Using JPF1 format

JOURNAL OF APPLIED PHYSICS

VOLUME 90, NUMBER 1

1 JULY 2001

In situ transmission electron microscopy study of Ni silicide phases formed on „001… Si active lines V. Teodorescu and L. Nistor National Institute of Materials Physics, P.O. Box MG-7 Magurele, 76900 Bucharest, Romania

H. Bender, A. Steegen,a) A. Lauwers, and K. Maexa) IMEC, Kapeldreef 75, B-3001 Leuven, Belgium

J. Van Landuytb) Universiteit Antwerpen, EMAT, Groenenborgerlaan 171, B-2020 Antwerpen, Belgium

共Received 2 January 2001; accepted for publication 12 April 2001兲 The formation of Ni silicides is studied by transmission electron microscopy during in situ heating experiments of 12 nm Ni layers on blanket silicon, or in patterned structures covered with a thin chemical oxide. It is shown that the first phase formed is the NiSi2 which grows epitaxially in pyramidal crystals. The formation of NiSi occurs quite abruptly around 400 °C when a monosilicide layer covers the disilicide grains and the silicon in between. The NiSi phase remains stable up to 800 °C, at which temperature the layer finally fully transforms to NiSi2. The monosilicide grains show different epitaxial relationships with the Si substrate. Ni2Si is never observed. © 2001 American Institute of Physics. 关DOI: 10.1063/1.1378812兴

I. INTRODUCTION

group Pnma. Both Ni2Si and NiSi growth are diffusion controlled.6 NiSi2 is the last silicide to be formed at much higher temperatures, i.e., around 800 °C.5,6 It can grow epitaxially on 共001兲 and 共111兲 Si and the reaction is controlled by a nucleation phenomena. According to marker experiments,3,5 Ni is the only moving species for the formation of all three silicides. NiSi2 has a cubic CaF2 structure with a lattice parameter a⫽0.5395 nm. There is only a small lattice mismatch with Si at room temperature 共⫺0.4%兲. Therefore, epitaxial growth of NiSi2 on Si is the easiest among all silicides.8 Thin layers of epitaxial NiSi2 are grown on 共111兲 Si even at room temperature.9 More recently, it has been shown that NiSi2 can grow epitaxially on 共001兲 Si at temperatures as low as 220 共Ref. 7兲 or 400 °C,8 if the annealed Ni layer is very thin 关55 共Ref. 7兲 and ⬍10 nm,8 respectively兴. The in situ production of the silicidation reaction in the electron microscope permits the direct structural investigation of the formation and transformation of the different silicide phases. The growth of Ti10 and Co11 silicide phases in active lines has been studied in this way. In this article we present an in situ transmission electron microscopy 共TEM兲 study of nickel silicidation processes in submicron active lines on patterned 共001兲 silicon substrates.

Metal silicide thin films grown in a controlled manner on silicon wafers play an important role in device fabrication for the microelectronic industry. Due to their low resistivity they are widely used in very large scale integrated technology. Among these materials, nickel monosilicide has attracted a special interest in the fabrication of advanced submicron devices because it can be produced at lower temperatures, thus limiting diffusion in the process of high density integrated circuits. It also results in less Si consumption than needed for the formation of CoSi2, allowing more shallow junctions. For these devices, the microstructure of silicides plays an important role. Understanding the formation and growth evolution of nickel silicides in patterned structures with fine lines is not only technologically important but it is also interesting from a pure scientific point of view. Nickel silicide formation and growth were quite extensively studied in the past.1–7 They are produced by depositing a rather thick 共⬎100 nm兲 Ni film on 共111兲 or 共001兲 Si wafers followed by annealing at different temperatures. It has been found that the first silicide to grow is Ni2Si. It starts forming at 200 共Ref. 2兲 or 250 °C. It takes approximately 1 h to grow 100 nm of Ni2Si on 共001兲 Si at 300 °C.3 When all the Ni is consumed for the formation of Ni2Si, NiSi starts growing at approximately the same temperature. As it is pointed out in Ref. 3, it also takes 1 h to grow 100 nm NiSi at 350 °C. NiSi has the lowest resistivity among the Ni silicides and therefore it is desirable for device fabrication. It has an orthorhombic structure with lattice parameters a ⫽0.5233 nm, b⫽0.3258 nm, and c⫽0.5659 nm, and space

II. EXPERIMENT

共001兲 silicon wafers are patterned with a field oxide mask by means of the polyencapsulated local oxidation process,12 yielding active lines with different widths and spacings in the range of 0.25–10 ␮m isolated by a 150 nm thick oxide. Also, larger unpatterned regions are present on the wafers. After cleaning, a thin chemical oxide is left on

a兲

Also at: E.E. Department, K. U. Leuven, Leuven, Belgium. Author to whom all correspondence should be addressed; electronic mail: [email protected]

b兲

0021-8979/2001/90(1)/167/8/$18.00

167

© 2001 American Institute of Physics

Downloaded 03 Sep 2001 to 146.103.254.11. Redistribution subject to AIP license or copyright, see http://ojps.aip.org/japo/japcr.jsp

168

J. Appl. Phys., Vol. 90, No. 1, 1 July 2001

the silicon surface. Subsequently, these wafers are covered with a metal layer consisting of 12 nm Ni or 12 nm Ni capped with a thin Ti layer. Specimens for the in situ TEM analysis are prepared in plan view by cutting 3 mm diameter discs from both the blanket and patterned regions with an ultrasonic cutter. The discs are polished from the back side of the Si wafer by dimpling and subsequent ion milling from the same side till a hole appeared in the center region. In situ heating experiments are performed in the Philips CM20 electron microscope equipped with a double tilt Gatan heating holder which works up to 1000 °C. The specimen for the in situ TEM analysis is fixed on its rim with a Pt washer and screw in a small furnace mounted in the tip of the heating holder. A thermocouple welded to the furnace measures the temperature during the experiments. The furnace is heated by supplying current at a manually controlled rate. The stage is water cooled so that the temperature can be maintained at a desired value. Therefore, a firm contact to the furnace is necessary. As this is realized, the difference between the actual temperature of the specimen and the measured one depends on the thermal conductivity of the specimen. Once the temperature on the specimen is stabilized this difference tends to vanish. It has been observed that at high temperatures 共higher than 800 °C兲 the contact between the specimen and furnace can become looser because of dilatation differences between specimen, washer, and screw. This can induce slight differences between the measured temperature and the actual temperature on the specimen of a few degrees centigrade. III. RESULTS A. In situ heating of blanket specimens

In order to determine the temperature ranges where different forms of Ni silicides are produced, several in situ heating experiments are performed on blanket specimens. Figure 1 shows the evolution of the electron diffraction patterns with temperature for the 12 nm thin Ni layer deposited on the 共001兲 Si substrate. At room temperature, 关Fig. 1共a兲兴, the electron diffraction pattern shows rings corresponding to the polycrystalline Ni film and spots of the 关001兴 oriented Si substrate. Note that the 200 and 020 reflections, forbidden for the Si structure, do not appear in the diffraction pattern. By heating the specimen in the electron microscope at 300 °C for 5 min, the diffraction pattern changes as shown in Fig. 1共b兲. Faint diffraction spots at the 200 and 020 positions appear, showing that epitaxial NiSi2 starts growing as the first silicide to be formed in these experiments. By continued heating, in a temperature interval of approximately 100 °C for 10 min, NiSi2 is completely formed 关Figs. 1共c兲 and 1共d兲兴. It grows fully epitaxial on the 共001兲 Si substrate. The crystallographic relationships are: 关 001兴 NiSi2 储关 001兴 Si and (200)NiSi2 储 (400)Si. The TEM images 共Fig. 2兲 of the blanket specimen at different temperatures corresponding to the diffraction patterns of Fig. 1 illustrate the formation and growth of NiSi2. The room temperature image of Fig. 2共a兲 reveals, of course, the morphology of the polycrystalline Ni thin film. The Ni

Teodorescu et al.

FIG. 1. The evolution with temperature of the electron diffraction patterns of Ni/共001兲Si blanket specimen: 共a兲 room temperature 共RT兲, 共b兲 300, 共c兲 350, and 共d兲 400 °C. Note the apparition of NiSi2, as first silicide to be formed.

layer is continuous and the crystallites are rather small: on the average less than 10 nm. When NiSi2 starts forming, the contrast of the TEM image, Fig. 2共b兲, changes. A gray mottled ‘‘flower’’-like contrast appears at 300 °C showing the formation of the NiSi2 phase. The possible epitaxial Ni silicide formation at these low temperatures is consistent with previous results.7 As the temperature is slightly increased 关Figs. 2共c兲 and 2共d兲兴, the NiSi2 crystallites double their size and many of them reveal square shapes. In Fig. 2共d兲, which corresponds to the diffraction pattern of Fig. 1共c兲, besides the Ni and NiSi2, some denuded zones of bright contrast can be observed. Hence, the thin film is no longer continuous. These denuded zones start appearing at 330 °C 关Fig. 2共c兲兴 and develop with increasing temperature. Note that approximately 5 min are spent on each heating step so

FIG. 2. TEM images of Ni/共001兲Si blanket specimen showing the formation of NiSi2: 共a兲 RT, 共b兲 300 °C NiSi2 appears, at 共c兲 330 °C and 共d兲 350 °C development of NiSi2.



FIG. 3. TEM image 共a兲 and corresponding diffraction pattern 共b兲 at 400 °C showing the formation of NiSi on the blanket specimen.

that, until Fig. 2共d兲 is taken, the specimen has been annealed for about 15 min. By keeping the specimen temperature around 400 °C for 10 min, another important modification appeared in the TEM images. A rather fast transformation front traverses the image, with a velocity which is estimated at approximately 200 nm/s, leaving the image as in Fig. 3共a兲. The corresponding diffraction pattern of this region of the film is given in Fig. 3共b兲. By comparison with Fig. 1共d兲, which shows the diffraction pattern of the same region of the specimen, just before the transformation front has passed, one can see that the newly appeared diffraction spots reveal the formation of another compound. Note that the lapse of time between the two diffraction patterns is approximately 1 min. As we shall see further the fainter diffraction reflections in Fig. 3共b兲 can be indexed as epitaxial NiSi along the 关001兴 NiSi zone axis. The TEM image in Fig. 3 also shows that the squareshaped NiSi2 crystallites grow in size and join each other. The monosilicide layer tends to grow continuously covering the NiSi2 crystallites and the empty spaces. The ‘‘wrinkled’’like contrast which appears might suggest the presence of interfacial dislocations between the two silicides. This contrast is present in the TEM images as long as NiSi is present in the specimen. At temperatures higher than 800 °C, NiSi transforms to NiSi2. This is obvious in the diffraction patterns 共not shown here兲, but also in the TEM image of Fig. 4共b兲. This image shows an increased sharpness of the NiSi2 grains. It reveals the pyramidal shape of NiSi2 crystallites, as

Teodorescu et al.

169

FIG. 4. TEM images of the blanket specimen at different temperatures: 共a兲 at 500 °C both NiSi and NiSi2 coexists, 共b兲 at 860 °C only the NiSi2 pyramids are present.

it was already observed in cross section TEM images by different authors.13,14 They showed that the NiSi2 pyramids grow with the tip inside the Si substrate and the 共111兲 faces parallel to the 共111兲 planes of silicon. The square base of the pyramids is parallel to the Si wafer surface. The NiSi2 pyramids are oriented on the Si cube faces with the edges of the square base parallel with the 具110典 directions of the Si. When these pyramids grow or join they form truncated pyramids 关Fig. 4共b兲兴 with a rectangular base. It is also worth mentioning that the NiSi2 crystallites did not grow in size by heating in the temperature range of 400–800 °C, where the NiSi phase still exists 共compare Figs. 3 and 4兲, showing that the NiSi2 phase has been formed at low temperatures 共300– 400 °C兲. Figure 5 reveals the diffraction patterns of different rather large regions 共5 ␮m in diameter兲 of the silicide film grown on the 共001兲 Si blanket. These diffraction patterns are taken at temperatures of 470 or 500 °C where the NiSi2 and NiSi coexist and are fully formed. They show that both silicides grow epitaxially. Note the similarities of the diffraction pattern from Fig. 5共a兲 with that from Fig. 3共b兲, which is taken just after the NiSi is formed. Because of the layer morphology of the specimen, double diffraction effects are present which give rise to a large number of diffraction spots in positions forbidden by the space group. By indexing the diffraction patterns of Figs. 5共a兲–5共c兲, the following crystallographic relations for the epitaxy of NiSi can be deduced. 共1兲 From Fig. 5共a兲: 关001兴 NiSi//关001兴 Si and 共020兲 NiSi// 共040兲 Si.


170

Teodorescu et al.


FIG. 6. Successive plan view TEM images showing the formation of Ni silicides at different temperatures, in an 800 nm thick active line: 共b兲–共c兲 NiSi2, 共d兲 NiSi and NiSi2. Note the formation of railways at the border of the lines.

共001兲 Si is realized in such a way that one of the faces of the NiSi rectangular prism is parallel with the Si cube face. The NiSi rectangular prism is oriented with one of its edges parallel to the face diagonal of the Si cube. No important changes occur in the TEM images and in the diffraction patterns upon heating the specimen in the temperature interval 500–800 °C. At temperatures higher than 800 °C, the diffraction reflections characteristic for the nickel monosilicide disappear. The disappearance of NiSi at high temperatures is indeed also observed in the TEM image from Fig. 4共b兲, as it has been shown previously. B. In situ heating of narrow structures

FIG. 5. Diffraction patterns of large grains on the blanket specimen showing three different epitaxial orientations for NiSi 共a兲, 共b兲, and 共c兲. The NiSi and epitaxial NiSi2 coexist at temperatures around 500 °C.

共2兲 From Fig. 5共b兲: 关100兴 NiSi//关001兴 Si and 共020兲 NiSi// ¯ 20) Si. (2 ¯ 兴 NiSi//关001兴 Si and 共020兲 NiSi// 共3兲 From Fig. 5共c兲: 关 101 ¯ (22 0) Si. For NiSi2 the epitaxial relations remain unchanged as mentioned before. From the above epitaxial relations, 共1兲 and 共3兲 are most often encountered. Random NiSi orientations are not observed on blanket specimens. The epitaxy of NiSi on

The next step for this in situ study is to find out how the silicidation reaction takes place in narrow structures 共submicron lines兲, which are used in semiconductor devices.14–16 In our particular case the narrow structures consist of groups of ten lines with equal nominal width and spacing, or of isolated lines. The thinning of the plan view specimens for TEM is done in such a way that a group of active lines crosses the region transparent to the electron beam more or less parallel with the slope direction of the specimens. Different specimens for TEM analysis have been prepared for different heating treatments in the electron microscope. From these experiments some general observations can be outlined. The growth of Ni silicides in active lines follows the regime of growth as observed for the blanket film. This is shown in Fig. 6 which illustrates in 共a兲 a plan view TEM image of a 800 nm wide active line. The line itself can be recognized from the bend contours produced by the Si substrate. A 12 nm Ni layer has been deposited on the line and on the amorphous SiO2 spacings. By heating to approximately 300 °C firstly NiSi2 is formed. The flower-like contrast is evident in the line 关Fig. 6共b兲兴, while the contrast on the oxide spacings remains unchanged. Also, the 共200兲 reflections appear in the diffraction pattern corresponding to the line, as in Fig. 1共b兲. At 370 °C the NiSi2 pyramids are well formed. They tend to align themselves at the borders of


Teodorescu et al.


the active line, while in the center of the line they grow more randomly, as in the blanket 共Fig. 2兲. After approximately 10 min at 400 °C 关Fig. 6共d兲兴 the transformation front passes and NiSi is formed and the whole active line is covered with silicides. The NiSi2 pyramids join and form, at the edges, a kind of ‘‘railway’’ which will exist no matter how high and how long the specimen is heated. As in the blanket specimens, in the active lines NiSi2 grows epitaxially by following the same crystallographic relations, i.e., 关 001兴 NiS2 //关001兴Si and (200)NiSi2 //共400兲Si. It is worth mentioning that the active lines on the 共001兲 Si wafer are along 关110兴 directions of Si. Because, as it has been previously mentioned, the edges of the NiSi2 pyramids are parallel with the 具110典 direction of Si, it is possible that this favors the alignment of NiSi2 pyramids along the borders of the active lines. Also, in the active lines NiSi grows in an oriented way during further heating. Large 300–400 nm crystallites are formed in crystallographic relationships with Si support. Most often the NiSi crystallites grow in such a way that one face of the NiSi prism is parallel with the Si cube face. Diffraction patterns like those from Figs. 5共a兲 or 5共b兲 on blanket specimens are common for the active lines too. Two other situations recorded at different temperatures and in different parts of a 500 nm wide active line are illustrated in detail in Figs. 7 and 8. When the rail joins the middle part of the line, the NiSi and NiSi2 crystallites join. The tendency is to create a continuous layer of silicides. This occurs around 400 °C when the transformation front passes. The diffraction pattern of Fig. 7共b兲 corresponds to the zone marked with an arrow on Fig. 7共a兲, with a large NiSi crystallite and two smaller grains near the rail. The diffraction spots corresponding to this large NiSi crystallite are marked with white lines on Fig. 7共b兲. Si and NiSi2 diffraction reflections are encircled. From this pattern one can deduce the ¯ 兴 NiSi//关001兴Si. following crystallographic relation: 关 011 There is a rotation of 2° between the two structures. In a similar way, Fig. 8共b兲 shows the diffraction pattern mostly corresponding to the large NiSi crystallite from Fig. 8共a兲. In this case the crystallographic relation is: ¯ 关 11 0 兴 NiSi//关001兴Si, with the NiSi and Si rotated by 3°. These last two crystallographic relations imply that for certain NiSi crystallites in the active lines the diagonal of the NiSi rectangular prism faces is parallel with the edges of the Si cube. On the borders of the active lines, i.e., where the NiSi2 rails are, the NiSi crystallites frequently showed different orientations. The most observed orientations are along ¯ 1 兴 or 关 ¯2¯3 1 兴 zone axes. the 关 13 The evolution of the silicidation reaction in lines where 12 nm Ni capped with Ti are deposited on 共001兲 Si is shown in Fig. 9. Again NiSi2 is the first silicide to form at temperatures around 300 °C, as rails on the borders of the active lines and in the lines as well 关Fig. 9共b兲兴. At 400 °C NiSi is also ¯ ) difformed 关Fig. 9共c兲兴. The dark field image with the (101 fraction spot of NiSi 关Fig. 9共d兲兴 reveals the presence of nickel monosilicide in the active line and on its borders. A further temperature increase results in an increase in size of the silicide crystallites which start to agglomerate.

171

FIG. 7. TEM image 共a兲 and corresponding diffraction pattern 共b兲 of Ni silicides in an active line. In 共b兲 the NiSi2 and Si reflections are encircled while the NiSi reflections are marked with black points and gratings.

Finally, we have to point out that Ni2Si is not found in any of the heating experiments, either on blanket specimens or on specimens with active lines. IV. DISCUSSION

In the literature it is reported that when a rather thick 共100–200 nm兲 Ni film is deposited on Si and annealed, the silicidation reaction follows the sequence Ni2Si 共orthorhombic兲→NiSi 共orthorhombic兲→NiSi2 共cubic兲, in a temperature range of 250–800 °C.6 NiSi forms only after the complete growth of Ni2Si with complete consumption of the Ni.3 It is observed that this sequence is very dependent on the thickness of the deposited Ni film.8,17,18 Very thin 共⬍3 nm兲 Ni films deposited on 共001兲 Si transform into NiSi2 at annealing temperatures in the range 350–400 °C.8,19 It is shown that NiSi2 nucleates directly from the deposited Ni, NiSi2 being the only silicide formed. The NiSi2 crystallites grow laterally across the film and epitaxial with the Si substrate. In the initial nucleation stage the very thin NiSi2 films consist only of islands limited by 共111兲 facets, called facet bars, without 共100兲 interfaces. As the islands grow by facet progression flat 共100兲 interfaces are left behind.


172


Teodorescu et al.

FIG. 9. Successive plan view TEM images showing the formation of Ni silicides in a 400 nm thick active line with Ti cap: 共b兲 NiSi2, 共c兲–共f兲 NiSi2 and NiSi. Note the presence of railways at all temperatures where the NiSi2 crystallites exist and the dark field image in 共d兲 revealing the NiSi crystallites.

FIG. 8. TEM image 共a兲 and corresponding diffraction pattern 共b兲 of Ni silicides in an active line. In 共b兲 the NiSi2 and Si reflections are marked with bullets and the NiSi reflections are marked with crosses and gratings.

In the in situ TEM heating experiments described here, the Ni layer is thicker 共12 nm兲, but yet much thinner than in the earlier annealing experiments 共⬎100 nm兲.1–6 Consequently, it becomes less surprising that NiSi2 is the first silicide to be formed. Since, as determined by marker experiments,5,6 Ni diffusion coefficients are 10 times larger that Si diffusion coefficients in the initial stage of NiSi2 formation Ni diffuses in Si and forms NiSi2. NiSi2 nucleation on 共111兲 Si occurs even at room temperature9 and the epitaxy is easiest on the 兵111其 Si planes. For this reason the growing NiSi2 crystallites take pyramidal shapes, oriented with the pyramid tip protruding into the Si substrate 关the 共111兲 planes of NiSi2 being parallel to the 共111兲 planes of Si兴. The pyramid bases are parallel with the wafer plane, since, as revealed in the diffraction patterns 关Fig. 1共d兲兴 the 共200兲 NiSi2 plane is parallel to the 共400兲 Si plane. If the Ni source is sufficient, these pyramids grow and neighboring crystallites join forming truncated pyramids 共Fig. 4兲. This process occurs in our case in the temperature interval of 300–400 °C. It is obvious from this description that the NiSi2 /Si interface has to be very rough. The atomic structure of this interface is analyzed by Cherns et al.20 and Chen and Chen.21 They showed that the junctions between the 共001兲 and 共111兲 facets

are made through dislocations with Burgers vector a/4具 111典 . In the in situ TEM experiments, in the temperature range 400–800 °C NiSi is found to coexist with NiSi2. This is the temperature range where NiSi is found as the stable silicide in all the previous experiments. NiSi can only be formed if unreacted Ni still existed. This can happen only if the Ni film is thick enough not to be completely consumed by the NiSi2 pyramids. Ni traces are indeed still observed in many of the diffraction patterns taken at temperatures around 400 °C, before the transformation front passed. After NiSi is formed no traces of Ni are observed in the diffraction patterns anymore. It is worth mentioning in this respect that CoSi2, which is isostructural to NiSi2, can transform by annealing to CoSi if additional Co is available for the reaction CoSi2⫹Co⫽2 CoSi2. This would imply a shrinkage or annihilation of the pyramids which is not observed in the present experiments. The NiSi film grown in our in situ heating experiments should be rather thin since only a slight increase of the NiSi2 crystallite size is observed at 860 °C, Fig. 4, where NiSi is finally completely transformed to NiSi2. One more problem is the presence of the railway structure at the border of the active lines in all the specimens investigated. In the attempt to understand this phenomenon cross section TEM specimens are prepared from the unheated samples. One set of cross section specimens is obtained in the standard procedure by slicing, gluing the pieces face to face, curing the glue at 150 °C for 1 h, polishing, and finally ion milling 共in a Balzers RES 010, 5 kV, 6° incidence from the sample surface兲 to get transparency to the electron beam. The temperature of the specimen during the ion milling is not well known, but from our experience it is not higher than 150 °C. Figure 10 a shows a TEM image of an active line in cross section. Several interesting features are revealed here. First, a very thin amorphous layer 共of approximately 1 nm兲 exists between the metallic layers and the silicon substrate.



FIG. 10. Cross section TEM images prepared in two different ways showing the profile of the active lines. In 共a兲 the sample is heated at 150 °C for 1 h while preparing the cross section specimen. Note the presence of NiSi2 pyramids in the line and on the borders. In 共b兲 the cross section specimen is prepared without heating. Note the borders of the line where the Ni film touches the Si substrate.

This is the chemical oxide left on the silicon surface after the cleaning step done before the metal deposition. This layer is not visible at the borders of the active line, where, on the other hand, the Ni layer becomes thinner. NiSi2 pyramids are formed at the borders and in the middle of the lines even at temperatures as low as 150 °C necessary for the cure of the glue. At higher magnifications 共Figs. 11 and 12兲 more details are revealed. Figure 11 is a high resolution cross section image of a middle of an active line. The three-layer periodicity in the contrast is typical for twinned material22 and indicates that the NiSi2 is at least partially formed in the twinned 共‘‘B-type’’ 19,21,23兲 orientation. The formation of the NiSi2 pyramids by diffusion of Ni through the thin SiOx layer is obvious. Figure 11 reveals a partial depletion 共consumption兲 of the Ni thin film in the region facing the pyramid base. Such a process of NiSi2 for-

FIG. 11. Cross section high resolution TEM images taken at the middle of an active line. The specimen is prepared in cross section by heating at 150 °C for 1 h. Note the amorphous SiOx layer at the interface between the metallic film and silicon and the NiSi2 pyramids protruding in the silicon substrate.

Teodorescu et al.

173

FIG. 12. Cross section high resolution TEM image taken at the edge of an active line. Note the disappearance of the amorphous SiOx layer and the significant decrease of the Ni film thickness at the line edge. A NiSi2 pyramid is also revealed at the line edge in the Si substrate.

mation might be similar to the oxide mediated epitaxy of the isostructural CoSi2 on Si, recently reported.24 Figure 12 is a high resolution image of the border of the active line. It actually shows how the railways are formed by diffusion of Ni in Si at the line border, with subsequent formation of NiSi2. This indeed confirms the results obtained by the in situ experiments performed on plan view specimens 共Figs. 6–9兲. The preferential growth of NiSi2 on the edges might be related to a smaller thickness of the chemical oxide in these regions or to the higher stress which might favor the epitaxial growth. In an attempt to study the initial condition of the lines, cross section TEM specimens are also prepared avoiding any heating. A single strip is glued on a TEM grid with cyanocrylate which does not need a curing step. The thinning is done by focused ion beam 共FIB兲 milling with the trench procedure.25 The temperature rise during FIB milling is estimated to be only on the order of 10 °C.25,26 Figure 10共b兲 shows a cross section TEM image of an active line in a specimen prepared in this way. The amorphous SiOx layer between the Ni film and Si substrate is, again, less visible at the borders of the line where the Ni film ‘‘touches’’ the Si substrate. This is better evidenced at higher magnifications in Fig. 13, which is a high resolution image of the active line edge. Figure 13 also shows that the Ni layer near the edge of the line is thinner, but to a lesser extent than in Fig. 12. Hence, although the minimized heating procedure followed for the FIB specimen preparation, the reaction still occurred at the edges of the active lines. Whether this reaction took place during the specimen preparation or before cannot be concluded. It is however clear that the configuration at the edges of the active lines, i.e., a possibly thinner chemical oxide and high stress are the origin of the railway formation during the in situ heating experiments.


174

Teodorescu et al.


In line structures, a preferential growth of the epitaxial grains near the edges of the lines is observed, resulting in a railway structure. This is related to the specific shape of the opened oxide windows. Also, the stress at the line borders might favor the epitaxial growth by minimizing the mismatch. ACKNOWLEDGMENT

This work was performed in the framework of the Bilateral Program BIL 97/93 between Flanders and Romania. D. J. Coe and H. Rhoderick, J. Phys. D 9, 965 共1976兲. S. S. Lau, J. W. Mayer, and K. N. Tu, J. Appl. Phys. 49, 4005 共1978兲. 3 T. G. Finstad, Phys. Status Solidi A 63, 223 共1981兲. 4 H. Foll, P. S. Ho, and K. N. Tu, Philos. Mag. A 45, 31 共1981兲. 5 F. D’Heurle, C. S. Petrsson, L. Slot, and B. Strizker, J. Appl. Phys. 53, 5678 共1982兲. 6 F. D’Heurle, C. S. Petrsson, J. E. E. Baglin, S. J. La Placa, and C. Y. Wong, J. Appl. Phys. 55, 4208 共1984兲. 7 L. J. Chen, C. M. Doland, I. W. Wu, J. J. Chu, and S. W. Liu, J. Appl. Phys. 62, 2789 共1987兲. 8 J. P. Sullivan, R. T. Tung, and F. Schrey, J. Appl. Phys. 72, 478 共1992兲. 9 R. T. Tung and F. Schrey, Appl. Phys. Lett. 55, 256 共1989兲. 10 L. M. Gignac, V. Svilan, L. A. Clevenger, C. Cabral, Jr., and C. Lavoie, Mater. Res. Soc. Symp. Proc. 441, 255 共1997兲. 11 C. Ghica, L. Nistor, H. Bender, J. Van Landuyt, A. Steegen, A. Lauwers, and K. Maex, J. Mater. Res. 16, 共2001兲. 12 G. Badenes, R. Rooyackers, I. De Wolf, and L. Deferm, Proc. Int Conf. Solid State, Devices and Materials, 26 –29 August, 1996, p. 46. 13 D. Fathy, O. W. Holland, and J. Narayan, J. Appl. Phys. 58, 295 共1985兲. 14 A. Lauwers et al., Microelectron. Eng. 50, 103 共2000兲. 15 D. X Xu, S. R. Das, C. J. Peters, and L. E. Erickson, Thin Solid Films 326, 143 共1998兲. 16 F. Deng, R. A. Johnson, P. M. Asbeck, and S. S. Lau, J. Appl. Phys. 81, 8047 共1997兲. 17 R. T. Tung, J. M. Gibson, and J. M. Poate, Phys. Rev. Lett. 50, 429 共1983兲. 18 J. M. Gibson, J. L. Bastone, R. T. Tung, and F. C. Uterwald, Phys. Rev. Lett. 60, 1158 共1988兲. 19 J. M. Gibson and J. L. Bastone, Surf. Sci. 208, 317 共1989兲. 20 D. Cherns, C. J. D. Hetherington, and C. J. Humphreys, Philos. Mag. A 49, 165 共1984兲. 21 W-J. Chen and F-R. Chen, Philos. Mag. A 68, 605 共1993兲. 22 H. Bender, A. De Veirman, J. Van Landuyt, and S. Amelinckx, Appl. Phys. A: Solids Surf. 39, 83 共1986兲. 23 W. J. Chen, F. R. Chen, and L. J. Chen, Appl. Phys. Lett. 60, 2201 共1992兲. 24 R. T. Tung, Appl. Phys. Lett. 68, 3461 共1996兲. 25 H. Bender, Inst. Phys. Conf. Ser. 164, 593 共1999兲. 26 T. Ishitani and T. Yaguchi, Microsc. Res. Tech. 35, 320 共1996兲. 1 2

FIG. 13. Cross section high resolution TEM image at the edge of an active line. The specimen is prepared in cross section by FIB without heating. Note that at the line edge the Ni film touches the Si substrate. The Ni layer thickness slightly decreases while the amorphous SiOx layer vanishes.

V. CONCLUSIONS

It is shown that during the in situ heating of thin Ni layers on silicon epitaxial NiSi2 is already formed at low temperatures. The formation of the disilicide as the first phase is related to the presence of a thin chemical oxide between the Ni and silicon substrate. The disilicide coexists with unreacted Ni until a temperature of 400 °C when a fast transformation of the unreacted Ni to NiSi occurs. This layer tends to cover the whole surface, i.e., over the disilicide pyramids as well as over the silicon in between. It shows different epitaxial relationships with the silicon. Up to 800 °C the NiSi and NiSi2 coexist and show no noticeable grain growth. At higher temperatures the monosilicide also transforms to the disilicide. In the whole temperature range used in this study, epitaxial NiSi2 grains are present. However, Ni2Si, reported as the first phase during the reaction of thick Ni layers on Si, is not observed during these in situ experiments.


Using JPF1 format - Universiteit Antwerpen

Using JPF1 format - Universiteit Antwerpen

Suggest Documents