Probing Charge Carrier Density in a Layer of Photodoped ZnO

Aug 18, 2010 - Probing Charge Carrier Density in a Layer of Photodoped ZnO Nanoparticles by Spectroscopic Ellipsometry ... Molecular Materials and Nan...
0 downloads 4 Views 2MB Size
14804

J. Phys. Chem. C 2010, 114, 14804–14810

Probing Charge Carrier Density in a Layer of Photodoped ZnO Nanoparticles by Spectroscopic Ellipsometry Girish Lakhwani,† Roel F. H. Roijmans,† Auke J. Kronemeijer,‡ Jan Gilot,† Rene´ A. J. Janssen,† and Stefan C. J. Meskers*,† Molecular Materials and Nanosystems, EindhoVen UniVersity of Technology, P.O. Box 513, 5600 MB, The Netherlands, and Molecular Electronics, Zernike Institute for AdVanced Materials, RijksuniVersiteit Groningen, Nijenborgh 4, 9747 AG Groningen, The Netherlands ReceiVed: May 27, 2010; ReVised Manuscript ReceiVed: July 23, 2010

Changes in the optical constants of a layer of ZnO nanoparticles (5 nm diameter) induced by UV illumination in O2-free atmosphere are determined by using spectroscopic ellipsometry. The onset of optical absorption of ZnO shifts to higher photon energy after illumination. This is interpreted in terms of a Moss-Burstein shift. From the magnitude of the shift, the charge carrier density in the conduction band after UV illumination was determined to be 2 × 1019 cm-3, about one carrier per particle. Kelvin probe measurements give a lower limit for the density of 1018 cm-3. The free carrier density after illumination is high enough to explain the formation of quasi-ohmic contacts between ZnO and the polymeric p-type conductor poly(3,4-ethylenedioxythiophene) (PEDOT). 1. Introduction The n-type semiconductor ZnO1,2 can be obtained in a variety of nanosized structures and shapes.3 ZnO nanocrystals with ∼5 nm diameter can be prepared routinely4 and have been tested in a variety of applications including gas sensors,5–9 UV sensors,10–16 solar cells,17,18 and electronic memories.19–23 An intrinsic advantage of the nanoparticle layers is that they have a high density of surface related electronic states because of a high internal surface area. For ZnO it is well-known that mobile electrons resulting from the intrinsic n-type doping can be trapped by gas molecules such as O2, NO2, and CO, that are adsorbed at the ZnO surface.24 This property allows for application of ZnO in resistive gas sensors. In addition, introduction of positive charge carriers in the ZnO, e.g., electrically or via illumination with UV light,23 can induce desorption of the adsorbed gas molecules, resulting in a persistent photoconductivity (Figure 1). This property allows for detection of UV light. Layers of ZnO nanoparticles have also been used in multijunction solar cells17 and in electronic memories.25 In the multijunction devices, a thin ZnO layer (thickness ∼30 nm) is used in combination with a heavily doped p-type layer to form an ohmic contact after illumination with UV light. The two layers serve as a charge recombination layer between photovoltaic cells placed in a tandem configuration. In the electronic memory application, the electrical conductivity is raised by electrically injecting positive charge carriers in the ZnO particles, resulting eventually in the formation of a contact between ZnO and a polymeric p-type conductor that has essentially ohmic characteristics.25 In these two examples, the density of electrons in the conduction band of the ZnO layer plays a crucial role. A question that has remained largely unanswered is what electron densities can actually be achieved in a layer of ZnO nanopar* To whom correspondence should be addressed. E-mail: s.c.j.meskers@ tue.nl. † Eindhoven University of Technology. ‡ Rijksuniversiteit Groningen.

Figure 1. Schematic illustration of the induction of free charge carriers in ZnO nanoparticles by either UV illumination or injection through electrical contacts and their removal by O2.23

ticles by UV illumination. The answer to this question plays a key role in understanding the formation of quasi-ohmic contacts between ZnO and other (semi)conductors. In this study we determine this charge carrier density from optical measurements. Apart from changing the conductivity, the presence of the mobile electrons also influences the optical properties of ZnO. Earlier studies on ZnO, where the level of n-type doping was varied systematically via intrinsic doping26 or extrinsic dopants such as In27 and Al28 have revealed a correlation between the optical properties of doped ZnO and the density of electrons in the conduction band. Changes in optical properties near the onset of the absorption corresponding to the band to band transition can be understood in terms of the Moss-Burstein effect, which is a shift of the band gap with increasing doping levels.29–31 The Moss-Burstein shift can be explained by considering that after partially filling the conduction band of ZnO with electrons, a higher photon energy is required to excite an electron from the valence band of ZnO to an empty level in the conduction band. This results in an apparent increase in the semiconductor band gap, marked by a blue shift in the absorption edge of the ZnO. The following simple relation has been proposed to describe the magnitude of the Moss-Burstein shift in ZnO:28

10.1021/jp104846h  2010 American Chemical Society Published on Web 08/18/2010

Probing Charge Carrier Density in ZnO Nanoparticles

Emin ) Eg +

( )

p2 1 (3π2nC)2/3 2 mred

J. Phys. Chem. C, Vol. 114, No. 35, 2010 14805

(1)

where Emin denotes the photon energy corresponding to the onset of absorbance arising from the transition of electrons from valence to conduction band in the direct band gap semiconductor. Eg refers to the band gap of the undoped semiconductor, mred to the reduced effective mass of the electron hole pair: 1/mred ) 1/mC + 1/mV, with mC and mV the effective masses of the carriers in the conduction and valence band (for ZnO: mC ) 0.28m0, mV ) 0.58m0, with m0 the electron rest mass). Finally nC denotes the density of charge carriers in the conduction band. Experiments on ZnO doped by various donors have indeed revealed the predicted nC2/3 dependence at low doping densities26,28,32 (nC e 2 × 1019 cm-3). These experiments also showed that, at high doping densities various corrections to eq 1 are necessary to quantitatively explain the observed Moss-Burstein shift.33 Here we use ellipsometry to study the optical properties of a layer of ZnO nanoparticles. In these experiments we use UV light to externally modify the effective degree of doping. By evaluating the magnitude of the Moss-Burstein shift we estimate the density of mobile electrons in UV-doped ZnO. Kelvin probe measurements on a junction between ZnO and tin doped In2O3 (ITO) are used as an independent way to estimate a lower limit for density of carriers. Finally, we apply ellipsometric measurements to investigate the formation of an ohmic contact in junctions between a layer of ZnO nanoparticles and a heavily doped, p-type polymeric conductor (poly(3,4ethylenedioxythiophene), PEDOT),34 upon photodoping the ZnO nanoparticle layer. 2. Experimental Section ZnO nanoparticles with an average diameter of 5 nm were synthesized as described before.4,35 Layers of nanoparticles were deposited by spin coating from a solution in acetone (10 or 25 mg/mL). ITO/PEDOT:PSS(50 nm)/ZnO(40 nm)/LiF(1 nm)/ Al(100 nm) diodes for electrical measurements were fabricated by spin coating a PEDOT:PSS dispersion in water (Clevios P VP AI 4083), followed by spin coating of ZnO nanoparticles from acetone and subsequent thermal evaporation of LiF and Al at a pressure of 10-7 mbar. Layers of nonacidic, pH-neutral, PEDOT (abbreviated na-PEDOT, Orgacon, batch 5541073, pH 7, 1.2 wt %, Agfa Gevaert NV) were spin coated from water after 1:1 dilution to obtain a 15 nm thick layer. Spin coating na-PEDOT on a bare Si wafer was hampered by poor wetting of the substrate. These dewetting problems were avoided by using glass as a substrate. To minimize reflections from the backside of the glass substrate in the ellipsometry measurements, the backside was covered with Scotch tape that effectively scatters the light, reducing reflection from the back surface. Optical constants of a layer of ZnO nanoparticles were determined via ellipsometry (300-1700 nm) in air with a WVASE32 ellipsometer (J. A. Woollam Co., Inc.). Ellipsometric measurements were performed with a Si wafer with a 2.28 nm thick oxide layer, as determined from ellipsometry on the bare wafer with use of known optical constants.36 Atomic force microscopy revealed that the surface of the ZnO layer has a roughness with a rms value of 10 nm. Surface roughness was taken into account in the modeling by using a convolution method; for more details see the Supporting Information. Illumination with UV light under O2 free conditions was performed by placing the sample film in a special sample chamber (VASE heat cell attachment, Heatcell.03 c) and

flushing with argon. A correction for attenuation and phase lag of the incident and reflected light when passing through the windows (held at a fixed angle of 70°) of the chamber was included when modeling the ellipsometric data Ψ and ∆. Illumination of the ZnO layer in the chamber was performed with a UV lamp (Spectroline, ENF 260C/F) with broad band emission centered on 3.5 eV with 0.4 eV fwhm, illuminating through the window of the chamber. Current-voltage characteristics and Kelvin probe measurements were performed in inert N2 atmosphere (O2, H2O < 10 ppm). Positive bias is defined as the Al electrode being charged negative. The Kelvin probe setup (Besocke Delta Phi, Kelvin Probe S and Kelvin Control 07) was enclosed in a home-built Faraday cage. ZnO samples were measured under inert atmosphere before and after 30 min of illumination with a solar simulator (Steuernagel, Solar Constant 1200). 3. Results and Discussion 3.1. Optical Constants of a Layer of ZnO Nanoparticles. Ellipsometric angles Ψ and ∆ for a thin layer of ZnO nanoparticles on a Si wafer were determined in air by spectroscopic ellipsometry (Figure 2). From these experimental data, the optical constants were calculated by fitting an oscillator model. In the oscillator model used, we assume that the ZnO layer behaves as a homogeneous and isotropic material. Crystalline ZnO is an anisotropic material with a hexagonal crystal structure and therefore one cannot a priori assume isotropic optical constants. However, the dielectric functions for crystalline ZnO show only a small anisotropy.37 In addition transmission electron and atomic force microscopy give no indication for preferential orientation or alignment of the approximately spherical nanocrystals. Henceforth, considering also the small size of the nanocrystals, an isotropic model would suffice to determine the optical constants. The oscillator model used is illustrated in Figure 2c for the imaginary part ε2 of the complex dielectric constant (ε˜ (ν) ) ε1(ν) + iε2(ν)). The model includes a Tauc-Lorentz38 oscillator to describe the band edge. Its contribution to ε2(ω) is defined as:

{

AEcB(E - Eg)2

ε2TL ) (E2 - Ec2)2 + B2E2 0

·

1 E>E g E

(2)

E e Eg

where A is a constant in units of energy, B is the energy bandwidth at half-maximum, Ec the resonance energy (peak transition energy), and Eg is the band gap energy (onset of absorbance). The real part of the Tauc-Lorentz can be obtained by Kramers-Kronig transformation of ε2.39 In addition to the Tauc-Lorentz model that is often used to account for the band edge of ZnO,40 we include three additional Gaussian oscillators in our model with resonance frequencies above that of the Tauc-Lorentz function, allowing for accurate reproduction of the ellipsometric angles at photon energies >3.5 eV. Fitting this oscillator model to the experimental data yields the parameters listed in Table 1. The broad optical transition with resonance frequency near 6 eV has been reported earlier for microcrystalline films;41 for the other two Gaussian oscillators we do not have a detailed assignment. The ε2 spectrum resulting from the fit is shown in Figure 2c. The mean square error of the fit is 6.42. The average thickness determined by the ellipsometry experiments was found to be 37.8 nm, in good accord with the thickness measured by surface profilometery (35 nm).

14806

J. Phys. Chem. C, Vol. 114, No. 35, 2010

Lakhwani et al. TABLE 1: Fit Parameters for the Oscillator Model for Thin Films of ZnO Nanoparticles on Sia oscillators

A (eV)

B (eV)

Ec (eV)

Tauc-Lorentz Gaussian Gaussian Gaussian

72.88 0.04 0.15 2.52

0.26 0.49 0.86 2.43

3.34 3.87 4.44 6.07

Additional fit parameter for the Tauc-Lorentz function: Eg ) 3.25 eV. a

Figure 2. Ellipsometric angles Ψ (a) and ∆ (b) for a 35 nm thin film of ZnO nanoparticles on a Si wafer at different angles of incidence together with the optical model fitted to the data. (c) Oscillators included in the model. (d) ε1 and ε2 spectra resulting from the fit.

The complex dielectric response function ε˜ (ν) ) ε1(ν) + iε2(ν) in the optical frequency range for the layer of ZnO nanoparticles as obtained from fitting the model is shown in Figure 2d. For the layer of ZnO nanoparticles, the imaginary part of the dielectric function ε2, describing dissipation of optical energy, shows an onset at 3.25 eV photon energy and a characteristic peak at ∼3.4 eV. This corresponds to the absorption edge of the direct band gap semiconductor ZnO. The dielectric function as shown in Figure 2 can be compared to those reported for crystalline ZnO.37,42 For this we use the maximal slope of the rising flank of the ε2(ν) spectrum as a semiempirical criterion to determine the onset of the interband transition.26 Thus, we find the onset at 3.36 eV in agreement with other studies on thin (