Cite this Article
Meng Xiao-Ran, Ping Yun-Xia, Yu Wen-Jie, Xue Zhong-Ying, Wei Xing, Zhang Miao, Di Zeng-Feng, Zhang Bo, Zhao Qing-Tai. Impact of Al addition on the formation of Ni germanosilicide layers under different temperature annealing. Chinese Physics B, 2017, 26(9): 098503  
Permissions
Impact of Al addition on the formation of Ni germanosilicide layers under different temperature annealing
Meng Xiao-Ran1, 2, Ping Yun-Xia1, †, Yu Wen-Jie2, Xue Zhong-Ying2, Wei Xing2, Zhang Miao2, Di Zeng-Feng2, Zhang Bo2, ‡, Zhao Qing-Tai3
Shanghai University of Engineering Science, Shanghai 201600, China
State Key Laboratory of Functional Material for Informatics, Shanghai Institute of Microsystem and Information Technology, Chinese Academy of Sciences, Shanghai 200050, China
Peter Grünberg Institute 9 (PGI 9-IT), and JARA-Fundamentals of Future Information Technology, Forschungszentrum Juelich, Juelich 52425, Germany

 

† Corresponding author. E-mail: xyping@sues.edu.cn bozhang@mail.sim.ac.cn

Project supported by the Natural Science Foundation of Shanghai, China (Grant No. 14ZR1418300) and the National Natural Science Foundation of China (Grant Nos. 61604094 and 61306126).

Abstract

Solid reactions between Ni and relaxed Si0.7Ge0.3 substrate were systematically investigated with different Al interlayer thicknesses. The morphology, composition, and micro-structure of the Ni germanosilicide layers were analyzed with different annealing temperatures in the appearance of Al. The germanosilicide layers were characterized by Rutherford backscattering spectrometry, cross-section transmission electron microscopy, scan transmission electron microscopy, and secondary ion mass spectroscopy. It was shown that the incorporation of Al improved the surface and interface morphology of the germanosilicide layers, enhanced the thermal stabilities, and retarded the Ni-rich germanosilicide phase to mono germanosilicide phase. With increasing annealing temperature, Al atoms distributed from the Ni/Si0.7Ge0.3 interface to the total layer of Ni2Si0.7Ge0.3, and finally accumulated at the surface of NiSi0.7Ge0.3. We found that under the assistance of Al atoms, the best quality Ni germanosilicide layer was achieved by annealing at 700 °C in the case of 3 nm Al.

PACS: 85.30.Kk;85.30.De;85.30.Hi
Keyword:germanosilicide;Al;Ni
1. Introduction

The nickel mono-silicide (NiSi) has been widely used in metal–oxide–semiconductor field effect transistor (MOSFET) due to its low resistivity, low Si consumption, and low formation temperature.[1–4] However, when Ni reacts with silicon–germanium alloy (Si1−xGex), it is difficult to form a uniform Ni–germanosilicide (NSG) layer due to the different heat formation between Ni–Si and Ni–Ge.[5,6] Furthermore, the poor thermal stability of the NSG layer and Ge-out diffusion from the NSG layer will induce the agglomeration of NSG grains, which brings large contact variations and leakage to the related Si1−xGex devices.[7] Si1−xGex is already employed in source/drain (S/D)[8] or channel areas of pMOSFETs,[9] so smooth and uniform NSG/Si1−xGex interfaces are needed for device applications. In recent years, the solid-phase reactions of Ni with Si1−xGex have been studied in detail.[10,11,12] The NSG/Si1−xGex contact properties could be improved by several methods, such as ions pre-implanted into Si1−xGex,[13] introducing some other metal elements,[14,15,16] and different annealing processes.[17,18] Especially, Al incorporation is considered as a promising way to mediate the complex reactions between Ni and Si1−xGex.[19–21,22] Our previous results showed that an Al interlayer or Ni1−xAlx alloy could greatly reduce the NSG/Si1−xGex interface roughnesses.[19,21] Moreover, it has been reported that Al atoms could tune the hole Schottky barrier height at the NSG/Si1−xGex junction to sub-0.1 eV.[23] Thus, the Al impact on germanosilicidation is interesting and needs further study. Liu et al. investigated the formation of Ni germanosilicide layers on strained Si1−xGex with different Al contents in Ni1−xAlx alloy.[24] However, in the case of Al interlayer, the effects of the Al thickness and annealing temperature on the formation of NSG are not fully understood. In this paper, we systematically investigate the solid reactions between Ni and relaxed Si0.7Ge0.3 substrate with different Al interlayer thicknesses under various annealing temperatures.

2. Experiment

The fully relaxed Si0.7Ge0.3 with a thickness of about 800 nm was deposited by reduced pressure chemical vapor deposition (RPCVD) on a Si (100) substrate with graded Si1−yGey (y < 0.3) buffer layers. After standard RCA cleaning, Si0.7Ge0.3 samples were dipped in dilute HF solution (1%) to remove the native oxide. 1 nm and 3 nm Al interlayers were deposited on the Si0.7Ge0.3 substrate followed by subsequent Ni layer with 10 nm thickness at room temperature, as schematically shown in Figs. 1(a) and 1(c). The metal thicknesses were confirmed by Rutherford backscattering spectrometry (RBS) using RUMP simulation[25] (see Figs. 1(b) and 1(d)). The Si1−xGex films were so thick that the individual spectra of Ni, Al, Si, and Ge could not be resolved. Meanwhile, 10 nm Ni layer without Al was also deposited directly on the Si0.7Ge0.3 substrate for comparison. Ni germanosilicidation was formed by rapid thermal annealing (RTA) in ambient N2 for 30 s at various temperatures ranging from 400 °C to 700 °C. The un-reacted Ni was selectively etched in solution H2SO4:H2O2 (1:4).

Fig. 1. (color online) Schematic diagrams of Ni/Al/SiGe systems and the corresponding RBS measurements: (a), (b) 1 nm Al; (c), (d) 3 nm Al.

The stoichiometry of the NSG layers was investigated by RBS. The simulations of the RBS spectra were made by means of RUMP. The morphology and microstructure were studied by cross-section transmission electron microscopy (XTEM) and scanning transmission electron microscopy (STEM). The elements redistributions were analyzed by secondary ion mass spectroscopy (SIMS).

3. Results and discussion
3.1. The stoichiometry of the NSG layers

The stoichiometry of the NSG layers was investigated by RBS with 1.4 MeV He+ at a scattering angle of 170°C. Figure 2 presents the RBS spectra of the NSG layers with or without (w/o) Al interlayer at different silicidation temperatures. For the samples w/o Al, based on the RUMP program (not shown), it is simulated that the proportion of Ni and Si0.7Ge0.3 is approximately 1:1 at 400 °C, indicating that the layer is in mono-germanosilicide phase (NiSi0.7Ge0.3). After high temperature annealing (600 °C), the composition of the NSG layer shows a great change indicated by the widening of the Ni signal and the raising of the Ge signal in the RBS spectrum, which means a rough NSG/Si0.7Ge0.3 interface and an inhomogeneous layer due to agglomeration. For the sample with 1 nm Al, at 400 °C, no significant difference to the sample w/o Al is found. At 600 °C, we can see some thermal stability improvements of the NSG layer due to the appearance of Al. However, according to the RUMP simulation (not shown), the Ge concentration in the NSG layer is higher than that in the mono NiSi0.7Ge0.3 phase, indicating that the Ge atoms are already diffused from NiSi0.7Ge0.3 polycrystalline grains. In the case of 3 nm Al, the RBS random spectrum of the sample annealed at 500 °C can be simulated by an 18.5 nm thick Ni2Si0.7Ge0.3 layer. After annealing at 600 °C, a mixture of phases (Ni2Si0.7Ge0.3 and NiSi0.7Ge0.3) is found according to the RUMP simulation. These phases can totally transfer to the mono-NiSi0.7Ge0.3 phase when the annealing temperature is increased to 700 °C. Most of the Al is found in a ∼4 nm thick oxidized Al top layer (see Fig. 2(c)). Based on these results, we conclude that the appearance of Al retards the Ni-rich germanosilicide phase to the mono-germanosilicide phase.

Fig. 2. (color online) RBS spectra of NSG layers formed at different temperatures: (a) w/o Al, (b) 1 nm Al, (c) 3 nm Al.
3.2. Morphology dependence on Al thickness

In order to study the difference in morphology between Ni/Si0.7Ge0.3 and Ni/Al/Si0.7Ge0.3 systems, XTEM was employed to analyze the samples with and w/o Al (as shown in Fig. 3). For the sample w/o Al and after annealing at 400 °C, polycrystalline NSG grains are formed, which cause rough surface and interface of the layer (Fig. 3(a)). The incorporation of Al can reduce the roughness of both the surface and the interface of the NiSi0.7Ge0.3 layer. 1 nm Al interlayer is sufficient for the suppression of NSG grains formation at the surface, as revealed by the TEM micrograph in Fig. 3(b). The layer uniformity and interface roughness are improved compared to the case of Al absence. 3 nm Al interlayer results in the NSG layer with a very sharp interface (Fig. 3(c)). Note that there are still a few Ni2Si0.7Ge0.3 grains inside the NiSi0.7Ge0.3 layer, which means that the Ni-rich germanosilicide phase is not totally transformed to the mono-germanosilicide phase even under high-temperature annealing. The existence of Ni2Si0.7Ge0.3 grains increases the surface roughness of the NiSi0.7Ge0.3 layer. These results are consistent with the RBS measurements shown in Fig. 2(c), which further proves the appearance of Al can retard the mono-germanosilicide phase formation.

Fig. 3. XTEM images of NSG layers: (a) w/o Al at 400 °C, (b) 1 nm Al at 400 °C, (c) 3 nm Al at 600 °C.
3.3. Dependence on annealing temperature

In this section, we discuss the annealing temperature dependence of the NSG formation with 3 nm Al interlayer. The STEM micrographs in Fig. 4 depict the morphology of the NSG layers formed at various annealing temperatures. The NSG layer formed at 500 °C has the average thickness of ∼ 19 nm with Ni2Si0.7Ge0.3 phase corresponding to the RBS simulation (Fig. 2(c)). After annealing at 550 °C, the number of polycrystalline NSG grains is reduced and the NSG layer uniformity and the interface morphology are improved. With the annealing temperature increasing to 700 °C, the NSG layer has totally transferred to the mono NiSi0.7Ge0.3 phase with the thickness of ∼ 24 nm, as demonstrated by the bright uniform layer in Fig. 4(c). Due to the lack of Ni2Si0.7Ge0.3 polycrystalline grains, the surface of the NiSi0.7Ge0.3 layer becomes very smooth with a sharp NiSi0.7Ge0.3/Si0.7Ge0.3 interface. In order to further investigate the temperature dependence, high-resolution XTEM (HRTEM) was performed for samples annealed at 500 °C, 550 °C, 600 °C, and 700 °C respectively. Figure 5(a) displays the sample annealed at 500 °C with several different orientation Ni2Si0.7Ge0.3 polycrystalline grains, which cause the surface and interface of the Ni2Si0.7Ge0.3 layer to be very rough. After 550 °C annealing, the Ni2Si0.7Ge0.3 layer starts to become uniform with smooth interface, although the atomic steps still exist (see Fig. 5(b)). At 600 °C, main Ni2Si0.7Ge0.3 has transformed to NiSi0.7Ge0.3 and only a few Ni2Si0.7Ge0.3 grains still exist in the NiSi0.7Ge0.3 layer, as shown in Fig. 3(c). After 700 °C annealing, the NiSi0.7Ge0.3 layer has a uniform thickness over the entire viewable area. The interface between NiSi0.7Ge0.3 and Si0.7Ge0.3 substrate is sharp and flat without evidence of the formation of any second phase. As reported earlier in Ref. [19], this NiSi0.7Ge0.3 layer is epitaxial growth on the (001) Si0.7Ge0.3 substrate with the (101) orientation.

Fig. 4. STEM images of NSG layers formed with 3 nm Al at different annealing temperatures: (a) 500 °C, (b) 550 °C, (c) 700 °C.
Fig. 5. High resolution XTEM images of NSG layers formed with 3 nm Al at different annealing temperatures: (a) 500 °C, (b) 550 °C, (c) 600 °C, (d) 700 °C.
3.4. Al redistribution after germanosilicidation

To clarify the exact role of Al atoms on the germanosilicidation process, we utilized SIMS to monitor the Al, Ni, Si, Ge depth profiles in the 3 nm Al samples annealed at different temperatures. Note that the results in Fig. 6(d) are a reproduction from Ref. [22] in order to facilitate a direct comparison. As shown in Fig. 6(a), at the annealing temperature of 500 °C, the flat Ni plateaus indicate the formation of single Ni2Si0.7Ge0.3 phase. The Al distribution is also very uniform in the NSG layer, indicating that main Al atoms are expelled from the interlayer to the NSG layer during the Ni-Si0.7Ge0.3 reaction. After 550 °C annealing, some Al atoms start to move to the surface and finally form an obvious pileup of Al on the upper of the NSG layer. Consequently, the NSG layer maybe splits into two sub-layers: one layer located closer to the un-reacted Si0.7Ge0.3 substrate, and the other layer adjacent to the surface with a much higher Al concentration. Moreover, we find that there are some Si and Ge atoms accumulations at the interface of NSG/Si0.7Ge0.3 in both 500 °C and 550 °C annealing samples. However, with further increasing annealing temperature to 600–700 °C, Si and Ge atoms accumulations disappear and most Al atoms segregate at the surface of the NSG layer. Furthermore, we note that the redistributions of the Ni/Al/Si/Ge signals in the NSG layers do not change after annealing at 600–700 °C due to the formation of the most stable mono-gemanosilicide phase. The minor difference of the SIMS depth files at 600 °C and 700 °C is the remaining Al atoms distributions in the oxide layer on the top of the NSG layers. Although a few Ni-rich germanosilicide grains are formed in the NSG layer at 600 °C (as shown in Fig. 5(c), it is found that they have marginal impacts on the Ni/Al/Si/Ge contents in the NSG layer.

Fig. 6. (color online) SIMS results of NSG layers formed with 3 nm Al at different annealing temperatures: (a) 500 °C, (b) 550 °C, (c) 600 °C, (d) 700 °C.
3.5. Sheet resistance measurements

The impact of the Al interlayer on the NSG electrical properties was investigated by sheet resistance measurements. The annealing temperature dependence of the sheet resistance for the samples with or w/o Al is shown in Fig. 7. Note that the data have been partially shown in Ref. [19] and they are reproduced here only for comparison with the case of 1 nm Al interlayer. For Ni/Si0.7Ge0.3 and Ni/Al(1 nm)/Si0.7Ge0.3 systems, the Rs keeps at the value of ∼ 7 Ω sq from 400 °C to 500 °C, indicating the formation of mono-germanosilicide phase. The layer suffers pronounced degradation when the temperature increases above 500 °C. However, for the Ni/Al(3 nm)/Si0.7Ge0.3 system, the sheet resistance remains at a higher value at 400–500 °C due to the formation of a Ni-rich germanosilicide phase, as shown in Figs. 5 and 6. The sheet resistance then decreases abruptly after annealing at 500 °C, which is attributed to the transformation of Ni-rich germanosilicide phase to mono-germanosilicide phase. The thermal stability of the NSG layer increases up to 700 °C. It is evident that the incorporation of 3 nm Al retards the phase formation and enhances the thermal stability of the germanosilicide layer.

Fig. 7. (color online) Sheet resistance of NSG layers formed with and w/o Al interlayer versus annealing temperature.
3.6. Discussion

Based on the structural and electrical analysis, we now discuss the effects of the Al interlayer on the growth mechanism of NiSi0.7Ge0.3 on Si0.7Ge0.3. In the case of Si substrate, the addition of Al decreases the disilicide formation temperature and leads to a uniform orientation of the disilicide layers, which depend on the Al content and annealing temperature.[26] However, for the Si0.7Ge0.3 substrate, the incorporation of Al increases the mono-germanosilicide formation temperature and finally mediates the NSG layer with smooth interface. The Al interlayer behaves as a diffusion barrier to reduce the diffusion of Ni and induces the formation of a uniform NSG layer. Considering the difference of 1 nm and 3 nm Al on the formation of NiSi0.7Ge0.3, as shown in Figs. 3(b) and 3(c), we find that the appearance of enough Al atoms could reduce and balance the reaction of Ni and Si0.7Ge0.3, inducing a uniform epitaxial growth of NiSi0.7Ge0.3 on Si0.7Ge0.3. In addition, Jin et al. found that the surface and interface energies of NiSiGe could be minimized for highly (200) orientation by the addition of 5% Pd in NiPd alloy.[16] We propose that Al atoms have the same effect, i.e., lowering the dominant interface and surface energies, indicating that NiSi0.7Ge0.3 with favorable orientation may grow faster than NiSi0.7Ge0.3 with other orientations.

4. Conclusion

The formation of NSG layers is systematically investigated with different annealing temperatures in the appearance of Al. It is shown that both the Al interlayer thickness and annealing temperature could greatly affect the Ni-Si0.7Ge0.3 reaction. The incorporation of Al improves the surface and interface morphology of the NSG layers, enhances the thermal stabilities, and retards the mono-germanosilicide phase formation. In the case of 3 nm Al, a very uniform and smooth NiSi0.7Ge0.3 layer is formed after 700 °C annealing. The Al atoms distribute from the Ni/Si0.7Ge0.3 interface to the total layer of Ni2Si0.7Ge0.3, and finally accumulate at the surface of the NiSi0.7Ge0.3 layer. These results indicate that the combined process of Al interlayer may serve as a potential method for improving the contact properties of NiSi1−xGex/Si1−xGex.

Reference
[1] Zhang S L Östling M 2013 Criti. Rev. Solid State 28 1
[2] Lavoie C d’Heurle F M Detavernier C Cabral J C 2003 Microelectron. Eng. 70 144
[3] Luo J Qiu Z Zha C Zhang Z Wu D Lu J Åkerman J Östling M Hultman L Zhang S L 2010 Appl. Phys. Lett. 96 031911
[4] Luo J Qiu Z Zha C Zhang Z Östling M Zhang S L 2010 J. Vac.Sci. Technol. 28 C1I1
[5] Zhang S L 2003 Microelectron. Eng. 70 174
[6] Li J Hong Q Z Mayer J W Rathbun L 1990 J. Appl. Phys. 67 2506
[7] Liu J Ozturk M C 2005 IEEE Trans. Electron Devices 52 1535
[8] Packan P Akbar S Armstrong M Bergstrom D Brazier M Deshpande H Dev K Ding G Ghani T Golonzka O Han W He J Heussner R James R Jopling J Kenyon C Lee S H Liu M Lodha S Mattis B Murthy A Neiberg L Neirynck J Pae S Parker C Pipes L Sebastian J Seiple J Sell B Sharma A Sivakumar S Song B Amour A St Tone K Troeger T Weber C Zhang K Luo Y Natarajan S 2009 IEDM Tech. Dig. 659
[9] Yu W Zhang B Zhao Q T Hartmann J M Buca D Nichau A Lupták R Lopes J M Lenk S Luysberg M Bourdelle K K Wang X Mantl S 2011 Solid State Electron. 62 85
[10] Jin L Pey K L Choi W K Fitzgerald E A Antoniadis D A Pitera A J Lee M L Chi D Z Rahman M A Osipowicz T Tung C H 2005 J. Appl. Phys. 98 033520
[11] Jarmar T Seger J Ericson F Mangelinck D Smith U Zhang S L 2002 J. Appl. Phys. 92 7193
[12] Pey K L Choi W K Chattopadhyay S Zhao H B Fitzgerald E A Antoniadis D A Lee P S 2002 J. Vac.Sci. Technol. 20 1903
[13] Zhang B Yu W Zhao Q T Buca D Holländer B Hartmann J M Zhang M Wang X Mantl S 2011 Electrochem. Solid-State Lett. 14 H261
[14] Xu Y Ru G Jiang Y Qu X Li B 2009 Appl. Surf. Sci. 256 305
[15] Liu Q Wang G Guo Y Ke X Liu H Zhao C Luo J 2015 Vacuum 111 114
[16] Jin L Pey K L Choi W K Fitzgerald E A Antoniadis D A Pitera A J Lee M L Tung C H 2005 J. Appl. Phys. 97 104917
[17] Setiawan Y Lee P S Pey K L Wang X C Lim G C Tan B L 2007 Appl. Phys. Lett. 90 073108
[18] Hu C Xu P Fu C Zhu Z Gao X Jamshidi A Noroozi M Radamson H Wu D Zhang S L 2012 Appl. Phys. Lett. 101 092101
[19] Zhang B Yu W Zhao Q T Mussler G Jin L Buca D Hollaender B Zhang M Wang X Mantl S 2011 Appl. Phys. Lett. 98 252101
[20] Zhao Q T Knoll L Zhang B Buca D Hartmann J Mantl S 2013 Microelectron. Engineering. 107 190
[21] Liu L Jin L Knoll L Wirths S Nichau A Buca D Mussler G Holländer B Xu D Di Z Zhang M Zhao Q Mantl S 2013 Appl. Phys. Lett. 103 231909
[22] Ping Y X Wang M L Meng X R Hou C L Yu W L Xue Z Y Wei X Zhang M Di Z F Zhang B 2016 Acta Phys. Sin. 65 036801 in Chinese
[23] Sinha M Lee R T P Lohani A Mhaisalkar S Chor E F Yeo Y C 2009 J. Electrochem. Soc. 156 233
[24] Liu L J Jin L Knoll L Wirths S Buca D Mussler G Hollaender B Xu D W Di Z F Zhang M Mantl S Zhao Q T 2015 Microelectron. Eng. 137 88
[25] Doolittle L 1985 Nucl. Inst. Meth. 9 344
[26] Mogilatenko A Beddies G Falke M Hausler I Neumann W 2012 J. Appl. Phys. 111 103512