Initial reaction of hafnium oxide deposited by remote

0 downloads 0 Views 295KB Size Report
Oct 5, 2014 - Downloaded to IP: ... as-deposited structure and chemical bonding were examined by TEM and XPS. The in situ XPS .... We used the GAUSSIAN98 quantum chemistry ... free energy change at the deposition temperature 250 °C. ... McQuarrie, Statitical Mechanics (University Science, Sausalito, 2000).
Initial reaction of hafnium oxide deposited by remote plasma atomic layer deposition method Youngdo Won, Sangwook Park, Jaehyoung Koo, Seokhoon Kim, Jinwoo Kim, and Hyeongtag Jeon Citation: Applied Physics Letters 87, 262901 (2005); doi: 10.1063/1.2150250 View online: http://dx.doi.org/10.1063/1.2150250 View Table of Contents: http://scitation.aip.org/content/aip/journal/apl/87/26?ver=pdfcov Published by the AIP Publishing Articles you may be interested in Al2O3 multi-density layer structure as a moisture permeation barrier deposited by radio frequency remote plasma atomic layer deposition J. Appl. Phys. 115, 073502 (2014); 10.1063/1.4866001 Comparative band alignment of plasma-enhanced atomic layer deposited high-k dielectrics on gallium nitride J. Appl. Phys. 112, 053710 (2012); 10.1063/1.4749268 Low temperature growth of high-k Hf–La oxides by remote-plasma atomic layer deposition: Morphology, stoichiometry, and dielectric properties J. Vac. Sci. Technol. A 30, 01A147 (2012); 10.1116/1.3665419 The reaction pathways of the oxygen plasma pulse in the hafnium oxide atomic layer deposition process Appl. Phys. Lett. 93, 124104 (2008); 10.1063/1.2991288 Relationships among equivalent oxide thickness, nanochemistry, and nanostructure in atomic layer chemicalvapor-deposited Hf–O films on Si J. Appl. Phys. 95, 5042 (2004); 10.1063/1.1689752

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: 166.104.31.57 On: Sun, 05 Oct 2014 04:58:16

APPLIED PHYSICS LETTERS 87, 262901 共2005兲

Initial reaction of hafnium oxide deposited by remote plasma atomic layer deposition method Youngdo Won Department of Chemistry, Hanyang University, Seoul 133-791, Korea

Sangwook Park, Jaehyoung Koo, Seokhoon Kim, Jinwoo Kim, and Hyeongtag Jeona兲 Division of Materials Science and Engineering, Hanyang University, Seoul 133-791, Korea

共Received 24 May 2005; accepted 17 November 2005; published online 19 December 2005兲 A remote plasma atomic layer deposition 共RPALD兲 method has been applied to grow a hafnium oxide thin film on the Si substrate. The deposition process was monitored by in situ XPS and the as-deposited structure and chemical bonding were examined by TEM and XPS. The in situ XPS measurement showed the presence of a hafnium silicate phase at the initial stage of the RPALD process up to the 20th cycle and indicated that no hafnium silicide was formed. The initial hafnium silicate was amorphous and grew to a thickness of approximately 2 nm. Based on these results and model reactions for silicate formation, we proposed an initial growth mechanism that includes adatom migration at nascent step edges. Density functional theory calculations on model compounds indicate that the hafnium silicate is thermodynamically favored over the hafnium silicide by as much as 250 kJ/ mol. © 2005 American Institute of Physics. 关DOI: 10.1063/1.2150250兴 The semiconductor industry currently demands both new materials and manufacturing techniques that ensure high performance and reliable operation of subnanometer scale devices.1 Because microelectronic features continue to become miniaturized, a SiO2 gate thickness below 1 nm will be required in the near future; at this size, tunneling current through the gate dielectric becomes a serious problem. A search for new high-k oxide materials found HfO2 to be a logical candidate to replace SiO2 due to its high dielectric constant and stability on Si substrates.2 In this study, we investigated the growth of HfO2 as a new gate oxide material on a Si substrate with the use of an atomic layer deposition 共ALD兲 system. The ALD process was applied to various materials to deposit nano-scaled thin films. The unique features of ALD warrants excellent conformity and accurate thickness control.3 The ALD method can be classified according to the choice of precursors and reactants. Based on the source material, ALD processes can be classified as halogen precursor ALD or metal organic 共MO兲 precursor ALD. The halogen precursor ALD was the first ALD process developed and has been studied for many years.4 In general, this ALD method provides good step coverage and low impurity concentrations in thin films.5 However, it exhibits problems such as the presence of halogen atom residues in films, corrosion of gas delivery lines, and particle generation. The use of a metal organic source, e.g., tetrakis-diethyl-amino-hafnium 共TDEAH兲 共Ref. 6兲 does not pose these problems. However, the MO source often produces low density films with high impurity contamination. Therefore, several methods have been developed to overcome these problems, such as the use of a plasma application in the ALD process. In this work, we utilized a downstream-type remote plasma ALD 共RPALD兲 system. The RPALD method has advantages such as decreased charged particle damage and increased reactivity of radicals.7. In order to apply the RPALD a兲

Author to whom correspondence should be addressed; electronic mail: [email protected]

method for nanoelectronic device fabrication, the chemical and structural details of the initial phase of the hafnium oxide deposition process needs to be understood. The main focus of this letter is the investigation of the growth mechanism of hafnium oxide during the RPALD process. The hafnium oxide films were deposited by the RPALD method on p-type Si substrates with 具100典 orientation and 6 – 12 ⍀ cm resistivity, using TDEAH as the hafnium precursor and oxygen plasma as the oxidizing agent. Si substrates were cleaned by soaking them in piranha solution 共H2SO4 : H2O2 = 4 : 1兲 for 10 min and in dilute HF solution 共HF : H2O = 1 : 100兲 for 2 min to remove organic and native oxides, respectively. A cycle of the deposition process consisted of 2 s of TDEAH supply, 1 min of Ar purge, 5 s of O2 plasma, and 1 min of Ar purge. The process window for HfO2 deposition was optimized at a temperature of 250 ° C, plasma power of 100 W. The RPALD reactor was connected to an in situ x-ray photoelectron spectroscopy 共XPS兲 system, which monitored the as-deposited state upon the completion of each cycle. All films were measured at take-off angles of 90°. High-resolution transmission electron microscopy 共HRTEM兲 was utilized determine to film thickness and to investigate interface morphologies and structures. Figure 1 shows the in situ XPS results of the hafnium oxide deposition process. Fig. 1共a兲, displays the in situ XPS spectrum of the as-deposited state for the first cycle of the

FIG. 1. 共a兲 In situ Si 2p XPS of HfO2 films as deposited with the oxygen gas and with the oxygen plasma at the first cycle. The XPS of the HFcleaned Si substrate is also shown for contrast. 共b兲 The XPS spectra as a function of the process cycle, showing data for the 1st, 5th, and 10th cycles.

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: 0003-6951/2005/87共26兲/262901/3/$22.50 87, 262901-1 © 2005 American Institute of Physics 166.104.31.57 On: Sun, 05 Oct 2014 04:58:16

262901-2

Appl. Phys. Lett. 87, 262901 共2005兲

Won et al.

FIG. 2. High resolution ex situ Hf 4f and O 1s XPS spectra of HfO2 films at the 5th, 10th, 20th, and 30th process cycles. The inset shows the deconvolution of the XPS line shapes into the hafnium silicate peak and the HfO2 peak at the 30th cycle.

TDEAH precursor and the oxygen plasma. The XPS specFIG. 3. Cross-sectional TEM images of as-deposited HfO2 films 共a兲 at the trum of the HF cleaned Si substrate and the as-deposited 5th cycle, 共b兲 at the 10th cycle, 共c兲 at the 20th cycle, and 共d兲 at the state with oxygen gas are also shown for comparison. While 30th cycle. the hafnium oxide film was not deposited with oxygen gas as the oxidizing agent, a significant change in the XPS specis approximately 2 nm thick. These high-resolution TEM imtrum was observed with the oxygen plasma even in the first ages indicate that the hafnium silicate layer is preferentially cycle. The XPS result exhibits the Si 2p feature of the bindformed in the beginning of the RPALD hafnium oxide depoing energy between 102 and 103 eV, which indicates the sition process and hafnium oxide starts to grow on top of the formation of hafnium silicate. A feature for hafnium silicide hafnium silicate layer. was not observed. The initial phase of the RPALD HfO2 Considering the experimental results on the basis of the deposition process is the formation of hafnium silicate from stoichiometric relations, we propose the following hafnium the first cycle. Figure 1共b兲 shows the XPS spectra at the 1st, oxide growth mechanism of the RPALD processes. At the 5th, and 10th cycles. As the deposition progresses, the intenmolecular orbital 共MO兲 precursor pulse of the first cycle, sity of the hafnium silicate peak increases, while the intenTDEAH reacts with the surface hydroxyl group, which is sity of the bulk Si peak at about 99 eV gradually decreases. generated during air exposure of the HF-cleaned Si surface. The XPS spectra as a function of the process cycle reveals The number of surface hydroxyl groups would also be afthat the hafnium silicate thin film grows in the early stages of fected by the following oxygen plasma pulses: the RPALD process. Si* – OH + Et2N – Hf – 共NEt2兲3 Figures 2共a兲 and 2共b兲 show the Hf 4f and O 1s XPS spectra, respectively, at the 5th, 10th, 20th, and 30th process → Si* – O – Hf – 共NEt2兲3 + Et2NH* , 共1兲 cycles. Wilk et al. reported that the Hf 4f 7/2 peak of Hf where Si* – OH denotes the surface group and Et2NH* represilicate is ⬃1 eV higher than that of HfO2, which is located sents Et2NH, CH3CH v NCH2CH3 and H2. Et stands for the at ⬃16.5– 17 eV.8 As shown in Fig. 2共a兲, the Hf 4f 7/2 peak at ethyl group, CH3CH2–. Equation 共1兲 is a feasible channel in the binding energy of 18.1 eV indicates the formation of multiple manifolds of reactions. At the oxygen plasma pulse, hafnium silicate up to the 10th cycle. The Hf 4f 7/2 peak at the possible reactions would be as follows: 30th cycle shifts to a lower binding energy of 17.5 eV due to HfO2 formation. The peak at 531.9 eV in Fig. 2共b兲 correSi* – O – Hf – 共NEt2兲3 + 3O* → Si* – O – Hf – 共OH兲3 sponds to the O 1s electronic state of the hafnium silicate, + 3CH3CH v NCH2CH3 , 共2兲 indicating that the as-deposited film contains the hafnium silicate phase up to the 30th cycle. Beyond the 20th cycle, another O 1s peak begins to develop around 530.2 eV, which Si* – O – Hf – 共NEt2兲3 + 3O* → Hf* – O – Si – 共OH兲3 causes a broadening of the initial O 1s peak and development + 3CH3CH v NCH2CH3 . 共3兲 of a shoulder feature on the higher binding energy side at the 30th cycle. The 530.2 eV peak corresponds to the O 1s elecO* represents the oxygen plasma source, which replaces ditronic state of the hafnium oxide. The inset of Fig. 2共b兲 ethylamine ligands and generates surface hydroxyl groups. shows the deconvolution of the XPS line shapes into the In Eq. 共3兲 the surface Si atom is highly excited by the oxygen hafnium silicate peak and the HfO2 peak at the 30th cycle. plasma energy and results in an exchange with the hafnium The O 1s XPS spectra indicate that HfO2 film is formed atom. Both reactions yield silicate sites. beyond the 20th cycle of the RPALD process. The deposition reactions 共1兲–共3兲 repeat through a few The cross-sectional TEM images of as-deposited more subsequent cycles and generate one or two monolayer hafnium oxide films are shown in Fig. 3 for the 5th, 10th, thick silicate step edges on the Si substrate surface. From the 20th, and 30th cycles of the RPALD process. Figure 3共a兲 oxygen plasma impact, some Si atoms may dissociate out of displays the amorphous hafnium silicate layer formed up to the substrate surface and ascend the step edges to be incorthe fifth cycle, which is confirmed by the in situ XPS specporated into the growing silicate layers. Jeong and Oshiyama trum of Fig. 1共b兲. Figure 3共b兲 shows that the amorphous carried out first-principles total-energy calculations of the hafnium silicate layer grows until the tenth cycle. Figures diffusion pathways and activation barriers for the adatom in 3共c兲 and 3共d兲 indicate that the HfO2 films are present at the epitxial growth on hydrogenated Si surfaces and demon20th and 30th cycles, respectively. These TEM images show strated that the Schwoebel barrier was absent near the singlethe HfO2 films grown on top of the initially deposited layer step.9 Zhu et al. performed density functional theory 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: hafnium silicate layer. The interfacial hafnium silicate layer calculations and kinetic Monte Carlo simulations to show 166.104.31.57 On: Sun, 05 Oct 2014 04:58:16

262901-3

Appl. Phys. Lett. 87, 262901 共2005兲

Won et al. TABLE I. Possible model reactions for silicate and silicide deposition processes.a Silicate formation reactions

E0

H523

G523

共4兲 SiH3OH + Hf共NH2兲4 → H3Si– O – Hf共NH2兲3 + NH3 共5兲 SiH4 + Hf共NH2兲3OH → H3Si– O – Hf共NH2兲3 + H2 共6兲 SiH4 + Hf共OH兲4 → H3Si– O – Hf共OH兲3 + H2 共7兲 H3Si– SiH3 + Hf共NH2兲3OH → H3Si– O – Hf共NH2兲3 + SiH4 共8兲 H3Si– SiH3 + Hf共OH兲4 → H3Si– O – Hf共OH兲3 + SiH4

−95.088 −84.927 −87.185

−93.683 −78.395 −80.708

−98.456 −80.453 −83.504

−98.354 −100.612

−99.950 −102.263

−98.603 −101.654

Silicide formation reactions 共9兲 SiH3OH + Hf共NH2兲4 → SiH2OH – Hf共NH2兲3 + NH3 共10兲 SiH4 + Hf共NH2兲3OH → H3Si– Hf共NH2兲3 + H2O 共11兲 SiH4 + Hf共OH兲4 → H3Si– Hf共OH兲3 + H2O 共12兲 SiH4 + Hf共NH2兲4 → H3Si– Hf共NH2兲3 + NH3

E0 109.473 179.267 171.842 99.139

H523 113.424 183.473 173.146 101.722

G523 106.853 175.851 175.869 91.215

Reactions 共1兲 through 共3兲 are in the text. E0 is the SCF energy difference, H523 is the enthalpy change, and G523 is the free energy change at the deposition temperature 250 ° C. The energies are given in kJ/mol. a

that an adatom could easily climb up monoatomic-layer-high steps on several fcc metal 共110兲 surfaces through a place exchange mechanism.10 They argued that the adatom ascending processes could apply to other related homo- and heteroepitaxial growth systems. In our RPALD cycles, the highenergy oxygen plasma might generate some Si and Hf atoms on the substrate surface to ascend into higher layers like adatoms. In order to elaborate on the growth mechanism of initial silicate layers, we established a series of possible model reactions in the most simplified form. Surface atoms are represented as hydrogen-terminated atoms: Si as SiH4, Si–Si as H3Si– SiH3. The diethylamino group is simplified into the amino group. We used the GAUSSIAN98 quantum chemistry package program to perform ab initio density functional theory calculations on the model compounds.11 The density functional theory incorporates 共B3LYP兲 functionals: Becke’s three parameter functionals and the Lee-Yang-Parr nonlocal correlation functionals.12,13 A 共LACVP**兲 basis set was employed and the Los Alamos effective core potential was used for the hafnium atom.14 The density functional theory calculations resulted in the energetic data for the compounds involved in the model reactions listed in Table I. The calculation yields the selfconsistent-field energy of the Hamiltonian E0 and vibrational frequencies of normal modes. Following a standard statistical mechanics procedure, the data were incorporated to generate enthalpies and free energies at the given temperature.15 As thermodynamic quantities are state functions, the reaction energetic data are easily obtained by subtracting the sum of reactant energies from the sum of product energies. In Table I, H523 is the enthalpy change and G523 is the free energy change at the deposition temperature 250 ° C. Reaction 共4兲 represents the suggested MO source deposition reaction. Reactions 共5兲 and 共6兲 correspond to “adatom” incorporation processes that are thermodynamically favored. Energetic adatoms could be generated on the substrate surface with the oxygen plasma impact. Reactions 共7兲 and 共8兲 also generate adatoms and yield silicate deposition. The energetic data of Table I definitely indicates that hafnium sili-

cate formation is a thermodynamically favored process. The large positive G523 values of the hafnium silicide formation reactions clearly show that silicide would never be formed in the hafnium oxide deposition processes. Our simple theoretical delineation describes the initial hafnium silicate layer formation of the hafnium oxide RPALD process. The growth process of hafnium oxide was experimentally investigated and theoretically modeled and it was found that the initial phase of the hafnium oxide deposition process was the formation of amorphous hafnium silicate with no silicide formation. The hafnium oxide grew on top of the initial hafnium silicate layer. This study was supported by the National Program for Tera-level Nano-devices of the Ministry of Science and Technology as one of the 21st century Frontier Programs. A. I. Kingon, J.-P. Maria, and S. K. Streiffer, Nature 共London兲 406, 1032 共2000兲. 2 M. Balog, M. Schieber, M. Michman, and S. Patai, Thin Solid Films 41, 247 共1977兲. 3 M. Ritala, M. Leskela, E. Rauhala, and P. Haussalo, J. Electrochem. Soc. 142, 2731 共1995兲. 4 T. A. Pakkanen, V. Nevalainen, M. Lindbald, and P. Makkonen, Surf. Sci. 188, 456 共1987兲. 5 M. Ritala and M. Leskela, in Handbook of Thin Film Materials, edited by H. S. Nalwa 共2002兲, p. 103. 6 H. K. Kim, J. Y. Kim, J. Y. Park, Y. Kim, W. M. Kim, Y. D. Kim, and H. Jeon, J. Korean Phys. Soc. 41, 739 共2002兲. 7 J. Y. Kim, S. Seo, D. Y. Kim, H. Jeon, and Y. Kim, J. Vac. Sci. Technol. A 22, 8 共2004兲. 8 G. D. Wilk, R. M. Wallace, and J. M. Anthony, J. Appl. Phys. 87, 484 共2000兲. 9 S. Jeong and A. Oshiyama, Phys. Rev. Lett. 81, 5366 共1998兲. 10 W. Zhu, F. B. de Mongeot, U. Valbusa, E. G. Wang, and Z. Zhang, Phys. Rev. Lett. 92, 106102 共2004兲. 11 J. A. Pople et al., Gaussian 98, Revision A. 4 共Gaussian, Pittsburgh, PA, 1998兲. 12 A. D. Becke, J. Chem. Phys. 98, 5648 共1993兲. 13 C. Lee, W. Yang, and R. G. Parr, Phys. Rev. B 37, 785 共1988兲. 14 P. J. Hay and W. R. Wadt, J. Chem. Phys. 82, 299 共1985兲. 15 D. McQuarrie, Statitical Mechanics 共University Science, Sausalito, 2000兲. 1

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: 166.104.31.57 On: Sun, 05 Oct 2014 04:58:16