Die elektronische Struktur des amorph-kristallinen Silizium-Heterostruktur-Kontakts

In der vorliegenden Arbeit wurde die elektronische Zustandsdichte hydrogenisierter amorpher Silizium (a-Si:H)-Schichten mit Dicken von 300 bis hinab zu ~2 nm untersucht. Die dazu eingesetzte Methode der Nah-UV-Photoelektronenspektroskopie (NUV-PES) liefert eine gemittelte Valenzband-Zustandsdichte d...

Full description

Saved in:
Bibliographic Details
Main Author: Korte, Lars
Contributors: Fuhs, Walter (Prof. Dr.) (Thesis advisor)
Format: Doctoral Thesis
Published: Philipps-Universität Marburg 2006
Online Access:PDF Full Text
Tags: Add Tag
No Tags, Be the first to tag this record!

In the present work, the electronic density of states of hydrogenated amorphous silicon (a-Si:H) layers in the thickness range from 300 down to ~2 nm was examined by Near-UV-photoelectron spectroscopy (NUV-PES). The measurements yield a mean density (averaged over all directions in k space) of the extended states in the valence band close to the band edge Ev, down to approximately Ev–1 eV, as well as the density of states in the band-gap between Ev and the Fermi level Ef. Using the constant final state yield spectroscopy (CFSYS) mode of NUV-PES, which was extended to a density of states spectroscopy in the present work, densities of states can be measured over more than 8 magnitudes, from > 10e22 down to 10e14 states/(eV cm3). The CFSYS detects photoelectrons with a constant final state energy, and the excitation-energy hv is varied, between ~4 and 7-8 eV in the experimental set-up used here. The detection energy is chosen so that only those electrons are counted whose kinetic energy is just high enough to overcome the work function barrier of the sample into the vacuum. This is in contrast to conventional techniques of photoelectron spectroscopy, where the excitation energy (e.g. the characteristic emission line of a gas discharge lamp) is kept constant and the detection energy is varied. Advantages of the CFSYS are the information-depth of ~5-10 nm, which is large compared to other photoelectron spectroscopies and matches optimally the a-Si:H layer thicknesses of the examined samples, and the mentioned span of 8 magnitudes in the detectable density of states. To avoid systematic errors in the the evaluation of the measured spectra, it is necessary to take into account the energy resolution of the PES system. Therefore, the resolution was measured and an (excitation energy dependent) transfer function was used to describe the influence of the apparatus on the PES measurement. The detailed analysis of the PES measurements was introduced using data obtained from a 300 nm thick, nominally undoped intrinsic a-Si:H layer that was deposited by PECVD on crystalline silicon. In order to evaluate the spectra, the measured photoelectron yield data have to be normalized to a density of states, and the valence-band mobility edge EVµ must be determined (the position of EF is known in the measured spectra, because the Fermi level is the common reference potential of energy analyzer and sample). To this end, an analytic model for the density of states was fitted to the measured yield data. The model describes the extended states close to the band edge as well as the localized states in the band gap. The gap states consist of the exponentially decreasing band tails (Urbach tails), which are commonly interpreted as strained Si-Si bonds, as well as dangling (broken) bonds, whose density of states distribution is approximately a Gaussian. The model spectrum was normalized to 2*10e21 states/(eVcm³) at the valence band mobility edge, which coincides within some 10 meV with the top of the exponential Urbach tail. The convolution of this model spectrum with the transfer function of the spectrometer could be used to obtain excellent fits to the data measured in all available PES modes (CFSYS as well as conventional UPS and total yield spectroscopy). The defect parameters obtained from the fits to the 300nm sample are elevated with respect to literature data: The exponential tail slope (Urbach energy) amounts to 62(3) meV, the density of the dangling bond defects to 3,5*10e18 cm-3. The Fermi level, EF–EVµ=1,1 eV, lies in the upper half of the band gap. The high density of dangling bonds was confirmed by photocurrent spectroscopy (CPM) measurements on 300nm a-Si:H films deposited on glass under identical deposition conditions, which could be described by the same density of states model as the PES data. In contrast to PES (which "sees" only the 5-10nm of the a-Si:H layer close to the surface) the photocurrent measurement yield the defect parameters averaged over the entire layer thickness. The PES and CPM results are consistent with ESR measurements on the same 300 nm films, which also show elevated values of the defect-density with respect to literature data for thick layers. It has to be concluded that either the deposition conditions were not optimized or the high defect density is due to contamination by impurity atoms. Finally, the photocurrent measurements can be evaluated in the Tauc plot to yield the optical band-gap, Egopt=1,76(5) eV. The methodology developed in the first part of the thesis (PES measurement and fit of the model density of states) was then applied to various series of approximately 10 nm thin a-Si:H layers on c-Si substrates, where the deposition temperature of the layers and the concentration of their doping both by phosphorus and boron were varied. Since for these layer thicknesses, the position of the Fermi level cannot be determined from measurements of the photo- or dark conductivity, the near-UV-PES is the only method to yield data on defect densities and the position of the Fermi level here. The experimental results can be summarized as follows: Ultrathin a-Si:H layers show an optimum of the deposition-temperature around 230°C, comparable to the well-established optimum temperature range for thick layers. The optimum is characterized by an Urbach energy of 66(1) meV and a defect-density of 2,9(3)*10e18 cm-3. For undoped layers, the Fermi level lies EF–EVµ=1,04(6) eV, the films are therefore slightly n-type. Conductivity measurements at identically prepared thick layers on glass allow to determine the distance of the Fermi level to the conduction band mobility edge, ECµ–EF. From these quantities, the mobility gap Egµ=Ecµ–Evµ can be determined under the assumption that the band gap of thin layers does not change in comparison to that of thick ones. It shows the same trend of a decrease of the band gap with rising deposition temperature as is known from thick layers. The Fermi level in the band gap can be shifted by doping for ultrathin layers in a similar way as for thick films; however, only a maximum variation of EF–Evµ from 0,55(6) eV at 1e4 ppm B2H6 in the gas phase to 1,49(6) eV (1e4 ppm PH3) is possible. Under the assumption that the band gap Egopt~Egµ is unchanged with respect to the thick, undoped layer, the latter corresponds to a distance of EF to the conduction band mobility edge of 0,27(8) eV. The small variations of the position of the Fermi level even for high doping levels correlate with reports in the literature, where from measurements by Near-UV-PES on thick a-Si:H, it was concluded that the position of the surface fermi level deviates from the bulk position. However, for the ultra-thin layers examined here, PES measures the mean Fermi level over the entire layer thickness; furthermore, it was estimated that there should be no considerable band bending in the ultra-thin layers. It seems plausible that the elevated density of dangling defects and band tail states (higher Urbach energy) should be at the origin of the smaller variation of Ef–Evµ. The cause for the poorer quality of the ultra-thin layers is certainly on the one hand the proximity of two interfaces (the vacuum/a-Si:H and the a-Si:H/c-Si interface) as well as the short deposition time of the ultra-thin layers, that hinder the relaxation of the amorphous network; on the other hand, suboptimal deposition conditions as well as a contamination as it was found for the thick layers cannot be ruled out. Both for the deposition temperature series of the undoped layers and the doping series with a substrate-temperature of 230°C, measurements of the transient surface-photovoltage (SPV) were carried out. For these measurements, excess charge-carriers are generated in the c-Si by a short light pulse, and the resulting change in the position of the Fermi level relative to the band edges is measured. These measurements yield through their decay behavior over time (which is an image of the bands relaxing to their dark equilibrium position) a measurement for the recombination velocity at the a-Si:H/c-Si interface. Furthermore, if high-injection conditions can be reached, i.e. a concentration of photogenerated excess charge-carriers in the c-Si that exceeds significantly both the minority as well as the majority charge-carrier density in the dark (high-injection case), then the maximum photovoltage is a measurement for the band bending in the c-Si in dark equilibrium (without illumination). If the valence band offset at the a-Si:H/c-Si interface is known, one can also calculate from PES measurements of Ef–Evµ the position of the Fermi level at the interface relative to the c-Si band edge. For optimal deposition-conditions, both calculations yield an identical position of the interface fermi level. this correlates with a slow decay of the SPV transient. For suboptimal a-Si:H deposition conditions (Tdep not 230°C, phosphorus doping >3000 ppm), the transient decays faster, and the calculated positions of the interface Fermi levels no longer agree. The faster decay as well as the discrepancy of the calculated Fermi levels are indications for an elevated recombination velocity: Obviously, the high excitation case is no longer reached. This is can be explained by the fact that for suboptimal deposition conditions the density of states both in the a-Si:H band-tails and deep in the band gap increase so that the photogenerated excess charge-carriers can tunnel from the c-Si into the amorphous layer where they find an efficient recombination path. If the a-Si:H layer thickness is reduced below the information depth of the Near-UV-PES, the density of states of the underlying substrate contributes to the measured PES signal. In this thesis, it was shown that the measured spectrum can be described as the sum of the spectra of a hydrogen terminated c-Si surface and a thick (~100 nm) a-Si:H layer. The shift of the two spectra relative to each other is the band offset DEv of the a-Si:H/c-Si interface, without further assumptions such as the ones necessary e.g. for the evaluation of C-V measurements to determine DEv. Such measurements and fits were carried out on film thickness series of intrinsic, n- and p-doped ultrathin a-Si:H layers on n- and p-doped c-Si wafers. They represent the first systematic examination of the a-Si:H/c-Si band offset with Near-UV-PES. The mean value of the band offset was determined to DEv=0,458(6) eV. It is independent of the doping both of the a-Si:H layer and the substrate; however, it decreases slightly with the a-Si:H layer-thickness. The absolute value of DEv, the independence of the doping and the decrease with the layer thickness are consistent with a model in which the band offset depends on the so-called charge neutrality level of the materials on both sides of the interface and on additional contributions by extrinsic interface dipoles.