Structural analysis of buried conducting CoSi2 layers formed in Si by high-dose Co ion implantation

Structural analysis of buried conducting CoSi2 layers formed in Si by high-dose Co ion implantation

Journal of Crystal Growth 187 (1998) 435—443 Structural analysis of buried conducting CoSi layers formed 2 in Si by high-dose Co ion implantation A.A...

958KB Sizes 0 Downloads 13 Views

Journal of Crystal Growth 187 (1998) 435—443

Structural analysis of buried conducting CoSi layers formed 2 in Si by high-dose Co ion implantation A.A. Galayev!, Yu.N. Parkhomenko",*, K.D. Chtcherbatchev", D.A. Podgorny", A.I. Belogorohov!, A. Die´guez#, A. Romano-Rodriguez#, A. Pe´rez-Rodrı´ guez#, J.R. Morante# ! Institute of Chemical Problems in Microelectronics, 86 Prospekt Vernadskogo, 117571 Moscow, Russian Federation " Department of Semiconductor Materials and Devices, Moscow Steel and Alloys Institute, 4 Leninsky Prospekt, 117936 Moscow, Russian Federation # EME, Dept. Fı& sica Aplicada i Electro% nica, Universitat de Barcelona, Avda. Diagonal 645-647, E-08028 Barcelona, Spain Received 7 October 1997

Abstract Buried cobalt disilicide layers have been synthesized in Si by high-dose Co` implantation and annealing. The implanted doses have been 1017 and 3]1017 ions/cm2, and after implantation the samples have been annealed in a one (600°C, 1 h) or two step (600°C, 1 h#1000°C, 30 min) process. The detailed characterization of the samples has been performed by SIMS, XPS, TEM, XRD and Raman scattering measurements. The obtained data show the formation of CoSi grains coherent with the Si lattice at the lowest dose, as well as the formation of a continuous CoSi epitaxial layer 2 2 at the highest one. After the two-step anneal, high crystalline quality heteroepitaxial Si/CoSi /Si structures (10 nm 2 Si/90 nm CoSi ) is obtained for the higher implantation dose while for the lower one a non-continuous buried layer with 2 octahedral CoSi precipitates is formed. ( 1998 Elsevier Science B.V. All rights reserved. 2

1. Introduction Transition metal silicides are extensively used in microelectronics due to their good adhesion to silicon, as well as to the similar values of the lattice parameter and the thermal expansion coefficient of both Si and conducting silicides. Moreover, silicide

* Corresponding author.

layers are characterized by a high electrical conductivity and thermostability. All this makes these materials suited for the fabrication of conducting layers and interconnections in VLSI technologies. Special attention is also paid to the applications of the silicides in RF devices, e.g., MOS transistor gates. There are different techniques for the synthesis of silicide layers, most of them being based on direct deposition of a film onto a substrate and subsequent thermal processing. In general, these

0022-0248/98/$19.00 ( 1998 Elsevier Science B.V. All rights reserved. PII S 0 0 2 2 - 0 2 4 8 ( 9 7 ) 0 0 6 0 0 - 3

436

A.A. Galayev et al. / Journal of Crystal Growth 187 (1998) 435—443

techniques impose serious requirements related to the vacuum conditions during deposition and the state of the surface before deposition. Furthermore, some problems arise concerning the oriented growth of single-crystal cobalt silicide layers on silicon [1,2]. In front of these, ion implantation constitutes a promising technique for the synthesis of silicide layers with optimum and stable electrical parameters [1—7]. Moreover, ion implantation allows the synthesis of buried silicide layers, which may lead to the fabrication of high-speed buried metal layer transistors, as well as the development of multilevel integrated circuits (ICs). In this work, the structural analysis of Si samples implanted with high doses of Co` ions is reported as a function of the processing parameters. The analysis corroborates the suitability of ion beam synthesis for the fabrication of thin film heteroepitaxial structures with buried CoSi layers, ob2 serving a strong dependence of the thickness of both the top Si and buried CoSi layers on the 2 implanted dose.

2. Experimental details Si(1 0 0) wafers (resistivity 4.5 ) cm) were implanted in a Balzers SCI-218 accelerator with 59Co` ions. The implantation energy was 180 keV, the current density was 13 lA/cm2, and the implanted doses were 1017 and 3]1017 ions/cm2. During implantation, channeling was avoided by tilting the beam through 7 arcdeg relative to the normal incidence direction. In order to anneal radiation defects and to initiate cobalt disilicide precipitation, the wafers were heated to 450°C during the implantation. After implantation, pieces from the wafers were annealed in a conventional furnace in a nitrogen ambient. Annealing was performed in two steps: the first one was a 600°C anneal (1 h) and the second, a 1000°C anneal (30 min). Annealing at 600°C is carried out in order to further promote the nucleation of CoSi precipitates in the implanted region. 2 As will be shown later, the second step leads to the coalescence of the CoSi precipitates and, in certain 2 cases, to a continuous buried layer, as well as damage recovery of the top and bottom Si layers.

The detailed characterization of the samples has been performed by secondary ion mass spectroscopy (SIMS), X-ray photoelectron spectroscopy (XPS), high-resolution X-ray diffraction (HRXRD), transmission electron microscopy (TEM) and Raman scattering measurements. SIMS measurements were performed on a Perkin-Elmer PHI-6600 spectrometer. The primary O` ion beam was incident onto the specimens at 60 arcdeg, the ion current was 300 nA, and the energy was 4 keV. The ion beam was rastered to a 400]400 lm2, and the data were collected from the etch center (6% of the total etch area) to exclude interference from the etch walls. XPS analysis was carried out on a Perkin-Elmer PHI-5500 spectrometer using Al K -radiation a (1486.6 eV) to study the valence band. The anode power was 300 W (1 kV, 12 mA). The Si spectra 24 were obtained with a pass energy of 93.90 eV with a step of 0.8 eV, and a signal-to-noise ratio of 200. For valence band measurements, a pass energy of 11.75 eV with a step of 0.1 eV and a signal-to-noise ratio of 250 were used. On the other hand, XRD rocking curves were measured with a Philips MRD diffractometer with Cu K -radiation (0.15406 nm) a in symmetrical (the (0 0 4)-reflection) and asymmetrical (the (21 21 4)- and (2 2 4)-reflections) geometries. The X-ray beam was collimated and monochromated using a Barrel’s Ge(2 2 0) 4-crystal monochromator. Cross-section TEM samples from the as-implanted and annealed pieces were prepared by mechanical grinding and ion milling. Observation of these samples was carried out in a Philips CM30 Supertwin microscope operated at 300 kV. Finally, Raman scattering measurements were performed using a Jobin—Yvon T64000 spectrometer, coupled with an Olympus metallographic microscope. The spectra were measured in backscattering geometry, and the excitation was provided by the 514 nm line from an Ar` laser.

3. Results and discussion 3.1. Samples as-implanted Fig. 1 shows the SIMS spectra measured in the samples implanted at doses of 1017 and

A.A. Galayev et al. / Journal of Crystal Growth 187 (1998) 435—443

Fig. 1. SIMS profiles measured on the as-implanted samples ((1) dose 1017 cm~2, (2) dose 3]1017 cm~2).

3]1017 cm~2. Significant differences are observed between these profiles: for the lower implanted dose, a Gaussian-shaped distribution is obtained. This is centered at a depth of about 120 nm, and has a peak Co atomic concentration of 8%. This is below the Co concentration for stoichiometric CoSi (33%). For the highest dose, the Co profile 2 becomes broader and shifts towards the surface. The shape of the profile is almost rectangular, with a relatively flat concentration and a sharp interface with the Si substrate. Moreover, the maximum value of Co concentration corresponds to the stoichiometric concentration in CoSi . 2 These data indicate a saturation of the Co concentration to the value corresponding to stoichiometric cobalt disilicide. The formation of this phase is corroborated by XPS measurements, which show the Si peak centered at the value corresponding to 24 Si bonding in CoSi (150 eV) [8]. The saturation of 2 the cobalt distribution takes place when the implanted dose is higher than a critical value which corresponds to the threshold dose [9]. This is defined as the dose for which the Co content in the implanted peak reaches the stoichiometric value and, therefore, a continuous buried CoSi layer is 2 formed. According to our data, this threshold dose is of about 2]1017 cm~2, for the conditions of implantation used in this work. Increasing the dose above this value leads to a redistribution of the excess Co ions towards the wings of the profile, and

437

the width of the silicide layer increases. The shift of the profile towards the surface is due to sputtering during implantation. Cross-section TEM images of the two samples are presented in Fig. 2. Fig. 2a is a low-magnification image of the lower implantation dose that shows the presence of a 90 nm thick Si layer, in which some threading dislocations can be observed, on top of a 230 nm thick buried layer, which contains CoSi . On the other hand, the effect of 2 implanting at a higher dose, presented in Fig. 2b, gives rise to an implanted layer that extends from the surface to a depth of about 400 nm. In both samples the implanted layer shows the presence of a high density of CoSi precipitates, which 2 are both A- and B-type, i.e., aligned and twinned compared to the silicon substrate, as can be seen in Fig. 2c for the lower implantation dose. These precipitates are limited by M1 1 1N planes and their size increases when approaching the implantation peak. Beneath these layers the presence of rod-like defects can also be detected, similar to the case of high-dose high-temperature ion beam synthesis of SiO and SiC layers in Si substrates [10]. 2 It is important to remark that none of both layers is continuous, in spite of the high implantation dose. Fig. 3a and Fig. 3b present the X-ray rocking curves obtained for the lower and for the higher implantation doses, respectively. In them the asimplanted as well as the two annealed samples are plotted. Regarding only the as-implanted samples, for the lower dose the presence of a second crystalline phase cannot be detected because of the lack of any peak corresponding to CoSi , although ob2 served by TEM. However, for the higher dose it is clear that CoSi has been directly formed in agree2 ment with TEM results.

3.2. Samples annealed Fig. 4 shows the evolution of the Co SIMS profiles with the annealing. For the lower dose (Fig. 4a) the implanted cobalt atoms redistribute from the profile wings to its maximum, i.e., are gettered at the maximum, already during the first-step annealing. Accordingly, the Co profile

438

A.A. Galayev et al. / Journal of Crystal Growth 187 (1998) 435—443

Fig. 2. Cross-section TEM images of the samples as-implanted at (a) 1017 cm~2 and (b) 3]1017 cm~2. (c) High-resolution TEM image of sample from (b) showing the presence of both A- and B-type precipitates.

becomes sharper and the concentration at the peak increases to values of 9]1021 cm~3 after annealing at 600°C and to 1.6]1022 cm~3 after annealing at 1000°C for the dose of 1017 cm~2, values which are

still below the stoichiometric Co concentration in CoSi (3.5]1022 cm~3). 2 For the samples implanted at the higher dose (Fig. 4b), the Co profile becomes more rectangular-

A.A. Galayev et al. / Journal of Crystal Growth 187 (1998) 435—443

439

Fig. 3. X-ray rocking curves measured under symmetrical conditions, (4 0 0) reflection, in the samples as-implanted and annealed in one and two steps ((1) as-implanted; (2) annealed at 600°C, 1 h; (3) annealed at 600°C, 1 h#1000°C, 30 min): (a) implanted at 1017 cm~2; and (b) implanted at 3]1017 cm~2.

like after annealing, and the transition to the substrate Si becomes sharper. The maximum concentration remains at the value corresponding to stoichiometric CoSi . 2 The rocking curves from the sample implanted at the lower dose shows a strong increase in the intensity of the (4 0 0) CoSi peak only after the two2 step annealing, as can be seen in Fig. 3a. This is related to the coalescence of the CoSi grains. 2 No more peaks corresponding to Co—Si compounds were observed in the rocking curves, indicating the absence of any other silicide phase. The absence of a diffraction peak in this position after the first anneal step points out the need for the high-temperature anneal in order to have a strong improvement of the crystalline quality of the CoSi . 2 On the other hand, the sample implanted at the higher dose presents an improvement of the crystalline quality of the CoSi already after the first 2

annealing step, evidenced through the increase of the peak of the silicide. The higher temperature annealing does not increase apparently the quality of the CoSi , but gives rise to a shift in the position 2 of the diffraction peak, which indicates the presence of strain in the layer. Furthermore, a reduction of the defect density after this second annealing can be deduced from this figure through the decrease of the diffraction intensity between the peaks corresponding to Si and CoSi . 2 X-ray rocking curves under symmetrical and asymmetrical reflections (Fig. 5) have allowed the determination of the angular distances between the Si and CoSi peaks and have been used to deter2 mine the CoSi lattice parameters in the directions 2 perpendicular (a ) and parallel (a ) to the surface M , and, thus, the strain. The results were recalculated to lattice periods of an unstrained crystal a%91 , and C0S*2 were used to calculate the relaxation coefficient R"a%91 !a /a"6-, !a . Moreover, from the C0S*2 S* C0S*2 S*

440

A.A. Galayev et al. / Journal of Crystal Growth 187 (1998) 435—443

Fig. 5. X-ray rocking curves measured under symmetrical (4 0 0) and asymmetrical ((21 21 4) and (2 2 4)) reflections for the sample implanted at 3]1017 cm~2 and annealed in two steps.

Fig. 4. SIMS profiles of the samples as-implanted and annealed for the doses of (a) 1017 cm~2 and (b) 3]1017 cm~2. ((1) asimplanted; (2) annealed at 600°C, 1 h; (3) annealed at 600°C, 1 h#1000°C, 30 min.)

peak shifts relative to the calculated positions, the perpendicular (e ) and parallel (e ) components of M , CoSi lattice strain in the layer were determined. 2 The results are summarized in Table 1. These data indicate that the layers are constrained in the perpendicular direction and extended in the parallel direction, which suggests the existence of a compressive biaxial stress, similar to that observed in heteroepitaxial multilayer structures [3]. After annealing at 600°C, the absolute strain changed only slightly, and after two-stage annealing strain increased by about 1.8 times and the relaxation R decreased by 50%. The increase in strain observed after the hightemperature anneal step suggests residual strain in

the silicide layer to be determined mainly by the differences in the thermal expansion coefficient between Si and CoSi (which have values of 2 9.4]10~6 and 2.3]10~6 K~1, respectively [5]) and in their lattice parameters (which differ by 1.2%). Moreover, we have to remark that after annealing the CoSi /Si interfaces become more 2 abrupt. Fig. 6 shows TEM images corresponding to the samples annealed after the two steps. The sample implanted at the lower dose is shown in Fig. 6a and the main features are the presence of a buried layer consisting of CoSi octahedra, with an average size 2 of about 140 nm, forming a non-continuous buried layer, centered at about 120 nm. Furthermore, in the top silicon layer, of about 40 nm thickness, threading dislocations can be observed, which cross the buried layer and even penetrate into the substrate to a depth of about 350 nm. All these results are in good agreement with the results measured by SIMS. For the higher implantation dose the twostep annealing leads to the formation of a continuous layer of about 90 nm thickness, as shown in Fig. 6b. The presence of steps at the lower interface of the buried layer can be clearly seen in this image. It is important to point out that the sample’s surface has been etched before the TEM preparation and, for this reason, the surface crystalline Si layer cannot be seen in this image. Fig. 6c is a

A.A. Galayev et al. / Journal of Crystal Growth 187 (1998) 435—443

441

Table 1 a is the experimental CoSi lattice parameter in the direction perpendicular to the surface, a is the experimental lattice parameter in M 2 , the direction parallel to the surface, and a"6-, is the lattice parameter of bulk (unstrained) CoSi 2 Nos

Specimen

a (nm) M

a (nm) ,

a"6-, (nm)

R (%)

e (%) M

e (%) ,

1 2 3 4

1017cm~2#600°C 1 h#1000°C 30 min 3]1017 cm~2, as-implanted 3]1017 cm~2#600°C 1 h 3]1017 cm~2#600°C 1 h#1000°C 30 min

0.5348 0.5335 0.5340 0.5318

0.5376 0.5381 0.5381 0.5395

0.5374 0.5364 0.5364 0.5364

82 74.6 74.6 53.7

!0.29 !0.54 !0.45 !0.86

0.22 0.32 0.32 0.58

Fig. 6. Cross-section TEM images of the samples annealed in two steps: (a) implanted at 1017 cm~2 and (b) implanted at 3]1017 cm~2. (c) High-resolution TEM image of sample from (b) showing the continuous CoSi layer, in which some steps are clearly visible. 2

442

A.A. Galayev et al. / Journal of Crystal Growth 187 (1998) 435—443

buried CoSi grains are not connected and, thus, 2 a continuous layer is not formed.

4. Conclusions

Fig. 7. XPS spectrum of the sample implanted at 1]1017 cm~2 and annealed in two steps.

high-resolution TEM image of this lower interface showing the perfect epitaxial relationship between both substrate and buried layer, as well as the presence of several steps. Fig. 7 shows the Co XPS spectrum measured 21 from the sample implanted at the higher dose after the two-steps annealing. This spectrum is characterized by the presence of a peak at 80 keV. This peak corresponds to plasmon losses, i.e., the energy spent on collective electron oscillations in atoms bound into an ordered structure. This corresponds to a plasmon line with hw"22 eV which is attributed to CoSi , being the plasmon of silicon 2 hw"17 eV [11]. The intensity and shape of this peak points out the high crystalline quality of the CoSi buried layer. This has been corroborated by 2 Raman scattering measurements, which show the presence of a Lorentzian peak at 285 cm~1, which agrees with the allowing vibration mode from bulk single-crystal CoSi [8]. 2 Finally, four-point probe measurements have also been carried out in order to determine the resistivity of the two-step annealed samples. The specimens implanted with a dose of 1017 cm~2 had o"28 m)/cm after two-stage annealing, while for 3]1017 cm~2, o"20 m)/cm [5—7]. Thus, the sample implanted at the highest dose have lower resistivity after two-step anneal than those implanted at the lower dose. This is because that the

High-dose cobalt implantation into heated single-crystal silicon wafers leads to the direct formation of CoSi precipitates. Annealing at 2 600°C causes the precipitates to grow and coalesce in the middle of the implanted layer and a further grow occurs when annealing at a higher temperature (1000°C), which form a continuous CoSi 2 layer with sharp boundaries only for an implantation dose of 3]1017 cm~2. The damaged superficial Si layer recrystallizes, and the radiation defects located under the buried CoSi layer are annealed. 2 Si/CoSi /Si heterostructures with thin buried 2 layers have been prepared in this way and their electrical characteristics show the formation of a highly conductive buried metallic layer which can be used for the production of rapid “hot”-electron transistors, and thick-layered structures are suitable for contact layers and multilevel IC interconnections.

Acknowledgements This work has been supported by the Russian GNTP “New Materials” (project 06.03) and project reference INTAS-95-1327 from the INTAS program from the European Union.

References [1] A.E. White, K.T. Short, R.C. Dynes, J.M. Gibson, R. Hull, Mater. Res. Soc. Symp. Proc. 107 (1988) 3. [2] R. Hull, A.E. White, K.T. Short, J.M. Bonar, J. Appl. Phys. 68 (1990) 1629. [3] J.C. Barbour, S.T. Picraux, B.L. Doyle, Mater. Res. Soc. Symp. Proc. 107 (1988) 269. [4] C.W.T. Bulle-Lieuwma, A.H. Van Ommen, D.E.W. Vandenhoudt, J. Appl. Phys. 70 (1991) 3093. [5] S. Mantl, I. Michel, D. Guggi, H.L. Bay, S. Mesters, Appl. Surf. Sci. 73 (1993) 102. [6] R.S. Spaggs, K.J. Reeson, R.M. Gwilliam, B.J. Sealy, A. De Veirman, J. Van Landuyt, Nucl. Instr. and Meth. B 55 (1991) 836.

A.A. Galayev et al. / Journal of Crystal Growth 187 (1998) 435—443 [7] S.V. Hutchinson, M.F. Finney, K.J. Reeson, M.A. Harry, R.M. Gwilliam, B.J. Sealy, Ion Implantation Technol. (1994) 934. [8] V.I. Nefedov, Handbook of X-ray Photoelectron Spectroscopy by Chemical Composition, “Chemistry”, Moscow, 1984.

443

[9] Y.-F. Hsieh, R. Hull, A.E. White, Appl. Phys. Lett. 58 (1991) 122. [10] A. Romano-Rodriguez, C. Serre, L. Calvo-Barrio, A. Perez-Rodriguez, J.R. Morante, R. Ko¨gler, W. Skorupa, Mater. Sci. Eng. B 36 (1996) 282. [11] G. Malegori, L. Miglio, Phys. Rev. B 48 (1993) 9223.