- Email: [email protected]
SiO2 system
402
Nuclear
Instruments
and Methods
in Physics Research
B27 (1987) 402-409
North-Holland,
ION BEAM MIXING EFFECTS G. BATTAGLIN
IN THE Ar +-IRRADIATED
I), S. LO RUSSO
‘) , A . PACCAGNELLA
Fe/SiO,
Amsterdam
SYSTEM
2), G. PRINCIPI
3, and P.Q. ZHANG
4)*
‘) Unit& CISM-GNSM
del CNR, Dipartimento di Fisica dell’liniversitri, Via Marzolo n. 8, 35131 Padova, Italy 2, Facoltd di Ingegneria, Universitd di Trento, Mesiano (Trento), Italy -‘, Istituto di Chimica Industriale dell’Universitci di Padova, Via Marzolo n. 9, 3.5131 Padova, Italy 4, Laboratori Nazionali di Legnaro dell’INFN, Legnaro (Padova), Italy. Received
10 December
1986 and in revised form 9 March
1987
Ion beam mixing effects at the metal/insulator interface have been studied by irradiating at room temperature the Fe film/SiO, substrate system with 100 keV Ar ions in the fluence range 3 x lo'*-1.3 X 10” ions/cm’. The aim of the work was to quantify the ion induced mixing of Fe atoms and clarify the mechanisms involved in the atomic transport processes. SEM analyses of the irradiated surfaces reveal ion beam induced surface morphological modifications as hole formation in the Fe film and Fe island growth. Computer controlled image analyses allowed us to evaluate the Fe coverage at any irradiation fluence, +, and to introduce a fluence
dependent “coverage” factor f(9) in calculations. The amount of mixed Fe atoms, Q,, has been measured by 1.8 MeV He+ Rutherford backscattering, after a chemical etching procedure to remove the unmixed Fe film. The mixing process is well described by the relation Q = Acp + Bfi, indicating that both collisional and diffusive mechanisms are effective. The relative influence of these two main mechanisms has been evaluated by determining the coefficients A and B, taking into account the morphological effects expressed by the “coverage” factor f(+).
1.Introduction Ion induced mixing effects on metal/insulator systems have been recently studied both in thin metal marker [l] and in bilayer metal film/SiO, configurations [2-61. The atomic transport process induced by the ion irradiation is ruled by two different physical mechanisms. The first is a collisional process based on the recoil implantation, which causes single long range displacements, with the amount of mixed atoms depending linearly on the ion fluence. The second is a diffusional process based on cascade mixing, which produces many short range displacements and gives rise to a square root dependence of the mixed amount on the ion dose [7-91. If both mechanisms are involved, the mixing efficiency can be related to the ion dose + by a phenomenological law such as: Q = A$ + B#” where Q is the amount of mixed atoms, A+ describes high energy primary recoils and Bd’* the diffusional mixing. The collisional, diffusional and chemical processes involved in the metal film/insulator ion beam mixing have been discussed in ref. [2] for transition metal (Ti, Cr or Ni)/SiO, couples bombarded with energetic (290 keV) Xe ions. The authors found that the amount of mixed metal atoms increases with the Xe dose + ap* Permanent address: Institute of High Energy Academia Sinica, Beijing, China.
Physics,
0168-583X/87/$03.50 0 Elsevier Science Publishers (North-Holland Physics Publishing Division)
B.V.
proximately as @, where /3 ranges from 0.56 to 0.73, for irradiations performed at room temperature and below. These data suggest that both collisional and diffusional transport processes are active in the ion mixing. the latter process being dominant for the couples with the lowest /3 value. Significant chemical effects on the mixing behaviour were observed at the irradiation temperature of 750 K, showing that only the cascade mixing process is affected by chemical driving forces. In order to better understand the ion beam mixing effects in metal/insulator systems, we report in this paper results concerning the Fe/%&, couple irradiated at room temperature with low current 100 keV Ar+ beam. The choice of Fe film can provide information complementary to those of ref. [2], being the Fe mass number intermediate to that of Ni and Cr. Moreover the use of the isotope 57Fe can provide structural information concerning the mixed region by using the Mossbauer spectroscopy, such measurements are in progress in our laboratory. The main analytical technique used in this work is the Rutherford backscattering spectrometry (RBS) of 1.8 MeV 4He ions, which can provide both the quantitative evaluation of the amount of mixed Fe from spectral areas and qualitative information about the mechanisms involved in the mixing process from the profile shape of mixed Fe. In fact it is known that recoil mixing far from the interface results in an exponential profile of mixed
G. Baitaglin et al. / Ion beam mixing
of the Fe/SO,
system
403
41 nm, the range straggling AR, is 16 nm. In order to overcome possible uncertainties in the analysis of the mixing behaviour, uncertainties which can be induced by sputtering and morphological modification of the surface and by the unmixed Fe layers on the top of the samples, one half of each implanted sample was chemically etched for 3 min in a 12M HCl solution at 85 ’ C for the removal of Fe which was not mixed, 3 min in a concentrated H,SO, solution at 85’ C for the removal of other surface contaminations, and finally washed in bidistilled water and in isopropyl alcohol. The amount of mixed Fe in the etched samples, as well as of residual Fe in the implanted non-etched samples, was determined by means of the 1.8 MeV 4He+ RBS analysis with 120” and 160” scattering angles. At an angle of 160° the depth resolution of our measurements was about 18 nm, much lower than the film thicknesses. Scanning electron microscopy (SEM) observations of both kinds of samples were performed in conjunction with the analysis of the emitted X-rays, by means of an
atoms N(x) = C+emax (x depth from the interface, Q ion dose, C and (Y constants) [lO,ll]. Care has been taken in this work to avoid or to recognize possible artifacts in the RBS analyses mainly due to the ion-induced morphological modifications of the irradiated surface.
2. Experimental 300 nm thick SiO, layers were grown by wet oxidation at 1000“ C of polished (100) Si wafers. Fe films with thickness of 26 nm, 32 nm and 51 nm were deposited by e-beam evaporation at a pressure of about 10K4 Pa over the SiO, layers, which were conventionally chemically cleaned before insertion in the deposition chamber. 100 keV 4”Ar’ implantations were performed at room temperature in the nominal fluence range 3 X 10”-1.3 X 10” ions cmm2, at 3 X lop4 Pa with a current density of 1 IJA cm-‘. The projected range R, of 100 keV Ar ions in Fe is
5
Ar i
I.U
ENERGY Fig. 1. RBS spectra
t MeVl
(4He+; 1.8 MeV; 0 = 160 o ) and SEM pictures for Fe(32 nm)/SiO, samples unirradiated (-, keV Art irradiated at 2.0 X lOI ions cm ~’ (+ + +, b) and 10.4~ lOI ions cme2 ( , c).
a) and 100
404
G. Battaglin et al. / Ion beam mixing of the Fe/SiO,
energy dispersion system (EDS). SEM micrographs were examined by using a TESAK equipped image analysis system.
3. Results The implantation effects in the Fe/SiO, system have been investigated in detail in samples with a 32 nm thick Fe film. The RBS spectra taken at a 160 o scattering angle of as evaporated, low dose (2 x lOi Ar+ cmm2) and high dose (10.4 x 1Ol6 Ar+ cmm2) implanted nonetched samples are shown in fig. 1. The low dose implanted sample shows very low Fe sputtering loss and mixing effects in comparison with
Fig. 2. SEM
system
the as deposited sample, and no noticeable morphological modification as shown in the SEM micrographs in the inset a, b. In the high dose implanted sample the seeming large mixing is mainly an artifact induced by the large morphological changes shown in the inset lc. In fact, the observed displacement of the Si and 0 yields toward higher energies is mainly due to the formation of holes in the Fe film, which leave uncovered the underlying SiO,. The EDS analysis have confirmed that the Fe K, line is not emitted uniformly from the sample surface, and that its intensity is the lowest when emitted from the hole positions. Moreover the RBS results reported in fig. 1 show that in the high dose implanted sample the amount of Fe is reduced to about the 60% of the as deposited, but
pictures showing the development of holes in the Ar + irradiated Fe(32 nm)/Si02 samples at 3.9 x lOI ions cm-’ (a), ions cm-* (c). The graph shows the fraction S of the sample surfaces which is not covered
6.5 X 1016 ions cm-* (b) and 13.0 X lOI by the Fe film after each Ar+ irradiation. lawS=&GtGGJ
For Ar ion fluences larger than a threshold dose & the trend follows the phenomenological line). The calculated values of m and (POare reported in the text.
( continuous
G. Battaglin et al. / Ion beam mixing of the Fe/SO,
the film thickness (evaluated from the fwhm of the RBS Fe signal) decreases less than 10%. The same behaviour has been observed also in samples with different Fe film thicknesses (26 and 51 nm). The Fe sputtering yield deduced from the RBS data is 0.7 (Fe atom)/(Ar ion), i.e. much lower than the theoretical value of 3.3 reported by Matsunami et al. [12]. Sputtering of SiO,, as deduced from RBS data, is very low also in high dose implanted samples. The SEM micrographs of the surface of samples implanted at Ar+ fluences, C#I,of 3.9, 6.5 and 13.0 x 1016 Ar+ cm-’ are reported in the insets a, b, c, of fig. 2, showing a decreasing coverage of the Fe film on the SiO, substrates as the ion fluence increases. The SEM micrographs have been analyzed by means of a computer controlled image analysis system, and the amount of the uncovered surface has been evaluated. Results are reported in fig. 2. A threshold fluence +a results for the hole formation in the Fe film. At higher fluences ($I > +,,) the uncovered surface, normalized to the total analyzed area, follows the trend (continuous line in the figure) described by the expression S( 0) = ,/m where m = (1.1 + 0.2) X lO_” cm* and $I,, = (3.7 + 0.6) x 1Or6 Ar+ cm-*. Moreover the image analysis has revealed that holes in the Fe film grow steadily in size at increasing fluences, but decrease in number for doses greater than 6.5 X 1Or6 Ar+ cm-*, when they impinge each other.
ENERGY
system
405
The morphological modifications develop simultaneously with a significant evolution of the Ar distribution inside the samples. The portion of the RBS spectra around the Ar peak positions is shown in fig. 3 for the same samples whose SEM micrographs have been reported in fig. 2. A very pronounced Ar peak at the Fe/SiO, interface appears when the early holes grow in the Fe film. At higher fluences the Ar peak is strongly lowered and more than 50% of the implanted Ar is lost, but no shift of the peak position appears. The formation of a sharp Ar peak indicates that Ar concentrates at the Fe/SiO, interface in a narrow region, with the likely formation of bubbles [13]. The displacement toward higher energies of the Si yields in fig. 3 is a consequence of the hole formation and of a slight thinning of the Fe film. The behaviour of the morphological modifications and Ar bubble formation has been investigated also in samples with a thicker Fe film, in order to clarify the physical mechanism ruling the hole formation in the Fe film. As a clarifying example, the SEM micrographs of a Fe(51 nm)/SiO, sample implanted with 1.3 x 10” Ar+ cme2 are shown before (fig. 4a) and after (fig. 4b) the etching procedure. While in the former no holes can be clearly detected in the Fe film, in the latter the SiO, substrate appears heavily blistered. The RBS analysis of the not etched sample shows a noticeable Ar peak at the Fe/SiO, interface, similar to those reported in fig. 3. Therefore holes grow in the thicker Fe film only after the formation of many Ar bubbles in SiO, at the metal interface.
1.1 [MeVl
Fig. 3. Partial RBS spectra (4He+; 1.8 MeV; 8 =160°) showing the Si edges and Ar signals for Fe(32 nm)/SiO, samples u&radiated () and 100 keV Ar+ irradiated at 3.9 X 10”’ ions cm -2 ( ), 6.5 X lOI ions cmm2 ( . . ) and 13.0 x lOI ions cmm2 (+ + +). Arrows indicate the surface position of Si and Ar and also the signals due to Ar present at the interface between the Fe films and the SiO, substrate.
Fig. 4. SEM pictures of the surface of the Fe(51 nm)/SiO, sample irradiated with 100 keV, 13.0~ lOI Arf cmm2 before (a) and after (b) the chemical etching which removes the unmixed Fe surfaces layer. The scale marker shows 1.6 pm.
406
G. Batraglin et al. / Ion beam mixing of the Fe/SiO,
system
Fig. 5. RBS spectra (4He+; 1.8 MeV; B = 120 ’ ) and SEM pictures for Fe(32 nm)/SiO, samples after the chemical etching which removes the unmixed Fe surface layer. Unirradiated (-, a) and 100 keV Ar+ irradiated at 2.0 X lOI ions cm-2 (+ + +, b) and 10.4 X lOI ions cmm2 ( . . . 2c). In the inset (d) the semilog plots of the Fe signal are reported, showing that the trailing edges can be well fitted by exponential trends.
The mixing effect, which can not be evaluated from the RBS spectra of fig. 1, has been measured on the etched samples. The etching time (3 min) in the HCl solution has been chosen in order to remove completely the Fe film from the as deposited samples, as verified by the RBS analysis. SiO, is at the surface also in the implanted etched samples. Increasing time up to 20 min, the SiO, thickness does not vary, within the sensitivity of the RBS technique; on the contrary, the Fe content in the implanted samples is about 10% lower in the long time etched than in the short time etched samples. The Fe removal is therefore carried on also in depth at the mixed interface in the implanted samples. The experimental uncertainty affecting the amount of mixed Fe has been evaluated as 10% in the high dose and 20% in the low dose implanted samples, i.e. much larger than the statistical errors in the measurement of the signal areas. In fig. 5 we report the RBS spectra, taken after the etching, of the same samples of fig. 1. The scattering
angle was 120 O. A slight channeling effect in the (100) Si support is discernible from the Si plateau. The different slopes of the Si and 0 edges of the spectrum of the higher dose implanted sample in comparison with spectra of the other specimens are induced by the surface roughness, as shown in the inset 5c. The Fe RBS yields are reported in the inset 5d on semilog scale; the trailing edge is well fitted by an exponential curve. Nevertheless, the actual Fe depth profile may differ from that deduced from the RBS spectra, being affected and altered by the sputtering effects and surface roughness, mainly at higher fluences. The measured amount of mixed Fe atoms as a function of the implanted Ar dose, $, is reported as Q, in fig. 6. The fitting curve will be discussed in the next section. Finally, few measurements have been performed on samples with different Fe film thickness (26 and 51 nm) and showed that the mixed Fe amount is higher the lower the thickness.
G. Baitaglin et al. / Ion beam mixing of the Fe/SiO,
407
system
processes (the nuclear stopping for Xe is about four times higher than that for Ar, in the examined systems). The hole growth could be attributed, as usual, to a surface migration of metal atoms, giving rise to the island formation often observed in metal/insulator couples upon irradiation [4]. However questions arise when trying to explain the experimental details in the present case. as: (1) metal islands actually grow in low adhesion couples, as Cu/SiO, or Au/SiO,, but do not in very adherent films; in our case an interface mixing occurs still at the lower doses before the hole formation, inducing a very effective bonding of the film with the substrate;
(2) the Fe film thickness if preserved
OJ 2
I
5
I
Id6
I
I
2 5 0 lions/cm2 I
I
lo"
I
I
2
Fig. 6. Measured Fe mixed amount Q, as a function of the Ar+ irradiation fluence +. The continuous line drawn through the experimental points has been obtained by a least-square fit using formula (3) reported in the text.
4. Discussion 4.1. Morphological
modifications
The observed behaviour of the morphological modifications is determined by two different processes: the hole formation and the hole growth. The former is promoted by the radiation induced modifications of the substrate. Ar bubbles are formed in the SiO, substrate (see fig. 4) below the Fe film in the interfacial region, as a consequence of the silica compaction and lateral stress relief upon irradiation [14]. In the regions of the Fe film lying just over the bubbles, stresses induced by the bubble formation increase as well as strains do, along directions parallel to the sample surface. While the 51 nm Fe film is thick enough to bear the induced stresses also at high fluences, in the 32 nm Fe film holes develop quickly also at low fluences. The formation of gas bubbles, and consequently of the holes in the metal film, depends more on the amount of implanted gas than on the radiation damage in the SiO, substrate. In fact, Nicolet et al. [3] have not observed surface modification in the Ni (25 nm)/SiO, couple implanted with 290 keV Xe+ 2 X 1Or6 cme2, although Xe is by far more effective than Ar in the mixing and substrate damage
also at the higher fluences when a large amount of Fe has been lost; (3) the sputtering yield is low, in particular at the lower fluences the Fe loss is nearly negligible, as a consequence of the protective effect of a surface carbon film which builds up on the sample surface during implantations, because of cracking of hydrocarbon molecules coming from the diffusion pumping system. In our opinion the two competitive mechanisms, surface migration-island growth and sputtering, could only fortituously preserve the Fe film thickness, and could do it hardly in samples with different film thicknesses, as previously reported. Therefore in the hole growth another physical mechanism is involved, which we tentatively indicate as an enhanced sputtering at the hole borders. The low angle incidence of the ion beam on the border surfaces can actually give a high sputtering yield, which counterbalances the protective effect of the surface carbon film, and supports the hole lateral growth. Since the holes are formed in the Fe film as a replica of the Ar bubbles in SiO,, the border sputtering supports also the hole lateral growth in SiO,, as confirmed by the SEM observations of the etched samples. 4.2. Mixing
effects
The mixing process can be phenomenologically scribed, as a function of the ion fluence 9, by
de-
Q=&+B&
(1)
where Q is the amount of mixed atoms, the A+ term describes primary recoils and the Bfi term the diffusion mechanisms. The effect of the decrease of the Fe film coverage with increasing irradiation fluence is taken into account by writing the differential expression for the mixed Fe amount: dQ,=j(+$$d+=j(O) where
Q,
is the actual
(2) mixed
Fe and the correction
G.Battaglin et al. / Ion beam mixing of the Fe/SO,
408 factor f( 9) is
f(G)=1 f(+) (see fig. 2) is a phenomenological expression for the fractional surface coverage of the Fe film; dQ/d$ = (A + B/2fi) is derived from eq. (1); cpo is the threshold fluence for the hole formation in the Fe film. By integration of eq. (2), the following expression for Q, has been obtained:
The logarithmic term plays a it gives a contribution to Q, therefore it is possible to neglect sion which is valid for any value
negligible role (in fact of the order of l%), it and write an expresof +:
Q,=++)+@+;),
(3’)
where S is the already area” :
s=o
defined
“normalized
uncovered
+ 5 Go
The sputtering of already mixed Fe from the areas becoming uncovered has been neglected in eq. (3), as: (1) no or a very low sputtering thinning is detected in the SiO, substrate; (2) the uncovered surface spans a limited extent; (3) sputtering data are lacking in the literature. Parameters A and B have been supposed not to vary with the ion fluence in eq. (3); in fact they are actually a function of the dose I$ only if the sample temperature strongly increases or the Fe film thickness strongly decreases during irradiation. For the former effect, the temperature increase does not exceed a few tens of degrees over room temperature, that is not enough to induce noticeable modifications in the mixing process; for the latter, the Fe thickness variations are very limited, as shown in section 3. The parameters A and B have been evaluated by means of a least-squares fit of the experimental points, reported in fig. 6, using formula (3). We obtained: B = (13.9 + 2.9) x lo6 cm-‘. A = (7.7 + 1.4) x 10-2, The interpolating curve, reported in fig. 6 as a continuous line, is in good agreement with experimental points. From eqs. (3) or (3’) we can evaluate that the effect of
system
the collisional mixing represents only f of the total amount of mixed Fe at the lower ion fluence considered, while it is 4 at the highest fluence, the remaining Fe mixing being induced by diffusional mechanisms. Moreover these results show that although the slope of the trailing edge of the Fe profile can be well fitted by an exponential curve (see inset d in fig. 5), the dominant effect of the recoil mixing cannot be asserted from this accordance; the overlap of the two mixing mechanisms and of the effect of surface features causes a Fe depth profile which cannot be unambiguously interpreted. The experimental data reported in fig. 6 could be reasonably fitted, as already done by other authors [3], also with a power law Q = const. X @, giving x = 0.70 + 0.04. However the physical meaning of such a result is not so straightforward as that previously reported, which allows to separate the contributions of collisional and diffusive mixing. Moreover, the mixing data presented in fig. 6 can be usefully compared with the results reported by Banwell et al. [2] for 290 keV Xe+ irradiation of the Cr/SiO, and Ni/SiO, systems. Among the data reported in [2], we selected those corresponding to the film thicknesses equal to R, - AR, (R, is the projected range of the incident ion, AR, the range straggling), as we have in the present work. By considering similar irradiation fluences (i.e. in the range 3 X 1015-2 X 1016 ions/cm2), the ratios of the measured mixed metal amounts, expressed as area1 densities, [Ni/Fe], [Cr/Fe], as well as [Ni/Cr], match the ratios of the nuclear stopping powers of Ar and Xe ions in the corresponding element 1151, averaged around the metal/SiO, interface, within about 20%. Such a satisfactory agreement among mixing efficiencies, scaled by the nuclear stopping powers relative to different ions, indicates that chemical effects in the mixing process are similar for all the reported systems and, as suggested in [2], likely to be almost negligible.
5. Conclusions From this study on the mixing effects induced by 100 keV Ar+ irradiation on the Fe/SiO, system we can conclude that: Ar+ irradiation induces both morphological modifications and atomic transport processes. Morphological modifications are activated by a buildup of Ar at the metal/SiO, interface, giving rise to bubble formation and consequent strain release. It is our opinion that the induced Fe island structure is not mainly due to surface Fe migration but to a highly efficient Fe sputtering at the island borders. SEM analyses are needed for a correct evaluation and interpretation of the mixing processes: by means of an image analysis system the Fe coverage has been
G. Battaglin et al. / Ion beam mixing of the Fe/SiO_,
evaluated, its dependence on the Ar+ fluence, 4, is f( C#J)= 1 - /m. The values of m and +a (threshold fluence for hole formation) have been determined with an accuracy better than 20%. To clarify the mixing mechanisms, the total amount of mixed Fe atoms, Q,, measured by RBS, has been related to the Ar+ fluence c$: the mixing is well described by the law Q = A$ + Bfi, indicating that both collisional and diffusive mechanisms are effective. Taking into account the “coverage” factor f( $), the coefficients A and B have been determined within 20%, allowing to determine the relative influence of the two main mechanisms. The authors are very grateful to Prof. P. Mazzoldi for stimulating discussions, Mr. E. BoLzan for technical assistance in ion implantation, Mr. P. Buso for chemical etching, Dr. P. Fabbri of Modena University for image analysis, Mr. A. Rampazzo for drawings, and Mrs. M.G. Dogliotti for typing the manuscript.
References [l] A.J. Barcz, B.M. Paine and M.-A. Nicolet, Lett. 44 (1984) 45.
Appl.
Phys.
system
409
PI T. Banwell, B.X. Liu, I. Golecki and M.-A. Nicolet, Nucl. Instr. and Meth. 209/210
(1983) 125. Mater.
[31 T.C. Banwell and M.-A. Nicolet,
Res. Sot. Symp. Proc. 27 (1984) 109. [41 G.J. Clark, J.E.E. Baglin, F.M. d’Heurle, C.W. White, G. Farlow and J. Narayan, Mater. Res. Sot. Symp. Proc. 27 (1984) 55. [51 M. Erola, J. Keinonen, H.J. Whitlow, A. Anttila and M. Hautala, Thin Solid Films 115 (1984) 125. L.A. Boatner, C.J. Mc161 G.C. Farlow, B.R. Appleton, Hargue and C.W. White, Mater. Res. Sot. Symp. Proc. 45 (1985) 137. 171 F. Besenbacher, J. Bettiger, SK. Nielsen and H.J. Whitlow, Appl. Phys. A29 (1982) 141. PI S. Matteson and M.-A. Nicolet, Mater. Res. Sot. Symp. Proc. 7 (1982) 3. Nucl, Instr. and Meth. [91 P. Sigmund and A. Gras-Marti, 182/183 (1981) 25. WI S. Dzioba and R. Kelly, J. Nucl. Mater. 76 (1978) 175. Pll L.A. Christel, J.F. Gibbons and S. Mylroie, Nucl. Instr. and Meth. 182/183 (1981) 187. WI N. Matsunami et al., At. Data and Nucl. Data Tables 31 (1984) 1. P31 Z.L. Liau, J.W. Mayer, W.L. Brown and J.M. Poate, J. Appl. Phys. 49 (1978) 5295. 1141 E.P. EerNisse, J. Appl. Phys. 45 (1974) 167. [151 J. Biersack and L.G. Haggmark, Nucl. Instr. and Meth. 174 (1980) 257.
Nuclear
Instruments
and Methods
in Physics Research
B27 (1987) 402-409
North-Holland,
ION BEAM MIXING EFFECTS G. BATTAGLIN
IN THE Ar +-IRRADIATED
I), S. LO RUSSO
‘) , A . PACCAGNELLA
Fe/SiO,
Amsterdam
SYSTEM
2), G. PRINCIPI
3, and P.Q. ZHANG
4)*
‘) Unit& CISM-GNSM
del CNR, Dipartimento di Fisica dell’liniversitri, Via Marzolo n. 8, 35131 Padova, Italy 2, Facoltd di Ingegneria, Universitd di Trento, Mesiano (Trento), Italy -‘, Istituto di Chimica Industriale dell’Universitci di Padova, Via Marzolo n. 9, 3.5131 Padova, Italy 4, Laboratori Nazionali di Legnaro dell’INFN, Legnaro (Padova), Italy. Received
10 December
1986 and in revised form 9 March
1987
Ion beam mixing effects at the metal/insulator interface have been studied by irradiating at room temperature the Fe film/SiO, substrate system with 100 keV Ar ions in the fluence range 3 x lo'*-1.3 X 10” ions/cm’. The aim of the work was to quantify the ion induced mixing of Fe atoms and clarify the mechanisms involved in the atomic transport processes. SEM analyses of the irradiated surfaces reveal ion beam induced surface morphological modifications as hole formation in the Fe film and Fe island growth. Computer controlled image analyses allowed us to evaluate the Fe coverage at any irradiation fluence, +, and to introduce a fluence
dependent “coverage” factor f(9) in calculations. The amount of mixed Fe atoms, Q,, has been measured by 1.8 MeV He+ Rutherford backscattering, after a chemical etching procedure to remove the unmixed Fe film. The mixing process is well described by the relation Q = Acp + Bfi, indicating that both collisional and diffusive mechanisms are effective. The relative influence of these two main mechanisms has been evaluated by determining the coefficients A and B, taking into account the morphological effects expressed by the “coverage” factor f(+).
1.Introduction Ion induced mixing effects on metal/insulator systems have been recently studied both in thin metal marker [l] and in bilayer metal film/SiO, configurations [2-61. The atomic transport process induced by the ion irradiation is ruled by two different physical mechanisms. The first is a collisional process based on the recoil implantation, which causes single long range displacements, with the amount of mixed atoms depending linearly on the ion fluence. The second is a diffusional process based on cascade mixing, which produces many short range displacements and gives rise to a square root dependence of the mixed amount on the ion dose [7-91. If both mechanisms are involved, the mixing efficiency can be related to the ion dose + by a phenomenological law such as: Q = A$ + B#” where Q is the amount of mixed atoms, A+ describes high energy primary recoils and Bd’* the diffusional mixing. The collisional, diffusional and chemical processes involved in the metal film/insulator ion beam mixing have been discussed in ref. [2] for transition metal (Ti, Cr or Ni)/SiO, couples bombarded with energetic (290 keV) Xe ions. The authors found that the amount of mixed metal atoms increases with the Xe dose + ap* Permanent address: Institute of High Energy Academia Sinica, Beijing, China.
Physics,
0168-583X/87/$03.50 0 Elsevier Science Publishers (North-Holland Physics Publishing Division)
B.V.
proximately as @, where /3 ranges from 0.56 to 0.73, for irradiations performed at room temperature and below. These data suggest that both collisional and diffusional transport processes are active in the ion mixing. the latter process being dominant for the couples with the lowest /3 value. Significant chemical effects on the mixing behaviour were observed at the irradiation temperature of 750 K, showing that only the cascade mixing process is affected by chemical driving forces. In order to better understand the ion beam mixing effects in metal/insulator systems, we report in this paper results concerning the Fe/%&, couple irradiated at room temperature with low current 100 keV Ar+ beam. The choice of Fe film can provide information complementary to those of ref. [2], being the Fe mass number intermediate to that of Ni and Cr. Moreover the use of the isotope 57Fe can provide structural information concerning the mixed region by using the Mossbauer spectroscopy, such measurements are in progress in our laboratory. The main analytical technique used in this work is the Rutherford backscattering spectrometry (RBS) of 1.8 MeV 4He ions, which can provide both the quantitative evaluation of the amount of mixed Fe from spectral areas and qualitative information about the mechanisms involved in the mixing process from the profile shape of mixed Fe. In fact it is known that recoil mixing far from the interface results in an exponential profile of mixed
G. Baitaglin et al. / Ion beam mixing
of the Fe/SO,
system
403
41 nm, the range straggling AR, is 16 nm. In order to overcome possible uncertainties in the analysis of the mixing behaviour, uncertainties which can be induced by sputtering and morphological modification of the surface and by the unmixed Fe layers on the top of the samples, one half of each implanted sample was chemically etched for 3 min in a 12M HCl solution at 85 ’ C for the removal of Fe which was not mixed, 3 min in a concentrated H,SO, solution at 85’ C for the removal of other surface contaminations, and finally washed in bidistilled water and in isopropyl alcohol. The amount of mixed Fe in the etched samples, as well as of residual Fe in the implanted non-etched samples, was determined by means of the 1.8 MeV 4He+ RBS analysis with 120” and 160” scattering angles. At an angle of 160° the depth resolution of our measurements was about 18 nm, much lower than the film thicknesses. Scanning electron microscopy (SEM) observations of both kinds of samples were performed in conjunction with the analysis of the emitted X-rays, by means of an
atoms N(x) = C+emax (x depth from the interface, Q ion dose, C and (Y constants) [lO,ll]. Care has been taken in this work to avoid or to recognize possible artifacts in the RBS analyses mainly due to the ion-induced morphological modifications of the irradiated surface.
2. Experimental 300 nm thick SiO, layers were grown by wet oxidation at 1000“ C of polished (100) Si wafers. Fe films with thickness of 26 nm, 32 nm and 51 nm were deposited by e-beam evaporation at a pressure of about 10K4 Pa over the SiO, layers, which were conventionally chemically cleaned before insertion in the deposition chamber. 100 keV 4”Ar’ implantations were performed at room temperature in the nominal fluence range 3 X 10”-1.3 X 10” ions cmm2, at 3 X lop4 Pa with a current density of 1 IJA cm-‘. The projected range R, of 100 keV Ar ions in Fe is
5
Ar i
I.U
ENERGY Fig. 1. RBS spectra
t MeVl
(4He+; 1.8 MeV; 0 = 160 o ) and SEM pictures for Fe(32 nm)/SiO, samples unirradiated (-, keV Art irradiated at 2.0 X lOI ions cm ~’ (+ + +, b) and 10.4~ lOI ions cme2 ( , c).
a) and 100
404
G. Battaglin et al. / Ion beam mixing of the Fe/SiO,
energy dispersion system (EDS). SEM micrographs were examined by using a TESAK equipped image analysis system.
3. Results The implantation effects in the Fe/SiO, system have been investigated in detail in samples with a 32 nm thick Fe film. The RBS spectra taken at a 160 o scattering angle of as evaporated, low dose (2 x lOi Ar+ cmm2) and high dose (10.4 x 1Ol6 Ar+ cmm2) implanted nonetched samples are shown in fig. 1. The low dose implanted sample shows very low Fe sputtering loss and mixing effects in comparison with
Fig. 2. SEM
system
the as deposited sample, and no noticeable morphological modification as shown in the SEM micrographs in the inset a, b. In the high dose implanted sample the seeming large mixing is mainly an artifact induced by the large morphological changes shown in the inset lc. In fact, the observed displacement of the Si and 0 yields toward higher energies is mainly due to the formation of holes in the Fe film, which leave uncovered the underlying SiO,. The EDS analysis have confirmed that the Fe K, line is not emitted uniformly from the sample surface, and that its intensity is the lowest when emitted from the hole positions. Moreover the RBS results reported in fig. 1 show that in the high dose implanted sample the amount of Fe is reduced to about the 60% of the as deposited, but
pictures showing the development of holes in the Ar + irradiated Fe(32 nm)/Si02 samples at 3.9 x lOI ions cm-’ (a), ions cm-* (c). The graph shows the fraction S of the sample surfaces which is not covered
6.5 X 1016 ions cm-* (b) and 13.0 X lOI by the Fe film after each Ar+ irradiation. lawS=&GtGGJ
For Ar ion fluences larger than a threshold dose & the trend follows the phenomenological line). The calculated values of m and (POare reported in the text.
( continuous
G. Battaglin et al. / Ion beam mixing of the Fe/SO,
the film thickness (evaluated from the fwhm of the RBS Fe signal) decreases less than 10%. The same behaviour has been observed also in samples with different Fe film thicknesses (26 and 51 nm). The Fe sputtering yield deduced from the RBS data is 0.7 (Fe atom)/(Ar ion), i.e. much lower than the theoretical value of 3.3 reported by Matsunami et al. [12]. Sputtering of SiO,, as deduced from RBS data, is very low also in high dose implanted samples. The SEM micrographs of the surface of samples implanted at Ar+ fluences, C#I,of 3.9, 6.5 and 13.0 x 1016 Ar+ cm-’ are reported in the insets a, b, c, of fig. 2, showing a decreasing coverage of the Fe film on the SiO, substrates as the ion fluence increases. The SEM micrographs have been analyzed by means of a computer controlled image analysis system, and the amount of the uncovered surface has been evaluated. Results are reported in fig. 2. A threshold fluence +a results for the hole formation in the Fe film. At higher fluences ($I > +,,) the uncovered surface, normalized to the total analyzed area, follows the trend (continuous line in the figure) described by the expression S( 0) = ,/m where m = (1.1 + 0.2) X lO_” cm* and $I,, = (3.7 + 0.6) x 1Or6 Ar+ cm-*. Moreover the image analysis has revealed that holes in the Fe film grow steadily in size at increasing fluences, but decrease in number for doses greater than 6.5 X 1Or6 Ar+ cm-*, when they impinge each other.
ENERGY
system
405
The morphological modifications develop simultaneously with a significant evolution of the Ar distribution inside the samples. The portion of the RBS spectra around the Ar peak positions is shown in fig. 3 for the same samples whose SEM micrographs have been reported in fig. 2. A very pronounced Ar peak at the Fe/SiO, interface appears when the early holes grow in the Fe film. At higher fluences the Ar peak is strongly lowered and more than 50% of the implanted Ar is lost, but no shift of the peak position appears. The formation of a sharp Ar peak indicates that Ar concentrates at the Fe/SiO, interface in a narrow region, with the likely formation of bubbles [13]. The displacement toward higher energies of the Si yields in fig. 3 is a consequence of the hole formation and of a slight thinning of the Fe film. The behaviour of the morphological modifications and Ar bubble formation has been investigated also in samples with a thicker Fe film, in order to clarify the physical mechanism ruling the hole formation in the Fe film. As a clarifying example, the SEM micrographs of a Fe(51 nm)/SiO, sample implanted with 1.3 x 10” Ar+ cme2 are shown before (fig. 4a) and after (fig. 4b) the etching procedure. While in the former no holes can be clearly detected in the Fe film, in the latter the SiO, substrate appears heavily blistered. The RBS analysis of the not etched sample shows a noticeable Ar peak at the Fe/SiO, interface, similar to those reported in fig. 3. Therefore holes grow in the thicker Fe film only after the formation of many Ar bubbles in SiO, at the metal interface.
1.1 [MeVl
Fig. 3. Partial RBS spectra (4He+; 1.8 MeV; 8 =160°) showing the Si edges and Ar signals for Fe(32 nm)/SiO, samples u&radiated () and 100 keV Ar+ irradiated at 3.9 X 10”’ ions cm -2 ( ), 6.5 X lOI ions cmm2 ( . . ) and 13.0 x lOI ions cmm2 (+ + +). Arrows indicate the surface position of Si and Ar and also the signals due to Ar present at the interface between the Fe films and the SiO, substrate.
Fig. 4. SEM pictures of the surface of the Fe(51 nm)/SiO, sample irradiated with 100 keV, 13.0~ lOI Arf cmm2 before (a) and after (b) the chemical etching which removes the unmixed Fe surfaces layer. The scale marker shows 1.6 pm.
406
G. Batraglin et al. / Ion beam mixing of the Fe/SiO,
system
Fig. 5. RBS spectra (4He+; 1.8 MeV; B = 120 ’ ) and SEM pictures for Fe(32 nm)/SiO, samples after the chemical etching which removes the unmixed Fe surface layer. Unirradiated (-, a) and 100 keV Ar+ irradiated at 2.0 X lOI ions cm-2 (+ + +, b) and 10.4 X lOI ions cmm2 ( . . . 2c). In the inset (d) the semilog plots of the Fe signal are reported, showing that the trailing edges can be well fitted by exponential trends.
The mixing effect, which can not be evaluated from the RBS spectra of fig. 1, has been measured on the etched samples. The etching time (3 min) in the HCl solution has been chosen in order to remove completely the Fe film from the as deposited samples, as verified by the RBS analysis. SiO, is at the surface also in the implanted etched samples. Increasing time up to 20 min, the SiO, thickness does not vary, within the sensitivity of the RBS technique; on the contrary, the Fe content in the implanted samples is about 10% lower in the long time etched than in the short time etched samples. The Fe removal is therefore carried on also in depth at the mixed interface in the implanted samples. The experimental uncertainty affecting the amount of mixed Fe has been evaluated as 10% in the high dose and 20% in the low dose implanted samples, i.e. much larger than the statistical errors in the measurement of the signal areas. In fig. 5 we report the RBS spectra, taken after the etching, of the same samples of fig. 1. The scattering
angle was 120 O. A slight channeling effect in the (100) Si support is discernible from the Si plateau. The different slopes of the Si and 0 edges of the spectrum of the higher dose implanted sample in comparison with spectra of the other specimens are induced by the surface roughness, as shown in the inset 5c. The Fe RBS yields are reported in the inset 5d on semilog scale; the trailing edge is well fitted by an exponential curve. Nevertheless, the actual Fe depth profile may differ from that deduced from the RBS spectra, being affected and altered by the sputtering effects and surface roughness, mainly at higher fluences. The measured amount of mixed Fe atoms as a function of the implanted Ar dose, $, is reported as Q, in fig. 6. The fitting curve will be discussed in the next section. Finally, few measurements have been performed on samples with different Fe film thickness (26 and 51 nm) and showed that the mixed Fe amount is higher the lower the thickness.
G. Baitaglin et al. / Ion beam mixing of the Fe/SiO,
407
system
processes (the nuclear stopping for Xe is about four times higher than that for Ar, in the examined systems). The hole growth could be attributed, as usual, to a surface migration of metal atoms, giving rise to the island formation often observed in metal/insulator couples upon irradiation [4]. However questions arise when trying to explain the experimental details in the present case. as: (1) metal islands actually grow in low adhesion couples, as Cu/SiO, or Au/SiO,, but do not in very adherent films; in our case an interface mixing occurs still at the lower doses before the hole formation, inducing a very effective bonding of the film with the substrate;
(2) the Fe film thickness if preserved
OJ 2
I
5
I
Id6
I
I
2 5 0 lions/cm2 I
I
lo"
I
I
2
Fig. 6. Measured Fe mixed amount Q, as a function of the Ar+ irradiation fluence +. The continuous line drawn through the experimental points has been obtained by a least-square fit using formula (3) reported in the text.
4. Discussion 4.1. Morphological
modifications
The observed behaviour of the morphological modifications is determined by two different processes: the hole formation and the hole growth. The former is promoted by the radiation induced modifications of the substrate. Ar bubbles are formed in the SiO, substrate (see fig. 4) below the Fe film in the interfacial region, as a consequence of the silica compaction and lateral stress relief upon irradiation [14]. In the regions of the Fe film lying just over the bubbles, stresses induced by the bubble formation increase as well as strains do, along directions parallel to the sample surface. While the 51 nm Fe film is thick enough to bear the induced stresses also at high fluences, in the 32 nm Fe film holes develop quickly also at low fluences. The formation of gas bubbles, and consequently of the holes in the metal film, depends more on the amount of implanted gas than on the radiation damage in the SiO, substrate. In fact, Nicolet et al. [3] have not observed surface modification in the Ni (25 nm)/SiO, couple implanted with 290 keV Xe+ 2 X 1Or6 cme2, although Xe is by far more effective than Ar in the mixing and substrate damage
also at the higher fluences when a large amount of Fe has been lost; (3) the sputtering yield is low, in particular at the lower fluences the Fe loss is nearly negligible, as a consequence of the protective effect of a surface carbon film which builds up on the sample surface during implantations, because of cracking of hydrocarbon molecules coming from the diffusion pumping system. In our opinion the two competitive mechanisms, surface migration-island growth and sputtering, could only fortituously preserve the Fe film thickness, and could do it hardly in samples with different film thicknesses, as previously reported. Therefore in the hole growth another physical mechanism is involved, which we tentatively indicate as an enhanced sputtering at the hole borders. The low angle incidence of the ion beam on the border surfaces can actually give a high sputtering yield, which counterbalances the protective effect of the surface carbon film, and supports the hole lateral growth. Since the holes are formed in the Fe film as a replica of the Ar bubbles in SiO,, the border sputtering supports also the hole lateral growth in SiO,, as confirmed by the SEM observations of the etched samples. 4.2. Mixing
effects
The mixing process can be phenomenologically scribed, as a function of the ion fluence 9, by
de-
Q=&+B&
(1)
where Q is the amount of mixed atoms, the A+ term describes primary recoils and the Bfi term the diffusion mechanisms. The effect of the decrease of the Fe film coverage with increasing irradiation fluence is taken into account by writing the differential expression for the mixed Fe amount: dQ,=j(+$$d+=j(O) where
Q,
is the actual
(2) mixed
Fe and the correction
G.Battaglin et al. / Ion beam mixing of the Fe/SO,
408 factor f( 9) is
f(G)=1 f(+) (see fig. 2) is a phenomenological expression for the fractional surface coverage of the Fe film; dQ/d$ = (A + B/2fi) is derived from eq. (1); cpo is the threshold fluence for the hole formation in the Fe film. By integration of eq. (2), the following expression for Q, has been obtained:
The logarithmic term plays a it gives a contribution to Q, therefore it is possible to neglect sion which is valid for any value
negligible role (in fact of the order of l%), it and write an expresof +:
Q,=++)+@+;),
(3’)
where S is the already area” :
s=o
defined
“normalized
uncovered
+ 5 Go
The sputtering of already mixed Fe from the areas becoming uncovered has been neglected in eq. (3), as: (1) no or a very low sputtering thinning is detected in the SiO, substrate; (2) the uncovered surface spans a limited extent; (3) sputtering data are lacking in the literature. Parameters A and B have been supposed not to vary with the ion fluence in eq. (3); in fact they are actually a function of the dose I$ only if the sample temperature strongly increases or the Fe film thickness strongly decreases during irradiation. For the former effect, the temperature increase does not exceed a few tens of degrees over room temperature, that is not enough to induce noticeable modifications in the mixing process; for the latter, the Fe thickness variations are very limited, as shown in section 3. The parameters A and B have been evaluated by means of a least-squares fit of the experimental points, reported in fig. 6, using formula (3). We obtained: B = (13.9 + 2.9) x lo6 cm-‘. A = (7.7 + 1.4) x 10-2, The interpolating curve, reported in fig. 6 as a continuous line, is in good agreement with experimental points. From eqs. (3) or (3’) we can evaluate that the effect of
system
the collisional mixing represents only f of the total amount of mixed Fe at the lower ion fluence considered, while it is 4 at the highest fluence, the remaining Fe mixing being induced by diffusional mechanisms. Moreover these results show that although the slope of the trailing edge of the Fe profile can be well fitted by an exponential curve (see inset d in fig. 5), the dominant effect of the recoil mixing cannot be asserted from this accordance; the overlap of the two mixing mechanisms and of the effect of surface features causes a Fe depth profile which cannot be unambiguously interpreted. The experimental data reported in fig. 6 could be reasonably fitted, as already done by other authors [3], also with a power law Q = const. X @, giving x = 0.70 + 0.04. However the physical meaning of such a result is not so straightforward as that previously reported, which allows to separate the contributions of collisional and diffusive mixing. Moreover, the mixing data presented in fig. 6 can be usefully compared with the results reported by Banwell et al. [2] for 290 keV Xe+ irradiation of the Cr/SiO, and Ni/SiO, systems. Among the data reported in [2], we selected those corresponding to the film thicknesses equal to R, - AR, (R, is the projected range of the incident ion, AR, the range straggling), as we have in the present work. By considering similar irradiation fluences (i.e. in the range 3 X 1015-2 X 1016 ions/cm2), the ratios of the measured mixed metal amounts, expressed as area1 densities, [Ni/Fe], [Cr/Fe], as well as [Ni/Cr], match the ratios of the nuclear stopping powers of Ar and Xe ions in the corresponding element 1151, averaged around the metal/SiO, interface, within about 20%. Such a satisfactory agreement among mixing efficiencies, scaled by the nuclear stopping powers relative to different ions, indicates that chemical effects in the mixing process are similar for all the reported systems and, as suggested in [2], likely to be almost negligible.
5. Conclusions From this study on the mixing effects induced by 100 keV Ar+ irradiation on the Fe/SiO, system we can conclude that: Ar+ irradiation induces both morphological modifications and atomic transport processes. Morphological modifications are activated by a buildup of Ar at the metal/SiO, interface, giving rise to bubble formation and consequent strain release. It is our opinion that the induced Fe island structure is not mainly due to surface Fe migration but to a highly efficient Fe sputtering at the island borders. SEM analyses are needed for a correct evaluation and interpretation of the mixing processes: by means of an image analysis system the Fe coverage has been
G. Battaglin et al. / Ion beam mixing of the Fe/SiO_,
evaluated, its dependence on the Ar+ fluence, 4, is f( C#J)= 1 - /m. The values of m and +a (threshold fluence for hole formation) have been determined with an accuracy better than 20%. To clarify the mixing mechanisms, the total amount of mixed Fe atoms, Q,, measured by RBS, has been related to the Ar+ fluence c$: the mixing is well described by the law Q = A$ + Bfi, indicating that both collisional and diffusive mechanisms are effective. Taking into account the “coverage” factor f( $), the coefficients A and B have been determined within 20%, allowing to determine the relative influence of the two main mechanisms. The authors are very grateful to Prof. P. Mazzoldi for stimulating discussions, Mr. E. BoLzan for technical assistance in ion implantation, Mr. P. Buso for chemical etching, Dr. P. Fabbri of Modena University for image analysis, Mr. A. Rampazzo for drawings, and Mrs. M.G. Dogliotti for typing the manuscript.
References [l] A.J. Barcz, B.M. Paine and M.-A. Nicolet, Lett. 44 (1984) 45.
Appl.
Phys.
system
409
PI T. Banwell, B.X. Liu, I. Golecki and M.-A. Nicolet, Nucl. Instr. and Meth. 209/210
(1983) 125. Mater.
[31 T.C. Banwell and M.-A. Nicolet,
Res. Sot. Symp. Proc. 27 (1984) 109. [41 G.J. Clark, J.E.E. Baglin, F.M. d’Heurle, C.W. White, G. Farlow and J. Narayan, Mater. Res. Sot. Symp. Proc. 27 (1984) 55. [51 M. Erola, J. Keinonen, H.J. Whitlow, A. Anttila and M. Hautala, Thin Solid Films 115 (1984) 125. L.A. Boatner, C.J. Mc161 G.C. Farlow, B.R. Appleton, Hargue and C.W. White, Mater. Res. Sot. Symp. Proc. 45 (1985) 137. 171 F. Besenbacher, J. Bettiger, SK. Nielsen and H.J. Whitlow, Appl. Phys. A29 (1982) 141. PI S. Matteson and M.-A. Nicolet, Mater. Res. Sot. Symp. Proc. 7 (1982) 3. Nucl, Instr. and Meth. [91 P. Sigmund and A. Gras-Marti, 182/183 (1981) 25. WI S. Dzioba and R. Kelly, J. Nucl. Mater. 76 (1978) 175. Pll L.A. Christel, J.F. Gibbons and S. Mylroie, Nucl. Instr. and Meth. 182/183 (1981) 187. WI N. Matsunami et al., At. Data and Nucl. Data Tables 31 (1984) 1. P31 Z.L. Liau, J.W. Mayer, W.L. Brown and J.M. Poate, J. Appl. Phys. 49 (1978) 5295. 1141 E.P. EerNisse, J. Appl. Phys. 45 (1974) 167. [151 J. Biersack and L.G. Haggmark, Nucl. Instr. and Meth. 174 (1980) 257.