Synthesis and optical properties of phosphorus doped ZnO: X-ray absorption, X-ray emission, and X-ray excited optical luminescence studies

Sample preparation

In hydrothermal synthesis, both undoped and P-doped ZnO nanostructures were grown on a Si substrate. Initially, a 10 nm thick film of ZnO was deposited on n-type Si (100) wafer by magnetron sputtering. Then the substrate was annealed under the ambient atmosphere at 400 °C for 1 h to crystalize the ZnO thin film that would act as a seed layer for the growth of ZnO nanostructures during the hydrothermal process. Ten millimolar zinc acetate dihydrate (≥98 %, Sigma-Aldrich), 10 mM hexamethylenetetramine (≥99.0 %, Sigma-Aldrich) and 1.5 mM NH₄H₂PO₄ (≥98 %, Sigma-Aldrich) were dissolved in deionized (DI) water to produce a 20 mL solution. Then 6.3 mL solution was transferred to a Teflon-lined stainless-steel autoclave of 10.5 mL capacity. A piece of ZnO thin film/Si wafer (5 × 5 mm) was put into the solution. The tank was heated in the oven at 100 °C for 24 h. After the reaction, the sample was washed with DI water and dried under ambient atmosphere at 60 °C for 4 h [13]. In the end, the sample was annealed in the oven at 400 °C for 2 h. For comparison, ZnO was also grown following the same procedures and conditions in the synthesis of P-doped ZnO, without the dopant NH₄H₂PO₄ [13].

In CVD synthesis, both undoped and P-doped ZnO nanostructures were also grown on the Si wafer. Initially, a 10 nm thick Au layer was deposited on n-type Si (100) wafer by sputtering. The Au layer would act as a catalyst layer for the growth of ZnO via the vapor-liquid-solid (VLS) mechanism [14]. The following steps were taken: First, 1 g of Zn powder (99.99 %, Alfa Aesar) and 0.03 g of P₂O₅ powder were mixed together and placed in a crucible boat. Second, the boat and the Si wafer were placed in a tube furnace. Eighty standard cubic centimeters per minute of Ar was used as carrier gas. During the growth, the furnace temperature was increased to 550 °C at a heating rate of 10 °C/min and was kept at 550 °C for 1 h [14], [15], [16]. ZnO without doping (P₂O₅) was also grown following the same procedures, as in the synthesis of P-doped ZnO. Figure S1 shows the schematic of the CVD synthesis. The products from the hydrothermal and CVD technique are henceforth denoted ZnO: n_hydro, ZnO: p_hydro, ZnO: n_CVD, and ZnO: p_CVD, respectively.

Characterization

The structure and morphology of the ZnO thus obtained were characterized by XRD (Rigaku, Cu K-alpha radiation) and SEM (LEO 1530XB SEM and LEO1540 XB FIB/SEM with EDX), respectively. XANES, XES and XEOL were used to reveal the electronic structure and optical properties of both undoped and P-doped ZnO nanostructures. XANES and XEOL measurements at O K-edge, Zn L_3,2-edge and P K-edge were conducted at the high-resolution Spherical Grating Monochromator (SGM) beamline [17] and the Soft X-ray Microcharacterization Beamline (SXRMB) [18] at Canadian Light Sources (CLS), respectively. XANES and XEOL measurements at Zn K-edge were performed at the sector 20-BM-B at Advanced Photon Source (APS) [19] and the X-ray Nanoprobe Beamline at Taiwan Photon Source (TPS) [20]. XES and XEOL at O K-edge and Zn L_3,2-edge were measured at Resonant Elastic and Inelastic X-ray Scattering (REIXS) beamline [21] at the CLS.

DFT calculation

DOS of ZnO (Wurtzite) were calculated using the Wien 2k code which is based on the DFT [22], [23], [24]. It has utilized the linearized augmented plane wave (LAPW) method [25] which is among the most accurate methods for performing electronic structure calculations for crystals. We use the generalized gradient approximation (GGA) [26] to calculate self-consistently the ground state energies for ZnO. The program starts with the crystal structures of the model ZnO with wurtzite symmetry and calculates the energy levels by iterations until the energies has been reached within 0.0001 Ry. We use the PBE-GGA (Perdew-Burke-Ernzerhof 96 [27]) as the exchange-correlation potential. The energy which separates the valence from core states is set to −6.0 Ry. The k-point sampling is 1000 k points. The partial DOS of Zn and O are calculated based on the distribution of the Eigen states of the ground levels. The XANES spectra can then be obtained by the transitions between these energy levels following the selection rules.

Morphology and structure from SEM/EDX and XRD

Figure S2 shows the morphology for both undoped ZnO and P-doped ZnO nanoparticles on the Si substrate prepared by the hydrothermal (Fig. S2(a) and (b)) and the CVD (Fig. S2(c) and (d)) methods. It is immediately apparent from Fig. S2 that the hydrothermal method produced nanorods while the CVD method yielded particles under our experimental conditions. It is interesting to note that the P-doped structures (Fig. S2(b) and (d)) are noticeably larger than their undoped counterparts. This observation indicates that the P source has played an important role in modifying the kinetics during crystal growth. The precursor concertation determines the density of the rod distribution, and the speed of nucleation determines the size of the nanostructures: the faster the growth, the smaller and denser the crystallites. When the first generation of nanorods appears, further nucleation would be more likely to participate in the growth of existing nanorods instead of forming a new nucleus on the Si substrate [28].

The chemical reaction for ZnO synthesised by CVD is 2Zn + O₂ → 2ZnO [1, 16]. The mechanism of ZnO synthesis can be divided into two stages: the formation of zinc vapor, and the nucleation and oriented growth of ZnO [29]. The Zn vapor would first react with residual oxygen in the system forming ZnO vapor that dissolve in liquid Au and forms alloy droplets. When the alloy droplet became supersaturated ZnO particles would be grown on the Si wafer surface [30]. When phosphorus participates in the synthesis, it can be present as both vapor and solute in Au droplet. If phosphorus substitutes the Zn in P-doping ZnO nanocrystals during growth, it will impede the process leading to large but less dense crystallites as seen in both Figs. S2b and S2d, where phosphorus was present.

EDX was also employed to determine the elemental distribution in the products. The bottom panel of Fig. S2 shows the EDX mappings for P-doped ZnO samples by hydrothermal (left panel) and CVD technique (right panel). The left panel shows that the P signal is weak and scattered in comparisons with those of Zn and O. This observation indicates the P doping is less effective with this technique, especially in the larger rods; the P pattern follows those of Zn and O albeit significantly weaker. On the contrary, the P map from the ZnO:p_CVD sample coincides with those of Zn and O perfectly. In the main time, the atomic percentage in Table S1 also suggests that P signals exist in both doped samples. Therefore the EDX results suggest that the phosphorus is likely doped into the ZnO crystal lattice in both ZnO:p_hydro and ZnO:p_CVD structures.

Figure S3 displays the XRD pattern for both undoped ZnO and P-doped ZnO nanoparticles on the Si substrate prepared by the hydrothermal and the CVD technique. All peaks are indexed to wurtzite-structured ZnO with lattice constant a = b = 0.3289 nm, and c = 2.1197 nm, confirming that all samples have HCP structure. By comparing P-doped ZnO samples and undoped ZnO samples, there is no obvious peak shifts among (100), (002) and (101). This is because of the atomic percentage of P is too low (Table S1). By combing the EDX results and XANES results to be discussed below, the P may be doped into the surface of ZnO particles.

XANES

XANES tracks the modulation of the absorption coefficient above the edge arising from the interference of the outgoing and backscattered photoelectron via multiple scattering. It contains all the information about the chemical surrounding of the absorbing atom [31]. Since the O K-edge and Zn L_3,2-edge fall into the soft X-ray region with very short one absorption lengths (Fig. S4), XANES is often measured with total electron yield (TEY), and partial fluorescence yield (PFY) with the added advantage that TEY is surface sensitive, and PFY (often collected with a SDD detector with moderate energy resolution) is bulk sensitive. In this section, both TEY and PFY modes are used at the O K-edge, P K-edge, and Zn L_3,2-edge. Zn K-edge is only recorded with FY since the measurement was conducted at ambient condition.

Figure 1(a) shows the Zn K-edge XANES of both undoped ZnO and P-doped ZnO by the hydrothermal and CVD technique in PFY mode and Fig. 1(b) displays the overlay of the normalized XANES. All the spectral features are remarkably similar with a noticeable broadening-observed in the doped samples, especially in the first two peaks. According to the dipole-transition selection rule, the major features A₁, B₁, and C₁ are Zn 4p characters in the conduction band in good agreement with the DOS shown in Fig. 1(e). The observation indicates that all samples have a wurtzite structure, and the doped samples exhibit some disordered locally. The disorder is indicative of doping, contracting the lattice constant slightly. Since XANES suggests that all nearest neighbors of Zn remain oxygen in a TD or slightly distorted TD environment, there is no evidence of direct Zn-P bonding.

Fig. 1:

Zn K-edge and L-edge XANES and partial DOS of Zn. (a) Zn K-edge XANES of ZnO:n_hydro, Zn:n_CVD, ZnO:p_hydro and ZnO:p_CVD; (b) Zn K-edge XANES normalized and overlaid; (c) and (d) Zn L_3,2-edge XANES of the same series in TEY and PFY. (e) Partial DOS of Zn in ZnO with region of interest above the Fermi level boxed.

Fig. 2:

O K-edge XANES and partial DOS of O. (a) TEY- and (b) PFY-XANES of ZnO: n_hydro, Zn:n_CVD, ZnO:p_hydro and ZnO:p_CVD at O K-edge; (c) O partial DOS in ZnO with region of interest boxed.

Fig. 3:

XANES spectra of ZnO: p_hydro, ZnO:p_CVD, NH₄H₂PO_4, P₂O₅ and black phosphorus at P K-edge under (a) TEY mode and (b) PFY mode.

Figure 1(c) and (d) display the Zn L_3,2-edge XANES in TEY and PFY mode, respectively. Zn L_3,2-edge XANES tracks unoccupied Zn 3d and 4s states in the conduction band. Since the Zn 3d orbital is fully occupied, the lowest unoccupied orbital of the Zn is 4s with some 4p mixed in via hybridization. This is the region marked with a rectangle in the partial DOS shown in Fig. 1(e), suggesting that the ZnO crystal remains intact upon P-doping, in accord with the Zn K-edge results. It should be noted that in the soft X-ray domain, thickness effect (self-absorption) sometimes plays a role in PFY, reducing the sharpness of the intense absorption. This occurs when the sample is thicker than the one absorption length (1/e attenuation of the fluorescence X-ray, ∼800 nm for Zn L_α, at 1011 eV). In this case, since the PFY mode has the same feature as the TEY mode, there is no noticeable thickness effect. Both Zn K-edge and L_3,2-edge XANES results suggest that the local chemical environment of zinc in P doped ZnO changes little, other than a bit more disordered. These is no spectroscopic evidence that P substitutes O in P doped ZnO.

Figure 2 shows the O K-edge XANES of the same series recorded in both TEY and PFY mode. A₃, B₃ and C₃ features match well with the unoccupied partial DOS of O 2p character as shown in Fig. 2(c). In Fig. 2(a), all P-doped ZnO samples show reduced intensity at peak A₃ compared to that of the undoped samples. This is indicative of hybridization of O 2p states with P 3sp states upon P doping and increase covalence when P substitutes Zn in the lattice. The PFY mode, Fig. 2(b), suffers from some thickness effect as the O K_α has shorter attenuation length in ZnO (∼200 nm) than that of the Zn L_α (∼800 nm) noted above (Fig. S4). Nonetheless, the ZnO characteristic features are unmistakable and exhibit the same trend as the TEY. Most important, the O K-edge TEY-XANES clearly shows features characteristic of ZnO with no sights of impurity features such as possible by-product with a tetrahedral PO₄ moiety, e.g., Zn₃(PO₄)₂, of which the O K-edge XANES is distinctly different from that of ZnO with a threshold, E₀, appears at ∼3 eV higher photon energy and a broad resonance appearing beyond the sharp peak position of ZnO [32].

The above discussed XANES results clearly indicate that P-Zn substitution has taken place. The determining evidence comes from the P K-edge XANES which probes the P neighboring atoms directly. Figure 3 shows the XANES of P-doped ZnO at the P K-edge in both TEY and PFY mode as well as the XANES of NH₄H₂PO₄, P₂O₅ and black phosphorus. It is apparent that both TEY and PFY XANES exhibit same spectral features which indicates that the P distribution is homogeneous and that there is no thickness effect because the P K_α (2017.4 eV) has an attenuation length of ∼ 1 μm in ZnO (Fig. S4). By comparing the white line (WL) position of the P-doped ZnO samples with reference compounds, we see that the WL is at the same position of B₄, the WL of P in NH₄H₂PO₄ and P₂O₅ in a tetrahedral environment at a +5 oxidation state. Thus, the oxidation state of P in ZnO can be assigned to +5. Thus, this result confirms that P substitutes Zn in the ZnO lattice in a tetrahedral site, because if the phosphorus substitutes the oxygen in ZnO, then the lattice structure would become PZn₄ like. Additional evidence comes from (1) the slight red shift in the WL of the P-doped samples which indicates increased covalence of P-O interaction in the ZnO lattice compared to the starting source materials with a PO₄ moiety, providing better screening of the P site in ZnO, and (2) similar but noticeably broadened features beyond the WL in the XANES indicates that P is in a Td or slightly distorted Td environment with O coordination.

XEOL

XEOL is an X-ray photon in/optical photon out technique, it tracks how a material converts the X-ray energy it absorbs into optical emission. It is a de-excitation spectroscopy that measures the optical response of the system with selected excitation energy. The thermalization of the photoelectrons and Auger electrons produced by the absorption of X-ray (as they leave the system) is primarily responsible for the energy transfer to the optical channel. XEOL can be site specific if the atom of interest in a chemical system is preferentially excited at its absorption edge. XEOL is also morphology and crystallinity dependent; this is especially true in nanostructures of which the dimension is shorter or comparable to the thermalization path (same as the Inelastic Mean Free Path, IMFP of electrons in solid) (Fig. S4). In a wide bandgap semiconductor, such as ZnO, XEOL spectra exhibit both near bandgap emission (NBE) and defect emission (DE). Therefore, it is a powerful tool to investigate both bandgap and defect information of crystal lattice [15, 31, 33]. XEOL excited with X-ray energy across the Zn K-edge, Zn L_3,2-edge and O K-edge have been used to obtain both bandgap and defect information of the undoped and P-doped ZnO samples prepared by the CVD method as will be described below.

Figure 4(a) and (c) show the XEOL from undoped and P-doped ZnO prepared by CVD, respectively. For the undoped ZnO sample, the sharp peak at 384 nm (3.23 eV) is the NBE and the broad peak at 510.7 nm (2.43 eV) is DE from oxygen vacancies. For the P-doped sample, the NBE and DE is at 386 nm (3.21 eV) and 507.7 nm (2.44 eV), respectively. Thus, doping has negligible effect on the emission energy. However, the optical yield, the intensity ratio of the NBE to DE, hence the branching ratio, changes as excitation energy changes from below to above the edge for a given sample. Take the undoped sample, ZnO: n_CVD, for example, the highest emission intensity for both NBE and DE occurs at the X-ray energy of 9.6705 keV which is the absorption maximum, and the DE is dominant in all excited energies (A_NBE/A_DE < 0.1, A is the area under the curve). This observation indicates that the crystallinity in ZnO: n_CVD is imperfect, and it contains oxygen vacancies.

Fig. 4:

XEOL of ZnO with excitation energies across Zn K-edge. (a) XEOL from ZnO:n_CVD excited with selected X-ray energies (9.640, 9.659, 9.665, 9.6705, 9.6765 and 9.700 keV) across the Zn K-edge. The insert is the corresponding XANES. (b) Corresponding area ratio (measured as the area under the peak), A_NBE/A_DE, (c) XEOL from ZnO: p_CVD excited with the same energies and (d) corresponding A_NBE/A_DE.

At the Zn K-edge, the 1s excitation channel is switched on, both Zn Kα fluorescence X-ray and the Zn KLL Auger will play a significant role in energy transfer to the optical channel via reabsorption and thermalization, respectively. Similar observation for both NBE and DE was also found by Lin et al. [34] in ZnO nanorods with hard X-ray energy excitation across the Zn K-edge. This is because the electron/hole pair completes thermalization when the electron reaches the bottom of the conduction band, and the hole reaches the top of the valence band. In the meantime, excess energy transferring would improve the crystallinity of ZnO in the shallow region of the structures [35].

Figure 4(b) and (d) show that A_NBE/A_DE < 0.1 for the undoped and A_NBE/A_DE ≥ 0.3 for the P-doped ZnO, which indicates that ZnO:p_CVD has better crystallinity, hence less oxygen-based surface defect (oxygen vacancy) than in ZnO:n_CVD. A_NBE/A_DE tracks the absorption coefficient across the edge indicating that the NBE dominates and the energy transfer process via thermalization has a significant larger effect on DE (slow decay) than NBE (fast decay). In addition, the ratio of A_NBE/A_DE across the edge yields different trends in ZnO:n_CVD and ZnO:p_CVD. In Fig. 4(b), A_NBE/A_DE shows an inverted trend compared to absorption, this is due to thermalization truncation. At the edge jump, the X-ray penetration depth decreases abruptly (Fig. S4), if the defect distribution concentrates near the surface, energy transfer to the DE channel will be less effective since the energetic electrons have a better chance of escaping the solid without complete thermalization.

XEOL at the Zn L_3,2-edge and O K-edge are shown in Fig. 5. At these soft X-ray energies, the attenuation length is significantly shorter than that at the Zn K-edge (Fig. S4). For ZnO:n_CVD at the Zn L_3,2-edge, the NBE appears at 382.9 nm (3.24 eV). The DE splits into two at 502.7 nm (2.47 eV), and 515.9 nm (2.40 eV), which are unresolved in the Zn K-edge XEOL. These two peaks are believed to be due to the distribution of oxygen vacancies [36]. It is apparent from Fig. 5 that A_NBE/A_DE < 0.1 for all X-ray energies in both doped and undoped ZnO, in contrast to the Zn K-edge in which A_NBE/A_DE ≥ 0.3 is observed for the P-doped samples. The NBE peak from ZnO:p_CVD at Zn L_3,2-edge is also at 382.9 nm (3.24 eV). The DE also splits into two at 502.7 nm (2.47 eV), and 515.9 nm (2.40 eV). From Fig. 5(c) and (d), we see that XEOL excited at the O K-edge exhibit similar behavior as those excited at the Zn L_3,2-edge. That is that DE dominates. Similar observation was reported by Wang et al. and Nie et al. [30, 37, 38] in ZnO nanorods excited with soft X-ray energy across the Zn L_3,2-edge and O K-edge. The NBE peak from ZnO: n_CVD excited at the O K-edge is at 382.9 nm (3.24 eV). The DE also splits into two at 502.7 nm (2.47 eV), and 515.9 nm (2.40 eV). The NBE peak from ZnO:p_CVD at O K-edge is also at 382.9 nm (3.24 eV). The DE also splits into two at 502.7 nm (2.47 eV), and 515.9 nm (2.40 eV). The A_NBE/A_DE intensity ratio for ZnO:n_CVD and ZnO:p_CVD excited across the Zn L_3,2-edge and O K-edge are summarized in Table S2. From Table S2, we see that the A_NBE/A_DE for a given sample is similar, this is due to total absorption since the X-ray penetration depth is shallow (Fig. S4). A closer look reveals that the A_NBE/A_DE ratio for the P-doped sample is consistently larger than that of the undoped sample at both edges. This observation indicate that P-doping appears to have improved the crystallinity of the ZnO and that there is a higher concentration of defects in the near surface region, consistent with K-edge XEOL observation. Since the DE from ZnO:n_CVD and ZnO:p_CVD originates from the oxygen vacancies, P doping must have either reduced the number of O vacancies responsible for the DE or impeded energy transfer to the DE channel. Another interesting observation is that excitation at the O K-edge appears to yield larger A_NBE/A_DE ratio than at the Zn L_3,2-edge, in favor of energy transfer to the NBG channel. Exploration of this site-specific energy transfer awaits further scrutiny.

Fig. 5:

XEOL of ZnO with excitation energies across Zn L-edge and O K-edge. (a) and (b): XEOL from ZnO:n_CVD and ZnO:p_CVD, respectively, excited across the Zn L_3,2-edge with X-ray energies of 1020, 1024, 1026, 1028, 1033, and 1041eV. The insert is the Zn L_3,2-edge XANES. (c) and (d): XEOL from ZnO:n_CVD and ZnO:p_CVD, respectively, excited across the O K-edge with energies of 525, 529.3, 533.9, 536.6 and 540 eV.

XES

XES is an element-specific method to probe the occupied electronic structure of materials in the valence band [39]. XES from low z elements or shallow cores recorded with high energy resolution spectrometers can be used to track X-ray fluorescence from the valence to core dipole transition. Thus, the elemental contribution to the valence band can be obtained. XES is best obtained in the RIXS (Resonant Inelastic X-ray Scattering) region where the fluorescence intensity is enhanced, and the core-hole lifetime broadening is supressed as described by the modified Kramer-Heisenberg equation [40]. X-ray fluorescence decay with the electronic transition from the valence band states to the core-hole states following the dipole selection rule. For examples, in XANES, the Zn 2p electron is excited into the 4s/3d states. In XES, the hole that was created during the XANES process can be filled with electrons from Zn 3d and 4s, of which the former is shallow core, and latter is in the valence band as shown in the DOS. At the Zn K-edge, XES from valence to core transition would be the 4p to 1s transition [41, 42]. XES excited at selected photon energy across the Zn L_3,2-edge and O K-edge have been obtained on both undoped and P-doped ZnO prepared by the CVD technique.

Figure S5 shows the XES spectra of ZnO:n_CVD ((a) and (c)) and ZnO:p_CVD ((b) and (d)) excited across the Zn L_3,2-edge and the O K-edge, respectively. The insert is the corresponding XANES. It should be noted that the valence band is primarily of O 2p character mixed with some Zn 3d/4s via hybridization so that the bandgap is best obtained by aligning the O K-edge XANES and O K-edge XES. The XES is comprised of O 2p states near the VBM. From Fig. S5(a) and (b), feature A₅ is due to the hybridization between O 2p and Zn 4p states. The energy shoulder (B₅) in the 522–524 eV region arises from the mixed states of O 2p-Zn 4s. The single band shape (feature C₅) is attributed to the O 2p hybridized with Zn 3d states [43]. There is no obvious feature A₅ and feature B₅ when the X-ray energy is at 529.4 eV which indicates that the core-hole is yet to be fully turned on and this is the RIXS which exhibits a dispersion (constant energy loss). There is no obvious difference among three X-ray energies at 533.9, 536.6 and 540 eV since this is now the characteristic fluorescence; small shift is due to adiabatic to sudden transition. XES at Zn L_3,2-edge are the same for the undoped and the P-doped samples (Fig. S5(c) and (d)). The XES is comprised of Zn 4s and 3d in the VBM. The selective excitation X-ray energies contribute to the expected admixture of 3d and 4s to 2p transitions.

The influence of doping to the electronic structure can be obtained from the band gap revealed from the O K-edge XANES-XES plot (Fig. S6). The excitation energy of the O K-edge XES for both samples is 536 eV. To avoid the uncertainties involved extrapolating the leading edges in the XANES and XES to the baselines, the second derivative of the XES and XANES spectra is plotted and used to determine the band gap value. The distance between the highest-energy peak of the XES 2nd derivative and the lowest-energy peak of the XANES 2nd derivative yields values of 3.25 eV for both undoped ZnO and P-doped ZnO. By comparing with the XEOL results, second derivative of the O K-edge XANES-XES plot yields a slightly larger bandgap values for both undoped and P-doped ZnO. Since the XEOL methodology is with less uncertainty, the NBE results are more reliable.