Efficient Closed-loop Maximization of Carbon Nanotube Growth Rate using Bayesian Optimization.

A major technological challenge in materials research is the large and complex parameter space, which hinders experimental throughput and ultimately slows down development and implementation. In single-walled carbon nanotube (CNT) synthesis, for instance, the poor yield obtained from conventional catalysts is a result of limited understanding of input-to-output correlations. Autonomous closed-loop experimentation combined with advances in machine learning (ML) is uniquely suited for high-throughput research. Among the ML algorithms available, Bayesian optimization (BO) is especially apt for exploration and optimization within such high-dimensional and complex parameter space. BO is an adaptive sequential design algorithm for finding the global optimum of a black-box objective function with the fewest possible measurements. Here, we demonstrate a promising application of BO in CNT synthesis as an efficient and robust algorithm which can (1) improve the growth rate of CNT in the BO-planner experiments over the seed experiments up to a factor 8; (2) rapidly improve its predictive power (or learning); (3) Consistently achieve good performance regardless of the number or origin of seed experiments; (4) exploit a high-dimensional, complex parameter space, and (5) achieve the former 4 tasks in just over 100 hundred experiments (~8 experimental hours) - a factor of 5× faster than our previously reported results.

Carbon Nanotubes and Related Nanomaterials: Critical Advances and Challenges for Synthesis toward Mainstream Commercial Applications.

Advances in the synthesis and scalable manufacturing of single-walled carbon nanotubes (SWCNTs) remain critical to realizing many important commercial applications. Here we review recent breakthroughs in the synthesis of SWCNTs and highlight key ongoing research areas and challenges. A few key applications that capitalize on the properties of SWCNTs are also reviewed with respect to the recent synthesis breakthroughs and ways in which synthesis science can enable advances in these applications. While the primary focus of this review is on the science framework of SWCNT growth, we draw connections to mechanisms underlying the synthesis of other 1D and 2D materials such as boron nitride nanotubes and graphene.

Discovery of wall-selective carbon nanotube growth conditions via automated experimentation.

Applications of carbon nanotubes continue to advance, with substantial progress in nanotube electronics, conductive wires, and transparent conductors to name a few. However, wider application remains impeded by a lack of control over production of nanotubes with the desired purity, perfection, chirality, and number of walls. This is partly due to the fact that growth experiments are time-consuming, taking about 1 day per run, thus making it challenging to adequately explore the many parameters involved in growth. We endeavored to speed up the research process by automating CVD growth experimentation. The adaptive rapid experimentation and in situ spectroscopy CVD system described in this contribution conducts over 100 experiments in a single day, with automated control and in situ Raman characterization. Linear regression modeling was used to map regions of selectivity toward single-wall and multiwall carbon nanotube growth in the complex parameter space of the water-assisted CVD synthesis. This development of the automated rapid serial experimentation is a significant progress toward an autonomous closed-loop learning system: a Robot Scientist.

Transition from direct to Fowler-Nordheim tunneling in chemically reduced graphene oxide film.

We investigate charge transport in a chemically reduced graphene oxide (RGO) film of sub-micron thickness. The I-V curve of RGO film shows current switching of the order of ∼10(5) above the threshold voltage. We found that the observed I-V curve is consistent with quantum tunnelling based charge transport. The quantum tunnelling based Simmons generalized theory was used to interpret the charge transport mechanism which shows that the current switching phenomenon is associated with transition from direct to Fowler-Nordheim (F-N) tunneling. The absence of current switching in the I-V curve after stripping away the oxygen functional groups from chemically RGO film confirms that the presence of these groups and reduced interaction between adjacent layers of RGO play a key role in charge transport. Such metal-based current switching devices may find applications in graphene-based electronic devices such as high voltage resistive switching devices.

Transparent stretchable single-walled carbon nanotube-polymer composite films with near-infrared fluorescence.

We report transparent stretchable single-walled carbon nanotube-polymer composite films that emit pronounced Raman and near-infrared fluorescence with a fine spatial resolution. The independent modulation in Raman and fluorescence spectra is demonstrated in response to touch and temperature. The optical signal transduction of transparent stretchable optoelectronic films may enable a paradigm shift in touch-sensing devices eliminating electrical interconnects.

MesoBioNano Explorer--a universal program for multiscale computer simulations of complex molecular structure and dynamics.

We present a multipurpose computer code MesoBioNano Explorer (MBN Explorer). The package allows to model molecular systems of varied level of complexity. In particular, MBN Explorer is suited to compute system's energy, to optimize molecular structure as well as to consider the molecular and random walk dynamics. MBN Explorer allows to use a broad variety of interatomic potentials, to model different molecular systems, such as atomic clusters, fullerenes, nanotubes, polypeptides, proteins, DNA, composite systems, nanofractals, and so on. A distinct feature of the program, which makes it significantly different from the existing codes, is its universality and applicability to the description of a broad range of problems involving different molecular systems. Most of the existing codes are developed for particular classes of molecular systems and do not permit multiscale approach while MBN Explorer goes beyond these drawbacks. On demand, MBN Explorer allows to group particles in the system into rigid fragments, thereby significantly reducing the number of dynamical degrees of freedom. Despite the universality, the computational efficiency of MBN Explorer is comparable (and in some cases even higher) than the computational efficiency of other software packages, making MBN Explorer a possible alternative to the available codes.

Enhancement of the electron spin resonance of single-walled carbon nanotubes by oxygen removal.

We have observed a nearly 4-fold increase in the electron spin resonance (ESR) signal from an ensemble of single-walled carbon nanotubes (SWCNTs) due to oxygen desorption. By performing temperature-dependent ESR spectroscopy both before and after thermal annealing, we found that the ESR in SWCNTs can be reversibly altered via the molecular oxygen content in the samples. Independent of the presence of adsorbed oxygen, a Curie law (spin susceptibility ∝ 1/T) is seen from ~4 to 300 K, indicating that the probed spins are finite-level species. For both the pre-annealed and post-annealed sample conditions, the ESR line width decreased as the temperature was increased, a phenomenon we identify as motional narrowing. From the temperature dependence of the line width, we extracted an estimate of the intertube hopping energy; for both sample conditions, we found this hopping energy to be ~1.2 meV. Since the spin hopping energy changes only slightly when oxygen is desorbed, we conclude that only the spin susceptibility, not spin transport, is affected by the presence of physisorbed molecular oxygen in SWCNT ensembles. Surprisingly, no line width change is observed when the amount of oxygen in the SWCNT sample is altered, contrary to other carbonaceous systems and certain 1D conducting polymers. We hypothesize that physisorbed molecular oxygen acts as an acceptor (p-type), compensating the donor-like (n-type) defects that are responsible for the ESR signal in bulk SWCNTs.

Unique origin of colors of armchair carbon nanotubes.

The colors of suspended metallic colloidal particles are determined by their size-dependent plasma resonance, while those of semiconducting colloidal particles are determined by their size-dependent band gap. Here, we present a novel case for armchair carbon nanotubes, suspended in aqueous medium, for which the color depends on their size-dependent excitonic resonance, even though the individual particles are metallic. We observe distinct colors of a series of armchair-enriched nanotube suspensions, highlighting the unique coloration mechanism of these one-dimensional metals.

Effect of vaporization temperature on the diameter and chiral angle distributions of single-walled carbon nanotubes.

Pulsed laser vaporization synthesis of single-wall carbon nanotubes on Co/Ni and Rh/Pd catalysts was explored with respect to variations in the production temperature. The nanotube type populations were determined via photoluminescence, UV-Vis-NIR absorption and Raman spectroscopy. It was found that lowered production temperature leads to smaller nanotube diameters and exceptionally narrow (n, m) type distributions, with marked preference towards large chiral angles for both catalysts. Interestingly, larger nanotube diameters tend to be associated with larger chiral angles. These results demonstrate that PLV production technique can provide at least partial control over the nanotube (n, m) populations. In addition, these results have implications for the understanding the nanotube nucleation mechanism in the laser oven. SWCNT synthesized at lower temperatures appear quite attractive as a starting material for nanotube type separation experiments.

Quantifying the semiconducting fraction in single-walled carbon nanotube samples through comparative atomic force and photoluminescence microscopies.

A new method was used to measure the fraction of semiconducting nanotubes in various as-grown or processed single-walled carbon nanotube (SWCNT) samples. SWCNT number densities were compared in images from near-IR photoluminescence (semiconducting species) and AFM (all species) to compute the semiconducting fraction. The results show large variations among growth methods and effective sorting by density gradient ultracentrifugation. This counting-based method provides important information about SWCNT sample compositions that can guide controlled growth methods and help calibrate bulk characterization techniques.

Strain measurements on individual single-walled carbon nanotubes in a polymer host: structure-dependent spectral shifts and load transfer.

The fluorescence spectra of individual semiconducting single-walled carbon nanotubes embedded in polymer films were measured during the application of controlled stretching and compressive strains. Nanotube band gaps were found to shift in systematic patterns that depend on the (n,m) structural type and are in excellent agreement with the predictions of theoretical models. Loss of nanotube-host adhesion was revealed by abrupt irregularities in plots of spectral shift vs strain.

A parametric study of single-wall carbon nanotube growth by laser ablation.

Results of a parametric study of carbon nanotube production by the double-pulse laser oven process are presented. The effect of various operating parameters on the production of single-wall carbon nanotubes (SWCNTs) is estimated by characterizing the nanotube material using analytical techniques, including scanning electron microscopy, transmission electron microscopy, thermo gravimetric analysis and Raman spectroscopy. The study included changing the sequence of the laser pulses, laser energy, pulse separation, type of buffer gas used, operating pressure, flow rate, inner tube diameter, as well as its material, and oven temperature. It was found that the material quality and quantity improve with deviation from normal operation parameters such as laser energy density higher than 1.5 J/cm2, pressure lower than 67 kPa, and flow rates higher than 100 sccm. Use of helium produced mainly small diameter tubes and a lower yield. The diameter of SWCNTs decreases with decreasing oven temperature and lower flow rates.

Gas-phase production of single-walled carbon nanotubes from carbon monoxide: a review of the hipco process.

The latest process for producing large quantities of single-walled carbon nanotubes (SWNTs) to emerge from the Rice University, dubbed HiPco, is living up to its promise. The current production rates approach 450 mg/h (or 10 g/day), and nanotubes typically have no more than 7 mol % of iron impurities. Second-generation HiPco apparatus can run continuously for 7-10 days at a time. In the HiPco process nanotubes grow in high-pressure, high-temperature flowing CO on catalytic clusters of iron. Catalyst is formed in situ by thermal decomposition of iron pentacarbonyl, which is delivered intact within a cold CO flow and then rapidly mixed with hot CO in the reaction zone. Upon heating, the Fe(CO)5 decomposes into atoms that condense into larger clusters. SWNTs nucleate and grow on these particles in the gas phase via CO disproportionation: CO + CO --> CO2 + C (SWNT), catalyzed by the Fe surface. The concentration of CO2 produced in this reaction is equal to that of carbon and can therefore serve as a useful real-time feedback parameter. It was used to study and optimize SWNT production as a function of temperature, pressure, and Fe(CO)5 concentration. The results of the parametric study are in agreement with current understanding of the nanotube formation mechanism.