Structure and thermal stability of MOCVD ZrO2 films on Si (100)

(1)

Structure and thermal stability of MOCVD ZrO

₂

ﬁlms on Si (1 0 0)

X. Wu

a,

*, D. Landheer

a

, M.J. Graham

a

, H.-W. Chen

b

, T.-Y. Huang

b

, T.-S. Chao

c

a

Institute for Microstructural Sciences, National Research Council of Canada, Ottawa, Ont., Canada K1A 0R6

b

Institute of Electronics Engineering, National Chiao-Tung University, Hsinchu 300, Taiwan

c

Department of Electrophysics, National Chiao-Tung University, Hsinchu 300, Taiwan Received 17 October 2002; accepted 23 December 2002

Communicated by D.P. Norton

Abstract

The structure and thermal stability of ZrO2 ﬁlms grown on Si (1 0 0) substrates by metalorganic chemical vapor

deposition have been studied by high-resolution transmission electron microscopy, selected area electron diffraction and X-ray energy dispersive spectroscopy. As-deposited ﬁlms consist of tetragonal ZrO2 nanocrystallites and an

amorphous Zr silicate interfacial layer. After annealing at 850_{C, some monoclinic phase is formed, and the grain size is}

increased. Annealing a B6 nm thick ﬁlm at 850C in O2 revealed that the growth of the interfacial layer is at the

expense of the ZrO2layer. In a 3.0 nm thick Zr silicate interfacial layer, there is a 0.9 nm Zr-free SiO2region right above

the Si substrate. These observations suggest that oxygen reacted with the Si substrate to grow SiO2, and SiO2reacted

with ZrO2to form a Zr silicate interfacial layer during the deposition and annealing. Oxygen diffusion through the

tetragonal ZrO2phase was found to be relatively easier than through the monoclinic phase.

PACS: 68.37; 61.14; 77.84; 81.15

Keywords: A1. Interfaces; A1. Transmission electron microscopy; A3. Metalorganic chemical vapor deposition; B2. Dielectric materials

1. Introduction

There is increasing interest in replacing silicon dioxide with high dielectric constant (high-k) materials as gate dielectrics in deep submicron complementary metal-oxide-semicon-ductor (CMOS) technology [1,2]. In the past few

years, extensive study on high-k gate dielectric materials has narrowed down the available choices to some prime candidates such as HfO2, ZrO2and their silicates. However, the integration of such high-k alternatives into the current CMOS tech-nologies remains a huge challenge mainly due to the stability and interfaces of these high-k di-electrics in contact with the Si substrate[1]. ZrO2 has a high dielectric constant (B25), a high bandgap energy (5.8 eV) and a suitable band offset on Si (1.4 eV) [3]. Although thermodynamic *Corresponding author. Tel.: 9937823; fax:

+613-9526337.

E-mail address:[email protected] (X. Wu).

(2)

calculations suggest that ZrO2 should not react directly with silicon substrates to form an SiO2 interfacial layer [4], in practice it is difﬁcult to avoid the formation of this low-k interfacial layer during the deposition and post-annealing[5,6].

ZrO2(as well as HfO2) is known to have three low-pressure structural phases, with monoclinic (m), tetragonal (t), and cubic (c) phases appearing with increasing temperature. The m–t transition takes place near 1100_{C, and t–c transition takes}

place near 2400_{C. It has been found that t- and}

c-ZrO2are fast ion conductors, while m- ZrO2is not a fast ion conductor [7,8]. The diffusivities of O in t- and c-ZrO2 are much higher than the diffusivity of O in m-ZrO2[7]. Annealing ZrO2in an oxygen-containing atmosphere may be bene-ficial for the interface structure of the films deficient in oxygen, but t-, c-ZrO2 are excellent ionic conductors. This could result in excess diffusion of oxygen to the interface, ultimately forming an excessively thick SiO2interfacial layer. From this point of view, obtaining an m-ZrO2film would have the advantage of limiting interfacial SiO2 layer formation during the post-annealing. Although m-ZrO2 is a room-temperature stable phase, nonequilibrium thermodynamic conditions during deposition result in t-, c-ZrO2 being formed.

There are some controversial results on the interfacial layer between the silicon substrate and the ZrO2 layer. A study of the structure and stability of ultrathin ZrO2 layers on Si(1 0 0) showed that the interfacial layer is pure SiO2, and ZrO2 is remarkably stable against silicate formation up to 900C, i.e. there is no reaction between SiO2and ZrO2[9]. Other studies showed that the interfacial layer is not pure SiO2, but includes Zr atoms [5,6,10]. This Zr-silicate interfacial layer could be formed as the result of the reaction between Si and ZrO2 [10], b ut could also be formed by the reaction between SiO2 and ZrO2. The latter requires the initial formation of an SiO2 layer by the reaction of Si with oxygen. This paper presents high-resolution transmission electron microscopy (HRTEM), selected area electron diffraction (SAED) and X-ray energy dispersive spectroscopy

(XEDS) studies of ZrO2 ﬁlms grown on Si(1 0 0)

by metalorganic chemical vapor deposition

(MOCVD). The emphasis is on the structure and thermal stability of the ZrO2 and interfacial layers.

2. Experimental procedures

Si (1 0 0) substrates, 100 nm diameter, n-type

(r ¼ 0:0220:06 O cm) were given an HF-last

RCA clean prior to ﬁlm deposition. The CVD chamber is equipped with a 360l/s turbomolecular pump and a liquid injection system (LDS-300B produced by ATMI). The latter consisted of a liquid pump to pump the precursor, a 0.15 molar solution of (C3H7O)2(C11H19O2)2Zr in octane, through a hot glass frit at a rate of 0.2 ml/min. The vapors were carried with a 50 sccm ﬂow of Ar to a gas distribution ring 10 cm from the substrate. The glass frit, the components of the vaporizer, the gas ring and the connecting tube were maintained at a temperature of 190C with heating tapes and blankets, while the substrate temperature was controlled in the range 390– 550_{C with quartz-halogen lamps and a}

thermo-couple. The ﬁlms used in this study were deposited

at 390_{C. Oxygen was introduced into the}

chamber at ﬂow rates of 0–150 sccm through a separate gas distribution ring 30 cm from the substrate. Just prior to deposition the wafers were heated for 10 min at 500C in 10 mTorr of O2to replace the surface hydrogen termination with oxygen.

The wafers were cut into pieces after deposition and annealed in a Heatpulse 610 (Steag RTP Systems) rapid thermal processing system. For the cross-section TEM sample preparation, two bars were cut out of a wafer, and were glued together with the ﬁlm sides face-to-face to make the central part of the 3 mm diameter cross-section sample disc. Then the disc was mechanically thinned to B100 mm thickness. The thinned disc was dimpled from both sides with 3 mm diamond paste until the center of the disc was B20 mm thick, and then polished from both sides with 1 mm diamond paste to get a very smooth surface. The ﬁnal thinning until perforation was conducted using Ar

(3)

ion-milling from both sides using an ion beam angle of 8, and a gun voltage of 6 kV. The TEM sample was examined in a Philips EM430 T and a Hitachi HD-2000 scanning transmission electron microscope equipped with a cold ﬁeld emission gun and an XEDS system.

3. Results and discussion

In order to study the ZrO2film morphology, a 60 nm thick film was grown. Fig. 1a is a cross-section HRTEM image of the film (note that only part of the ZrO2 layer is shown in this figure).

t (111) t (200) t (202) t (311) Tetragonal ZrO 2 Zr silicate Si (a) (b) (c)

Fig. 1. (a) HRTEM image of an as-deposited ZrO2ﬁlm; (b) SAED pattern of the ﬁlm shown in (a); and (c) SAED pattern of the same

ﬁlm after annealing at 850_{C in N}

(4)

Clearly, the ZrO2ﬁlm is polycrystalline consisting of many nanocrystallites. There is also a 6.1 nm thick amorphous interfacial layer.Fig. 1bshows a SAED pattern obtained from the ﬁlm. Diffraction spots are due to the silicon substrate, and the silicon substrate was titled to the [0 1 1] zone axis. This [0 1 1] silicon diffraction pattern is used as a calibration standard to determine the plane spa-cing of ZrO2. The rings are due to nanocrystalline ZrO2. The relationship between a certain plane spacing (d) and the corresponding diffraction pattern (either rings or spots) is given by

Rd ¼ lL; ð1Þ

where R is either the ring radius or the separation of the direct (center) and diffracted beams as measured on the diffraction pattern, L is the camera length, and l is the wavelength of electrons at the given accelerating voltage. lL is a constant for a certain TEM operation condition. Since we know the spacing (d) of plane (h k l) of Si (d ¼ a=pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffih2_{þ k}2_{þ l}2_{; where a ¼ 0:5431 nm, is}

the lattice constant of Si), the constant lL can be determined using Eq. (1) with the diffraction spots of the silicon. Using a simple data extraction technique called electron diffraction image processing [11], the plane spacings

correspond-ing to the rcorrespond-ings were determined as d1 =

0.2954 nm, d2¼ 0:2544 nm, d3¼ 0:1836 nm and

d4¼ 0:1553 nm. Comparing the plane spacing data

with the tabulated tetragonal ZrO2plane spacing [12], the rings are indexed completely as tetragonal ZrO2 (t-ZrO2) with the lattice constants a ¼ 0:512 nm, c ¼ 0:525 nm, and the correspond-ing planes are indicated in Fig. 1b. The result agrees with the recent study by Perkins et al.[13].

The SAED pattern of an 850C, 2 min N2

annealed ﬁlm (Fig. 1c) contains two features of interest. Firstly, the contours of the rings due to

tetragonal ZrO2 are still there, but much

less continuous compared to the SAED pattern of the as-deposited ﬁlm (Fig. 1b), which is due to the larger grain size after annealing. Secondly, in addition to the diffraction spots of the Si substrate, there appear many distinct spots unlike any features shown in Fig. 1b. These spots are due to monoclinic ZrO2 with the lattice constants a ¼ 0:515 nm, b ¼ 0:521 nm,

c ¼ 0:531 nm, a ¼ 90, b ¼ 99:2 and g ¼ 90 [14]. Thus, some metastable tetragonal ZrO2phase transformed into the stable monoclinic phase, and the ﬁlm contains mixed tetragonal and monoclinic ZrO2phases after annealing.

An ultrathin film was grown for further structural analysis. The as-deposited film (Fig. 2a) consists of a 4.6 nm thick, polycrystalline ZrO2 layer and a 1.4 nm thick, amorphous interfacial layer. It is extremely difficult to obtain diffraction patterns from such a thin film by TEM. Instead, we performed fast Fourier transforms on the recorded HRTEM image to obtain the electron diffraction pattern from the ZrO2 layer. These diffraction patterns revealed that the ZrO2is in the metastable tetragonal phase. The plane spacings of individual grain were also measured directly from the HRTEM. Most of the measured values of plane spacings are between 0.295 and 0.296 nm and this is the (1 1 1) plane placing of tetragonal ZrO2 [12]. X-ray photoelectron spectroscopy (XPS) analysis showed that the interfacial layer is not a pure SiOx, but a Zr silicate layer[5], which is in agreement with a recent study[10].

Two pieces of wafer were annealed in O2 at 850_{C. One annealed for 20 s, and the other was}

annealed for 2 min. Figs. 2band c are HRTEM images of these two samples. The two annealed samples showed a similar structure to the as-deposited film: a polycrystalline ZrO2film with an amorphous interfacial layer. The total thickness of the polycrystalline ZrO2 layer and amorphous interfacial layer was 6.2 and 5.9 nm for the two samples, that is essentially unchanged compared with the as-deposited film, 6.0 nm. The thickness of the individual ZrO2layer and interfacial layer, however, was changed: the interfacial layer thick-ness increased from 1.4 to 2.7 nm for 20 s anneal-ing, and to 3.7 nm for 2 min annealanneal-ing, while the ZrO2layer thickness decreased from 4.6 to 3.4 nm for 20 s annealing and to 2.2 nm for 2 min annealing. The growth of the interfacial layer is at the expense of the ZrO2layer.

The Zr and O distributions of the annealed ﬁlm shown inFig. 2bwere determined using XEDS. A line scan was performed from the surface to the substrate to record the O, Zr and Si K edges simultaneously. The scan path is perpendicular to

(5)

the interface. The electron probe diameter is 0.5 nm, so the layer thicknesses obtained from XEDS measurement may not be exactly same as by HRTEM. The intensities of the O, Zr and Si K edges are extracted from each point.Fig. 3shows

the O and Zr profiles. The intensity of each element shown inFig. 3has not been corrected to reflect the relative concentrations in the film. In the ZrO2 layer (from 0 to 3.4 nm), both Zr and O counts exhibited an inverse ‘‘V’’ shape distribu-tion, the highest counts being reached around the middle of the film. This type of distribution of Zr and O counts has been reported in a Zr silicate film [15]. The ratio of the counts of Zr to O is not a constant at each position, which means that the composition is not stoichiometric ZrO2 through-out the layer. In the amorphous interfacial layer (from 3.4 to 6.4 nm), Zr was detected and this interfacial layer was revealed by XPS to be a Zr silicate. However, Zr is present only in the upper part (3.4–5.5 nm) of the interfacial layer, and there is no Zr in the lower part (from 5.5 to 6.4 nm). This indicates that there are two distinct layers in the amorphous interfacial layer: a 2.1 nm thick Zr silicate layer and a 0.9 nm thick Zr-free silicon oxide region adjacent to the Si substrate. The presence of this thin pure silicon oxide layer is a significant result. This thin silicon oxide could serve as an ‘‘ideal’’ SiO2-like interface, which is essential for ZrO2 or any other high-k gate dielectric to be integrated into CMOS technology. However, the 0.9 nm thick Zr-free silicon oxide is thicker than an ‘‘ideal’’ interface, and it would compromise the capacitance gain from the high-k

0 1 2 3 4 5 6 0 50 100 150 200 250 Interfacial layer ZrO 2 film Counts

Distance from surface (nm)

Zr O

7

Fig. 3. XEDS proﬁles of O and Zr of the annealed ZrO2ﬁlm

shown inFig. 2b. Si Zr silicate ZrO2 14Å 46Å (a) ZrO2 Zr silicate Si 27Å 35Å ZrO2 Zr silicate Si 22Å 37Å (b) (b)

Fig. 2. HRTEM images of a ZrO2ﬁlm: (a) as-deposited; (b)

annealed at 850_{C in O}

2for 20 s; and (c) annealed at 850C in

(6)

ZrO2 layer. The thickness of Zr-free silicon oxide should be controlled during the deposition and annealing treatments, and kept less than 0.5 nm.

The above observations suggest that the most likely mechanism of formation of the Zr silicate interfacial layer on top of the Zr-free silicon dioxide region is that oxygen reacted with the Si substrate to grow SiO2, and SiO2 reacted with ZrO2 to form Zr silicate. During the deposition and annealing treatments, any excess of oxygen will lead to rapid oxygen diffusion through the ZrO2, resulting in SiO2and Zr silicate formation.

In the precursors used for the deposition,

(C3H7O)2(C11H19O2)2Zr, the Zr is co-ordinated to six O atoms, while only two are required to form the ZrO2. This leaves four oxygen atoms per molecule that may release active oxygen during the decomposition of the precursor. Using O-free Zr precursors such as nitrogen co-ordinated precur-sors, and controlling the post-annealing conditions by using spike anneals and reduced oxygen partial pressure anneals may reduce the interfacial layer thickness.

In order to study the influence of ZrO2structure on the growth of the interfacial layer thickness, another piece of wafer was annealed first in N2for 2 min, then in O2 for 20 s (sample A). HRTEM observation revealed that the interfacial layer thickness of this sample is 2.3 nm, slightly less than 2.7 nm of the sample (Fig. 2b) annealed in O2 for 20 s without a prior N2anneal (sample B). The thickness difference between the two samples can be explained by considering the ZrO2 crystal structures before O2 annealing for both samples. After first annealing in N2, sample A consists of both tetragonal and monoclinic ZrO2 phases, while before O2annealing, sample B contains only the tetragonal ZrO2phase. As we discussed earlier, the diffusivities of O in t- and c-ZrO2 are much higher than the diffusivity of O in m-ZrO2 [7]. Therefore, O diffusion through the ZrO2 layer in sample B is relatively easier than in sample A, which results in the thicker interfacial layer in sample B. This observation suggests that before O2 annealing, obtaining m-ZrO2may help reduce the unwanted interfacial layer thickness. For the 60 nm thick sample film, the grain size was

observed to increase after 850C N2 annealing; sample A may also have a larger grain size than sample B before O2annealing, which would help reduce the interfacial layer thickness as well. However, the increase in grain size is limited by the ﬁlm thickness in this case. The interfacial layer thickness difference is not signiﬁcant in this study due to the ultrathin ZrO2 layer thickness and ultrasmall ZrO2grain size.

4. Conclusions

(1) As-deposited ﬁlms consist of a tetragonal ZrO2nanocrystallite layer and an amorphous Zr silicate interfacial layer. Some tetragonal ZrO2 phase transformed to monoclinic after annealing at 850C, and the grain size became larger.

(2) Annealing aB6.0 nm thick ﬁlm revealed that the growth of the interfacial layer is at the expense of the ZrO2layer. The total thickness of the polycrystalline ZrO2 layer and the amorphous interfacial layer remains un-changed compared with the as-deposited ﬁlm. A 3.0 nm thick interfacial layer was deter-mined by XEDS as a Zr silicate layer with a 0.9 nm thick Zr-free silicon oxide region adjacent to the Si substrate. These observa-tions suggest that oxygen reacted with the Si substrate to grow SiO2, and SiO2reacted with ZrO2 to form a Zr silicate interfacial layer during the deposition and annealing.

(3) Comparing the interfacial layer thickness after annealing aB6.0 nm thick ﬁlm in O2with and without prior N2 anneal suggests that O diffusion through the tetragonal ZrO2 phase is relatively easier than through the mono-clinic phase.

Acknowledgements

K. McIIwrath is acknowledged for the XEDS data acquisition, and the authors are grateful to J.W. Fraser, J.R. Phillips, X. Tong and T. Quance for their technical support.

(7)

References

[1] G.D. Wilk, R.M. Wallace, J.M. Anthony, J. Appl. Phys. 89 (2001) 5243.

[2] R.M. Wallace, G. Wilk, Mater. Res. Bull. 27 (2002) 192. [3] J. Robertson, J. Vac. Sci. Technol. B 18 (2000) 1785. [4] K.J. Hubbard, D.G. Schlom, J. Mater. Res. 11 (1996)

2757.

[5] H.-W. Chen, T.-Y. Huang, D. Landheer, X. Wu, S. Moisa, G.I. Sproule, T.-S. Chao, J. Electrochem. Soc. 149 (2002) F49.

[6] Y.Z. Yao-Zhi Hu, S.P. Sing-Pin Tay, J. Vac. Sci. Technol. B 19 (2001) 1706.

[7] U. Brossmann, R. Wurschum, U. Sodervall, H.-E. Schaefer, J. Appl. Phys. 85 (1999) 7646.

[8] B.W. Busch, W.H. Schulte, E. Garfunkel, T. Gustafsson, Phys. Rev. B 62 (2000) R13.

[9] M. Copel, M. Gribelyuk, E. Gusev, Appl. Phys. Lett. 76 (2000) 436.

[10] T. Yamaguchi, H. Satake, N. Fukushima, Appl. Phys. Letts. 80 (2002) 1987.

[11] J.P. McCaffrey, E.B. Svedberg, J.R. Phillips, L.D. Madsen, J. Crystal Growth 200 (1999) 498.

[12] Diffraction data, JCPDS ﬁle 17–923.

[13] C.M. Perkins, B.B. Triplett, P.C. McIntyre, K.C. Saraswat, E. Shero, Appl. Phys. Letts. 81 (2002) 1417.

[14] Diffraction data, JCPDS ﬁle 36–420.

[15] D.A. Muller, G.D. Wilk, Appl. Phys. Lett. 79 (2001) 4195.