You are on page 1of 5

ARTICLE IN PRESS

Journal of Crystal Growth 299 (2007) 218222 www.elsevier.com/locate/jcrysgro

Atomic layer deposition of aluminum oxide on hydrophobic and hydrophilic surfaces


Nobuhiko P. Kobayashia,b,, Carrie L. Donleya, Shih-Yuan Wanga, R. Stanley Williamsa
a

Quantum Science Research, Advanced Studies, Hewlett-Packard Laboratories, 1501 Page Mill Road, Palo Alto, CA, USA b School of Engineering, University of California, Santa Cruz, 1156 High Street, Santa Cruz, CA, USA Received 26 September 2006; received in revised form 3 November 2006; accepted 7 November 2006 Communicated by M. Kawasaki

Abstract Aluminum oxide was deposited at 45 1C by atomic layer deposition onto an atomically smooth gold surface coated with a CH3terminated alkanethiolate self-assembled monolayer (SAM) and onto an OH-terminated silicon dioxide surfaces. The growth of the resulting lms was characterized with reection absorption infrared spectroscopy, contact-angle measurement, and atomic force microscope. Aluminum oxide lms on the SAMs exhibited a growth instability, while the lms on the OH-terminated silicon dioxide maintained an atomically smooth surface. r 2007 Elsevier B.V. All rights reserved.
PACS: 81.15.Gh Keywords: A1. Atomic force microscopy; A1. Roughening; A1. Surfaces; A3. Atomic layer deposition; B1. Oxides

In optimizing devices in which organic materials are used as active components, employing encapsulation lms that act as gate insulators and shield the organic materials during processing and effectively prevent water and oxygen from penetrating into the organic materials during device operation is often desirable. Although the formation of organic lms on inorganic surfaces (organic/inorganic) has been studied extensively, in particular, in the development of organic eld effect transistors [1], detailed studies on the deposition of inorganic lms onto organic surfaces (inorganic/organic) have been dominated by metals deposited on self-assembled monolayers (SAMs) [25]. However, various metal oxides are also of interest in developing the encapsulation lms on organic surfaces [69]. In this paper, we describe the growth of aluminum oxide (AlOx) on CH3terminated SAMs (hydrophobic surfaces) by atomic layer deposition (ALD) (Model Savanna 100, Cambridge NanoTech, Inc., Cambridge, Massachusetts, USA) at low temperature to minimize the chemical and physical impact
Corresponding author. Tel.: +1 650 857 4660.

E-mail address: nobuhiko.kobayashi@hp.com (N.P. Kobayashi). 0022-0248/$ - see front matter r 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.jcrysgro.2006.11.224

on the organic layers. The characteristics and evolution of the AlOx lms are compared to those grown on OHterminated SiO2 (hydrophilic surfaces). First, a gold lm with an atomically smooth surface was prepared on glass substrates via our template-stripping process [10]. Subsequently SAMs of CH3(CH2)17SH were formed on the template-stripped gold (TS-gold) lm by immersion into an ethanol solution containing the alkanethiolate (molar concentration of 0.01 M) for extended times (424 h) at room temperature. The TS-gold surface having atomically smooth surface proles ensures that CH3SAMs are formed with a minimum number of structural defects [11]. This surface is expected to be strongly hydrophobic in the early stage of the deposition of AlOx. Hydrophilic OH-terminated silicon oxide surfaces (OHSiO2) were also prepared on Si(1 0 0) substrates, which were compared to the SAM surfaces as a substrate for the deposition of AlOx lms. The AlOx lms were deposited using water and trimethylaluminum (TMAl) as the sources for oxygen and aluminum, respectively. The ALD process was performed by alternatively supplying pulses of nitrogen

ARTICLE IN PRESS
N.P. Kobayashi et al. / Journal of Crystal Growth 299 (2007) 218222 219

gas containing either water or TMAl. The substrate temperature was set to 45 1C for all samples, which is lower than those temperatures reported to cause SAMs to degrade structurally [12]. We implemented an incubation period at the beginning of the AlOx deposition that consisted of a 140 ms water pulse followed by a 60 s reaction period. After that a 140 ms TMAl pulse was followed by a 60 s reaction period to promote the formation of AlOx nuclei. After the initial incubation period, one cycle of the ALD process consisted of a 140 ms water pulse, a 160 s nitrogen purge period, a 140 ms TMAl pulse, and a 16 s nitrogen purge period. These specic deposition conditions were previously optimized for AlOx lms grown at 45 1C in order to maintain a self-limited deposition rate of 0.12 nm/cycle on hydrogen terminated silicon surfaces. Relatively longer purge periods were necessary at lower temperatures [13]. Dissociative reaction, including ligand exchange, of TMAl on SAMs with a CH3 terminal group has been found to be nearly quenched because of a small thermodynamic driving force and a large kinetic barrier [8]. This relative inertness of TMAl on CH3SAMs is potentially an advantage for growing an AlOx barrier layer, i.e. the SAMs will not be impacted by TMAl. Furthermore, water used as an oxygen source is not reactive with alkane chains, but is expected to form droplets on the hydrophobic surface of the densely ordered SAMs. A reaction of these droplets with pyrophoric TMAl should form AlOx nuclei. The evolution of the AlOx lms was characterized by reectionabsorption infrared spectroscopy (RAIRS), water contact angle measurement and a non-contact mode atomic force microscope (AFM). Shown in Fig. 1(a) and (b) are RAIRS spectra taken from AlOx samples after specied number of ALD cycles. For the RAIRS method used here, the strong electromagnetic reection of the metal substrates eliminates electric elds parallel to the surface so that only the surface-normal components of the molecular vibrations in the samples can be detected [14]. Fig. 1(a) displays the progressive changes observed in the range of 30002800 cm1 for the four peaks numbered 14 associated with four main CH stretching modes, asymmetric CH3 (a-CH3), asymmetric CH2 (a-CH2), symmetric CH3 (s-CH3), and symmetric CH2 (s-CH2), of the alkanethiolate chains, respectively. The spectra taken from the samples with more than 200 ALD cycles are essentially identical to the 200 ALD cycle spectrum, and therefore they are not shown in Fig. 1(a). The spectrum of the 0 ALD cycle sample indicates that the as-formed SAMs were wellordered [15], judged by the positions and widths of the a-CH2 and s-CH3 peaks. Among several notable features in Fig. 1(a), the CH3 peaks (peak 1 and 3) virtually disappeared at the onset of the AlOx deposition, which suggests that the CH3 terminal group of the SAMs was perturbed during the early stage of the AlOx deposition. This contrasts to the case where 0.37 nm of elemental aluminum was sublimated onto CH3-terminated alkanethiolate SAMs in ultra high vacuum in that all four

a
5.0x10-3 1 4.0x10-3 Absorbance 200 3.0x10
-3

3 4

150 100

2.0x10-3

65 35 25 0 2900 Wavenumber (cm-1) 2800

1.0x10-3

0.0 3000

b
0.24 400 350 Absorbance 0.20 300

0.16

250 200 150 100 65 35 25 1200 1000 Wavenumber (cm-1) 800 0

0.12

0.08

Fig. 1. RAIRS spectra collected from AlOx samples after specied number of ALD cycles. (a) displays the progressive changes observed for the four peaks, numbered 14, associated with four CH stretching modes of the SAMs. (b) shows progressive changes of a RAIRS peak associated with several overlapping AlO modes.

CH peaks remained essentially unchanged. In one case, the aluminum penetrated the monolayer with the CH3 terminal group to reach the SAM/Au interface, leaving the molecules intact [3]. In our ALD process, however, we implemented an induction cycle in which water and pyrophoric TMAl necessarily reacted with each other on top of the SAM. These reactions may have disordered the lm, and it is possible that reactive species created by the waterTMAl reaction attacked the alkanethiolate chain, which may have resulted in a change in the orientation of the CH3 terminal group. In contrast to the instantaneous disappearance of the CH3 peaks, the CH2 peaks were found to be present even on the 400 ALD cycle sample. As seen in Fig. 1(a), both the a-CH2 and s-CH2 stretching mode peaks exhibited signicant broadening and peak shift to higher wavenumbers even

ARTICLE IN PRESS
220 N.P. Kobayashi et al. / Journal of Crystal Growth 299 (2007) 218222

after only 25 ALD cycles, which is another indication of structural disorder within the alkanethiolate chains. This picture contrasts strongly with the behavior for the deposition of elemental aluminum onto CH3-terminated SAMs [3]. The large shift in both a-CH2 and s-CH2 stretching mode peaks observed in Fig. 1(a) may have indicated that the alkane chains underwent a structural phase transition from an ordered to an amorphous solid [16]. Progressive changes of a RAIRS peak associated with several AlO modes are presented in Fig. 1(b). The peak positions shifted slightly to higher wavenumbers between 25 and 250 ALD cycles, and later, they stayed at 919 cm1 through 400 ALD cycles. These broad spectra are due to several overlapping AlO modes, but the peak position at 919 cm1 would imply that a dominant component is g-Al2O3 [17], which has been identied to be a primary phase formed via the oxidation of aluminum at low temperatures [18]. Plotted in Fig. 2(a) is the integral intensity of the AlO RAIRS peaks as a function of the equivalent thickness, which was obtained from two cross-reference samples, as referenced to ALD deposited AlOx on hydrogen terminated Si, with known thickness measured by spectroscopic ellipsometry. Shown in Fig. 2(b) are water contact angle (yc) measurements on the surfaces of the AlOx both on the hydrophobic CH3SAM surfaces (open circles) and the hydrophilic OHSiO2 surfaces (solid triangles). The contact angle yc on the surface of freshly prepared alkanethiol SAMs was $1061, consistent with the numbers previously reported [19]. On the CH3SAM surfaces, once AlOx was deposited, yc decreased steadily as the equivalent thickness increased until it reached 2530 nm. Subsequently, yc appeared to increase slowly. In contrast, yc on the AlOx deposited on the OHSiO2 remained almost unchanged from the 0 ALD cycle sample. The trend seen in Fig. 2(b) will be discussed in accordance with the AFM results later. In Fig. 3, representative non-contact mode AFM images collected on the AlOx both on the CH3SAM in (a)(d) and the OHSiO2 in (e)(h) for the 0, 100, 200 and 400 ALD cycle samples, respectively, are displayed. The surface of the freshly prepared SAM in (a) exhibited a root-mean square roughness (Rrms) of 0.24 nm. As shown in (b), after 100 ALD cycles, the surface was found to be covered with numerous islands with a mean height of 3.170.7 nm and width of 13.673.2 nm measured at half-height (several islands are marked with arrows). In addition to numerous islands, several aggregates (marked 14) that seem to consist of 36 islands each were also observed. On the 200 ALD cycle sample in (c), well-developed domes with a mean height of 11.770.3 nm and width of 42.571.5 nm were seen (two such domes are marked with arrows), which presumably are the successors of the aggregates seen in (b). In (d), the domes appeared to have merged. The Rrms increased from 0.24 nm on the fresh SAMs to 1.23, 2.16 and 8.09 nm for 100, 200 and 400 cycle samples, respectively. A gradual increase in the coverage of the hydrophilic ALD AlOx on the hydrophobic CH3SAM surface seen in Fig. 3(a)(d) accounts for the monotonic

35

a
30 25 Integral absorbance

20

15

10

b
100

Water contact angle (degree)

80

60

40

20

0 0 10 20 30 40 Equivalent AlOx thickness (nm)


Fig. 2. (a) shows the integral intensity of the AlO RAIRS peaks (Fig. 1(b)) as a function of the equivalent thickness of the AlOx layers. (b) is a plot of water contact angle (yc) on the surfaces of the AlOx both on the hydrophobic CH3SAM surfaces (open circles) and the hydrophilic OHSiO2 surfaces (solid triangles). On the CH3SAM surfaces, once AlOx was deposited, yc decreased steadily as the equivalent thickness increased until it reached 2530 nm. Subsequently, yc appeared to increase slowly. In contrast, yc on the AlOx deposited on the OHSiO2 remained almost unchanged from the 0 ALD cycle sample.

decrease in the contact angle shown in Fig. 2(b). In addition, as seen in Fig. 2(b), the water contact angle slowly increases as the surface further coarsens (Fig. 3(c) and (d)), which could be attributed to a dependence of the water contact angle on the surface morphology. Many systems exhibit an approximately linear dependence of the contact angle on the rms roughness of the substrate [20], which suggests that there is a feed-back mechanism that

ARTICLE IN PRESS
N.P. Kobayashi et al. / Journal of Crystal Growth 299 (2007) 218222 221

operates in the AlOx ALD process (or any other material that uses water as a source) on an initially hydrophobic surface. Once the growing lm becomes rough, the tendency of liquid lms to have a higher contact angle

Fig. 3. Representative non-contact mode AFM images collected on the AlOx both on the CH3SAM in (a)(d) and OHSiO2 in (e)(h) for the 0, 100, 200 and 400 ALD cycle samples, respectively.

drives the growth instability. In contrast, the surface of the OHSiO2 in (e), exhibited an atomically smooth morphology with Rrms of 0.11 nm, and the surface qualitative morphology of the AlOx on the OHSiO2 substrate was essentially unchanged, as shown in panels (f)(h), with slight increases in Rrms. Table 1 summarizes the characteristics of the AlOx deposition on both the OHSiO2 and CH3SAM. The fth column represents the ratio of Rrms to the thickness (d) after a specic ALD cycle. As clearly seen, the ratio decreases for the OHSiO2 while the ratio increases for the CH3SAM substrate, which shows that there is a growth instability for AlOx ALD on the CH3SAM surfaces. The trend seen for the OHSiO2 can be attributed normal kinetic roughening often seen at low temperatures [21]. Table 1 is further visualized in Fig. 4 where representative line scans collected from the AFM images in Fig. 3 are displaced upward by the average lm thickness. Fig. 4(a) shows that there is a growth instability for AlOx ALD on the CH3SAM surfaces, while panel (b) shows normal kinetic roughening on the OHSiO2 surfaces. As a conclusion, a qualitative picture of the AlOx ALD on hydrophobic CH3SAM surfaces can be drawn as follows. The linear correlation seen between the integral intensity and the equivalent AlOx thickness in Fig. 2(a) indicates that the total amount of AlOx accumulated on the SAM is proportional to the equivalent thickness, that is, surface morphology at any given total amount of AlOx can be examined in terms of how a given total amount of AlOx is distributed over the SAM surfaces. On hydrophobic CH3SAM surfaces, TMA is expected to react spontaneously with water droplets pre-existing on the surfaces, which results in densely packed islands Fig. 3(b). The formation of well-developed domes then, as in Fig. 3(c), initiates as a result of merging islands. Beyond this evolution stage, the surface develops a noticeably rough morphology as in Fig. 3(d). In contrast, water droplets do not form on hydrophilic OHSiO2 surfaces, therefore, the morphology remains smooth as seen in Fig. 3(f)(h). In designing ALD processes, formulating appropriate precursors is critical [22], however, our comparative studies of

Table 1 A comparative summary of the characteristics of the AlOx deposition on the hydrophilic OHSiO2 and hydrophobic CH3SAM surfaces Surface CH3SAM Number of ALD cycles N (cycles) 100 200 400 100 200 400 Thicknessa d (nm) 8.2 17.6 39.2 10.6 20.3 35.9 rms roughnessb Rrms (nm) 1.23 2.16 8.09 0.24 0.27 0.37 Rrms/d 1.5 101 1.2 101 2.1 101 2.3 102 1.3 102 1.0 102 d/N (nm/cycle) 8.2 102 8.8 102 9.8 102 1.1 101 1.0 101 9.0 102

OHSiO2

The fth column represents the ratio of Rrms to the thickness (d) after a specic ALD cycle. The ratio decreases for the OHSiO2 surface while it increases for the CH3SAM surface, which shows that there is a growth instability for AlOx ALD on the latter surface. a Measured by spectroscopic ellipsometry. b Measured by atomic force microscope.

ARTICLE IN PRESS
222 N.P. Kobayashi et al. / Journal of Crystal Growth 299 (2007) 218222

50

a
400

40 Thickness (nm)

the precursors in the ALD process, which suggests that surface wettability must be explicitly controlled for specic reactants to maintain atomically smooth surface morphologies during lm deposition.

30

References
20 200 10 100 0 0 50
[1] C.D. Dimitrakopoulos, D.J. Mascaro, IBM J. Res. Dev. 45 (2001) 11. [2] A.S. Killampalli, P.F. Ma, J.R. Engstrom, J. Am. Chem. Soc. 127 (2005) 6300. [3] A. Hooper, G.L. Fisher, K. Konstadinidis, D. Jung, H. Nguyen, R. Opila, R.W. Collins, N. Winograd, D.L. Allara, J. Am. Chem. Soc. 121 (1999) 8052. [4] J.J. Senkevich, F. Tang, D. Rogers, J.T. Drotar, C. Jezewski, W.A. Lanford, G.-C. Wang, T.-M. Lu, Chem. Vapor Deposition 9 (2003) 258. [5] G.L. Fisher, A.E. Hooper, R.L. Opila, D.L. Allara, N. Winograd, J. Phys. Chem. B 104 (2000) 3267. [6] H. Shin, R.J. Collins, M.R. De Guire, A.H. Heuer, C.N. Sukenik, J. Mater. Res. 10 (1995) 692. [7] P. Wohlfart, J. Weib, J. Kashammer, M. Kreiter, C. Winter, R. Fisher, S. Mittler-Neher, Chem. Vapor Deposition 5 (1999) 165. [8] Y. Xu, C.B. Musgrave, Chem. Mater. 16 (2004) 646. [9] J.P. Lee, Y.J. Jang, M.M. Sung, Adv. Funct. Mater. 13 (2003) 873. [10] J.J. Blackstock, Z. Li, M.R. Freeman, D.R. Stewart, Surf. Sci. 546 (2003) 87. [11] J. Yang, J. Han, K. Isaacson, D.Y. Kwok, Langmuir 19 (2003) 9231. [12] N. Prathima, M. Harini, N. Rai, R.H. Chandrashekara, K.G. Ayappa, S. Sampath, S.K. Biswas, Langmuir 21 (2005) 2364. [13] M.D. Groner, F.H. Fabreguette, J.W. Elam, S.M. George, Chem. Mater. 16 (2004) 639. [14] M.K. Debe, J. Appl. Phys. 55 (1984) 3354. [15] M. Lestelius, I. Engquist, P. Tengvall, M.K. Chaudhury, B. Liedberg, Colloids Surf. B 15 (1999) 587. [16] M.D. Porter, T.B. Bright, D.L. Allara, C.E.D. Chidsey, J. Am. Chem. Soc. 109 (1987) 3559. [17] J.Y. Ying, J.B. Benziger, H. Gleiter, Phys. Rev. B 48 (1993) 1830. [18] N.P. Kobayashi, J.T. Kobayashi, W.-J. Choi, P.D. Dapkus, Appl. Phys. Lett. 73 (1998) 1553. [19] C.D. Bain, E.B. Troughton, Y.-T. Tao, J. Evall, G.M. Whitesides, R.G. Nuzzo, J. Am. Chem. Soc. 111 (1989) 321. [20] S.J. Hitchcock, N.T. Carroll, M.G. Nicholas, J. Mater. Sci. 16 (1981) 714. [21] W.M. Tong, R.S. Williams, Annu. Rev. Phys. Chem. 45 (1994) 401. [22] P. de Roufgnac, A.P. Yousef, K.H. Kim, R.G. Gordon, Electrochem. Solid State Lett. 9 (2006) F45.

40 Thickness (nm) 400 30

20

200

10 100 0 0 100 200 300 400 0 500

Distance (nm)
Fig. 4. AFM line scans collected from the images in Fig. 3 for the corresponding ALD cycles as indicated. (a) and (b) represent the line scans for the AlOx lms deposited on the hydrophobic CH3SAM and hydrophilic OHSiO2 surfaces, respectively. The line scans are displaced upwards by the average lm thickness. (a) clearly shows that there is a growth instability for ALD AlOx on the hydrophobic CH3SAM surfaces, while (b) shows normal kinetic roughening on the hydrophilic OHSiO2 surfaces.

the hydrophobic and hydrophilic surfaces demonstrate that the observed growth instability on the CH3SAM is associated with the characteristics of surface wetting by

You might also like