Unveiling the Evolutions of Nanotube Diameter Distribution during the

Mar 12, 2017 - *E-mail: [email protected]. Cite this:ACS ... Digital Isotope Coding to Trace the Growth Process of Individual Single-Wa...
1 downloads 0 Views 10MB Size
Unveiling the Evolutions of Nanotube Diameter Distribution during the Growth of Single-Walled Carbon Nanotubes Hugo Navas,† Matthieu Picher,† Amandine Andrieux-Ledier,‡ Frédéric Fossard,‡ Thierry Michel,† Akinari Kozawa,§ Takahiro Maruyama,§ Eric Anglaret,† Annick Loiseau,‡ and Vincent Jourdain*,† †

Laboratoire Charles Coulomb, CNRS, Univ. Montpellier, 34095 Montpellier, France Laboratoire d’étude des microstructures, CNRS-ONERA, 92322 Châtillon, France § Department of Applied Chemistry, Meijo University, 468-8502 Nagoya, Japan ‡

S Supporting Information *

ABSTRACT: In situ and ex situ Raman measurements were used to study the dynamics of the populations of singlewalled carbon nanotubes (SWCNTs) during their catalytic growth by chemical vapor deposition. Our study reveals that the nanotube diameter distribution strongly evolves during SWCNT growth but in dissimilar ways depending on the growth conditions. We notably show that high selectivity can be obtained using short or moderate growth times. High-resolution transmission electron microscopy observations support that Ostwald ripening is the key process driving these seemingly contradictory results by regulating the size distribution and lifetime of the active catalyst particles. Ostwald ripening appears as the main termination mechanism for the smallest diameter tubes, whereas carbon poisoning dominates for the largest ones. By unveiling the key concept of dynamic competition between nanotube growth and catalyst ripening, we show that time can be used as an active parameter to control the growth selectivity of carbon nanotubes and other 1D systems. KEYWORDS: carbon nanotubes, diameter, selectivity, growth, CVD, Raman, growth time

A

distribution of SWCNT samples is still missing. Several studies have been devoted at understanding the mechanisms controlling the diameter distribution during the CCVD growth of SWCNTs.8 Using iron nanoclusters with average diameters of 3, 9, and 13 nm, Lieber et al. grew CNTs with average diameters of 3, 7, and 12 nm, respectively,9 thus highlighting a strong correlation between the catalyst size and the CNT diameter. High-resolution transmission electron microscopy (HRTEM) observations by Dai et al. confirmed the nanoparticle/nanotube size correlation in the case of SWCNTs with 1−3 nm diameters.10 These results supported that controlling the nanotube diameter distribution is first a matter of controlling the size distribution of catalyst nanoparticles at the stages of preparation, pretreatment, and growth.11−13 However, secondary influences were also shown to impact the size relationship between the catalyst particle and the

n amazing feature of single-walled carbon nanotubes (SWCNTs) is the significant variation of optical, electrical, or chemical properties induced by a small change of structure.1 A typical example is the band gap of a semiconducting SWCNT, which scales inversely with the nanotube diameter. Controlling the structure, and notably the diameter distribution, of SWCNTs during their growth therefore constitutes a prerequisite for many envisioned applications of these one-dimensional nanostructures. Catalytic chemical vapor deposition (CCVD) is currently the most popular method for synthesizing SWCNTs because it offers both more defined and more versatile growth conditions (catalyst, substrate, temperature, carbon feedstock, and gaseous additives), thus allowing the growth of SWCNTs with a better control of their orientation, surface density, purity, defect density, and structure.2 A still major drawback is that CCVD usually produces SWCNT samples with a large distribution of diameters and chiral angles. Although many groups reported catalyst systems and growth conditions yielding high selectivity for specific diameters and/or chiral angles,3−7 a profound understanding of the processes impacting the structural © 2017 American Chemical Society

Received: January 5, 2017 Accepted: March 11, 2017 Published: March 12, 2017 3081

DOI: 10.1021/acsnano.7b00077 ACS Nano 2017, 11, 3081−3088

Article

www.acsnano.org

Article

ACS Nano

Figure 1. (a) Sketch of the experimental approach based on in situ Raman spectroscopy. (b) SWCNT growth kinetics measured in different growth conditions by monitoring the G band area. (c) RBM spectra of SWCNT samples obtained by stopping the growth (T = 575 °C, P = 20 Pa EtOH, Ni/SiO2) at the specific states of progress indicated by red arrows in (b). The laser wavelength is 532 nm.

Figure 2. Time evolution of the Raman spectra (λ = 532 nm) for Ni/SiO2 in various growth conditions. (a) Type I: toward smaller diameters (575 °C, 59 Pa EtOH). (b) Type II: toward smaller then larger diameters (575 °C, 533 Pa EtOH). (c) Type III: toward larger then smaller diameters (525 °C, 59 Pa EtOH). (d) Type IV: negligible evolution with time (800 °C, 59 Pa EtOH). Beside each spectrum, the corresponding time taken after injection of the carbon precursor is indicated. Insets show the evolution of the diameter indicator (see main text) as a function of the growth duration.

nanotube. For instance, Liu et al. reported that a higher carbon feeding rate leads to the growth of larger diameter tubes, which they interpreted by assuming an optimal carbon feeding rate for each particle size, with smaller particles being poisoned and larger ones being underfed.5 In contrast, Chen et al. reported a

decrease of the mean diameter when increasing the pressure of the carbon precursor,14 suggesting a more complex dependence on the carbon precursor supply. Picher and Navas et al. actually showed that the growth of small-diameter SWCNTs is restricted to a narrow window of precursor pressure− 3082

DOI: 10.1021/acsnano.7b00077 ACS Nano 2017, 11, 3081−3088

Article

ACS Nano

growth conditions, one can validate or invalidate assumptions based on previous observations. The most striking result is that, in most growth conditions, the RBM profile strongly evolves with the growth time (see Figures 1c, 2, and S2 for Ni and Figure S1 for Pt). Importantly, this cannot be accounted for by only a difference of growth rates between different types of nanotubes, which would not modify the RBM intensity ratio, but requires a difference in the nucleation and/or termination times. Significant evolutions of the 2D (G′) band profiles are also observed and, to a lesser extent, evolutions in the G band profile. Four different types of evolutions were observed as a function of the growth conditions (Figure 2). Type I corresponds to an increase of the proportion of high-frequency RBMs and of low-frequency 2D components (Figures 2a and S2a). The RBM frequency is known to be inversely proportional to the SWCNT diameter, whereas the 2D frequency was shown to increase with the mean diameter of the SWCNT sample.23 Both features are therefore consistent with an evolution of the nanotube population toward smaller diameters. Type II also displays evolution toward smaller diameters at short times, followed by an evolution toward larger diameters at longer times (Figures 2b and S2b). An opposite trend is observed for type III with an initial evolution toward larger diameters followed by an evolution toward smaller diameters (Figures 2c and S2c). Finally, type IV corresponds to large diameters displaying essentially no evolution of Raman profile with time (Figure 2d). To check whether these evolutions may alternatively be interpreted as an evolution of the metallic/semiconducting ratio, we compared the results at 532 nm with measurements at other wavelengths (488, 647, and 785 nm); as shown in Figure S2, similar trends are obtained, although less marked for the 2D band, which may be attributed to the effect of air exposure. The conclusion that the evolution of the RBM profiles mostly reflects an evolution of the diameter distribution rather than of the metallic/ semiconducting ratio is additionally supported by the simultaneous onset of small metallic and semiconducting tubes (see high-frequency RBMs in Figure 1 for Ni and in Figure S1 for Pt). We highlight that Raman data collected at several wavelengths are perfectly appropriate to evidence evolutions of the SWCNT diameter distribution but not to quantify them because the effective Raman cross sections of each (n,m) tube are not known. The type of evolution is strongly dependent on the growth conditions, as summarized in Figure 3. Type I is essentially observed at intermediate temperature and precursor pressure, which also corresponds to the conditions leading to the highest nanotube yields and to the highest proportions of small diameters. This latter point is illustrated in Figure 3 by a diameter indicator corresponding to the barycenter of the RBM profile measured at 532 nm and converted in diameter units using a standard RBM law24 (this relative indicator should not be confused with the absolute mean diameter of the SWCNT sample, which cannot be easily determined from Raman spectroscopy due to the differences of Raman cross sections and of experimental resonance conditions between tubes of different types). Type II is observed at precursor pressures higher than those of type I and also at higher temperatures, although less markedly. Type III is observed at temperatures lower than that of type I and, less markedly, at lower precursor pressures. Finally, type IV is observed at the highest growth temperatures, where mostly large-diameter tubes are observed to grow. The influence of the growth conditions on the final

temperature due to the deactivation by catalyst coarsening at high temperature and by carbon poisoning at high precursor pressure.4,15 Studying ultralong individual SWCNTs, Yao et al. showed a reversible decrease/increase of diameter along a given tube (e.g., by up to 0.4 nm for an initial tube diameter of 1.6 nm) when increasing/decreasing the growth temperature by a few tens of °C.16,17 This last result provides strong evidence that, at a constant amount of catalyst, the nanotube diameter can be reversibly modulated by the growth conditions, although the underlying mechanism remains debated. For instance, Fiawoo et al. accounted for the observations of SWCNTs connected to much larger catalyst particles by a “perpendicular growth” mode18 favored at high carbon supply as later supported by the experimental studies of He et al.19 Surprisingly, the influence of the synthesis time on the structural distribution of SWCNTs has been overlooked in the vast majority of works, with a few rare exceptions.20,21 For instance, Kato and Hatakeyama reported that, by using short growth durations of a few seconds, SWCNTs with a narrow chirality distribution (i.e., a majority of (7,6) and (8,4) tubes) could be obtained.20 Of course, a major drawback of this approach was the very short length (