1

Silicidation of Co/Si Core Shell Nanowires

Han-Bo-Ram Leea, Gil Ho Gub, Chan Gyung Parkb, and Hyungjun Kima*

aSchool of Electrical and Electronic Engineering, Yonsei University, 262 Seongsanno, Seodaemun-gu, Seoul, Korea (120-749)

bDepartment of Materials Science and Engineering, POSTECH, San-31, Nam-gu, Pohang, Korea (790-784)

ABSTRACT

The formation of CoSi2 nanowires (NWs) via the annealing of Si NWs conformally coated with atomic layer deposited(ALD) Co layerswas investigated. Co ALD was carried out using Co(iPr-AMD)2 and NH3 as a precursor and a reactant, respectively. Rapid thermal annealing (RTA) of Co/Si core-shell NWs produced Co oxide instead of CoSi2 due to oxygen contamination during annealing. To prevent oxygen contamination, a highly conformal ALD Ru layer was used as a capping layer. X-ray diffraction showed the formation of CoSi2when the Co/Si core-shell NWs were annealed at over800 °C. To investigate the deposition and silicidation process on a nanometer scale, high resolution Cs-corrected scanning transmission electron microscopywas utilized with electron energy loss spectroscopy and energy dispersion spectroscopy. In contrast to previous reports on silicidation of NWs, the dominant diffusion species was found to be Si instead of Co based ona nucleation-controlled reaction.

Corresponding email:

1. Introduction

Silicidesare important materials forcontactsin modern Si devices due to theirlow contact resistance and good compatibility. Silicide is usually formed by metal deposition on a Si layer followed by annealing at the appropriate temperature through a solid-state reaction.1Recently, silicide nanowires (NWs) have attracted great attentionsince they havethe high conductivity and compatibility with metalrequired for contacts in nanowire devices. Silicide NWsare usually synthesized bychemical vapor deposition (CVD). For instance, three kinds of cobalt silicide NWs, Co2Si, Co3Si, and CoSi, were synthesized by the reaction of CoCl4 vapor and Si substrate.2Also, SiH4 gas was reacted with Ni thin film to form NiSi NW.3, 4 Similarly, several silicide NWs have been reported, such as TiSi25, Ti5Si36, TaSi27,Fe5Si38, FeSi9, and Ni2Si10.Another approach of synthesizing silicide NWs is viasolid-state reaction between metal and Si NWs. This method has a benefit of enabling contact formation for NW based device fabrication. However, there have been few studieson the silicide NW synthesis using solid-state reaction between Si NWs and metal films.

Recently, the formation of Co and Ni silicide NWs by in situ annealing of point contact between Si NWs and metal in a UHV-TEM system was reported, focusing on the silicide formation kinetics.11-13The epitaxial layer-by-layer growth of CoSi2 along the Si NW axis was observed, which was explained by the limited supply of Co atoms due to point contact.12In that report, it was found that having a lower interface energy between the native oxide and Si than between native oxide and silicide prevented heterogeneous nucleation, resulting in homogeneous nucleation.13Meanwhile, silicidation of a single Si NW partially covered with Pt deposited by evaporation was reported focusing on the effects of annealing environment.14However, due to the limited coverage of evaporation technique, uniform silicide NWs could not be synthesized.

Atomic layer deposition (ALD) is a promising deposition technique for nanoscale 3D device fabrication due to its excellent conformality and thickness controllability.15 Recently, we reported high quality Co ALD using bis(N,N’-diisopropyl-acetamidinato)-cobalt(II) [Co(iPr-AMD)2] and NH3 as a precursor and a reactant, respectively.16 In the present study, we utilized Co ALD to conformally coat Si NWs and investigated the Si NWs silicidation via annealing through solid-state reaction between Si NW and ALD Co layer. The microstructure change was analyzed by X-ray diffraction (XRD) and scanning transmission electron spectroscopy (STEM). To prevent oxygen contamination during annealing, we employedan ALD Ru film as a capping layer.

2. Experimental Details

In this study, a commercial ALD chamber (Quros Plus150TM) with a loadlock chamber was used. Two metal-organic precursors for Co and Ru ALD processes were in separatebubblers. For Co ALD, Co(iPr-AMD)2 and NH3 were used as theprecursor and reactant, respectively. One ALD cycle was composed of four steps consisting of precursor exposure (ts), purging (tp), reactant exposure (tr), and purging, where the flow of Ar purging gas was 50 sccm. The ts, tr, and tpwere 3 s, 6 s, and 1 s, respectively, and typical deposition temperature was 300 °C. The precursor for Ru ALD was DER (Ru(DMPD)(EtCp)) with O2counter reactant. The DER precursor diluted with ethylcyclohexane was delivered into the chamber with Ar carrier gas through liquid injection system and vaporizer. More information and detailed experimental setup for Co and Ru ALD can be found in our previous reports.16, 17

The Si(001) substrates were routinely used for silicidation experiments on planar substrate. The substrates were cleaned by dipping in buffered oxide etchant (6 : 1) for 10 s to remove native oxide, and followed by deionized (DI) water rinsing and N2 blowing resulting in H-terminated Si. After cleaning, Si(001) substrateswere immediately loaded into the chamber to prevent the formation of a native oxide layer. Si NWs were fabricated by CVD with Au catalysis and SiH4 gas exposure. The Au catalyst was formed by agglomeration of very thin Au film deposited on SiO2 substrate by thermal evaporation. When the Co film was deposited on Si NWs, the same cleaning step was repeated to remove the native oxide formed around Si NWs. For silicidation experiment, the Ru capping layer was deposited in situ after Co ALD without breaking vacuum. The specimens were annealed byrapid thermal annealing (RTA) at various annealing temperatures (Ta) from 400 °C to 900 °C for 30 s with N2 ambient. To remove residual O2 inside RTA chamber, N2 flushing was carried out for 5 min after specimen loading. Most RTA experiments were performed at atmosphere but several silicidation was carried out under vacuum to prevent O2 contamination.

Field emission scanning electron microscopy (FE-SEM) was used to analyze the morphology. The resistivity of the Co films was measured by 4-point probe. The change of microstructure to silicide phases was analyzed by XRD with a synchrotron source (Pohang acceleration laboratory, 3C2 beam line) as well as Cu k1. For the nanoscale structural and chemical composition analysis, high resolution Cs-corrected STEM (JEOL JEM-2100F) with electron energy loss spectroscopy (EELS) and energy dispersive spectroscopy (EDS) was employed. The STEM was operated at 200 kV and its probe size was reduced to 0.1 nm for precise EELS analysis.

3. Results and Discussions

Figures 1a and 1b show SEM images obtained before and after Co ALD on Si NWs, respectively. The diameters of individual Si NWsare approximately 20 nm as shown in Figure 1a before Co deposition. 833 growth cycles were carried outto deposit 20 nm-thick Co film on the Si NWs, based on the growth rate of 0.24 Å/cycle. The Co ALD produced Si NWs uniformly coated with a Co layer as shown in Figure 1b.The final diameter of Si NWs wasapproximately 60 nm, indicating that Co layer with 20 nm of thickness was conformally deposited all around Si NWs. Although the Si NWsare each over a few micrometers long and entangled with one another,it was confirmed that the diameter of Co coated Si NW is almost constant due to the excellent conformality of the current ALD Co process.16In this case, the conformality was defined by the ratio of Co/Si NW diameters at upper part and lower part of NWs. Higher magnification SEM image can be found in supplementary materials (Figure S1a).18 Because ALD Co has relatively rough surface morphology on Si surface as shown in Figure S1b, the Si NWs covered with ALD Co look like rugged wire.18

The core-shell Co/Si NWs were annealed using RTAunder N2 ambient. Typically, in the annealing of Co films on Si substrates, phase transformation occurs by the following sequence: metal-rich silicide (Co2Si), monosilicide (CoSi), and disilicide (CoSi2).1 Sometimes, the formation of Co2Si is not observed. The formation temperature of CoSi is between Ta = 400 C and 600 C, and CoSi2begins to form at Ta 700 C. Also, in our previous study on silicidation of Si substrates with PE-ALD Co, CoSi2 formation was observed at over Ta = 700 C.19Thus, the specimen was annealed at Ta = 900 C which is known to be a sufficiently high temperature to form CoSi2. Figure 2 presents the XRD spectrum of the annealed sample showing two XRD peaks at 31.2 and 36.8 , corresponding to Co3O4(220) and Co3O4(331), respectively.

The Co/Si core-shell NWsannealed at Ta = 900 C were further analyzed by STEM-EELS. The TEM image in Figure 3a shows that the inner core seems to remain intact with a 20 nm diameter.STEM-EELS was carried out along the black line perpendicular to the NW shownin Figure 3a. The STEM-EELS shows that the Si signal was only observed at the center of NW within the 20 nm core region,as seen in Figure 3b. Co was detected along the whole scan line and the relative intensity was higherin the outer layer region than in the core region. A similar trend was observed for O, indicating that Co and O existed in the outer layer due to the circular shape of the sample cross section. Thus, we infer that Si and Co did not react under this condition and Co reacted with O to form a Co oxide shell, which is corroborated by the XRD result. We tried to remove the oxygen contamination by annealing under vacuum. However, similar results were observed.

It is well known that the oxygen contamination disturbs the formation of Co silicides. Thus, it is a common practice that a Ti capping layer, prepared by sputtering or evaporation,is used for preventing O2 contamination and scavenging residual oxygen in the film. However, conventional Ti deposition process cannot be used as a capping layer for annealing core-shell NWs due to the limited conformality. It should be noted that Ti ALD processeshave rarelybeen reported. Thus, an alternate capping material, which can be deposited by ALD, should be employed for the current process scheme. Ru is a stable material which can be easily prepared by the ALD process.17 Although Ru has little tendency to scavenge oxygen like Ti, Ru has high resistance to oxygen at high temperatures. Thus, ALD Ru was chosen as an alternative capping layer to Ti. Separate experiments for the feasibility of Ru capping layer were carried out for ALD Co film on Si substrate, confirming the silicide formation at Ta 800 C, which will be published elsewhere.

Ru/Co/Si core-shell NWs were prepared by carrying out an additional 500 cycles of ALD Ru following the Co ALD without vacuum break. Figure 4a shows the SEM image of Si NWs sequentially coated with ALD Co and ALD Ru. The STEM bright field image in Figure 4b shows a cross-section view of a core-shell NW composed of two polycrystalline layers. The STEM-EDS mapping on this NW shown in Figure 4cclearly shows that the inner layer is Co and outer layer is Ru. The thickness of each layer is 20 nm, and total diameter of this NW including two layers is measured to be around 100 nm. The outer most Ru shell has a thickness of 20 nm, which agrees with the expected value calculated from the growth rate of Ru ALD, 0.4 Å/cycle.

Ru/Co/Si core-shell NWs were annealed using RTA at various temperatures from Ta = 400 ºC to 900 ºC and synchrotron radiation XRD analysis was carried out as shown in Figure 5. In contrast to annealed NWs without Ru capping, two strong diffraction peaks were observed at 28.1 º and 35.2 º, corresponding to CoSi2(111) and (200), respectively, at Ta≥800 ºC.20, 21 The three additional peaks at 38.4 º, 42.2 º, and 44.1 º were indexed as Ru(100), (002), and (101), respectively.Figure 6a shows SEM image of Ru/Co/Si core-shell NW after RTA at Ta = 900 ºC. The surface morphology of the annealed NW is more textured and rougher compared to the as-prepared NW in Figure 4. In Figure 5, Ru XRD peak becomes stronger and narrower with increasing Ta. Thus, the change of morphology can be understood by the recrystallization of Ru capping layer during annealing. The diameter change of Ru/Co/Si NWsafter RTA was plotted with annealing temperature in Figure 6b. The diameters of core-shell NWs were measured from SEM images, and the average diameter and deviation were calculated by measurement on different NWs in the same SEM image.The average diameter and deviation of as-preparedRu/Co/Si NWs was 102.4 ± 5.6 nm, and are denoted by the dashed line and shading in Figure 6b. Diameters of core-shell NWs remained almost same up to Ta = 700 ºC. At Ta = 800 ºCand 900 ºC, however, diameters increased abruptly reaching 130 nm. Together with XRD data in Figure 5, it can be concluded that phase transformation into CoSi2 leads to a volume expansion of Ru/Co/Si NWs along the radial direction.

The Ru/Co/Si core-shell NWsannealed at Ta = 900 ºC were further analyzed by STEM with EDS mapping. Figure 7a shows a cross sectional STEM bright field image of the annealed core-shell NWs. The image shows that the cross section is composed of a dark ring-shape region between bright core and shell regions. The diameter of NWs is larger than the pre-annealed NWs, agreeing with the SEM results. Figure 7bshows a dark field STEM image and Figure 7c displays the EDS mapping result. Two hollow regions are clearly observed in the center of cross-section area in the DF image, which are denoted by the black arrows.In addition, the other NW in the bottom of the image in Figure 7a shows a hollow region as well. The EDS mapping indicates that the ring-like shape is composed of Ru and the core and shell are composed of Co and Si. Figure 7d is an HR-TEM image of the region denoted by a square in Figure 7b. Various lattice fringes from crystal orientations were observed, indicating that the annealed NWs have a polycrystalline microstructure. The spacing of lattice fringes in Figure 7d is 3.1 Å, corresponding to the d-spacing of CoSi2(111) plane, which is consistent with the XRD results shown in Figure 5.

The presence of both a shell and a core region of CoSi2indicates that there may be interdiffusion between Ru coatingand Co/Si core during silicidation. The detailed mechanisms on the inward-diffusion of Ru and outward-diffusion of Co/Si are still not clear. The formation enthalpy ofCoSi2 (-98.7 kJ/mol)is larger than that of Ru silicides (RuSi (-58.1 kJ/mol), and Ru2Si3 (-60.7 kJ/mol)) in the reported thermodynamic data.22, 23In addition, a solid solution exists in the phase diagram of Ru and Co form.24Thus, it is plausible that the interdiffusion occurs during the annealing, which results in the formation of CoSi2 in core and shell.

Kinetically, silicidationoccurs through two different reaction modes: diffusion-controlled and nucleation-controlled reactions. Laterally uniform growth is typically observed in diffusion-controlled reactions in which reaction rate at the interface between the metal (or silicon) and the silicide is much higher than the diffusion rate. For instance, epitaxial CoSi2 fabricated by utilizing a diffusion-controlled reaction shows atomically sharp interface.25, 26 On the contrary, non-uniform nucleation occurs in nucleation-controlled reactions. In the previous reports on silicidation of NWsthrough point contact under UHV system, the reaction occurs through a diffusion-controlled reaction due to the limited contact area between Si and metal, resulting in the formation of epitaxial CoSi2 with atomically sharp edge.13, 27 Since the current core-shell NW system has large contact area between Si NW and Co layer, however, there is little barrier for diffusion. So, the silicidation is thought to be governed by a nucleation-controlled reaction resulting in polycrystalline CoSi2 formation.

Kirkendall voids are formed when there is large difference in diffusion flux.28 Thus, when the outward-diffusion from the core part of core-shell NW is faster than the inward-diffusion from the outer layer,voids can be formed inside of the inner shell. The formation of nanocavities when annealing oxide coreshell NWs has previously been reported.29, 30Also, in the previous reports on silicidation, voids were observed in nanoparticles as well as planar system.28, 31, 32Thus, the formation of voids in the coreshellcan be explained by the faster Si outward-diffusion than Co inward-diffusion, resulting in the Kirkendall void formation.If Co atoms diffuse out to Ru, voids can be generated at the interface between Co and Ru. However, since the voids were mainly observed in the CoSi2 region, it is likely that the voids are formed by Si outward-diffusion. Until now, researchers have intensively investigated what material is the dominant diffusion species (DDS) during the formation of silicides in a thin film system. Usually, marker experiments have been employed to determine the DDS.33-35 While it is generally accepted that the DDS for CoSi is Si, there are contradicted reports on DDS for CoSi2 and bothCo36, 37 or Si38-40 have been reported. DDS can be changed depending on various factors, such as diffusion-controlled or nucleation-controlled reaction mechanisms and strain caused by volume mismatch.41In this system, it appears thatCo and Si interdiffuse into each sides and that Si is DDS in our system.

Also, we need to consider the volume change during silicidation. Because lattice constant of Si (5.43 Å) is almost same as that of CoSi2 (5.36 Å), the volume change in the Si corecaused by Co atoms is not significant. However, the volume changeduring the transformation from Co (3.51 Å) to CoSi2 in Co layerby Si diffusionissignificant.If Si atoms in a NW diffuse into Co shell side and the space occupied by Si NW is preserved in core region despite of Si outward-diffusion, it can be calculated with assuming single crystalline microstructure thata 1 nm-thick Co shell deposited around a 20 nm diameter Si core is changed to 3.2 nm-thick CoSi2. This volume change is more significant in polycrystalline microstructure system. This volume expansion at the Co outer layer may induce tensile stress to the outermost Ru layer, resulting in breaking of Ru layer as shown in Figure 7.In addition, a recrystallization and grain growth of Ru during annealing can contribute the total volume change. Because our system is 1-D NW system, the growth of grains to the tangential direction is limited leading to volume expansion to the radial direction, resulting in large volume change.

Therefore, the CoSi2 NWswere formed by dominant diffusion of Si atom into Co shell side, resulting in the formation of void in core region. In addition, the diameter of core-shell NW is changed after annealing because the silicidation mainly takes place in the Co shell region by diffusing Si atoms.

4. Summary

The formation of CoSi2 NWs was investigated by direct silicidation of Ru/Co/Si core shell NW prepared by ALD followed by RTA. The ALD Co film was deposited from Co(iPr-AMD)2 and NH3 as a precursors and reactant, respectively. After RTA of Co/Si core-shell NW, only Co oxide was observed instead of CoSi2 due to oxygen contamination. To prevent oxygen contamination, a highly conformal ALD Ru film was used for capping layer. CoSi2 diffraction peaks were observed in XRD results, and confirmed from STEM images. In the contrast to previous NW silicidation results, the dominant diffusion species is Si instead of Co atom.

5. Acknowledgements