(272d) Machine Learning Algorithms for High-Throughtput Chemical Sensing Using Liquid-Crystals

Zavala, V. M., University of Wisconsin-Madison
Cao, Y., University of Wisconsin-Madison
Yu, H., University of Wisconsin-Madison
Abbott, N. L., University of Wisconsin-Madison
Security, health, and environmental monitoring requires the development of chemical sensing technologies that can be used in-situ and with limited equipment and human intervention. The impact of such technologies when deployed at a large scale can be significant; for instance, the U.S. Department of Energy Savannah River laboratory analyzes over 40,000 groundwater samples per year at a cost of $1,000 per sample ($40 million per year) [1]. Liquid crystals (LCs) are fluid phases with preferred molecular orientations, a so-called director, that undergo surface-driven ordering transitions in the presence of chemical species such as organophosphonates [2, 3, 4, 5, 6], chlorine, ammonia, and hydrogen sulfide [7]. The optical characteristics (features) of the LC transitions can be tailored and exploited to design chemical sensors. For instance, LCs can be designed to assume homeotropic (perpendicular) orientations on surfaces decorated with metal salts [2, 4, 8]. Chemical species that diffuse into the LCs and bind more strongly to the metal cations than the LC functional groups will trigger a transition of the LCs orientation from homeotropic to planar (see Figure 1) [9, 10, 11]. It is possible to manipulate the selectivity and response characteristics (e.g., dynamics) of the LC by tuning the binding energies of the LC functional groups (e.g., nitrile and pyridine groups) to the surface (e.g., Fe+3, La3+) [9, 10]. For instance, ordering transitions of LC sensors fabricated using a nematic LC called 4-cyano- 4’-penthylbiphenil (5CB) and surfaces presenting aluminum perchlorate salts have been studied in [12, 13, 14].

In this work, we present a machine learning (ML) framework to optimize the specificity and speed of LC-based chemical sensors. Specifically, we demonstrate that ML techniques can uncover valuable feature information from surface-driven LC orientational transitions triggered by the presence of different gas-phase analytes (and the corresponding optical responses) and can exploit such feature information to train accurate and automatic classifiers. We demonstrate the utility of the framework by designing an experimental LC system that exhibits similar optical responses to a stream of nitrogen containing either 10 ppmv dimethyl-methylphosphonate (DMMP) or 30% relative humidity (RH). The ML framework is used to process and classify thousands of images (optical micrographs) collected during the LC responses and we show that classification (sensing) accuracies of over 99% can be achieved. For the same experimental system, we demonstrate that traditional feature information used in characterizing LC responses (such as average brightness) can only achieve sensing accuracies of 60%. We also find that high accuracies can be achieved by using time snapshots collected early in the LC response, thus providing the ability to create fast sensors. We also show that the ML framework can be used to systematically analyze the quality of information embedded in LC responses and to filter out noise that arises from imperfect LC designs and from sample variations. We evaluate a range of classifiers and feature extraction methods and conclude that linear support vector machines are preferred and that high accuracies can only be achieved by simultaneously exploiting multiple sources of feature information.


[1] Clifford K Ho, Alex Robinson, David R Miller, and Mary J Davis. Overview of sensors and needs for environmental monitoring. Sensors, 5(1):4–37, 2005.

[2] K-L Yang, Katie Cadwell, and Nicholas L Abbott. Contact printing of metal ions onto carboxylate-terminated self-assembled monolayers. Advanced Materials, 15(21):1819–1823, 2003.

[3] Kun-Lin Yang, Katie Cadwell, and Nicholas L Abbott. Mechanistic study of the anchoring behavior of liquid crystals supported on metal salts and their orientational responses to dimethyl methylphosphonate. the Journal of Physical Chemistry B, 108(52):20180–20186, 2004.

[4] Katie D Cadwell, Mahriah E Alf, and Nicholas L Abbott. Infrared spectroscopy of competitive interactions between liquid crystals, metal salts, and dimethyl methylphosphonate at surfaces. The Journal of Physical Chemistry B, 110(51):26081–26088, 2006.

[5] Maricar L Bungabong, Peng Bin Ong, and Kun-Lin Yang. Using copper perchlorate doped liquid crystals for the detection of organophosphonate vapor. Sensors and Actuators B: Chemical, 148(2):420–426, 2010.

[6] Santanu Kumar Pal, Claribel Acevedo-Velez, Jacob T Hunter, and Nicholas L Abbott. Effects of divalent ligand interactions on surface-induced ordering of liquid crystals. Chemistry of Materials, 22(19):5474–5482, 2010.

[7] Sheila E Robinson, Bart A Grinwald, Laura L Bremer, Kurt A Kupcho, Bharat R Acharya, and Patrick D Owens. A liquid crystal-based passive badge for personal monitoring of exposure to hydrogen sulfide. Journal of occupational and environmental hygiene, 11(11):741–750, 2014.

[8] Jacob T Hunter and Nicholas L Abbott. Adsorbate-induced anchoring transitions of liquid crystals on surfaces presenting metal salts with mixed anions. ACS applied materials & interfaces, 6(4):2362–2369, 2014.

[9] Tibor Szilv´asi, Luke T Roling, Huaizhe Yu, Prabin Rai, Sangwook Choi, Robert J Twieg, Manos Mavrikakis, and Nicholas L Abbott. Design of chemoresponsive liquid crystals through integration of computational chemistry and experimental studies. Chemistry of Materials, 29(8):3563– 3571, 2017.

[10] Huaizhe Yu, Tibor Szilv´asi, Prabin Rai, Robert J Twieg, Manos Mavrikakis, and Nicholas L Abbott. Computational chemistry-guided design of selective chemoresponsive liquid crystals using pyridine and pyrimidine functional groups. Advanced Functional Materials, 2018.

[11] Luke T Roling, Jessica Scaranto, Jeffrey A Herron, Huaizhe Yu, Sangwook Choi, Nicholas L Abbott, and Manos Mavrikakis. Towards first-principles molecular design of liquid crystalbased chemoresponsive systems. Nature communications, 7:13338, 2016.

[12] Daniel S Miller, Rebecca J Carlton, Peter C Mushenheim, and Nicholas L Abbott. Introduction to optical methods for characterizing liquid crystals at interfaces. Langmuir, 29(10):3154–3169, 2013.

[13] Alireza Hassanzadeh and Robert G Lindquist. Liquid crystal sensor microchip. IEEE Sensors Journal, 12(5):1536–1544, 2012.

[14] Rahul R Shah and Nicholas L Abbott. Orientational transitions of liquid crystals driven by binding of organoamines to carboxylic acids presented at surfaces with nanometer-scale topography. Langmuir, 19(2):275–284, 2003.