| |
| |
Spatially-resolved, energy-filtered imaging of core level and valence band photoemission of highly p and n doped silicon patterns
(Result of the month 09/2009) |
|
|
Image 1 shows energy-filtered threshold images from the two samples. One can clearly distinguish the heavily doped N+ (P+) regions on the P- (N) substrate for N+/P- (P+/N). On the right the area-averaged Si 2p spectra acquired on N+/P- and P+/N using the NanoESCA in the XPS mode. In this mode the instrument functions as a spectrometer using only the first hemispherical analyser which projects the photoelectrons onto a channeltron detection system. The spectra have two main peaks, one due to the native oxide at a binding energy of 104 eV and another, strongly attenuated, from the underlying bulk Si substrate at 100 eV. The separation of the two peaks, about 4 eV, corresponds to what one would expect for oxidized silicon. However, the spectral features are too large to be attributed to single peaks. This is a first indication that there is mixing of the electronic levels in the same field of view. FoV 25 m, h=127 eV, extraction voltage 12 kV, E~0.15 eV. |
|
Accurate description of spatial variations in the energy levels of patterned semiconductor substrates on the micron and sub-micron scale as a function of local doping is an important technological challenge for the microelectronics industry. Spatially-resolved surface analysis by photoelectron spectromicroscopy can provide an invaluable contribution thanks to the relatively non-destructive, quantitative analysis. We present results on highly doped n and p type patterns on, respectively, p and n type silicon substrates. Using synchrotron radiation and spherical aberration-corrected energy filtering we have obtained spectroscopic image series at the Si 2p core level and across the valence band. Local band alignments are extracted accounting for doping, band bending and surface photovoltage. |
|
|
Five images from the stack of 91 images I(X,YEk) acquired across the Si 2p core-level for the N+/P- sample as a function of binding energies. At high binding energy the contrast inversion over the Si4+ peak from the N+ and P- doped regions is clearly seen, indicating that the Si4+ peaks are shifted as a function of the doping type and concentration. There is a similar but much weaker inversion for the substrate peak. The small squares represent the areas of interest (AOIs) defined for spectral extraction (400×400 nm²). The resulting core level spectra are displayed underneath together with the best least square fits to the data. The colour code identifies the AOIs from which the spectra were extracted. Standard core level fitting criteria were applied to the extracted spectra. After subtraction of a Shirley background function, five component deconvolution was applied, simulating the Si0, Si sub-oxides and Si4+ components. Gaussian line shapes with a 2p spin-orbit splitting of 0.61 eV and a 2:1 branching ratio were used. The full width half maxima used for the Si0, Si+, Si2+, Si3+ and Si4+ were 0.6, 0.9, 1.0, 1.1 and 1.4 eV, respectively. The experimental intensities extracted from AOIs distributed horizontally across the FoV are remarkably constant; the root mean square deviation is less than 5%. The ~2×102 counts per binned pixel give a count rate from each AOI of ~ 2×103. The point X2 residual between the experimental data and the fit is 2.2 (~ 0.1%) for spectra from the N+ regions and 3.0 for spectra from the P- regions. The standard deviation in the binding energy values obtained from spectra from different AOIs was 30 and 50 meV for the N+ and P- regions, respectively. Subtle changes occur in both the positions and the internal weightings of the Si 2p components. The substrate signal is more strongly attenuated over the P- substrate than over the N+ pattern, indicating that the native oxide thickness depends on the doping. Using the core level intensities and applying a simple model of exponential attenuation of the substrate signal due to the finite electron inelastic mean free path we can estimate the thickness of the oxide layer. With a mean free path of the Si 2p electrons (0.75 nm) the N+(P-) oxide thickness is 1.81 (2.60) nm, giving an oxide thickness difference over N+ and P- regions of 0.79 nm. |
|
|
|
Image series were also acquired near the top of the valence band up to the Fermi level. The extracted spectra for N+/P- and P+/N are shown in image 3 with for convenience a reminder of the AOIs used. The Fermi level is marked by a dotted line. Two main features are visible in all spectra: the O 2p orbital at 6-7 eV below the Fermi level and a small but significant contribution from the Si 3p orbitals from the underlying substrate, attenuated by the native oxide, but is still visible. As for the core level spectra, the attenuation is stronger in the spectra extracted from the p doped region than in those extracted from the n type pattern. Extrapolating the upper edges of the O 2p and the Si 3p to zero allows a first determination of the valence band onsets and correlation with the core level results. |
|
|
Authors:
N Barrett (1), L F Zagonel (1), O Renault (2), A Bailly (2)
Institutes:
(1) CEA DSM/IRAMIS/SPCSI, CEA Saclay, 91191 Gif sur Yvette, France, http://iramis.cea.fr/en
(2) CEA-LETI, MINATEC, 17 rue des Martyrs, 38054 Grenoble Cedex 9, France, http://www.leti.fr/
Publication(s) J. Phys.:
Condens. Matter 21 (2009) 000000
|
|
This result has been obtained with :
NanoESCA
|
|
|
|
|
|
