Relationships among equivalent oxide thickness

Relationships among equivalent oxide thickness, nanochemistry, and nanostructure in atomic layer chemical-vapor-deposited Hf–O films on Si S. K. Dey, A. Das, M. Tsai, D. Gu, M. Floyd et al. Citation: J. Appl. Phys. 95, 5042 (2004); doi: 10.1063/1.1689752 View online: http://dx.doi.org/10.1063/1.1689752 View Table of Contents: http://jap.aip.org/resource/1/JAPIAU/v95/i9 Published by the AIP Publishing LLC.

Additional information on J. Appl. Phys. Journal Homepage: http://jap.aip.org/ Journal Information: http://jap.aip.org/about/about_the_journal Top downloads: http://jap.aip.org/features/most_downloaded Information for Authors: http://jap.aip.org/authors

Downloaded 24 Sep 2013 to 202.116.1.148. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jap.aip.org/about/rights_and_permissions

JOURNAL OF APPLIED PHYSICS

VOLUME 95, NUMBER 9

1 MAY 2004

Relationships among equivalent oxide thickness, nanochemistry, and nanostructure in atomic layer chemical-vapor-deposited Hf–O films on Si S. K. Deya) Department of Chemical and Materials Engineering, Department of Electrical Engineering, and Center for Solid State Science, Arizona State University, Tempe, Arizona 85287-6006

A. Das, M. Tsai,b) and D. Gu Department of Chemical and Materials Engineering, Arizona State University, Tempe, Arizona 85287-6006

M. Floyd and R. W. Carpenter Center for Solid State Science, Arizona State University, Tempe, Arizona 85287-6006

H. De Waard, C. Werkhoven, and S. Marcus ASM America, Inc., Pheonix, Arizona 85034-7200

共Received 14 October 2003; accepted 4 February 2004兲 The relationships among the equivalent oxide thickness 共EOT兲, nanochemistry, and nanostructure of atomic layer chemical-vapor-deposited 共ALCVD兲 Hf–O-based films, with oxide and nitrided oxide interlayers on Si substrates, were studied using x-ray photoelectron spectroscopy 共XPS兲, high-resolution transmission electron microscopy 共HRTEM兲, scanning transmission electron microscopy 共STEM兲 in annular dark-field imaging 共ADF兲, and parallel electron energy-loss spectroscopy 共PEELS兲, capacitance–voltage, and leakage-current–voltage measurements. The XPS (Hf 4 f binding energy shift兲 studies indicated the formation of Hf–O–Si bonds in as-deposited amorphous films, the amount of which was influenced by the interlayer composition and annealing conditions. After post-deposition annealing in N2 and O2 , the Hf–O layers were nanocrystalline. Although HRTEM images showed a structurally sharp interface between the Hf–O layer and the interlayer, angle-resolved XPS, ADF imaging, and PEELS in the STEM revealed a chemically diffused HfSiOx region in between. This interdiffusion was observed by the detection of Si 共using Si L edge兲 and Hf 共using Hf O2,3 edge兲 in the Hf–O layer and the interlayer. For an annealed Hf–O/interlayer stack, with an ALCVD target thickness of 4.0 nm for the Hf–O layer on 1.2 nm of nitrided chemical oxide, the experimentally measured EOT and leakage current 共at ⫺1 V兲 were 1.52 nm and ⬃10⫺8 A/cm2 . A three-layer 共1.2 nm interlayer of nitrided chemical oxide/compositionally graded, 2 nm region of HfSiOx /2 nm HfO2 layer兲 capacitor model was used to determine the respective contributions to the measured EOT, and the dielectric permittivity of the interlayer was found to be 6.06. These studies clearly indicate that a total EOT of 1 nm and below is attainable in the Hf–N–O–Si/Si–N–O system. © 2004 American Institute of Physics. 关DOI: 10.1063/1.1689752兴

I. INTRODUCTION

Our recent study of ALCVD ZrO2 /native oxide/Si stack demonstrated that in addition to surface chemical reactions during ALCVD, there are parallel kinetic processes that aid in the nanochemical and nanostructure evolution.7 These processes, during ALCVD and during post-deposition annealing, include SiOx growth at the native oxide/Si interface, interdiffusion of Zr, Si, and O species, formation of Zr–O–Si bond linkages and amorphous Zr silicates through reactions at internal interfaces, and nucleation and growth of ZrO2 . Due to this intermixing, an intermediate layer with graded composition and K is formed, and which must be taken into account to model the measured equivalent oxide thickness (EOTtot) of the stack. In order to quantitatively analyze the EOTtot and to eventually control the electrical properties of a CMOS gate stack based on the HfO2 dielectric, the study of its nanochemistry and nanostructure is imperative. Equipped with the knowledge of the precise thickness and spatial dependence of Si

High permittivity 共K兲 hafnia (HfO2 ) is one of the most promising candidates to replace oxide (SiOx ) or oxynitride (SiOx Ny )-based gate dielectrics in future generation complementary metal-oxide-semiconductor 共CMOS兲 devices.1,2 Typically, HfO2 films derived from atomic layer chemical vapor deposition 共ALCVD兲, using halide and water precursors, have excellent thickness uniformity over ⬃3–5 nm.3 Due to thermodynamic stability and electrical issues,4,5 HfO2 films are deposited on top of a surface layer (t ⬃0.5– 1.2 nm) of SiOx or SiOx Ny rather than directly on Si.6 This becomes an interlayer after Hf–O deposition. a兲

Author to whom correspondence should be addressed; electronic mail: [email protected] b兲 Currently at: Department of Physics, National Sun Yat-Sen University, Kaohsiung 804 Taiwan. 0021-8979/2004/95(9)/5042/7/$22.00

5042

© 2004 American Institute of Physics


Dey et al.

J. Appl. Phys., Vol. 95, No. 9, 1 May 2004

and Hf in the dielectric sublayers, an appropriate equivalent circuit can then be developed. Therefore, in this article, the extent of intermixing of Si and Hf in as-deposited ALCVD Hf–O-based films, on different interlayers and with varying annealing conditions, is discussed first in light of x-ray photoelectron spectroscopy 共XPS兲 data. From this preliminary study, a suitable process recipe was determined. Next, to determine the relationships among the EOTtot , nanochemistry, and nanostructure of ALCVD Hf–O-based films, detailed nanochemical and nanostructure studies 共on a sample that was annealed following this process recipe兲 were carried out using angle-resolved XPS, high-resolution transmission electron microscopy 共HRTEM兲, and analytical electron microscopy in imaging and parallel electron energy-loss spectroscopy 共PEELS兲 modes. Finally, quantum-corrected EOTtot of this annealed sample was estimated from C – V measurements, and using a three-layer capacitance model, that reflects the spatial dependence of composition 共and K兲 in the dielectric sublayers, the respective contributions to EOTtot and the K of the interfacial layer were determined. II. EXPERIMENT

Dielectric films (t⬃3 – 8 nm) in the Hf–O system were deposited at 300 °C on 200 mm p-Si 共100兲 wafers, with interlayers of chemical oxide and oxynitride, by ALCVD 共ASM Pulsar 2000 module兲. Separated by a purge flow of N2 , precursor gases of hafnium tetrachloride (HfCl4 ) and water (H2 O) were sequentially pulsed into the reactor at ⬃1 Torr. This cycle was repeated at a rate of 4.3 s/cycles until the desired film thickness was obtained. Annealing of the as-deposited films was carried out in a tube furnace with flowing nitrogen or oxygen at 900 °C and 1 atm. The XPS surface analysis system 共Kratos XSAM 800兲 used a monochromatic and standard Mg x-ray 共12 KeV兲 source. The mean analysis depth was 3–5 nm. To collect a calibration spectrum, a small piece of silver foil was placed on top of the samples. After the sample was tilted to different angles with respect to the detector, the angle-resolved spectrum was collected.8 From the analysis of these XPS results, a suitable process recipe 共Fig. 1兲 was selected to subject an as-deposited Hf–O-based film, of ALCVD target thickness of 4 nm, to post-deposition annealing treatments. This film was used to carry out the nanochemical and nanostructure analysis by electron microscopy. Cross-sectional TEM specimens were prepared by mechanical polishing and dimpling, followed by Ar ion-beam thinning to electron transparency. All microscopy were done at 200 KeV, with the incident beam along the 关110兴 direction of the substrate, in the plane of various interfaces and multilayers of interest. HRTEM images were used to examine the structure of the multilayers, and annular dark field scanning TEM 共STEM兲 was used to position the probe for EELS nanospectroscopy with Gaussian probe images of full width at half-maximum (FWHM)⫽0.25 nm. The PEELS was done on a JEOL 2010F microscope with a Gatan Enfina spectrometer at a resolution of 1 eV. The electrical and dielectric properties of the film, for which the nanostructure and nanochemistry were analyzed,

5043

FIG. 1. Flow chart for the deposition and annealing of a Hf–O film with ALCVD target thickness of 4 nm on nitrided chemical oxide.

were measured at room temperature. Platinum 共Pt兲 top electrodes 共100 nm thick, 0.0001 cm2兲 were deposited onto the samples by e-beam evaporation. The high-frequency 共100 KHz, acosc 20 mV兲 capacitance–voltage (C – V) and the dc leakage current density–voltage (J L – V) characteristics of the capacitors were measured using a multifrequency LCR meter 共HP impedance analyzer 4284 A兲 and current meter 共Sony Tektronix™ 372兲, respectively. The quantumcorrected EOTtot was estimated using the Hauser program.9 III. RESULTS AND DISCUSSION A. Evidence of the formation of Hf–O–Si bonds in as-deposited and annealed ALCVD Hf–OÕ„SiOx or SiOx Ny …ÕSi

An HRTEM image of a typical as-deposited specimen is shown in Fig. 2. The ALCVD Hf–O layer was ⬃3 nm thick, and the chemical oxide interlayer was ⬃1.2 nm thick. Note that both layers were amorphous in the as-deposited condition. The peak positions of the XPS Hf 4 f doublet 共due to spin–orbit coupling兲 for a series of as-deposited samples

FIG. 2. HRTEM bright-field image of an as-deposited ALCVD Hf–O on chemical Si–O/共100兲 vicinal Si surface. Note the near atomically smooth interfaces, compared, for example, to sputtered thin films.


5044

Dey et al.


TABLE I. Hf XPS peak positions of as-deposited samples after deconvolution of the Hf 4 f doublet. Sample HfO2 filma HfSiO4 filma 3.0 nm Hf–O⫹1.2 nm SiOx Ny 8.0 nm Hf–O⫹1.2 nm SiOx Ny 3.0 nm Hf–O⫹1.2 nm chemical oxide 8.0 nm Hf–O⫹1.2 nm chemical oxide a

Hf 4 f 5/2 共eV兲

Hf 4 f 7/2 共eV兲

18.88 20.24 19.71 19.52 19.86 19.67

17.16 18.92 18.03 17.8 18.18 18.02

References 11 and 12.

with different thickness 共3 and 8 nm, respectively兲 and different interlayers 共chemical oxide and oxynitride兲 are shown in Table I. Due to the formation of Hf–O–Si bonds in Hf silicate during ALCVD, the Hf 4 f peaks shift towards higher binding energies with respect to pure HfO2 ; the ionic character of the more ionic cation 共Hf兲 increases in Hf silicate due to charge transfer effects.10 In addition, the peak positions indicate the formation of nonstoichiometric Hf silicates. Note that the higher degree of peak shifts in Hf–O layers deposited on an interlayer of chemical oxide illustrate that these Hf–O layers exhibit a higher tendency towards Hf–O–Si bond formation than Hf–O layers deposited on an interlayer of oxynitride. Thus, the species in the Hf–O/SiOx Ny stack is less prone to diffusion and interfacial reactions. Comparing films with different thicknesses of the Hf–O layer, the 3 nm film shows a higher 4 f peak shift with respect to the 8 nm film. Since the mean sampling depth of photoelectrons is 3–5 nm, in the 3 nm film, the signals are predominantly from the interfacial region and the extent of Hf–O–Si bond formation appears to be more. In contrast, the extent of Hf–O–Si bond formation appears to be less in the 8 nm film because the signals are collected from a depth of 3–5 nm from the free surface. The annealing environment also shows a significant effect on the extent of Hf–O–Si bond formation. In Fig. 3, the highest Hf 4 f peak shifts are observed in O2 -annealed sample compared to as-deposited and N2 -annealed samples with an interlayer of chemical oxide. Note that, due to the absence of Hf 4 f peaks at 14.9 and 16.5 eV in annealed

FIG. 3. Hf 4 f XPS spectra of as-deposited and annealed samples 共3 nm Hf–O on chemical oxide/Si兲.

FIG. 4. Angle-resolved Hf 4 f XPS spectra of an oxygen annealed sample 共3 nm Hf–O on chemical oxide/Si兲. The various sample tilt angles are indicated.

samples,13 the formation of Hf–Si bonds is discounted. In an interfacial chemistry study of HfO2 /SiO2 using synchrotron radiation XPS, Renault et al.14 also found an intermediate layer of HfSiOx and no evidence of hafnium silicide. Although, following the desorption of SiO gas in reducing environments, the formation of hafnium silicide has been reported,12 the absence of hafnium silicide in the current ALCVD Hf–O samples 共even after annealing at 900 °C in N2 ) may be due to the presence of residual oxygen in the nitrogen gas.13,15 Figure 4 shows the angle resolved Hf 4 f doublet peaks for an oxygen-annealed Hf–O film 共3 nm兲 on chemical oxide interlayer. Note that a higher tilt angle of the sample represents signals that are predominantly from the free surface rather than the interfacial region. At 0° tilt, the Hf 4 f chemical shift is higher with respect to signals collected at higher tilt angles; the peak positions being similar for 40° and 60° tilts. These results indicate that the extent of Hf–O–Si bond formation is higher near the Hf–O/interlayer interface. B. Nanostructure and nanochemistry of annealed ALCVD Hf–OÕNitrided SiOx ÕSi

The nanostructure of the post-deposition annealed 共recipe in Fig. 1兲 film, with a target ALCVD Hf–O thickness of 4 nm, is shown in Fig. 5. The Hf–O layer is polycrystalline and had no particular orientation relationship with the substrate from grain to grain, indicating that nucleation did not consistently occur at the Hf–O/interlayer interface. In addition, the large grain size relative to the film thickness is attributed to the low nucleus density relative to the film thickness. Diffraction data showed that the monoclinic structure was present in this area of the film, which is supported by x-ray diffraction studies 共not shown here兲 and is consistent with reports of other groups.16 The 1.2-nm-thick interlayer did not crystallize during annealing. Figure 6 shows an annular dark-field 共ADF兲 STEM image of an adjacent area, marked to show regions used to acquire EELS spectra. The contrast variations in Fig. 6 correspond mainly to composition differences among the layers; the brightest region was Hf rich (Z⫽72 for Hf兲, and the Si-rich interlayer, with the smallest average atomic number, was the darkest. Energy loss nanospectra, taken from points



Dey et al.

5045

FIG. 5. HRTEM showing multilayer nanostructure after annealing.

a, b, c, and d in Fig. 6, are shown in Fig. 7. In Fig 7共A兲, the Hf O 2,3 peak at about 47 eV in the low-loss region is of most interest. Notice that this edge is not present in spectrum a, collected from the substrate just adjacent to the interlayer, but it appears in spectra b, c, and d with increasing strength. The Si-L2,3 edge at ⬃100 eV loss is shown in Fig. 7共B兲. It is strongest in spectrum a, from the substrate, and decreases successively in spectra b and c, and is not observable in d, from near the middle of the Hf–O layer. The spectral region containing the oxygen and nitrogen K edges is shown in Fig. 7共C兲. Neither of these edges is present in curve a, from the substrate adjacent to the intermediate layer. Strong O K edges occur in the other three spectra, as expected, and a weak N K edge is seen in curve b, from the interlayer. This small amount of nitrogen is present because the chemical oxide interlayer was plasma nitrided before HfOx deposition, and it did not detectably migrate from this region during annealing. However, no nitrogen was detected in the Hf–O layer after a post-deposition annealing in nitrogen atmosphere. Further analysis of the spectra from b showed that the interlayer was an amorphous suboxide (SiOx ) with x ⬃0.35, similar to the native oxide-based interfacial layer under high-K ZrO2 films.7 Nanospectra for points b and c 关Figs. 7共A兲 and 7共B兲兴 indicated that significant Si and Hf interdiffusion occurred across the Hf–O/interlayer interface. Other

FIG. 6. ADF STEM image showing probe locations for spectra shown in Fig. 7.

FIG. 7. A. EELS nanospectra showing the valence and Hf O2,3 loss regions. B. EELS nanospectra showing the Si L region. C. EELS nanospectra showing nitrogen and oxygen K-edge regions.

nanospectra 共not shown兲 showed that Si had diffused into the Hf–O layer for ⬃2 nm beyond interlayer/Hf–O interface, so that only the outer region of this layer was close to stoichiometric HfO2 composition. Similarly, Hf diffused into the 1.2 nm 共suboxide兲 interlayer during annealing. The probe size (FWHM⫽0.25 nm) broadens the observed chemical distributions, and convolution of the probe with a step function chemical distribution broadens the distribution to 0.5 nm. This indicates that interdiffusion did occur on a length scale of 0.5 nm. Analysis showed that the composition of point c, 1.2 nm from the Si/interlayer interface, is closer to hafnon (HfSiO4 ) than to HfO2 . The Si content linearly decreases to zero 共undetectable兲 halfway 共2 nm兲 into the Hf–O layer,


5046

Dey et al.


tain significant concentrations of Hf and Si 关cf. Fig. 7共B兲兴 but they do not correspond to equilibrium hafnium dioxide (HfO2 ) or hafnon (HfSiO4 ).

C. The effect of nanostructure and nanochemistry of annealed ALCVD Hf–OÕnitrided SiOx ÕSi on EOT

FIG. 8. Energy loss spectra from bulk standard HfO2 and HfSiO4 . Spectra displaced vertically for visibility.

which is useful information to model the dielectric behavior of the layer for this chemically diffuse region. Figure 8 shows the low-loss regions of EELS spectra from bulk standard specimens of hafnium dioxide and hafnium silicate 共hafnon兲. These spectra, with much larger numbers of counts because they are from bulk specimens, are useful for comparison with the nanospectra from the multilayers shown in Fig. 7共A兲. The bulk spectra show that the intensity of the first low-loss peak in the dioxide at about 16 eV is larger than the broader adjacent valence peak with a maximum at about 27 eV. This intensity asymmetry is reversed in the hafnon spectra. The Hf O 2,3 regions are similar for both materials, including the energy loss 共48 eV兲 of the delayed maximum. The shape of the EELS spectrum d in Fig. 7共A兲 agrees very well with the HfO2 spectrum of Fig. 8, indicating that this region of the layer is HfO2 . In spectra from b and c 关Fig. 7共A兲兴, the Hf O 2,3 delayed maximum agrees with both standard spectra 共it occurs at the same energy loss, within 1 eV, in both兲 but the valence band intensity distribution differs from both standards. These regions con-

EOTtotal t IL ⫽ ⫹ ␧ 0 K SiO2 ␧ 0 K IL

冕

IL⫹ HfSiOx

IL

冋

␧ 0 K HfSiO4 ⫹

冉

1 K HfO2 ⫺K HfSiO4 t HfSiOx

where ␧ 0 is the permittivity of free space, and K and t are the permittivity and thickness, respectively, of various regions indicated in the subscripts. Following integration of Eq. 共2兲 and some rearrangement, the total EOT 共i.e., EOTIL ⫹EOTHfSiOx ⫹EOTHfO2 ) is given by EOTtotal⫽

K SiO2 K IL ⫹

t IL⫹

K SiO2 K HfO2

K SiO2

␥

t HfO2 ,

冋

ln 1⫹ ␥

t HfSiOx K HfSiO4

册

共3兲

where ␥ ⫽(K HfO2 ⫺K HfSiO4 )/t HfSiOx . The measured C – V characteristic of the MOSCAP (Pt/Hf–O/IL/p-Si stack兲 is plotted in Fig. 9. The applied bias is referred to the top Pt electrode contact. From the

The three distinct regions of the annealed stack 共targeted 4 nm ALCVD Hf–O layer/1.2 nm interlayer 共IL兲 of nitrided SiOx /p-Si) that contribute to the total equivalent SiO2 thickness 共EOT兲 are 共a兲共b兲

共c兲

1.2-nm-thick, bottom IL of nitrided SiOx , 2-nm-thick, intermediate, and linearly graded HfSiOx region, with composition ranging from HfSiO4 共at the IL/HfSiOx interface兲 to HfO2 , up to a depth of half the thickness of the original Hf–O layer, and HfO2 composition in the top 2 nm of the original Hf–O layer.

Therefore, the aforementioned stack is modeled as three capacitors (CIL , CHfSiOx , and CHfO2 ) in series, for which the total capacitance (C tot) can be written as 1 C total

⫽

1 1 1 ⫹ ⫹ . C IL C HfSiOx C HfO2

共1兲

Since the compositional gradient within the 2-nm-thick HfSiOx region is linear with respect to Hf or Si concentration, the permittivity 共K兲 can be assumed to be linearly graded from K HfSiO4 共at the IL/HfSiOx interface兲 to K HfO2 共at the HfSiOx /HfO2 interface兲. Therefore, Eq. 共1兲 may be written in terms of the total EOT as follows:

冊

共 x ⬘ ⫺t IL兲

册

dx ⬘ ⫹

t HfO2 ␧ 0 K HfO2

,

共2兲

measured capacitance density in accumulation (C acc /A), the total quantum-corrected EOTtot 共i.e., EOTtot ⫽␧0KSiO2 /(C acc /A) was found to be 1.52 nm. In Eq. 共3兲, all the quantities are known except the permittivity (K II) of the IL. With K HfSiO4 ⫽11,12,15 K SiO2 ⫽3.9,12 and K HfO2 ⫽26.1 from existing data17 with corresponding ␥ of 7.5, and t IL ⫽1.2 nm, t HfSiOx ⫽2 nm, and t HfO2 ⫽2 nm from current nanochemical and nanostructure observations, K IL is calculated to be 6.06. The higher K IL value for the interlayer of SiOx , with respect to stoichiometric SiO2 , is indicative of a variety of oxidation states of Si,7 coupled with some unmeasured quantities of N 共due to nitridation of SiOx ) and Hf 共due to in-diffusion in SiOx ) in the IL. Figure 10 is the calculated plot of EOT versus the target ALCVD HfO2 thickness. For a target HfO2 thickness of 4


Dey et al.


FIG. 9. The C – V curve for a MOS 共100 nm Pt/2 nm HfO2 /2 nm HfSiOx /1.2 nm IL of nitrided SiOx /Si) configuration.

nm on a 1.2 nm IL, the contributions of EOTIL , EOTHfSiOx , and EOTHfO2 to the experimentally determined EOTtot of 1.52 nm are 0.77, 0.45, and 0.3 nm, respectively. These contributions indicate that a total EOT of 1 nm and below is attainable in the Hf–N–O–Si/Si–N–O dielectric stack through dielectric thickness control and appropriate annealing. Moreover, for a target HfO2 thickness ⬍2.0 nm, the calculated plot exhibits a negative inflection or concavity, which is indicative of the breakdown of a bilayer capacitance model and a need for the use of a trilayer model given by Eq. 共3兲. In addition, the increasing negative slope and decreasing EOT with decreasing target ALCVD Hf–O thickness suggests that although the effective K reduces with respect to the K HfO2 , an intermediate HfSiOx layer with higher K 共compared to SiOx ) develops.18,19 This observation is a direct consequence of the aforementioned chemical interdiffusions, which only manifests at a low Hf–O thickness. Note that these results are also consistent with experimental data reported by Dey et al.7 for a target ALCVD ZrO2 thickness

FIG. 10. Calculated EOT versus the target ALCVD HfO2 thickness. Note the various contributions to the experimentally determined EOTtot of 1.52 nm for a target 共4 nm兲 HfO2 thickness are identified by arrows and given by Eq. 共3兲.

5047

⬍5.0 nm. In these EOT versus the target ALCVD dielectric 共Hf–O and Zr–O兲 thickness plots, the deviation from linearity with decreasing dielectric thickness occurs at a higher value in the ZrO2 system 共Fig. 1 in Ref. 7兲 compared to the HfO2 system 共Fig. 9兲. This is apparently due to more interdiffusion in the former. Finally, some comments on the leakage current density (J L ) of the Pt/Hf–O/IL/p-Si stack. The J L at ⫹1 and ⫺1 V were found to be 2⫻10⫺7 and 8⫻10⫺8 A/cm2 , respectively. The low magnitude of J L is attributed to the reduced probability of tunneling through the HfSiOx layer and tunneling through defects within the bandgap of the HfO2 layer.20,21 However, the asymmetry in the magnitude of J L for positive and negative gate voltages stems from the asymmetry of the band alignment and band bending at the Si/SiOx and Pt/HfO2 interfaces, and the consequent asymmetry in the transmission probability.20,22,23 However, in such transition-metal-oxide systems, both the valence and conduction band components may contribute to J L . 24 Therefore, to quantitatively model the higher J L for the positive gate voltages 共in inversion兲 over the corresponding negative gate voltages 共in accumulation兲, the additional contribution from the valence band component must be taken into account.25

IV. SUMMARY

The relationships among the measured EOTtot , nanochemistry, and nanostructure of ALCVD Hf–O-based films, with oxide and nitrided oxide interlayers on p-Si substrates, were studied using XPS, HRTEM, STEM 共ADF and PEELS兲, C – V, and J L – V measurements. The as-deposited amorphous films showed a tendency towards the formation of Hf–O–Si bonds during ALCVD, the amount of which was enhanced by an interlayer of chemical oxide and postdeposition O2 annealing. After post-deposition annealing 共in N2 at 600 °C followed by O2 at 700 °C for 1 min each兲, the Hf–O layers were nanocrystalline. Although HRTEM images showed a near atomically smooth and structurally sharp interface between the Hf–O layer and the interlayer, angleresolved XPS, ADF imaging, and PEELS 共using Si-L and Hf-O2,3 edges兲 in the STEM revealed a chemically diffused HfSiOx region in between. For an annealed Hf–O/interlayer stack, with an ALCVD target thickness of 4.0 nm for the Hf–O layer on 1.2 nm of nitrided chemical oxide, the experimentally measured EOT and leakage current 共at ⫺1 V兲 were 1.52 nm and ⬃10⫺8 A/cm2 . A three-layer capacitor model indicated that the contributions to the measured EOTtot from 1.2 nm interlayer of nitrided chemical oxide, 2 nm HfSiOx region in the lower half of the targeted Hf–O layer, and 2 nm HfO2 layer in the upper half of the targeted Hf–O layer were 0.3, 0.45, and 0.77 nm, respectively. This indicates that a total EOT of 1 nm and below is attainable in the Hf–N–O– Si/Si–N–O dielectric system through dielectric thickness control and appropriate annealing. Finally, the dielectric permittivity of the interlayer was found to be 6.06, which is indicative of a variety of oxidation states of Si, coupled with very small quantities of N 共due to nitridation of SiOx ) and Hf 共due to in-diffusion in SiOx ) in the interlayer.


5048


ACKNOWLEDGMENTS

The authors would like to acknowledge the help of Tim Karcher and David Wright of the center for solid state sciences, Arizona State University. This research was supported in part by the College of Liberal Arts and Sciences at ASU, US Dept of Energy under grant No. DE-FG03-94ER45510, and NSF-ECS 共contract no. ECS-0000121兲. 1

G. D. Wilk, R. M. Wallace, and J. M. Anthony, J. Appl. Phys. 89, 5243 共2001兲. 2 B. H. Lee, L. Kang, R. Neih, W.-J. Qi, J. C. Lee, Appl. Phys. Lett. 76, 1926 共2000兲. 3 M. Ritala, M. Leskala, L. Niinisto, T. Prohaska, Friedbacher, and M. Grasserbauer, Thin Solid Films 72, 250 共1994兲. 4 Y.-S. Lin, R. Puthenkovilakam, and J. P. Chang, Appl. Phys. Lett. 81, 2041 共2002兲. 5 M. Cho, J. Park, H. B. Park, C. S. Hwang, J. Jeong, K. S. Hyun, Y.-W. Kim, C.-B. Woh, and H.-S. Kang, Appl. Phys. Lett. 81, 3630 共2002兲. 6 M. L. Green, M.-Y. Ho, B. Busch, G. D. Wilk, T. Sorsch, T. Conard, B. Brijs, W. Vandervorst, P. I. Raisanen, D. Muller, M. Bude, and J. Grazul, J. Appl. Phys. 92, 7168 共2002兲. 7 S. K. Dey, C.-G. Wang, D. Tang, M. J. Kim, R. Carpenter, C. Werkhoven, E. Shero, J. Appl. Phys. 93, 4144 共2003兲. 8 G. Renault, L. G. Gosset, D. Rouchon, and A. Ermolief, J. Vac. Sci. Technol. A 20, 1867 共2002兲. 9 J. R. Hauser and K. Ahmed, in Characterization and Metrology for ULSI Technology: 1998 International Conference, edited by D. G. Seiler, A. C.

Dey et al. Diebold, W. M. Bullis, T. J. Shaffner, R. Mcdonald, E. J. Walters 共The American Institute of Physics, Melville, NY, 1998兲, pp. 235–239. 10 M. J. Guittet, J. P. Crocombette, and M. Gautier-Soyer, Phys. Rev. B 63, 125117 共2001兲. 11 V. Cosnier, M. Olivier, G. Theret, and B. Andre, J. Vac. Sci. Technol. A 19, 2267 共2001兲. 12 G. D. Wilk, R. M. Wallace, and J. M. Anthony, J. Appl. Phys. 87, 484 共2000兲. 13 S. Sayan, E. Garfunkel, T. Nishimura, W. H. Schulte, T. Gustafsson, and G. D. Wilk, J. Appl. Phys. 94, 928 共2003兲. 14 O. Renault, D. Samour, J.-F. Damlencourt, D. Biln, F. Martin, S. Marthon, N. T. Barett, and P. Besson, Appl. Phys. Lett. 81, 3627 共2002兲. 15 A. I. Kingon, J. P. Maria, and S. K. Streiffer, Nature 共London兲 406, 1032 共2000兲. 16 H. Kim, P. C. McIntyre, and K. C. Saraswat, Appl. Phys. Lett. 82, 106 共2003兲. 17 S. K. Dey, unpublished data on pure ALCVD HfO2 films. 18 G. Lucovsky and G. B. Rayner Jr., Appl. Phys. Lett. 77, 2912 共2000兲. 19 H. A. Kurtz and R. A. B. Devine, Appl. Phys. Lett. 79, 2342 共2001兲. 20 M. Houssa, M. Naili, C. Zhao, H. Bender, M. M. Heyns, and A. Stesmans, Semicond. Sci. Technol. 16, 31 共2001兲. 21 K. Masuno, T. Murakami, and S. Waku, Ferroelectrics 3, 315 共1972兲. 22 S. K. Dey, J. J. Lee, and P. Alluri, Jpn. J. Appl. Phys. 34, 3142 共1995兲. 23 R. S. Johnson, G. Lucovsky, and I. Baumvol, J. Vac. Sci. Technol. A 19, 1353 共2001兲. 24 S. K. Banerjee, University of Texas at Austin 共private communication兲. 25 S.-H. Lo, D. A. Buchanan, Y. Taur, and W. Wang, IEEE Electron Device Lett. 18, 209 共1997兲.
