II. EXPERIMENT

In crystalline Si and Si_1–xGe_x, the presence of strain causes a shift of the Raman peak (phonon peak) positions and deterioration in crystallinity (due to defects and/or dopant concentration increase), which affects the full width at half-maximum (FWHM) of the Raman peak. Spectral resolution and repeatability of measurement in conventional monochromator-based Raman systems are not suitable for the accurate and precise measurements required for characterizing high quality crystalline semiconductor materials due to certain aspects of their system design, including the use of moving parts. For proper characterization and monitoring of high quality crystalline semiconductor materials, very high spectral resolution (typically <0.1 cm⁻¹ before curve fitting), measurement repeatability, and long-term stability of better than <0.1 cm⁻¹ are required. The requirements of such a Raman system for Si-based semiconductors, and the necessity of designing a polychromator-based, long focal length (2.0 m) multiwavelength Raman spectroscopy system are reported by Yoo et al.Reference Yoo, Ueda, Ishigaki and Kang^4–Reference Yoo, Ueda, Ishigaki and Kang⁶ and Tzeng et al.Reference Tzeng, Ku, Chang, Yang, Chern, Lin, Hasuike, Harima, Ueda, Ishigaki, Kang and Yoo⁷

A multiwavelength imaging Raman spectroscopy (MRS-300) system was designed to overcome the aforementioned deficiencies in commercially available conventional micro-Raman systems.Reference Yoo, Ueda, Ishigaki and Kang^4–Reference Tzeng, Ku, Chang, Yang, Chern, Lin, Hasuike, Harima, Ueda, Ishigaki, Kang and Yoo^7, Reference Yoo, Kang, Ueda and Ishigaki¹¹ Schematic illustration of the MRS-300 system is shown in Fig. 1. Stress/strain and Ge content in Si_1–xGe_x/Si at different depths can be monitored in-line. The MRS-300 was designed with no moving parts and a minimum of optical components for improved stability and repeatability, along with a longer focal length for enhanced spectral resolution. The design concept and performance of the MRS-300 in stressed Si and Si_1–xGe_x/Si characterization applications have been reported by Yoo, et al.Reference Yoo, Kang, Ueda and Ishigaki^11, Reference Yoo, Ueda and Kang¹²

FIG. 1.

Schematic illustration of multiwavelength Raman spectroscopy (MRS-300) system. (color online)

Undoped and B-doped, ultrathin (20~35 nm) Si_1–xGe_x epitaxial layers were grown on 300 mm diameter p ⁻-Si(100) wafers by chemical vapor deposition. The ranges of target Ge content and B concentration in Si_1–xGe_x epitaxial layers were 25 to ~40 at.% and 1.0 × 10¹⁸ ~1.0 × 10²¹ cm⁻³, respectively. The Si_1–xGe_x/Si(100) wafers were characterized using micro-Raman spectroscopy under UV and visible wavelengths (363.8, 457.9, 488.0, and 514.5 nm) of Ar⁺ lasers. The approximate probing depth of these excitation wavelengths in p ⁻-Si is estimated to be ~5, ~290, ~490, and ~645 nm (respectively) based on the light absorption coefficient of Si at these wavelengths. The Raman probing depth was estimated to be one-half of the penetration depth of the excitation wavelength because the Raman signal from the sample attenuates at the same rate as the excitation light, as it escapes from the sample. The probing depth in Si_1–xGe_x becomes shallower as Ge content is increased. In Si_1–xGe_x (x = 0.30), the penetration depth is estimated to be 20% shallower than that in p ⁻-Si. The probing depth dependence of the excitation wavelength for Raman measurements on undoped Si_0.70Ge_0.30 epitaxial layers on Si(100) is illustrated in Fig. 2.

FIG. 2.

Excitation wavelength dependence of probing depth of Raman measurement on undoped Si_0.70Ge_0.30 epitaxial layers on Si(100). (color online)

The Ge content and B concentration in ultrathin Si_1–xGe_x films and the thickness of the Si_1–xGe_x films were measured by SIMS for verification and compared with the Raman measurement results.

III. RESULTS AND DISCUSSION

Raman spectra from ~60 nm thick, undoped Si_0.70Ge_0.30 epitaxial layers on Si(100) and a reference Si(100) wafer in the (150 ~550 cm⁻¹) wave number range under 457.9 nm excitation are shown in Fig. 3. The main peak at 520.3 cm⁻¹ is the Si–Si stretching (optical phonon) mode from the Si wafer. Three other peaks in the ranges of 280~310 cm⁻¹ and 390~450 cm⁻¹, correspond to the Ge–Ge and Si–Ge phonon stretching modes of the Si_1–xGe_x epitaxial layer.Reference Hikavyy, Nguyen, Loo, Ryan, Wormington and Hopkins^1, Reference Alonso and Winer¹³ The shift of the Si–Si peak to the lower wave number side is proportional to the Ge content in the Si_1–xGe_x layer and has been well studied.Reference Hikavyy, Nguyen, Loo, Ryan, Wormington and Hopkins^1, Reference Yoo, Ueda, Ishigaki and Kang^4–Reference Tzeng, Ku, Chang, Yang, Chern, Lin, Hasuike, Harima, Ueda, Ishigaki, Kang and Yoo^7, Reference Alonso and Winer¹³ From the position of the Si–Si peak of Si_1–xGe_x (i.e., shift of the Si–Si peak of Si_1–xGe_x from the Si–Si peak for a Si wafer), Ge content can be estimated. For accurate Ge content measurement, high spectral resolution (~0.1 cm⁻¹ before curve fitting) is required.Reference Yoo, Ueda, Ishigaki and Kang^4–Reference Tzeng, Ku, Chang, Yang, Chern, Lin, Hasuike, Harima, Ueda, Ishigaki, Kang and Yoo⁷ For thin Si_1–xGe_x epitaxial layers, shorter excitation wavelength Raman measurement results in stronger Raman signals from the Si_1–xGe_x epitaxial layers due to the shallower probing depth. Thus, for thorough and complete analysis, high spectral resolution and multiwavelength Raman measurement capability are necessary.

FIG. 3.

Raman spectra from ~60 nm thick, undoped Si_0.70Ge_0.30 epitaxial layers on Si(100) and a reference Si(100) wafer in the wave number range (150~550 cm⁻¹) under 457.9 nm excitation. (color online)

Figure 4 shows Raman spectra from undoped and B-doped Si_1–xGe_x (x = 0.28) under various excitation wavelengths (363.8, 457.9, 488.0, and 514.5 nm). The thickness of Si_1–xGe_x and B concentration of doped Si_1–xGe_x (x = 0.28) were ~60 nm and ~5.0 × 10¹⁹ cm⁻³, respectively. As excitation wavelengths shorten (and probing depth becomes shallower), the Si–Si peak in Si becomes weaker and the Si–Si peak in Si_1–xGe_x becomes stronger. Under the 363.8 nm excitation only the Si–Si peak in Si_1–xGe_x was observed due to the shallow probing depth of the UV light. UV Raman spectra from a reference Si(100) wafer was plotted for comparison. The B-doped Si_1–xGe_x epitaxial layer showed a smaller and broader Si–Si peak compared to that of the undoped Si_1–xGe_x epitaxial layer on Si at all excitation wavelengths. The results are consistent with the previous studies by Pérez-Rodríguez et al.Reference Pérez-Rodríguez, Romano-Rodríguez, Cabezas, Morante, Jawhari and Hunt¹⁴ and Yoo et al.Reference Yoo, Ueda, Ishigaki and Kang¹⁵ Because the Si_1–xGe_x epitaxial layer is relatively thin (~60 nm), shorter excitation wavelength Raman spectra provides a stronger Si–Si peak from the Si_1–xGe_x epitaxial layer and clearer broadening of the Si–Si peak due to B doping. The shorter excitation wavelength gives a better signal-to-noise (S/N) ratio for the thin Si_1–xGe_x epitaxial layer on Si(100) wafers. Appropriate excitation wavelengths must be chosen on the basis of the structure of samples and desired probing depths.

FIG. 4.

Multiwavelength (363.8, 457.9, 488.0, and 514.5 nm) Raman spectra from undoped and B-doped (target B concentration = 5.0 × 10¹⁹cm−3) Si_1–xGe_x (x = 0.28). (color online)

Multiwavelength imaging Raman characterization of B-doped, ultrathin Si_1–xGe_x epitaxial layers on Si(100) wafers was used to find correlation between Raman spectra and B concentration. The thickness of Si_1–xGe_x epitaxial layer was in the range of 20 ~35 nm. Figure 5 shows Raman spectra [(a) as measured and (b) after curve fitting] from ultrathin, B-doped Si_1–xGe_x epitaxial layers with a target B concentration range of 1.0 × 10¹⁸ ~1.0 × 10²¹ cm⁻³ under 363.8 nm excitation. The B concentration measured by SIMS ranged from 7.8 × 10¹⁷ to 9.1 × 10²⁰ cm⁻³. Large intensity variation of the Si peak was observed due to the thickness variations in ultrathin Si_1–xGe_x epitaxial layers and the light absorbance of excitation wavelength at 363.8 nm. The thinner Si_1–xGe_x epitaxial layers resulted in a stronger Si peak from the Si(100) wafer. As B concentration increases, the Si–Si peak from the Si_1–xGe_x epitaxial layers becomes weaker and broader. For Si_1–xGe_x epitaxial layers with B concentration greater than 5.0 × 10¹⁹ cm⁻³, the position of the Si–Si peak from the Si_1–xGe_x epitaxial layer shifted toward the lower wave number side due to strain relaxation. A similar trend was observed under the 457.9 nm excitation Raman measurement. However, the S/N ratio was generally poor due to the weak Si–Si signal from the ultrathin Si_1–xGe_x epitaxial layers. Raman measurements with 488.0 and 514.5 nm excitation did not show reliable measurement of the Si–Si peak from Si_1–xGe_x epitaxial layers due to the nature of the deeper probing depths.

FIG. 5.

UV Raman spectra (a) as measured and (b) after curve fitting from 20~35 nm thick, B-doped Si_1–xGe_x (x = 0.30) epitaxial layers with various B concentrations. The large intensity variation of the Si peak at 520.3 cm⁻¹ is due to the thickness variations in ultrathin Si_1–xGex epitaxial layers and light absorbance at this particular excitation wavelength of 363.8 nm. (color online)

The FWHM of the Si–Si peak from a 20 ~35 nm thick Si_1–xGe_x (x = 0.30 and 0.35) epitaxial layers under 363.8 and 457.9 nm excitation was plotted in Fig. 6 as a function of the B concentration in the Si_1–xGe_x epitaxial layer measured by SIMS. The B doping level was varied from undoped to ~1.0 × 10²¹ atoms/cm³. The FWHM of the Si–Si peak, from ultrathin Si_1–xGe_x epitaxial layers, broadens as B concentration increases, regardless of Ge content. The FWHMs of undoped Si_1–xGe_x epitaxial layers of x = 0.30 and 0.35 are ~4.5 cm⁻¹. At a B concentration of ~5.0 × 10¹⁹ atoms/cm³, the FWHMs of Si_1–xGe_x epitaxial layers for x = 0.30 and 0.35 were ~9.0 cm⁻¹ and 8.7 cm⁻¹, respectively. The FWHM values continued to broaden as the B dopant concentration increased, and almost doubled at a B concentration of ~5.0 × 10¹⁹ atoms/cm³. In this study, the lowest B concentration detected by FWHM broadening was 8.7 × 10¹⁷ atoms/cm⁻³. Data acquisition times range from a few seconds to a few minutes (per point), depending on film thickness, B concentration, and desired signal to noise and productivity tradeoffs.

FIG. 6.

FWHM of the Si–Si peak from 20~35 nm thick Si_1–xGe_x (x = 0.30 and 0.35) epitaxial layers as a function of B concentration in the Si_1–xGex epitaxial layer (as measured by SIMS). FWHM estimation error during curve fitting is less than 0.2 cm⁻¹. (color online)

Both FWHM curves, plotted against B concentration, under 363.8 and 457.9 nm excitation are nearly identical for B concentration up to ~1.3 × 10²⁰ cm⁻³. The 457.9 nm excitation FWHM curve begins to deviate from the 363.8 nm FWHM curve beyond this region. We suspect this is caused by the low S/N ratio of 457.9 nm Raman spectra from ultrathin Si_1–xGe_x epitaxial layers and strain relaxation by heavy B doping. We have verified the strain relaxation in heavily doped (B > 7.8 × 10¹⁹ cm⁻³) Si_1–xGe_x epitaxial layers by HRXRD. The HRXRD rocking curve of SiGe layers with higher B concentration (B > 7.8 × 10¹⁹ cm⁻³) become narrower. This indicates strain relaxation in SiGe and matches well with Raman spectra under longer wavelength (457.9 nm or longer) excitation. Although HRXRD gives information on the average characteristic of the film, the multiwavelength feature of the MRS-300 system allows us to compare the film characteristics at various depths.

HRXRD detects the lattice spacing of the Si_1–xGe_x epitaxial layer on the Si wafer. The thickness and Ge content of the Si_1–xGe_x epitaxial layer are estimated from x-ray rocking curve simulations. For a B impurity to cause a change in lattice spacing, a significant quantity of B atoms must occupy substitution sites in the lattice. Due to the lattice mismatch between the Si_1–xGe_x epitaxial layer and Si wafers, the Si_1–xGe_x/Si interface is strained, and the amount of strain is mainly determined by the Si_1–xGe_x epitaxial layer thickness, Ge content, and crystallinity of the Si_1–xGe_x epitaxial layer. The relatively thick Si_1–xGe_x layers on Si, with high Ge content, experience less strain, depending on crystal quality, and result in a decrease of the lattice constant of Si_1–xGe_x perpendicular to the wafer surface. This often causes Ge content to be estimated lower from HRXRD than the actual Ge content.Reference Zhao, Zuo, Cheng, Yu and Wang³ The presence of strain causes two-dimensional lattice distortion and results in changes in lattice spacing. This change in lattice spacing due to strain is significantly larger than that from typical levels (5.0 × 10¹⁸ ~5.0 × 10¹⁹ atoms/cm³) of B doping. For these reasons, it can be impractical to use HRXRD to quantify B dopant concentration and as an in-line B concentration monitoring technique.

On the basis of the lattice constants of Si (a _Si = 0.543059 nm) and Ge (a _Ge = 0.564613 nm), atomic densities of Si and Ge are calculated to be 5.00 × 10²² atoms/cm³ and 4.42 × 10²² atoms/cm³.Reference Sze¹⁶ For Si_1–xGe_x (x = 0.30), the lattice constant and atomic density is estimated to be 0.5490 nm and 4.835 × 10²² atoms/cm³, respectively.Reference Dismukes, Ekstrom, Steigmeier, Kudman and Beers^{17, 18} The B concentration of 1.0 × 10¹⁸ atoms/cm³ is equivalent to 20.7 ppm. For comparison, the reported HRXRD detection limit is ~2.7 × 10¹⁹ atoms/cm³ (or ~558 ppm) of B concentration in Si_1–xGe_x (x = 0.16). Thus, Raman spectroscopy has a B detection limit that is at least 25 times more sensitive than the B detection limit of HRXRD. In addition, HRXRD requires a relatively large measurement area (~100 μm × 100 μm) and HRXRD can only provide an average value for the entire layer. On the other hand, micro-Raman spectroscopy only requires a measurement area larger than the beam size (1 μm or less) and can provide depth profiling capability when multiple excitation wavelengths are used. To characterize device wafers with multilayer structures, high spatial resolution with depth profiling capability is very useful. The combination of features, including the noncontact and nondestructive nature of micro-Raman spectroscopy, high B concentration sensitivity, high spatial resolution, and depth profiling capability make multiwavelength micro-Raman spectroscopy very attractive for use as an in-line monitoring technique. In-line monitoring examples of Ge content and Si_1–xGe_x epitaxial layer thickness for epitaxial film characterization and Si_1–xGe_x epitaxial process optimization have been reported elsewhere.Reference Tzeng, Ku, Chang, Yang, Chern, Lin, Hasuike, Harima, Ueda, Ishigaki, Kang and Yoo⁷

Considering the penetration (probing) depth (<5.0 nm) of the 363.8 nm excitation wavelength in Si_1–xGe_x (x = 0.30) and laser beam diameter <0.5 μm, the probing volume and total number of atoms in that probing volume, can be estimated to be <3.9 × 10⁻¹⁵ cm³ and <1.89 × 10⁸ atoms, respectively. Thus, for B concentration of 1.0 × 10¹⁸ cm⁻³ in Si_1–xGe_x (x = 0.30), the total number of B atoms, in the probing volume of 363.8 nm excitation, is ~3900 atoms. The presence of only ~3900 B atoms (~20 ppm) can cause broadening of the Si–Si Raman peak (distribution of optical phonon vibrations) of Si_1–xGe_x (x = 0.30) layer and result in an increase in FWHM. For thicker Si_1–xGe_x epitaxial layers, a combination of UV and visible wavelength excitations for Raman measurements can provide noncontact, nondestructive depth profile information for B concentration, Ge content, strain, and crystallinity.

IV. SUMMARY

We have found a very strong correlation between B content and the nature of the Si–Si Raman peak from Si_1–xGe_x epitaxial layers. As B concentration increased from undoped to ~1.0 × 10²¹ atoms/cm³, the FWHM of the Si–Si Raman peak widened and the peak height became smaller, for a given Ge content. Boron concentration of ultrathin Si_1–xGe_x films, determined by noncontact micro-Raman spectroscopy, was in good agreement with SIMS analysis results. Boron concentration as low as 8.7 × 10¹⁷ atoms/cm³ can be detected by micro-Raman spectroscopy, which is ~30 times lower detection limit than the detection limit (2.7 × 10¹⁹ atoms/cm³) of HRXRD. Multiwavelength Raman measurements are extremely useful as an in-line characterization technique for depth profiling of B concentration, Ge content, strain, and crystallinity of Si_1–xGe_x epitaxial layers. With its unique advantages, multiwavelength Raman spectroscopy has an important place among characterization techniques such as HRXRD, SE, AES, SIMS, and PL.