Study of the interface stability of the metal ( Mo , Ni , Pd ) / HfO 2 / AlN / InGaAs MOS devices

The degeneration of the metal/HfO2 interfaces for Mo, Ni, and Pd gate metals was studied in this paper. An unstable PdOx interfacial layer formed at the Pd/HfO2 interface, inducing the oxygen segregation for the Pd/HfO2/InGaAs metal oxide capacitor (MOSCAP). The low dissociation energy for the Pd-O bond was the reason for oxygen segregation. The PdOx layer contains O2− and OH− ions which are mobile during thermal annealing and electrical stress test. The phenomenon was not observed for the (Mo, Ni)/HfO2/InGaAs MOSCAPs. The results provide the guidance for choosing the proper metal electrode for the InGaAs based MOSFET.

The degeneration of the metal/HfO 2 interfaces for Mo, Ni, and Pd gate metals was studied in this paper.An unstable PdO x interfacial layer formed at the Pd/HfO 2 interface, inducing the oxygen segregation for the Pd/HfO 2 /InGaAs metal oxide capacitor (MOSCAP).The low dissociation energy for the Pd-O bond was the reason for oxygen segregation.The PdO x layer contains O 2 and OH ions which are mobile during thermal annealing and electrical stress test.The phenomenon was not observed for the (Mo, Ni)/HfO 2 /InGaAs MOSCAPs.The results provide the guidance for choosing the proper metal electrode for the InGaAs based MOSFET.© 2017 Author(s).All article content, except where otherwise noted, is licensed under a Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).[http://dx.doi.org/10.1063/1.4986147]2][3][4][5][6] The HfO 2 /InGaAs interface, which affects the power consumption of complementary metal-oxide-semiconductor (CMOS), has been greatly improved recently. 7,8The reactions between the metal electrode and the oxide layer may lead to the degeneration of the oxide layer.To prevent these interface reactions, the multilayer metal alloy of TiNi was used as the gate electrode for the HfO 2 /InGaAs metal oxide semiconductor capacitor (MOSCAP). 9Also, a thin AlN layer was inserted between the metal electrode and the HfO 2 layer to suppress this interface reactions. 10Yoshida et al. 11 recently reported that the reactions between the high-k material and InGaAs were due to the interaction at the metal/oxide interface.Their study indicated that the metal electrode affects not only on the metal/HfO 2 interface but also on the HfO 2 /InGaAs interface.To suppress this interfacial interaction, the passivation of HfO 2 /InGaAs interface using AlN layer was reported. 7,12herefore, the mechanism suggested by Yoshida et al. should be complemented when a non-oxide layer is used to passivate the HfO 2 /InGaAs interface.Additionally, the oxide degeneration depends not only on the electrode-high-k reaction but also on the inter diffusion between the oxide layer and the metal electrode after post metal-deposition annealing (PMA).In this study, the oxygen segregation at the (Mo, Ni, Pd)/HfO 2 interfaces of the HfO 2 /AlN/InGaAs MOSCAP is investigated.Mo was chosen due to its low work function applicable for NMOS InGaAs devices, while Ni and Pd were chosen because of their high work function suitable for PMOS InGaAs devices.
The MOSCAP was fabricated on the 100-nm epitaxial In 0.53 Ga 0.47 As layer (5 × 10 17 /cm 3 Sidoped n-type wafer) which was grown on the n+-InP substrate (using molecular beam epitaxial (MBE) growth technique).First, the wafer went through an HCl:H 2 O (3.8%) solution treatment for 2 minutes before it was loaded into the ALD chamber.Then, 0.8-nm AlN layer and 50 cycles HfO 2 were deposited on the InGaAs layer. 13The AlN layer serves as the passivation layer to prevent the a Author to whom correspondence should be addressed.Electronic mail: edc@mail.nctu.reaction between oxygen atoms in HfO 2 and In, Ga, and As atoms in the InGaAs channel. 7After that, the samples were annealed in forming gas (FG) at 450 o C for 5 minutes using rapid thermal annealing (RTA).Three types of metal including Mo, Ni, and Pd (with the same thickness of 50 nm) were studied as the gate metals on the HfO 2 layer.The gate metals were deposited through E-gun evaporation with the based pressure about 3 × 10 7 torr.The deposition rate was controlled by quartz crystals (gold, 6 MHz) and kept to be 1 Å/s.The AuGeNiAu was deposited on the backside of the wafer to form the ohmic contact.Finally, the MOSCAPs were annealed in three different ambiences including FG, nitrogen (N) and oxygen (O) at 350 o C for 30s using RTA for comparison.The metal/HfO 2 interfaces were characterized by a Field Emission High-resolution transmission electron microscopy (HRTEM) JEM-2100F, while the thickness of HfO 2 was determined by using a SOPRA GES5 Ellipsometer and confirmed by HRTEM measurement.The capacitance-voltage (C-V) and current-voltage (J-V) were performed by using an HP4284A LCR meter.The X-ray photoelectron spectroscopic (XPS) were measured in a commercial Microlab 350 XPS system with Al Kα source.
The core levels were determined by using XPSPEAK (version 4.1) with Gaussian-Lorentz line shape and a Shirley background.Sample charging effects were corrected by placing the Au 4f 7/2 peak at binding energy of 84.0 eV and shifting the rest of the regions accordingly.The uncertainty of the core position was 0.05 eV. Figure 1(a) shows the capacitance-voltage hysteresis (∆V FB ) of the (Mo, Ni, Pd)/HfO 2 /AlN/ InGaAs MOSCAPs.The flat band voltage (V FB ) was determined from the comparative method. 14he fact that the ∆V FB of the Pd MOSCAP is very large (∼ 1V), while the ∆V FB of the Mo MOSCAP is very small (∼ 0.11 V) indicates that a large number of bulk oxide traps was created after Pd was deposited on HfO 2 and annealed in FG.The large ∆V FB was due to the interaction of Pd atoms and O atoms at the Pd/HfO 2 interface, which creates oxygen vacancies inside the HfO 2 layer.This reaction leads to the increase of the oxide thickness and reduces the accumulation capacitance (C acc ) of the Pd MOSCAP as shown in Fig. 1(b).Based on the thermodynamics, the ∆C acc of the Mo MOSCAP should be larger than the ∆C acc of the Pd MOSCAP because the Mo-O reaction (oxidation formation enthalpy ∆H ∼ 178 kCal/mol) is more exothermal than the Pd-O reaction (∆H 49.9 kCal/mol). 15wever, the opposite results were observed in Fig. 1 The low D it values (∼ 4 × 10 12 cm 2 eV 1 determined from Terman method 16 ) of the Ni MOSCAP and the Mo MOSCAP, shown in Fig. 1(d), are in agreement with the previous reports. 6,7,10Note that Terman method gives both slow and fast traps and provides a reliable value for Dit at mid-gap. 16The In out-diffusion 9,[17][18][19][20] from InGaAs substrate to the oxide layer (Al 2 O 3 or HfO 2 ) was also reported to affect the quality of the oxide layer.The D it values of Pd MOSCAPs are much higher than those of Ni and Mo MOSCAPs as shown in Fig. 1(d), indicating that the PdO x layer at Pd/HfO 2 interface reduces the passivation effects of AlN layer.The increase of oxygen vacancies inside HfO 2 due to the reaction of Pd and HfO 2 may facilitate the out-diffusion of In atoms from InGaAs substrate to HfO 2 layer, leading to the increase of D it for the Pd MOSCAPs.Fig. 1 also shows that the PMA process has little effect on the electrical characteristics of the Ni MOSCAP; thus, Ni is a more suitable gate metal for the InGaAs based p-MOSFET. 7he dissociation energies (D S ) of the metal-oxygen bonds for the metals used in this study are Mo-O bond (119.9 kCal/mol), Ni-O bond (87.42 kCal/mol), and Pd-O bond (53.87 kCal/mol). 21It can be seen that PdO x is easier to decompose during the PMA process compared to other oxides in this study following the equation: The O 2 ions can easily diffuse into the Pd layer, leading to the formation of a thick PdO x IL at the Pd/HfO 2 interface as shown in Fig. 2(a).This IL was not found at the Ni/HfO 2 interface as shown in Fig. 2 To investigate the stability of the PdO x IL due to the oxygen segregation at the HfO 2 layer, the XPS analysis was conducted on the clean HfO 2 surface, the (2-nm) Mo/HfO 2 interface, and the (2-nm) Pd/HfO 2 interface; the results are shown in Fig. 3.For the clean HfO 2 surface, the XPS data is shown in Fig. 3(a).The strongest peak O I (530.57eV) is attributed to the oxygen in a metal-oxygen bond without oxygen vacancy; the peak O II (532.54 eV) and the peak O III (533.83 eV) are attributed to the oxygen in the metal-oxygen bond with oxygen vacancy and the OH group attached to Hf 2+ ions, respectively. 22The O 1s peak of the Mo/HfO 2 interface after PMA in FG at 350 o C for 30s is same as the O 1s peak of the clean HfO 2 surface, indicating that the Mo/HfO 2 interface is quite stable due to the large dissociation energy of the Mo-O bond.For the Pd/HfO 2 interface, due to the overlap of the Pd 3p 3/2 peak with the O 1s peaks, the Pd 3p 1/2 peak was firstly set at 562.20 eV 23 as a standard for the Pd 3p 3/2 , O I , O II , and O III peaks fitting.The spin-orbit splitting of the Pd 3p peak was chosen at 27.8 eV, which was in agreement with the previous report. 24It was found that the O I intensity decreased from 85.4 % for the clean HfO 2 surface to 33 % for the Pd/HfO 2 interface as shown in Table I.The considerable increase of the intensities of the O II and O III peaks at the Pd/HfO 2 interface indicates a significant increase of the oxygen vacancies in HfO 2 , due to the diffusion of O 2 ions from the HfO 2 layer to the Pd layer.The same phenomenon was observed in Fig. 3(b).The intensity of the Pd 4d peak (∼ 1.4 eV) clearly decreased after the Pd/HfO 2 interface was annealed in FG at 350 o C for 30s, while the intensity of O 2p peak (∼ 7eV) increased after the sample was annealed.The Hf 4f peaks shifted after Pd deposition on the HfO 2 surface; this shift was due to the metal induced gap state phenomenon 25 at the Pd/HfO 2 interface.After the Pd MOSCAP was annealed, the increases of the O 2 and OH ion densities, as can be seen in Fig. 3(a), induced a dipole at the Pd/HfO 2 interface which increased the binding energy of Hf 4f peak as shown in Fig. 3(b).
Figures 4(a) and (b) show the C-V characteristics of the Pd MOSCAP and the Mo MOSCAP, respectively, with the gate stress at 2 V and + 2 V for 1800s; the C-V behavior can be used to investigate the effects of the O 2 and OH ions diffusion on the oxide film.To track the degeneration of the oxide layer due to a movement of the O 2 and OH ions, the C-V measurements were suddenly conducted during the constant-voltage-stress. 26,27 For the Pd MOSCAP, the C-V curve was found to have negative shift, and the C acc was found to increase after it went through the negative stress

7 FIG. 1 . 3 Do
FIG. 1.The electrical characteristics of (Mo, Ni, Pd)/HfO 2 /AlN/InGaAs MOSCAPs after PMA in FG, N and O ambient gases at 350 o C for 30s, (a) the hysteresis at V FB of the C-V curves after PMA, (b) the C acc variation of the MOSCAPs after PMA measured at +2 V and 1MHz, (c) the leakage currents at V FB + 1 (V) after PMA, and (d) the D it at E -E C = 0.3 eV from the conduction band edge after PMA.
(b).Fig.2(c) shows the Energy-dispersive X-ray spectroscopy (EDS) spot profile of the PdO x IL.An oxygen percent of ∼ 24 % was found in the PdO x layer, indicating a significant internal diffusion of the O 2 ions from the HfO 2 layer into the Pd layer.Fig.2(d) and (e) show the scanning electron microscope (SEM) images of the Pd MOSCAP and the Ni MOSCAP, respectively, after RTA in FG at 400 o C for 5 minutes.Some bubbles occurred at the surface of the Pd MOSCAP, indicating that the Pd MOSCAP was destroyed due to the diffusion of O 2 ions from the HfO 2 layer into the Pd layer.

085208- 4 Do 5 Do
FIG. 2. (a) and (b) TEM cross sections of Pd/HfO 2 /AlN/InGaAs and Ni/HfO 2 /AlN/InGaAs structures after RTA in FG at 350 o C for 30s, respectively, (c) the EDS spot profile of the interfacial layer PdO x at the Pd/HfO 2 interface, (d) and (e) SEM images of the patterns of Pd/HfO 2 and Ni/HfO 2 MOSCAPs, respectively, after RTA in FG at 400 o C for 5 minutes.

FIG. 4 .
FIG. 4. The effects of O 2 and OH ions on the C-V behaviors of Pd MOSCAP and Mo MOSCAP, (a) and (b) the C-V characteristics of Pd MOSCAP and Mo MOSCAP, respectively, after gate stressed at -2 V and + 2 V for 1800s; (c) the schematic PdO x interfacial layer after gate stressed at -2 V and + 2 V for 1800s.

TABLE I .
The area percent (%) of O I , O II , and O III peaks of clean HfO 2 surface, Mo/HfO 2 interface and Pd/HfO 2 interface.O i /(O I +O II +O III )