The goal of this work was to optimize the existing test methods for measuring bulk copper and nickel impurities in the heavily doped p-type silicon wafers and to assess their sensitivity, recovery rate, and correlation. Three test methods were studied; low-temperature out-diffusion, polysilicon ultratrace profiling, and wafer digestion. The bulk copper and nickel recovery rates of low-temperature out-diffusion were improved by one order of magnitude in the concentration around 2 x 10(12) atoms/cm(3). Among the three test methods, polysilicon ultratrace profiling and wafer digestion were found to be most sensitive and the method detection limit of 5 x 10(11) atoms/cm(3) or better was achieved after optimization. The optimized wafer digestion and polysilicon ultraprofiling also correlated well across six laboratories with a bias less than +/- 50% for both bulk copper and nickel. (C) 2006 The Electrochemical Society.

For the determination of inorganic and organic anions, the pyromellitic acid (PMA) electrolyte is widely used. The pH adjustment of the self-prepared electrolyte was very challenging to satisfy the pH of specification of pH 7.8 +/- 0.1. A modification was proposed to provide a more simple electrolyte by buffering the PMA electrolyte with triethanolamine (TEA) only instead of adjusting the pH by NaOH and TEA. Thus, the proposed electrolyte consisted of 2.25 mmol l(-1) PMA, 0.75 mmol 1(-1) hexamethonium hydroxide and 12 mmol 1(-1) TEA. The performance of the PMA electrolyte buffered by TEA only was compared to a commercial available PMA and statistically validated in accordance with the methodology of Taguchi. No statistically significant difference could be found for both electrolytes assessing the performance and detection limits of hydrodynamic, stacking and electrokinetic injection with transient isotachophoretic preconcentration as well as repeatability of migration times, peak resolutions and peak symmetries.

Synchrotron radiation induced TXRF allows the nondestructive investigation of low Z contaminations on Si wafer surfaces at trace levels required by the semiconductor industry. The PTB (Physikalisch Technische Bundesanstalt) U180 undulator beamline at BESSY II, equipped with a plane grating monochromator ensuring an energy resolving power E/ΔE between 500 and 5000, can be operated either in wiggler mode for photon energies up to 1.7 keV to excite Al, Mg and Na efficiently, or in undulator mode, i.e. using one of the first odd U180 harmonics, to obtain intensive low energy radiation below 0.7 keV to excite carbon, nitrogen and oxygen. The specific feature of the beamline is its high spectral purity that allows for fundamental investigations. The TXRF wafer chamber of the Atominstitut was used for the experiments with a sidelooking Si(Li) detector with the wafer arranged vertically to take advantage of the linear polarization for background reduction. The energy dependence of the resonant Raman scattering, which is a limiter for the determination of Al at ultra trace levels excited with energies just below the Si absorption edge was studied as well as the influence of the incidence angle on the Raman peak. Droplet samples containing boron were measured and the detection limit of 3 ng determined. A single Carbon layer (5 nm) and a C–Ni–C multilayer sample on a Si wafer were characterized and it was shown that the thickness and density of these layers could be determined.

Total reflection X-ray fluorescence analysis (TXRF) using monochromatized undulator radiation from the synchrotron radiation facility BESSY 11 has been employed to investigate light element impurities on silicon wafer surfaces. Dedicated droplet samples on silicon wafer surfaces were prepared and for the TXRF arrangements employed lower limits of detection (LOD) for C, N, Na, Mg and Al ranging from 0.3 pg to 1.3 pg. Calculated vapor phase decomposition (VPD) detection limits, derived from the assumption that the sample droplets were collected from a 200 mm wafer, ranged from 2 10(7) atoms/cm(2) to about 10(8) atoms/cm(2) for Na, Mg and Al. Using a thin window Si(Li) detector, the resonant Raman scattering effect limiting the LOD of Al was studied, and an initial LOD value for B was found to be 7 ng. In addition, a TXRF investigation of only a few nm thick C-Ni-C structure deposited on a wafer could be performed. In near edge X-ray absorption fine structure (NEXAFS) investigations in conjunction with TXRF analysis, several organic contaminants on silicon wafer surfaces were studied with respect to their speciation. A novel instrumentation for analyzing contamination on 200 mm and 300 mm silicon wafer surfaces by synchrotron based total reflection X-ray fluorescence (TXRF) has been designed by the Physikalisch-Technische Bundesanstalt, This instrumentation is also suited for energy-dispersive X-ray fluorescence (EDXRF) analysis of thin structures deposited on silicon wafers. The most prominent features are a high vacuum load-lock combined with an equipment front end module (EFEM) and a UHV irradiation chamber with an electrostatic chuck (ESC) mounted on an 8-axis manipulator. The whole surface of a 200 mm or a 300 mm wafer can be scanned by monochromatized undulator radiation. The commissioning of the instrumentation was initiated in November 2002 by a TXRF investigation of droplets containing 50 pg to 500 pg Na deposited on a 200 rum wafer, in which promising LOD values for Na were found. Off-line reference TXRF and EDXRF analyses will start in the spring of 2003. For this accumulation time of 100 s, a conservative estimate of the LOD value for Na ranged from 1.0 pg to 1.7 pg depending both on the absolute Na deposition in the droplet and the pile-up probability of O-Kalpha, overlapping with the Na-Kalpha fluorescence radiation. The 500 pg Na spectrum is also plotted in a logarithmic scale in the upper right part of figure 6, clearly showing the actual pile-up events of the O-Kalpha and Na-Kalpha at about 1.5 keV. Taking into account the ratio of the respective Na deposition present in the lateral maximum position to the total Na deposition in the droplets, the LOD values for Na range from about 170 fg to 250 fg.

Total reflection X-ray fluorescence analysis (TXRF) using monochromatized undulator radiation from the synchrotron radiation facility BESSY II has been employed to investigate light element impurities on silicon wafer surfaces. Dedicated droplet samples on silicon wafer surfaces were prepared and for the TXRF arrangements employed lower limits of detection (LOD) for C, N, Na, Mg and Al ranging from 0.3 pg to 1.3 pg. Calculated vapor phase decomposition (VPD) detection limits, derived from the assumption that the sample droplets would have been collected from a 200 mm wafer, range from 2 10(7) atoms / cm(2) to about 10(8) atoms / cm(2) for Na, Mg and Al. Using a thin window Si(Li) detector, the resonant Raman scattering effect limiting the LOD of Al at trace le-vels was studied and an initial LOD value for B was found to be 7 ng. In addition, a TXRF investigation of only a few run thick C-Ni-C structure deposited on a wafer could be performed. In a near edge X-ray absorption fine structure (NEXAFS) investigation in conjunction with TXRF analysis several organic contaminants on silicon wafer surfaces were studied with respect to their speciation.

An abstract is not available for this content so a preview has been provided. Please use the Get access link above for information on how to access this content.

Organic contamination is starting to play an important role in the production and quality control of Si wafers. For the traceability of the source of contamination, information on the chemical binding conditions is very valuable. A near edge X-ray absorption fine structure (NEXAFS) investigation is the natural development of total reflection X-ray fluorescence (TXRF) analysis of the wafer surfaces able to solve the problem of speciation. The plane grating monochromator beamline for undulator radiation of the Physikalisch-Technische Bundesanstalt at the electron storage ring BESSY II, which provides photon energies between 30 eV and 1.9 keV for the specimen excitation, is an ideal excitation source for TXRF-NEXAFS experiments that require a high resolving power and a sufficient photon flux for trace analysis of low Z elements. The contaminants have been diluted and deposited as droplets on wafer pieces thoroughly cleaned after the cutting. The K edges of C, N, O have been examined. Some discrepancies have been found in the analysis of the same compounds in two different beamtimes; molecular orientation is pointed to as the cause for the difference in magnitude of the resonances. The unintentional contamination has been identified as mainly composed of aliphatic chains.

The unique properties of synchrotron radiation, such as high incident flux combined with low divergence, its linear polarization and energy tunability, make it an ideal excitation source for total reflection X-ray fluorescence (TXRF) spectroscopy in order to non-destructively detect trace impurities of transition metals on Si wafer surfaces. When used with a detector suitable for the determination of low energy radiation this technique can be extended to the detection of low-Z elements, such as Al, Na and Mg. Experiments have been performed at SSRL Beamline 3-3, a bending magnet beamline using monochromatic radiation from a double multilayer monochromator. The wafer was mounted vertically in front of the detector, which was aligned along the linear polarization vector of the incoming synchrotron radiation. This configuration allows the detector to accept a large solid angle as well as to take advantage of the reduced scattered X-ray intensity emitted in the direction of the linear polarization vector. A comparison between droplet samples and spin coated samples was done, in order to compare the capabilities of vapor phase decomposition (VPD-TXRF) with conventional SR-straight-TXRF. Detection limits in the range of 50 fg corresponding to 2E10 atoms/cm2 have been obtained for Na. The spin coated samples, prepared from solutions containing an equal amount of Na, Mg and Al showed an unexpected result when performing a scan of the angle of incidence of the incoming X-rays suggesting a different adsorption behavior of the elements in a multielement solution on the wafer surface. The observation of this behavior is important because the spin coating technique is the standard method for the preparation of surface standards in semiconductor quality control. This effect could be characteristic of the Na, Mg, Al solution used, but the angle dependence of the fluorescence signal of a standard should always be investigated before using the standard for calibration of the apparatus and quantification.

TXRF is routinely used and suited to inspect Si wafer surfaces for possible impurities of metallic elements at the level of pg and below. Lightweight, compact sized, high-resolution Silicon drift detectors (FWHM=148 eV at 5.9 keV) electically cooled and with high throughput are ideally as the new spectrometer and for clean room application. A KETEK 5 mm2 Si drift detector was compared with a NORAN 80 mm2 SiLi in a previously commercially available ATOMIKA 8010 wafer analyzer. Results are presented and show that almost the same detection limits for both detector types were achieved analyzing a droplet sample containing 1 ng Ni on a Si wafer. Also, the performance to detect low Z elements like Na, excited with monochromatic Cr Kα radiation in a vacuum chamber was tested and detection limits of 600 pg obtained.

We performed measurements of gettering efficiencies for Cu in silicon wafers with competing gettering sites. Epitaxial wafers (p/p+) boron-doped with a polysilicon back side allowed us to compare p+ gettering with polysilicon gettering. We further measured metal distributions in p+/p- epitaxial test wafers, with the p- substrate wafers pretreated for oxygen precipitation to compare p+ gettering with oxygen precipitate gettering. Our test started with a reproducible spin-on contamination in the 1012 atoms/cm2 range, followed by thermal treatment in order to redistribute the metallic impurity. Wafers were then analyzed by a novel wet chemical layer-by-layer etching technique in combination with inductively coupled plasma mass spectrometry. This led to “stratigraphical” concentration profiles of the impurity, with typical detection limits of 5–10×1012 atoms/cm3. Twenty-five percent of the total Cu contamination in the p/p+/poly wafer was found in the p+ layer, whilst 75% was gettered by the polysilicon. Obviously, polysilicon exhibits a stronger gettering than p+ silicon, but due to the large distance from the front surface, polysilicon was less effective in reducing impurities from the front side of a wafer compared with p+ gettering. An epitaxial layer p+ on top of p- substrates with oxygen precipitates gettered 50% of the total Cu; while the other 50% of the Cu was measured in the p- substrate wafer with oxygen precipitates. Without oxygen precipitates, 100% of the spiked Cu contamination was detected inside the p+ layer. Gettering by oxygen precipitates thus occurs in the same temperature range as that where p+ silicon begins to getter Cu.

Driving forces of gettering reactions can be structured in a cause-effect diagram. For industrial applications, the classification of causes requires quantitative data on gettering efficiencies (GE), which can be provided by analyzing spiked wafers by using the ultra-trace profiling/inductively coupled plasma mass spectrometry (UTP/ICP-MS) method. Segregation gettering of 3d-metals in p/p+ and n/n+ epitaxial wafers was modeled by summarizing the stabilization energies of donors (p+), acceptors (n+) and dopant-metal pairs. Calculations and experimental findings were in quantitative agreement. GE of wafers with poly-silicon backsides were measured for Cu and modeled by a kinetic model within the uncertainty of the measurement. GE of bulk micro defects (BMDs) for Cu and Ni were measured for different thermal treatments and compared with a reaction- limited gettering mechanism, thus, a dependence of the GE on the total inner surface (TIS) of the BMDs. The TIS was reliably obtained by a combination of measured delta Oi values and modeled BMD-sizes. Our results of Ni and Cu gettering agree with a reaction limited mechanism, and the obtained threshold levels agree with the recently published value of Sueoka. et. al. The influence of the contamination method or the crystal growth technique on the gettering efficiencies is understood qualitatively, but still lacks a quantitative model.

Based on experimental findings we set up calculations of numerical modeling of gettering efficiencies for Cu in various silicon wafers. Gettering efficiencies for Cu were measured by applying a reproducible spin-on contamination in the 1012 atoms/cm2 range, followed by a thermal treatment to redistribute the metallic impurity. Subsequently, the wafers were analyzed by a novel wet chemical layer-by-layer etching technique in combination with inductively coupled plasma mass spectrometry. We investigated p/p+ and n/n+ epitaxial wafers with different doping levels and different substrate-doping species. We have also investigated gettering efficiencies of phosphorus-diffused p- and n-type wafers. Heavilyboron doped silicon exhibited a gettering efficiency of ∼100%, while gettering by n+ silicon occurred for doping levels >3×1019 atoms/cm3 only. In another set of experiments we investigated the dependence of the gettering efficiency of p-type wafers with poly-silicon back sides for different cooling rates and Cu spiking levels. A strong dependence on both parameters was found. Cu gettering in p/p+ epitaxial wafers was modeled by calculating the increased solubility of Cu in p+ silicon compared to non-doped silicon taking into account the Fermi-level effect, which stabilizes donors in p+ silicon, and the pairing reaction between Cu and B. Calculated gettering efficiencies were in very good agreement with experimental results. Gettering in n+ silicon was similarly modeled in terms of pairing reactions and the Fermi-level effect. But, for n-type silicon, many experimental uncertainties existed; thus, we applied our expressions to solubility data of Hall and Racette to obtain the unknown parameters. The empirical calculations were in good agreement even with results on n/n+ wafers. For phosphorus-diffused wafers we had to consider an excess vacancy concentration of 1.2–5.5 times the equilibrium concentration to explain the experimental findings by the model. Gettering by poly-silicon back sides was simulated by solving the time-dependent diffusion equation with boundary conditions that take into account different surface reaction rates of silicon point defects. Using this advanced model, the experimentally measured gettering efficiencies were reproduced within the uncertainty of the measurement.

We have performed measurements on the gettering efficiencies for Ni in different silicon wafers. Gettering efficiencies were measured of wafers grown by different crystal-growth techniques, such as Czochralski-grown (CZ) and floating zone (FZ), as well as wafers containing crystal-originated particles (COPs) of different size and density. Lightly boron doped CZ wafers covered with an epitaxial layer were also evaluated. In another set of experiments, we compared different back-side-gettering techniques, like poly-silicon, stacking faults and He-implanted back sides and the dependence of back-side gettering on cooling rate and contamination level. Internal surfaces of oxygen precipitates were also investigated. The gettering test started with a reproducible spin-on contamination in the range around 1012 atoms/cm2 and was followed by a thermal treatment to redistribute the Ni impurity in the wafer. Subsequently, wafers were analyzed for their surface and bulk contamination by a novel layer-by-layer etching, stratigraphical technique in combination with inductively coupled plasma mass spectrometry. No detectable gettering effect of COPs was found. FZ wafers differed remarkably in their gettering behavior from CZ wafers, obviously due to differences in aggregated self-point defects. Most remarkably, the deposition process of an epitaxial layer changed the gettering behavior of p/p- wafers. Comparing the gettering efficiencies of different back sides, an extraordinarily high gettering efficiency of He-implanted voids can be anticipated, which was higher than the gettering efficiency of poly-silicon and stacking faults. High cooling rates at the end of the drive-in cycle and low contamination levels lowered the gettering efficiencies of back-side-gettering techniques, suggesting a diffusion-limited gettering process. Based on the dependence of the gettering efficiencies on different drive-in cycles, a surface reaction as a mechanistic initiation of the drive-in must be assumed. Oxygen precipitates exhibited a high gettering effect for Ni contamination. All experimental results are interpreted by available active surfaces in the gettering phases.

We have measured the gettering efficiencies for Cu and Ni of various silicon wafers, such as MeV-boron-implanted p- polished wafers treated with two different implantation doses of 3×1013 atoms/cm2 B and 1×1015 atoms/cm2 B, respectively. A third kind of wafer was covered with a poly-silicon back side and thermally pretreated before the gettering test to form oxygen precipitates in the bulk. The gettering test started with a reproducible spin-on spiking on the front side of the wafers in the range around 1012 atoms/cm2, followed by a thermal treatment to redistribute the metallic impurities in the wafer. Then the gettering efficiencies were measured by a novel wet chemical layer-by-layer etching technique in combination with inductively coupled plasma mass spectrometry. This led to “stratigraphical concentration profiles” of the metallic impurities in the wafer with typical detection limits of (5–10)×1012 atoms/cm3. The concentration profiles were compared with concentration profiles found after testing the gettering efficiency of p/p+ epitaxial wafers. Almost 100% of the total intentional Cu spiking was recovered in the boron buried layer for both implantation doses. On the front surface and in the region between the front surface and the buried layer a Cu concentration ∼20 times higher than on/in p/p+ epitaxial wafers/layers was measured for the implanted specimen. The lower implantation dose led to higher Cu-concentration levels on the front surface compared to the higher implantation dose. The wafer containing a MeV-boron-implanted layer as well as oxygen precipitates and a poly-silicon back side exhibited a Cu distribution of 30/∼0/70%, respectively. Thus, the gettering by poly-silicon exceeded both the gettering effects by the buried layer and by the oxygen precipitates. Ni gettering in MeV-boron-implanted wafers exhibited other characteristics. The gettering efficiency of the buried layer was 65%, while the remaining Ni contamination was equally distributed between the front-side region and the wafer back side. A wafer containing a buried layer obtained by a 1×1015 atoms/cm3 B dose and oxygen precipitates exhibited 17% of the total Ni contamination in the boron layer, while ∼80% of the total Ni contamination was gettered by oxygen precipitates. In the case of buried layer/oxygen precipitates/poly-silicon back side the distribution was found to be 13/37/45%, thus exhibiting equal gettering strengths for oxygen precipitates and the poly-silicon back side for Ni contamination. The results were discussed in terms of segregation and relaxation-induced gettering mechanismsincluding different reaction rates.

We have measured the gettering efficiencies for Cr, Mn, Fe, Co, Ni and Cu in p/p+ epitaxial wafers. The gettering test started with a reproducible spin-on contamination on the front side of the wafers in the 1012–1014 atoms/cm2 range, followed by thermal treatment to redistribute the metallic impurities in the wafer. The gettering efficiencies were measured by a novel wet chemical stratigraphic etching technique in combination with inductively-coupled plasma mass spectrometry. The residual bulk metal contamination was also measured by this method. This procedure led to global distributions of the 3d elements on the wafer’s front side, in the bulk of the wafer and on the wafer’s back side. Recovery rates were found to be 34%, 2.3%, 100%, 85%, 100% and 100% for Cr, Mn, Fe, Co, Ni and Cu, respectively. An impurity segregation effect in the wafer bulk was measured for Cu (100%) and Cr (34%), while no detectable segregation-induced gettering mechanisms were detected for the other elements in the applied concentration range. The segregation-induced gettering mechanisms were interpreted from the electronic structure of the metallic impurities. For segregation gettering by increased solubility in p+ silicon, the metallic species must form donors. Only Cu+ (3d 10) and Cr+ (3d 5) can form singly positively charged species that exhibit a spherical electronic distribution. It is well known from spinell structures that 3d 10 and, to a smaller extent 3d 5, are stable configurations in tetrahedral structures like the silicon lattice. Thus, we link the segregation-induced gettering mechanism in p/p+ epitaxial wafer to the electronic configuration of the 3d elements.

Several methods are presented for the routine ultra-trace analytical monitoring of inorganic and organic anions and cations on the surface and in the native oxide of silicon wafers – the wafer-surface water-extraction method, the vapor-phase-decomposition method, and the re-dissolving method. Electrokinetic injection, sample stacking, and electrolyte composition were, therefore, optimized and made robust. For electrokinetic injection with transient isotachophoretic preconcentration a linear range of 0.05 to 0.5 µmol L–1 was obtained; for sample stacking the linear range was 0.5 to 10 µmol L–1, even in the presence of up to 750 µmol L–1 hydrofluoric acid. Inorganic anions and monovalent carboxylic acids are predominately dissolved in the aqueous layer on the wafer surface whereas dicarboxylic acids are chemically bonded to the silanol groups and form esters.

Several different total reflection X-ray fluorescence (TXRF) experiments were conducted at the plane grating monochromator beamline for undulator radiation of the Physikalisch-Technische Bundesanstalt (PTB) at the electron storage ring BESSY II, which provides photon energies between 0.1 and 1.9 keV for specimen excitation. The lower limits of detection of TXRF analysis were investigated for some low Z elements such as C, N, O, Al, Mg and Na in two different detection geometries for various excitation modes. Compared to ordinary XRF geometries involving large incident angles, the background contributions in TXRF are drastically reduced by the total reflection of the incident beam at the polished surface of a flat specimen carrier such as a silicon wafer. For the sake of an application-oriented TXRF approach, droplet samples on Si wafer surfaces were prepared by Wacker Siltronic and investigated in the TXRF irradiation chamber of the Atominstitut and the ultra-high vacuum TXRF irradiation chamber of the PTB. In the latter, thin C layer depositions on Si wafers were also studied.

. We have measured the gettering efficiencies for Cu and Ni in p/p-Si epitaxial wafers. The wafers were pretreated to obtain oxygen precipitates of different sizes and densities in the bulk. Gettering tests started with a reproducible spin-on spiking in the range of 10 12 atoms/cm 2 , followed by thermal treatment to drive-in and redistribute the impurities in the wafer. Subsequently, the wafers were analyzed by a novel stratigraphical layer-by-layer etching technique in combination with inductively coupled plasma mass spectrometry. Gettering efficiencies for Ni did not depend on oxygen precipitate sizes and densities as long as ΔO i was larger than 0.2×10 17 atoms/cm 3 and the bulk micro defect densities were detectable by preferential etching (10 7 cm -3 ). In these cases, gettering efficiencies were 96–99% for Ni, while wafers not containing any measurable BMDs exhibited no detectable gettering. Cu exhibited a more complex behavior because the total Cu contamination was found to be divided into two species, one mobile and the other immobile species. A dependence on BMD size and BMD density of the Cu distributions in the wafers was also detected. Gettering effects were increased with increasing BMD densities and sizes. For BMD densities <10 9 cm -3 , Cu was not efficiently gettered by oxygen precipitates. Even for BMD densities >10 10 cm -3 , gettering effects due to oxygen precipitates were one order of magnitude lower than in heavily boron-doped silicon.