Localization of Ag Dopant Atoms in CdSe ... - ACS Publications

Jul 6, 2015 - Nanoparticle Process Technology and Center for Nanointegration Duisburg-Essen (CENIDE), University of Duisburg-Essen, 47057. Duisburg ...
0 downloads 0 Views 7MB Size
Article pubs.acs.org/JPCC

Localization of Ag Dopant Atoms in CdSe Nanocrystals by Reverse Monte Carlo Analysis of EXAFS Spectra Alexander Kompch,† Ayaskanta Sahu,‡,§ Christian Notthoff,† Florian Ott,§ David J. Norris,§ and Markus Winterer*,† †

Nanoparticle Process Technology and Center for Nanointegration Duisburg-Essen (CENIDE), University of Duisburg-Essen, 47057 Duisburg, Germany ‡ Department of Chemical Engineering, University of Minnesota, Minneapolis, Minnesota 55455, United States § Optical Materials Engineering Laboratory, ETH Zurich, 8092 Zurich, Switzerland ABSTRACT: The structure of CdSe nanocrystals doped with 0.2%−2.5% Ag corresponding to 1.1−13.6 Ag atoms per nanocrystal is studied in detail by a combination of X-ray diffraction (XRD) and X-ray absorption spectroscopy at the Ag−K, Cd−K, and Se−K edges. X-ray absorption near-edge structure (XANES) data are compared with ab initio multiple scattering simulations. Extended X-ray absorption fine structure (EXAFS) spectra are analyzed by reverse Monte Carlo (RMC) simulations. The XANES data provide evidence that Ag is located inside the CdSe nanocrystals, and the EXAFS spectra show that the local structure of Ag can be described by tetrahedral interstitial sites in either wurtzite or zinc blende lattices similar to the coordination of Ag in Ag2Se.

1. INTRODUCTION Formally, in semiconductor physics, doping is the incorporation of foreign atoms into a semiconductor that can act as donors or acceptors and increase the charge carrier density. Typically, the number density of dopants is rather small: of the order of parts per million. In bulk semiconductors like silicon, dopants like boron or phosphorus provide p- or n-type conductivity, respectively, with charge carrier densities well above the intrinsic level at ambient temperature.1 Typical dopant concentrations are about 1022−1024 m−3. At a bulk number density of 5 × 1028 m−3 for silicon this corresponds to a doping level of parts per million. A 5 nm silicon particle has a volume of about 6.5 × 10−26 m3 corresponding to about 3250 atoms and on average 10−4−10−2 dopant atoms per particle at the bulk doping level. Therefore, most particles will be undoped in a corresponding ensemble, and the doping level for particles actually containing dopant atoms is orders of magnitude higher compared to the bulk level. Hence, doping of semiconducting nanoparticles has a completely different character compared to bulk systems. In order for the dopant atom to be electrically functionalto provide mobile charge carriersit should form shallow impurity levels in the host semiconductor. Deep impurity centers typically generate a larger (local) distortion of the host lattice and trap charge carriers.2 Therefore, the challenge in obtaining and understanding functional doping of nanoparticles is to control and determine the distribution of the dopant atoms relative to the particle surface and the location of the dopant atoms relative to the crystal lattice. Because standard diffraction techniques average the structural information over length scales much larger compared to the © 2015 American Chemical Society

lattice constant, information about the dopant location is usually obtained within the so-called virtual crystal approximation.3 Extended X-ray absorption fine structure (EXAFS) and X-ray absorption near-edge structure (XANES) in X-ray absorption spectra, however, contain element-specific information on the local structure. Only very few investigations of the local structure in doped CdSe nanocrystals by XAFS have been published so far. Meulenberg et al. investigated the Cu doping of CdSe.4 XANES spectroscopy, X-ray photoluminescence spectroscopy (XPS), and photoluminescence (PL) spectroscopy revealed a statistical doping mechanism where Cu(I) may act as a deep electron trap. Synthesis of Cu-doped CdSe quantum dots was recently reported by Jawaid et al. where organometallic clusters containing four copper atoms are used as nucleation seeds.5 This ensures that every nanocrystal contains the same number of dopants. However, XANES and EXAFS spectra at the K-edge of Cu reveal that the original cluster seeds remain partially intact and that the Cu species can trap n- and p-type charge carriers. Analogous investigations were performed for ZnS and InAs. Gul et al. successfully doped ZnSe with Cu.6 EXAFS spectra reveal that Cu atoms are surrounded by four neighbors in the lattice but are located close to the nanoparticle surface with Cu participating in the trapping of charge carriers. Recently, Amit et al. investigated the impurity location and binding in InAs nanocrystals heavily doped with Received: May 7, 2015 Revised: July 2, 2015 Published: July 6, 2015 18762

DOI: 10.1021/acs.jpcc.5b04399 J. Phys. Chem. C 2015, 119, 18762−18772

Article

The Journal of Physical Chemistry C

Ag atoms can potentially enhance the fluorescence of the CdSe nanocrystals dramatically. The nonmonotonic trends in fluorescence and shifts in the Fermi level with doping suggest that Ag changes from an interstitial (n-type) to a substitutional (p-type) impurity with increasing dopant content. However, we lacked structural information regarding the exact location of the dopants inside the nanocrystals. Here we investigate in detail the sites on which the dopant atoms are located by a combination of X-ray absorption fine structure (XAFS) spectroscopy, which provides elementspecific information about the local structure around the dopant atoms, and a data analysis method that extracts this key structural information through partial pair distribution functions, namely, reverse Monte Carlo analysis (RMC). Both, XAFS and RMC do not rely on long-range, periodic order which is not present for the dopant atoms in randomly doped materials. The analysis of EXAFS spectra using reverse Monte Carlo (RMC) simulations9 makes it possible to refine spectra of different absorbing atoms with a single, mutual model (a configuration of atoms), and it is possible to get consistent structural information on the dopant and the host.10 The final structural results are the partial pair distribution functions (PDFs), which can be analyzed to obtain information about mean coordination numbers, bond distances, mean square displacements, and higher moments of the distribution. Cadmium selenide doped with silver has the principal advantage that K-edge XAFS spectra are easily accessible for all three elements, namely Cd, Se, and Ag. Therefore, we are able to use RMC analysis of the complete set of all three EXAFS spectra with a single, mutual model to determine the local structure, i.e., the dopant site, of Ag in CdSe nanocrystals.

Figure 1. X-ray diffractogram of the 0.67% Ag-doped CdSe sample fitted by Rietveld refinement using GSAS and MAUD (number of data reduced to 10% for clarity).

Cu (200 to 2500 Cu atoms per nanocrystal) by a combination of XAFS and DFT with a focus on the concentration dependence in heavily doped systems.7 The Cu atoms are located on hexagonal interstitial sites in the zinc blende lattice leading to n-type conductivity. At high doping level Cu−Cu pairs are observed. In our previous publication we discussed the incorporation of Ag dopants in CdSe nanocrystals.8 We found that even a few

Table 1. Results of Rietveld Refinement with GSAS and MAUD (Figure 1) of X-ray Diffraction Data of Nano-CdSe Doped with 0.67% Ag Measured at Room Temperaturea phase; (wt %)

CdSe, wurtzite, P63mc; 58(1) (GSAS, MAUD)

a (Å) 4.274(2) (GSAS); 4.317(7) (MAUD) b (Å) c (Å) 6.933(7) (GSAS); 7.24(3) (MAUD) V (Å3) 109.7 (3) (GSAS); 116.9 (9) (MAUD) GSAS: S O x y Cd 2b 0.99 1/3 2/3 Ag 2b 0.01 1/3 2/3 Se 2b 1.00 1/3 2/3 MAUD: S O x y Cd 2b 1.00 1/3 2/3 Se 2b 1.00 1/3 2/3 d (nm) 0.88(2) (GSAS); 35(10) (MAUD) MAUD: ε 0.041(1) I 0(0.001) E 0(1) T 0.057(7) Δ2θ, δ2θ 10−60, 0.04 # Bragg 180 # param. 10 GSAS: Rp = 4.95%, Rwp = 5.91% MAUD: Rw = 5.01%, Rwnb = 4.39%, R = 3.97%, Rnb = 3.80%

CdSe, zinc blende, F−43m; 42(1) (GSAS, MAUD) 6.062(2) (GSAS); 6.054(9) (MAUD)

z 0 0 0.375 z 0 0.369(8)

U (A2) 0.025 0.025 0.025 B (A2) 0.3(1) 0.3(1)

222.8 (2) (GSAS); 221.9 (10) (MAUD) S O x y 4b 0.99 1/4 1/4 4b 0.01 1/4 1/4 4a 1.00 0 0 S O x y 4b 1.00 1/4 1/4 4a 1.00 0 0 0.89(2) (GSAS); 35(10) (MAUD)

z 1/4 1/4 0 z 1/4 0

U (A2) 0.025 0.025 0.025 B (A2) 0.3(1) 0.3(1)

0.02(1) 0.10(1) 0(0.001) 0(0.001)

a a, b, c: lattice constants; V: volume of unit cell; Cd, Ag, Se: atoms in asymmetric unit; S, O: site and occupation, x, y, z: fractional coordinates; U: GSAS isotropic temperature factor (not refined); B: MAUD isotropic temperature factor (coupled for the different sites); d: coherent diffracting domain size; ε: microstrain, I: internal stacking fault probability; E: external stacking fault probability; T: twin fault probability (the fault probability is the fractional area of all atomic close-packed planes which are faulted); Δ2θ: fitted interval; δ2θ: step size; # Bragg: number of Bragg reflections in measured interval; # param.: number of refined parameters; R: Rietveld indices of GSAS and MAUD.

18763

DOI: 10.1021/acs.jpcc.5b04399 J. Phys. Chem. C 2015, 119, 18762−18772

Article

The Journal of Physical Chemistry C

Figure 2. continued

18764

DOI: 10.1021/acs.jpcc.5b04399 J. Phys. Chem. C 2015, 119, 18762−18772

Article

The Journal of Physical Chemistry C

Figure 2. Experimental XANES data compared with FEFF simulations using zinc blende and wurzite structural models. (a) Cd K-edge spectrum of 0.67% Ag in CdSe (number of data reduced to 20% for display clarity). (b) Se K-edge spectra of 0.67% Ag in CdSe (number of data reduced to 50% for clarity). (c) XANES Ag K-edge spectra of CdSe with different Ag contents and Ag2Se. Comparison of experimental Ag K-edge spectra of 0.67% Ag in CdSe with FEFF simulations for Ag on different sites in the zinc blende lattice (d) and wurtzite (e) lattice. (f) Comparison of Ag K-edge spectra of 0.67% Ag in CdSe with FEFF simulations for Ag in CdSe on different sites in the wurtzite lattice as determined from relaxed configurations obtained by DFT simulations. (g) XANES Ag K-edge spectra of Ag metal and Ag2Se compared with corresponding FEFF simulations (number of data reduced to 50% for clarity).

E0 = 26 711 eV, and Pb L3 for Se with E0 = 13 035.2 eV.13 The spectra of commercial powders of CdSe and Ag2Se (