November_EDFA_Digital

edfas.org 13 ELECTRONIC DEVICE FAILURE ANALYSIS | VOLUME 22 NO. 4 behind due to the difficulty in making low resistance contacts to n-type Ge. [8] Contact resistance is a strong functionof the carrier concentration at the semiconductor surface, and it is extremely difficult to achieve high n-type dopant activation in Ge. Point defects in Ge are known to behave as acceptor sites and reduce the overall electron concentration through compensation, particularly near the surface. [9] Minimization of these defects is essential to achieving high performance Ge NMOS which makes the process sensitive to the implant and anneal processing steps. Co-doping has been shown to reduce the defect density and enhance dopant activation. [10] In a study employing DHEM, the impact of co-doping was directly measured and characterized. Two sets of samples were prepared using epi-Ge layers (3 μm thick) on Si wafers. For the first set a Ge layer was implanted with only P (dose: 6x10 14 cm -2 , energy: 90 keV). For the second set, Ge was implanted with Sb (dose: 6x10 14 cm -2 , energy: 65 keV) followed by P (dose: 6x10 14 cm -2 , energy: 90 keV). Both samples were annealed at 500°C for 10 s in N 2 to activate the dopants. DHEM carrier concentration profiles were determined after direct measurements of sheet resistance and Hall mobility values as described before. As shown in Fig. 3a, for the Ge sample with only P-doping the active dopant concentration was ~5x10 18 cm -3 at the surface despite the P chemical concentration being ~1x10 20 cm -3 as measured by SIMS. This indicated that only ~5% of the implanted dopants were electrically active. For the co-doped sample, however, surface carrier concentration values increased to ~2x10 19 cm -3 with a similar P chemical concentration, as seen in Fig. 3b. [11] It should be noted that prior work by others had shown similar increase in activation with co-doping but did not capture the significant decrease in the electron concentra- tion by 4x within the top 10 nm of the surface region. [10] The measured dip in active carriers in the DHEM profiles of both samples strongly suggested that the Ge substrate was still highly defective even after co-doping. The sub-nm depth resolution of DHEMallowed accuratemeasurement of differences incarrier concentrationvalueswithin the top 10 nmof the surface that could not be resolved previously in co-doped n-type Ge films. In a follow-up study, DHEM was used to analyze the impact of SiO 2 and Al 2 O 3 caps on co-doped samples. Dielectric caps are typically used to prevent dopant out- diffusion during implant and anneal process steps and should ideally not interfere with the doping and activa- tion processes apart from containing the dopants. It was found that contact resistivity (ρ c ) on co-doped samples processed with Al 2 O 3 showed very high values compared to samples processed with SiO 2 caps. DHEM data showed differences in surface carrier density values, where the active dopant concentration decreased by an order of magnitude from ~2x10 19 cm -3 for the SiO 2 capped sample to ~2x10 18 cm -3 for the Al 2 O 3 capped sample. Correlation of poor contact resistivity with decreased dopant activa- tion lead to a search for additional contamination sources that would decrease such activation, and as a result it was found that Al had diffused into Ge from the Al 2 O 3 cap after dopant implants and anneal. It is important to note that DHEM is able to obtain depth profiles of electrical param- eters of semiconductor layers under thin insulating films alsowithout the need to remove such insulating films. The data partially summarized above will be presented at the upcoming 237th ECS Meeting. [12] Fig. 3 DHEM (carrier concentration and resistivity) and SIMS (chemical concentration) depth profiles in; a) P-doped Ge and, b) Sb + P co-doped Ge. Resistivity is shown on the right axis in both plots. (b) (a)