Structural and electrical properties of atomic layer

Structural and electrical properties of atomic layer deposited Al-doped ZrO2 films and of the interface with TaN electrode S. Spiga, R. Rao, L. Lamagna, C. Wiemer, G. Congedo, A. Lamperti, A. Molle, M. Fanciulli, F. Palma, and F. Irrera Citation: Journal of Applied Physics 112, 014107 (2012); doi: 10.1063/1.4731746 View online: http://dx.doi.org/10.1063/1.4731746 View Table of Contents: http://scitation.aip.org/content/aip/journal/jap/112/1?ver=pdfcov Published by the AIP Publishing

[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP: 151.100.120.139 On: Mon, 13 Jan 2014 15:45:29

JOURNAL OF APPLIED PHYSICS 112, 014107 (2012)

Structural and electrical properties of atomic layer deposited Al-doped ZrO2 films and of the interface with TaN electrode S. Spiga,1,a) R. Rao,2 L. Lamagna,1 C. Wiemer,1 G. Congedo,1 A. Lamperti,1 A. Molle,1 M. Fanciulli,1,3 F. Palma,2 and F. Irrera2 1

Laboratorio MDM, IMM-CNR, Via C. Olivetti 2, 20864 Agrate Brianza (MB), Italy Information, Electronics and Communications Department and IUNET (Italian University NanoElectronics Team), Sapienza University, Via Eudossiana 18, 00184 Roma, Italy 3 Dipartimento di Scienza dei Materiali, Universita` degli studi di Milano Bicocca, Via R. Cozzi 53, 20125 Milano, Italy 2

(Received 13 January 2012; accepted 31 May 2012; published online 9 July 2012) Al-doped ZrO2 (Al-ZrO2) films deposited by atomic layer deposition onto silicon substrates and the interface with the TaN metal gate are investigated. In particular, structural properties of asgrown and annealed films in the 6–26 nm thickness range, as well as leakage and capacitive behavior of metal-oxide-semiconductor stacks are characterized. As-deposited Al-ZrO2 films in the mentioned thickness range are amorphous and crystallize in the ZrO2 cubic phase after thermal treatment at 900 C. Correspondingly, the dielectric constant (k) value increases from 20 6 1 to 27 6 2. The Al-ZrO2 layers exhibit uniform composition through the film thickness and are thermally stable on Si, whereas chemical reactions take place at the TaN/Al-ZrO2 interface. A transient capacitance technique is adopted for monitoring charge trapping and flat band instability at short and long time scales. The role of traps nearby the TaN/Al-ZrO2 interface is discussed and compared with other metal/high-k oxide films. Further, analytical modeling of the flat band voltage shift with a power-law dependence on time allows extracting features of bulk traps close to the silicon/oxide interface, which exhibit energy levels in the 1.4–1.9 eV range above the valence band C 2012 American Institute of Physics. [http://dx.doi.org/10.1063/1.4731746] of the Al-ZrO2. V I. INTRODUCTION

In the wide scenario of high dielectric constant (high-k) materials, zirconium oxide (ZrO2) has been attracting most attention in the last decade for various applications in micro and nanoelectronics devices.1,2 ZrO2 features a relatively high k, fairly 20 in the stoichiometric and thermodynamically stable configuration of the monoclinic phase;2,3 k value can be increased by stabilization of crystallographic cubic or tetragonal polymorphs of ZrO2. In particular, by theoretical calculations, k values of 36 and 47 were reported for the cubic fluorite and the tetragonal polymorphs, respectively.3 Experimental and theoretical evidences of stabilization of the high-k values in polymorphs of ZrO2 have been reported by several authors by proper doping of the oxide by incorporating elements such as La, Ge, Al, Y, and/or by controlling oxygen vacancies incorporation during the film deposition.4–11 It is worth noting that among the various dopant elements used to stabilize the different polymorphs of ZrO2, Al is of particular interest, since it is more compatible than rare earth elements with the complementary metal oxide semiconductor (CMOS) technology. Recently, pure and doped ZrO2 films are emerging as potential candidate materials in various types of volatile and non-volatile memory devices, as well as highperformance transistors.1,12–20 ZrO2 has been proposed as storage layer in FLASH charge trapping memory either as a single layer,13,14 as well as in conjunction with Si3N4 in the dualcharge storage level scheme to increase the storage capability a)

Author to whom correspondence should be addressed. Electronic mail: [email protected].

0021-8979/2012/112(1)/014107/9/$30.00

of each single cell.15 Recently, it has been used in metal-insulator-metal (MIM) capacitors for dynamic random access memory (DRAM)10,15 and for crossbar write-once read-many-times (WORM) applications.17 Furthermore, pure and doped ZrO2 is attracting an increasing interest for resistive memories (ReRAM), where the elemental cell is a MIM structure, and the film resistance changes under the effect of an external bias.18–21 In particular, doping with trivalent atoms, such as Al and La, has been demonstrated to reduce the dispersion of programming voltages and currents.21 Al-doped ZrO2 (Al-ZrO2) has been also investigated for logic devices in combination with Si and III-V substrates.9,12 All these considerations prompted us to investigate AlZrO2 in terms of structural properties, thermal stability, conduction and trapping, as well as the interface with silicon and metal electrodes. In particular, as-grown and annealed Al-ZrO2 films were thoroughly characterized to evaluate the structural and chemical properties after high-temperature thermal treatment at 900 C and to monitor the corresponding variation in the dielectric constant value. Further, ad hoc electrical studies were performed on samples with W/TaN gate electrode, which is of current interest for practical applications since features high work function, is compatible with CMOS processes, and is already used in microelectronics for both memory devices and high performance transistors.22–25 The TaN/Al-ZrO2 interface was studied extensively to evaluate possible chemical reactions after thermal treatment and to characterize electrical properties. In particular, we adopted a transient technique for the electrical characterization of the metal/oxide interface in terms of charge trapping and

112, 014107-1

C 2012 American Institute of Physics V


014107-2

Spiga et al.

consequent fluctuation of the flat band voltage. Indeed by using conventional steady-state techniques, it is impossible to quantify charge trapping at the electrode/oxide interface, since electron capture is extremely fast and, in general, this process is hidden by other transient phenomena with longer characteristic times. This fact holds also when the amount of trap centers at the metal/oxide interface is higher than at the semiconductor/oxide interface. The transient capacitance technique used in this work monitors trapping behavior on a timescale from one hundred of microseconds to hundreds of seconds, and an analytical model allows extraction of the trap features. II. FILM DEPOSITION AND EXPERIMENTAL METHODS

Al-ZrO2 films with thickness in the 6–26 nm range were grown by atomic layer deposition (ALD) at 300 C in a Savannah 200 reactor (Cambridge Nanotech, Inc.) on p-type Si(100) wafers (substrate Boron doping around 4.7 1015 cm3) featuring 1.5 nm thick-SiO2 chemical oxide films grown on the top. For the ALD process, Al(CH3)3 was used in combination with the Zr precursor (CpMe)2Zr(OMe)Me (also called ZrD-04), and O3 was chosen as oxygen source. ZrD-04 bubbler was heated at 110 C; N2 was employed as carrier gas and as inert gas for purging the chamber. The ALD cycle ratio was set to Al2O3:ZrO2 ¼ 1:10, as sketched in Fig. 1(a); therefore, one ALD super-cycle was composed by one Al2O3 step and ten ZrO2 steps. Pulse lengths were set as following: Al(CH3)3 0.15 s, purge 6 s, O3 0.015 s, purge 6 s, ZrD-04 10 s, purge 6 s, O3 0.015 s, purge 6 s. A M2000-F spectroscopic ellipsometer (J.A. Woollam Inc.) was employed to evaluate the stack thickness. The beam incidence angle and photon energy range were set at 70 and

FIG. 1. (a) ALD super-cycle structure used for the deposition of Al-ZrO2 films. (b) Linear relationship between the Al-ZrO2 film thickness and the number of ALD super-cycles.

J. Appl. Phys. 112, 014107 (2012)

1–5 eV, respectively. In Fig. 1(b), the linear relation between the number of ALD super-cycles and film thickness is shown. As-grown films were subjected to a post deposition annealing (PDA) at 900 C for 60 s in N2, leading to film crystallization. As-deposited and annealed Al-ZrO2 films were characterized by x-ray photoelectron spectroscopy (XPS), x-ray reflectivity (XRR), grazing incidence x-ray diffraction (GIXRD), and time of flight secondary ion mass spectrometry (ToF-SIMS). XPS was generated by a standard Mg Ka (1253.6 eV) source with a take-off-angle (TOA) of 37 and 70 . Sufficiently thick films (12 nm) were used to avoid interface effects. Photoelectrons were recorded in an Omicron EA 125 HR analyzer (endowed with an energy resolution of 0.6 eV and with an angle acceptance of 68 in the high-resolution mode) using a pass energy of 20 eV during acquisition of single XPS lines. XRD and XRR were performed with a laboratory diffractometer equipped with a copper sealed tube, a scintillator detector, a position sensitive detector, and a four circles goniometer.26 XRR data were acquired in Bragg-Brentano configuration with a step of the incidence angle of 0.01 . XRD data were acquired in grazing incidence geometry and collecting the diffracted radiation at wide angle on a position sensitive detector with a 2H angular resolution of 0.029 . Experiments were also carried out with synchrotron radiation in order to check the presence of diffracted peaks belonging only to tetragonal phase of the ZrO2. ToF-SIMS depth profiling was performed in a ION-TOF IV instrument (Ion Tof GmbH) using 500 eV Csþ ions for sputtering a 200 200 lm2 area and 25 keV Gaþ ions for analyzing a 50 50 lm2 area, centered on the sputtered zone, operating in dual beam interlaced mode. Secondary ions were collected in negative polarity. All the secondary ion intensities were normalized to the 30Si signal intensity in bulk Si. In order to assess the electrical properties, three different types of metal/Al-ZrO2/SiO2/p-Si capacitors were characterized: samples S1 featured W/TaN metal gate and 900 C annealed Al-ZrO2; samples S2 featured Al metal gate and 900 C annealed Al-ZrO2; samples S3 featured Al metal gate and as-deposited Al-ZrO2. As for W/TaN metal gates, 15 nm-thick TaN followed by 30 nm thick layer of W were deposited on annealed Al-ZrO2 by sputtering and followed by a second annealing step in N2 at 900 C. The W/TaN metal gate area (8 104 cm2) was defined by lift-off procedure. A 100 nm thick Al film was deposited on the other side of wafer as back metal contact. All the capacitors were finally annealed in forming gas (4% H2, 96% Ar, pressure: 200 mbar) at 400 C for 15 min to passivate oxide/silicon interface and improve the electrode contact resistance. The thermal treatments used for TaN, including the forming gas step, are often used during device integration and it is, therefore, important to measure devices properties subjected to the required thermal steps. Additional dedicated samples were prepared without patterning the W/TaN layers and keeping experimental conditions and thermal budget as used for the capacitors fabrication. These W/TaN/Al-ZrO2 stacks were analyzed by XRR, GIXRD, and ToF-SIMS to address the structural/chemical properties of the electrode and oxide/metal interface, and the


014107-3

Spiga et al.

J. Appl. Phys. 112, 014107 (2012)

FIG. 2. Al 2p3/2, Zr 3d, and O 1s XPS lines taken on a 12 nm thick Al-ZrO2/Si sample with a TOA ¼ 70 . XPS counts of the lines have been normalized to the respective atomic sensitivity factors.

results were correlated to the electrical ones obtained on capacitors. Finally, Al metal gates (area of 8 104 cm2) were also deposited at room temperature by thermal evaporation through a shadow mask, on both as-deposited and annealed Al-ZrO2 films, in order to evaluate the film k value before and after PDA. III. RESULTS AND DISCUSSION A. Film composition and structural properties

(Fig. 3) depth profiles of 26.4 nm thick Al-ZrO2 films, as-deposited and after PDA at 900 C, were acquired. Uniform chemical composition is found over the film thickness, as shown by both the Al and Zr related signals constant intensity value, without significant difference before and after PDA (Figs. 3(a) and 3(b)). Moreover, no Si diffusion from the substrate is detected in films subjected to PDA, as evidenced by the Si-related signals, which are only marginally smeared out at the interface with the oxide. This finding proves a good thermal stability at 900 C of the Al-ZrO2 films upon Si/SiO2.

We studied the chemical properties and thermal stability at 900 C of Al-ZrO2 layers (before metal gate deposition) by XPS and ToF-SIMS in order to elucidate the oxide film composition and the compositional depth profiling of the Al-ZrO2/ Si stack. To assess the oxide composition, the Al 2p3/2, Zr 3d, and O 1s XPS lines were recorded from a 12 nm-thick AlZrO2 film as reported in Fig. 2. XPS lines were fitted using pseudo-Voigt functions (products of Lorentzian and Gaussian functions). A doublet of Voigt functions with an energy separation of 2.4 eV was used for the Zr 3d line to account for the spin orbit splitting. The O 1s line was decomposed in two components, one at lower binding energy, BE ¼ 531.3 eV which is assigned to O bonding in the Al-ZrO2, and one at higher binding energy, BE ¼ 532.7 eV which is assigned to hydroxide formation at the surface level. The latter assignment is supported by XPS data taken at TOA ¼ 37 , i.e., for an enhanced sensitivity to the sample surface, where the ratio between the two O 1s components is comparatively higher (larger contribution from the hydroxide at the surface level) than that extracted for a TOA ¼ 70 . After normalization to the relevant atomic sensitivity factors (ASF),27 the relative concentration of each elemental constituent (Table I) was deduced from the XPS lines in Fig. 2 by disregarding the hydroxide contribution to the O 1s line. It can be noticed that the Al concentration in the film amounts to 9.7%. To verify the compositional uniformity throughout the overall Al-ZrO2 film (without metal gate on top), ToF-SIMS TABLE I. Atomic concentrations extracted from the XPS lines shown in Fig. 2 upon normalization on the relevant ASF values. The percentages were computed by disregarding the hydroxide component in the O 1s line.

Al Zr O

ASF

%

0.185 2.1 0.66

9.7 30.0 60.3

FIG. 3. ToF SIMS depth profiles of as deposited (a) and 900 C annealed (b) 26.4 nm thick Al-ZrO2 films on Si. (c) ToF-SIMS depth profiles of the W/TaN/Al-ZrO2 stack after annealing at 900 C (full symbols), with focus on the TaN/Al-ZrO2 interface. The TaO2 signal from as-deposited stack (open cyan symbols) is also shown as a reference.


014107-4

Spiga et al.

Finally, ToF-SIMS profiles were acquired for the W/TaN/AlZrO2/SiO2/Si stack (as deposited and after PDA) to study TaN/Al-ZrO2 interface. The profiles drawn in Fig. 3(c) indicate that Al-ZrO2 uniformity is preserved after metal gate deposition and thermal treatments, and that the metal/oxide interface is quite sharp. Nevertheless, after annealing the increased intensity of TaO2 signal within the TaN layer indicates an oxygen diffusion, which ultimately chemically modifies the metal itself as well as TaN/Al-ZrO2 interface. As a result of the described chemical reaction, defects at the interface could have been induced with annealing. It is worth noting that TaO2 relative maximum in intensity at W/TaN interface is due to matrix effect and does not reflect TaO2 real increment at the interface with respect to the bulk. The influence of Al incorporation on the structural properties of the Al-ZrO2 films was investigated by means of GIXRD. In the investigated thickness range (6-26 nm), the asgrown films are amorphous, while GIXRD analysis performed after annealing at 900 C revealed the cubic phase only, with no sign of monoclinic phase. The diffraction pattern of the 24.4 nm thick sample is shown in Fig. 4. As compared with the pattern reported in database for cubic ZrO2,28 the diffracted peaks are shifted towards higher angles, corresponding to lower lattice parameters. In fact, the lattice parameter of the ˚ , whereas the value reported for bulk cubic phase is a ¼ 4.96 A ˚ .28 This difference can be cubic ZrO2 phase is a ¼ 5.1 A ascribed to the incorporation of Al within the metallic sublattice of cubic ZrO2. Actually, the Al-O distance in aluminum oxide structures is lower than the Zr-O distance in cubic ZrO2,28 therefore leading to a reduced lattice parameter. The absence of diffracted maxima related to the tetragonal phase was also checked with dedicated synchrotron radiation experiments, aimed at the detection of the well isolated, tetragonal (102) peak, as shown in the inset of Fig. 4.

J. Appl. Phys. 112, 014107 (2012)

FIG. 5. XRR data and fitting lines for as deposited (circles) and 900 C annealed (triangles) Al-ZrO2 layers.

XRR analyses of as-grown and annealed Al-ZrO2 stacks are presented in Fig. 5. From the simulation of the XRR data, uniform Al-ZrO2 layers with very low surface and oxide/silicon interface roughness (0.5 6 0.1 nm) are revealed. Annealing caused a reduction of the layer thickness and an ˚ 3 to increase of the electron density (from 1.64 6 0.05 e/A ˚3 1.75 6 0.05 e /A ) that corresponds to a densification of the Al-ZrO2 layer. The latter effect might be related to a thermal rearrangement of the atoms which leads to a shortening of the atomic distances as observed in a crystallization process. Combining the spectroscopic ellipsometry evaluation of the stack thickness and the XRR-deduced thickness of Al-ZrO2, the SiO2 interfacial layer (IL) is 1.5 6 0.3 nm for the asgrown and 2.3 6 0.3 nm for the 900 C annealed stacks. The W/TaN/Al-ZrO2 stack was also measured by XRR before and after annealing at 900 C (data not shown). For as-deposited films, the metal/oxide interface is very smooth, with roughness of 0.3 6 0.1 nm. After annealing, no significant variations are detected, except an increase of the TaN/ Al-ZrO2 roughness up to 0.6 6 0.1 nm, consistently with ToF-SIMS depth profiles, which evidenced a chemical reaction at the TaN/Al-ZrO2 interface. The as-deposited TaN layer is polycrystalline with grain size of the order of 3 nm (GIXRD results, not shown), and no significant variation is detected after annealing at 900 C in N2 and after forming gas annealing. B. Electrical characterization: Steady-state C-V (SCV) characteristics and leakage current

FIG. 4. GIXRD analysis of 24.4 nm thick Al-ZrO2 films after annealing at 900 C. The diffraction patterns of the monoclinic, tetragonal, and cubic ZrO2 phases are also added for comparison.28 The inset shows the details of the t(102) region as probed by synchrotron radiation diffraction measurements.

Steady-state capacitance-voltage (C-V) measurements were performed on capacitors S1, S2, and S3. Fig. 6 shows several C-V curves obtained at different frequencies on S1 samples featuring 25 and 10 nm thick Al-ZrO2 films. The C-V curves are well shaped with no frequency dispersion in the accumulation region. Sweeping voltage up-wards and downwards a small hysteresis is appreciable only in the thick sample. The slight dispersion of C-V curves in the depletion region points out the presence of oxide/silicon interface traps, whose density is evaluated to be in the 5-7 1011 eV1 cm2 range by the Hill-Coleman method.29


014107-5

Spiga et al.

FIG. 6. C-V characteristics, acquired at 1, 50, 100, and 300 kHz, of TaN/AlZrO2/SiO2/p-Si capacitors including 25-nm-thick and 10-nm-thick Al-ZrO2 films annealed at 900 C.

The equivalent oxide thickness (EOT) of the investigated Al-ZrO2 films was extracted from the simulation of the C-V curves at 100 kHz by using the metal-insulating-semiconductor fitting software (MISFIT), which takes into account quantum mechanical correction [refer to Ref. 30 for details on the fitting model]. The simultaneous fit of the C-V and conductance curves using MISFIT yielded an interface trap density in the 6.5–8.0 1011 eV1 cm2 range, which is consistent with the value extracted with the Hill-Coleman method. Figs. 7(a) and 7(b) report the plot of the EOT vs. Al-ZrO2 thickness (as measured by XRR) for as-deposited and annealed films. The k value is extracted from the slope of the linear fit of the experimental data. A k value of 20 6 1 is found for the as-deposited Al-ZrO2 films (S3 samples), and an increase of the k value up to 27 6 2 is obtained in annealed samples (both with W/TaN or Al metal gate electrodes, S1 and S2 samples). The intercept of the plots shown in Figs. 7(a) and 7(b) gives the EOT of the interfacial layer (EOTIL). For the Al metal gate, in the case of as-deposited S3 samples, the EOTIL was 1.3 6 0.2 nm, consistently with the presence of a 1.5 nm thick SiO2 chemical oxide grown on Si. After PDA at 900 C (S2 samples), an increase of the

FIG. 7. EOT vs. Al-ZrO2 thickness as-deposited films (Al gate) and annealed Al-ZrO2 films (Al or W/TaN metal gate).

J. Appl. Phys. 112, 014107 (2012)

EOTIL up to 2.1 6 0.2 nm was found, which is consistent with the value measured by XRR (2.3 6 0.3 nm), within the experimental errors. Therefore, within experimental errors, the detected increase of the EOTIL after annealing correlates with the observed increase of the IL physical thickness measured by XRR and spectroscopic ellipsometry before metal gate deposition. However, the EOTIL after annealing extracted in the case of S1 samples with TaN metal gates is 1.7 6 0.2 nm, slightly lower than in the case of Al gate. The small difference measured for EOTIL in samples S1 and S2 could arise from various effects. First of all, S1 and S2 samples were subjected to different thermal treatments, and the resulting different metal/oxide or oxide/silicon interfaces could give a diverse contribution to the total measured EOTIL.31,32 Moreover, it cannot be excluded an oxygen scavenging effect by the TaN metal layer, as already reported for other electrodes such as Ti, TiN, and Hf.33–35 Indeed, ToF-SIMS analyses evidenced a chemical reaction at the TaN/Al-ZrO2 interface, with an increase of oxygen signal inside the TaN layer after annealing. Due to the interest in CMOS applications, samples with TaN metal gates (S1 samples) where further investigated from the electrical point of view. Current vs voltage (I-V) measurements were performed in steady-state conditions. Curves referring to S1 samples with 10 and 25 nm-thick Al-ZrO2 films are reported in Fig. 8. The plateau at small positive voltages was due to minority carrier saturation. In the accumulation condition (negative voltages), two regimes were distinguished, namely low and high fields. The knees of the curves were approximately at the same value of electric field (1.6 MV/cm). On the contrary, breakdown occurred with different electric field (2.1 MV/cm and 2.8 MV/cm, corresponding to 5 V and 10 V, respectively for the thin and the thick samples). In both cases, leakage current was ascribed to holes conduction (this will be clarified in the band diagrams discussed below) and was modeled by a combination of elastic multi-trap assisted tunnel (MTAT) and Poole-Frenkel (PF) emission. The calculated curve of leakage current of the 10 nm-thick sample is displayed in the inset of Fig. 8 (line)

FIG. 8. I-V curves of 25-nm-thick (solid line) and 10-nm-thick (dashed) samples with W/TaN electrode. The inset shows interpolation of leakage data at low fields of the 10 nm-thick sample: symbols are experimental data, lines interpolations accounting for MTAT and PF emission, and assuming a density of bulk traps NT 61019 cm3.


014107-6

Spiga et al.

together with experimental data (symbols). At extremely low fields, the MTAT mechanism prevailed with respect to PF emission. The PF emission played a more significant role when the gate voltage approached 3 V, because the effective trap barrier height was sufficiently reduced to allow emission. The best interpolation of experimental data was obtained with a density of traps NT 6 1019 cm3 and a trap energy location around 250 meV below the valence band. Regarding the 25 nm-thick sample, we observed again a prevalent role of MTAT at extremely low fields and an increasing contribution of PF emission beyond 6 V (i.e., at the same electric field as in the 10 nm-thick samples) before the curve knee at 6.5 V. For a better comprehension of this picture, the band diagrams of the thin and the thick S1 samples under accumulation are reported in Fig. 9. The W/TaN work-function value used in band diagram simulations was 5.3 6 0.1 eV, as extracted by the flat-band voltage (VFB) versus EOT plot (data not shown) for films in the 5 to 25 nm thickness range, and assuming that oxide fixed charge (estimated to be 1.2 1012 cm2) is localized mostly at the IL/Al-ZrO2 interface. The VFB values were extracted from C-V curves acquired in pulsed conditions (see also Sec. III C) and neglecting oxide charge contribution. Indeed, C-V measurements performed in pulsed condition guarantee that the contribution of charge trapped into the oxide is negligible in the evaluation of the flat band voltage. In

FIG. 9. Sketch of the band diagrams of the 25-nm and the 10-nm-thick samples at: VG ¼ 4.4 V (a); VG ¼ 8 V (b). In these diagrams, the W/TaN electrode work-function is 5.3 6 0.1 eV.

J. Appl. Phys. 112, 014107 (2012)

Fig. 9(a), the applied voltage is VG ¼ 4.4 V and in Fig. 9(b), it is VG ¼ 8 V. Two curves are drawn in each plot, referring to the thin (10 nm) and the thick (25 nm) samples. As can be argued, only holes are responsible for conduction in both cases. With VG ¼ 4.4 V, the barrier is triangular, but still thick enough to inhibit Fowler-Nordheim (FN) tunnel for both thin and thick samples. At VG ¼ 8 V, the thin samples have already experienced breakdown, while FN tunnel may occur in the 25 nm-thick samples. In accordance with band diagrams, FN interpolation of experimental data of the 25-nmthick sample was obtained between 6.5 V and 10 V. This is shown in the FN plot in Fig. 10(a). Also, we verified that in the 10-nm-thick samples, before the breakdown, conduction is dominated by PF emission. This is shown in the PF plot in the range 3.3 V, 5 V, reported in Fig. 10(b). Those samples never enter the FN tunnel regime. C. Investigation of trapping at the interfaces with a pulsed capacitance technique

In order to evaluate trapping at the TaN/Al-ZrO2 interface, electron injection from the gate was forced applying a negative voltage to the electrode of S1 samples (VSTRESS). In this condition (accumulation for the p-type silicon substrate), electrons can be captured nearby the metal/oxide interface and holes nearby the silicon/oxide interface. Moreover, it is

FIG. 10. Interpolation of current data at high fields of samples with W/TaN electrode: (a) FN plot for 25-nm-thick samples in the voltage range between 6.5 V and 10 V and (b) PF plot for 10-nm-thick samples in the voltage range between 3.3 V and 5 V.


014107-7

Spiga et al.

worth noting that electron trapping is extremely fast and, in particular, it is much faster than hole trapping. Since conventional SCV measurement takes several seconds to complete a voltage sweep, it may detect only the long-time tail of trapping transients and totally ignores the short time transients due to electron trapping. On the contrary, the pulsed C-V technique (PCV) allows acquisition of a C-V curve in some tens of microseconds and (at least a portion of) the electron trapping transient can be monitored.36 In our experiments, VSTRESS was varied between 4.5 V and 6 V for the 25-nm-thick samples and between 2.5 V and 3.3 V for the 10-nm-thick ones. The flat band voltage shift (defined as DVFB ¼ VFB(t) VFB(0)) was evaluated as a function of the stress time in the 50 ls-100 s range (Fig. 11). At extremely short stress times (1 ms), pointing out to positive total net trap charge in the film. Therefore, we can model these experimental findings as following: at short stress time, electrons are trapped in electronic states at the metal/oxide interface and under the effect of the applied electric field, they move toward the silicon substrate; thus the distribution of negative charge in the bulk oxide determines a net positive DVFB. For long stress times, the contribution of hole trapping close to silicon interface become relevant, and the two distributions of charge (electrons nearby the metal and holes nearby the substrate) contribute contemporarily, but in opposite ways, to the total DVFB. However, since positive trapped charge is located close to silicon substrate, its effect on the total DVFB shift is higher than the one of negative charge. Therefore, even if after a given stress time, the amount of captured holes at the silicon interface is lower than that of captured electrons at the metal interface, the hole contribution may prevail on total DVFB. For this reason, in the case of thick samples, the measured DVFB became more and more negative increasing the stress time [Fig. 11(b)]. On the contrary, in case of thin samples, the DVFB remained positive [see Fig. 11(a)] because in the investigated time interval, the negative charge propagated inside the high-k oxide as to cover most of the film, and this negative charge was not compensated by holes on that timescale. Finally, the observed saturation at long times shown in Fig. 11(b) for thick film was due to filling of the available trap sites for holes at the silicon/oxide interface and consequent Coulomb repulsion of additional positive charge trapping. This effect occurred at longer times with higher values of VSTRESS. To summarize, the shape of DVFB vs. stress time curve is the result of the balance between the contribution due to the negative and the positive trapped charges, and therefore it is related to the electronic quality of both the metal/high-k and of high-k/substrate interfaces. If no traps are present at

J. Appl. Phys. 112, 014107 (2012)

FIG. 11. Data of flat band voltage shift measured during electrical stress in (a) 10 nm and (b) 25 nm thick samples with W/TaN electrode. Interpolating curves calculated with the model proposed in Sec. III D are drawn in (b) with solid lines.

the metal/high-k interface, the whole DVFB vs time curve will be negative. If the metal/high-k interface features an important density of traps, a bell shaped curve with a peak or plateau (with positive flat band values) is observed. Indeed, at extremely short times, electron trapping leads to positive and increasing values of DVFB, then the negative charge contribution begins to be balanced by the positive charge, and at longer stress times, the curve decreases towards a negative VFB value due to the prevalence of the contribution associated with the trapped holes. The results here obtained for the TaN/Al-ZrO2 interface are, therefore, consistent with the above proposed picture, as well as reported behavior on metal/high-k stacks (TiN/GdSiO (Ref. 37) and TaN/Al2O3 (Ref. 38)), pointing out to the presence of electron trapping originating by trap states near the metal interface. These interface defects could be also related to the observed TaN/ Al-ZrO2 chemical reaction, as discussed in Sec. III A. D. Analytical model of trapping and extraction of trap energy levels

Experimental data of DVFB in Fig. 11 were successfully interpolated by a trapping model recently proposed in literature,39 which predicts a DVFB power law dependence on time. For the sake of clarity, the key steps of that model are


014107-8

Spiga et al.

J. Appl. Phys. 112, 014107 (2012)

recalled here. The starting point of the model is that only a fraction (N T ) of the total density of traps reacts with free charge. This portion depends on the electric field in the highk dielectric (FHK) and follows a Fermi-like distribution. Denoting with sC the characteristic capture time, the rate equation for the concentration of trapped charge (nT) can be written as @nT ðx; tÞ N T nT ðx; tÞ ¼ ; @t sC

(1)

whose solution is inserted into the general equation DVFB

q ¼ CHK

tð ox

tHK x nT ðx; tÞdx; tHK

(2)

0

where tHK is the thickness of the high-k film. Denoting with ET0–EF the energy level of the reacting trap with respect to the Fermi level when bands are flat, with ET–EF its energy level when an external voltage is applied, and with x0 its space position, we write x0 ¼

ðET0 EF Þ ; q FHK

ET ðxÞ ¼ ET0 q FHK x:

(3) (4)

The most severe assumption of the model is that the total density of traps in the oxide bulk (NT) is uniform through the whole stack. This assumption is necessary in order to fully analytically solve the problem. Finally, one gets the power-law expression DVFB ¼ a tb þ c;

(5)

where a¼A

KT NT qFHK x0 e KT ; q FHK

(6)

FIG. 12. Band diagram of the W/TaN/Al-ZrO2/SiO2/Si system under VG ¼ 4.5 V and VG ¼ 6 V. Dashed lines represent the energy levels of the traps reacting with holes, as extracted from the model proposed in Sec. III D.

with solid lines. Values of the trap energy distance from the valence band (ET-EV) of the Al-ZrO2 are listed in Table II, together with the calculated values of the electric field in the high-k film (FHK). The energy level depends on VSTRESS, in fact increasing the band bending traps with energy levels closer and closer to the Al-ZrO2 valence bands interact. A picture is sketched in Fig. 12. IV. CONCLUSIONS

KT NT qFHK x0 e KT ; c¼ q FHK

(7)

and A and b are constants. Making use of this analytical model, the energy position of the traps involved in the capturing mechanism in the different stress conditions was extracted. Again, starting from the interpolation of leakage current data, NT ¼ 6 1019 cm3 was assumed. For clarity, in Fig. 11(b) results from interpolations of data are shown TABLE II. Energy level of Al-ZrO2 traps reacting with holes at long times in different experimental conditions. VSTRESS (V) 4.5 4.8 5.2 6.0

FHK (MV/cm)

ET-EV (eV)

1.22 1.30 1.40 1.60

1.87 1.79 1.60 1.37

Uniform thin Al-ZrO2 films were grown by ALD on Si(100) wafers. Structural and chemical characterizations were performed by XPS, XRR, GIXRD, and ToF-SIMS. As-grown films were found amorphous in the investigated thickness range (6–26 nm), while the cubic phase developed after annealing at 900 C. ToF-SIMS profiles revealed that Al-ZrO2 chemical uniformity was preserved after annealing together with a remarkable stability towards Si; however, an interaction at the TaN/Al-ZrO2 interface was found. Metal/Al-ZrO2/SiO2/p-Si capacitors with different thickness of the high-k film were investigated from the electrical point of view. Measured k value was 20 6 1 for the asdeposited films and increased to 27 6 2 after annealing, consistently with the stabilization of the ZrO2 cubic phase. Leakage current in TaN-gated capacitors was modeled by MTAT and PF emission in all samples, whereas conduction at high fields was due either to FN tunnel or PF emission for the thick and the thin samples, respectively. Systematic


014107-9

Spiga et al.

investigation of trapping effects during electrical stress was performed by the pulsed C-V technique in a wide time interval, from 50 ls to 100 s. As a result, the TaN/Al-ZrO2 interface exhibited quite relevant electron trapping. High density of traps at the latter interface can be the result of chemical reactions in agreement with ToF-SIMS results. Concomitant trapping of electrons at the metal/oxide interface and holes at the oxide/substrate interface gave rise to a dynamical instability of the flat band voltage, which followed a power-law with the stress time. Experimental data were interpolated by an analytical model. Within the assumption of an uniform distribution of bulk traps in the Al-ZrO2 film, a band of trap energy levels in the 1.4-1.9 eV range above the valence band of the high-k oxide was found. ACKNOWLEDGMENTS

This work was partially funded by the FIRB Project No. RBIP06YSJJ funded by the Italian MIUR. Mario Alia (Laboratorio MDM, IMM-CNR) is acknowledged for help in device fabrication. The authors are also grateful to Oier Bikondoa at BM28, ESRF, Grenoble, France for supporting synchrotron XRD measurements. 1

See www.itrs.net for ITRS roadmap, 2011. J. Robertson, Eur. Phys. J. Appl. Phys. 28, 265 (2004). 3 X. Zhao and D. Vanderbilt, Phys. Rev. B 65, 075105 (2002). 4 D. Sangalli and A. Debernardi, Phys. Rev. B 84, 214113 (2011). 5 D. Tsoutsou, L. Lamagna, S. N. Volkos, A. Molle, S. Baldovino, S. Schamm, P. E. Coulon, and M. Fanciulli, Appl. Phys. Lett. 94, 053504 (2009). 6 P. Tsipas, S. N. Volkos, A. Sotiropoulos, S. F. Galata, G. Mavrou, D. Tsoutsou, Y. Panayiotatos, A. Dimoulas, C. Marchiori, and J. Fompeyrine, Appl. Phys. Lett. 93, 082904 (2008). 7 Y.-H. Wu, M.-L. Wu, J.-R. Wu, and L.-L. Chen, Appl. Phys. Lett. 97, 043503 (2010). 8 A. Lamperti, L. Lamagna, G. Congedo, and S. Spiga, J. Electrochem. Soc. 158, G221 (2011). 9 L. Lamagna, A. Molle, C. Wiemer, S. Spiga, C. Grazianetti, C. Congedo, and M. Fanciulli, J. Electrochem. Soc. 159, H220 (2012). 10 A. Paskaleva, M. Lemberger, A. J. Bauer, W. Weinreich, J. Heitmann, E. Erben, U. Schro¨der, and L. Oberbeck, J. Appl. Phys. 106, 054107 (2009). 11 K.-C. Lau and B. I. Dunlap, J. Phys.: Condens. Matter 21, 145402 (2009). 12 A. Molle, L. Lamagna, C. Wiemer, S. Spiga, M. Fanciulli, C. Merckling, G. Brammertz, and M. Caymax, Appl. Phys. Express 4, 094103 (2011). 2

J. Appl. Phys. 112, 014107 (2012) 13

G. Congedo, A. Lamperti, L. Lamagna, and S. Spiga, Microelectron. Eng. 88, 1174 (2011). 14 Y.-H. Wu, L.-L. Chen, J.-R. Wu, M.-L. Wu, C.-C. Lin, and C.-H. Chang, IEEE Electron Device Lett. 31, 1008 (2010). 15 G. Zhang, W. Hwang, S.-H. S. Lee, B.-J. Cho, and W.-J. Yoo, IEEE Trans. Electron Devices 55, 2361 (2008). 16 D. Zhou, U. Schroeder, J. Xu, J. Heitmann, G. Jegert, W. Weinreich, M. Kerber, S. Knebel, E. Erben, and T. Mikolajick, J. Appl. Phys. 108, 124104 (2010). 17 Q. Zuo, S. Long, S. Yan, Q. Liu, L. Shao, Q. Wang, S. Zhang, Y. Li Wang, and M. Li, IEEE Electron Device Lett. 31, 344 (2010). 18 Q. Liu, W. Guan, S. Long, M. Liu, S. Zhang, Q. Wang, and J. Chen, J. Appl. Phys. 104, 114514 (2008). 19 S.-Y. Wang, D.-Y. Lee, T.-Y. Huang, J.-W. Wu, and T.-Y. Tseng, Nanotechnology 21, 495201 (2010). 20 X. Wu, P. Zhou, J. Li, L. Y. Chen, H. B. Lv, Y. Y. Li, and T. A. Tang, Appl. Phys. Lett. 90, 183507 (2007). 21 H. Zhang, B. Gao, B. Sun, G. Chen, L. Zeng, L. Liu, X. Liu, J. Liu, R. Han, J. Kang, and B. Yu, Appl. Phys. Lett. 96, 123502 (2010). 22 C.-H. Lee, K.-C. Park, and K. Kim, Appl. Phys. Lett. 87, 073510 (2005). 23 C. H. Cheng, A. Chin, and F. S. Yeh, IEEE Electron Device Lett. 31, 1020 (2010). 24 M. Jamil, J. Oh, M. Ramon, S. Kaur, P. Majhi, E. Tutuc, and S. K. Banerjee, IEEE Electron Device Lett. 31, 1208 (2010). 25 G. Congedo, S. Spiga, U. Russo, A. Lamperti, O. Salicio, E. Cianci, and M. Fanciulli, J. Vac. Sci Technol. B 29, 01AE04 (2011). 26 C. Wiemer, S. Ferrari, M. Fanciulli, G. Pavia, and L. Lutterotti, Thin Solid Films 450, 134 (2007). 27 D. Brigg and M. P. Seah, Auger and X-ray Photoelectron Spectroscopy, Practical Surface Analysis, 1st ed. (Wiley, New York, 1990), Vol. 1, p. 635–636 28 Inorganic crystal structure database, file 53998 for cubic, 18190 for monoclinic, and 9993 for tetragonal ZrO2, Fachinformationzentrum Karlsruhe (2011). 29 W. A. Hill and C. C. Coleman, Solid State Electron. 23, 987 (1980). 30 G. Apostolopoulos, G. Vellianitis, A. Dimoulas, J. C. Hooker, and T. Conard, Appl. Phys. Lett. 84, 260 (2004). 31 L. Kornblum, J. A. Rothschild, Y. Kauffmann, R. Brener, and M. Eizenberg, Phys. Rev. B 84, 155317 (2011). 32 J. Kwon and Y. J. Chabal, Appl. Phys. Lett. 96, 151907 (2010). 33 H. Kim, P. C. McIntyre, C. O. Chui, K. C. Sarawat, and S. Stemmer, J. Appl. Phys. 96, 3467 (2004). 34 C. Choi and J. C. Lee, J. Appl. Phys. 108, 064107 (2010). 35 ˚ . Ragnarsson, T. Chiarella, M. Togo, T. Schram, P. Absil, and T. L.-A Hoffmann, Microelectron. Eng. 88, 1317 (2011). 36 G. Puzzilli, B. Govoreanu, F. Irrera, M. Rosmeulen, and J. Van Houdt, Microelectron. Reliab. 47, 508 (2007). 37 R. Rao, R. Simoncini, H. Gottlob, M. Schmidt, and F. Irrera, J. Vac. Sci. Technol. B 29, 01A902 (2011). 38 R. Rao, P. Lorenzi, G. Ghidini, F. Palma, and F. Irrera, IEEE Trans. Electron Devices 57, 637 (2010). 39 R. Rao, R. Simoncini, and F. Irrera, Appl. Phys. Lett. 97, 163502 (2010).
