Real‐time Voltammetric Anion Sensing Under Flow

Abstract The development of real‐life applicable ion sensors, in particular those capable of repeat use and long‐term monitoring, remains a formidable challenge. Herein, we demonstrate, in a proof‐of‐concept, the real‐time voltammetric sensing of anions under continuous flow in a 3D‐printed microfluidic system. Electro‐active anion receptive halogen bonding (XB) and hydrogen bonding (HB) ferrocene‐isophthalamide‐(iodo)triazole films were employed as exemplary sensory interfaces. Upon exposure to anions, the cathodic perturbations of the ferrocene redox‐transducer are monitored by repeat square‐wave voltammetry (SWV) cycling and peak fitting of the voltammograms by a custom‐written MATLAB script. This enables the facile and automated data processing of thousands of SW scans and is associated with an over one order‐of‐magnitude improvement in limits of detection. In addition, this improved analysis enables tuning of the measurement parameters such that high temporal resolution can be achieved. More generally, this new flow methodology is extendable to a variety of other analytes, including cations, and presents an important step towards translation of voltammetric ion sensors from laboratory to real‐world applications.


S1.1 General Information
All experiments were performed at room temperature in the presence of oxygen. All commercially available chemicals and solvents were used as received without further purification. All hygroscopic tetrabutylammonium (TBA) salts were stored in vacuum desiccators at room temperature. Ultrapure water was obtained from a Milli-Q system (18.2 MΩcm). Supporting electrolyte (TBAClO4 from Sigma Aldrich) was of electrochemical grade.

S1.2 Electrochemical Measurements & SAM Formation
All experiments were conducted using an Autolab Potentiostat (Metrohm) or PalmSens4 Potentiostat with a three-electrode setup equipped with a gold disc working electrode (BaSi, 1.6 mm diameter) and platinum wire counter electrode. A non-aqueous Ag|AgNO3 reference electrode (with an inner filling solution of 10 mM AgNO3, 100 mM TBAClO4 in ACN) was utilised for all experiments and all potentials are reported wrt. to this reference electrode. All experiments were carried out with 100 mM TBAClO4 as a supporting electrolyte with additional 10 mM HClO4, as indicated. In all cases, including sensing studies, the ionic strength was maintained at a constant 100 mM TBA-anion (+ 10 mM HClO4) throughout.
Au disc electrodes were cleaned according to previously reported protocols. 2 Immediately following the cleaning procedure, the Au disc electrodes were rinsed thoroughly with water and ethanol and immersed in a solution of 0.25 mM 1.XB/HB in ACN overnight in the dark.
Subsequently, the Au disc electrodes were rinsed with copious amounts of ACN and then used immediately. Detailed surface characterisation of the so-formed SAMs is reported elsewhere. 1

S1.3 Voltammetric Measurements
Static square wave voltammetry (SWV) measurements were conducted with a step potential of 2 mV, amplitude of 20 mV and frequency of 25 Hz. Continuous flow SWV measurements were carried out between -0.25 to 0.3 V with a step potential of 2 or 5 mV (typically 5 mV, unless otherwise indicated), amplitude of 20 mV and frequency of 50 Hz, with all pre/postequilibration times set to 0 s.

S1.4 Binding Isotherm Analysis and LOD Determination
All anion-induced shifts are reported with respect to the potential in the baseline preceding each injection for flow experiments or with respect to the initial SWV preceding the first addition for static titrations. All binding isotherm fitting was carried out with OriginPro 2017.
All binding constants are rounded to three significant figures and were obtained by fitting of the sensing isotherms to the Langmuir-Freundlich model (Eqn. 1). For quantitative analysis, the sensing isotherms for H2PO4were corrected for full protonation of H2PO4by the acidified electrolyte, by correcting isotherms by -10 mM.
LODs were calculated according to Eqn. S1 where σ is the standard deviation of the baseline/blank and S is the slope of the linear region of the sensor response (herein also referred to as "sensitivity"). For continuous voltammetric measurements under flow, σ was determined from the root-mean-square deviation of a linear fit of ten data points (E1/2 values) in the initial baseline of the sensograms (determined by the respective analysis method as used for the measurement, PeakPick or AsymFit), immediately preceding the response to the first addition of analyte. For static voltammetric measurements, σ was determined as the standard deviation of ten data points (E1/2 values) from ten SWVs performed immediately preceding the first analyte addition. S was determined from the slope of a linear fit to the pseudo-linear regime of each respective binding isotherm, with a range between 0-11 mM for HSO4and Cl -, and either 9.5-13.4 or 10.5-13.4 mM for H2PO4 -(for static and continuous measurements, respectively).

=
Eqn. S1 Experimental protocols for flow measurements are detailed in Section S3. In the presence of 10 mM HClO4 the binding isotherms for all anions are affected to such a degree that standard binding models cannot accurately describe them anymore. As the of the anions in the solvent system are not known, the concentration of free anion cannot be calculated. Consequently, no corrections (or adjusted binding models) can be applied.
We thus utilise the Langmuir-Freundlich isotherm as a semi-empirical model to describe the sensor response. "n" is typically interpreted as a heterogeneity factor and herein accounts for the "heterogeneity" imposed by partial anion protonation.
Eqn. 1 was chosen as an empirical model which relates the shift in potential (ΔE) to coverage, whereby the maximum shift in potential induced (ΔEmax) by specific target binding correlates to a maximum coverage (θ = 1). Additionally, it is able to account for some of the inhomogeneities observed, that alternative, simpler models (e.g. Langmuir, Nernstian) cannot.  Figure S5.3). Similarly, the LODs given for H2PO4are "apparent" LODs in the "absence" of acid, i.e. are also corrected by -10 mM.

S3.1 Continuous Voltammetric Measurements
A 3D printed microfluidic cell (see Figures     Inset depicts the two running modes for the microfluidic system: load and inject. Running electrolyte is continuously pumped throughout. When in load mode, only fresh running electrolyte passes through the system but when switched to inject mode, electrolyte passes through the sample loop which was filled with the analyte solution prior to switching.

S3.4 Optimisation of SWV Parameters for Flow Measurements
To With these parameters a single SW scan took ≈12.6 s to record, which, under the experimental flow conditions represents a temporal resolution that is slightly too low to accurately measure the anion specific cathodic shifts (see Figure S3.6). With this in mind, we optimised the SWV parameters in order to reduce tscan, which is theoretically given by Eqn.  . Notably, as a result of the comparably low temporal resolution of ≈12.6 s not many data points fall within each response peak arising from analyte addition, such that errors will be larger.

S4.1 AsymFit Method
Initialisation parameters for each variable in Eqn. S3 were estimated for each SWV as follows: • y0 (alpha), is the initial current value from the isolated data set around each peak.
• A (beta), is the peak height given by the difference between the maximum current and y0.
• xc (Guessxpeak) is the estimated peak potential, as determined via the PeakPick method.
• c1, c2 and c3 (full width at half maximum, FWHM) are all given by an estimation for the full width half maximum value, here as the difference between the potential at the 6 th and 16 th data points from the isolated data set around each peak. ) ] Eqn. S3 Figure S4.1. Schematic depiction of the steps for the AsymFit data analysis method. 1. An initial estimate for E1/2 was obtained from the PeakPick method of the raw data. 2. and 3. All data points ±50 mV around this estimated value were isolated (i.e. the baseline was removed). 4. The isolated peak was fitted according to the asymmetric double sigmoidal function (Eqn. S3). 5. The E1/2 of this fitted, continuous peak distribution was obtained by the PeakPick method (as E1/2 at Imax).   As a result of the washing steps following each addition whereby the baseline is reestablished, minor baseline drifts can be accounted for, which is not possible for static titrations. This may explain why the static and continuous isotherms differ slightly in some cases, in particular for chloride (see Figure S5.5).