Improvement of Fermi-Level Pinning and Contact Resistivity in Ti/Ge Contact Using Carbon Implantation

Abstract

Effects of carbon implantation (C-imp) on the contact characteristics of Ti/Ge contact were investigated. The C-imp into Ti/Ge system was developed to reduce severe Fermi-level pinning (FLP) and to improve the thermal stability of Ti/Ge contact. The current density (J)-voltage (V) characteristics showed that the rectifying behavior of Ti/Ge contact into an Ohmic-like behavior with C-imp. The lowering of Schottky barrier height (SBH) indicated that the C-imp could mitigate FLP. In addition, it allows a lower specific contact resistivity (ρ_c) at the rapid thermal annealing (RTA) temperatures in a range of 450–600 °C. A secondary ion mass spectrometry (SIMS) showed that C-imp facilitates the dopant segregation at the interface. In addition, transmission electron microscopy (TEM) and electron energy loss spectroscopy (EELS) mapping showed that after RTA at 600 °C, C-imp enhances the diffusion of Ge atoms into Ti layer at the interface of Ti/Ge. Thus, carbon implantation into Ge substrate can effectively reduce FLP and improve contact characteristics.

1. Introduction

As a channel material for the next-generation field-effect transistors (FETs), Germanium (Ge) is considered a promising alternative to silicon (Si) owing to its higher carrier mobility and the process compatibility with the advanced Si microfabrication. However, the low-solid solubility and the high-diffusion coefficient of n-type dopants in Ge hinder the realization of low specific contact resistivity (ρ_c) [1]. Moreover, Fermi-level pinning (FLP) caused by the metal-induced gap states (MIGS) at the metal/Ge interface is another problem to be solved [2,3,4,5]. FLP strongly occurs near the Ge valence band (E_v) and forces the electron Schottky barrier height (e-SBH) above 0.5 eV irrespective of the metal workfunction [6]. Several approaches, including dopant segregation [7], dipole formation [8], and surface treatment [9] were proposed to mitigate FLP phenomena. Recently, the use of an ultra-thin insulator between the metal and Ge showed an effective reduction of FLP but the degradation of ρ_c due to a high tunneling resistance [10,11,12,13]. The formation of metal germanide can be another approach because the MIGS from metal dangling bond states in germanide can lead to an FLP reduction [14,15].

Ion implantation is another approach to achieving low ρ_c and suppressing dopant-diffusion behaviors. For example, Germanium implantation before silicidation induces surface amorphization to aid an epitaxial regrowth on the semiconductor surface [16]. Carbon implantation (C-imp) has been introduced in Ni-silicide and Ni-germinide contacts to reduce contact resistivity [17,18]. However, Ti/Ge contact with carbon implantation has been rarely reported.

Here, we investigated the effects of C-imp on the FLP reduction of a Ti/Ge contact and the related contact characteristics. Electrical characteristics were measured using the multiring-circular transmission line model (MR-CTLM) structure and Schottky barrier diode (SBD). Physical and structural properties of Ti/Ge contact with C-imp were analyzed using scanning electron microscopy (SEM), transmission electron microscopy (TEM), electron energy loss spectroscopy (EELS), and secondary ion mass spectrometry (SIMS).

2. Materials and Methods

N-type Ge wafers moderately doped with phosphorus (~10¹⁸ cm⁻³) were cleaned in a 1:100 diluted HF (dHF) solution and deionized (DI) water to remove native oxide. Subsequently, C⁺ ions were implanted into the Ge substrate at a dose of 1 × 10¹⁵ cm⁻² and an implantation energy of 10 keV. A reference sample without C-imp was also prepared. A SBD of Ti/Ge structure and a MR-CTLM structure were fabricated on the Ge substrate to characterize electrical properties. First, a 100 nm thick SiO₂ was deposited to isolate the contact holes using a plasma-enhanced chemical vapor deposition (PECVD). Then, the metal contact was formed using the conventional photolithography process. Sequentially, the oxide was etched using a dry etcher, and a Ti (5 nm)/TiN (5 nm) was deposited using a DC sputtering system. After a metal lift-off process, rapid thermal annealing (RTA) was performed in N₂ ambient for 60 s at 450–600 °C. Finally, a 100 nm thick Al was deposited as contact pad metal. The electrical measurements of current (I)–bias voltage (V) were performed using Keithley 4200-SCS. TEM images of the Ti/Ge structure without and with C-imp were obtained using a JEOL JEM 2200FS with an image Cs-corrector.

3. Results

Figure 1 shows the J-V characteristics of the Ti/Ge contacts with and without C-imp at RTA temperatures in a range of 450–600 °C for 60 s in N₂ ambient. The Ti/Ge contact without C-imp shows a typical rectifying behavior attributed to a strong FLP near the E_v, which leads to a significantly high e-SBH and reduces the reverse current density. On the other hand, the Ti/Ge contact with C-imp shows an Ohmic-like behavior with relatively high current density under the reverse regime, indicating the alleviation of FLP.

Figure 2a shows the extracted e-SBHs of the Ti/Ge contacts without (blue box) and with (red box) C-imp after RTA at 550 °C and 600 °C, respectively, for 60 s in N₂ ambient. The e-SBHs were extracted from the current-temperature (I-T) curves in a range of 300–378 K. The I-V relationship of a Schottky barrier diode is represented by [19]

$$I=A{A}^{*}{T}^2{e}^{-q{\varnothing }_B/kT}\left(e^{qV/nkT}-1\right)=I_{S1}{e}^{-q{\varnothing }_B/kT}\left(e^{qV/nkT}-1\right)=I_S\left(e^{qV/nkT}-1\right)$$

(1)

where I_s is the saturation current, A is the diode area, A* = 4πqk²m*/h³ = 120 (m*/m) A/cm²∙K² Richardson’s constant, Φ_B is the barrier height, and n is the ideality factor. For $V\gg kT/q$ Equation (1) can be written as follows:

$$\ln \left(I/T^2\right)=\ln \left({\mathrm{AA}}^{*}\right)-q\left({\varnothing }_B-V/n\right)/kT$$

(2)

$${\varnothing }_B=\frac{V}{n}-\frac{k}{q}\frac{d\left[\ln \left(I/T^2\right)\right]}{d\left(1/T\right)}=\frac{V}{n}-\frac{2.3k}{q}\frac{d\left[\log \left(I/T^2\right)\right]}{d\left(1/T\right)}$$

(3)

Therefore, the barrier height is calculated from the slope (=d[ln(I/T²)]/d(1/T)). The bandgap and electron affinity in eV of Ge at 300 K are 0.66 and 4.0 eV, respectively. The workfunction of Ti metals is about 4.3 eV. When Fermi level is pinned near E_v of Ge, Φ_B of ~0.6 eV is calculated. If there is negligible FLP, Φ_B of ~0.3 eV is obtained.

Without C-imp, the SBH of ~0.48 eV was obtained for both 550 °C and 600 °C RTA, indicating the occurrence of FLP. In contrast, the SBH with C-imp was significantly reduced from 0.31 eV at 550 °C to 0.27 eV at 600 °C.

Figure 2b,c show schematics of the energy band diagrams for Ti/Ge contacts. Without C-imp, Fermi-level on the Ti side is pinned with the charge neutrality level (E_CNL) due to FLP [6].

Figure 3 shows a top-view SEM image of the fabricated MR-CTLM structure to extract ρ_c and the sheet resistance beneath the metal (R_S). The current flows through multiple metal-semiconductor structures from the center region to the outer-circle region. From the I-V curve of MR-CTLM, the total resistance (R_tot) is expressed as the sum of the effective resistance (R_eff) and the parasitic resistance (R_pr) as follows [20]:

$$R_{tot}=R_{eff}+R_{pr}$$

(4)

$$R_{eff}=\frac{R_s}{2\pi }{{\displaystyle \sum }}_{i=0}^9\left[\ln \left(\frac{r_i+S_m}{r_i}\right)+L_t\left(\frac{1}{r_i}+\frac{1}{r_i+S_m}\right)\right]$$

(5)

$$R_{pr}=\frac{R_m}{2\pi }\left[{{\displaystyle \sum }}_{i=1}^9\ln \left(\frac{r_i-L_t}{r_i-S_s+L_t}\right)\right]$$

(6)

where r₀~r₉ are the inner radius of the serial CTLM. S_m and S_s are the spacing among metal rings and dielectric rings, respectively. L_t is the transfer length. S_s = 10 μm, r₀ = 50 μm, and S_m, from 0.5 to 10 μm were defined using an i-line stepper. ρ_c was calculated from the L_t ($=\sqrt{{\rho }_c/R_s}$) which was extracted by fitting a set of R_t-S_m data using Equations (4)–(6).

Figure 4 shows the extracted ρ_c values versus RTA temperature. ρ_c was obtained using a MR-CTLM test structure [20]. A relatively high ρ_c value seems mainly because of the low activation of a substrate doping concentration of ~1 × 10¹⁸ cm⁻³ [21,22]. After RTA annealing at 600 °C, the ρ_c values of the Ti/Ge with and without C imp were 1.3 × 10⁻⁵ and 8.4 × 10⁻⁴ Ω∙cm², respectively. Owing to the FLP effect, the Ti/Ge contact without C-imp shows ρ_c values higher than 1.0 × 10⁻⁴ Ω∙cm².

To further analyze the effect of C-imp on the Ti/Ge composition, TEM and SIMS were conducted. The decrease of ρ_c is mainly attributed to the dopant segregation in the Ti/Ge interface [23]. In particular, for the Ti/Ge contact with C-imp after RTA at 600 °C, a further reduction of ρ_c is observed. These results can be expected by TiGe_x formation. The low resistive C54-TiGe_x is formed at a temperature above 600 °C [24], which mitigates FLP and improves the contact resistivity [14,15].

Figure 5a,b show SIMS profiles for Ti/Ge contacts without and with C-imp, respectively. At the Ti/Ge interface with C-imp, the peak P concentration increases from 1.6 × 10¹⁸ cm⁻³ to 3.6 × 10¹⁸ cm⁻³, attributed to the dopant segregation facilitated by carbon [18]. This dopant segregation can increase the tunneling current by reducing the depletion thickness at the interface and lowering the contact resistivity.

To directly observe the microstructure of Ti/Ge contact, the cross-sectional TEM images and the corresponding EELS were analyzed. The samples were prepared after RTA at 600 °C for 60 s in N₂ ambient, as shown in Figure 6. In EELS maps, a bright region represents the area that the element of interest is abundant. With C-imp, Ge element is considerably observed in the Ti layer (red box in Figure 6b). The diffused Ge reacts with Ti and forms the Ti-germanide during the RTA process, which is beneficial to reduce the contact resistivity [14,15]. These results show that the C-imp is a promising approach to lower the contact resistivity in Ti/Ge contact by inducing the dopant segregation and Ge diffusion into the Ti layer.

4. Conclusions

We investigated the electrical and material characteristics of a Ti/Ge contact with C-imp. The current-voltage behavior shows that the carbon implantation changes the Ti/Ge rectifying behavior into an Ohmic-like behavior above RTA at 450 °C. The extracted Schottky barrier height was also decreased due to the mitigation of Fermi-level pinning. The specific contact resistivity of the Ti/Ge contact with C-imp was significantly reduced by approximately two orders of magnitude. Transmission electron microscopy and secondary ion mass spectrometry showed that carbon element at the Ti/Ge interface facilitates the dopant segregation and induces the diffusion of Ge into Ti layer. Therefore, the carbon implantation is promising to improve the Ti/Ge contact properties for high-performance Ge-FET applications.