Artificial neural network for predictive synthesis of single-walled carbon nanotubes by aerosol CVD method

Abstract

We propose to use artificial neural networks to process the experimental data and to predict the performance of the aerosol CVD synthesis of single-walled carbon nanotubes based on Boudouard reaction. We employ five key input parameters of the growth (pressures of CO, CO2 and ferrocene as well as residence time and temperature) to control the performance of produced nanotube films (yield, mean and standard deviation of diameter distribution, and defectiveness). The prediction errors were found to be comparable with the corresponding experimental errors. We believe the proposed approach is of great interest for the synthesis of nanocarbons with tailored characteristics.

High-performance transparent conducting films of long single-walled carbon nanotubes synthesized from toluene alone

“…Hence, it is inadvisable to claim that nanotube bundle diameter, nanotube quality, diameter or length alone contribute most to the conductivity (or sheet resistance) of a SWCNT film. A systematic experimental investigation of growth parameters assisted by big data analysis, like machine learning [38,39], might effectively locate truly optimal conditions for producing SWCNTs for the fabrication of TCFs with an optoelectronic performance approaching or even outperforming that of ITO film.…”

Applications of Carbon Nanotubes in the Internet of Things Era

“…Furthermore, the chemical reaction processes could be calculated with machine learning [ 108 , 109 ]. Two mainstreaming algorithms, i.e., support vector regression [ 110 ] and artificial neural networks [ 111 ], are being developed for optimizing the chemical processes, including the catalysis [ 112 ] and carbon nanotube growth [ 113 ].…”

“…Secondly, an artificial neural network algorithm-based machine learning has been utilized for the guide of experimental parameters over the synesthetic CNT quality [ 111 ]. Indeed, five synthesis parameters, e.g., the pressure of feedstock, types of feedstocks, substrate temperature, and synthesis time, have been chosen as input for the machine learning.…”

“…3 The upcoming machine learning algorithms for obtaining the properties, quality, and growth rate of carbon nanotube synthesis as well as the target of wafer-scale carbon nanotube synthesis Furthermore, the chemical reaction processes could be calculated with machine learning [108,109]. Two mainstreaming algorithms, i.e., support vector regression [110] and artificial neural networks [111], are being developed for optimizing the chemical processes, including the catalysis [112] and carbon nanotube growth [113].…”

Red GaPAs/GaP Nanowire-Based Flexible Light-Emitting Diodes

“…Transparent flexible contact pads to the top n-doped segments of NWs (Figure 4a) were based on SWCNT films [40,41], the size of square contact pads was about 1 mm 2 . The SWCNTs were synthesized by the aerosol (floating catalyst) CVD method described in detail elsewhere ( [42,43]) and downstream of the reactor in the form of randomly oriented nanotube film on microporous nitrocellulose filters. After the top contact application, the PDMS/NW membranes were mechanically peeled from Si substrate with the help of a razor blade.…”