**1. Introduction**

The corrosion protection of an organic coating depends largely on the intrinsic barrier properties of the polymeric film towards oxygen, water, and aggressive species, but can be reinforced by entrapped corrosion inhibitors [1–3]. In the case of bronze artworks exposed outdoors, commercial coatings such as Incralac ® and Soter ® contain benzotriazole (BTA) that operates in the dual functions of an inhibitor of bronze corrosion and an anti-UV additive [4]. In these coatings, the inhibitor dissolves in the electrolyte (in the rain) and penetrates through the coating, so producing an inhibited solution that exerts a protective action at the coating/metal interface. In general, the positive influence of direct inhibitor addition in coatings may be limited by solubility problems, by a decrease in barrier properties and adherence to the substrate, and by the rapid leaching of small inhibitor molecules induced by rainfalls, so determining a fast drop of the overall protection performance [5].

In recent years, grea<sup>t</sup> e fforts were devoted to overcoming these problems and increasing the coating durability by encapsulating corrosion inhibitors in coatings through the adoption of suitable carriers, which make them more compatible with the coating network [6,7]. Di fferent carrier types were investigated, such as inhibitor-filled porous particles, or nanocapsules, coated by polyelectrolytes [8–10]

or layered anion-exchange particles containing intercalated corrosion inhibiting anionic species [11]. The former polyelectrolyte-coated carriers can release inhibitor molecules due to local variations of solution pH linked to the onset of corrosion phenomena, while the latter carriers with intercalated inhibiting anionic species limit the access of aggressive anions, such as chlorides, to the metal substrate and release corrosion inhibiting species by ion-exchange phenomena.

Beside these carriers, cyclodextrins (CD) were also used to host corrosion inhibitors due to their complexing capability towards small organic molecules with corrosion inhibition properties. Cyclodextrins (CDs) [12] are water-soluble macrocyclic oligosaccharides consisting of at least six α -D-glucopyranose units linked via α (1–4) glycosidic bonds. The most common ones contain 6, 7 and 8 glucopyranose residues and are known as α-, β-, and γ-CD, respectively. CDs are among the most widely used host molecules thanks to their unique and specific structure that creates an internal cavity that is less hydrophilic than the external aqueous environment and therefore able to accommodate a large variety of hydrophobic molecules inside it [13]. The strong propensity of CDs to form inclusion complexes has been exploited not only for the production of smart coatings for corrosion protection, but also in many other fields of science, for example, as drug transport systems, [14], to increase the solubility of some chemical species [15], in separation technology [16], and in other areas [17].

Addition of complexes of α- and β-CDs with dibenzylthiourea (DBT) in acid solutions improved the corrosion resistance of carbon steel due to the enhanced solubility of the complexes in comparison to that of DBT alone [18]. The controlled release of BTA from β-CD–BTA complexes was investigated in order to achieve effective bronze corrosion protection in chloride solutions. The use of the complex instead of pure BTA was intended to reduce the toxicity of the additives used for corrosion protection [19]. Moreover, complexes of β-CD and γ-CD with mercaptobenzothiazole (MBT), mercaptobenzimidazole (MBI), mercaptobenzimidazole sulfonate, and thiosalicylic acid [20–23] were incorporated in coatings and improved the corrosion resistance of Al alloys and zinc. In fact, these complexes represent a reservoir of corrosion inhibitor molecules that, at the onset of corrosion phenomena and in correspondence of regions of coating delamination, tend to adsorb on the metal surfaces, shifting the complexation equilibrium towards the release of further inhibitor molecules.

In this research, the effective formation and the stability of a complex between β-CD and 5-mercapto-1-phenyl-tetrazole (MPT) (Figure 1) were assessed. Among non-toxic corrosion inhibitors [24–26], MPT was chosen because of its outstanding inhibiting properties towards copper and bronze corrosion [27–29]. Its complex with β-CD was incorporated in a 3-mercapto-propyltrimethoxy-silane (PropS-SH) coating, and the protectiveness of the obtained coating was assessed on bronze during both immersions in concentrated acid rain and exposures to alternated acid rain spray. These tests were also performed on plain silane coating and coatings with only β-CD or only MPT additions. For the continuous immersion tests, thin sub-micrometric silane films prepared by the dip coating method were used, in order to better differentiate their protectiveness. During the alternated acid rain spray test, the self-healing properties were assessed on thicker coatings (about 5 μm) produced by spraying after introduction of cross cut scratches. Spraying and brushing are the application methods most commonly adopted by restorers [30] and produce coatings more representative of those actually applied in the field.

It is important to stress that according to occupational hazard tests, the silane formulations here investigated were less toxic to restorers than Incralac® (as both pure product and ready to use 30% solution in nitro diluent) [31].

**Figure 1.** Molecular structure of (**a**) 5-mercapto-1-phenyl-tetrazole (MPT) and (**b**) β-cyclodextrin (β–CD). β–CD is a CD type constituted by 7 glucopyranoside units linked by 1,4 glycosydic bonds.

#### **2. Materials and Methods**

#### *2.1. Chemicals, Aggressive Environment and Alloy*

All reagents and solvents used in this study were purchased from commercial sources. In particular, the chemicals used for the inhibitor complex and the coating production were β-cyclodextrin (β-CD, ≥ 97% purity), 5-mercapto-1-phenyl-tetrazole (MPT, 98% purity), and 3-mercapto-propyl-trimethoxy-silane (PropS-SH, purity 95%), all purchased from Sigma-Aldrich (Darmstadt, Germany).

The coating protectiveness was tested on as-cast bronze with composition 91.9 Cu, 2.4 Sn, 1.0 Pb, Zn 2.9, 0.8 Sb, wt.%, and a microstructure reproducing those of Renaissance bronze artefacts with cored dendrites of Cu-solid solution characterized by Sn and Sb local enrichment and also including Pb globules in the interdendritic spaces, as reported in previous papers [29,31–33].

Concerning the environments where the coating protectiveness was assessed, the cyclic acid rain (AR) spray test was performed using a synthetic AR, prepared with Sigma-Aldrich ACS reagents, according to the recipe reported in [34] and containing the following ion concentrations: SO4<sup>2</sup>− 1.90 mg· <sup>L</sup>−1, Cl− 1.27 mg· <sup>L</sup>−1, NO3− 4.62 mg· <sup>L</sup>−1, CH3COO− 0.23 mg· <sup>L</sup>−1, HCOO− 0.05 mg· <sup>L</sup>−1, NH4+ 1.05 mg· <sup>L</sup>−1, Ca2+ 0.34 mg· <sup>L</sup>−1, Na<sup>+</sup> 0.53 mg· <sup>L</sup>−1, and pH 4.35. During the electrochemical tests, accelerated corrosion conditions were obtained by using tenfold concentrated AR (ARX10, pH = 3.3) at 30 ◦C.

#### *2.2. [*β*-CD–MPT] Complex Stability Analysis*

#### 2.2.1. Nuclear Magnetic Resonance (NMR) Measurements

The NMR spectra were recorded in D2O solution using 5 mm tubes, at 296 K, with a Varian Mercury Plus 400 (Palo Alto, CA, USA), operating at 400 (1H) and 100 MHz (13C), respectively. The chemical shifts were referenced to the DOH signal: δ (H) 4.65 ppm. The 1D-Rotating frame Overhauser effect spectroscopy (ROESY) NMR spectra were acquired using standard pulse sequences from the Varian library. The relaxation delay between successive pulse cycles was 1.0 s.

The Diffusion-Ordered Spectroscopy (DOSY) experiments were performed using the Dosy Bipolar Pulsed Pair STimulated Echo (DBPPSTE) pulse sequence [35] from the Varian library, using 15 different gradient values varying from 2% to 95% of the maximum gradient strength. A 500 ms diffusion time was chosen, and the gradient length was set to 2.0 ms. The analysis of all NMR spectra was performed with MestreNova (by Mestrelab Research, S.L., Santiago de Compostela, Spain), version: 6.0.2–5475 and for the DOSY analysis, the Baysian DOSY transform from MestreNova, version: 6.0.2–5475 was used.

#### 2.2.2. H NMR Titration

The following two solutions were prepared in D2O: Solution A: 2.8 mM MPT. Solution B: 2.8 mM MPT and 12.0 mM β-CD. A 0.8 mL aliquot of solution A was placed in a 5 mm NMR tube. A measured amount of solution B was added, changing the molar fraction of the host to about 0, 0.39, 0.71, 0.98, 1.21, 1.68, 2.02 and 2.29. Spectra were recorded after each addition. The chemical shift variation of the gues<sup>t</sup> signals was collected, and the binding constants β (as log K) were calculated by the curve fitting method [36] using the commercial HypNMR2008 [37] program (details are given in the Supplementary Information (SI) file).

#### 2.2.3. Electrospray Ionization (ESI) Mass Spectra

ESI mass spectra were obtained using an LCQ Duo (ThermoQuest, San Jose, CA, USA) in negative-ion mode. Instrumental parameters: capillary voltage –10 V, spray voltage 4.50 kV, mass scan range was from m/z 100 to 2000 amu, for 30,000 ms scan time; N<sup>2</sup> was used as the sheath gas. The samples were injected into the spectrometer through a syringe pump at a constant flow rate of 8 mL/min.

#### 2.2.4. Fourier Transform Infra-Red (FTIR) Analysis

Di ffuse reflectance FTIR spectra were recorded on β-CD–MPT complex powder and on MPT and β-CD powders, as references. The instrument used was a Thermo-Scientific Nicolet iS50 spectrometer (Thermo Fisher Scientific, Waltham, MA, USA), operating in dry CO2-free air flow generated by a Balston 75-52 unit. It was equipped by a deuterated triglycine sulfate (DTGS) detector, which allowed for the investigation of the 4000–400 cm<sup>−</sup><sup>1</sup> wavenumber region with a resolution of 4 cm<sup>−</sup>1.

#### *2.3. Silane Coating Production*

Silane hydrolysis was carried out by dissolving PropS-SH in a hydroalcoholic solution (90/5/5 v/v ethanol/water/PropS-SH), acidified to pH 4 by the addition of some drops of diluted sulphuric acid solution, according to the methodology refined in previous research works [29,38–40].

Plain PropS-SH coatings were directly produced from this solution after 24 h room-temperature ageing. The coating additives were introduced in the aged silane hydroalcoholic solution; in particular, 3 mL aqueous solutions of either 5.94 mM MPT or 5.94 mM β-CD or 5.94 mM β-CD + 5.94 mM MPT were added to 30 mL of silane solution, so that the final molar concentration of the additives in the coating formulations was 0.54 mM. These solutions were sonicated for 3 min and then applied to the substrate either by dip coating (1 h immersion and then fast withdrawal, reaching a final coating thickness of about 300 nm; for accelerated electrochemical tests) or by spraying (to reach a final constant coating specific weight of 6 ± 1 <sup>g</sup>·m<sup>−</sup><sup>2</sup> and thickness of 5 ± 1 μm; for cyclic AR spray tests). Finally, the coatings were cured for 24 h at 50 ◦C. This low temperature curing was compatible with the requirements for cultural heritage bronze artworks.

#### *2.4. Silane Coating Protectiveness*

Electrochemical tests were performed under accelerated corrosion conditions, that is in ARX10 (pH 3.3) at 30 ◦C, on thin dip-coated bronze electrodes. As a reference, tests were also carried out on bare bronze electrodes.

The evolution of corrosion conditions was monitored over 20 days of immersion by Electrochemical Impedance Spectroscopy (EIS, performed by a PARTSTAT 2273, from Ametek, Berwyn, PA, USA) tests performed at intervals, under the following experimental conditions: corrosion potential (*E*cor) ± 10 mV rms, 10 kHz–1 mHz frequency range and 10 frequencies/decade. Polarization resistance ( *R*p) values were estimated from the spectra in the Nyquist form, as the di fference between the limit of the real part of the impedance at frequency tending to 0 ( *<sup>R</sup>*p') and the solution resistance ( *R*s) value (*R*p = *<sup>R</sup>*p' – *R*s) [40]. *<sup>R</sup>*p values are inversely proportional to the corrosion currents (*i*cor), as indicated

by the Stern and Geary relationship [41]: *Rp* = B*icor* , with B a constant depending on the Tafel slopes of the anodic and cathodic polarization curves. The time evolutions of average *<sup>R</sup>*p and *E*cor values were obtained from triplicate experiments.

Ohmic drop-compensated polarization curves were collected at the end of the 20 day immersion period. Separate anodic and cathodic potential scans, always starting from *E*cor, were carried out at a rate of 0.1667 mV·s<sup>−</sup>1. These tests were performed in triplicate and representative curves were reported.

The self-healing capability of PropS-SH coatings was assessed by exposing coated coupons with cross cut scratches to a cyclic AR spray test at 35 ◦C for 4 weeks. Each cycle consisted in 8 h spraying and 16 h waiting. During the test, the coupons were supported with an angle of 30◦ from the vertical. Micrographs documented the extent of the final corrosion attack.
