Precise Catalyst Production for Carbon Nanotube Synthesis with Targeted Structure Enrichment

Abstract

The direct growth of single-walled carbon nanotubes (SWCNTs) with a narrow distribution of diameter or chirality remains elusive despite significant benefits in properties and applications. Nanoparticle catalysts are vital for SWCNT synthesis, but how to precisely manipulate their chemistry, size, concentration, and deposition remains difficult, especially within a continuous production process from the gas phase. Here, we demonstrate the preparation of W₆Co₇ alloyed nanoparticle catalysts with precisely tunable stoichiometry using electrospray, which remain solid state during SWCNT growth. We also demonstrate continuous production of liquid iron nanoparticles with in-line size selection. With the precise size manipulation of catalysts in the range of 1–5 nm, and a nearly monodisperse distribution (σ_g < 1.2), an excellent size selection of SWCNTs can be achieved. All of the presented techniques show great potential to facilitate the realization of single-chirality SWCNTs production.

1. Introduction

The rational design and synthesis of nano-catalysts with precisely controlled structures, morphologies, and chemical compositions strongly impact the activity and selectivity of catalysts in applications ranging from the growth of single-walled carbon nanotubes (SWCNTs) to clean energy electrochemical reactions [1] such as the oxygen reduction reaction (ORR), oxygen evolution reaction (OER), or hydrogen evolution reaction (HER). Nevertheless, it is often unachievable to produce nano-catalysts with a targeted diameter, narrow size distribution, and specific stoichiometry, particularly if the diameter of the particles is smaller than 5 nm or if the particles should contain an alloyed chemistry. These factors especially hinder the growth of SWCNTs with specific chiralities, which is needed for applications such as next-generation electronics and optoelectronics [2,3].

For the chemical vapor deposition (CVD) of SWCNTs, alloyed catalysts such as W₆Co₇ [4,5], Co_xMo_y [6] or Co_xNi_y [7] introduced new hope for the in situ enrichment of certain SWCNT chiralities. However, among the reported methods, synthesis of key alloyed catalysts is still restricted by the random nucleation from the colloidal solution [6] and aerosol [7], or use of specific organic molecules [4]. Precisely manipulating the chemistry of every single tiny alloyed nanoparticle (<5 nm) then becomes critical but difficult, because of the common stoichiometry deviation between catalysts during catalyst nucleation and growth, which can affect the crystalline structure and the uniformity of catalyst performance.

On the other hand, because of the diameter correlation between the SWCNT and catalyst, a chemistry-independent method to precisely control catalyst sizes smaller than 2 nm is a long-sought goal. Recently, our group showed the first study to realize the continuous synthesis of solid catalysts (W-, Mo-, and Re-based) with a size smaller than 2 nm [8]. These solid catalysts remain in a solid state in the high-temperature CNT growth environment. However, although regarded more efficient on catalyzing CNT growth with special enhanced enrichment of near-armchair chiralities, the corresponding continuous synthesis of liquid catalysts (Fe-, Ni-, and Co-based) also with precise size manipulation is still missing. Additionally, the size-dependent influence on SWCNTs also needs to be confirmed.

Here, we present the first study on the use of electrospray techniques to continuously produce W₆Co₇ alloyed nanoparticles with precisely manipulated stoichiometry and size. On the other hand, we also present a continuous production method for typical liquid catalysts—Fe-based nanoparticles—using straightforward gas-phase synthesis and in-line size selection techniques. This allows for a precisely manipulated size to any target value in the region of 1–8 nm and a narrow size distribution. The constraint effect on the diameter of SWCNTs (d_t) can be clearly seen. These new advances are important to develop future continuous synthesis methods for chirality-controlled SWCNTs.

2. Results

2.1. Production of Alloyed Nano-Catalysts and Their Growth of CNTs

2.1.1. Nanoparticle Production from Electrospray Method with Precise Stoichiometry

With W-Co alloyed catalysts as an example, synthesis begins with the generation of gas-phase W-Co compound salt nanoparticles (sNPs) suspended in a carrier gas, as shown in Figure 1. With the gas-phase transport, sNPs pass through size selection and controlled deposition. The W-Co alloyed catalyst (refractory-ferromagnet metal alloyed catalyst) is a typical solid catalyst to support the growth of SWCNTs, which remain rigid and crystalline in a CNT growth environment, believed to strongly affect the chirality enrichment in the final product [4,6].

As shown in Figure 1a, the mixed solution of W- and Co-based compounds passes through a capillary tube driven by the high gas pressure. Under optimized parameters, the solution cone formed at the tip of the capillary facilitates the jet of nano-droplets driven by the high voltage difference between the cone and the orifice plate. The jetted highly charged nano-droplets are timely neutralized and self-dried in the carrier gas, forming the sNPs.

The size distributions of sNPs are first characterized by a scanning mobility particle spectrometer (SMPS) (Figure 1b). Owing to the nanoscale of the droplet jet, the nanoparticles (sNPs) formed after solvent removal are well constrained to < 8 nm with a mean diameter (d_mean) ~ 3.96 nm, which is much smaller than sNPs formed by a common atomizer (>10 nm) [8,9].

After constraining the sNP size with cone-jetting, the size distributions of the sNPs are further narrowed by a differential mobility analyzer (DMA). In the following section of this work (for liquid catalysts), the size selection principle of the DMA will be illustrated in detail. With the final controlled disposition onto various target substrates by an electrostatic precipitator, sNPs are evenly distributed on the collection zone with a tunable areal density (Figure 2a). The size distribution of selected NPs follows a log-normal distribution with a geometric standard deviation (σ_g) of 1.21 based on AFM statistics, much smaller than the that before DMA size selection (σ_g = 1.40). σ_g can be further lowered to 1.05 based on our previous research [8] as long as the aerosol concentration is high enough to guarantee efficient collection.

After calcination and reduction at high temperature in H₂, the sNPs are reconstructed into single-crystal alloyed NPs (alloyed-mNPs, Figure 2b–e). Based on our previous work, the diameter is observed to shrink by ~40% (60% remaining) during this process after losing O, N, and H and reconstruction [8]. Energy-dispersive X-ray spectroscopy (EDS) elements mapping verifies that both W and Co elements are uniformly distributed across the alloyed-mNPs. The atomic ratio of W/Co measured by EDS is (42.3 ± 7.1): (57.8 ± 8.3). Combined with the interplanar distances of mNPs measured from HRTEM, our particles’ characteristics are in good agreement with those of W₆Co₇ (PDF exp.2-1091). Figure 2b inset shows interplanar distances 0.216 and 0.228 nm of the (1 0 10) and (1 1 3) planes in the W₆Co₇ alloy, respectively. Moreover, no accumulation of oxygen is seen in the area of the NPs, verifying the alloyed metal state of the NPs. We suggest the alloyed-mNPs are W₆Co₇ alloyed single crystals (details of the crystal structure identification can be found in the Supplementary Material).

Importantly, with this electrospray method, each nanoparticle is formed from one nano-droplet, thus the stoichiometry of the sNPs and mNPs is precisely equal to the solution from which the nano-droplets are formed. Here, the atomic ratio of W/Co in the prepared mixed solution is 6:7, matching the stoichiometry of W₆Co₇ catalysts and the EDS results.

2.1.2. CNTs Grown from the Alloyed Nano-Catalysts

The NPs produced in this work are based on a continuous gas-phase process. In our previous work, we demonstrated its potential to be extended into a floating catalyst CVD process (FCCVD) [8]. However, for the sake of convenience, and a better control on the synthesis parameters of NPs and CNTs, all the CNTs in this work are grown using a batch process from substrates loaded by produced NPs.

Due to the gas-phase nature of the sNPs formation process, by modifying the receiving target substrates, CNTs can be facilely grown on various targets without any ex situ transfer processes. Figure 3 shows random CNTs grown on SiO₂/Si (Figure 3a) and an aligned array (Figure 3b) on the ST-cut quartz substrate because of the guidance from the atom steps of the substrate [10]. By decoupling the catalyst size from the areal density, further growth of high-density SWNT arrays for electronics becomes possible [11,12]. Moreover, it is worth noting that in SEM images of CNTs laying on an insulating substrate (SiO₂/Si) obtained with a low accelerating voltage (1–3 kV), the bright line in the images (Figure 3a,b) originates from not only the CNT itself, but also from the adjacent zone of substrate affected by the CNT [13]. Therefore, the diameter of the bright lines cannot represent that of CNTs.

With radial breathing mode (RBM) peaks from the resonant Raman spectroscopy of SWCNTs detected with 532, 638, and 785 nm lasers, the chiralities abundances are identified based on the Kataura plot and summarized in a graphene map (Figure 3c). To guarantee unambiguous identification, the color density is indicative of solely SWCNTs within 0.81–1.53 nm (160–295 cm⁻¹, marked by the blue dash-dotted line). Chiralities near (2n,n) most enriched (corresponding range of 19.1 ± 5°) are marked by a black dashed line, and few can be found near the zigzag region. The most enriched visible chiralities are (12,6), (10,5), (13,6), and (12,5), which are all on or near (2n,n), with the chiral angle near 19.1°. This result matches the solid catalysts performance we reported recently [8]. The highest abundance of (12,6) reached approximately 22% among the diameter range of 0.81–1.53 nm. The slightly lower abundance compared with a previous report is mainly because of the comparatively larger size of the W-Co alloyed catalyst with the electrospray method, which is a compromise to reach a higher catalyst production rate and collection efficiency. Currently, the yield of our electrospray method is around 10¹–10³ #/cm³/min, which set up many difficulties on characterization and size selection. However, significant advancements in electrospray have been made in recent years as a part of a large effort by the synthesis community [14]. The yield can also be improved in the future by using electrospray with multi cone jets, instead of a single one used here. Considering the unique advantage of the electrospray method on the manipulation of stoichiometry in the nano-catalyst, further work could effectively reduce the size of the alloyed catalyst and further constrain chiralities.

2.2. Production of Size-manipulated Fe-based Catalyst and the SWCNTs Diameter Constraint

2.2.1. Fe₂O₃ Nanoparticle Production with Precise Size Manipulation

For liquid iron catalysts, we follow our reported techniques [8] to produce metal oxide nanoparticles (oNPs) in the gas phase. Then, to grow SWCNTs, the Fe-oNPs are reduced into Fe metal nanoparticles (Fe-mNPs), forming the final catalyst upon the supply of the carbon feedstock. As shown in Figure 4, the ferrocene vapor taken by the carrier air gas enters the high-temperature furnace (700 °C), and is calcinated into Fe-oNPs. The dominant sizes of Fe oNPs are adjustable from 2 to 10 nm by varying the input of the Fe precursor (temperature of the ferrocene container and the flow rate of the carrier air gas).

Although small and highly concentrated, Fe-oNPs’ size is still too broadly distributed to effectively constrain the diameter of SWCNTs. Similar to the techniques above, proactive NP mobility selection is integrated into the mainstream to transform the distribution of the entire aerosol flow from polydisperse to monodisperse. In this process, after being charged by a radioactive ionizer, the oNPs are passed between two concentric cylinders in the DMA at an electrical potential difference. Only oNPs with the prescribed mobility-equivalent diameter pass downstream, which results in the narrow size distribution of the selected Fe-oNPs.

Using AFM, the size distribution of these selected Fe NPs is measured (Figure 5a–d). Their sizes are precisely tuned in the range of 1 to 8 nm. From relatively large Fe-oNPs (Figure 5a), it is obvious that the as-produced Fe-oNPs are nanoaggregates with irregular morphologies. After being processed in H2 at 400 °C, the nanoaggregates are reduced and reconstructed into rounded NPs (and inevitably re-oxidized in the air when characterized by AFM). Meanwhile, for all sizes, the size distribution after selection is inherited to that of NPs after reduction and reconstruction (Figure 5a–d, σ_g ~ 1.2, and σ_g = 1 represents ideal monodisperse particles). This implies that processing in H₂ at 400 °C for 5 min will not broaden the size distribution although the aggregation and Ostwald ripening effect are enhanced.

It is also worth noting that the absolute FWHM of these selected distributions scales with the selected midpoint size. Therefore, the selection of smaller-diameter particles also corresponds to a smaller FWHM (insets of Figure 5b–d, FWHM decreases from 3 to 0.5 nm when d_mean decreases from 7.6 to 1.4 nm). This is the key feature for achieving nearly monodisperse Fe catalysts for SWCNT growth.

2.2.2. Constrain SWCNTs Diameter by Fe Catalysts

In Figure 6, we compare the diameter distribution of SWCNTs grown from size-selected 1,8 (Figure 6a–i) and 3.6 nm (Figure 6j–l) Fe-oNPs catalysts. Resonant Raman spectroscopy shows the RBM peaks of SWCNTs in the laser spot excited with laser phonons [3,15]. We follow our reported method [8] on Raman mapping characterization and data analysis. In Figure 6a–c, we show the raw Raman spectra of 1.8 nm Fe-oNPs on the RBM region detected by 532, 638, and 785 nm lasers.

The peak position of each spectrum is identified after background removal (SiO₂/Si signal, Figure 6d–f). Based on the RBM peaks frequency (ω_RBM) to the diameter of SWCNTs (d_t) relationship for samples grown on SiO₂/Si substrates [16], ω_RBM = 235.9/d_t + 5.5, we obtain the diameter distribution statistics from the peak position abundance statistics results (Figure 6g–l).

Here, with Raman $x-y$ 2D mapping results from 532, 638, and 785 nm lasers, the resonant peak position abundance statistics results are summarized with a normalized scale so that peak counts can be compared between different lasers. However, as is well known, the scale of detectable SWCNTs in each Raman mapping result depends on both the diameter distribution of the sample and whether resonance happens with the laser wavelength, according to the Kataura plot [17,18], as shown in Figure 6m.

In our results, for abundance detected by the 638 nm laser (Figure 6h vs. Figure 6k), it is clear that 1.8 nm Fe-oNPs catalysts preferentially produce more 1.2–1.4 nm SWCNTs and fewer for d_t > 1.4 nm.

A similar trend can also be seen by the 532 nm laser (Figure 6g vs. Figure 6j), where SWCNTs with d_t > 1.4 nm are much suppressed with 1.8 nm Fe-oNPs catalysts, although 1.2 < d_t < 1.3 nm cannot be detected by the 532 nm laser because of a lack of resonance and the known bias of the chirality distribution to the armchair chiralities with liquid catalysts like Fe [19].

For the 785 nm laser results (Figure 6i vs. Figure 6l), the disappearance of 1.2–1.4 nm SWCNTs enrichment is also due to the poor resonance condition and the chirality distribution propensity for armchair chiralities. The suppressed abundance peak around 1.6 nm tends to verify the constraint effect from the size of the catalysts.

3. Discussion

To pursue single-chirality SWCNT growth, the mechanisms underlying the chirality selection of SWCNTs are of great importance. However, many theoretical contributions have been proposed, but always lack sound experimental verification. Among them, interface energy-driven nucleation thermodynamics [20,21,22] combined with CNT growth kinetics [20], as well as catalyst epitaxy with symmetry matching [4,23,24], are the most favorable directions.

Before the realization of chirality enrichment, normally specific diameter enrichment in the SWCNT product is the precondition. However, not all the grown SWCNTs possess diameters to match those of their catalysts, which is known as tangential growth (d_t ~ d_catalyst). Sometimes, the diameter of a grown SWCNT can be smaller than that of its catalyst (d_t < d_catalyst), and this is known as perpendicular growth.

With in-line size selection using a DMA, we can directly select small-diameter catalysts. Although CNT growth is not fixed to either perpendicular growth or tangential growth, small catalysts provide a limited upper bound to the diameter of CNTs regardless of the growth mode. Here, we verified, with 1.8 nm Fe-oNPs, that the diameter of SWCNTs can be enriched <1.4 nm. Consequently, the resultant chirality choices are significantly reduced. Further, combined with the declined chiral angle distribution, the chirality choices are again reduced.

The SWCNT diameter distribution results from 1.8 nm Fe-oNPs also imply that the size distribution of catalysts set by NP mobility size selection can still be inherited by SWCNTs, even though Fe is believed to remain in a liquid state during CNT growth and lead to more aggregation and Ostwald ripening than solid catalysts.

With the electrospray technique, owing to its chemistry-independent property, as long as mixed solutions of the corresponding soluble compounds are prepared, similar alloyed catalysts such as Mo-Fe, W-Fe, and Re-Ni can also be produced with precise manipulation of the stoichiometry.

Further, combined with proper processing, the formed single-crystal nano-catalysts with different chemistry and stoichiometry give researchers, for the first time, an ability to manipulate the interface between the CNT and catalyst, where the interfacial energy has been reported to determine the chiral preference of SWCNTs by affecting the possibility of nucleation, the growth speed, and even the growth time.

4. Conclusions

With the mechanism of one nanoparticle formed from one atomized droplet, we produce W₆Co₇ alloyed nanoparticles successfully with the same target chemistry in a continuous gas-phase method. They act as the solid catalyst to grow SWCNTs with chirality control. As an example, for liquid catalysts, we produce the gas-phase size-selected Fe-based nanoparticles with a size down to 1.5–1.8 nm. The small diameter of these liquid catalysts shows an obvious diameter constraint effect on SWCNTs, with the diameter highly enriched to <1.4 nm.

We assert that the two techniques shown here can be used to provide a useful set of experimental designs to support or rule out potential theoretical mechanisms, facilitating measuring the physical parameters, and finally forming a cohesive and useful theory that can guide the development of models, experiments, and the production of chirality-controlled CNTs.

5. Materials and Methods

5.1. Nanoparticle Synthesis, Size Selection, and Collection

Production of alloyed W₆Co₇-nanoparticles: An aqueous mixed metal salt solution was prepared with 25 μM (NH₄)₆H₂W₁₂O₄₀ (Sigma-Aldrich 463922), 350 μM Co(NO₃)₂·6H₂O (Sigma-Aldrich 203106), and 20 mM CH₃CO₂NH₄ (Sigma-Aldrich A1542, as a buffer agent which will vaporize before size selection). The mixed solution was atomized into nano-droplets suspended in clean air as a carrier gas through an electrospray aerosol generator (TSI Inc. 3480, Minnesota USA). For the electrospray process, a silica capillary with an inner diameter of 40 μm was used to obtain a cone-jetting mechanism with a proper flow rate of carrier gas (~2 slpm) and 2 kV high voltage. A radioactive charge neutralizer (TSI 3077) was used to discharge the nano-droplet to avoid splitting with the removal of water in the nano-droplet, leaving precipitated sNPs. The sNPs were then sent through a DMA (TSI 3085). The DMA selected particles with a prescribed mobility-equivalent diameter and a very narrow range. The DMA-selected nanoparticles were sent into an electrostatic precipitator for collection onto the target substrate.

Production of size-selected Fe-oNPs: Clean air (20 sccm) was set to pass through the container of ferrocene powder (Sigma-Aldrich F408) at room temperature. Another clean air (1 slpm) carried the vapor of ferrocene through an alumina tube in a furnace whose temperature was set to 700 °C, and the nucleated polydisperse Fe-oNPs were first charged using a radioactive charge neutralizer (TSI 3077). Production of alloyed W₆Co₇-nanoparticles: An aqueous mixed metal salt solution was prepared with 25 μM (NH₄)₆H₂W₁₂O₄₀ (Sigma-Aldrich 463922), 350 μM Co(NO₃)₂·6H₂O (Sigma-Aldrich 203106), and 20 mM CH₃CO₂NH₄ (Sigma-Aldrich A1542, as a buffer agent which will vaporize before size-selection). The mixed solution was atomized into nano-droplets suspended in clean air as a carrier gas through an electrospray aerosol generator (TSI Inc. 3480, Minnesota USA). For the electrospray process, a silica capillary with an inner diameter of 40 μm was used to obtain a cone-jetting mechanism with a proper flow rate of carrier gas (~2 slpm) and 2 kV high voltage. A radioactive charge neutralizer (TSI 3077) was used to discharge the nano-droplet to avoid splitting with the removal of water in the nano-droplet, leaving precipitated sNPs. The sNPs were then sent through a DMA (TSI 3085). The DMA selected particles with a prescribed mobility-equivalent diameter and a very narrow range. The DMA-selected nanoparticles were sent into an electrostatic precipitator for collection onto the target substrate.

To avoid or reduce possible aggregation, oNPs were first reduced in H₂ with the temperature slowly rising to 400 °C to obtain mNPs. Then, the temperature was raised quickly to 850 °C (100 K/min). Ethanol vapor as the carbon feedstock was introduced immediately when the final temperature was reached (850 °C) in an Ar carrier gas stream. The growth time was set to only 10 min. After growth, the carbon-rich environment was expelled by H₂, and the samples were cooled to room temperature.

5.2. Aerosol Characterization, AFM, HRTEM, EDS, and Raman Mapping

Aerosol size distributions of all NPs were analyzed using a scanning mobility particle spectrometer (SMPS, TSI 3938, Minnesota USA), which consists of a combination of a DMA (TSI 3085, Minnesota USA) and a condensation particle counter (CPC, TSI 3756, Minnesota USA). AFM was conducted on a Veeco Dimension Pro AFM in peak force mode. Both transmission electron microscopy (TEM, Thermo Scientific Talos) and energy-dispersive X-ray spectroscopy (EDS) elements mapping were conducted on a Thermo Scientific (FEI) Talos F200X G2 TEM operating at 200 kV. TEM images were acquired using a Ceta 16M CMOS camera. EDS was performed in scanning transmission electron microscopy (STEM) mode with images acquired using the high-angle annular dark field (HAADF) detector with EDX spectra/maps collected using the Super-X EDS detector system which consists of 4 windowless silicon drift detectors. Prior to TEM, the NPs were dispersed onto Si₃N₄ TEM grids. Raman mapping was conducted in the RBM range, with 532, 638, and 785 nm lasers (HORIBA XploRA PLUS). The step size is set to 3 μm in both the x and y directions.

6. Patents

The pertaining UK patent has been filed with the application number as 1917638.7., which partially results from the work reported in this manuscript.