Structural and Electrical Characterizations of Yttrium Oxide Films after ...

Journal of The Electrochemical Society, 152 共12兲 F217-F225 共2005兲

F217

0013-4651/2005/152共12兲/F217/9/$7.00 © The Electrochemical Society, Inc.

Structural and Electrical Characterizations of Yttrium Oxide Films after Postannealing Treatments C. Durand,a C. Dubourdieu,b,z C. Vallée,a E. Gautier,c F. Ducroquet,d D. Jalabert,e H. Roussel,b M. Bonvalot,a and O. Jouberta a

Laboratoire des Technologies de la Microélectronique (LTM/CNRS), CEA-LETI, 38054 Grenoble Cedex 9, France Laboratoire des Matériaux et du Génie Physique, UMR-CNRS 5628, ENSPG, 38402 Saint Martin d’Hères Cedex, France c Institut des Matériaux de Nantes (IMN-CNRS), 44322 Nantes Cedex 3, France d Laboratoire de Physique de la Matière, INSA de Lyon, 69621 Villeurbanne Cedex, France e DRFMC, CEA-Grenoble, 38054 Grenoble Cedex 9, France b

We investigated annealing effects on yttrium oxide films elaborated at 350°C by pulsed injection plasma-enhanced metallorganic chemical vapor deposition. A certain amount of carbon is present in the as-deposited layers due to incomplete metallorganic precursor decomposition. Because the carbon is known to be disadvantageous for the dielectric permittivity, several annealing post-treatments 共under Ar or O2 at 700°C and under H2 at 450°C during 30 min at atmospheric pressure兲 were performed. Chemical modifications in the yttrium oxide layer and also at the interface were investigated combining several analyses 关as infrared spectroscopy, 共angle-resolved兲 X-ray photoelectron spectroscopy, X-ray diffraction, transmission electron microscopy, electron energy loss spectroscopy, and Rutherford backscattering spectrometry兴 carried out on thick and thin films 共respectively, ⬃40 and ⬃5 nm兲. High-temperature post-treatments result in a significant reduction of carbon content. However, the presence of oxygen at high temperature 共even for low residual oxygen partial pressure兲 involves SiO2 formation at the interface and leads to the transformation of the yttrium oxide into Y-based silicates, which, for the thin films, results in the absence of Y2O3 crystalline phase. The H2 /450°C anneal does not modify the film’s chemical nature except for a significant reduction of hydrogen content. After annealing under H2 /450°C, a strong improvement of the capacitance behavior is observed, especially on the thin films. © 2005 The Electrochemical Society. 关DOI: 10.1149/1.2109487兴 All rights reserved. Manuscript submitted April 18, 2005; revised manuscript received August 3, 2005. Available electronically October 27, 2005.

The aggressive scaling of complementary metal oxide semiconductor 共CMOS兲 devices beyond the 0.1-␮m generation is driving the SiO2-gate dielectrics to its physical limits due to excessive direct tunneling leakage current. High-dielectric-constant 共high-k兲 metal oxides appear to be promising candidates to replace SiO2.1 For a few years, substantial efforts have been done on many materials, especially on ZrO2, HfO2, and their alloys with SiO2 and Al2O3. Rare earth oxides and Y2O3 have also attracted attention thanks to their reasonably high permittivities 共␬ = 12–15 for Y2O3兲, thermal stabilities with Si, relatively large conduction band offsets 共2.3 eV for Y2O3兲, and low leakage currents.2 Although previous studies have revealed that rare earth oxide films absorb water,3-5 共films can be capped or annealed to avoid the water contamination兲, interesting results have been found on Y2O36-12 and Y-based silicate.13-15 In previous studies, we showed that yttrium silicate and yttrium oxide films could be grown at low temperature by pulsed-injection plasma-enhanced metallorganic chemical vapor deposition.16-18 However, these films contain a significant amount of carbon due to the low processing temperature that does not allow the metallorganic precursor to fully decompose. Because carbon is detrimental to the electrical properties, it is desirable to remove the carbon contamination, which can be carried out by postdeposition treatments such as annealings. Annealings are often necessary to improve the overall properties of dielectric films. Moreover, the CMOS process includes high-temperature annealing steps for dopant diffusion. Interface regrowth after high-temperature annealing is one of the most important parameters to control as it greatly increases the equivalent oxide thickness 共EOT兲. The key challenge is thus to control the nature of the interface and to limit its formation. In the present study, postannealing effects are studied on thick and thin yttrium oxide films 共⬃40 and ⬃5 nm兲, respectively, deposited at 350°C by pulsed liquid-injection plasma-enhanced metallorganic chemical vapor deposition 共PE-MOCVD兲. Structural and chemical modifications upon annealing are investigated combining Fourier transform infrared spectroscopy 共FTIR兲, X-ray photoelectron spectroscopy 共XPS兲, X-ray diffraction 共XRD兲, Rutherford back-

z

E-mail: [email protected]

scattering spectroscopy 共RBS兲, nuclear reaction analysis 共NRA兲, and elastic recoil detection analysis 共ERDA兲. Furthermore, the changes at the interface are analyzed on thin films by means of angleresolved X-ray photoelectron spectroscopy 共AR-XPS兲, attenuated total reflection-FTIR 共ATR兲, transmission electron microscopy 共TEM兲, and energy electron loss spectroscopy 共EELS兲. Finally, the effects of annealing on the electrical properties are discussed. Experimental Yttrium oxide films were grown by pulsed liquid-injection PEMOCVD, which combines a plasma assistance and a pulsed liquid delivery system. The technique is precisely described elsewhere.16 The plasma allows deposition at much lower temperature 共below 380°C兲 compared to a classical MOCVD process. The liquid system allows precise control of the vapor pressure of precursor in the gas phase and thus a good reproducibility can be achieved.19 The liquid is formed from Y共tmhd兲3 关for tri共2,2,6,6-tetramethyl-3,5heptanedionate兲 yttrium兴 precursor powder diluted in a solvent 共cyclohexane兲 at 0.02 mol/L. The opening time and the frequency of the injector were kept constant and equal to 2 ms and 5 Hz, respectively. The injected volume per injection was equal to 6.4 共±0.5兲 mm3. The injection number was equal to 5000 and 700 for the thick and the thin films, respectively. The plasma power supply was operated at low power 共70 W兲. The plasma gas was oxygen 共flow rate of 250 sccm兲, and the deposition pressure was maintained at 0.27 kPa 共2 Torr兲. The substrate chuck was heated at 350°C for deposition. Films were deposited on thermally oxidized p-type 8-in. Si共100兲 wafers with a thermal SiO2 thickness of 0.8 nm. The chemical characterizations of the as-deposited layers were presented in details in Ref. 18. According to this previous work, the as-deposited films elaborated with an injection frequency of 5 Hz present an interfacial layer 共SiO2 and Y-silicate mixture兲 with a yttrium oxide layer on top of it, which contains carbon 关Si/SiO2 + SixOyYz 共⬃2 nm兲/YxOyCz兴 as shown in Fig. 9.18 The film thickness was estimated at ⬃40 and ⬃5 nm for the thick and thin layers, respectively. The films were subsequently cut into several pieces in order to carry out several postdeposition annealings: under Ar or O2 flux at 700°C and under H2 flux at 450°C. All annealings were done at atmospheric pressure during 30 min.


F218

Table I. Binding energy, fwhm, and peak shape (Lorentzian– Gaussian percentage) for each Si oxidation state to fit the Si 2p XPS region.

a

Oxidation state

Binding energy 共eV兲

fwhm

Lorentzian 共L兲–Gaussian 共G兲 percentage 共%兲共L:100 and G:0兲

Si0 Si1+ Si2+ Si3+ Si4+

99.3 100.3 101.1 102.0–102.3a 103.3

0.9 1.0 1.2 1.3–1.5a 1.4–1.5a

25 0 0 0 0

Variable values which depend on fitting results.

X-ray photoelectron spectroscopy 共XPS兲 was used to identify bonds in the film and in the interface. The qualitative depth profile was carried out by angle-resolved AR-XPS. The effective probed depth decreases as the take-off analysis angle decreases. The surface contributions are thus enhanced at grazing angle compared to analyses near normal take-off angles. XPS spectra were collected with an Al K␣ X-ray source 共h␯ = 1486.6 eV兲, and a hemispherical electron analyzer was used in fixed pass energy mode 共20 eV兲. The plot resolution was set at 0.1 eV, except for the Si 2p regions, for which it was fixed to 0.05 eV in order to improve the fitting procedure. The photoelectron take-off angle was fixed at 45°, except for AR-XPS analyses. The surfaces were treated by an Ar+ ion beam at 1.5 keV during 30 s with a flux of 10 ␮A. Nevertheless, no Ar sputtering was done prior to measurements for thin films in order to prevent film or interface degradation. Energy shift calibration was performed by positioning the Si 2p core level from the Si substrate at 99.3 eV. This correction positions the C 1s peak from surface contamination at 285.7 eV, so when no Si 2p contribution from the substrate was detected, the charging correction was carried out with the surface C 1s peak. The Si 2p region provides worthwhile information on the interface between the substrate and the high-k top layer. Five contributions 共Si0, Si1+, Si2+, Si3+, and Si4+兲 were used to fit accurately the Si 2p signal. Table I presents the fitting characteristic values of each contribution, which are derived from typical values of the literature.20 These values were fixed to fit the whole Si 2p signal 关i.e., binding energy, full width at half maximum 共fwhm兲, and Lorentzian/Gaussian peak shape percentage兴. The inset in Fig. 1

shows an example of the Si 2p fitting model. The well-known method of the SiO2 thickness calculation is given by the following formula21

冉

tox = ␭ox sin共␪兲ln

R +1 R⬁

冊

关1兴

where ␪ is the photoelectron take-off angle, R is the ratio Aox /ASi of the measured sample 共ASi being the relative areas of the Si0 component and Aox the sum of the relative area of the Sin+ components兲, and R⬁ is the ratio Aox /ASi in the case of a thick thermal SiO2 and bulk Si sample. ␭ox is the inelastic mean-free path for Si 2p photoelectrons coming from the Si substrate and the SiO2 共supposing no elastic scattering effects兲. However, ␭ox is difficult to estimate for our interface because of the silicate and yttrium oxide presence. For that reason, only the area ratio R evolution is given to study the interface modification. This thickness representation of the oxidized Si-based compounds is valuable because ␭ox, ␪, and R⬁ can be considered constant. Infrared measurements were performed using a Bruker IFS 55 Fourier-transform infrared spectrometer 共FTIR兲. The classical FTIR technique is based on transmission of the infrared beam through the wafer, whereas in the ATR technique, the infrared beam is reflected on the substrate surface by means of a germanium prism pressed against the silicon sample. The ATR characterization method using a p-polarized infrared beam gives useful information on the bonding environment for thin films 共⬍50 nm兲 and also for the interface 共if the film is thin enough兲.22 The resolution and number of scans were set at 4 cm−1 and 500, respectively. Chemical and structural modifications were also studied by XRD in terms of phases, crystalline structure, and orientation. As the XRD analysis is limited by the film thickness, this characterization was done only on thick samples. The composition of thick films was determined by RBS using a He+ ion beam. The detection angle was 165° and the detector resolution was 16 keV. NRA and elastic recoil detection analysis 共ERDA兲 were also carried out on the same samples in order to evaluate, respectively, the carbon and hydrogen contents. TEM analyses were performed with a JEOL 2010F equipped with a field emission gun operating at 200 kV. Point-to-point resolution is 0.19 nm, and the minimum probe size is 0.5 nm. This microscope is also equipped with a Gatan imaging filter 共GIF兲 that allows for EELS. For spectra shown in this study, only the central area of the charge-coupled device 共CCD兲 was used 共200 ⫻ 1024 pixels兲. Analyses across various interfaces were obtained with the GIF controlling to pilot the TEM beam. The EELS spectra were obtained at interfaces by positioning the probe automatically along a line scan using a spot size of around 1 nm. The samples were prepared in cross-sectional orientation by mechanical polishing to a thickness of 40 mm. They were then thinned by low-voltage 共3 kV兲 argon ion milling of both surfaces.23 Electrical measurements were performed on MOS structures. Gold electrodes were evaporated though shadow masks just after annealing treatments. The surface area of the MOS capacitors was equal to 0.694 ⫻ 10−2 mm2. Measurements of capacitance vs applied voltage 共C-V兲 were conducted on a HP 4284 A precision LCR meter at three frequencies 共1, 10, and 100 kHz兲. The equivalent oxide thickness 共EOT兲 was extracted from the C-V accumulation region 共quantum mechanical correction was not applied兲.

Table II. Chemical composition of thick films determined by RBS. The carbon and hydrogen contents were extracted from NRA and ERDA analyses, respectively.

Figure 1. Area ratios calculated from the Si 2p fitting procedure vs annealing post-treatment process. The graph gives the area ratio Si–O/Si and also the area ratio of each Si suboxide state 共Sin+ with n = 1–4兲. The inset shows an example of decomposed spectrum using the Si 2p fitting procedure.

Y O C H Si

As-deposited 共%兲

H2 /450°C 共%兲

Ar/700°C 共%兲

O2 /700°C 共%兲

23.1 56.6 16.0 4.2 —

24.2 58.8 16.1 0.9 —

36.0 52.6 9.0 2.4 —

29.0 40.7 5.7 0.7 23.9


F219

Figure 2. FTIR spectra for the thick as-deposited and annealed samples.

Current–voltage curves 共I-V兲 were measured using a HP4156A meter. Results and Discussion Annealing effects on thick layers (⬃40 nm).— The film composition extracted from RBS measurements before and after annealings is given in Table II. The composition takes into account the carbon and the hydrogen contents determined respectively by NRA and ERDA analyses. RBS analyses show that the main consequence of 700°C annealings is a reduction of carbon content in the layer. The 关C兴/关Y兴 ratio calculated from RBS percentage is equal to ⬃0.7 for the as-deposited film and after H2 /450°C annealing, whereas the ratio value decreases down to around ⬃0.2 for 700°C annealings 共both Ar and O2兲. The FTIR spectra, shown in Fig. 2, confirm also the large carbon reduction after high-temperature treatments. Absorption bands at 840, 1520, and 1440 cm−1, assigned to C–O and C–C stretching modes,24-27 are significantly attenuated. After O2 /700°C annealing, Si atoms are detected in the films by RBS. These Si atoms are engaged in SiO2 or Y-based silicate compounds as shown by FTIR analyses 共the peak around 1080 cm−1 on FTIR spectra typically corresponds to SiO2 or Y-based silicate signatures兲. The silicide presence is not expected after annealing under O2 at high temperature. Chambers and Parsons have shown that a Y-based silicide film is completely transformed into silicate film upon oxi-

dized atmosphere annealing at 900°C.13 The silicate formation is strongly favored by the oxygen presence in the annealing atmosphere at 700°C. After H2 /450°C annealing, the carbon impurities are not significantly reduced, because the intensity of bands from organic bonds is

Figure 3. XRD diagrams for 共1兲 as-deposited, 共2兲 H2 /450°C annealed, 共3兲 Ar/700°C annealed, and 共4兲 O2 /700°C annealed samples.

Figure 5. The ATR absorbance spectra for as-deposited and annealed films compared with the bare SiO2 /Si substrate 共0.8-nm thermal oxide兲, which is traced in solid line. Note the silicate signature at 1040 cm−1.

Figure 4. XPS Y 3d 共a兲 and O 1s 共b兲 regions for thick as-deposited and annealed samples. Ar sputtering was done before analysis in order to remove the surface contamination.

F220

Journal of The Electrochemical Society, 152 共12兲 F217-F225 共2005兲 Ar/700°C annealing, the 关O兴/关Y兴 ratio is equal to 1.46, which is very close to the expected value of 1.5 for pure Y2O3 compound. The ␪/2␪ diagrams are shown in Fig. 1. As-deposited films are amorphous 共or nanocrystallized兲 and remain amorphous 共or nanocrystallized兲 after annealing at 450°C, whereas crystalline Y2O3 is formed upon high-temperature treatments. Y2O3 formation at 700°C is also confirmed by the XPS spectra, shown in Fig. 3. For the as-deposited films, the Y 3d doublet is shifted to higher binding energy compared to the Y2O3 reference position 共Y 3d5/2 and Y 3d3/2 at 156.7 and 158.7 eV, respectively兲.28 This shift was attributed to Y–O–C bonds in the films. Carbon bonding to oxygen is expected to shift the Y 3d peak toward higher bindings energies as compared to Y–O in Y2O3 by considering the electronegativity of carbon.16 The peak of O 1s at 530.6 eV is consistent with this picture. Annealing under H2 at 450°C results in no change in the Y 3d region nor on the O 1s peak at 530.6 eV. For samples annealed at 700°C 共both Ar and O2兲, the Y 3d5/2 peak position at ⬃156.7 eV and the O 1s position at ⬃529.5 eV clearly indicate the presence of Y2O3 compounds. The second peak on O 1s spectra located at around 533 eV mainly corresponds to an O–C bond related to carbon content in the film. The residual peak shoulder at ⬃532 eV observed for 700°C annealings is attributed to residual carbon content and also to OH bonding or oxygen trapping in the materials, as already observed in the literature for XPS analysis of Y2O3 materials.13,16,29 This study on thick films has allowed us to determine the annealing effects on yttrium oxide films. In particular, we have shown that the silicate formation is strongly enhanced by the oxygen presence in the annealing atmosphere at 700°C. Moreover, 700°C annealing induces the formation and the crystallization of Y2O3 with low carbon content, whatever the annealing atmosphere 共Ar or O2兲. Finally, the H2 /450°C does not significantly modify the film in terms of crystallization and carbon contamination. We have only detected a reduction of hydrogen content.

Figure 6. XPS Si 2p 共a兲, Y 3d 共b兲, and O 1s 共c兲 regions for as-deposited, H2 /450°C, and 700°C annealed 共Ar and O2兲 samples. The Si 2p XPS spectrum of the bare SiO2 /Si substrate 共0.8 nm thermal oxide兲 is traced in a solid line. Si 2p Spectra is normalized with respect to the Si0 peak area.

similar to those from the as-deposited samples. However, the RBS analysis points out a clear reduction of hydrogen content 共−3,3%兲. The as-deposited and H2 /450°C-annealed films are oxygen-rich 共关O兴/关Y兴 Ⰷ 1.5兲, which is most probably related to the organic impurities 共C–O bonds兲. Upon high-temperature annealing, carbon reduction is accompanied by a decrease of oxygen content. After

Annealing effects on thin layers (⬃5 nm): Effects on the interface layer and silicate formation.— The same types of annealings were performed on thin layers 共⬃5 nm兲. Such a thickness is typical of that used for CMOS applications. Changes at the interface were particularly investigated here. The initial structure of the asdeposited samples is extracted from Ref. 18 and shown in Fig. 9. The films consist of an interfacial layer 共made of a mixture of SiO2 and yttrium silicate兲 and of a yttrium oxide film with carbon contamination on top of it. ATR spectra of as-deposited and annealed layers are presented in Fig. 4. The ATR spectrum of the bare Si/SiO2 共0.8 nm兲 substrate is also shown. The band at 840 cm−1 and two other bands centered on 1520 and 1440 cm−1, assigned to C–O and C–C stretching modes, are decreased after H2 /450°C annealing and become very weak after the 700°C post-treatments. Thus, carbon content in the films is weakly reduced after H2 /450°C 共which was not the case for thick films兲 and is almost completely cancelled out for high-temperature treatments 共no significant difference appears whether O2 or Ar is used兲. XPS spectra of the Si 2p, Y 3d doublet 共Y 3d5/2−3/2兲 and O 1s are shown in Fig. 5 for as-deposited and annealed films. As the XPS method for measuring oxide thickness is based on the area ratio between oxidized silicon peaks 共denoted Si–O兲 and substrate silicon peak 共denoted Si0兲, Si 2p spectra normalized on the Si0 peak area provides a good estimation of the oxidized Si-based compounds thickness evolution. None of the annealing treatments tends to form the pure Y2O3 compound, because no Y 3d5/2 peak expected at 156.7 eV28 is observed in the Y 3d region 共Fig. 5b兲. For annealing performed at 450°C under H2, the Si–O peak remains almost unchanged, thus showing that the interface remains almost unaltered 共clearly no silicate or SiO2 is formed at 450°C兲. The Y 3d doublet is shifted toward lower binding energy, and the Y 3d split is more pronounced, which indicates a reduction of the


F221

Figure 7. Angle-resolved XPS analyses 共Si 2p and Y 3d regions兲 at three take-off angles 关90°共䊏兲, 45°共쎲兲, and 20°共䉱兲兴 of the 700°C Ar annealing 共a and b兲 and of the 700°C O2 annealing 共c and d兲. Note that the intensity of Si 2p spectra is normalized on the Si–O peak intensity.

bonding environment multiplicity for yttrium atoms. The shift is attributed to the reduction of carbon contamination, in particular, in the yttrium oxide layer as observed in the ATR spectrum. In the O 1s region, the peak located around 533 eV for as-deposited samples is attributed to three main contributions 关Si–O bonds from SiO2 共533.0 eV兲, carbon surface contamination 共O–C bonds兲, and bulk carbon contamination 共O–C bonds兲兴, whereas the peak located at ⬃530.6 eV is mainly attributed to Y–O–C bonds 共a small part of the signal also comes from Y–O–Si bonds兲.16 A slight shift 共−0.3 eV兲 toward lower binding energies is observed on this peak after annealing at 450°C in the O 1s region. It is again coherent with the reduction of carbon content in the film. In the case of annealing at 700°C, the Si 2p region exhibits a strong increase of the Si–O peak amplitude. The Si–O peak for as-deposited films, which is shifted toward lower binding energy as compared to the initial substrate Si–O peak, is the signature of the interfacial Y-based silicate layer 关Si 2p peak for Y–O–Si bond is measured at lower binding energy 共⬃102 eV兲 than SiO2 due to a lower electronegativity of yttrium 共1.22兲 compared to that of silicon 共1.90兲兴. No shift is observed after annealing, but the large increase of the peak amplitude shows that yttrium silicate is further formed upon annealing. The yttrium silicate is formed by reaction of the YxOyCz top layer with the Si/SiO2 substrate. The Y 3d5/2 peak position near 158 eV 共shifted toward lower binding energy as compared to the as-deposited films or H2 /450°C annealed films兲 is consistent with yttrium silicate phases and a reduction in carbon content 共as shown by ATR-FTIR in Fig. 4兲. On the ATR spectra 共Fig. 4兲, the band at 1240–1250 cm−1 originating from longitudinal optical 共LO兲 phonon mode of the Si–O–Si stretching vibration is strongly enhanced for 700°C annealings compared to the as-deposited and H2 annealing samples. It indicates that the number of Si–O bonds is significantly increased 共i.e., the density and/or the thickness are enhanced兲. The peak shape also changes, which results from a SiO2 interface modification in terms of density, Si suboxide presence, and thickness.22 The peak at 1040 cm−1, which is strongly enhanced

upon high-temperature annealing, is attributed to the Y-based silicate bonding 共Si–O–Y兲. For more quantitative analysis of the silicate phases, a detailed study of the Si 2p peak was done. In Fig. 6, we plot the area ratio between each Sin+ component 共n = 1–4兲 and the silicon substrate Si0. Areas are extracted from the Si 2p fitting procedure described previously. The inset in Fig. 6 shows an example of fitted curve, and Table I gives the fixed characteristic values of each Si-oxidized contribution used for the fitting procedure. The area ratios for the initial Si/SiO2 共0.8-nm兲 substrate are also shown for comparison. The huge increase upon high-temperature annealing of the Si–O/Si0 area ratio, which is the sum of all Sin+ components, illustrates evidently that such treatments favor oxidized Si-based compound formation. The Si1+ /Si0 and Si2+ /Si0 ratios remain roughly constant, whereas both Si3+ /Si0 and Si4+ /Si0 ratios are enhanced, especially under O2 flux. This huge increase of both Si3+ /Si0 and Si4+ /Si0 ratios indicates the increase of both silicate and SiO2 content. The H2 /450°C annealing does not modify the oxidized Si-based compound thickness. The high-temperature annealed samples were further investigated by AR-XPS analyses. Figure 7 shows the Si 2p and Y 3d signals collected at three take-off angles 共90, 45, and 20°兲 for the 700°C Arand O2-annealed samples, respectively. On Fig. 7a and c, Si 2p spectra were normalized with respect to the Si–O peak intensity. In both cases, a weak shift of the Si–O peak toward lower binding energy is observed as the angle is decreased, which indicates a silicate signal enhancement compared to the SiO2 one. The Sin+ 共with n = 1–4兲 relative contribution area calculated from the Si 2p fitting procedure shows that the Si3+ is clearly highlighted at grazing takeoff angle compared to the Si4+ contribution, whatever the annealing atmosphere. These observations indicate that the films are composed of a SiO2 interfacial layer with a yttrium silicate layer above. Concerning the Y 3d doublet 共Fig. 7b and d兲, no Y2O3 signature is visible 共i.e., Y 3d5/2 at 156.7 eV兲 whatever the take-off angle. No significant shift is observed on the Y 3d doublet; thus, the yttrium


F222

Figure 9. Diagram summarizing the chemical nature and thickness modifications induced by the O2 /700°C annealing post-treatment compared to the initial as-deposited sample microstructure. The as-deposited microstructure is extracted from a previous work published in Ref. 18.

Figure 8. High-resolution TEM image and EELS measurements of the 700°C O2 annealing, 关共a兲 and 共b兲, respectively兴. The film nature and thickness are noted on TEM micrograph. The schematic line scan for EELS measurements is also shown on the TEM image. EELS measurements shows Si K and Y L2,3 edges measured across the film.

atoms are only involved in silicate bonds. The Y 3d doublet split is obscured, which is indicative of multiple bonding environments of yttrium atoms. To conclude, the film structure deduced from ARXPS is Si/SiO2 /YxOySiz without any Y2O3 signature for both Ar and O2 /700°C annealings.

Table III. Y/Si ratios obtained from EELS spectra at several probe positions. Ratios are calculated from position 1, corresponding to the first EELS measure done in the Si substrate, beyond position 12, corresponding to the last measure done in the extreme top Y-based layer. Although this ratio calculated from EELS spectra must not be considered as an exact value, the ratio variations are clearly related to the concentration. Probe positions 1–5 共Si Substrate兲 6 7 8 9 10 11 12 共Extreme top Y-based layer兲

Y/Si Ratios 0 0.0011共±0.0022兲 0.045 共±0.0076兲 0.14 共±0.024兲 0.21 共±0.035兲 0.21 共±0.035兲 0.22 共±0.036兲 0.32 共±0.054兲

These analyses were cross-checked with EELS measurements and HR-TEM images performed on an O2 /700°C-annealed sample 共Fig. 8兲. The TEM image 共Fig. 8a兲 shows clearly a thick interfacial layer 共⬃3.0 nm兲 and a dark film on top of it 共⬃4.1 nm兲. The top layer presents crystalline regions. Electron diffraction confirms the crystalline nature of the films 共XRD could not be performed on such thin films兲. The schematic line scan for the EELS analysis is marked on the TEM micrograph. The ionization Si K, Y L3, and L2 edges 共respectively, at 1839, 2080, and 2155 eV兲 were used to detect simultaneously the presence of silicon and yttrium atoms in the same spectral region. The spectra are shifted for clarity 共Fig. 8b兲. EELS spectra were sequentially done along the line scan using identical step size. On the start point, only the Si K-edge is detected 共and the extended fine structure兲, because the line scan begins in the Si substrate. On the following EELS spectra, when yttrium is detected 共Y L2 and L3 edge兲, the Si K-edge is still observed. Therefore, EELS measurements confirm that yttrium is incorporated in the form of a silicate. Thus, the dark layer is composed of Y-based silicate compounds. The contrast of the top layer 共silicate兲 is not uniform on the TEM image, which could indicate a gradient in the silicate composition, with a rich Y-silicate on the extreme top layer. From EELS spectra, quantitative Y/Si ratios were calculated for each step along the line scan. We used the Hartree–Slater model with convergence angle correction to evaluate the cross section in the Gatan-ELP software.30 Their values are reported in Table III. Ratios are calculated from position 1, corresponding to the first EELS measure done in the Si substrate beyond position 12, corresponding to the last measure done in the extreme top Y-based layer. Although this ratio calculated from EELS spectra must not be considered as an exact value, the ratio variations are clearly related to a concentration change. The Y/Si ratio is found to increase when the probe position is going from the substrate to the top surface. This indicates a silicate composition gradient with a rich Y-silicate on the extreme top layer. This composition gradient was not detected by the previous AR-XPS analysis 共Fig. 7兲, because no shift is observed on the Y 3d doublet with varying take-off angle. Because the Si 2p peak from the silicon substrate is detected even for the most grazing angle 共20°兲, the silicate composition gradient located on the top surface layer is likely not to be detected. The obscured Y 3d doublet split for 700°C treatments measured in AR-XPS analysis 共Fig. 7兲 is consistent with a multiple bonging environment induced by the silicate composition gradient. Moreover, the XPS resolution using a twin anode source is probably insufficient to detect fine shifts. To conclude this part, we have sketched the structure of the asdeposited and O2 /700°C annealed films in Fig. 9. Unlike previous results on thick layers, no Y2O3 phase is observed on thin layers after high-temperature post-treatments. The stratified structure Si/SiO2 /Si–O–Y is deduced for 700°C post-treatments whatever the annealing atmosphere 共Ar and O2兲. In the case of Ar atmosphere, the SiO2 and silicate formation certainly comes from the residual oxy-


F223

Figure 10. C–V curves for thick and thin as-deposited, 450°C-H2 and 700°C-Ar annealed samples measured at three frequencies 共1, 10, and 100 kHz兲. Graphs 共a兲 and 共c兲: C-V curves for thick films 共⬃40 nm兲. Graphs 共b兲 and 共d兲: C-V curves for thin films 共⬃5 nm兲.

gen partial pressure in the furnace. According to these results, yttrium oxide is not stable on SiO2. The silicate formation after hightemperature post-treatments has already been reported for yttriumbased compounds deposited on Si3,13,31 and on SiO2.15 Yttrium silicate is thermodynamically favorable when Y2O3 is present on the SiO2 layer.14 As the silicate formation is detrimental for EOT reduction, yttrium oxide may be difficult to integrate as a transistor gate oxide. Note that Copel et al. take advantage of the silicate formation to develop a solid-state technique using a postannealing treatment at 900°C under ultrahigh vacuum. In this case, the unwanted interfacial SiO2 is consumed by yttrium oxide to form the silicate layer without SiO2 interface. The silicate films have already presented interesting electrical properties 共EOT equal to 1.2 nm兲.13,14 These results seem to indicate that Y-based silicate is a most promising material, compared to yttrium oxide, as a high-k material for CMOS applications. Electrical characterizations on thick and thin films.— The C-V curves measured at several frequencies 共1, 10, and 100 kHz兲 are given in Fig. 10 for the as-deposited, H2 /450°C, and Ar/700°C samples. An effective permittivity of 7.5 共±1.2兲 for as-deposited films is deduced from C-V curves. This low value compared to the usual permittivity of Y2O3 material 共12–15兲 is explained by the carbon and silicate presence in the films. C-V curves of the as-deposited samples exhibit a large flatband voltage 共VFB兲, which is much more pronounced for high-frequency measurements 共−4 V at 100 kHz for the thick films兲. The theoretical VFB is expected to be equal to 0.1–0.2 V when no charge is present in the oxide 共the work function of the gold gate is taken as ⌽Au = 5.1 eV32兲. The VFB shift is attributed to a significant number of positive charges into the oxide, some of which are mobile as indicated by the hysteretic behavior 共reaching 0.20 V兲 of the C-V curves 共not shown here兲. For both thin and thick films, the H2 /450°C annealing improves significantly the electrical characteristics. First, the accumulation capacitance is increased; thus, the EOT is decreased compared to as-deposited films. It is more pronounced for the thin films

Figure 11. Leakage current density vs applied voltage 共J-V兲 curves for the as-deposited, 450°C-H2, and 700°C-Ar annealed films. 关The graphs 共a兲 and 共b兲 represent J-V curves for thick and thin films, respectively.兴


F224

Table IV. Electrical characteristics list extracted from electrical curves measured on thin films (Fig. 10b and d and 11b): flatband voltage at 10 kHz „VFB…, EOT at 10 kHz, gate leakage current at VFB-1 V „JG…, interfacial state density „Dit…, and voltage hysteretic shift „⌬VHys…. EOT VFB 共10 kHz兲共10 kHz兲共V兲共nm兲 As deposited H2 /450°C Ar/700°C

−1.3 −0.3 −0.5

6.8 5.4 10.4

JG 共VFB-1 V兲共J/cm2兲 4.0 ⫻ 10−8 3.1 ⫻ 10−8 5.8 ⫻ 10−8

Dit ⌬VHys 共eV−1 cm−1兲共V兲 — 2 ⫻ 1011 —

0.20 0.07 —

共EOT = 6.8 and 5.4 nm, respectively, for the as-deposited and H2-annealed samples shown in Fig. 10兲, which could be due to the carbon removal in thin films. Second, the VFB shift is strongly reduced as well as its dependence with frequency. For the thin film, there is even no more shift and frequency dependence. Therefore the H2 annealing reduces significantly the charge density in the films, which is consistent with the reduction of the hysteretic behavior 共reaching in this case 0.07 V兲 also measured on the C-V curves 共not shown here兲. It may be related to the decrease of H content in the film. Because of VFB stability with the frequency observed for the H2 annealing, it is possible to use the combined high- and lowfrequency method33 to extract the interfacial density of state 共Dit兲. This method allows calculating the Dit profile in the bandgap thanks to the capacitance difference between low- and high-frequency C-V curves. From this capacitance difference related to the Dit, we deduced an average Dit value for H2 annealing films equal to 2 ⫻ 1011 eV−1 cm−1. The high-temperature annealings do not improve the electrical properties. For thin films, the formation of SiO2 and Y-silicates leads to a decrease of the capacitance 共for example, the EOT is 6.8 nm for as-deposited films and 10.4 nm after Ar/700°C annealing兲. For thick films, the densification and crystallization of Y2O3 counterbalance the increase of the interfacial layer thickness and results in a slight increase of the accumulation capacitance. The capacitance increase in the positive voltage is probably due to a low-quality interface, which can generate minor carriers. The VFB shift and also the frequency dependence are reduced, which is a sign of decreasing charge density in the oxide, due probably to carbon removal. Moreover, a stretching effect is observed on the C-V curves, especially on thin film. This phenomenon is generally attributed to a nonabrupt interface, which could arise from the Y-based silicate gradient previously observed by EELS analysis. Leakage current density vs applied voltage 共J-V兲 curves are presented in Fig. 11. A low leakage current density is measured on the as-deposited thick films 共⬍10−7A/cm2 for the bias voltage ranging from 0 to −20 V兲. The breakdown voltage happens at −34 V, which corresponds to an electrical field 共Ebd兲 of 8 MV/cm 共assuming a 40-nm film thickness兲. This Ebd value corresponds to the one expected from the McPherson’s model that links Ebd with the permittivity ␬ by the following relationship: Ebd ⬀ 1/冑␬.34 These electrical properties in terms of leakage current density and electrical breakdown values are quite promising, especially for MIM applications.17 No significant modification is observed on J-V curves for the H2 /450°C annealed samples, which is consistent with previous chemical characterizations. On the contrary, Ar/700°C annealings are clearly detrimental 共increase of the leakage current density and decrease of the breakdown voltage兲, which can be attributed to the occurrence of grain boundaries upon crystallization. For thin films, the as-deposited and the H2 /450°C annealed films show a leakage current density lower than 3 ⫻ 10−8 A/cm2 at −1 V. The Ar/700°C annealings exhibit a higher current density, attributed to the film crystallization. Table IV, which summarizes the main electrical characteristics for thin films, clearly shows that H2 /450°C annealing is an effective

postdeposition treatment for reducing the positive charges in the oxide film and reducing the EOT with an acceptable interfacial state density 共2 ⫻ 1011 eV−1 cm−1兲, without modifying the leakage current properties. Positive electrical effects on high-k material induced by hydrogen annealing have already been reported in the literature such as the reduction of interface state density, the frequency dispersion, and the hysteresis phenomenon.35,36 Conclusion Annealing post-treatments were carried out on yttrium oxide films grown at 350°C by pulsed injection PE-MOCVD in order to evaluate their effects on both chemical nature and electrical properties. High-temperature annealings at 700°C efficiently reduce the organic contamination 共carbon and hydrogen兲. However, these annealings involve SiO2- and Y-based silicate formation 共depending on the available oxygen quantity in the annealing atmosphere兲. The high-temperature treatments are clearly detrimental in terms of EOT reduction, due to low-permittivity compound formation 共silicates and SiO2兲. On the contrary, we point out that the annealing at 450°C under H2 flux considerably improves the electrical properties 共EOT reduction, charge density decrease, Dit ⬃ 2 ⫻ 1011 eV−1 cm−1兲 of thin films. These improved properties are attributed to a reduction of hydrogen and carbon content. Moreover, electrical characterizations show promising current properties for as-deposited thick films in terms of leakage current 共⬍3 ⫻ 10−8 A/cm2 at −1 V兲 and electrical breakdown field 共8 MV/cm兲, which are attractive for metal– insulator–metal capacitors.17 Acknowledgments The HRTEM/EELS work was conducted at CP2M, an electron microscopy and microanalysis center at the University of AixMarseille III, where state-of-the-art equipment is accessible and available to all. Many thanks to its director and to the technical persons who helped in obtaining these results. The authors would like to thank warmly O. Renault for fruitful discussions on XPS, N. Rochat for helpful expertise on IR analysis, L. Vallier for XPS expertise, and T. Luciani for deposition equipment support. CNRS assisted in meeting the publication costs of this article.

References 1. G. D. Wilk, R. M. Wallace, and J. M. Anthony, J. Appl. Phys., 89, 5243 共2001兲. 2. Y. C. Yeo, T. J. King, and C. Hu, Appl. Phys. Lett., 81, 2091 共2002兲. 3. B. W. Busch, J. Kwo, M. Hong, J. P. Mannaerts, B. J. Sapjeta, W. H. Schulte, E. Garfunkel, and T. Gustafsson, Appl. Phys. Lett., 79, 2447 共2001兲. 4. T. Gougousi, M. J. Kelly, and G. N. Parsons, Appl. Phys. Lett., 80, 4419 共2002兲. 5. D. Niu, R. W. Ashcraft, and G. N. Parsons, Appl. Phys. Lett., 80, 3575 共2002兲. 6. S. Guha, E. Cartier, M. A. Gribelyuk, N. A. Bojarczuk, and M. C. Copel, Appl. Phys. Lett., 77, 2710 共2000兲. 7. E. K. Evangelou, C. Wiemer, M. Fanciulli, M. Sethu, and W. Cranton, J. Appl. Phys., 94, 318 共2003兲. 8. F. Paumier and R. J. Gaboriaud, Thin Solid Films, 441, 307 共2003兲. 9. Y. Tsuchiya, M. Endoh, M. Kurosawa, R. T. Tung, T. Hattori, and S. Oda, Jpn. J. Appl. Phys., Part 1, 42, 1957 共2003兲. 10. J. Kwo, M. Hong, A. R. Kortan, K. L. Queeney, Y. J. Chabal, R. L. Opila, D. A. Muller, S. N. G. Chu, B. J. Sapjeta, T. S. Lay, J. P. Mannaerts, T. Boone, H. W. Krautter, J. J. Krajewski, A. M. Sergnt, and J. M. Rosamilia, J. Appl. Phys., 89, 3920 共2001兲. 11. D. Niu, R. W. Ashcraft, Z. Chen, S. Stemmer, and G. N. Parsons, J. Electrochem. Soc., 150, F102 共2003兲. 12. A. Dimoulas, G. Vellianitis, A. Travlos, V. Ioannou-Sougleridis, and A. G. Nassiopoulou, J. Appl. Phys., 92, 426 共2002兲. 13. J. J. Chambers and G. N. Parsons, J. Appl. Phys., 90, 918 共2001兲. 14. M. Copel, E. Cartier, V. Narayanan, M. C. Reuter, S. Guha, and N. Bojarczuk, Appl. Phys. Lett., 81, 4227 共2002兲. 15. M. Copel, Appl. Phys. Lett., 82, 1580 共2003兲. 16. C. Durand, C. Dubourdieu, C. Vallée, V. Loup, M. Bonvalot, O. Joubert, H. Roussel, and O. Renault, J. Appl. Phys., 96, 1719 共2004兲. 17. C. Durand, C. Vallée, V. Loup, O. Salicio, C. Dubourdieu, S. Blonkowski, M. Bonvalot, P. Holliger, and O. Joubert, J. Vac. Sci. Technol. A, 22, 655 共2004兲. 18. C. Durand, C. Vallée, C. Dubourdieu, E. Gautier, M. Bonvalot, and O. Joubert, J. Vac. Sci. Technol. A, 22, 2490 共2004兲. 19. C. Dubourdieu, J. P. Sénateur, and F. Weiss, Crystal Growth in Thin Solid Films: Control of Epitaxy, M. Guilloux-Viry and A. Perrin, Editors, Research Signpost, Kerala, India, p. 169 共2002兲. 20. M. L. Green, E. P. Gusev, R. Degraeve, and E. L. Garfunkel, J. Appl. Phys., 90,

Journal of The Electrochemical Society, 152 共12兲 F217-F225 共2005兲 2057 共2001兲. 21. Z. H. Lu, J. P. McCaffrey, B. Brar, G. D. Wilk, R. M. Wallace, L. C. Feldman, and S. P. Tay, Appl. Phys. Lett., 71, 2764 共1997兲. 22. N. Rochat, A. Chabli, F. Bertin, M. Olivier, C. Vergnaud, and P. Mur, J. Appl. Phys., 91, 5029 共2002兲. 23. R. Anderson and S. D. Walck, Mater. Res. Soc. Symp. Proc., 480, 480 共1997兲. 24. H. K. Ryu, J. S. Heo, S. I. Cho, and S. H. Moon, J. Electrochem. Soc., 146, 1117 共1999兲. 25. S. Momose, T. Nakamura, and K. Tachibana, Jpn. J. Appl. Phys., Part 1, 39, 5384 共2000兲. 26. T. Nakamura and K. Tachibana, Jpn. J. Appl. Phys., Part 1, 40, 338 共2001兲. 27. S. Momose, T. Shara, T. Nakamura, and K. Tachibana, Jpn. J. Appl. Phys., Part 1, 40, 5501 共2001兲. 28. Handbook of X-ray Photoelectron Spectroscopy J. Chastain, Editor, Perkin-Elmer Corporation, Eden Prairie, MN 共1992兲.

F225

29. N. Bassim, V. Cracium, J. Howard, and R. K. Singh, Appl. Surf. Sci., 205, 267 共2003兲. 30. Electron Energy Loss Spectrocopy in the Electron Microscope, 2nd ed., R. F. Egerton, Editor, Plenum Press, New York and London 共1996兲. 31. M. D. Ulrich, J. E. Rowe, D. Niu, and G. N. Parsons, J. Vac. Sci. Technol. B, 21, 1792 共2003兲. 32. H. B. Michaelson, J. Appl. Phys., 48, 4729 共1977兲. 33. E. H. Nicollian and J. R. Brews, MOS (Metal Oxide Semiconductor) Physics and Technology, John Wiley & Sons, New York 共1992兲. 34. J. McPherson, J.-Y. Kim, A. Shanware, and H. Mogul, Appl. Phys. Lett., 82, 2121 共2003兲. 35. H. Harris, K. Choi, N. Mehta, A. Chandolu, N. Biswas, G. Kipshidze, S. Nikishin, S. Gangopadhyay, and H. Temkin, Appl. Phys. Lett., 81, 1065 共2002兲. 36. I. S. Jeon, J. Park, D. Eom, C. S. Hwang, H. J. Kim, C. J. Park, H. Y. Cho, J. H. Lee, N. I. Lee, and H. K. Kang, Jpn. J. Appl. Phys., Part 1, 42, 1222 共2003兲.