Role of Water in Super Growth ofSingle-Walled Carbon Nanotube CarpetsPlacidus B. Amama,†,‡ Cary L. Pint,§ Laura McJilton,| Seung Min Kim,⊥Eric A. Stach,⊥ P. Terry Murray,¶ Robert H. Hauge,| and Benji Maruyama*,†

Air Force Research Laboratory, Materials and Manufacturing Directorate, AFRL/RX,Wright-Patterson Air Force Base, Ohio 45433, Richard E. Smalley Institute forNanoscale Science and Technology, Department of Physics and Astronomy, andDepartment of Chemistry, Rice UniVersity, Houston Texas 77005, UniVersalTechnology Corporation, Dayton, Ohio 45432, School of Materials Engineering andBirck Nanotechnology Center, Purdue UniVersity, West Lafayette, Indiana 47907, andUDRI, UniVersity of Dayton, Dayton, Ohio 45469

Received June 28, 2008; Revised Manuscript Received October 27, 2008

ABSTRACT

The Ostwald ripening behavior of Fe catalyst films deposited on thin alumina supporting layers is demonstrated as a function of thermalannealing in H2 and H2/H2O. The addition of H2O in super growth of single-walled carbon nanotube carpets is observed to inhibit Ostwaldripening due to the ability of oxygen and hydroxyl species to reduce diffusion rates of catalyst atoms. This work shows the impact of typicalcarpet growth environments on catalyst film evolution and the role Ostwald ripening may play in the termination of carpet growth.

Vertically oriented single-walled carbon nanotube (SWNT)carpets or forests grown by catalytic chemical vapor deposi-tion (CVD) have received enormous attention because oftheir suitability in a growing number of important techno-logical applications.1 Although the ultralong, aligned SWNTSwhich can be attained by this growth method are promisingfor such applications, the lack of understanding associatedwith growth termination after micron-tall carpets are attainedlimits their widespread appeal.2 This problem is compoundedby the limited understanding of the complicated processassociated with carpet growth. So far, termination of carpetgrowth has typically been framed as a “poisoning” effectthat occurs at high temperatures.3 Although the mechanismbehind poisoning is not yet understood, there have beenstudies which attribute the termination of carpet growth tothe accumulation of amorphous carbon on active catalystsites4,5 or the interdiffusion of the catalyst with the substrateresulting in the formation of a silicide.6 Recent work byHarutyunyan et al. emphasizes that the size of the catalystparticle can play a key role in stabilizing a carbide phasebased on pressure dependence in accordance with the

Young-Laplace equation.7 On the other hand, a recent studyby Han et al.8 demonstrates that both diffusion limitationsand spontaneous catalyst deactivation are not the dominantmechanisms of growth termination in their SWNT carpets.They hypothesize that growth termination occurs as a resultof a chemical-mechanical coupling of the top surface of thefilm, which causes an energetic barrier to the relativedisplacement between neighboring nanotubes. Without men-tion, this topic of growth termination remains as one of themost important and elusive concepts in the field of carpetgrowth, leading to many hypothetical explanations for“poisoning,” with no dominant mechanism emerging.

To promote and sustain the growth of SWNT carpets,Iijima’s group advanced an approach whereby H2O was usedas a protective agent against amorphous carbon coating. Thewater-assisted CVD growth (also referred to as “supergrowth”) revealed that the activity and lifetime of thecatalysts are dramatically enhanced by introducing a well-defined, limited amount of H2O into the growth chamber,resulting in the rapid growth of highly dense, verticallyaligned SWNT carpets of high purity with heights up to 2.5mm after 10 min.9 Zhang et al.10 also reported a molecularoxygen-assisted growth of ultrahigh-yield SWNT carpetswith heights up to 10 µm after 10 min by plasma-enhancedCVD. These results emphasize that oxidants are generallycapable of increasing the activity and lifetime of the catalystfor growth of carpets. Super growth represents a majorbreakthrough in the field; the catalyst activity (i.e., number

* To whom correspondence should be addressed. E-mail: Benji.Maruyama@wpafb.af.mil.

† Air Force Research Laboratory.‡ Universal Technology Corporation.§ Department of Physics and Astronomy, Rice University.| Department of Chemistry, Rice University.⊥ School of Materials Engineering and Birck Nanotechnology Center,

Purdue University.¶ UDRI, University of Dayton.

NANOLETTERS

2009Vol. 9, No. 1

44-49

10.1021/nl801876h CCC: $40.75 2009 American Chemical SocietyPublished on Web 12/01/2008

of catalysts which grow nanotubes) in super growth isestimated to be 84% ((6%), the highest ever recorded fornanotube growth, and it is expected that growth with catalystactivity approaching 100% may be achieved in the future.11

This motivates our desire to understand the mechanism bywhich catalytic activity is preserved when an oxidizing agentis present during growth, and specifically to understand therole that a dynamic catalyst ripening process may play inthis growth technique.

Ostwald ripening is a phenomenon whereby larger particlesgrow in size while smaller particles, having higher strainenergy, shrink in size and eventually disappear via atomicinterdiffusion. As Ostwald ripening proceeds, the numberof particles decreases while the average catalyst diameterand the spread in the particle size distribution increases.Previous studies have shown that the catalyst size distributionand nanotube growth are strongly dependent on conditionsin which the catalyst film is annealed.12,13 To complimentthese studies, the data we present here provide evidence thatan important difference exists between the catalyst morphol-ogy evolution during thermal annealing with and withoutwater, and that this can be fully interpreted in the frameworkof Ostwald ripening.

The thermal annealing of the catalyst in hydrogen alone(H2) and with water (H2/H2O) and carpet growth with water(H2/H2O/C2H2) and without water (H2/C2H2) were performedin the same CVD chamber. A detailed description of theCVD chamber and the reaction conditions for SWNT carpetsare given elsewhere.14 The substrate consists of boron dopedsilicon (Si) (100) wafers with a 100 nm Al2O3 film depositedby atomic layer deposition (ALD) supporting a 0.5 nm thickFe film deposited ex-situ by e-beam evaporation. The furnacewas preheated to 750 °C, and the substrates were rapidlyinserted into the furnace and annealed in both H2 and H2/H2O at 1.4 Torr for 30 s and 5 min (without C2H2). Therespective flow rates were 400 sccm H2 and 2 sccm H2O.Prior to carrying out thermal annealing of the substrates inH2, the incoming gas lines were resistively heated for 60min at a background pressure of 1 × 10-6 Torr. This ensuredthe removal of residual H2O during these experiments. Afterthermal annealing, the samples were rapidly cooled. Thesame process was carried out when the catalyst was exposedafter carpet growth, except that 2 sccm C2H2 were includedin the reaction gas mixture, and a tungsten hot filament wasutilized in the first 30 s in order to quickly reduce the catalystparticles for growth. Following SWNT carpet growth, thecatalyst layers were exposed by removing the carpets usingtwo different techniques: oxidation and carpet lift-off, andoxidation and carpet burnoff in air. Prior to lift-off, thesamples were briefly heated in air to oxidize the SWNT ends,breaking the SWNT-catalyst bonds to aid the lift-off pro-cess.15 We found this to be effective in leaving the catalystlayer fully intact on the catalyst support. After the oxidationstep, the carpets were removed by using an adhesive tape.In order to substantiate that no catalyst was being selectivelyremoved from the substrate by the lift-off process, we alsooxidized the SWNTs in air at 600 °C to expose the catalyst.

A field-emission transmission electron microscope/scan-ning transmission electron microscope (S/TEM) (80-300Titan) from FEI Corporation was employed for studying theripening behavior of the catalyst. For TEM sample prepara-tion, the sample was cut to a 3 mm disk and the backside ofthe sample was hand-polished and dimpled down to about5-10 µm at the center of the sample. Then, the sample wasion-milled from the backsides at a 4.5° angle and at 4.5 kVusing a Gatan PIPSTM until the small hole at the center ofthe sample was made. The samples were further characterizedby atomic force microscopy [(AFM) Digital InstrumentNanoscope IIIa] operating in the tapping mode with a scanrate of 1.5 Hz.

To demonstrate the lifetime of the catalysts, the depen-dence of the SWNT carpet height and the continued growthof the carpet on the presence and absence of water in theCVD chamber is shown in Figure 1a. Using the super growthconditions, highly dense SWNT carpets of high quality weregrown; a representative FESEM image of a typical SWNTcarpet grown using 0.5% C2H2 with H2O is shown in Figure1b. Detailed characterization of the SWNT carpets is reportedelsewhere.16,17 In the presence of H2O, the carpet heightincreased linearly with growth time up to 120 min implyinga long catalyst lifetime, whereas in the absence of H2O, the

Figure 1. (a) Plot of the SWNT carpet height as a function of thegrowth time for reactions with H2O and with no oxidant. (b) FESEMimage of SWNT carpet grown using 0.5% C2H2 with H2O.

Nano Lett., Vol. 9, No. 1, 2009 45

carpet height remained roughly constant (∼10-20 µm),suggesting the early termination of growth occurs before 15min. Repeated experiments under super growth conditionsreveal that the carpet height continued to increase linearlyeven up to six hours. It is clear from these results and others9

that H2O plays a crucial role in sustaining SWNT carpetgrowth.

In order to understand the role of water during supergrowth, we have studied the morphological restructuring ofthe Fe catalyst film under typical CVD growth conditionsin H2 and H2/H2O at 750 °C as described previously. Theplan-view TEM images of the Fe2O3 catalyst nanoparticlesformed on the Al2O3 substrate after thermal annealing in H2

and H2/H2O for 30 s and 5 min are shown in Figure 2a-d.The TEM images reveal a significant difference in theripening behavior of the catalysts in these ambient even after30 s. Severe ripening of the catalyst is observed after thermalannealing in H2 for 5 min while the catalyst remainconsiderably stable even after 5 min of annealing in H2/H2O.This observation is consistent with the data presented in

Figure 1a, where we demonstrate that the presence of H2Oextends the lifetime of active catalyst particles during carpetgrowth under the same conditions.

The plots of the catalyst particle diameter distributions afterthermal annealing in H2 and H2/H2O are presented in Figure3a,b, respectively. The results were obtained from theirrespective TEM images by measuring the diameters of thenanoparticles in a 200 × 200 nm area. The number densitiesof catalyst nanoparticles for samples annealed for 30 s and5 min in H2 were 348 and 264 particles per (200 nm)2 whilethe number densities for samples annealed in H2/H2O for30 s and 5 min were 544 and 420 particles per (200 nm)2.The number of particles reduces with time, which isconsistent with Ostwald ripening, and the Ostwald ripeningrate is lower for the sample annealed in H2/H2O comparedto that annealed in H2 alone. From these results, it is clearthat the number densities of nanoparticles formed afterthermal annealing in H2/H2O are higher than those obtainedafter annealing in H2, consistent with Ostwald ripening. Sucha decrease in the number of particles observed over time

Figure 2. Plan-view TEM images of catalyst nanoparticles formed on the substrate after annealing in H2 for (a) 30 s and (b) 5 min, andin H2/H2O for (c) 30 s and (d) 5 min. A reduction in the number of catalyst particles occurs from 30 s to 5 min. Each catalyst particle thatdisappears can no longer support nanotube growth. The number of catalyst particles in the H2/H2O is higher than that of H2 alone, implyinga higher number of growing nanotubes.

46 Nano Lett., Vol. 9, No. 1, 2009

with larger particles growing at the expense of smaller oneshas also been observed for Au catalyst droplets on Si (111).18

Further, the mean diameters of the nanoparticles forcatalysts annealed in H2 after 30 s and 5 min were 10.1 (2.1 and 10.5 ( 5.3 nm while those annealed in H2/H2O were6.2 ( 1.4 and 6.8 ( 2.2 nm, respectively. The meandiameters were obtained from a measurement of 120 nano-particles. Unlike thermal annealing in H2, annealing in H2/H2O resulted in nanoparticles with smaller mean diametersand the standard deviations (narrower peak widths) as shownabove were also lower. From the TEM results shown inFigure 2, it is apparent that for samples annealed in H2 thereis significant coarsening of the nanoparticle catalysts withboth an increase in the mean diameter as well as a significantbroadening of the diameter distribution. These resultsdemonstrate that H2O suppresses the ripening of the catalystsand may also account for the narrow diameter distributionof SWNTs observed previously.14

Figure 4 shows the AFM topography images of the Fe2O3

nanoparticles formed on the substrate after thermal annealingin H2 and H2/H2O for 30 s. Consistent with the TEM results,the AFM scans reveal that the nanoparticles formed on thesubstrate upon thermal treatment in H2/H2O ambient aresignificantly smaller in size than those formed in H2 ambient.In fact, for samples thermally annealed in H2/H2O, it wasdifficult to resolve some of the smaller nanoparticles due tothe limited resolution of the AFM. AFM analysis revealedthat the mean feature heights of the nanoparticles formed

from Fe films thermally treated in H2 and H2/H2O are 3.5and 2.6 nm, respectively. It is apparent from the TEM andAFM results that the presence of H2O in the reaction chamberinhibits the ripening of the catalyst, thus preserving morecatalyst nanoparticles that have feature heights and diametersless than 3 and 6 nm, respectively. Note that 5-6 nm Fe2O3

nanoparticles correspond to Fe nanoparticles in the range of3-4 nm, which fall within the size range selective for SWNTgrowth. SWNTs with mean diameter of ∼3 nm are charac-teristic of growth by this technique and have been reportedand characterized by Pint et al.14,15,17 and Hata et al.9 TheAFM images presented in Figure 4 provide further evidenceof the inhibitive role of H2O as the ripening of the catalystnanoparticles is low in the case of H2/H2O.

Having established the Ostwald ripening phenomenonduring thermal annealing of catalyst films and the effect ofwater, we then decided to further substantiate that thisphenomenon can take place during SWNT carpet growth.We performed growth in the presence of 2 sccm C2H2 andH2 both with and without the presence of H2O. This identifieswhether ripening can take place after a catalyst particle hassuccessfully nucleated a SWNT. Figure 5 shows plan-viewTEM images of the catalyst layer exposed by both catalystoxidation and SWNT lift-off using an adhesive (a and b)and SWNT burnoff in air (c and d). Both sets of results forthe catalyst layer exposed by lift-off and burnoff yield thesame result: that Ostwald ripening occurs during carpetgrowth, and that the presence of H2O inhibits the ripeningeffect. For instance, for the catalyst layers exposed byburnoff, the number densities of particles in a 100 × 100nm area for carpets grown in H2/C2H2 and H2/H2O/C2H2 were48 and 113, respectively. It should be noted that particle sizesin the catalyst layer exposed by burnoff are larger due toripening that occurs during the burnoff process. Nonetheless,these images leave little ambiguity in the notion that theripening effect observed in Figures 2-4 also occurs duringcarpet growth. This emphasizes the role that water plays insuppressing the ripening of the catalysts during the growthprocess, and further bolsters the concept that Ostwaldripening may play a significant role in the eventual termina-tion of growth itself.

Figure 3. Plots of the catalyst particle diameter distributions afterthermal annealing in (a) H2 and in (b) H2/H2O.

Figure 4. AFM images of catalyst nanoparticles formed on the substrate after thermal annealing in (a) H2 and (b) H2/H2O for 30 s.

Nano Lett., Vol. 9, No. 1, 2009 47

On the basis of these results, we propose that the additionof water in super growth is in fact a means of inhibitingOstwald ripening through the ability of oxygen and hydroxylspecies to reduce diffusion or migration rates of catalystatoms from one catalyst to another across the samplesubstrate. As important, or perhaps more important than theresistance to coarsening, or increasing of catalyst size, is theability of water to stabilize the small catalyst particles andthus delay the collective termination of growth. Hannon etal.18 have demonstrated that Ostwald ripening during VLSgrowth of Si nanowires can lead to the termination of growthdue to the complete disappearance of the catalyst.

The schematic in Figure 6 illustrates the role water playsin inhibiting the ripening of the catalyst and how it isexpected to affect carpet growth. Our growth results indicatethat as the ripening of the catalyst particles is inhibited withthe addition of H2O, the lifetime of the active catalysts isdramatically enhanced. We hypothesize that during carpetgrowth, metal catalyst atoms desorb from the catalystparticles and diffuse across the substrate. Larger particlesare energetically favored over smaller particles due to the

lower fractional surface energy, and therefore absorb moreatoms than the smaller particles. The larger particles thengrow and the smaller particles shrink and disappear entirely.Ultimately, when a catalyst particle disappears, or when toomuch catalyst is lost, the nanotube growing from it stops

Figure 5. Plan-view TEM images of catalyst nanoparticles exposed by SWNT carpet removal by catalyst oxidation and lift-off (a and b),and SWNT burnoff in air (c and d); SWNT carpet growth was carried out in H2/C2H2 and H2/H2O/C2H2 for 30 s.

Figure 6. Schematic of Ostwald ripening of catalysts, and how itis expected to affect carpet growth with and without the additionof water vapor.

48 Nano Lett., Vol. 9, No. 1, 2009

growing. We propose that the reduction in the number ofparticles during heat treatment correlates with the numberof nanotubes that stop growing. For carpet growth, eachterminated nanotube imparts a mechanical drag force onadjacent growing nanotubes due to van der Waals forces andinterlocking. Therefore, it is possible that when enoughnanotubes stop growing, the carpet growth collectivelyterminates. Water vapor or hydroxyl groups impede diffusionby temporarily complexing with the catalyst atoms. Thisreduces the ripening rate, thus extending the life of thecatalysts and resulting in taller carpets.

In summary, the TEM and AFM results support ourhypothesis that a major role of H2O during super growth isthe inhibition of the Ostwald ripening behavior of thecatalyst. Our work demonstrates that water vapor impedesOstwald ripening of catalysts, thus extending their activelifetime. We note that the results of our work are only basedupon the observation of catalyst ripening and do not allowus to rule out any of the other mechanisms proposed so farto explain super growth. It is likely that the amorphouscarbon etching phenomena proposed by Hata et al.9 is acomplimentary effect to the inhibition of Ostwald ripeningin water-assisted growth. However, the ability to retard theeffect of Ostwald ripening to enhance catalyst lifetime inSWNT carpet growth can be utilized as a powerful tool toimprove the growth, yield, and purity of carbon nanotubesynthesis. In addition, the ability to negate effects that resultin eventual SWNT growth termination, such as Ostwaldripening, can provide a direction for future rational catalystdesign in order to achieve the most highly efficient growthpossible.

Acknowledgment. This work was funded by the Air ForceOffice of Scientific Research (AFOSR). S.M.K. and E.A.S.gratefully acknowledge support from the Army ResearchOffice (ARO).

References(1) (a) Chattopadhyay, D.; Galeska, I.; Papadimitrakopoulos, F. J. Am.

Chem. Soc. 2001, 123, 9451. (b) Jo, S. H.; Tu, Y.; Huang, Z. P.; et al.Appl. Phys. Lett. 2003, 82, 3520. (c) Yildirim, T.; Ciraci, S. Phys.ReV. Lett. 2005, 94, 175501. (d) Hinds, B. J.; Chopra, N.; Rantell, T.;Andrews, R.; Gavalas, V.; Bachas, L. G. Science 2004, 303, 62.

(2) Murakami, Y.; Chiashi, S.; Miyauchi, Y.; Hu, M.; Ogura, M.; Okubo,T.; Maruyama, S. Chem. Phys. Lett. 2004, 385, 298.

(3) (a) Puretzky, A. A.; Geohegan, D. B.; Jesse, S.; Ivanov, I. N.; Eres,G. Appl. Phys. A 2005, 81, 223. (b) Wood, R. F.; Pannala, S.; Wells,J. C.; Puretzky, A. A.; Geohegan, D. B. Phys. ReV. B 2007, 75, 235446.

(4) Bower, C.; Zhou, O.; Zhu, W.; Werder, D. J.; Jim, S. Appl. Phys.Lett. 2000, 77, 2767.

(5) Yoshihara, N.; Ago, H.; Tsuji, M. J. Phys. Chem. C 2007, 111, 11577.(6) De los Arcos, T.; Vonau, F.; Garnier, M. G.; Guggenheim, R. Appl.

Phys. Lett. 2002, 80, 2383.(7) Harutyunyan, A. R.; Awasthi, N.; Jiang, A.; Setyawan, W.; Mora, E.;

Tokune, T.; Bolton, K.; Curtarolo, S. Phys. ReV. Lett. 2008, 100,195502.

(8) Han, J.-H.; Graff, R. A.; Welch, B.; Marsh, C. P.; Franks, R.; Strano,M. S. ACS Nano 2008, 2, 53.

(9) (a) Hata, K.; Futaba, D. N.; Mizuno, K.; Namai, T.; Yumura, M.;Iijima, S. Science 2004, 306, 1362. (b) Futaba, D. N.; Hata, K.;Yamada, T.; Mizuno, K.; Yumura, M.; Iijima, S. Phys. ReV. Lett. 2005,95, 056104.

(10) Zhang, G.; Mann, D.; Zhang, L.; Javey, A.; Li, Y.; Yenilmez, E.;Wang, Q.; McVittie, J. P.; Nishi, Y.; Gibbons, J.; Dai, H. Proc. Natl.Acad. Sci. U.S.A. 2005, 102, 16141.

(11) Futaba, D. N.; Hata, K.; Namai, T.; Yamada, T.; Mizuno, K.;Hayamizu, Y.; Yumura, M.; Iijima, S. J. Phys. Chem. B 2006, 110,8035.

(12) Pisana, S.; Cantoro, M.; Parvez, A.; Hofmann, S.; Ferrari, A. C.;Robertson, J. Phys. E 2007, 37, 1.

(13) Sohn, J. I.; Choi, C.-J.; Lee, S.; Seong, T.-Y. Jpn. J. Appl. Phys. 2002,41, 4731.

(14) Pint, C.; Pheasant, S.; Nicholas, N.; Horton, C.; Hauge, R. J. Nanosci.Nanotech. 2008, 8, 6158.

(15) Pint, C. L.; Xu, Y.; Pasquali, M.; Hauge, R. H. ACS Nano 2008, 2,1871.

(16) Pint, C. L.; Pheasant, S. T.; Pasquali, M.; Coulter, K. E.; Schmidt,H. K.; Hauge, R. H. Nano Lett. 2008, 8, 1879.

(17) Pint, C. L.; Nicholas, N.; Pheasant, S. T.; Duque, J. G.; Parra-Vasquez,A. N. G.; Eres, G.; Pasquali, M.; Hauge, R. H. J. Phys. Chem. C 2008,112, 14041.

(18) Hannon, J. B.; Kodambaka, S.; Ross, F. M.; Tromp, R. M. Nature2006, 440, 69.

NL801876H

Nano Lett., Vol. 9, No. 1, 2009 49