscopically relevant point defects.
• ii)Process related changes of the chemical environment (addition or exchange of solvents, pH changes, etc.) surrounding the nanoparticle ensemble can lead to significant adsorption induced variations in PL emission. In addition, they can alter size and morphology of the ZnO nanoparticles as changes in the solvent composition may induce dissolution-recrystallization processes in ZnO nanoparticles.^[14]
• iii)Different light sources are used for photoexcitation experiments to probe PL emission and/ or to measure the formation of paramagnetic product states. Associated variations in photon energy (energy value and excitation profile) and photon flux determine nature and intensity of respective spectroscopic features.

Similar to the field of heterogeneous photocatalysts^[24,27], uniform guidelines for the analysis and characterization of pure and modified ZnO nanostructures are needed for a robust characterization of ZnO based hybrids developed for applications in optoelectronic devices. A broader materials community will benefit from this information.

Experimental Section

ZnO nanoparticles were synthetized from zinc acetate dihydrate (Sigma-Aldrich, ≥ 99.0 %) precursor via chemical vapor synthesis. Details about synthesis conditions and the subsequent surface purification procedure can be found in references ^[18] and ^[14]. Aqueous dispersions of phase pure wurtzite ZnO nanocrystals with an average particle size of 14 nm, as determined by TEM analysis ^[14], were prepared with a particle concentration of 0.1 mg·ml^-1 for photoluminescence and of 0.1 mg·ml^-1 and 6 mg·ml^-1 for EPR experiments. For the generation of colloidal dispersions, the dry nanoparticles were brought into contact with high purity water (ρ = 18 MΩ·cm). Trisodium citrate dihydrate (Merck, ≥ 99.0 %) and zinc acetate dihydrate (Sigma-Aldrich, ≥ 99.0 %) at concentrations of 50 and 30 w/w % relative to the particle weight, respectively, were added as aqueous salt solutions to stabilize the colloidal dispersion. The pH values recorded for the dispersions were all in the range between 7.0 and 8.1 slightly increasing with the concentration of citrate added. The solid phase was dispersed by direct sonication using an ultrasonic finger (Hielscher Ultrasonics UP200St, Ti-sonotrode Ø 2 mm, amplitude 25 %, 15 min). During the ultrasonic treatment the dispersion was stirred continuously and cooled by an ice bath.

The zeta potential values of the dispersions were determined via electrophoretic light scattering using a Zetasizer Nano ZSP ZEN5600 (Malvern Instruments). Scattering was measured using a red laser light (λ = 632.8 nm). Electrophoresis inside the sample cell was induced by 40 V applied voltage. The Smoluchowski approximation was chosen for the zeta potential calculations, assuming an extended solvation shell of particles in the aqueous dispersion.

The photoluminescence emission spectra were recorded with a FLS980 spectrometer system (Edinburgh Instruments) equipped with a 450 W Xe-arc lamp. For sample excitation a mono-chromatic UV light beam with a wavelength of λ_ex = 270 nm was used. After transmission through the entrance slit (with a corresponding spectral width of Δλ = 10 nm) a measured light power of 0.102 mW reaches the sample. Correspondingly, the number of photons impinging on the colloidal samples is 1.4·10¹⁴ cm^-2·s^-1.

EPR spectroscopy was performed with a Bruker EMXplus-10/12/P/L X-band spectrometer equipped with a waveguide Cryogen-Free System from Oxford Instruments. The spectra were recorded at 10 K with a field modulation frequency of 100 kHz, a modulation amplitude of 0.1 mT and a microwave frequency of 9.30 GHz. Spin quantification was carried out by number of spins calculations with the Xenon software from Bruker.

Sample illumination during EPR measurements was carried out with a 300 W Xe-arc lamp. The lamp was equipped with a water and an interference filter (280FS25-50 (P004-01)) to set the wavelength of the excitation light to 272 nm (hν = 4.6 eV). The measured light power density of 0.464 mW·cm^-2 corresponds to a photon flux of photons of 6.4·10¹⁴ cm^-2·s^-1 reaching the 1 cm² front face area of the sample. This value is comparable to the excitation conditions related to PL experiments. Complementary EPR experiments were performed with polychromatic excitation light, where a 395 nm longpass filter was used to approximately determine the intensity of the excitation light in the range of energies that are larger than the ZnO bandgap. With a light power density of 30 mW·cm^-2 the number of photons reaching the resonator’s 1 cm² front face area in front of the sample is 5.6·10¹⁶ s⁻¹. Evolution and concentration build-up of paramagnetic species were tracked with EPR spectroscopy both, during UV light exposure and after its discontinuation. For the experiments described along Figures 3 and 4 the error margin related to the overall method corresponds to x ± 50 %. This includes the reproducible preparation of the colloidal specimen, sample distribution inside the cavity, alignment of the UV excitation source and the EPR signal integration and spin counting.

EPR spectrum analysis was carried out by simulation using the computational and Matlab-based package/software EasySpin.^[23] Details of the simulation parameters can be found in the Supporting Information.