Investigation on interfacial and electrical properties of Ge MOS capacitor with different NH3-plasma treatment procedure

    Corresponding author: Jingping Xu, jpxu@hust.edu.cn
  • School of Optical and Electronic Information, Huazhong University of Science and Technology, Wuhan 430074, China

Key words: Ge MOSNH3-plasma treatmentTaON interlayerstacked gate dielectric

Abstract: The effects of different NH3-plasma treatment procedures on interfacial and electrical properties of Ge MOS capacitors with stacked gate dielectric of HfTiON/TaON were investigated. The NH3-plasma treatment was performed at different steps during fabrication of the stacked gate dielectric, i.e. before or after interlayer (TaON) deposition, or after deposition of high-k dielectric (HfTiON). It was found that the excellent interface quality with an interface-state density of 4.79×1011 eV-1cm-2 and low gate leakage current (3.43×10-5A/cm2 at Vg=1 V) could be achieved for the sample with NH3-plasma treatment directly on the Ge surface before TaON deposition. The involved mechanisms are attributed to the fact that the NH3-plasma can directly react with the Ge surface to form more Ge-N bonds, i.e. more GeOxNy, which effectively blocks the inter-diffusion of elements and suppresses the formation of unstable GeOx interfacial layer, and also passivates oxygen vacancies and dangling bonds near/at the interface due to more N incorporation and decomposed H atoms from the NH3-plasma.

    HTML

1.   Introduction
  • Recently, Ge-based MOSFET with high-$k$ gate dielectric has been widely investigated mainly due to its higher carrier mobility compared with Si and potential for continuous dimensional scaling[1]. For this purpose, a high-quality interface between the high-$k$ dielectric and Ge is key to realizing the application of high-performance Ge MOS devices. However, it is difficult to achieve a high-quality oxide/Ge interface with integrity like SiO$_{\rm{2}}$/Si due to the unfavorable surface properties and water-soluble native oxides of Ge[2]. In order to solve the problem, generally, there are two ways to improve the interface quality: first, various interfacial passivation layers (IPL), e.g. TaON[3, 4], AlON[5, 6], GeO$_{\rm{2}}$[7, 8], La$_{\rm{2}}$O$_{\rm{3}}{}$[9, 10] and Y$_{\rm{2}}$O$_{\rm{3}}{}$[11, 12] were introduced between high-$k$ dielectric and Ge, which demonstrated perfect interface properties. Among them, TaON with a relatively high $k$ value (~26) was fabricated as the passivation layer on the germanium surface prior to deposition of high-$k$ dielectric, which can effectively suppress the growth of unstable GeO$_{{x}}$, thus reducing interface states and increasing carrier mobility in the inversion channel of Ge-based transistors[4]. Second, surface passivation of Ge substrate such as nitrogen-related plasma[13-19], F-plasma[20] and S treatment[21] were also employed to improve the interface quality. The plasma nitridation has attracted a great deal of attention as a promising approach for improving the interface and electrical properties of MOS capacitors. Among the nitrogen-related plasma treatments, it is well known that the NH$_{\rm{3}}$-plasma nitridation can terminate the oxide/Ge interface with Ge nitride or oxynitride, resulting in low density of interface traps and high thermal stability[22, 23]. It is expected that the further improvements of the interfacial and electrical properties of the Ge MOS capacitor could be achieved by combining the TaON as IPL with NH$_{\rm{3}}$-plasma treatment. Also, it is expected that different passivation effects would be obtained if the NH$_{\rm{3}}$-plasma treatment is performed at different steps during fabrication of the stacked high-$k$ dielectric. So in this work, different passivation effects of the NH$_{\rm{3}}$-plasma treatment procedure, i.e., before or after IPL deposition, or after deposition of high-$k$ dielectric, on the interface quality and electrical properties of the Ge MOS device are compared through fabricating a HfTiON gate dielectric Ge MOS capacitor with TaON as IPL, and the involved mechanisms are analyzed by means of XPS results.

2.   Experiment
  • Ge MOS capacitors were fabricated on n-type (100) Ge wafers with a resistivity of 0.02-0.10 $\Omega \cdot$cm. The wafers were in turn cleaned using ethanol, acetone and trichloroethylene, followed by dipping in diluted HF ($1:50$) for 30 s, and then rinsing in deionized water for several times to remove the native oxides.

    After drying by high pure N$_{\rm{2}}$, four groups of samples were prepared with different procedures of the NH$_{\rm{3}}$-plasma treatment, as shown in Fig. 1. For the Ge-NH$_{\rm{3}}$ sample, the Ge substrate was firstly exposed to the NH$_{\rm{3}}$-plasma, followed by a deposition of 2-nm TaN interlayer by sputtering Ta target in Ar/N$_{\rm{2\thinspace }}$(15 sccm/6 sccm) ambient and then $in-situ$ deposition of 8-nm HfTiN high-$k$ layer by co-sputtering Hf and Ti targets in the same ambient. In Figs. 1(b) and 1(c), the NH$_{\rm{3}}$-plasma treatment was carried out after TaN deposition (denoted as the TaON-NH$_{\rm{3}}$ sample) or after HfTiN deposition (denoted as the HfTiON-NH$_{\rm{3}}$ sample). In addition, a sample without any NH$_{\rm{3}}$-plasma treatment in Fig. 1(d) was fabricated as the control sample. The NH$_{\rm{3}}$-plasma treatment was performed in a plasma-enhanced chemical vapor deposition chamber at 300 ℃, with an RF power of 120 W. Post-deposition annealing was carried out at 500 ℃ for 300 s in N$_{\rm{2}}$ (500 sccm) $+$ O$_{\rm{2}}$ (50 sccm) to transform HfTiN and TaN into HfTiON and TaON. Finally, Al was thermally evaporated and patterned as a gate electrode with an area of 7.85 $\times$ 10$^{\mathrm{-5}}$ cm$^{\mathrm{2}}$, followed by an N$_{\rm{2}}$ annealing at 300 ℃ for 20 min to reduce the contact resistance.

    High-frequency (HF, 1-MHz) capacitance-voltage ($C$-$V)$ curve and gate leakage current-voltage ($J_{\rm g}$-$V_{\rm g})$ measurements were measured by using an Agilent 4284A precision LCR meter and Keithley 4200-SCS semiconductor characterization system, respectively. The physical thickness of the gate dielectric was determined by spectroscopic ellipsometry. All electrical measurements were carried out under a light-tight and electrically shielded condition.

3.   Results and discussion
  • Fig. 2 shows the HF $C$-$V$ curves of the four samples, swept in two directions (from inversion to accumulation and back). Obviously, the control sample without any NH$_{\rm{3}}$-plasma treatment exhibits poor $C$-$V$ behavior with the largest "stretch out" along the voltage axis and the smallest slope in the depletion region. While the other three samples with NH$_{\rm{3}}$-plasma treatment show better $C$-$V$ behavior and larger accumulation capacitance than the control sample, implying that the NH$_{\rm{3}}$-plasma treatment at any step during fabrication of the gate stack can improve the interfacial quality of the gate stack/Ge and electrical characteristics of the devices to different extents. Obviously, the TaON-NH$_{\rm{3}}$ and Ge-NH$_{\rm{3}}$ samples show comparable $C$-$V$ behavior with larger accumulation capacitance than the HfTiON-NH$_{\rm{3}}$ sample, in which the Ge-NH$_{\rm{3}}$ sample exhibits the sharpest transition from depletion to accumulation. So, it can be suggested that the NH$_{\rm{3}}$-plasma treatment on the Ge surface before TaON deposition is probably the most effective method for improving the interfacial and electrical properties. This may be attributed to the fact that the native oxides on Ge substrates can be transformed into GeO$_{{x}}$N$_{\rm {y}}$ by NH$_{\rm{3}}$-plasma treatment, which exhibits good thermal stability and interface properties with Ge, thus suppressing the growth of Ge suboxide (GeO$_{{x}}$, $1 < x < 2$)[14, 17]. Huang[19] has pointed out that nitrogen incorporated in the interlayer by N$_{\rm{2}}$-plasma treatment could effectively enhance its blocking capability against the intermixing between Ge and HfTiON, and thus suppress the growth of unstable low-$k$ GeO$_{{x}}$ at the GGON/Ge interface. The improvement of the electrical characteristics would be attributed to the fact that the oxygen vacancies in HfTiON can be passivated by the NH$_{\rm{3}}$-plasma treatment, which suppresses further oxygen diffusion and the formation of oxygen vacancies, improving the quality and reliability of the high-$k$ dielectric[24].

    From the bi-directionally swept $C$-$V$ curves in Fig. 2, the hysteresis voltage is extracted to be 60, 140, 160, and 240 mV for the Ge-NH$_{\rm{3}}$, TaON-NH$_{\rm{3}}$, HfTiON-NH$_{\rm{3}}$ and control samples, respectively. The smallest hysteresis voltage (60 mV) is obtained for the Ge-NH$_{\rm{3}}$ sample, implying there are the fewest slow states or deep-level traps in the HfTiON/TaON stacked dielectric and near the TaON/Ge interface due to the reduction of oxygen defects and Ge out-diffusions induced by the NH$_{\rm{3}}$-plasma treatment on the Ge surface.

    The flat-band voltage ($V_{\rm fb})$ of the samples are determined from their flatband capacitance[25], and equivalent oxide-charge density ($Q_{\rm ox})$ is calculated as $-C_{\rm ox}(V_{\rm fb} -\phi_{\rm ms})$/$q$, where $\phi_{\rm ms}$ is the work-function difference between Al and Ge, and $C_{\rm ox}$ is the oxide capacitance (accumulation capacitance) per unit area. The capacitance equivalent thickness (CET) is calculated (CET$= $ $ k_{\rm{0}} k_{\rm{SiO2}}$/$C_{\rm ox}$, where $ k_{\rm{0}}$ and $k_{\rm{SiO_2}}$ are the vacuum permittivity and relative permittivity of SiO$_{\rm{2}}$; $C_{\rm ox}$ is the accumulation capacitance per unit area). The interface-state density near midgap ($D_{\rm it})$ is extracted from the 1-MHz $C$-$V$ curve using the Terman method[26] for the purpose of comparison, as listed in Table 1.

    The negative shift of $V_{\rm fb}$ indicates that positive oxide charges exist in the gate stacked dielectric[27]. Obviously, the samples with NH$_{\rm{3}}$-plasma treatment have smaller $V_{\rm fb}$ and $D_{\rm it}$ than the control sample, with the smallest for the Ge-NH$_{\rm{3}}$ sample ($V_{\rm fb} =$ 0.08 V and $D_{\rm it} =$ 4.79 $\times$ 10$^{\mathrm{11}}$ eV$^{\mathrm{-1}}$cm$^{\mathrm{-2}})$, which should be attributed to the fact that the nitrogen incorporated near the TaON/Ge interface by the NH$_{\rm{3}}$-plasma treatment can effectively occupy the oxygen vacancies near the interface and in the TaON IPL, resulting in a reduction of relevant defect states, and also the passivation role of H and N atoms decomposed from the NH$_{\rm{3}}$-plasma on the dangling bonds near/at the interface is another possible cause of improving the interface properties for the Ge-NH$_{\rm{3\thinspace }}$sample due to the direct action of the NH$_{\rm{3}}$-plasma on the Ge surface. This further indicates that the NH$_{\rm{3}}$-plasma treatment on the Ge surface can effectively remove the nature oxides and relevant defective bonds on the Ge surface, thus reducing the defects at/near the interface. The largest accumulation capacitance of the Ge-NH$_{\rm{3\thinspace }}$sample is also attributed to the effective incorporation of nitrogen at/near the TaON/Ge interface, which can enhance the suppression of the low-$k$ interlayer growth and give a high-quality interface, as confirmed by the XPS result below.

    Fig. 3 shows the $C$-$V$ curves of all the samples measured at 1 MHz and 100 kHz. The frequency dispersion can be estimated from the percentage of accumulation-capacitance change, i.e. ($C_{\rm \rm ox, 100\, kHz}-C_{\rm \rm ox, 1\, MHz})$/$C_{\rm \rm ox, 1\, MHz}$ at a gate voltage of 2 V, to be 4.18%, 6.33%, 11.45%, and 16.51% for the Ge-NH$_{\rm{3}}$, TaON-NH$_{\rm{3}}$, HfTiON-NH$_{\rm{3}}$ and control samples, respectively. The frequency dispersion in accumulation is caused by the border traps near the oxide/semiconductor interface[17]. Obviously, the samples with NH$_{\rm{3}}$-plasma treatment exhibit smaller frequency dispersion than the control sample, indicating less slow states at/near the TaON/Ge interface for the former than the latter. Again, the smallest frequency dispersion is obtained for the Ge-NH$_{\rm{3}}$ sample, implying the fewest slow states and deep-level traps at/near the TaON/Ge interface, attributed to the effective suppression of low-$k$ GeO$_{{x}}$ and reduction of the oxygen defects and Ge out-diffusion induced by the NH$_{\rm{3}}$-plasma treatment on the Ge surface, as mentioned above.

    Fig. 4 shows the gate leakage properties of all the samples. Large gate leakage current density is observed when positive gate voltage is applied on the sample without the NH$_{\rm{3}}$-plasma treatment, e.g., 7.19 $\times$ 10$^{\mathrm{-4\thinspace }}$A/cm$^{\mathrm{2}}$ at $V_{\rm g} =1$ V. This is likely associated with a large amount of GeO$_{{x}}$ at/near the Ge substrate[28]. Fortunately, the gate leakage current density is reduced for the NH$_{\rm{3}}$-plasma treated samples. For the HfTiON-NH$_{\rm{3}}$ sample, as described above, the oxygen vacancies in HfTiON can be passivated by the NH$_{\rm{3}}$-plasma treatment, which suppresses further oxygen diffusion and the formation of the oxygen vacancies, improving the quality and reliability of the high-$k$ dielectric and thus leading to the decrease of the gate leakage current density. While the gate leakage current density is greatly decreased for the Ge-NH$_{\rm{3}}$ and TaON-NH$_{\rm{3}}$ samples, with the smallest for the Ge-NH$_{\rm{3}}$ sample (3.43 $\times$ 10$^{\mathrm{-5\thinspace }}$A/cm$^{\mathrm{2}}$ at $V_{\rm g} = 1$ V), which should be attributed to their high-quality interface due to the greatly suppressed growth of Ge suboxide and passivation of oxygen vacancies near the interface due to more N incorporation.

    To identify the composition and chemical status at/near the TaON/Ge interface, and further analyze the effect of the NH$_{\rm{3}}$-plasma treatment on interface quality, the high-$k$ layer is etched to a distance of about 3 nm from the Ge surface using an in-situ Ar$^+$ ion beam in the XPS chamber.

    Fig. 5 is the N 1s and O 1s spectra of the three samples with NH$_{\rm{3}}$-plasma treatment, by which the presence of nitrogen and oxygen can be confirmed. As shown in Fig. 5(a), the Ge-NH$_{\rm{3}}$ and TaON-NH$_{\rm{3}}$ samples exhibit an obvious N 1s peak at 397.0 eV, which should be from the N-Ge bonding[29], since the N-Ta, N-H and N-O bonds have a higher binding energy at 398.2[30], 398.8[31], and 402.9 eV[32], respectively, indicating that the NH$_{\rm{3}}$-plasma can effectively react with the Ge surface to form N-Ge bond when it is used to treat the Ge surface or the TaON IPL, which is consistent with the results in Ref. [17]. However, N 1s peak becomes very weak when using NH$_{\rm{3}}$-plasma to treat the high-$k$ layer, as shown by the HfTiON-NH$_{\rm{3}}$ sample, indicating that very few N are incorporated at the Ge surface. On the other hand, the intensity of the O 1s peak of the three samples exhibits totally opposite results to that of the N 1s peak, i.e. the strongest for the HfTiON-NH$_{\rm{3}}$ sample and the weakest for the Ge-NH$_{\rm{3}}$ sample. In other words, GeO$_{{x}}$ effectively gets suppressed by N incorporation to transform it into oxynitride, i.e. forming more GeO$_{{x}}$N$_{\rm {y}}$, as supported by a shift of O 1s peaks toward the lower binding energy owing to the presence of more nitrogen in the interlayer and at/near the TaON/Ge$_{\rm{\thinspace }}$interface, especially for the Ge-NH$_{\rm{3}}$ sample.

    For investigating Ge oxides or oxynitride near the gate stack/Ge interface, the XPS spectra of Ge 2p are shown in Fig.6. The main peak from Ge-bulk locates at 1217.46 eV, which is in good agreement with the values of 1217.4[33] or 1217.77 eV[28]. In addition to the Ge 2p peak, the spectra were deconvoluted into two subpeaks by using XPS analysis software (Thermo Scientific Avantage): one located at 1218.82 eV with the peak splitting of ~1.36 eV relative to Ge bulk, showing the existence of GeO$_{{x}}$ ($x < 2$)[34], and another located at 1220.01 with binding-energy shift of 2.55 eV from Ge bulk, indicating nitrogen incorporation and the formation of GeO$_{{x}}$N$_{\rm {y}}$[28, 34], because nitridation can change the chemical state and result in the subpeak shifted to a lower binding energy[19]. To further analyze the effects of NH$_{\rm{3}}$-plasma treatment on the Ge oxides or oxynitride, the area ratio of GeO$_{{x}}$N$_{\rm {y}}$ or GeO$_{{x}}$ peak to Ge 2p peak is listed in Table 2. Obviously, the Ge-NH$_{\rm{3}}$ sample has the largest/smallest area ratios of 12.65%/2.06% for GeO$_{{x}}$N$_{\rm {y}}$/GeO$_{{x}}$ peaks, i.e. more GeO$_{{x}}$N$_{\rm {y}}$ and less GeO$_{{x\thinspace }}$are formed in the Ge-NH$_{\rm{3}}$ sample than other samples. From the XPS results, it can be suggested that the amount of Ge-N bonds depends on the reaction intensity between Ge substrate and NH$_{\rm{3}}$-plasma. As a result, more Ge-O bonds are broken and more Ge-N bonds are formed for the Ge-NH$_{\rm{3}}$ sample because the NH$_{\rm{3}}$-plasma can directly react with the native GeO$_{{x}}$ on the Ge substrate. For the TaON-NH$_{\rm{3}}$ sample, the 2-nm TaON is like a hard non-continuous mask layer, so that the NH$_{\rm{3}}$-plasma can partly react with the Ge layer, and for HfTiON-NH$_{\rm{3}}$ sample, the 2-nm TaON plus an 8-nm HfTiON layer serve as a thick hard mask layer, so that the NH$_{\rm{3}}$-plasma cannot react with the Ge surface any more. In addition, the NH$_{\rm{3}}$-plasma can produce high-density reactive species such as H atoms and NH radicals[35, 36], which are expected to terminate the dangling bonds and passivate oxygen vacancies at/near the interface. This is why the Ge-NH$_{\rm{3}}$ sample has a better interface quality and electrical properties.

4.   Conclusions
  • In summary, the HfTiON gate dielectric Ge MOS capacitor with TaON as IPL has been fabricated, and the impacts of the NH$_{\rm{3}}$-plasma treatment procedure on the interfacial and electrical properties of the devices were investigated. Totally, it is an effective way to passivate the gate dielectric and its interface with Ge substrate using the NH$_{\rm{3}}$-plasma treatment during fabrication of the stacked gate dielectric, in which the NH$_{\rm{3}}$-plasma treatment on the Ge surface before IPL deposition is the most effective approach to obtain excellent interface quality and electrical properties with low interface-state and oxide-charge densities, small hysteresis and frequency dispersion, and low gate leakage current. The involved mechanisms lie in the fact that the NH$_{\rm{3}}$-plasma treatment on the Ge surface can incorporate more nitrogen to passivate oxygen vacancies and suppress the formation of the unstable GeO$_{{x}}$ at/near the interface, and also the NH radicals and H atoms decomposed from the NH$_{\rm{3}}$ plasma can passivate the dangling bonds near/at the interface. This makes it a promising method for preparing high-quality Ge MOS devices with high-$k$ gate dielectric.

Figure (6)  Table (2) Reference (36) Relative (20)

Journal of Semiconductors © 2017 All Rights Reserved