Non-contact monitoring of Ge and B diffusion in B-doped epitaxial Si_1-xGe_x bi-layers on silicon substrates during rapid thermal annealing by multiwavelength Raman spectroscopy Open Access

For performance enhancement of Si-based complementary metal-oxide-semiconductor (CMOS) devices, device scaling has been the most important principle for more than last three decades. Following the 90 nm technology node, performance enhancement by standard device scaling alone became nearly impossible without introducing new materials and new device structures.^1–5 

Carrier mobility of [110] channel in Si (100) substrates can be modulated by uniaxial strain along the [110] direction.^3,5 Electron and hole mobilities improve under tensile and compressive uniaxial stress, respectively. To improve hole mobility in p-channel devices, Ge is embedded to form Si_1-xGe_x in the source/drain (S/D) region. For tensile stressor material for electron mobility enhancement in n-channel devices, C-doped Si is considered to be one of the most promising candidate materials.^6,7 Since the performance of CMOS devices is limited by hole mobility of p-channel devices, significant emphasis has been placed on stressor material (Si_1-xGe_x) engineering for p-channel devices. Optimization of Ge content, film structure, and dopant levels in the Si_1-xGe_x layer, is being pursued in developing next generation devices.^1–6 

Undoped and B-doped Si_1-xGe_x layers are epitaxially grown on Si (100) substrates in the early stage of device fabrication. They are typically exposed to subsequent thermal processing steps with significantly higher temperature than their epitaxial growth conditions. Understanding of thermal behavior of Ge and B in B-doped Si_1-xGe_x layers is very important in designing appropriate device structures.^6,8–11 Nondestructive monitoring of Ge content and B concentration, before and after subsequent high temperature process steps, would be very beneficial for device structure optimization and early detection of potential process related problems. Characterization of Si_1-xGe_x epitaxial layers using conventional techniques such as cross-sectional transmission electron microscopy (XTEM), Auger electron spectroscopy (AES), secondary ion mass spectroscopy (SIMS), double crystal X-ray diffraction and photoluminescence (PL) has been reported by Lafontaine et al.¹² The reported techniques are invasive and time consuming, and not suitable for real time process monitoring which is a theme of this work.

In this paper, feasibility of non-contact, in-line monitoring of Ge and B diffusion behaviors of undoped and B-doped Si_1-xGe_x layer(s) on Si (100) device wafers during rapid thermal annealing (RTA) steps was evaluated using multiwavelength micro-Raman spectroscopy. The Raman characterization results were compared with Ge and B depth profiles from secondary ion mass spectroscopy (SIMS) for verification of effectiveness of multi-wavelength mirco-Raman spectroscopy as a viable, non-contact, in-line Ge and B monitoring technique

Undoped and B-doped Si_1-xGe_x layers on 300 mm Si (100) device wafers were epitaxially grown by low pressure chemical vapor deposition (LPCVD). The epitaxial temperature and pressure were 710°C and 1300 Pa, respectively. Single layer and bi-layer Si_1-xGe_x thin epitaxial films with various Ge content, B concentration and thicknesses were prepared. Ge content was varied between 13 ∼ 35 at%. B doping was done up to 1.4 × 10²⁰ cm^-3 by adding GeH₄ gas during epitaxial growth. The thickness of a single Si_1-xGe_x layer was varied from 45 ∼ 60 nm. The thickness of bi-layer Si_1-xGe_x was varied between 60 ∼ 85 nm (top layer: 15 ∼ 25 nm, bottom layer: 45 ∼ 60 nm). Figure 1(a) shows the micro-Raman and SIMS measurement locations on one of the monitor pads (denoted as +) and micro-Raman line scan locations from monitor pads A to B. Cross-sectional illustration of measurement locations on a typical sample is illustrated in Fig. 1(b).

Wafers were annealed using a commercially available RTA system at Taiwan Semiconductor Manufacturing Company. RTA temperature was varied from 950°C to 1050°C in 50°C intervals. Annealing time was fixed at 60 s. To prevent adverse effects of oxidation, annealing was done in O₂ free, 100% N₂ ambient.

Micro-Raman measurements were made at the center of monitor pads (50 μm × 80 μm) under excitation wavelengths from ultra-violet (UV) to visible (363.8, 441.6, 457.9, 488.0 and 514.5 nm). Probing depth increases with increasing excitation wavelength which enables virtual depth profiling in the multiwavelength micro-Raman measurements. Micro-Raman measurement area is approximately 1 μm diameter. Raman measurement was mainly done using a polychromator-based, high resolution multiwavelength micro-Raman spectroscopy system with a focal length of 2.0 m from WaferMasters, Inc. (Fig. 2)^13–18 The polychromator-based Raman system enables multi-wavelength Raman measurements at the same measurement site without time consuming and unreliable system calibration between excitation wavelength switching. However, Raman measurement wavenumber ranges of individual excitation wavelengths are prefixed in narrow ranges of interest (typically 450 cm^-1 ∼ 550 cm^-1) for measurement accuracy and repeatability.

A stand alone, conventional monochromator-based, medium resolution Raman spectroscopy system with a focal length of 0.8 m was also used for wide range (100 cm^-1 ∼ 550 cm^-1) Raman measurements to locate the Ge-Ge peak (∼200 cm^-1), Si-Ge peak (∼400 cm^-1) and Si-Si peak (∼500 cm^-1) from the Si_1-xGe_x layer and the Si-Si peak (∼520 cm^-1) from the Si wafer. The wide range measurement requires scanning and rotation of a diffraction grating inside the monochromator. Backslash of mechanical components used in the monochromator induces positioning errors greater than measurement accuracy required for Si_1-xGe_x film characterization even with frequent wavenumber calibrations, thus was not used for characterization of spectra. The more accurate, polychromator-based Raman measurement data were mainly discussed in this paper.

Ge and B depth profiles were measured at the center of monitor pad A (denoted as + in Fig. 1(a)) by SIMS after micro-Raman measurements for verification and correlation purposes. Thermal diffusion characteristics of Ge and B atoms within Si_1-xGe_x bi-layers and Si_1-xGe_x/Si interfaces were characterized before and after RTA for 60s under various temperatures.

Since the thermal diffusion behavior at the Si_1-xGe_x/Si interfaces of single and bi-layer Si_1-xGe_x/Si wafers were quite similar, experimental results on Si_0.85Ge_0.15 (B: ∼1.4 × 10²⁰ cm^-3, ∼20 nm)/Si_0.70Ge_0.30 (B: ∼4.5 × 10¹⁹ cm^-3, ∼50 nm) / Si(100) wafers (see Fig. 1(b)) are mainly discussed throughout this paper. Results from single layer Si_1-xGe_x on Si wafers are briefly described towards the end.

Multiwavelength micro-Raman spectra, measured from as-grown and annealed bi-layer Si_1-xGe_x/Si wafers, are summarized in Fig. 3. Annealing temperature dependence (a) and the excitation wavelength dependence (b) of multiwavelength Raman spectra are shown. The Si-Si peak positions of Si_0.85Ge_0.15, Si_0.70Ge_0.30 and Si substrate were indicated on the top horizontal axis. All Raman spectra measured under 363.8 nm excitation showed a Si-Si peak from the Si_1-xGe_x layer, since its probing depth is shallower than the total thickness of the Si_1-xGe_x bi-layer (∼75 nm). As-grown Si_1-xGe_x bi-layers showed two Si-Si peaks corresponding to two Si_1-xGe_x layers with different Ge content. Si_1-xGe_x epitaxial layer on Si pads produces a tensile strain in the Si substrate, resulting in a shift in the Si-Si peak to lower wavenumbers. The shift is larger for higher Ge concentrations. The small peak at ∼516.5 cm^-1 (denoted as inverted green triangle) is corresponding to Ge content of ∼13.7 at%. The second broad peak at ∼510.0 cm^-1 (denoted as inverted pink triangles) corresponds to Ge content of ∼29.4 at%. The first and the second Si-Si peaks from the as-grown Si_1-xGe_x bi-layer correspond to the Ge content of the top and bottom Si_1-xGe_x layers. The heavy B-doping in the Si_1-xGe_x layer both weakens and broadens Si-Si peak. The first broad peak of the as-grown Si_1-xGe_x layers on Si at ∼516.5 cm^-1 is the result of heavy B-doping (1.4 × 10²⁰ cm^-3) in the top Si_1-xGe_x layer, with a nominal Ge content of 15 at%. The second broad Si-Si peak from the bottom B-doped (4.5 × 10¹⁹ cm^-3) Si_1-xGe_x layer, with a nominal Ge content of 30 at%, was measured through the top Si_1-xGe_x layer and has more modest intensity. After RTA, all wafers showed one strong Si-Si peak at ∼509.1 cm^-1 regardless of annealing temperature. The peak became very strong, and its position was slightly shifted from 510.0 cm^-1 to 509.1 cm^-1, towards lower wavenumbers. The change of the Si-Si peak before and after RTA suggests significant B diffusion, from the top Si_1-xGe_x layer, and slight increase in average Ge concentration within the probing depth of the 363.8 nm excitation laser beam. Decrease of B concentration in the top Si_1-xGe_x layer by thermal diffusion during annealing, results in significant increase of Si-Si peak height (SIMS profiles shown later).

For Raman spectra below 441.6 nm, the Si peak from the Si substrate becomes dominant due to deeper probing depths. As the excitation wavelength gets longer, the relative intensity of Si-Si to the Si peak intensity gets smaller. Under 441.6 nm excitation, the Si peak is slightly (∼0.2 cm^-1) shifted towards lower wavenumbers from the stress free Si peak of 520.3 cm^-1, indicating that the Si immediately below the Si_1-xGe_x layer is slightly under tensile stress (∼87 MPa). It agrees well with general expectation. As-grown Si_1-xGe_x bi-layer on Si showed weak, broad Si-Si peaks (vertical dotted line and inverted green triangle) corresponding to the top and bottom Si_1-xGe_x layers.

Comparison of spectra behavior from wafers annealed at different temperatures gives insight into diffusion taking place at common annealing temperatures. The wafer annealed at 950°C for 60 s shows two Si-Si peaks at the same positions (see the dotted red line and dotted green line) as the as-grown wafer. The major Si-Si peak from the thick bottom Si_1-xGe_x layer became stronger due to significant B diffusion from the thin (∼20 nm thick) top Si_1-xGe_x layer. For the wafer annealed at 1000°C, the major Si-Si peak, from the thick bottom Si_1-xGe_x layer, became even stronger suggesting further B diffusion within Si_1-xGe_x layers. The wafer annealed at 1050°C showed the Si-Si peak intensity decreased and its peak position shifted towards higher wavenumbers (denoted as an inverted red triangle in Fig. 3) by 1.0 cm^-1 (from the as-grown value of 509.0 cm^-1 to 510.0 cm^-1 after RTA). This suggests the decrease of average Ge content and increase of B concentration in the bottom Si_1-xGe_x layer, due to Ge diffusion from the bottom Si_1-xGe_x layer and B diffusion from the top Si_1-xGe_x layer into the bottom Si_1-xGe_x layer.

For Raman excitations at 457.9 nm and longer, the trends are quite similar except for the relative intensity of Si-Si peaks to the Si peak and the disappearance of the Si-Si peak from the heavily B-doped, thin (∼20 nm thick) top Si_1-xGe_x layer. If we can verify and correlate the change in shape and position of the Si-Si peak with the degree of Ge and B diffusion after RTA, multiwavelength Raman measurement can be a very useful, non-destructive Si_1-xGe_x layer monitoring technique in manufacturing environments.

Figure 4 shows 457.9 nm excitation line scan spectra between monitor pads A and B of as-grown and annealed Si_1-xGe_x layers on Si(100) wafers. The line scan was started 10 nm from the edge of monitor pad A and ended 10 nm into monitor pad B (see Fig. 1(a)). A twenty-six (26) point line scan was done in 1 μm intervals. Deconvoluted Si-Si (pink) and Si peaks (blue) of the first spectra were plotted in front of the as-measured spectra for easy comparison. As seen in all line scan summaries, the boundary between monitor pads A and B only shows a single peak (blue tall peaks in the middle) from the Si substrate due to the lack of a Si_1-xGe_x layer.

As-grown Si_1-xGe_x/Si showed a very weak and broad Si-Si peak along with a moderate Si peak. The weak and broad Si-Si peak is due to heavy B-doping in the top Si_1-xGe_x layer. After RTA, all wafers showed significant intensity increase and sharpening of the Si-Si peak indicating B diffusion from the heat treatment. The intensity of the Si-Si peak almost doubled (increased from ∼500 counts to ∼1000 counts) after RTA, while the Si-Si FWHM narrowed to 6.1∼6.2 cm^-1 compared to the as grown value of 8.8 cm^-1. The wafers annealed at 950°C and 1000°C showed stronger Si peaks between monitor pads due to crystallinity improvement. The wafer annealed at 1050°C showed slight decrease of Si-Si intensity and slight shift towards higher wavenumbers (not obvious from the graph due to three dimensional plotting). The decrease of intensity and the peak shift towards higher wavenumbers are due to the decrease of average Ge content and increase of B concentration in the bottom Si_1-xGe_x layer due to Ge outdiffusion from the bottom Si_1-xGe_x layer and B diffusion from the top Si_1-xGe_x layer into the bottom Si_1-xGe_x layer. The Si peak between monitor pads slightly decreased. Lateral B diffusion is speculated to be one of the causes for the decrease of Si peak intensity. Multiwavelenth micro-Raman scans are sufficiently sensitive to detect the Ge and B redistribution from as-grown material in subsequent process steps, and, as such, can be used as an in-line monitoring technique.

The same wafers were tested using a conventional monochromator-based Raman spectroscopy system with a focal length of 0.8 m. The general shape of Raman spectra remained the same as the ones measured by the polychromator-based, multiwavelength high resolution Raman spectroscopy system. However, detailed features (denoted as inverted green triangles from thin heavily B-doped Si_1-xGe_x top layer in Fig. 3) were missing due to the limited spectral resolution (2.5 times lower spectral resolution than in the 2.0 m focal length system). The conventional system requires wavenumber calibration using a reference Si sample or plasma lines between excitation wavelength switching. This frequent system calibration makes Raman measurement on the exact same measurement location uncertain and very difficult and causes productivity to be very low for in-line monitoring applications. The variation of Raman peaks from the reference Si sample often exceeds 1 cm^-1 even with frequent system calibration. The uncertainty of 1 cm^-1 in Raman shift is equivalent to ∼3 at% Ge content conversion error for fully strained Si_1-xGe_x/Si and a 434.5 MPa Si stress conversion error.^14,19,20 The sensitivity of Raman shift on Ge content and B concentration in Si_1-xGe_x is easily lost, especially forgoing depth analysis with fixed excitation wavelength operation. A properly designed Raman system makes meaningful measurements difficult for in-line monitoring applications.^17,18

Figure 5 shows linear and semi-log SIMS depth profiling results on Ge and B atoms of as-grown and annealed B-doped Si_1-xGe_x bi-layers on Si (100) wafers. The linear plots show the details of Ge and B diffusion behavior, whereas the conventional semi-log plots appear abrupt for Ge and B dopant and diffusivity. As-grown Si_1-xGe_x bi-layer on Si(100) has total Si_1-xGe_x thickness of ∼75 nm (Si_0.85Ge_0.15 (B: ∼1.4 × 10²⁰ cm^-3, ∼20 nm) / Si_0.70Ge_0.30 (B: ∼4.5 × 10¹⁹ cm^-3, ∼50 nm) / Si(100) as illustrated in Fig. 1(b)). Almost no Ge diffusion was found at 950°C RTA. As RTA temperature increases, Ge diffusion becomes significant.^8,9,21 While Ge diffusion slightly increases with increasing RTA temperature, significant B diffusion took place, even at 950°C RTA.^9–11,22 B concentration in the top Si_1-xGe_x layer was decreased by ∼43% after 950°C RTA for 60 s. Boron piled up at the surface and at the interface between the top and bottom Si_1-xGe_x layers. The B-enhanced region, in particular the interface between the top and bottom Si_1-xGe_x layers, becomes the source of B diffusion. As RTA temperature increases, B continues to redistribute by thermal diffusion, the B concentration of the top Si_1-xGe_x layer becomes lower than that of the bottom Si_1-xGe_x layer. The B concentration of the bottom Si_1-xGe_x layer ended up higher than the as-grown B concentration. The B concentration was inverted from the as-grown (designed) profile. It cannot be explained by a simple thermal diffusion model. It may be due to the difference in solid solubility of B in Si_1-xGe_x alloys, with different Ge content. It is known that the solid solubility of B in Si_1-xGe_x increases with Ge content.²³ The interaction between Ge and B plays an important role in depletion of B in the top Si_1-xGe_x layer, with the highest initial B concentration. It behaves like a capillary reaction. B concentration in Si_1-xGe_x not only affects resistivity, but also changes strain in the Si_1-xGe_x layer.²⁴ In any event, modulation of Ge and B profiles after thermal exposure, during subsequent process steps, may totally change actual device structure from the as-designed structure and may cause significant deviation from the expected device performance.⁹ The results of the SIMS measurements and the multiwavelength micro-Raman analysis showed excellent general agreement. The relevance of this result will be discussed later.

Thermal diffusivity of Ge in Si is approximately two orders of magnitude smaller than that of B in the temperature range of 950°C ∼ 1050°C.^25–27 Figures 5(a) and 5(b) show the Ge thickness increase and B depth profile change before and after RTA, in linear scale. Due to the small diffusituvity of Ge in Si, the Ge profile did not result in visible change compared to the B profile. Figures 5(c) and 5(d) show the semi-log Ge and B depth profiles, before and after RTA. From the slope of Ge and B in the Si region, abruptness and diffusivity of Ge and B in the Si region can be compared.

The abruptness and diffusivity of Ge and B in Si, were estimated and summarized in Fig. 6. To estimate the abruptness of Ge and B profiles, atomic concentrations of Ge 1.0 × 10¹⁹ cm^-3 and B 1.0 × 10¹⁸ cm^-3 were used (Fig. 6(a)). Both Ge and B profile abruptness of as-grown Si_1-xGe_x/Si interfaces were 4.5 nm/decade (from 1.0 × 10¹⁹ cm^-3 to 1.0 × 10¹⁸ cm^-3). The slope of the B depth profile changed at a Ge concentration of ∼1 at%, regardless of RTA temperature, in the range of 950°C to 1050°C. The maximum B concentration in Si was ∼1.0 × 10¹⁹ cm^-3 which is lower than the solid solubility (9 × 10¹⁹ ∼ 4 × 10²⁰ cm^-3) of B in Si in the temperature range of 950°C to 1050°C.²⁸ The maximum B concentration at the Si_1-xGe_x/Si interface is balanced between the B injection rate from B-doped Si_1-xGe_x layer into the Si and the B diffusion rate in Si. As RTA temperature is increased, the Ge and B abruptness is deteriorates. While the rate of Ge abruptness degradation is moderate, the rate of B abruptness degrades several times faster than that of Ge. To calculate the activation energy of Ge and B diffusion, the diffusivity, after RTA for 60 s in the range of 950°C ∼ 1050°C, of Ge and B from as-grown Si_1-xGe_x/Si, were plotted (Fig. 6(b)). For estimating diffusion length of Ge and B profiles, atomic concentrations of Ge 10 at % and B 1.0 × 10¹⁸ cm^-3 were used. The activation energy for Ge diffusion in this temperature range is estimated to be 10.0 eV. The rates of change for B diffusivity in two 50°C intervals (950°C ∼ 1000°C and 1000°C ∼ 1050°C) were different. The activation energy of B diffusion was increased from 1.0 eV (950°C ∼ 1000°C) to 6.0 eV (1000°C ∼ 1050°C) suggesting different diffusion mechanisms take place. The average activation energy of B diffusion in the temperature range of 950°C ∼ 1050°C is estimated to be ∼ 3.4 eV, which is similar to the reported values.^29,30 As seen in Figs. 5(c) and 5(d), B atoms initially diffuse from the abrupt Si_1-xGe_x/Si interface with a high Ge content (x ∼0.3) and then diffuse into the Ge-free Si wafer. The diffusivity of B in Si_1-xGe_x is strongly influenced by Ge content, presence of strain and crystallinity imperfection.^28,29 The B diffusivity reduction with increasing Ge content has been reported.³⁰ In the early stage of B diffusion through the Si_1-xGe_x/Si interface, the Ge content and strain levels are constantly changing as the diffusion process progresses. Once B atoms pass the diffused Ge layer, they then face a different host material (Si instead of Si_1-xGe_x). Very complex B diffusion and segregation processes are to be expected.

The Ge and B depth profiles of B-doped, single layer Si_1-xGe_x/Si(100) after RTA simply follow typical Ge and B diffusion characteristics near the Si_1-xGe_x/Si interface and Si substrate (60 nm ∼ 200 nm depth of Figs. 5(a) ∼ 5(d)). Increase in B dopant concentration, seen in the B-doped Si_1-xGe_x bi-layers on Si(100), was not observed. Absence of an additional B source to cause B injection into the Si_1-xGe_x layer made B diffusion from B-doped Si_1-xGe_x layer into Si very simple and predictable As seen in Figs. 5 and 6, B diffuses 1.5 ∼ 3.5 orders of magnitude faster than Ge in Si_1-xGe_x over the temperature range of 950 and 1050°C. Since the intensity of the Si-Si peak of B-doped Si_1-xGe_x layers increases significantly (and FWHM decreases) after annealing at 950°C for 60 s, the increase of intensity and narrowing of the Si-Si peak of B-doped Si_1-xGe_x layers, after annealing, can be interpreted as B redistribution in the doped Si_1-xGe_x layers and/or B diffusion into the Si substrate. The change of intensity and FWHM of the Si-Si peak from B-doped Si_1-xGe_x layers can be used as an effective monitoring signal for B redistribution and diffusion.

Raman study of C⁺-implanted Si_1-xGe_x and Si_1-x-yGe_xC_y layers by M.Ya. Valakh et al.^31,32 reported shifts in the wavenumber (stress relaxation) with annealing in the temperature range of 600 ∼ 1050°C for 10 min. The C⁺-implanted Si_1-xGe_x and Si_1-x-yGe_xC_y layers have significant density of threading dislocations and misfit discloations after annealing 900°C. In contrast, no significant shift from the B-doped Si_1-xGe_x double layer on Si was observed after annealing. This is due to the presence of a significant amount of fast diffusing B atoms in the Si_1-xGe_x double layers and relatively short thermal exposure, compared to the previous reports (60 s versus 10 min). The fast diffusing, small size B atoms move freely within, and between, layers under elevated temperatures and slows the generation of dislocations and the strain relaxation. As a result, no significant strain relaxation was observed.

As stated earlier, complex Si_1-xGe_x structures with high B concentration on Si(100) substrates, used in advance CMOS devices (45 nm technology node and beyond), must be monitored in-line to prevent unexpected material property variations. In this work, the different techniques of multiwavelength Raman and SIMS were used together for their respective capabilities. Raman spectra from various Si_1-xGe_x/Si structures showed good agreement with highly quantitative SIMS analysis results. Multiwavelength Raman spectra provide additional insights into material properties in depth directions (i.e. virtual depth profiling) and the relative ease and speed of measurements, along with the non-contact, non-destructive nature of the analysis, makes the use of this technique for process uniformity/repeatability and material homogeneity, as well as production monitoring, feasible. In comparison, SIMS provides excellent quantitative analysis, but is destructive and much more time consuming. Together, they can be used very effectively, depending on the needs and circumstances. Multiwavelength micro-Raman spectroscopy is a very promising non-contact, in-line monitoring technique for Si_1-xGe_x/Si, used in advanced devices.

In an effort to develop a non-contact, in-line monitoring technique for process-induced Ge and B concentration variations in Si_1-xGe_x, multiwavelength micro-Raman characterization was done on B-doped Si_1-xGe_x bi-layers on Si(100) device wafers. The wafers were then characterized by SIMS for verification of Raman characterization results. Ge content and B concentration of two Si_1-xGe_x layers were varied. As-grown Si_1-xGe_x bi-layer on Si(100), with a Si_1-xGe_x thickness of ∼75 nm (Si_0.85Ge_0.15 (B: ∼1.4 × 10²⁰ cm^-3, ∼20 nm)/Si_0.70Ge_0.30 (B: ∼4.5 × 10¹⁹ cm^-3, ∼50 nm)/Si(100) were used as an example for this study. The wafers were heat treated by RTA in the temperature range of 950°C to 1050°C in 50°C intervals to study the thermal diffusion behavior of Ge and B atoms in subsequent thermal exposures of downstream process steps. Multiwavelength micro-Raman and SIMS measurements were performed to identify and monitor Ge and B redistribution during RTA. Significant B diffusion was observed after rapid thermal annealing as low as 950°C. Noticeable Ge diffusion was observed after RTA at 1050°C. Pile up of B atoms at the surface and at the boundary between Si_1-xGe_x bi-layers was observed in the early stages of thermal diffusion, suggesting complex diffusion and segregation processes. Multi-wavelength micro-Raman spectroscopy is found to be very sensitive to the change in Ge content and B concentration in thin Si_1-xGe_x layers on Si wafers. It is proving to be a very promising and novel, non-contact, in-line monitoring technique for process-induced Ge and B concentration variations in thin epitaxial Si_1-xGe_x/Si, used in advanced mobility-enhanced devices.

The authors would like to thank Chih-Mou Huang, Ruey-Lian Hwang and Fu-Yuan Tseng of the TSMC Failure Analysis Division (FAND) for SIMS analysis. The authors also would like to thank Prof. Tzong-Yow Tsai of National Cheng Kung University for assistance in this project.