Physical techniques for silicon layer analysis

MICl~Ogi.lICfllOI¢~ ELSEVIER

Microelectronic Engineering 40 (1998) 223-237

Physical techniques for silicon layer analysis A. P6rez-Rodrfguez, A. Cornet, J.R. Morante. E.M.E., Departament d'Electrbnica, Universitat de Barcelona, Avda. Diagonal, 645-647, 08028 Barcelona, Spain This paper reviews the application of physical techniques (SIMS, RBS, XPS, AES, XRD, Raman and FTIR spectroscopies and AFM) for the characterisation of layers, structures and processes in Si technology. The principles of the techniques are given, and their use for the analysis of Si related structures and processes is discussed. Structural features related to the device and system performances are considered. These include features related to amorphous, polycrystalline and single crystal layers, as grain size and crystal orientation, disorder, defects, strain, composition ..... as well as the quality and nature of surfaces and interfaces in Si related layers (Si, SiGe and ternary related alloys, SiC, dielectric layers, silicides...). Examples taken from the current research in these areas are presented. I.-~TRODUCTION The development of complex microelectronic systems strongly needs for the solution of fundamental problems on the materials and structures. The increasing complexity of the used systems implies stronger requirements on the characteristics of the structures to be made. Features as interface and surface quality, roughness, strain, doping profiles, chemical composition and crystalline structure of the different phases are critical for the performance of the devices and systems to be developed. These c~in modify the electrical and optical performance of electronic devices and optoelectronic components. In addition, the increasing interest on integrated sensors and microsystems (MEMS) arise new needs including mechanical and thermal properties. Materials for packaging and substrates have also become determining for many applications such as high frequency, high temperature or harsh environment. All this is conditioned to the availability of suitable analytical tools to determine these features and properties. The knowledge and development of these tools is a key point in Microelectronics research. This paper reviews the application of physical techniques for the characterisation of structures and processes in Si technology. These include ion beam, mass spectroscopy and electron emission spectroscopy techniques, X-Ray Diffraction, vibrational spectroscopies and nanometric imaging [1-3]. In the next sections, the principles of the techniques are given and their use in Si technology

is presented. Emphasis has been given to activities developed in the group of characterisation of microelectronic materials from the Electronic Materials and Engineering Laboratory (EME).

2.- SIMS Bombarding the surface of a solid with an energetic ion beam generates a variety of secondary transitions, including electrons, photons and ions. The analytical technique Secondary Ion Mass Spectroscopy (SIMS) is based on the detection of the secondary ions ejected as consequence of the energy and momentum transfer processes occurring after the collision [3,4]. The essential features of a SIMS system are the ion source and the secondary ion mass analyser. The primary ions produced on the ion source are separated by a primary ion mass analyser and focused by a condensing lens on the sample. Charge deflection plates are used to position the beam or raster it over the sample surface. The primary ion beam currents used in typical SIMS analysis range from 10 nA to 15 p.A, having energies between 1 and 20 keV. Depending on the application and mode of operation, lateral resolution between 50 nm and 2 gm can be obtained. The SIMS spectrometer is used to separate the secondary ions with respect to their charge to mass ratios. The atomic ions give an elemental identification, whereas the clusters can provide information on the chemical groups present on the region analysed. The depth of the mixing zone, which limits the depth resolution of the SIMS

0167-9317/98/$-see front matter Copyright © 1998 Elsevier Science B.V. All rights reserved. PII: S0167-93 17(98)00273-1

224

A. P~rez-Rodr{guez et al. / Microelectronic Engineering 40 (1998) 223-237

analysis to 2-30 nm, is function of the energy, angle of incidence, mass of the primary ions and sample material. Since generation of ions from the mixing zone requires removal of the material, SIMS is basically a depth profiling technique. The rate at which the mixing zone is advanced is called the sputtering rate. To improve the depth resolution, it is necessary to generate the ions from a uniform depth• The sputtering yield (atoms removed by incident ion) depends on the energy of the primary ion beam. Moreover, the ion yield (ionisation efficiency of an element) is enhanced by adequate selection of the primary ions. On this basis, the most favourable primary ions are Cs + to enhance the yield of negative secondary ions or O to enhance the yield of positive secondary ions. Nevertheless, the accumulation on the surface of the sample of charge carried by positive primary ions can affect the measurements, defocusing or moving the primary ion beam out from the analytical area of interest or changing the energy of the secondary ejected ions. The effect of sample charging can be eliminated or reduced by compensating the positive charge with an additional electron beam. The SIMS spectrum consists in the plot of the secondary ion intensity versus the charge to mass ratio. The spectrum from even high purity elements can be very complex, because it includes as well ionised atoms and isotopes as ionic complexes or multiply ionised species. The assignment of each peak is difficult, due to the overlap in the mass spectrum of ionised atomic clusters or multiply charged ions with the same mass-to-charge ratio. This can cause erroneous assignment of an element or degrade the detection limit of the element of interest. A voltage offset technique where the mass spectrometer is adjusted to accept only ions in a certain kinetic energy range or high mass resolution techniques are used to discriminate the contribution of interfering ions. Despite these difficulties, SIMS is one of the most sensitive analytical techniques, with elemental detection limits in the ppm to sub-ppb range. The interpretation of the SIMS spectrum yields a great deal of information about the sample components or the presence of any surface impurities or contaminants. In Si technology, SIMS is mainly used in the depth profiling mode, to measure the concentration of specific preselected elements as a function of the

10 24

•

10 23

i

'=

-

i

-

o

C as-implanted (SIMS

o

C (600°C, lh) (SIMS)

•

C (900°C, lh) (SIMS

........ C (1100°C, lh) (SIMS

10 22

10 u MP

;;

,,"

L)

1020 0

,;

ii 100

200

300

400

\

51)0

d e p t h (rim) Figure 1. Carbon SIMS profiles from a SiO2 film implanted with C + and Si + ions and annealed. depth from the surface. This allows accurate measurements on the in depth profile of dopants and impurities, and the analysis of their evolution under processing of the structures. So, SIMS has been used to characterise diffusion, segregation and pile-up effects in multilayered structures, gettering of impurities, intermixing .... Fig. l shows an example of this application for the study of the distribution of C atoms in a S i O 2 layer implanted with C ÷ and Si ÷ ions and annealed [5]. The Si implanted profile is simulated by TRIM [6]. The aim of these processes is to achieve carbon containing nanoparticles with luminescence in the visible region. As it is shown, the carbon content decreases with annealing. After annealing at 1100°C there is only some remaining carbon around the Si projected range. Disappearance of carbon with the annealing is likely related to the formation of CO and CO2 species, which diffuse away from the 8 i 0 2 layer. Excess of Si atoms in the network inhibits carbon oxidation, and siliconcarbon ]amorphous complexes are formed. The accurate measurement by SIMS of carbon implanted profiles in Si has also been used for determining the minimum carbon content needed for obtaining an etch stop layer in Si technology, which is about 1-2x1022 cm -3. Fig. 2 shows the SIMS profiles measured before and after chemical etching (TMAH) [7]. These layers were implanted at temperature high enough to avoid amorphisation

A. PFrez-Rodrfguez et al. I Microelectronic Engineering 40 (1998) 223-237

225

C concentration (em "~) I0 I'ilndOlll

.......... 1 ...................

i... , ~

.......

..... unetehed

lO:" "....... ""etch-stop iplane --"~'

m. = O L)

.

400. . . . . . . . . . . . . . . . . . . . . . . . . . . .600 800 ........................... Depth (nm)

=

io 00

ahgnc

d

~

6bO

400 channel

Figure 2. SIMS profiles from a Si wafer implanted with C + ions, before and after chemical etching.

Figure 3. RBS spectra from a SiGe amorphous film on Si.

(500°C), and their etch-stop behaviour has been explained assuming a percolation process in a SiC/Si binary system. SIMS can also work in: (i) bulk analysis mode, to achieve maximum sensitivity to trace-level components, (ii) mass scan mode to survey the entire mass spectrum within a certain volume of the specimen, and (iii) imaging mode to determine the lateral distribution of specific elements. When used to provide quantitative analysis, the dependence of the ion yield on the primary beam and sample (matrix effects) must be taken into account. Accurate quantification strongly requires the comparison to standards or reference materials.

On the other hand, the energy lost in the elastic collision is given by the momentum transfer in the event, which is characterised by the mass of the nucleus. Accordingly, the overall energy spectrum of backscattered ions gives information on both the elemental composition and depth distribution of these elements in the host material. Fig. 3 shows an RBS spectrum measured with He + ions at 1.5 MeV from an amorphous SiGe0.25 thin film on Si (thickness 70 nm) (random spectrum). Ions scattered by the Si atoms give a continuous signal from channel 464. This is the maximum energy of the ions scattered by the Si atoms and corresponds to scattering with a surface atom. The yield measured at lower channels is given by the amount of the Si atoms at the corresponding depth. The dip observed in the surface region is related to the lower content of Si in the surface SiGe film. Scattering of the ions with a lighter atom (as Ge) gives a lower energy loss in the elastic collision. Hence, scattered ions will give rise to a peak in the RBS spectrum at higher energies• This is the peak at around chanel 670 in Fig. 3. In a similar way, ions scattered by heavier atoms will give rise to a peak superimposed to the continuum spectrum from the substrate. In both cases, the shape of the peak will be determined by the impurity distribution. Hence, the analysis of the different peaks in the RBS spectra allows to quantify the sample composition and its distribution in depth. The detailed knowledge of the interaction mechanisms

3.-RUTHERFORD BACKSCATTERING Rutherford Backscatttering (RBS) constitutes a powerful ion scattering technique for compositional depth profiling and structural analysis of crystalline layers in Si technology [8,9]. Typically, an RBS spectrum is obtained by bombarding the surface of the sample with a beam of high energy light ions (He +, H ÷ at 1 - 3.4 MeV). The spectrum is obtained by detecting ions backscattered with a certain angle as a function of their energy. The loss of energy of these ions is determined by two mechanisms: i) inelastic electronic interactions and' ii) an elastic backward collision with a nucleus. The amount of energy lost by electronic interactions depends on the depth travelled by the ion, which is determined by the depth at which the elastic collision takes place.

226

A. P~rez-Rodrfguez et al. / Microelectronic Engineering 40 (1998) 223-237

between the scattered ions and the host atoms allows a high accuracy in compositional determination, typically of a few percent. Detection limits depend on the mass of the element, ranging from a few parts per million for heavy elements to a few percent for light elements. The thickness of the analysed layer is typically below 2 btm, although implanting lighter ions (as protons) allows an increase of the sampling depth (up to one order of magnitude). Lateral resolution for most instruments is on the order of 1-2 mm 2, but with microbeam systems lateral resolutions down to 1-10 ~tm can be achieved. Fig. 3 also shows the spectrum obtained when the incident beam is aligned in a low density crystalline direction of the substrate. Channelling of incident ions in low density paths determines a decrease of the scattered yield, and the comparison between both random and aligned spectra gives information on the crystalline quality of the layer. Defects as impurity or host atoms in interstitial positions will give rise to an increase of the aligned yield. This allows to quantify by RBS the amount o f impurities in substitutional sites, in addition to the total content (measured form the random spectrum). The comparison between both spectra also allows to quantify process induced residual damage. This has been extensively used for the analysis of damage induced by ion implantation processes and its recovery by post implant annealing. The 100% damage is obtained when both random and scattered yields coincide, which corresponds to the amorphisation of the layer (loss of long range order). This is the case in the 315 nm depth surface region of Fig. 3, which shows the existence of a 245 nm thick amorphous Si film below the SiGe one. In this case, the Si substrate was preamorphised by ion implantation with Si ÷ ions before deposition of the SiGe film [10]. Fig. 4 shows the partial damage recovery on the top Si layer from a Silicon-OnInsulator (SOI) structure implanted with a high dose of carbon atoms after annealing at 1250°C [ 11 ]. As profiling technique, RBS is free from matrix effects that are normally associated with techniques using sputtering, as SIMS, XPS or AES. In contrast with these, RBS does not destroy the sample surface, and due to this the technique is usually referred as not destructive. However, one has to be cautious with this, as the analysing RBS beam generates defects in the matrix, which degrade its crystalline quality. Damage related to the analysing beam appears at depths close to the projected range of the

counts (x 100) lmpl'at5()O°C

,0 t [

•

I

'

~ . r - , - -

i'

I

' ^~/Q.. -

'

-

I

r

'

"

~.k.

[

440

46O

480

5OO

/

~/aligne d

Impi. at 500OC .~. 12500 C/2he~./--V-~_

0

'.

:

.

520

channel Figure 4. RBS spectra of the top Si layer of a SOI substrate implanted with C. ions (of the order of 1-10 ~m) and in some cases even amorphisation of the buried damaged layer has been observed [ 12]. Other factor limiting the application of the technique is related to the convoluted nature of the mass and depth information available in an RBS spectrum, which in some cases results in spectral interference between the peak of a light element and a buried heavier one. In these cases, the correlation between measurements performed at different angles can give further information for the interpretation of the peaks 4.- E L E C T R O N E M I S S I O N T E C H N I Q U E S X-Ray Photoelectron Spectroscopy (XPS) and Auger Electron Spectroscopy (AES) techniques are based in the measurement of the energy distribution of electrons ejected from a material [ 1,3,13]. As the electronic energy levels of each orbital are discrete and are different for the same orbital in different atoms, the measurement of the electron binding energies can provide atomic or chemical state identification, or both. Moreover, since the energy of the concerned electrons falls in the range where they can travel only very short distances without being inelastically scattered, all the techniques are very surface sensitive. Profiling analysis can also been obtained by combining ion-beam sputtering with the continuous analysis of the electron ejected as the surface moves into the sample.

227

A. Pdrez-Rodr[guez et al. / Microelectronic Engineering 40 (1998) 223-237

4.1. X-Ray Photoelectron Spectroscopy In this technique, an X-ray photon of sufficiently high energy (hv), usually 1486.6 eV or 1256.6 eV (corresponding to Alkc~ or Mgkc~ radiation lines), is used to ionise an atom, producing an ejected electron coming from the core levels. The kinetic energy of this electron (KE) can be related with its binding energy (BE) to the atom concerned by the Einstein photoelectric law: KE=hv-BE. Since hv is know, a measurement of KE determines BE. As core-level electrons of an atom have energies nearly independent of the chemical species in which the atom is bound, the identification of core level BEs provides unique signatures of the elements. All the elements in the periodic table can be identified, except H and He, which have no core levels. Moreover, small shifts in the BEs provide additional chemical state information and the relative concentrations of the different elements present can be determined from the peak intensities. Due to the loss of energy in the inelastic scattering process, only electrons coming from atoms near the surface, in the range between 2 atomic layers to 15-10 layers, can be detected. The probed depth can also be made smaller for any given XPS peak and material by detecting at grazing emission angle 0 (angle resolved XPS technique). So, XPS is a very surface sensitivity technique. This characteristic, combined with quantitative and chemical state analysis and the tact that it is a non destructive technique, have made XPS the most broadly applicable surface analysis technique when no spatial resolution is required (greater than 150 gm). The most appropriate applications are: evaluation of surface material processing steps, as cleaning procedures, plasma etching, thermal oxidation, silicide thin-film formation, evaluation of thin-film coatings, surface composition differences for alloys, oxidation, corrosion .... Nevertheless, it is not suitable for trace analysis because the absolute sensitivity is only between 0.01-0.3% at., depending on the elements. In some applications where information from deeper regions is required, or a contaminant surface is covering the surface of interest, an argon ion sputtering system is used to strip the surface while taking data, or prior to taking data. Although this system can be used to depths of I.tm, it is most effective over much shorter distances (hundreds or thousands of ,~,) because sputtering introduces artefacts that get worse as the sputtering depth increases.

c/s 1.500

)

110

i

|

105

100

m

|

l~

95

I~O

Binding Energy (eV)

Figure 5.Si2p XPS spectrum from a Si(100) surface. The sensitivity of the XPS spectrum to the surface chemical state can be seen in Fig. 5, where the Si2p spectrum from a (100) Si surface is plotted. The spectrum shows two clear contributions, which correspond to Si in both Si and SiO2 chemical coordination. The significant Si02 signal in the spectrum (peak at higher energies) is due to the native oxide layer (thickness of about 2-4 nm), which drastically depends on surface processing and characteristics.

4.2. Auger Electron Spectroscopy AES is closely related to XPS. The hole left in a core level after the XPS process, is filled by an electron dropping from a less tightly bound level. The energy released can be used to eject another electron, the Auger electron, whose energy depends only on the energy levels involved and not on whatever made the initial core hole. This allows, unlike XPS, to use an electron beam as a probe of the sample surface. However, since all the energy levels involved are either core or valence levels, the type of information supplied, like XPS, is elemental identification from peak positions and chemical state information from chemical shift and line shapes. The advantages of using AES is to improve both lateral (as low as 300 ,~., if using field emission electron beams) and depth resolutions (5 to 100 ]k) and to increase the speed to collect information. It has also a good absolute detectability, as low as 100 ppm. When the incident beam is scanned over the sample (Scanning Auger Microprobe, SAM) mapping at high spatial resolution is obtained. On the contrary,

228

A. P~rez-Rodriguez et al. / Microelectronic Engineering 40 (1998) 223-237

the major disadvantages to use electrons are the beam damage and, in the case of analysing insulators, sample charging related problems. As in the case of XPS, it can be combined with ion-beam sputtering to remove material from the surface and to continue to monitor the composition and chemistry of the remaining surface as this surface moves into the sample. Whereas the extraction from qualitative information from the Auger spectra is quite simple, quantitative analysis to obtain the concentration of a particular species is very hard to implement, and the data are subject to several artefacts or interferences which might lead to misinterpretation.

Log. Intensity (a..u.) /

loSL:-

F

1el-

10:

10~

5.- X-RAY D I F F R A C T I O N X-Ray Diffraction (XRD) constitutes a powerful non-destructive technique for the structural analysis of crystalline and polycrystalline layers. It is based on the diffraction by the crystalline planes of a collimated beam of X-rays with wavelengths typically in the range 0.7-2 ,~. The XRD spectrum is usually obtained by measuring the diffracted intensity as a function of the diffraction angle 20 (angle between the incident and diffracted beams) and the orientation of the specimen. Different maxima, corresponding to the contribution of the different crystalline planes in the crystal, are obtained when constructive interference of diffracted X- rays occurs according to the Braggs law, ~, = 2 d sin 0, being 3, the X-rays wavelength and d the spacing between the atomic planes in the crystal. The technique is extremely sensitive to variations of d determined by strains in the crystal. Strain values in the range 10-5-10-6 are routinely measured, and values as low as 10-8 can be detected. In general XRD does not provide spatial resolution, being the measured area of several mm 2, although for special applications spatial resolution about 10 ~m can be achieved with a microfocus source for films thick enough (1 I.tm). The penetration depth of the X rays is determined by the absorption length, which for Si has a value of 66 ~tm at ~, = 1.54 A. These characteristics give a strong interest to XRD for the non-destructive analysis of epitaxial films and heterostructures [14]. This is usually performed by the analysis of the rocking curves, which are obtained by fixing the diffracted angle and rotating the sample around a certain orientation. Fig. 6 shows a (004) reflection measured from a SiGexCy

10 0

33.8

,

=

I

34.0

34.2

,

,

I

34.4

=

I

34.6

=

34.8

Angle (seconds)

Figure 6. XRD rocking curve from a SiGexCy layer on Si ( x=23%, y=2%). film on Si (x=23%, y=2%). SiCy and SiGexCy are systems which have received a strong interest in the last years [15], due to the possibility to incorporate a significant fraction of carbon in substitutional sites in the crystalline network, which allows to partially compensate the compressive strain in the SiGex crystal. By this way, thicker layers with higher Ge content could be achieved without relaxation. The possibility to obtain high crystalline quality layers with equilibrium lattice constant smaller than that of Si opens up additional band gap engineering options in the SiGe area The spectrum in Fig. 6 shows two intense peaks, which correspond to the diffraction of the (004) planes from the Si substrate and the epitaxial film. The angular separation between both peaks is related to the difference in the lattice constant of the film and the substrate. By combining spectra measured with different orientations of the sample, the lattice constants of the film parallel and perpendicular to the growth plane can be determined. Then, the lattice constant of the corresponding strain free alloy can be obtained from a straightforward application of elasticity theory.

A. Pdrez-Rodrfguez et al. / Microelectronic Engineering 40 (1998) 223-237

Fig. 6 is also characterised by the presence of interference fringes, in addition to the diffracted peaks. These are determined by the high crystalline quality of the film, which is fully strained. Assuming a uniform film, the film thickness t can be derived from the differential form of Bragg's law [16]: t = ~, sin ~/(Am sin 20), where e is the angle between the diffraction plane and the surface plane. On the other hand, strain relaxation is accompanied by the generation of misfit dislocations which degrade the crystalline quality of the layer. This determines a decrease on the intensity of the XRD peaks, as well as the disappearance of the interference fringes The measurement of the strain free lattice constant allows to determine the composition of the layer. In general, the lattice parameter of a crystalline alloy is given by the linear interpolation of lattice constants from the parent semiconductors, according to Vegard's law [17]. For SiGex, some deviation from this simple law has been observed, mainly for values of x about 0.5 [18]. In this case, the experimental data have been fitted with the expression: asiBe = x abe + (l-x) asi - 0.00436 x 3 + 0.03265 x2 - 0.02829 x, being all lattice parameters in angstroms [14]. This can also be used to determine the amount of substitutional carbon in SiCy and SiGe×Cy alloys. However, the validity of Vegard's law in these systems is still not clear, and different authors have reported theoretical calculations which deviate considerably from this law [19,20] Nevertheless, XRD measurements on SiGeC films grown by CVD (Chemical Vapour Deposition) show that the lattice constant tends to follow Vegard's taw [21]. This could be related to the influence of not substitutional carbon on the average lattice constant in the film. Although further analysis is still needed to clarify this, these data suggest that, in spite of the theoretical calculations, X-ray diffraction can also be used for the chemical analysis of these films. XRD is also a very powerful technique for the analysis of polycrystalline films. Data related to texture and preferred crystalline orientations, strain, gx~ain size, and crystalline quality of the film can be obtained from the XRD spectra. In principle, and for a given orientation of the film, diffraction will occur from any crystallite with the proper orientation satisfying diffraction conditions. If the crystals are randomly oriented, peaks corresponding to the different planes will appear in the XRD spectrum, and the relative intensity of the different peaks is

229

characteristic of each material. The amount of preferred orientation can be estimated in a straightforward way by comparing the integrated intensities of the different peaks (after correcting from geometrical factors) from the calibrated data. The size of the crystalline grains can also be derived in a straightforward way from the broadening of the peaks, according to the Scherrer's formula: L _= M(A20 cos0), where A20 is the width of the peak. However, this is only valid in the absence of inhomogeneous strain, as strain gradients in the film will also determine a broadening of the peak. Moreover, the peak width is also affected by the presence of extended defects in the crystals as dislocations or stacking faults. An interesting way for strain and stress measurements in polycrystalline films is the sin 2 ~t method [22,23]. In this method, the average stress associated with a certain grain orientation in the film is obtained by performing measurements with different values of the Euler angles (% ~). Assuming linear stress-strain relations and isotropic material for each crystallographic orientation, the XRD peaks show a linear dependence on the average biaxial stress components and sin 2 ~. On the other hand, analysis of the in depth distribution of the XRD data can be obtained by performing the measurements on layers etched at different depths. This has been performed by chemical etching [23], and the correlation of the data allowed to determine the stress and texture profile in polycrystalline Si layers. Disadvantages of this approach are the increase in measuring time and its destructive character. XRD profiling can also be achieved by fitting the spectra with their theorerical simulation. This has been performed for single crystal ion ,implanted layers and epitaxial films to determine the strain profile in the structure [14,24]. Modelling of strain (usually based in Vegard's law) allows to relate the strain profile to a concentration profile. Finally, XRD is extensively used for phase identification. This has been also applied in Si technology for the identification and structural assesment of phases formed by ion beam synthesis, as SiC, AIN and CoSi2. The analysis of the diffracted peaks in the spectra from the formed phases and the host crystal has allowed to determine the existence in some cases of preferential crystal orientation of the grains in relation with the matrix (SIC, AIN) [25,26]. This strongly depends on the

230

A. Pdrez-Rodrfguez et al. / Microelectronic Engineering 40 (1998) 223-237

implantation conditions, and is likely related to the occurrence of a topotactic transformation of the implanted crystal [25]. 6.VIBRATIONAL RAMAN AND FTIR

SPECTROSCOPIES:

Raman and Infrared spectroscopies are vibrational techniques well suited for the characterisation of structures and processes in Si technology. Firstly, Raman scattering [27-29] provides information on the structural assessment of crystalline and polycrystalline layers, in a way similar to XRD. In addition, Raman scattering is also sensitive to the structure of amorphous Si, which is important for the analysis of recrystallisation processes as well as the study of the amorphous tissue in nanocrystalline systems. Fourier Transform Infrared (FTIR) spectroscopy complements well this technique for the structural analysis of dielectric layers in Si technology (as silicon nitride and silicon oxide, which have very low Raman efficiency) [30]. 6.1.- Raman spectroscopy Raman spectroscopy is based on the inelastic scattering of photons with elemental excitations in the material. For first order processes, the main interactions correspond to the creation (Stokes process) or annihilation (Antistokes process) of a phonon. Due to the conservation of energy, the wavenumber of the scattering photon is shifted in relation to that of the incident photon in an amount which corresponds to the energy of the phonon. Conservation of momentum implies that only zone centre phonons can be Raman active [27-29]. In Si, Raman spectra are usually measured in backscattering geometry, for which the penetration depth of the scattered volume is determined by optical absorption. Assuming the same value of optical absorption for the incident and scattered light, the penetration depth has been estimated for the different visible lines of an Ar ÷ laser, being the obtained values between 300 nm (L= 457.9 nm) and 800 nm (X=- 514 nm). In MicroRaman configuration [31,32], excitation of the sample and collection of scattered ions is performed through an optical microscope. This allows to achieve a high lateral resolution. According to Rayleigh diffraction ctriteria, the diameter of the light spot on the sample is given by 1.22 L/NA, being NA the numerical

aperture of the microscope objective. Using a X100 objective with NA=0.95, the size spot in the sample can be as low as 0.6 gm 0~=457.9 nm). Thus, submicron lateral resolution is achieved. The first order Raman spectrum of crystalline Si has a single line at about 520 cm -I, which corresponds to a triply degenerated phonon. This line has a lorentzian shape, with a Full Width at Half Maximum (FWHM) of about 3 cm -l. The position and shape of this line is sensitive to the presence of strain, structural defects and damage in the lattice, temperature, chemical composition and concentration of carriers. Strain in the Si lattice determines a shift of the Raman line which, in the elastic regime, depends linearly on the strain value [33]. For tensile strains, the increase of the lattice parameter determines a decrease (red shift) of the vibrational modes. Compressive strains give rise to a decrease of the lattice parameter and a consequent increase of the frequency of the vibrational mode is to be expected. For biaxial stress, stress is given by: e -- 250 MPa/cm -j A~. This determines a relatively poor resolution in the stress measurement (about 10 MPa, for shift resolution of about 0.05 cm-l). The presence of stress gradients in the scattering volume would also determine a broadening of the peak. Damage in the lattice leads to a decrease of the intensity of the first order modes, related to the breaking of bonds and atomic disorder and, hence, a decrease of the Raman polarizability tensors. The decrease in the intensity of the modes allows the quantification of the residual damage in ion implanted layers, determining the threshold dose for amorphization of the layer and the degree of damage recovery by annealing [31,34]. Besides, structural defects determine the appearance of disorder effects, which are related to the breaking of the conservation of momentum due to spatial confinement of phonons by the defects. This has been modelled by introducing a correlation length L, which corresponds to the average size of the crystalline domains in which phonons are confined [35]. Fig. 7 shows the spectra simulated for different values of L and assuming spherical confinement: for L < 200 A, a red shift and asymmetric broadening of the peak is observed, which increases as L decreases. Different authors have reported a modification of this method to take into account the possible existence of strain effects in the spectra [36,37]. These models usually assume uniform strain in the scattering volume.

231

A. Pdrez-Rodr(guez et al. / Microelectronic Engineering 40 (19981 223-'237

Raman intensity (u. a.)

Raman intensity (u. a.) 1.2

substrate

1.0

0.8

Si-Si

(dl

0.6

0.4 411

0.2

Si-Ge Ge-Ge

0°0

505

~

51o Wavenumber (cm"I)

20 . . . . . . . .

21111

Figure 7.First order Si Raman spectrum simulated assuming spherical confinement: (a) L= 60 /~, (b)

i

. . . . . . . . .

31111

I

400

. . . . . .

, 1 , I , , ,

51111

. . . . . .

I

. . . .

600

Wavenumber (cm "I)

L= 100 A., (c) L= 200 ]k. They have been applied to the analysis of nanostructure and disordered systems as porous [36], nanocrystalline [38] or highly damaged Si films [37]. However, we have to bear in mind the simplicity of the used model, in which strain is assumed not to affect the shape of the Raman line and both strain and disorder effects are assumed as uniform in the scattering volume. Besides, Raman scattering offers an interesting alternative to XRD for the characterisation of heteroepitaxial layers. For SiGex alloys, the first order Raman spectrum shows three main lines, ;related to the Si-Si (around 500 cml), Si-Ge (around 400 cm l ) and Ge-Ge (around 300 cm 1) vibrational modes (fig. 8). These modes show a linear dependence on both the chemical composition (related to mass disorder and microscopic strain effects) and average strain, and their fitting gives the following relationships [39]: O)si_si (cm -1) = 520 - 68 x - 830 c 6Osi_ce(cm -l) = 400.5 + 14.2 x -575 o COoe-Oe(cm 1) = 282.5 + 16 x - 384 o, being o the strain parallel to the substrate (o = (aparasiGe)/asiGe).Similar relationships have been reported

Figure 8. First order Raman spectrum from a SiGe0.3 iayer on Si, showing the Si-.Si, Si-Ge and Ge-Ge vibrational modes.The peak at 520 cm -I is due to the substrate (Si) contribution. for SiGeC [21] and SiGeB [40] ternary alloys, where an additional term (related to the dependence of the modes on the carbon or boron content) is added. For SiGeB, an additional complexity arises from the high doping of the layers (>_ 1019 cm-3). This causes a Fano-like interaction, and the fitting of the first order Si line allows to determine the Fano parameters, which strongly depend on the concentration of carriers. For amorphous materials, the lack of long-range order determines a breakdown of the momentum conservation rule, and all phonons become Raman active. Then, the Raman spectrum reflects the density of vibrational states. For amorphous Si, the spectrum shows 4 broad bands which are centred at about 150 cm 4 (TA), 310 cm -1 (LA), 380 cm -I (LO) and 480 cm -1 (TO). It has been shown that the width of the TO peak is a good measure of local order, being related to parameters as the bond angle distortion in the network [41]. Moreover, the quantification of the Raman efficiencies from both amorphous and crystalline phases allows to

232

A. P~rez-Rodrfguez et al. / Microelectronic Engineering 40 (1998) 223-237

determine the crystalline fraction in partially amorphous systems [38]. One of the main advantages of Raman scattering in Si integrated circuits technology over other structural characterisation techniques as XRD is the high lateral resolution achieved by the Raman Microprobe. This has been applied for the stress assessment of the active region in CMOS devices [42], as well as for the measurement of self-heating effects in integrated transistors [43]. In the last case, the temperature in the active region of the device under electrical operation can be measured from both the Stockes/Antistockes intensity ratio or from the shift of the first order Raman line, which is related to the increasing contribution of anarmonic effects w i t h the temperature [44]. MicroRaman measurements have also been used for the depth profile analysis of complex structures as Silicon-OnInsulator, by performing the measurements on the surface of low angle bevelled samples [45]. For thick enough layers (thickness higher than 5-10 I.tm), depth profiling can also be directly made by moving the probe through the cross-section of the layer [23]. Finally, MicroRaman has also been used for the measurement of the thickness and uniformity of thin films on Si [46]. The thickness of the film can be determined from the decrease of the intensity of the Si substrate Raman line, due to light absorption in the film. By this way, CoSi2 films with thickness as low as 1-2 nm have been detected, and the technique can be used for the measurement of silicide films in the thickness range 10 nm to 100 nm with an uncertainty below 10% 6.2.- F T I R

Fig. 9 shows the FTIR transmission spectrum (angle of incidence 30 °) from SiO2 films implanted with different doses of Ar ÷ ions (from 3x10 lz cm 2 (sample 1) to 3x1014 cm "2 (sample 5). Sample 6 is not implanted) [47]. The spectra are characterised by the presence of 4 transverse optical (TO) and longitudinal optical (LO) vibrational modes of the Si-O-Si unit, which are sensitive to the stoichiometry, strain and structural features in the film. The higher absorption band corresponds to the asymmetric stretching motion of the Si-O-Si unit (TO3), located around 1070 cm 1 Stoichometry of SiOx layers (x<2) has been studied from the broadening and shift of the TO3 vibrational mode towards lower wavenumbers [48]. The position of this mode is also strain dependent.

Absorbance .36

TO3

6

.26

.t6 .06

6

5

L~ ~ ~]

0~

~A

~

-.04 . . . . . . . . 1450 1187.5

,

,

i

i

925,

q

i

,

,

|

,

i

662.5

Wavenumbers (era"z)

Figure 9. FTIR transmission spectrum (angle of incidence 30 °) from SiO2 films implanted with different doses of Ar ÷ ions, from 3x1012 cm -2 (sample 1) to 3x1014 cm -2 (sample 5). Sample 6 is not implanted) This is due to the dependence of the vibrational frequency on the average Si-O-Si bond angle [49]. Distortions in the Si-O-Si bond angle determine the broadening and low frequency shift of the TO3 mode from the implanted oxides, in Fig. 9. In a more general way, changes in the IR spectrum are related to changes of the dielectric function, which reveal structural differences in the oxide network. In this sense, disorder effects in oxide layers have been related to an increase in the contribution of the TO4LO4 vibrational modes [50]. The analysis of the dielectric layers by FTIR has also to take into account the dependence of the spectra on the geometrical arrangement of the structure. This is specially important for complex multilayer structures, as Silicon on Insulator (SOl) [51]. Fig. 10 shows the features of the T03 mode from SO1 wafers as a function of the thickness of the top Si layer (buried oxide thickness 390 nm), together with the theoretical fitting assuming the dielect5ric function of a thermal SiOz.[52] In this case, the interpretation of the spectra needs for the modelling of the optical behaviour of the structure. This has to take into account multiple reflection and

233

A. P~rez-Rodrfguez et al. / Microelectronic Engineering 40 (1998) 223-237

~=I°'°I"---

IR intensity (a.u.)

m-lm

'~ ~ / ....

Experimental Fitted spectra

I

1080 I I

i

•

•

t ~---- I B I E C

•

75 55~B._

900

z o>-

z

800

700

600

Wavenumber (cm "1)

.

g~

m'-.

1

1000

Si THICKNESS

2000

(A)

Figure 10.Theoretical and experimental features (peak position, FWHM, absorption intensity) of the TO3 mode from the buried oxide of SOI structures versus thickness of the top Si layer-. transmission events at the different interfaces, as well as the parameters of the dielectric function of the different layers [52]. For complex layers (formed by more than one constituent), effective medium approximation models are used. Hence, the fitting of the spectra can provide information on the nature, structure, dimensions and composition of the different layers and interfaces. FTIR also allows the accurate quantification of concentration of impurities in Si wafers, as interstitial oxygen or substitutional carbon in the Si lattice [30]. This is related to the measurement of the integral intensity of the respective local vibrational modes (LVM), as their dependence on the impurity content has been precisely calibrated from well established standards. FTIR is specially interesting for the investigation of carbon in silicon: Mel6ndezLira et al [53] have shown the validity of the calibration of the intensity of the carbon mode at about 605 cm -1 for carbon contents much higher than the solubility of carbon in Si (of the order of 1017 cm-3). These authors have measured SiCy layers made by ion implantation and laser annealing with effective atomic carbon ratios up to 0.5%. According

Figure 11. SiC absorption band from an amorphous layer as-implanted and after partial (thermal) and full (IBIEC) recrystallisation. to these data, FTIR can be used to quantify the concentration of substitutional carbon in epitaxial films. Besides, the FTIR spectrum is also very sensitive to the presence of SiC, due to the absorption band at about 800 cm -I related to the Si-C stretching mode. This allows to observe precipitation of SiC when relaxation of the SiGexCy layer occurs. Moreover, the shape and position of this band depends on the crystalline nature of SiC: for crystalline SiC; the band has a lorentzian shape, with a FWHM of the order of 50 cm -]. For amorphous SiC, the band has a gaussian shape, being centred at lower wavenumbers (in the range 700-800 cm -1) and with higher values of FWHM., Fig.ll shows the evolution of the SiC absorption band from an amorphous layer after partial recrystallisation (thermal annealing at 700°C) and complete recrystallisation by an IBIEC (Ion Beam Induced Epitaxial Crystallisation) process [54]. The fitting of the band with gaussian and lorentzian curves allows to quantify the fraction of crystalline and amorphous SiC in the processed film. In general, quantification from FTIR requires the knowledge of reflection and transmission at the back surface of the wafer. This is specially difficult for not polished back surfaces, which is the most usual case. To avoid this, FTIR measurements are performed using a reference sample with an identical back surface (usually a not processed substrate

234

A. P~rez-Rodr[guez et al. / Microelectronic Engineering 40 (1998) 223-237

wafer) and the spectra from the processed samples are compared with those measured under the same conditions on the reference ones. Then, the accuracy of these measurements is conditioned by the availability of a suitable reference. FTIR spectrometers can also work in combination with an Infrared microscope, which allows to achieve lateral resolution down to about 20 gin. For very thin layers, grazing angle incidence measurements allow to reduce the probing depth. For surface and interface analysis, the Attenuated Total Reflection (ATR) configuration can be used. In this case, reflections in the studied surface occur along all the probed surface, as the light is guided by the ATR crystal in the surface of the sample, and the spectrum can become sensitive to the presence of even submonolayers on the sample. However, the quality of the interface between the ATR crystal and the sample is critical, and quantification of the results is difficult. 7.- A T O M I C F O R C E M I C R O S C O P Y Among all microscopic techniques, Scanning Tunnelling Microscopy (STM) and Atomic Force Microscopy (AFM, also known as Scanning Force Microscopy) have unique characteristics for surface roughness measurements at atomic scale resolution. The basic structure of the AFM microscope is a flexible cantilever with a sharp tip. When the tip approaches the surface of the sample (at a distance of few A), Van der Waals repulsive forces between the atoms of the tip and those of the surface cause the deflection of the cantilever, being this deflection function of the distance between the tip and the sample. A piezoelectric transducer is used to scan the tip over the surface of the sample, and the deflection of the cantilever is usually measured by a laser beam reflected on it. The main operation mode of the AFM microscope isto use a feedback loop to keep a constant distance between the tip and the surface. By monitoring the position of the scanner in three dimensions, an image of the surface is directly obtained, achieving a depth resolution of 0.01 nm, with lateral resolution of 0.1 nm. In relation with the STM (which measures the tunnel current between the tip and the surface), AFM has the advantage that it does not depend on the electronic or optical properties of the sample, and almost any solid surface can be studied by the AFM. This microscope was proposed by Binning, Quate

Figure 12. SEM image of the AFM cantilever. and Gerber in 1985 to avoid the problems of STM to study insulators [55]. The instrument also allows the study of surfaces in air, which greatly simplifies sample preparation (almost no sample preparation being required). On the other side, the dependence of the deflection of the cantilever on the tip-tosurface distance follows a power law, which is weaker than the exponential dependence of the tunnelling current on the distance in the STM. Thus, several atoms on the AFM tip will interact with several atoms on the surface, and only with an unusual sharp tip and an extremely flat surface lateral resolution becomes truly atomic, being normally of about 1 nm. Fig. 12 shows a Scanning Electron Microscope (SEM) image of an AFM cantilever with an integrated pyramidal tip. The pyramid is usually obtained by CVD deposition of Si3N4 on an etch pit in (100) Si bounded by (111) faces. In some cases Si tips are more suitable, due to their higher stiffness and lower sensitivity to moisture, which might cause sticking problems. The knowledge of the topography of the tip and the mechanical properties of the cantilever (as its force constant) are important for the correct interpretation of the AFM images. For a given feature in the surface of the sample, the microscope gives an image determined by a convolution of the shape of the tip and that of the feature. When the tip is much sharper than the feature, the image corresponds to the real shape of the feature. However, if this requirement is not satisfied the image is affected by the shape of the tip

A. P~rez-Rodrfguez et al. / Microelectronic Engineering 40 (1998) 223-237

J,

,~

P' I +

• "~' I

i+" ¢

•

~,~*t.~ . , - . t .

it, .,|,,I 7,i,

~.

h

"¢

r,i,+~

, • at~.'~

"| ' ,

=+

. p

+,,

, " .~

"I

""

,..%,:

235

Young's modulus measurements based on the deflection of an end-loaded cantilever beam by the AFM tip have been performed on micromachined polysilicon rims and ultrathin 13-SIC cantilever beams. This method combines a very high load resolution with a nanometric precision in the measurement of the cantilever deflection. Once the system has been calibrated, the technique provides a straightforward and simple method, which enables accurate, easy and quick measurements of the mechanical properties of the films. REFERENCES

• i t ; ~ ,,~ :

2

1

1.

10~). 0tiff r l ~ / d l v

Figure 13. AFM image of the surface of an etched ~SiC layer on Si. and tip imaging can occur. Other artefact characteristic of AFM imaging is the deformation of the surface which might occur when measuring soft materials. Applications of AFM include measurement of surface roughness, study of cleaning and surface processing and analysis of first stages of deposition processes. In fig. 13, AFM is used to study the roughness on the surface of an etched SiC layer [7]. This layer was obtained by carbon ion implantation in Si, and roughness in this case is critical for the fabrication of a SiC on Insulator (SiCOI) structure by thermal bonding of the surface on a SiO2 film [56]. According to fig, 13, this surface has an RMS roughness value of 6-7 nm.. AFM can also been applied for the measurement of mechanical properties of micromachined test structures with nanometric resolution [57]. Miniaturisation of microelectronic devices and micro-electro-mechanical systems requires for accurate methods to test and evaluate the mechanical properties of the films in the structures, which strongly depend on their fabrication process. Accurate measurements of these properties can only be achieved by miniaturising the test probe to a size approaching the film thickness and the microstructure dimensions. In this context, the nanometric spatial resolution achieved by the AFM is interesting for the micromechanical assesment of test structures.

C.R. Brundle, C.A. Evans, S. Wilson (eds), Encyclopedia of Materials Characterisation: Surfaces, Interfaces, Thin Films. ButterworthHeinemann, Stoneham, 1992. 2. T.E. Jenkins, T.Jenkins (eds), Semiconductor Science: Growth and Charcaterization, Prentice Hall, 1995. 3. D.J. O'Connor, B.A. Sexton, R.St.C. Smart (eds), Surface Analysis Methods in Materials Science, Springer-Verlag, Berlin, Heidelberg, 1992. 4. D. Briggs, M.P.Seah (eds), Practical Surface Analysis. Vol. II: Ion and Neutral Spectroscopies, Wiley, New York, 1992. 5. M. L6pez, B. Garrido, C. Bonafos, O. Gonz~ilez-Varona, A. P6rez-Rodrfguez, R. Rodrfguez, P. Ruterana, J.R. Morante, Mater. Res. Soc. Symp. Proc., 486 (1998) (in press). 6. J.F. Ziegler, J.P. Biersack, U. Littmark, The Stopping and Range of Ions in Solids, Vol. I, Pergamon, New York, 1985 7. C. Serre, A. Pdrez-Rodrfguez, A. RomanoRodrfguez, L. Calvo-Barrio, J.R. Morante, J. Esteve, M.C. Acero, W. Skorupa, R. K6gler, J. Eiectrochem. Soc., 144 (1997) 2211. 8. W.K. Chu, J.M. Mayer, N.A. Nicolet, Backscattering Spectroscopy, Academic Press, New York, 1978. 9. J.R. Bird, J.S. Williams, Ion Beams for Materials Analysis, Academic Press, Australia, 1989. 10. A. P6rez-Rodrfguez, A. Romano-Rodrfguez, C. Serre, L. Calvo-Barrio, R. Cabezas, J.R. Morante, J. Calderer, H. Reuther, W. Skorupa, Nucl. Instr. and Meth. in Phys. Res. B, 120 (1996) 151.

236

A. P~rez-Rodriguez et al. / Microelectronic Engineering 40 (1998) 223-237

11. R. K6gler, H. Reuther, M. Voelskow, W. Skorupa, A. Romano-Rodrfguez. A. P6rezRodrfguez, C. Serre, L. Calvo-Barrio, J.R. Morante, in "Proceedings of the 1lth Int. Conf. on Ion Implantation Technology", E. Ishida et al (eds), The Institute of Electrical and Electronics Engineering Inc., 1997, p. 709. 12. N. Frangis, J. Van Landuyt, M.G. Grimaldi, L. Calcagno, Nucl. Instr. and Meth. in Phys. Res. B, 120 (1996) 186. 13. D. Briggs, M.P.Seah (eds), Practical Surface Analysis. Vol. I: Auger . and X-Ray Photoelectron Spectroscopies, Wiley, New York, 1983. 14. P.F. Fewster, Semicond. Sci. Technol., 8 (1993) 1915. 15. S.C. Jain, H.J. Osten, B. Dietrich, H. Rticker, Semicond. Sci. Technol., 10 (1995), 1289. 16. W.J. Bartels, W.J. Nijman, J. Crystal Growth, 44 (1979) 518. 17. L. Vegard, Z. Phys., 5 (1921) 17. 18. J.P. Dismukes, L. Ekstrom, R.J. Paff, J. Phys. Chem., 68 (1964) 3021. 19. P.C. Kelires, Phys. Rev. Lett., 75 (1995) 1114. 20. D.H. Whysong, J. Undergrad. Res. Phys., 12 (1994) 51. 21. M. Mel6ndez-Lira, J. Men6ndez, W. Windl, O.F. Sankey, G.S. Spencer, S. Sego, R.B. Culbertson, A.E. Bair, T.L. Alford, Phys. Rev. B, 54 (1996) 12866. 22. D. Rats, L. Bimbault, L. Vandenbulcke, R. Herbin, K.F. Badawi, J. Appl. Phys., 78 (1995) 4994. 23. J.R. Morante, J. Samitier, M.S. Benrakkad, Microelectronics Journal, 27 (1996). 24. A. Pesek, Appl. Phys. A, 58 (1994) 141. 25. A. Romano-Rodrfguez, C. Serre, L. CalvoBarrio, A. P6rez-Rodrfguez, J.R. Morante, R. K6gler, W. Skorupa, Mater. Sci. Eng., B36 (1996) 282. 26. L. Calvo-Barrio, A. P6rez-Rodrfguez, A. Romano-Rodrfguez, J.R. Morante, J. Montserrat, Nucl. Instr. and Meth. in Phys. Res. B, 84 (1994) 214 27. M. Cardona, G. Gtintherodt (eds), Light Scattering in Solids I to V, Springer, Berlin, Heidelberg, 1975 to 1989. 28. B. Pr6vot, J. Wagner, Prog. Crystal Growth and Charact., 22 (1991) 245.

29. F.H. Pollak, R. Tsu, in "Spectroscopic Characterization Techniques for Semiconductor Technology", Proc. SPIE 452 (1983), 26. 30. J.R. Ferraro, K. Krishnan (eds), Practical Fourier Transform Infrared Spectroscopy, Academic Press, Sanm Diego, 1990. 31. S. Nakashima, M. Hangyo, IEEE J. Quantum Elect., 25 (1989), 965. 32. T. Jawhari, in "Handbook of Advanced Materials Testing", N.P. Cheremisinoff, P.N. Cheremisinoff (eds), Marcel Dekker, Mew York, 1995, p. 105. 33. E. Anastassakis, in "Physical Problems in Microelectronics", J. Kassabov (ed.) (1985), 128. 34. A. P6rez-Rodrfguez, Y. Pacaud, L. CalvoBarrio, C. Serre, W. Skorupa, J.R. Morante, J. Electron. Mater., 25 (1996) 541. 35. P.M. Fauchet, I.H. Campbell, CRC Crit. Rev. Solid State Mater. Sci., 14 (1988), $79. 36. M. Yang, D. Huang, P. Hao, F. Zhang, X. Hou, X. Wang, J. Appl. Phys., 75 (1994) 651. 37. J. Macfa, E. Martin, A. P6rez-Rodrfguez, J. Jim6nez, J.R. Morante, B. Aspar, J. Margail, J. Appl. Phys.; 82 (1997) 3730. 38. A. Achiq, R. Rizk, F. Gourbilleau, R. Madelon, B. Garrido, A. P6rez-Rodrfguez, J.R. Morante, J. Appl. Phys., 83 (1998) (in press). 39. J.C. Tsang, P.M. Mooney, F. Dacol, J.O. Chu, J,. Appl. Phys., 75 (1994) 8098. 40. A. P6rez-Rodriguez, A. Romano-Rodrfguez, R. Cabezas, J.R. Morante, T. Jawhari, C.E. Hunt, J. Appl. Phys., 80 (1996) 5736. 41. J. Bullot, M.P. Schmidt, Phys. Stat. Sol. (b), 143 (1987) 345. 42. I. De Wolf, Semicond. Sci. Technol., 11 (1996) 139. 43. O. Gonz~ilez-Varona, A. P6rez-Rodrfguez, J.R. Morante, in "Actas 1a Conf. de Dispositivos Electr6nicos", R. Alcubilla J, Pons (eds.), Univ. Polit~cnica de Catalunya, Barcelona, 1997, p. 301 44. M. Balkanski, R.F. Wallis, E. Haro, Phys. Rev. B, 28 (1983) 1928. 45. E. Martin, J. Jim6nez, A.P6rez-Rodrfguez, J.R. Morante, Mater. Letters, 15 (1992) 122. 46. A. P6rez-Rodriguez, E. Roca, T. Jawhari, J.R. Morante, R.J. Schreutelkamp, Thin Solid Films, 251 (1994) 45. 47. B. Garrido, PhD Thesis, Universitat de Barcelona, 1993.

A. P~rez-Rodr[guez et al. / Microelectronic Engineering 40 (1998) 223-237

48. G. Harbeke, E.F. Steigmeier, P. Hemment, K.J. Reeson, L. Jastrzebski, Semiocnd. Sci. Technol., 2 (1987) 687. 49. G. Lukovsky, M.J. Manitini, J.K. Srivastava, E.A. Irene, J. Vac. Sci. Technol. B, 5 (1987) 530. 50. P. Lange, J. Appl. Phys., 66 (1989) 201. 51. J. Samitier, S. Martinez, A. El Hassani, A. P6rez-Rodrfguez, J.R. Morante, Appl. Surf. Sci., 63 (1993) 312. 52. M.L. Naiman, C.T. Kirk, R.J. Aucoin, F.L. Terry, P.W. Wyatt, S . D . Senturia, J. Electrochem. Soc., 131 (1984)637. 53. M. Mel6ndez-Lira, J. Men6ndez, K.M. Kramer, M.O. Thompson, N. Cave, R. Liu, J.W. Christiansen, N.D. Theodore, J.J. Candelaria, J. Appl. Phys., 82 (1997) 4246. 54. C. Serre, L. Calvo-Barrio, A. P6rez-Rodr/guez, A. Romano-Rodrfguez, J.R. Morante, Y. Pacaud, R. K6gler, V. Heera, W. Skorupa, J, Appl. Phys., 79 (1996) 6907 55. G. Binnig, C.F. Quate, C. Gerber, Phys. Rev. Lett., 54 (1986) 930. 56. C. Serre, A. Romano-Rodrfguez, A. P6rezRodrfguez, J.R. Morante, L. Fonseca, J.A. Plaza, M.C. Acero, E-MRS 1998 Spring Meeting (Strasbourg, France, June 1998). 57. C. Serre, P. Gorostiza, A. P6rez-Rodrfguez, F. Sanz, J.R. Morante, Sensors and Actuators A (1998) in press.

237