Growth, structure, and morphology of TiO2 films deposited by molecular beam epitaxy in pure ozone ambients
1. ARTICLE IN PRESS
Microelectronics Journal 37 (2006) 1493–1497
www.elsevier.com/locate/mejo
Growth, structure, and morphology of TiO2 films deposited by
molecular beam epitaxy in pure ozone ambients
Patrick Fishera, Oleg Maksimovb, Hui Dua, Volker D. Heydemannb,
Marek Skowronskia, Paul A. Salvadora,Ã
a
Department of Materials Science and Engineering, Carnegie Mellon University, Pittsburgh, PA 15213, USA
b
Electro-Optics Center, Pennsylvania State University, Freeport, PA 16229, USA
Available online 14 August 2006
Abstract
TiO2 films were grown using a reactive molecular beam epitaxy system equipped with high-temperature effusion cells as sources for Ti
and an ozone distillation system as a source for O. The growth mode, characterized in-situ by reflection high-energy electron diffraction
(RHEED), as well as the phase assemblage, structural quality, and surface morphology, characterized ex-situ by X-ray diffraction and
atomic force microscopy (AFM), depended on the choice of substrate, growth temperature, and ozone flux. Films deposited on (1 0 0)
surfaces of SrTiO3, (La0.27Sr0.73)(Al0.65Ta0.35)O3, and LaAlO3 grew as (0 0 1)-oriented anatase. Both RHEED and AFM indicated that
smoother surfaces were observed for those grown at higher ozone fluxes. Moreover, while RHEED patterns indicated that anatase films
grown at higher temperatures were smoother, AFM images showed presence of large inclusions in these films.
r 2006 Elsevier Ltd. All rights reserved.
PACS: 81.15.Hi; 61.14.Hg; 61.10.Nz
Keywords: Titanium dioxide; MBE; Thin film epitaxy
There is a wide technological interest in TiO2 because it Most previously reported films were grown by pulsed
has interesting physical and chemical properties, which laser deposition [14,15], metal organic chemical vapor
make it appealing for optical, dielectric, electrochemical, deposition [6,7,11,16,17], or sputtering [5,10,18]. Molecular
and photocatalytic applications [1–3]. TiO2 is also of beam epitaxy (MBE), a technique that differs in thermo-
scientific interest since it has polymorphic structures dynamics and kinetics from the other approaches and that
(anatase, rutile, and brookite) that exhibit different can produce high quality films, has been limited to growth
stabilities and properties [4,5]. Certain applications, such of TiO2 on LaAlO3 [8], SrTiO3 [8], and GaN [13]
as integrated dielectrics or photoelectrochemical cells, substrates. MBE also allows the integration of charge
require thin films of TiO2 that exhibit a specific crystal neutral TiO2 monolayers with other chemical and structur-
structure, orientation, and/or morphology. Thus, there are al layers, such as SrO in SrTiO3 [19]. Importantly,
a number of reports on the growth of TiO2 films on various differences in sources used in MBE can vary the thermo-
substrates, including Al2O3 [5–7], LaAlO3 [8], (La, Sr)(Al, dynamic/kinetic aspects of the growth and, therefore,
Ta)O3 (LSAT) [9], MgO [5,10], SrTiO3 [11], BaTiO3 [12], impact film properties.
(Zr, Y)O2 [9], and GaN [13]. Depending on the choices of Oxide MBE requires a special oxygen source to
substrate and deposition conditions, both rutile and/or completely oxidize deposited metal, to prevent source
anatase can be grown. oxidation, and to allow the use of in-situ electron probes.
For this purpose, RF or microwave plasma sources are
ÃCorresponding author. Tel.: +1 412 268 2702; fax: +1 412 268 7596. often used instead of molecular oxygen (O2) [8,20].
E-mail address: paul7@andrew.cmu.edu (P.A. Salvador). However, plasma sources produce highly energetic species
0026-2692/$ - see front matter r 2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.mejo.2006.05.010
2. ARTICLE IN PRESS
1494 P. Fisher et al. / Microelectronics Journal 37 (2006) 1493–1497
and have limited flux control [21]. An alternative approach
is to use pure ozone (O3) or an O3/O2 gas mixture. Ozone
flux is easily controlled with a mass flow controller and can
be directed towards the heated substrate surface where it is
thermally cracked into O and O2.
There have only been a limited number of reports
describing the influence of ozone flux on the growth of
oxide films, particularly for the cases with no obvious
oxidation problem, such as TiO2. In this study, we have
grown TiO2 films on various single crystal substrates using
a MBE system equipped with a high-temperature effusion
cell as a source of Ti and an ozone distillation system to
provide pure ozone. This paper describes the effects of
substrate, growth temperature, and ozone flux on crystal
structure, orientation, phase assemblage, and morphology
of the films.
Commercial SrTiO3(1 0 0), LSAT(1 0 0), and LaAlO3
(1 0 0) substrates were etched in a 3:1 HCl:HNO3 solution
for 2–3 min [22], rinsed in deionized water, and then
degreased prior to the growth. Quarters of 2-inch wafers
were mounted into an Inconel sample holder and
introduced into the growth chamber (SVT Associates).
Substrates were heated to 750 1C using a resistive boron
nitride heater (temperature is measured with a thermo-
couple located in close proximity to the substrate) and
annealed at 750 1C for 1 h under the deposition ozone flux.
Ozone was generated with a commercial unit (Ozone
Solutions) capable of producing 6% O3 in O2. It was
distilled by passing the O2/O3 mixture through the liquid-
nitrogen cooled dewar filled with silica gel; the O3 was
adsorbed while the remnant O2 was pumped away. After
storing sufficient amount, the pure ozone stream was
generated by warming the dewar. The high-concentration
O3 flux was introduced into the chamber (base pressure of
10À10 Torr) using a mass flow controller at a rate of
0.25–2 sccm, which corresponded to a pressure of 6 Â 10À6–
2.5 Â 10À5 Torr. The Ti flux was produced with a high-
temperature effusion cell operated at a fixed temperature
between 1500 and 1600 1C. During each growth run, the
process was monitored using a differentially pumped
RHEED system (Staib Instruments) operated at 12.0 kV
with an incident angle of 31. X-ray diffraction (XRD) Fig. 1. RHEED images taken along the [1 0 0] azimuth at the end of
measurements [23] were made in y–2y, o, and f-scans growth for TiO2 films deposited at 750 1C using an ozone flux of 1 sccm
modes. Atomic force microscopy (AFM) was carried out in and a Ti source temperature of 1550 1C on (a) SrTiO3(1 0 0), (b)
contact mode [23]. X-ray reflectance was performed to LSAT(1 0 0), (c) and LaAlO3(1 0 0). The film in (d) is similar to (c) but
was deposited on LaAlO3(1 0 0) using a Ti source temperature of 1525 1C.
determine film thickness.
Fig. 1 shows RHEED images, taken along the [1 0 0]
azimuth, of anatase films grown at 750 1C using a Ti-cell more clearly evident than those on SrTiO3 or LSAT,
temperature of 1550 1C and an ozone flux of 1 sccm on (a) indicating better surface morphology and higher crystal-
SrTiO3(1 0 0), (b) LSAT(1 0 0), and (c) LaAlO3(1 0 0). linity for the films on LaAlO3. This is likely a result of the
Fig. 1(d) is a similar RHEED image to Fig. 1(c) but for excellent lattice match between anatase and LaAlO3(1 0 0)
a film grown on LaAlO3(1 0 0) using a Ti-cell temperature (mismatch ¼ f ¼ 0.3%).
of 1525 1C (all other conditions were the same). All A comparison of the RHEED patterns given in Figs. 1(c
RHEED patterns were consistent with anatase (0 0 1) and d) suggests that film quality can be improved by a
growth and exhibited a characteristic four-fold reconstruc- growth rate reduction (as was observed in [8]). However,
tion [8,20]. Fig. 1 shows that the RHEED patterns for the ˚
the growth rate for the film in Fig. 1(d) was about 0.017 A/s,
films grown on LaAlO3 are sharper and Kikuchi lines are and for most applications a higher rate would be preferred.
3. ARTICLE IN PRESS
P. Fisher et al. / Microelectronics Journal 37 (2006) 1493–1497 1495
Fig. 3. RHEED images taken along the [1 0 0] azimuth at the end of film
growth for TiO2 deposited on LaAlO3 at 550 1C using a Ti source
temperature of 1550 1C and an ozone flux of (a) 0.25 sccm O3 and (b)
1.00 sccm O3.
Fig. 2. (a) RHEED intensity oscillations during TiO2 film growth on
LaAlO3(1 0 0), using an ozone flux of 1 sccm and a Ti source temperature
present RHEED images taken along the [1 0 0] azimuth for
of 1550 1C (taken along the [1 0 0] azimuth). (b) X-ray reflectivity pattern
for the film whose RHEED oscillations were shown in (a). anatase films grown on LaAlO3 under different ozone
fluxes (at T ¼ 550 1C and at a Ti cell temperature ¼
1550 1C). In Fig. 3(a), a spotty pattern is observed for the
To study if the growth could be optimized without film deposited at 0.25 sccm O3. The RHEED pattern of a
sacrificing growth rate, the effects that other parameters film deposited at the increased O3 flux of 1.0 sccm is given
had on the growth of TiO2 on LaAlO3(1 0 0) substrates in Fig. 3(b). In this higher ozone flux case, the RHEED
(which exhibited the sharpest 2D RHEED images in Fig. 1) pattern is more streaky than that in Fig. 3(a). On raising
were investigated. the temperature from 550 to 750 1C, one obtains the
RHEED intensity oscillations, one set of which is given RHEED pattern given in Fig. 1(c), wherein all character of
in Fig. 2(a) for the film whose RHEED pattern was given a transmission spot pattern is lost, and the pattern is
in Fig. 1(c), were observed consistently for the films grown dominated by spots on a circle and minor streakiness,
on LaAlO3 substrates. The thickness of this film was indicating an atomically smooth surface. Furthermore, the
determined using X-ray reflectance and the scan for this four-fold reconstruction becomes evident for the high-
film is given in Fig. 2(b). This data was refined using temperature and high-ozone flux-grown film.
Philips’ WinGixa software and the film was determined to To better understand surface morphologies, films were
˚
be 180 A thick. Combining this with the RHEED oscilla- studied ex-situ with AFM. In Figs. 4(a)–(c), we show AFM
tions, we calculated that one RHEED intensity oscillation images of the TiO2 films whose RHEED patterns were
˚
corresponded to the growth of E4.5 A of anatase; that given in Figs. 1(a)–(c), which were grown on SrTiO3,
˚
value is very close to a bilayer of anatase (E4.75 A), in LSAT, and LaAlO3, respectively. Two types of features are
agreement with a previous report [8]. Similar behavior observed in the AFM images: a uniform grayish contrast
(meaning growth via bilayer units) was obtained for the that represents the matrix phase (and the major surface
films grown with different Ti cell temperatures (i.e., growth feature) and significantly brighter spots that stand out from
rates), substrate temperatures, and ozone flux values. The the uniform contrast and represent inclusions in the films.
absence of RHEED oscillations for films grown on SrTiO3 When comparing the Figs. 4(a)–(c), the grayscale changes
and LSAT is reflective of the more diffuse nature and in the uniform background decrease on going from SrTiO3
increased spottiness of the RHEED patterns, indicating to LSAT to LaAlO3. The overall grayscale is similar,
that those films surfaces are rougher than surfaces of films however, between all the images because both LSAT and
on LaAlO3. LAO exhibit bright white features indicative of large
Our goal was to understand what effects other growth protruding objects from the film. Electron microscopy
parameters had on structural/surface quality. In Fig. 3, we experiments (not shown here) carried out by us on films
4. ARTICLE IN PRESS
1496 P. Fisher et al. / Microelectronics Journal 37 (2006) 1493–1497
Fig. 5. y–2y scans of TiO2 films on SrTiO3(1 0 0), LSAT(1 0 0), and
LaAlO3(1 0 0).
though the surface of the anatase matrix is tending to
flatten out at higher temperatures. For films deposited on
LaAlO3 using the lower ozone flux of 0.25 sccm (Figs. 4(d
and f)), the rms roughness increased somewhat with
˚
substrate temperature, going from 62 A at T ¼ 550 1C
(Fig. 4(f)) to 102 A ˚ at T ¼ 750 1C (Fig. 4(d)).
Importantly, films grown under higher ozone fluxes were
smoother at all substrate temperatures. Figs. 4(e and f)
Fig. 4. (a–c) are AFM images of TiO2 films grown at 750 1C using a Ti correspond to the RHEED images in Figs. 3(b and a),
source temperature of 1550 1C and an ozone flux of 1 sccm on (a)
respectively. The change in film surface with ozone pressure
SrTiO3(1 0 0); (b) LSAT(1 0 0); and on LaAlO3(1 0 0); (d–f) are AFM
images of TiO2 films grown on LaAlO3(1 0 0) using a Ti source that we previously observed by RHEED corresponded to a
temperature of 1550 1C at (d) 750 1C, 0.25 sccm O3; (e) 550 1C, 1.00 sccm major change in surface roughness, going from a rms
O3; and (f) 550 1C, 0.25 sccm O3. ˚
roughness of 62 A at 0.25 sccm O3 to an rms roughness of
8A ˚ at 1.00 sccm. In general for the films deposited using
1.00 sccm O3, the rms roughnesses were much lower than
deposited on LaAlO3 agreed with Chambers et al’s [8] for films grown at 0.25 sccm, increasing from 8 A at ˚
observations that these objects are rutile inclusions that T ¼ 550 1C (Fig. 4(e)) to 22 A ˚ at T ¼ 650 1C (not shown)
have coherent interfaces with, and protrude out from the ˚
to 52 A at T ¼ 750 1C (Fig. 4(c)). Note again that the
flat surface of, the anatase matrix. increase in rms roughnesses with increasing temperature
Pronounced relationships between both substrate tem- appears to arise from the increase in the large inclusions.
perature and ozone flux were observed, as illustrated in Over the regions of the surface where there are no
Figs. 4(c–f). Figs. 4(d–f) are AFM images of TiO2 films inclusions, the rms roughnesses were very low: 8 A at ˚
grown on LaAlO3(1 0 0) using a Ti source temperature of ˚ ˚
T ¼ 550 1C, 4 A at T ¼ 650 1C, and 7 A at T ¼ 750 1C.
1550 1C at (d) 750 1C, 0.25 sccm O3; (e) 550 1C, 1.00 sccm Clearly the inclusions play a major role in the overall
O3; and (f) 550 1C, 0.25 sccm O3. At higher temperatures surface roughness.
(650 1C and above), large inclusions became increasingly The XRD spectra of TiO2 films deposited on all 3
common, as observed in Figs. 4(c and d). At higher substrates at 750 1C using a Ti-cell temperature of 1550 1C
temperatures, the stable rutile phase is more likely to and an ozone flux of 1 sccm are shown in Fig. 5. These
nucleate during growth and the metastable anatase is more XRD patterns reveal that highly (0 0 1)-oriented anatase
likely to transform irreversibly to the stable rutile. Based films grew on SrTiO3 (1 0 0), LSAT (1 0 0), and LaAlO3
on our experimental evidence from TEM and the above (1 0 0) substrates. In fact, similar XRD results were
given thermodynamic arguments, it seems likely that the observed by us for anatase films deposited over a wide
large inclusions observed in the AFM images for the high- range of conditions (4501oTdo750 1C and 0.25 o flux O3
temperature films are rutile. The inclusions greatly increase o 2.00 sccm). f-scans showed that these films were
the overall root-mean-square (rms) roughness values, even all epitaxial and all had the same relationship:
5. ARTICLE IN PRESS
P. Fisher et al. / Microelectronics Journal 37 (2006) 1493–1497 1497
(0 0 1)TiO2 ||(1 0 0)Subs; [0 1 0]TiO2 ||[0 1 0]Subs. The full-widths N00014-03-1-0665. Any opinions, findings, and conclusions
at half-maximum (FWHM) of the rocking curve on the or recommendations expressed in this material are those of
anatase (0 0 4) peaks were largest for the films on the authors and do not necessarily reflect the views of the
SrTiO3(1 0 0) (E0.71), intermediate on LSAT (0.11), and Office of Naval Research. It also made use of the facilities
lowest (E0.031) on LaAlO3(1 0 0). The FWHM for the maintained through the MRSEC program of the National
substrates themselves were all E0.021 under our diffraction Science Foundation under Award no. DMR-0520425.
conditions, indicating that the films on LaAlO3 substrates
were of superior crystalline quality to the other two films,
in spite of the inherent twinning in the LaAlO3 substrate.
These diffraction observations are similar in nature to References
those observed in other reports using other deposition
methods [11,16] and using MBE deposition [8], although [1] J.G.E. Jellison, F.A. Modine, L.A. Boatner, Opt. Lett. 22 (1997)
no other reports exist on MBE growth of TiO2 films on the 1808–1810.
[2] M.R. Hoffman, S.T. Martin, W. Choi, D.W. Bahnemann, Chem.
LSAT(1 0 0) surface.
Rev. 95 (1995) 69–96.
It should be pointed out that under the growth [3] A.L. Linsebigler, G. Lu, J. John, T. Yates, Chem. Rev. 95 (1995)
conditions investigated here, temperature seemed to play 735–758.
the largest role on inclusion formation; substrate choice, [4] M. Koelsch, S. Cassaignon, J.F. Guillemoles, J.P. Olivet, Thin Solid
ozone flux, and Ti-cell temperature played secondary roles, Films 403–404 (2002) 312–319.
if any. Interestingly, the inclusions did not appear to affect [5] P.A.M. Hotsenspiller, G.A. Wilson, A. Roshko, J.B. Rothman, G.S.
either the RHEED pattern or the XRD pattern in an Rohrer, J. Cryst. Growth 166 (1996) 779–785.
[6] D.R. Burgess, P.A.M. Hotsenspiller, T.J. Anderson, J.L. Hohman,
obvious manner. Nevertheless, electron microscopy experi-
J. Cryst. Growth 166 (1996) 763–768.
ments carried out by us on films grown on LaAlO3 (not [7] H.L.M. Chang, H. You, Y. Gao, J. Guo, C.M. Foster,
shown here) agreed with observations given in Ref. [8] R.P. Chiarello, T.J. Zhang, D.J. Lam, J. Mater. Res. 7 (1992)
demonstrating that rutile inclusions protrude out of the 2495–2506.
matrix anatase phase in films grown at high temperatures. [8] S.A. Chambers, C.M. Wang, S. Thevuthasan, T. Droubay, D.E.
It should be noted that no rutile inclusions are observed in McCready, A.S. Lea, V. Shutthanandan, J.C.F. Windisch, Thin Solid
pure TiO2 films deposited using pulsed laser deposition on Films 418 (2002) 197–210.
LaAlO3(1 0 0) substrates at T ¼ 750 1C in elevated pres- [9] S. Yamamoto, T. Sumita, T. Yamaki, A. Miyashita, H. Naramoto,
J. Cryst. Growth 237–239 (2002) 569–573.
sures of pure oxygen (PE200 m Torr O2). Further work is
[10] T. Aoki, K. Maki, Q. Tang, Y. Kumagai, S. Matsumoto, J. Vacuum
required to understand the thermodynamics and kinetics of Sci. Technol. A 15 (1997) 2485–2488.
the nucleation and growth of these rutile inclusions during [11] S. Chen, M.G. Mason, H.J. Gysling, G.R. Paz-Pujalt, T.N. Blanton,
thin film deposition. T. Castro, K.M. Chen, C.P. Fictorie, W.L. Gladfelter, A. Franciosi,
In conclusion, TiO2 thin films were grown on SrTiO3 P.I. Cohen, J.F. Evans, J. Vacuum Sci. Technol. A 11 (1993)
(1 0 0), LSAT (1 0 0), and LaAlO3 (1 0 0) substrates by 2419–2429.
reactive MBE and characterized in-situ by RHEED and ex- [12] V.A. Burbure, P.A. Salvador, G.S. Rohrer, J. Am. Ceram. Soc.
(2006), in press, Doi:10.1111/j.1551-2916.2006.01168.x.
situ by XRD and AFM. Films grown on SrTiO3 (1 0 0),
[13] P.J. Hansen, V. Vaithyanathan, Y. Wu, T. Mates, S. Heikman, U.K.
LSAT (1 0 0), and LaAlO3 (1 0 0) grew as (0 0 1) anatase, Mishra, R.A. York, D.G. Schlom, J.S. Speck, J. Vacuum Sci.
which was found to grow over a range of temperatures Technol. B 23 (2005) 499–506.
(450–750 1C). It was observed that the films’ structural [14] X. Liu, X.Y. Chen, J. Yin, Z.G. Liu, J.M. Liu, X.B. Yin, G.X. Chen,
quality and morphology highly depended on substrate J. Vacuum Sci. Technol. A 19 (2001) 391–393.
temperature and ozone flux. Films were found to grow with [15] Sugiharto, S. Yamamoto, T. Sumita, A. Miyashita, J. Phys. Condens.
a smoother surface at higher ozone fluxes, and films grown Matter 13 (2001) 2875–2881.
[16] Y. Gao, S. Thevuthasan, D.E. McCready, M. Engelhard, J. Cryst.
at higher substrate temperatures exhibited large inclusions.
Growth 212 (2000) 178–190.
It is interesting that these results indicate that the values for
[17] G.S. Herman, Y. Gao, T.T. Tran, J. Osterwalder, Surf. Sci. 447
ozone flux, Ti flux, and temperature have not been (2000) 201–211.
optimized for the growth of high-quality anatase TiO2 [18] M. Kamei, T. Mitsuhashi, Surf. Sci. 463 (2000) L609–L612.
exhibiting very narrow rocking curves. To realize such films, [19] J.H. Haeni, C.D. Theis, D.G. Schlom, W. Tian, X.Q. Pan, H. Chang,
a better understanding is required of the thermodynamic I. Takeuchi, X.D. Xiang, Appl. Phys. Lett. 78 (2001) 3292–3294.
and kinetic processes that occur in MBE and that govern [20] W. Gao, C.M. Wang, H.Q. Wang, V.E. Henrich, E.I. Altman, Surf.
rutile nucleation and growth on perovskites surfaces. Sci. 559 (2004) 201–213.
[21] E. Coleman, T. Siegrist, D.A. Mixon, P.L. Trevor, D.J. Trevor,
J. Vacuum Sci. Technol. A 9 (1991) 2408–2409.
Acknowledgements [22] V. Leca, G. Rjinders, G. Koster, D.H.A. Blank, H. Rogalla, Mater.
Res. Soc. Symp. Proc. 587 (2000) O3.6.1–O3.6.4.
This work was supported by the Office of Naval Research [23] A. Asthagiri, C. Niederberger, A.J. Francis, L.M. Porter, P.A.
under grants N0001-05-1-0238, N00014-05-1-0152, and Salvador, D.S. Sholl, Surf. Sci. 537 (2003) 134–152.