swift heavy ion induced effects at mo/si interface and silicide formation

74

6

Research ArticleReceived: 5 February 2009 Revised: 25 April 2009 Accepted: 10 May 2009 Published online in Wiley Interscience: 12 June 2009

(www.interscience.wiley.com) DOI 10.1002/sia.3083

Swift heavy ion induced effects at Mo/Siinterface and silicide formationGarima Agarwal,a∗ Vaibhav Kulshrestha,a Rajkumar Jain,a D. Kabiraj,b

Indra Sulania,b Pawan Kulriyab and I. P. Jaina

Swift heavy ion (SHI) induced modification at metal/Si interfaces has emerged as an interesting field of research due to its largeapplications. In this study, we investigate SHI-induced mixed molybdenum silicide film with ion fluences. The molybdenum thinfilms were deposited on silicon substrates using e-beam evaporation at 10−8 torr vacuum. Thin films were irradiated with Auions of energy 120 MeV to form molybdenum silicide. The samples were characterized by grazing incidence X-ray diffraction(GIXRD) technique for the identification of phase formation at the interface. Rutherford backscattering spectrometry (RBS) wasused to investigate the elemental distribution in the films. The mixing rate calculations were made and the diffusivity valuesobtained lead to a transient melt phase formation at the interface according to thermal spike model. Irradiation-induced effectsat surface have been observed and roughness variations at the surface were calculated using atomic force microscopy (AFM)technique. Copyright c© 2009 John Wiley & Sons, Ltd.

Keywords: SHI; molybdenum silicide; RBS; GIXRD; interface; AFM

Introduction

The reaction of metal thin films on silicon substrates has itssignificance for electrical contacts in Si-device technology becauseof the increase in scaling down of modern devices to submicrondimensions below 100 nm. Wide applications of silicides incombination with more stringent performance requirements invery large-scale integrated circuit devices increase the needto understand the process of silicide formation. In particular,the refractory metal silicides, such as MoSi2, WSi2 and TaSi2,have increased the interest in regard to the possibility of theirapplications as gate and interconnect metallization materialsin fabricating semiconductor devices,[1,2] because of their lowelectrical resistance and high chemical and thermal stability. Ionbeam mixing (IBM) has been considered as an alternative meansto form metal silicide contacts in microelectronic devices usingion beam to trigger the reaction at the interface of metal and Si.[3]

The technique is particularly pertinent for inducing the formationof refractory metal silicides because of considerable difficultiesencountered in obtaining reproducible electrical contacts andgood surface morphology by conventional furnace annealing,which would also influence treatments in device fabrication. IBMis a local rearrangement of atoms across the interfaces induced byenergetic ions, which is well understood for ions at low energiesin the range of keV where energy loss takes place due to elasticcollisions (nuclear energy loss Sn). However, swift heavy ions(SHIs) having energy ≥1 MeV induced IBM mainly via inelasticcollisions (electronic energy loss Se), leading to the excitationof the target electrons. In Fig. 1(a), the results of Monte Carlosimulation for nuclear and electronic energy losses for 120 MeVAu ions in Si have been shown. Two theoretical models, Coulombexplosion[4] and thermal spike,[5] are invoked to explain such amixing. According to the thermal spike model for semiconductors,SHI looses their energy predominantly by electronic excitationand ionization of target atoms; after about 10−14 s, part of theexcitation energy is transferred to the lattice, by electron–phonon

coupling which leads to the defects and atomic displacementsat the interface. Greater the e–p coupling factor, more will bethe energy transferred to the lattice, causing more vibrations ofthe atoms at the interface due to the interdiffusion in molten iontracks.

A number of studies on reaction kinematics between Mo andSi have been done by solid-state reaction and ion implantationtechniques, for the formation of molybdenum silicide. d’Heurleet al.[6] compare the observations made with As ion implantationof energy 250 keV through Mo/W/Si, resulting in the formationof two superimposed silicide layers of MoSi2 and WSi2, with thatof made on similar system by ordinary heat treatment. Samesystem was studied by Kwang et al.[7] using As+ ion implantationof 140–180 keV energies, which was useful to observe rapiddiffusion of the implanted ions during the formation of silicide.Cheng et al.[2] have applied XTEM to discuss the growth kineticsof MoSi2 on Si (001) including growth rates, interface structures,and microstructures after annealing process. They also discussedthe same problem by applying the effects of As+-ion-inducedmixing[8] and observed using TEM. Cheng et al.[3] have showndirect synthesis of equilibrium MoSi2 phase with a well-crystallinestructure on Si surface by Mo ion implantation using MEVVA ionsource, followed by thermal annealing. A quite similar system wasalso studied by Liang et al.[9] at a substrate temperature of 100 ◦Cduring Mo ion implantation, followed by Rapid Thermal Annealing(RTA), and the results revealed a flat and continuous molybdenum

∗ Correspondence to: Garima Agarwal, Centre for Non-Conventional EnergyResources, 14-Vigyan Bhavan, University of Rajasthan, Jaipur-302004, India.E-mail: [email protected]

a Centre for Non-Conventional Energy Resources, 14-Vigyan Bhawan, Universityof Rajasthan, Jaipur-302004, India

b Inter University Accelerator Centre, Aruna Asaf Ali Marg, New Delhi-110067,India

Surf. Interface Anal. 2009, 41, 746–752 Copyright c© 2009 John Wiley & Sons, Ltd.

74

7

Swift heavy-ion-induced effects

0 2 4 6 8 10 12 14 16 180

2

4

6

8

10

12

14E

nerg

y lo

ss p

er io

n (k

eV/n

m)

Depth (µm)

(a)

(b)

Electronic energy lossNuclear energy loss

120 MeV Au -> Si

Figure 1. Results of (a) Monte Carlo simulation for nuclear and electronicenergy losses for 120 MeV Au ions in Si and (b) energy loss distributionfrom 120 MeV Au ions in the Si/Mo/Si system. This figure is available incolour online at www.interscience.wiley.com/journal/sia.

silicide film with superior properties. Recently SHI-induced mixinghas been discussed by Bhattacharya et al.[10] in Mo/Si system byinvoking thermal spike model and they calculated the upper limitfor the threshold of defect creation by electronic energy loss inmolybdenum.

In this paper, the experimental observations and the charac-terization of the observed patterns by Au ion irradiation at Mo/Siinterfaces are reported and the possible mechanism of mixing isdiscussed.

Experimental Details

A trilayer system, Si (50 nm)/Mo (50 nm)/Si (30 nm), has beendeposited onto Si (100) substrate using electron beam evaporationat vacuum of 4 × 10−8 torr at Inter University Accelerator Centre(IUAC), New Delhi. The deposition rate was 0.2 Å/s for Si and1.0 Å/s for Mo. The top Si layer was to protect the whole systemby oxidation. This system was then irradiated with 120 MeV Au

1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4 2.6101

102

103 As-deposited Mo/Si System

Ref

lect

ivity

(lo

g sc

ale)

q (deg)

Figure 2. X-ray reflectivity curve of the as-deposited Mo/Si system.

30 40 50 60 70 800

40

80

120

160

200

240

280

320

As Deposited

a3

a2

a1

Mo (400)

Mo (211)

Mo (110) Mo5Si3 (521)In

tens

ity (

conu

ts)

2q

as deposited

a1=5x1012ions/cm2

a2=5x1013ions/cm2

a3=1.5x1014ions/cm2

Figure 3. Grazing incidence x-ray diffraction curves of the as-depositedand irradiated Mo/Si system. This figure is available in colour online atwww.interscience.wiley.com/journal/sia.

ions using 15 UD Pelletron Accelerator at IUAC New Delhi, India,in material science chamber at different fluences from 1012 to1014 ions/cm2 at room temperature under 10−6 torr vacuum.Irradiation flux was kept low to avoid sample heating duringirradiation. The electronic and nuclear energy losses were 36.64and 0.69 keV/nm for Mo and 12.94 and 0.21 keV/nm for Si, whichhave been calculated using SRIM 2003 program.[11] In the case ofSHI-induced interaction with solid, the electronic energy loss isdominant over the nuclear energy loss. Figure 1(b) shows thedistribution of 120 MeV Au ions within the Si/Mo/Si system.An x-ray reflectivity (XRR) pattern has been made to ensurethe deposited thickness of as-deposited Mo/Si system. Grazingincidence x-ray diffraction (GIXRD) measurements were performedusing CuKα radiation. The elemental distribution in all sampleswas determined by Rutherford backscattering spectroscopy (RBS)using 3-MeV He+ ion beam at backscattering angle 170◦ of theDynamitron Accelerator, at the University of Stuttgart, Germany.

Surf. Interface Anal. 2009, 41, 746–752 Copyright c© 2009 John Wiley & Sons, Ltd. www.interscience.wiley.com/journal/sia

74

8

G. Agarwal et al.

150 200 250 3000

1000

2000

3000

4000

5000

Oxygen

Top SiSi

Yie

ld

Channel

150 200 250 3000

1000

2000

3000

4000

5000

Oxygen

Top Si

Si

Yie

ld

Channel

150 200 250 3000

1000

2000

3000

4000

5000

Oxygen

Top Si

Si

Yie

ld

Channel

350 375 400 425 4500

2000

4000

6000

8000

10000

12000

14000Mo

---- unirradiated

---- irradiated at 5e12

Yie

ld

Channel

350 375 400 425 4500

2000

4000

6000

8000

10000

12000

14000

yield reduction

---------- irradiated at 1.5e14

Mo

Yie

ld

Channel

(a)

(b)

(c)

350 375 400 425 4500

2000

4000

6000

8000

10000

12000

14000

yield reduction

----------

Mo

Yie

ld

Channel

unirradiated

irradiated at 5e13

unirradiated

Figure 4. RBS spectra of as-deposited and irradiated Mo/Si system at fluences: (a) 5 × 1012, (b) 5 × 1013, and (c) 1.5 × 1014 ions/cm14. This figure isavailable in colour online at www.interscience.wiley.com/journal/sia.

The irradiated and the as-deposited thin films were characterized

in a single RBS run. Fits to the spectra were obtained with the

help of the RUMP computer code.[12] Irradiation effect on the

surface morphology of the Mo/Si thin film system has been

studied using atomic force microscopy (AFM) in the tapping

mode.

Results and Discussion

X-ray reflectivity

Figure 2 shows the XRR profile of the as-deposited sample inwhich Bragg reflection due to trilayer periodicity is clearly visibleup to third order. Mo thickness calculated from XRR profile was

www.interscience.wiley.com/journal/sia Copyright c© 2009 John Wiley & Sons, Ltd. Surf. Interface Anal. 2009, 41, 746–752

74

9


0 2 4 6 8 10 12 14 160

5

10

15

20

25

30

35

∆s2

(nm

2 )

Ion fluence (1013 ions/cm2)

Figure 5. Variation of interface variance �σ 2 with increase in ion fluenceas calculated from RBS data.

Table 1. The calculated structural parameters of the Mo and t-Mo5Si3thin films on Si (100) irradiated with 120 MeV Au ions at differentfluences

Fluence (ions/cm2) 2θ d (Å) hkl D (Å)δ × 1015

(lin/m2)

Mo 0 (Pristine) 40.42 2.256 110 165 3.67

1.5 × 1014 40.47 2.253 110 173 3.34

Mo5Si3 5 × 1013 55.03 1.698 521 66 22.96

1.5 × 1014 55.13 1.695 521 78 16.44

found to be ∼50 nm and was the same for Si layer also. Hence, thedeposition was in the right order of accuracy.

Structural information by GIXRD

Grazing incidence was employed to prevent the radiationpenetrating far in to the substrate, thus enabling to hide thesubstrate peak of Si. Figure 3 shows the x-ray diffraction patternsof the as-deposited as well as irradiated thin films with differentfluences. The as-deposited XRD pattern gives separate crystallineBragg peaks of Mo corresponding to planes (110, 400, and 211).This system is irradiated at different fluences as a1 = 5 × 1012,a2 = 5×1013, and a3 = 1.5×1014 ions/cm2. The sample irradiatedat the lowest dose of 5 × 1012 ions/cm2 shows a broad peak ofMo5Si3 with minimum intensity and it increases with increasingthe fluence. At the highest dose of 1.5 × 1014 ions/cm2, a sharppeak of tetragonal-Mo5Si3 was formed, whereas the other twocrystalline peaks of Mo with planes (110) and (211) remain at thesame position. The diffraction peak at 2θ = 55.13◦ correspondingto (521) plane allows the positive identification of Mo5Si3 phase(JCPDS X-Ray Powder file data file 34–0371). The calculated dvalues for the different ion fluences are given in Table 1. Thecrystallite size (D) was calculated using the Scherrer formula.[13]

The dislocation density (δ) was evaluated from the followingrelation:[14]

δ = 1

(D2)(1)

Table 2. Mixing width, interface variance, and mixing rate with ionfluence

Fluence� (ions/cm2)

Mixingwidth X

(nm)

Interface variance�σ 2 (�) = σ 2

(�) − σ 2 (0)(nm2)

Mixingrate k = �σ 2

(�)/�(nm4/ion)

Unirradiated 0 – ∼21.3

5 × 1012 1.5 2.25

5 × 1013 3.4 11.56

1.5 × 1014 5.7 32.49

It is to be noted that the weak peak of Mo5Si3 is also observedin the films irradiated at the lowest fluence 5 × 1012 ions/cm2.The dominant peaks of Mo5Si3 phase confirm that films aremainly composed of Mo5Si3. Mo5Si3 has a tetragonal structurewith the same a and b axes (a = b = 9.648 Å, c = 4.913 Å).From the evaluated microstructural parameters (Table 1), it is clearthat the crystallite size of Mo increases in the irradiated sample incomparison of pristine one, while the dislocation density decreaseswith the increase in crystallite size. In addition, the crystallite sizeof Mo5Si3 increases in the case of irradiation reaching to higherfluences. This may be due to the sudden increase in temperature atthe irradiation regime. However, the dislocation density decreaseswith increasing ion fluences. Such a decrease in dislocation densityis the consequence of the decrease in concentration of latticeimperfection due to the increase in crystallite size[15] or it may bedue to the lattice mismatch between t-Mo5Si3 and Si substrate.Hence, XRD measurements revealed that irradiation promotes themovement of atoms at the interface, leading to the formation of amixed compound in the form of metal silicide.

RBS analysis

The RBS spectra of as-deposited and irradiated Mo/Si thin filmsystem at various fluences are given in Fig. 4. Sample irradiated atthe lowest dose of 5 × 1012 ions/cm2 shows a very little shift in thepeaks of Mo and Si, which are not clearly observable but with thehelp of RBS data and its calculation using RUMP code a little mixingof the order of 1.5 nm is determined. In the case of irradiation atfluences 5 × 1013 and 1.5 × 1014 ions/cm2, a little decrease inthe slope and shift toward the lower edge of Mo peak and infront edge of Si peak indicate that the mixing occurred at theseinterfaces. This peak-shifting-induced mixing is clearly visible inFig. 4, where yield was taken at y-scale. These profiles for Mo/Sisystem were extracted from the RBS spectra with the help of RUMPsimulation code. They were fitted with a Gaussian error functionand the values for mixing length, interface variance, mixing rate,and mixing efficiency corresponding to each irradiation dose aretabulated in Table 2. The observed increase in the variance wascalculated by

�σ 2 = σ 2 (φ) − σ 2 (0) (2)

where σ 2(�) and σ 2(0) are the variances of irradiated sample atfluence � and unirradiated sample, respectively. The maximummixing length corresponds to the formation of t-Mo5Si3 at Mo/Siinterface. It is clear from Fig. 5 that the value of �σ 2 increaseslinearly with increase in ion fluence, hence supporting the results

of GIXRD. The mixing rate is defined as k = σ 2(φ)φ

, which was

calculated from the slope of the linearly fitted line of Fig. 5 as shownin Table 2. Assuming the thermal spike model for semiconductor


75

0

G. Agarwal et al.

Figure 6. Roughness analysis of pristine Mo/Si system. This figure is available in colour online at www.interscience.wiley.com/journal/sia.

Figure 7. Roughness analysis of Mo/Si system irradiated at fluence 5 × 1012 ions/cm2. This figure is available in colour online at www.interscience.wiley.com/journal/sia.

system,[16] each incident ion loses its energy during its transit acrossthe material mainly by electronic excitation. The temperature ofelectronic system rises for a typical time span of 10−15 s, whichis followed by transfer of energy from electrons to the target viathe electron–phonon coupling, inducing the local temperaturerise (≥1000 K) in a very narrow zone. The rise and decay ofthe lattice is governed by two coupled differential equations.[5]

For the temperatures above the melting point, the lattice melts,followed by fast quenching. The melting of lattice facilitatesatomic motion through this zone and an enhanced diffusion takesplace dominantly in each cylindrical melt zone, as the diffusivityin molten state is several orders of magnitude higher than that inthe corresponding solid. Thermal spike model has also been usedpreviously to explain IBM in various metal/Si systems.[17 – 19] Thediffusion during transient thermal spike follows the equation to

calculate the diffusion coefficient as

D = �σ 2(φ)MI

2ts(3)

where ts is the duration of the transient melt phase and �σ 2 (�)MI

is monoion mixing effect,[20] which is defined as

�σ 2(φ)MI = k

(π r2)(4)

where k is the mixing rate and r is the radius of the molten iontrack. Taking the typical track radius to be 1.5 nm and ts to be∼1 ps for another metal Ni, the data for which is available,[21]

the diffusivity D at the interface turns out to be of the order of∼10−6 m2/s. The value is a characteristic and is clearly in the range


75

1


Figure 8. Roughness analysis of Mo/Si system irradiated at fluence 5 × 1013 ions/cm2. This figure is available in colour online at www.interscience.wiley.com/journal/sia.

Figure 9. Roughness analysis of Mo/Si system irradiated at fluence 1.5 × 1014 ions/cm2. This figure is available in colour online at www.interscience.wiley.com/journal/sia.

of the interdiffusion constants known from liquids and hencesupports the thermal spike model of mixing.

Surface morphology of Mo/Si thin films

The AFM measurements were carried out by using NanoscopeIIIE model from Digital Instruments, USA, at IUAC, New Delhi. Thetapping mode was used to study the surface modification on theAu-ion-irradiated Mo/Si system. AFM examinations revealed someinteresting features that emerged on the top Si surfaces afterSHI irradiation. For the surface morphology of all samples, thescanning area was taken to be 1 µm × 1 µm. Surface morphologyof pristine Mo/Si system is shown in Fig. 6; there is no correlatedstructure growth. It shows smooth surface; however, the rmssurface roughness was measured to be 0.604 nm, which is a small

quantity. Roughness analysis for the samples irradiated at a doseof 5 × 1012 ions/cm2 is depicted in Fig. 7, which shows somekind of grain formation, which is just started at this lowest dose.The rms surface roughness is measured to be 0.987 nm, whichis almost equal with the pristine one. It can be said that thesurface is almost smooth for the lowest dose irradiation. In thesection analysis, the grain size is measured to be 34.249 nm. Itcan be seen from the roughness analysis or surface morphologyof the samples irradiated at a dose 5 × 1013 ions/cm2 (Fig. 8)that the grain formation has now increased and the grains gotelongated to form cylindrical shape on increasing the dose. Therms surface roughness is measured to be 3.843 nm, which is higherthan the value obtained from previous irradiated sample. In thesection analysis, the grain size is measured to be 100.33 nm. Itcan be concluded that considerable amount of mixing has taken


75

2

G. Agarwal et al.

place at the surface also, which grows with the increased fluence.From Fig. 9 it is seen that when the sample is irradiated at thehighest dose of 1.5×1014 ions/cm2, the calculated rms roughnessincreased to 4.065 nm. However, from the section analysis, grainsize decreased up to 84.894 nm. It can be concluded that onirradiation the mixing has taken place at the interface, leadingto the surface. The grain formation started at a lower dose andthen it is increased as the system initially gets amorphized and atthe highest dose it is converted into pure crystalline form for theformation of compound, resulting in the reduced grain size. Theabove discussion gives us an idea that IBM is an advance tool toproduce irradiation-induced effects at the surface as well as at theinterface in the form of compound formation.

Conclusion

This work describes the mixing behavior of Mo and Si under SHIirradiation using 120 MeV Au ions. XRD results indicate that SHI-induced mixing has taken place at the interface of Mo/Si system,which results in the formation of t-Mo5Si3 that is not a stable phasein the series of Mo/Si interface reacted phases, but it is suitable formicroelectronics. From RBS analysis, the diffusivity value suggestsa transient molten phase at the interface following the thermalspike model. AFM results revealed an increase in surface roughnessand grain size, which increased on ion irradiation up to a certainlevel, and then grain shape converted to a particular size showingthe mixing effect at the surface also.

Acknowledgements

The authors would like to thanks S. R. Abhilash for his valuablehelp during the sample preparation. The authors are also thankfulto the Pelletron Group of Inter University Accelerator Centre (IUAC)for providing a stable beam for uniform irradiation during the

experiment. The financial support of CSIR in the form of fellowshipis also gratefully acknowledged. The authors would also like tothank H. Paulus of University of Stuttgart, Germany, for providingRBS facility.

References

[1] A. K. Vasudevan, J. J. Petrovic, Mater. Sci. Eng., A 1992, 155, 1.[2] J. Y. Cheng, H. C. Cheng, L. J. Chen, J. Appl. Phys. 1987, 61, 2218.[3] X. Q. Cheng, R. S. Wang, X. J. Tang, B. X. Liu, Nucl. Instrum Methods

Phys. Res., Sect. B 2003, 207, 301.[4] S. Klaumunzer, G. Ming-dong Hou, Schumacher Phys. Rev. Lett. 1986,

57, 850.[5] Z. G. Wang, C. Dufour, E. Paumier, M. Toulemonde, J. Phys.: Condens.

Matter 1994, 6, 6733.[6] F. M. d’Heurle, M. Y. Tsai, C. S. Petersson, B. Stritzker, J. Appl. Phys.

1982, 53, 3067.[7] D. L. Kwang, D. C. Mayers, N. S. Alvi, L. W. Li, E. Norbeck, Appl. Phys.

Lett. 1985, 47, 688.[8] J. Y. Cheng, H. C. Cheng, L. J. Chen, J. Appl. Phys. 1987, 62, 3722.[9] J. H. Liang, C. T. Tsai, Nucl. Instr. Meth. Phys. Res. B 2006, 242, 598.

[10] D. Bhattacharya, G. Principi, A. Gupta, D. K. Avasthi, Nucl. InstrumMethods Phys. Res., Sect. B 2006, 244, 198.

[11] http://www.srim.org/.[12] L. R. Doolittle, Nucl. Instrum Methods Phys. Res., Sect. B 1985, 9, 344.[13] G. B. Williamson, R. C. Smallman, Philos. Mag. 1956, 1, 34.[14] G. Gardillo, J. M. Florez, L. C. Hernandez, Sol. Energy Mater. Sol. Cells

1995, 37, 273.[15] N. El-Kadry, A. Ashour, S. A. Mahmood, Thin Solid Films 1995, 269,

112.[16] D. K. Avasthi, W. Assman, A. Tripathi, S. K. Srivastava, S. Ghosh,

F. Gruner, M. Toulemonde, Phys. Rev. B 2003, 68, 153106.[17] K. Diva, D. Kabiraj, B. R. Chakraborty, S. M. Shivaprasad, D. K.

Avasthi, Nucl. Instrum Methods Phys. Res., Sect. B 2004, 222, 169.[18] G. Agarwal, P. Sharma, I. P. Jain, Appl. Surf. Sci. 2006, 252, 1165.[19] G. Agarwal, P. Sharma, A. Jain, C. Lal, D. Kabiraj, I. P. Jain, Vacuum

2009, 83, 397.[20] S. Kumar, R. S. Chauhan, S. A. Khan, W. Bolse, D. K. Avasthi, Nucl.

Instrum Methods Phys. Res., Sect. B 2006, 244, 194.[21] F. Shi, W. Bolse, K. P. Leib, J. Appl. Phys. 1995, 78, 230.


swift heavy ion induced effects at mo/si interface and silicide formation

Documents