The Effects of Native Oxides on the Selective Photo-Reduction Property of Cu Nanoparticles Induced by the Localized Surface Plasmon

Copper nanoparticles (CuNPs) in air are easily oxidized into CuNPs@Cu2O core–shell structure. The localized surface plasmon resonance (LSPR) of CuNPs can be damped seriously by the Cu2O shell. The effect of this native Cu2O layer on the LSPR-induced selective photocatalysis properties of CuNPs is often ignored. In this paper, this effect was explored by comparison of the evolution of the photo-reduction efficiency (PRE) with the nanoparticle size (d) of the as-deposited CuNPs to that of AgNPs via monitoring the transformation from 4-nitrothiophenol (PNTP) to 4,4 trdimercaptoazobenzene (DMAB) using surface enhanced Raman scattering. It was found that the PRE of the as-deposited CuNPs increases first and then decreases with increasing d, much different from that of AgNPs: increases with increasing d. Based on the transmission electron microscope images, X-ray photoelectron spectroscopy, and by monitoring the PRE stability in air of the as-deposited CuNPs, the mechanisms of the PRE evolution with d and the laser illumination time (t) were suggested. Cu2O plays a dominated role when the nanoparticles’ sizes were small. Both Cu2O and CuNPs play roles when the nanoparticles’ sizes were large, which brought in the low PRE of the as-deposited CuNPs.


Introduction
Non-noble Cu nanoparticles (CuNPs) have localized surface plasmon resonance (LSPR) property comparable to that of AuNPs in intensity and wavelength [1,2] and thus have a promising prospect of large-scale applications in biology, catalysis, sensing, and solar energy utilization [3,4].However, CuNPs in air are easily oxidized into CuNPs@ Cu 2 O core-shell structure [5,6]. What's more, due to the non-limiting oxidation [7], the CuNPs can be completely oxidized into Cu 2 O nanoparticles [8], and eventually hollow Cu 2 O nanoparticles are formed [9]. The LSPR of CuNPs can be damped severely by the formed Cu 2 O layer. For example, the LSPR of CuNPs fabricated via sputtering in vacuum was nearly disppeared within 24 h [10]. How this native Cu 2 O affects the LSPR-induced photocatalytic properties of CuNPs is rarely reported [11], but it is of significance for understanding the mechanism of photocatalysis of CuNPs.
Surface-enhanced Raman spectroscopy (SERS) is a powful tool with high sensitivity in monitoring LSPRinduced selective photocatalysis processes [12], such as the transformations from 4-nitrothiophenol (PNTP) to 4,4′-dimercaptoazobenzene (DMAB) [13][14][15]. AgNPs, stable in dry and clean air, possessing strong LSPR, can induce selective photocatalysis effectively and stably. The LSPRinduced photocatalytic properties of AgNPs were demonstrated to increase with increasing mean nanoparticle size 1 3 (d) [16,17]. This can be used as a reference. In this paper, the selective photocatalytic property of native Cu 2 O covered CuNPs (CuNPs@Cu 2 O) was studied by using SERS monitoring the photo-reduction from PNTP to DMAB. It was found that the CuNPs@Cu 2 O showed much different sizedependent photo-reduction property from that of AgNPs. Based on the data of transmission electron microscope (TEM), X-ray photoelectron spectroscopy (XPS), and by monitoring the photo-reduction efficiency (PRE) changing with the time exposed to air, the meachnisms of the photoreduction property of the CuNPs@Cu 2 O were explored.
CuNPs and AgNPs were normally fabricated via chemical solution methods where precursors and surfactants were used [18]. The surfaces of the CuNPs inevitably contained residual organic molecules, which affected the hot carriers tranferring to the target molecules. Thermal evaporation of high pure Ag or Cu wire in a vacuum chamber is a simple method to fabricate clean AgNPs and CuNPs [19]. Such nanoparticles ensure the target molecules to be directly adsorbed on the real surface of the nanoparticles. Under this situation, the hot electrons and holes produced due to the decay of the LSPRs of AgNPs and CuNPs can be captured by the target molecules effectively; thus, selective photo-transformation reaction can take place with high efficiency.

Experimental
High purity Ag and Cu wires (99.999%) were evaporated onto glass substrates respectively in a vacuum chamber which was pumped by a turbo molecular pump backed by a rotary pump. The background pressure was 6.0 × 10 −4 Pa. The deposited Ag and Cu thin films were composed of CuNPs and AgNPs due to the island growth mode, and thus in this paper, the Ag and Cu thin films are denoted as xnmAg and ynmCu, or xnmAgNPs and ynmCuNPs, where x and y are the mean mass thickness values recorded by the quartz crystal microbalance (QCM). The mean deposition rates of Cu and Ag were both about 0.3 Å/s. Some of the as-deposited ynmCuNPs were annealed in a fused tube furnace at 300 °C for 7 min in nitrogen ambient of a flux of 1600 sccm to decompose the Cu 2 O layer, denoted as ynmCuNPs-AN.

Characterizations
The morphology and structural characterizations of the asdeposited AgNPs, CuNPs, and the annealed CuNPs were carried out by transmission electron microscopy (TEM) using a JEM-2100F (JEOL). The accelerating voltage was 200 kV. The colors of the samples due to plasmonic effect were characterized by photo-images recorded by cell phone, as shown in Fig. S1. The composition was characterized by X-ray photoelectron spectroscopy (XPS) using Thermo ESCALAB 250Xi under 5 × 10 −8 Pa with a monochromatic Al Kα X-ray source (1486.6 eV). Charge neutralization was deemed to have been fully achieved by monitoring the C 1 s signal for adventitious carbon. The samples were not sputtered prior to analysis. The LSPRs of CuNPs and AgNPs were characterized using UV-Vis absorption spectroscopy (Shmadzu-2500).

TEM
The TEM images of the xnmAgNPs (x = 0.5, 1.0) and the ynmCuNPs (y = 0.5, 1.0) are shown in Fig. 1a and b and Fig. 1d and e, respectively. Most AgNPs were round and isolated. The mean nanoparticles size (d) increased with the mean mass thickness (x) of Ag thin film. The nanoparticle histograms are shown in Fig. S2 and the mean nanoparticles sizes and the densities are shown in Table S1. The mean nanoparticles sizes were about 3.8 and 6.7 nm for the 0.5 nm AgNPs and the 1.0 nm AgNPs, and the corresponding nanoparticles densities were about 0.8 and 1.1 per 100 nm 2 . The nanoparticles' shapes of the 0.5 nm CuNPs in Fig. 1d are irregular as reported by Popok et al. [20]. In fact, both the nanoparticles of the 0.5 nm CuNPs and the 1.0 nm CuNPs were round as indicated by the insets in Fig. 1d and e. The mean nanoparticles sizes were estimated to be about 3.4 and 5.4 nm, but the nanoparticles densities were about 1.7 and 2.3 per 100 nm 2 , twice those of the xnmAgNPs with same mean thickness, indicating that CuNPs are flatter than AgNPs. The TEM images of the 0.5 nm CuNPs and the 1.0 nm CuNPs after annealing in nitrogen ambient of 1600 sccm for 7 min are shown in Fig. 1g and h. Both the mean nanoparticle sizes, 4.2 and 6.8 nm, and the nanoparticles densities, 0.9 and 1.2 per 100 nm 2 , were similar to these of the xnmAgNPs with same mean thicknesses. The HRTEM image of CuNPs of the 1.0 nm CuNPs is shown in Fig. S3.

Absorption Spectra
The absorption spectra of the xnmAgNPs and the ynmCuNPs (x, y = 0.2, 0.5, 1.0, and 2.0) are shown in Fig. 1c and f, respectively. All the xnmAgNPs possess obvious LSPR. With the increasing of x; the LSPR intensity increases and the peak shifts to long wavelength. But the absorption spectra of the ynmCuNPs are complex. The 0.2 nm CuNPs show very weak absorbance in 360-450 nm. The other three show similar spectra, with large dip at 575 nm. According to our recent study [10], the absorption spectra were composed of three components, the absorbance of Cu 2 O (wavelength shorter than 575 nm), the interband transition of electrons from 3 s to 4 s (wavelength shorter than 575 nm), and the LSPR of CuNPs@ Cu 2 O extending from 400 nm to wavelength over 800 nm due to the coupling of the LSPR of CuNPs@Cu 2 O.

3
The 0.2 nm CuNPs did not show noticeable absorbance of CuNPs@Cu 2 O, which indicates that the 0.2 nm CuNPs might have been oxidized completely into Cu 2 O or its LSPR was damped completely by the native Cu 2 O on its surface. For the other three, y = 0.5, 1.0, and 2.0, illuminating with 532-nm laser, both the excitons of Cu 2 O and the LSPR of CuNPs@Cu 2 O could be excited.
The absorption spectra of the annealed ynmCuNPs are shown in Fig. 1i. All samples show strong LSPR. As y increases from 0.2 to 1.0, the peak of the LSPR shifts from 620 to 650 nm, and its intensity increases gradually due to the increased nanoparticles size. The absorbance in 360-575 nm increases simultaneously due to partial oxidation of Cu atoms and the interband transition.

XPS
XPS spectra of the xnmAgNPs, ynmCuNPs (x, y = 0.2, 0.5, 1.0, and 2.0), and the 0.5 nm CuNPs, 1.0 nm CuNPs after annealed in nitrogen ambient of flux 1600 sccm at 300 °C for 7 min were measured. All date was calibrated by C1s with binding energy of 284.6 eV. The Ag3d 5/2 lines were fitted with two components of binding energies 367.8 (silver in Ag x O) and 368.3 eV (metallic silver) [21]. The Cu2p 3/2 lines were fitted with four components of binding energies 932.5, 932.7, 933.7, and 934.8 eV, which were corresponding to the copper atoms in Cu 2 O, Cu, CuO, and Cu-OH or Cu(OH) 2 [22], respectively. The spectral fitting details are shown in Fig. S4a-d for the xnmAgNPs, and Fig. S4e-h for the ynmCuNPs. The peak energy, the full width at half maximum (FWHM), and the content of each component are listed in Table 1.
With the increasing of the mean thickness from 0.2 to 2.0 nm, the content of metallic Ag was increased from 78.2 to 92.9%, while that of metallic Cu was increased from 14.6 to 42.7%, indicating that CuNPs were easy to be oxidized, and AgNPs were much more stable than CuNPs. When y = 0.2, the content of metallic copper was less than 20%, indicating that small size CuNPs were oxidized seriously. The 0.2 nm CuNPs contained metallic Cu about 14.6% but did not show LSPR (Fig. 1f). It was reported [23] that CuNPs with a size smaller than 2 nm did not own LSPR due to insufficient electrons. Therefore, in our case, the size of the 0.2 nm CuNPs should be less than 2 nm. The LSPR was too weak to be observed and thus the observed absorbance should be from Cu 2 O.
The 0.5 nm CuNPs and the 1.0 nm CuNPs contained metallic Cu of about 33.4 and 42.0% respectively. Both possessed LSPR in 575-800 nm. After annealing in nitrogen ambient, the contents of metallic Cu were increased to 39.7 and 48.8% respectively, indicating that decomposition of Cu 2 O took place during annealing.

Selective Photo-Reduction from PNTP to DMAB
The photocatalytic properties of the samples were evaluated by photo-reduction of PNTP (~ 95%, J&K Scientific) to DMAB by SERS. The molecular structures of PNTP and DMAB are shown in Fig. 2a and b [25,26]. The S atoms were absorbed on the surface of CuNPs as shown in Fig. S5a. The reaction from PNTP to DMAB are expressed in Fig. 2c [25] and Fig. S5. The Raman spectra of PNTP and DMAB are shown in Fig. 2d. The Raman peaks at 1335 cm −1 were chosen as the characteristic peaks of PNTP and these at 1142, 1388, and 1432 cm −1 were the characteristic peaks of DMAB. The photo-reduction efficiencies (PREs) were characterized by the intensity ratio of the peak at 1432 cm −1 of DMAB to that at 1335 cm −1 of PNTP (I 1432 /I 1335 ) [27,28].

The Selective Photo-Reduction Property of AgNPs
The SERS spectra of PNTP on the xnmAgNPs (x = 0.1, 0.2, 0.5, 1.0, and 2.0) were taken. The evolution of the SERS spectrum with the laser illumination time (t) for the 0.5 nm AgNPs is shown in Fig. 3a. With the laser illumination, the intensities of the characteristic peaks of DMAB at 1142, 1388, and 1432 cm −1 increase gradually while that of the characteristic peak of PNTP at 1335 cm −1 decreases correspondingly. The results demonstrate the photo-reduction from PNTP to DMAB. The PREs, characterized by I 1432 /I 1335 , were calculated and are shown on the right side of each spectrum. The PREs of other samples were calculated in the same way. The evolutions of PRE with t (PRE-t) of all samples are shown in Fig. 3b. All the samples show the same trends that the PRE increases with the laser illumination (t), but the PREs are very low during the initial 80 s. We denote this style of PRE-t as type I. For the same t, the PRE increases as x increases.

The Selective Photo-Reduction Property of CuNPs
The SERS spectra of PNTP on the ynmCuNPs (y = 0.1, 0.2, 0.5, 1.0, and 2.0) were collected. The evolution of the SERS spectrum with t for the 1.0 nm CuNPs is shown in Fig. 3c.
With the laser illumination, the intensity of the characteristic peak of DMAB at 1432 cm −1 increases gradually while that of PNTP at 1335 cm −1 decreases correspondingly, demonstrating the photo-reduction from PNTP to DMAB. The PREs of all samples were calculated and are shown in Fig. 3d. However, it is much different from that of AgNPs. Three obvious features can be noticed. First, the PRE increases as the laser illumination, which is same as that of the xnmAgNPs. Second, during the first 80 s, the samples with thicknesses smaller than 0.5 nm show relatively larger PREs than the others. We denote this style PRE-t as type II, which is a parabolic function. Third, for the same laser illumination time, as the thickness (y) of the Cu thin film increasing, the PRE increases first and then decreases. The 0.2 nm CuNPs, whose Raman spectrum evolution with the laser illumination time is shown in Fig. 3e, has the highest PRE. This PRE-y relation is much different from that the PRE-x of AgNPs.
The PRE-t of the ynmCuNPs just after annealing (denoted as ynmCuNPs-AN) is shown in Fig. 3f, which keeps as type I. The PRE increases with the mean thickness (y) of the deposited Cu thin film. Both the relations of the PRE-t and the PRE-y are similar to these of the xnmAgNPs.
Since the nanoparticle sizes (d) increase with the increasing of the mean thicknesses of Ag (x) and Cu (y) thin films (Fig. 1), both the PRE-y of CuNPs and the PRE-x of AgNPs can be substituted by PRE-d. Therefore, only the as-deposited CuNPs show different PRE-t and PRE-d relations from that of xnmAgNPs and ynmCuNPs-AN. This might be caused by the thick native oxides shell formed on the surface of the as-deposited CuNPs.

Energy Levels
In order to explain the mechanisms, energy levels of CuNPs, AgNPs, PNTP, and DMAB were determined. The energy levels of the LUMO and HOMO of PNTP and DMAB were calculated with density functional theory (DFT) [29], mPW91PW91 functional [30], and 6-31G(d) basis set, using Gaussian 09 software [31], and are shown in Fig. 4a.
The Fermi levels of AgNPs and CuNPs, the conduction bands (CBs) and valence bands (VBs) of Ag 2 O and Cu 2 O determined according to refs [32,33] and the references there, are shown in Fig. 4a. The Fermi level of AgNPs is about − 4.7 eV, and the CB and VB of Ag 2 O are at − 4.6 and -5.9 eV respectively. Their Fermi energy alignments can be described by Fig. 4b. The Fermi level of CuNPs is about − 4.9 eV; the CB and VB of Cu 2 O are at − 4.22 and -6.24 eV, respectively. The Fermi energy levels alignment between CuNPs and Cu 2 O can be described in Fig. 4c. According to Fig. 4b and c, both the energy bands of Ag 2 O and Cu 2 O bend down when combined with AgNPs and CuNPs, respectively.

The Mechanisms of the Selective Photo-Reduction Property of AgNPs
According to Fig. 4a, the energy difference between the LUMO of PNTP (− 2.29 eV) and the Fermi level of Ag (− 4.7 eV) is about 2.41 eV (514 nm). Since the excitation wavelength 532 nm matches the energy difference 514 nm well, the hot electrons produced in AgNPs excited by 532 nm have a large chance to enter into the unoccupied orbit of PNTP, inducing the reduction of PNTP to DMAB.
Since the LSPR intensity of the xnmAgNPs increases with x increasing (Fig. 1c), the number of hot electrons should increase with x increasing when the AgNPs were excited by a laser of 532 nm. This definitely contributes to the PRE from PNTP to DMAB. This trend coincides with the evolution of the PRE with x (Fig. 3b); thus, the mechanism responsible for the photo-reduction from PNTP to DMAB was mainly due to the LSPR of AgNPs [34].
We explain the low PREs in Fig. 3b during the initial 80-s illumination. We attribute this mainly to the thin Ag 2 O layers on the surfaces of AgNPs. According to the XPS results (Table 1), when AgNPs were exposed to air,they were covered by a thin layer of Ag 2 O (denoted as AgNPs@Ag 2 O). Excited by visible light, both the LSPR of the Ag core and the excitons of Ag 2 O were excited. Due to the downward band bending of Ag 2 O (Fig. 4b), the hot electrons drop back to Ag core and the holes were left on Ag 2 O [35]. Most hot electrons cannot reach the PNTP on the surfaces of AgNPs@ Ag 2 O; therefore, the PREs were very low during the first 80 s. However, as laser illumination time increasing, Ag 2 O was reduced into Ag by some of the hot electrons reaching Ag 2 O as what was reported that Cu 2 O was reduced to Cu by hot electrons [36]. Therefore, the PREs increase with t increasing when t was longer than 80 s. On the other hand, the defects inside the nanoparticles and on the surface capture hot electrons, which brought in the low PRE during the initial 80-s illumination.

The Mechanisms of the Selective Photo-Reduction Property of CuNPs
The selective photo-reduction property of the as-deposited CuNPs cannot be explained only by LSPR of CuNPs. According to Fig. 1f, when the ynmCuNPs were excited by a laser of wavelength of 532 nm, the absorbance at 532 nm increases with y increasing, but as shown in Fig. 3d, the PRE increases first and then decreases. Except for the LSPR of CuNPs, another factor should take a role either. We will explain this based on the data of XPS.
The fitting of Cu2p demonstrates that most of the 0.2 nm CuNPs has been oxidized into Cu 2 O. The size of the Cu core was less than 2 nm and thus did not possess LSPR. Ref [37] reported that the reaction from PNTP to DMAB could be trigged by electrons from excitons of Cu 2 O; therefore, the PRE of the 0.2 nm CuNPs was caused by Cu 2 O. Since Cu 2 O is stable in air at room temperature, the absorption of the 0.2 nm CuNPs should not change with the time the sample exposed to air and thus, the PRE-t should be stable. In order to demonstrate, the stabilities of the absorbance and the PRE-t of the as-deposited 0.2 nm CuNPs were monitored for 24 h. The results are shown in Fig. 5a-f. Both the PRE-t and the absorbance of the 0.2 nm CuNPs ( Fig. 5a and d) are stable, and the PRE-t keeps as type II without any change. Therefore, the PREs of the ynmCuNPs when y = 0.1 and 0.2 are can be considered due to Cu 2 O.
The Cu2p lines of XPS spectra of the 1.0 nm CuNPs after exposed to air for 24 h were fitted and the results are shown in Table 1. The 1.0 nm CuNPs after exposed to air for 24 h still contains Cu 0 of about 26.2%. These Cu 0 should own LSPR. The stabilities of the absorbance and the PRE-t of the 1.0 nm CuNPs for 24 h are shown in Fig. 5c and f. The PRE-t keeps as type I within 24 h (Fig. 5c). The LSPR of the 1.0 nm CuNPs between 575 and 800 nm decreases gradually, and 24 h later, the LSPR still exists (Fig. 5f). Therefore, the PRE can be considered mainly caused by the LSPR of CuNPs.
The 0.5 nm CuNPs, after exposed to air for 24 h, the content of metallic Cu was reduced to about 14.5%. According to the case of the 0.2 nm CuNPs, when the content of metallic Cu was 14.6%, the Cu core size was smaller than 2.0 nm, thus it can be inferred that the Cu core size of the 0.5 nm CuNPs after exposed to air for 24 h be reduced to a size less than 2 nm, and the LSPR should be nearly disappeared. As shown Fig. 5 The PRE-ts (a, b, c) and the absorbance (d, e, f) of ynmCuNPs (y = 0.2, 0.5, and 1.0) after exposed to air for 0, 2, and 24 h 1 3 in Fig. 5b and e, the absorbance between 575 and 800 nm (Fig. 5e) decreases gradually. Twenty-four hours later, it was near 0. The color of the 0.5 nm CuNPs changes from light cyan to light yellow after 24 h (Fig. S7). Accompanied with this decrease, the PRE-t was changed from type I (∆t = 0, 2 h) to type II (∆t = 24 h) (Fig. 5b).
In summary, when CuNPs were exposed to air, they were oxidized into CuNPs@Cu 2 O. There existed two kinds of mechanisms responsible for the evolutions of the PRE with y. When y was smaller than 0.5, the CuNPs were seriously oxidized. No LSPR was excited due to insufficient electrons in CuNPs of size smaller than 2 nm [23]. Under this situation, the electrons causing the PNTP reduction should be the excitons of Cu 2 O; the PRE-t was a parabolic function, as type II. When y was not smaller than 0.5, both LSPR and Cu 2 O plays a role. Due to the downward band-bending, the PRE-t is type I.
We try to discuss why the PRE of the as-deposited ynmCuNPs decreases as the thickness of the Cu thin film increases from 0.5 to 2.0 nm (Fig. 3d). According to Fig. 1f, the thicker the deposited Cu thin film was, the stronger the absorbance of Cu 2 O (wavelength shorter than 575 nm) was, which means the thicker the Cu 2 O was. As shown in Fig. 6, fewer hot electrons could tunnel thicker Cu 2 O layer to reach the PNTP on the sample's surface. Therefore the thicker the deposited Cu thin film is, the smaller the PRE is (Fig. 3d).
We further demonstrate the mechanisms we proposed by annealing CuNPs in nitrogen ambient and monitoring the PRE-t evolution. As shown in Fig. 7, the PRE-t of the annealed 0.5nm CuNPs was kept as type I till exposed to air for 36 h, then it was changed to type II. During annealing at 300 °C for 7 min in nitrogen ambient, a part of Cu 2 O was decomposed (Table 1). At the same time, agglomeration of CuNPs took place ( Fig. 1g and  h), which enhanced the intensity of LSPR. Furthermore, annealing eliminated the defects inside the CuNPs, which can improve the anti-oxidation ability. Therefore, the effect of the native Cu 2 O was weak and the LSPR of CuNPs became the dominant one. The selective photoreduction property of the annealed CuNPs (the black line in Fig. 7) shows similar size dependent relation with that of AgNPs (Fig. 3f). After exposed to air for more than 48 hours, the PRE-t (the origin and green lines in Fig. 7) was similar to that of the as-deposited CuNPs (Fig. 3d, type-II).

Conclusion
When CuNPs were exposed to air, they were oxidized into CuNPs@Cu 2 O. If the CuNPs@Cu 2 O showed absorbance in 575-800 nm, which was from LSPR of Cu core with size larger than 2 nm; then, the PRE-t relation should be type I; if the absorbance in 575-800 nm was 0 which indicates that the core size was smaller than 2 nm, the absorbance was mainly from Cu 2 O. The PRE-t relation should be type II. For the as-deposited CuNPs, with the core size increasing, the role of LSPR was enhanced, which brings to the distinct relation of PRE-d, changing from type II to type I. (c) 1.0nmCuNPs Cu 0 -42.0% Fig. 6 The sketch of the configurations of the ynmCuNPs Fig. 7 The PRE-t evolution with the time exposed to air of the 0.5 nm CuNPs after annealed in nitrogen ambient