Search

PVD CuMn self-forming barrier (SFB) approach was investigated on 300mm wafers as an alternative to conventional PVD Ta/Cu metallization. Cu fill on very aggressive dual damascene features targeted for beyond 32nm node was evaluated along with integrated electrical and reliability performance on low-k (k=2.6) interlayer dielectric (ILD).

The role of band bending in affecting surface recombination velocity measurements has been evaluated by combining barrier height data with charge-carrier lifetime measurements for Si(111) surfaces in contact with a variety of acidic aqueous electrolytes. Charge-carrier lifetimes and thus surface recombination velocities have been measured by contactless radio frequency photoconductivity decay techniques for long bulk lifetime n-Si(111) samples in contact with 11 M (40% by weight) NH_4F(aq), buffered (pH = 5) HF(aq), 27 M (48% by weight) HF(aq), or concentrated 18 M H_2SO_4. Regardless of the sample history or surface condition, long charge-carrier lifetimes were observed for n-Si(111) surfaces in contact with 11 M NH_4F(aq) or buffered HF(aq). On the basis of previous barrier height measurements, this behavior is consistent with the formation of an electrolyte-induced surface accumulation layer that reduces the rate of steady-state surface recombination even in the presence of a significant density of surface trap sites. A straightforward evaluation of the surface trap state density from the measured surface recombination velocities, S, is thus precluded for such Si/liquid contacts. In contrast, a wide range of S values, depending on the history of the sample and the state of the surface, were observed for n-Si(111) surfaces in contact with 27 M HF(aq). These results in conjunction with previously measured barrier height data indicate that the charge-carrier lifetimes measured for n-Si(111) in contact with 27 M HF(aq) can be directly correlated with the surface condition and the effective surface-state trap density. These conclusions were confirmed by measurements of the apparent S values of n-Si(111) surfaces in contact with various solutions in the presence of the known deep trap, Cu. For Si(111)/HF(aq) contacts, very high (≥920 ± 270 cm s^(-1)) surface recombination velocities were observed when 0.16 mM (10 ppm) Cu^(2+) was in the solution and/or adsorbed onto the Si(111) surface as Cu^0 deposits, whereas low (100 ± 75 or 225 ± 20 cm s^(-1)) apparent surface recombination velocities were measured for Cu-contaminated Si(111) samples in contact with 0.16 mM (10 ppm) Cu^(2+)-containing 11 M NH_4F(aq) or BHF(aq) solutions, respectively.

Open-circuit impedance spectra, channel impedance spectroscopy on solution-gated field-effect devices, and differential capacitance vs potential (Mott−Schottky) measurements were used to determine the energetics of n-Si(111), n-Si(100), and p-Si(111) electrodes in contact with aqueous 11 M (40% by weight) NH_4F, buffered HF (BHF), 27 M (48%) HF(aq), and concentrated (18 M) H_2SO_4. A Mott−Schottky analysis of A_s^2C_(sc)^(-2)-vs-E (where As is the interfacial area, and C_(sc) is the differential capacitance as a function of the electrode potential, E) data yielded reliable barrier heights for some silicon/liquid contacts in this work. Performing a Mott−Schottky analysis, however, requires measurement of the differential capacitance under reverse bias, where oxidation or etching can occur for n-Si and where electroplating of metal contaminants can occur for p-Si. Hence, open-circuit methods would offer desirable, complementary approaches to probing the energetics of such contacts. Accordingly, open-circuit, near-surface channel conductance measurements have been performed using solution-gated n^+-p-Si(111)-n^+ and p^+-n-Si(100)-p^+ devices. Additionally, open-circuit impedance spectra were obtained for silicon electrodes in contact with these solutions. The combination of the three techniques indicated that the surfaces of n-Si(111) and n-Si(100) were under accumulation when in contact with either 11 M NH_4F(aq) or BHF(aq). The barrier heights for n-Si(111) and n-Si(100) in 11 M NH_4F(aq) were −0.065 ± 0.084 V and −0.20 ± 0.21 V, respectively, and were −0.03 ± 0.19 V and −0.07 ± 0.24 V, respectively, for these surfaces in contact with buffered HF(aq). Consistently, p-Si(111) surfaces were determined to be in inversion in contact with these electrolytes, exhibiting barrier heights of 0.984 ± 0.078 V in contact with 11 M NH_4F(aq) and 0.97 ± 0.22 V in contact with buffered HF(aq). In contact with 27 M HF(aq), n-Si(111) and n-Si(100) were in depletion, with barrier heights of 0.577 ± 0.038 V and 0.400 ± 0.057 V, respectively, and p-Si(111) was under inversion with a barrier height of 0.856 ± 0.076 V. Measurements performed in 18 M H_2SO_4 revealed barrier heights of 0.75 ± 0.11 V, 0.696 ± 0.043 V, and 0.889 ± 0.018 V for n-Si(111), n-Si(100), and p-Si(111), respectively, demonstrating that in 18 M H_2SO_4, the band edge positions of Si were different for different doping types. The barrier height data demonstrate that the observed low recombination rates of silicon in contact with 11 M NH_4F, BHF, or 18 M H_2SO_4 cannot necessarily be attributed to a reduction in the number of surface trap states. In part, low surface recombination rates are expected for such systems because the very large or very small barrier height for silicon in contact with these liquids provides a potential barrier that prevents one type of photogenerated carrier (either electrons or holes) from reaching the surface, thereby producing a low steady-state surface recombination rate.

Near-surface channel impedance measurements, open-circuit impedance spectra, and differential capacitance vs potential measurements have been used to determine the barrier height of liquid contacts formed with n-type and p-type Si electrodes. Barrier heights were measured as the redox potential, E(A/A^-), of a metallocene-based, one-electron, outer-sphere, acceptor/donor (A/A^-) pair was varied in CH_3CN solvent. The barrier heights of p-Si(111) electrodes in contact with CH_3CN−Me_(10)Fc^(+/0) (where Me_(10)Fc is decamethylferrocene) or CH_3CN-CoCp_2^(+/0) (where CoCp_2 is cobaltocene) were 0.69 ± 0.1 and 1.1 ± 0.1 V respectively. In contrast, barrier heights for n-Si(111)/CH_3CN−Me_(10)Fc^(+/0) and n-Si(111)/CH_3CN-CoCp_2^(+/0) contacts were 0.66 ± 0.1 and 0.09 ± 0.01 V, respectively. These measurements indicate that the barrier heights closely track changes in the electrochemical potential of the contact, instead of being relatively invariant to changes in the Fermi level of the contacting phase, as is observed for Si/metal Schottky barriers. These measurements also demonstrate that the low effective surface recombination velocity, S, for silicon in contact with CoCp_2^(+/0) is primarily the result of an accumulation layer rather than solely being due to a low density of surface electrical defects.

The electron scattering at surfaces, interfaces, and grain boundaries is investigated using polycrystalline and single-crystal Cu thin films and nanowires. The experimental data is described by a Fuchs--Sondheimer (FS) and Mayadas--Shatzkes (MS) model that is extended to account for the large variation in the specific resistivity of different grain boundaries as well as distinct top and bottom surfaces with different scattering specularity $p$. Textured polycrystalline Cu(111) thin films with thickness $d$ $=$ 25--50 nm are deposited on a stack of 7.5-nm Ta on SiO${}_{2}$/Si(001). Subsequent annealing results in small-grain (SG) thin films with an average grain size $\overline{D}$ that increases from 90 to 120 nm with increasing $d$. Corresponding large-grain (LG) thin films with $\overline{D}$ $=$ 160--220 nm are obtained by depositing 100--200-nm-thick films, followed by an in-situ anneal and a subsequent etch to match the thickness of the SG samples. Nanowires are fabricated from the SG and LG thin films using a subtractive patterning process, yielding wire widths of 75--350 nm. Single-crystal and LG layers exhibit a 18--22$%$ and 10--15$%$ lower resistivity than SG layers, respectively. The resistivity decrease from SG to LG Cu nanowires is 7--9$%$. The thickness and grain size dependence of the resistivity of polycrystalline and single-crystal Cu layers is well described by an exact version of the existing FS $+$ MS model but is distinct from the commonly used approximation, which introduces an error that increases with decreasing layer thickness from 6.5$%$ for $d$ $=$ 50 nm to 17$%$ for $d$ $=$ 20 nm. The case of nanowires requires the FS $+$ MS model to be extended to account for variation in the grain boundary reflection coefficient $R$, which effectively increases the overall resistivity by, for example, 16$%$ for 50 \ifmmode\times\else\texttimes\fi{} 45 nm${}^{2}$ wires. The overall data from single and polycrystalline Cu layers and wires yields $R$ $=$ 0.25 \ifmmode\pm\else\textpm\fi{} 0.05, and $p$ $=$ 0 at Cu-air and Cu-Ta interfaces.