All chemicals were of analytical grade and used as received. High-molecular-weight poly(vinyl chloride) (PVC), potassium tetrakis(4-chlorophenyl)-borate (KTPB), dibenzo-18-crown-6 (DB18C6), and tetrahydrofuran (THF) were obtained from Fluka AG (Switzerland). PM hydrochloride, dioctylphenylphosphonate (DOPP), bis(2-ethylhexyl) adipate (BEHA), dibenzo-30-crown-10 (DB30C10), α-, β-, γ-CD and tri-n-butyl phosphate (TBP) were purchased from Sigma-Aldrich (USA). PM syrups and tablets were purchased from local pharmacies. Doubly deionised water was used to prepare all solutions. A stock solution of 0.1 M PM was prepared by dissolving an appropriate amount of PM in 50 mL water. Standard solutions (1 × 10⁻² to 1 × 10⁻⁶ M) PM were prepared fresh by diluting the appropriate amount of PM in water. Each stock solution was protected from light by wrapping the flasks with aluminium foil and stored in a refrigerator when not in use. Stock solutions (0.1 M) of various salts and sugars were also prepared. Other solutions were prepared by the subsequent dilutions of the stock solutions.

The FIA system (Figure 2) consisted of a multi-channel peristaltic pump (Gilson Miniplus 3) and an injection valve (Rheodyne Type 500 Teflon rotary selection valve), with an exchangeable sample loop. The rest of the FIA tubings were of 0.80 mm i.d. A sampling volume of 150 μL and carrier flow rate of 2.83 mL min⁻¹ were used throughout. The flow-through cell was of wall-jet design with a built-in Ag/AgCl reference electrode (Model FIP-3, Chemflow Devices, Australia) [26]. The sample was injected into a carrier stream of acetate buffer (0.1 M, pH 6.0). Potassium chloride (0.1 M) was used as reference solution. Potentiometric measurements were performed at 25 ± 2 °C using an Orion model 701A digital ionanalyser that was connected to the flow cell. A glass electrode, connected to an Orion expandable ionanalyser (model EA 940) was used to measure the pH of aqueous solutions.

Potentiometric PM selective sensors was prepared by dissolving PVC (176.0 mg), plasticising solvent (356.4 mg), ionophore (11.0 mg) and KTPB where appropriate (6.6 mg) in THF (5 mL). Four drops of the mixture were deposited on the electrode body [26] and the system was allowed to evaporate for 3 h. The membrane was then conditioned in 1 × 10⁻² M PM for 1 h prior to use. PM standard solutions covering the range 1×10⁻⁶ to 1×10⁻¹ M was injected into the FIA unit. Each solution was measured in triplicate. The average potentials at maximum heights were plotted against log [PM].

Selectivity coefficients, K A , B pot, of the fabricated sensors towards several cations and sugars were measured using the separate solution method [27] and the following equation was applied: (1) log   K A , B pot = E B − E A S + [ 1 − Z A Z B ]   log   a Awhere E_A and E_B are the potential readings of PM and interferring ions, respectively; a_A is the activity or concentration of PM. Z_A and Z_B are the charge of PM and interferring ions, respectively and S is the slope of the calibration graph.

The United States Pharmacopoeia method [28] was used for the determination of PM in certain samples as comparison to the proposed method.

The starting geometries of PM and the host structures (α-CD, β-CD, γ-CD, DB18C6 and DB30C10) were built based on the structures that were generated from the crystallographic parameters provided by the Cambridge Structural Database (CSD) [29–33] and were separately optimised using the semi empirical method, PM3 using Gaussian03 software package [34]. The starting geometries of the inclusion complexes were constructed using HyperChem (Version 7.0, Hypercube, Gainesville, FL, USA). The previously optimised structures of PM and host molecules were allowed to approach each other along the symmetric axis (the X-axis) passing through the center of the host cavity. For example, in the case of β-CD, the coordinate system used to define the process of complexation was based on constructing the CD with the seven identical glucose units positioned symmetrically around the Z-axis, such that all the glycosidic oxygens are in the XY plane and their center was defined as the center of the coordination system [35]. The PM molecule was docked into the cavity of the CD with the central nitrogen atom connecting the two benzene rings that coincides with the Z-axis. Docking was initially done to maximise the electrostatic and hydrophobic interactions between the host and the guest molecules. Multiple starting points were generated by moving the guest molecules along the – and + Z-axis from 10 to −10 Å, at 1 Å intervals, and by rotating the guest molecules from 0°–360° at 45° intervals. Three different inclusion orientations, i.e., the nitrogen from the alkyl group of the PM vertically, facing up and down horizontally into the cavity of the host, were considered for each case (Figure 3).

The inclusion interactions were simulated in vacuum and the presence of water molecules were ignored to save computational time especially for large molecules. The complexation energy, ΔE_comp, was calculated for the minimum energy structures by the following equation: (2) Δ E comp = E comp − E G − E Hwhere E_comp, E_G, and E_H represent the total energy of the host-guest complex, the free guest molecule and the free host molecule, respectively. The magnitude of the energy change is an indication of the driving force towards complexation. The more negative the complexation energy change, the more thermodynamically favourable is the inclusion complex.

Initially, the effect of the type of ionophore on the sensor performance was investigated using BEHA as plasticiser. Membranes using α-CD and DB18C6 (sensors 2 and 5, Table 1) resulted in non-functional sensors.

The cavities of these hosts are probably too small [α-CD (174 Å)] [35] to accommodate the guest molecule. When hosts with larger cavity sizes ([β-CD (262 Å), γ-CD (427 Å) and DB30C10] were used, promising PM sensors were obtained. The characteristics of these sensors were further improved by adding KTPB as membrane additive (sensors 10–12), and near Nernstian responses were obtained. Several different membrane compositions that contained 1.0, 1.2, 1.6, and 2.0% of KTPB were also investigated. The best composition of KTPB, resulting in better slope, correlation coefficient and wide concentration range was 1.2%, therefore this amount of KTPB was used in all membranes. The beneficial effects of adding additives such as KTPB have been noted for reducing the membrane resistance and suppressing the permeation of counter anions in aqueous phase into the membrane phase [25,39]. On the contrary, sensors without the additive (sensors 3, 4 and 6) exhibited shorter linear range and produced sub-Nernstian slopes. It has been reported that in the absence of an additional lipophilic anion exchanger, poor performance may be caused by possible trace anionic impurities within the PVC used [39]. Non-functional sensors were obtained with membranes that only contained plasticisers (sensors 7–9). The observed effects of the membrane compositions on the sensor response are clear indicators that inclusion complexes were formed between the larger cavity hosts and the PM guest molecule.

The lifetime of the sensors were studied by continuously pumping for 300 min and repeatedly injecting 1 × 10⁻³ M PM at 30 minute intervals (Figure 5). No noticeable reduction in peak heights over this period was observed for sensor 10. In contrast, sensors that were plasticised with DOPP and TBP (sensors 13 and 16, respectively) experienced a marked reduction in peak height after 60 minutes of continuous use. Indeed, it was found that sensor 10 remained operational over one month of daily use.

Selectivity is an important characteristic of a sensor that delineates the extent to which the device may be used in the estimation of an analyte ion in the presence of other ions and extent of utility of the sensor in real sample measurements. The selectivity coefficient ( K A , B pot ) of the sensor was determined using the separate solution method [27] at a fixed concentration of 1 × 10⁻³ M for both drug and interfering ions. Table 2 shows that there are no significant interferences from the cations, trifluoperazine and the sugars tested. Insignificant interferences from the alkali and alkaline earth metal ions suggest the good prospect of the sensor to be used in biological fluids. The overall selectivity of sensor 10 is better than that of sensors that are based on γ-CD or DB30C10 (sensors 11 and 12).

Under the optimum conditions, response time of not more than 5 s was obtained for an injected sample, translating into a sample throughput of about 96 samples h⁻¹. The FIA system was characterised by a dispersion coefficient of 1.9, which was calculated using the equation: (3) D = h o / hwhere h_o and h is the peak height for undispersed and dispersed sample, respectively. The limited dispersion obtained indicate the limited mixing of analyte and solutions in the flow system, ensuring a sensitive detector output.

To further understand how molecular recognition takes place at the atomic level, molecular modelling techniques were used to complement the experimental studies. The complexation energies (E_comp) of the PM inclusion complexes are listed in Table 4. The optimised geometries for the lowest energy conformation for the inclusion complexes of PM with CDs and dibenzo-crown ethers are shown in Figure 6, respectively. In general, the inclusion complexes of CDs with PM have larger binding energies when compared to their dibenzocrown ethers counterpart. The overall lowest binding energies with the different hosts decrease in the order of: γ-CD > β-CD > α-CD > DB30C10 > DB18C6.

The dominant driving force for the complexation is evidently H-bondings interaction. Close observation on the structure of the inclusion complexes shows that γ-CD complexes have more H-bonding interactions compared to the β-CD complexes. Various authors have also reported that intermolecular hydrogen bondings between the host-guest interactions resulting from the fitting of small molecules into the CD cavities play an important role in the binding energies of a number of CD inclusion complexes [40,41]. An almost complete inclusion of the guest molecule was observed for the promethazine/γ-CD complex (Figure 6(c)). Meanwhile, for the promethazine/β-CD complex, only the N-alkyl group is included leaving the aromatic rings protrude outside the cavity (Figure 6(b)). However, for the α-CD the N-alkyl group barely touches the upper rim of the α-CD (Figure 6(a)). Consistently, no H-bondings were observed for α-CD and that of the crown ether complexes. The high and positive complexation energy displayed by promethazine/DB18C6 complex (Table 4), shows that inclusion complex between the two molecules are highly unfavourable from the theoretical point of view, consistent with the experimental observations.