A Machine-Learning Approach to Correcting Non-Ideal Background Currents in Disordered Self-Assembled Monolayers

Abstract

In traditional studies of electroactive molecules tethered to metal surfaces through well-ordered self-assembled monolayers (SAMs), cyclic voltammetry is the primary tool used to measure the coverage of the tethered molecules. The charge under the peak in the voltammogram is directly proportional to the coverage of the redox-active molecules. Accurate coverage measurements are possible because well-ordered SAMs exhibit ideal background currents which are easy to subtract from the total current to isolate the SAMs faradaic current. However, for disordered electroactive SAMs with poor surface blocking properties, specific interactions with the metal substrate create non-ideal background currents which are difficult to separate from the SAMs faradaic current. Although analytical approaches exist to account for background currents in other electrochemical systems, they have some fundamental limitations that impede their practical application to electroactive SAMs. In our work, we propose a simple machine-learning approach to empirically correct for background currents in disordered SAMs, and we apply this approach to determine the surface coverage of partially-desorbed ferrocene-terminated monolayers (Fc-SAMs) from gold surfaces. We use LASSO regression and an ensemble of background scans from partially-desorbed unfunctionalized decanethiol SAMs to approximate the Fc-SAM background currents. Using the approximated background, we estimate the coverage of Fc-SAMs after background correction. This technique allows us to explore Fc-SAM coverage as a function of potential when applying potentials near and beyond the onset of the so-called reductive desorption potential of the SAM which induces non-ideal background currents. To verify the accuracy of the coverages determined using our LASSO method, we used two verification methods: 1) backfilling with decanethiol and 2) chemical stripping of the disordered SAM followed by inductively-coupled plasma-mass spectrometry (ICP-MS). The pros and cons of both confirmation methods will be discussed along with a comparison of error rates and limits of detection between traditional background subtraction method and the LASSO method. We propose that this new approach will enable a more quantitative description of disordered SAM films after reductive desorption events.