New insights into methanol and formic acid electro-oxidation on Pt: Simultaneous DEMS and ATR-SEIRAS study under well-defined flow conditions and simulations of CO spectra

There continues to be great interest in the electro-oxidation of methanol and formic acid on Pt since they are fuels for direct methanol/formic acid oxidation fuel cells. Although extensive studies of the oxidation of methanol and formic acid on Pt-based catalysts have been carried out, their mechanisms remain a subject of continued debate.^1–7 It is generally accepted that methanol and formic acid oxidation reactions proceed through two pathways: an indirect pathway through an adsorbed CO intermediate, and a direct oxidation pathway.^2,5–8 In addition, methanol oxidation also generates numerous intermediates including formaldehyde, formic acid, and methylformate, which can be further oxidized to CO₂, but their further oxidation is dependent on convection, electrode materials, electrode size and surface roughness, and electrolyte and methanol concentration.^{3,5–7,9–11} Linear-bonded CO, bridge-bonded CO, and formate have been identified as adsorbed species during methanol and formic acid decomposition/oxidation on Pt by infrared spectroscopy.^12,13 These adsorbed CO species are difficult to be oxidized on pure Pt and, thus, can, and often do, poison Pt surfaces.² PtRu is generally considered to be the most active catalyst for methanol oxidation due to the bifunctional mechanism.^6,9,14 In contrast, Pb-modified Pt, PtPb alloys, and intermetallics have been found to be the most active catalysts for formic acid oxidation due to the “third-body” effect, as well as electronic effects.^2,7,15 It has also been proposed that adsorbed formate could be an active intermediate during methanol and formic acid oxidation based on the observation that the oxidation current is proportional to its infrared intensity.^12,13,16 In contrast, adsorbed formate was also reported to be a spectator species, such as bisulfate/sulfate and acetate.^17,18 Other adsorbed species, such as CH₂OH_ad, CHOH_ad, COH_ad, and COOH_ad, have been proposed to be intermediates during methanol oxidation on Pt.^2,4 However, there is no direct spectroscopic evidence to support such statements.¹⁹ These adsorbed intermediates might, and likely do, have very short lifetimes, and, thus, their coverage might be too low to be detected under operando conditions. Although methanol and formic acid oxidation on Pt has been extensively studied, the detailed mechanism is still far from being fully understood.

In order to advance our understanding of these processes, we have studied methanol and formic acid oxidation on a Pt film electrode in 0.1M HClO₄ by a combined spectroscopic system of differential electrochemical mass spectrometry (DEMS) and attenuated total reflection-surface enhanced infrared absorption spectroscopy (ATR-SEIRAS) under well-defined flow conditions in a dual thin-layer flow cell,²⁰ which can avoid the influence of generated intermediates/products on methanol and formic acid oxidation. We have also investigated the effects of formic acid concentration and Pb modification on formic acid oxidation on Pt. Adsorbed HCOOH and CHO species were first identified using ATR-SEIRA spectroscopy. The effects of the Pt film thickness and morphology on infrared spectra were also discussed. The anomalous infrared spectra of linearly bonded CO were simulated with the Fresnel equations and Bruggeman effective medium theory.^21,22 These studies have provided new observations and valuable insights into methanol and formic acid oxidation mechanisms on Pt-based electrodes.

DEMS and ATR-SEIRAS were combined through a dual thin-layer flow cell (Fig. 1), which was made of Kel-F.²⁰ The thickness of the upper and lower compartments was ∼200 and 100 µm, respectively. In the top compartment of the dual thin-layer flow cell, a semi-cylindrical Si prism with a chemically deposited Pt film on the flat rectangular side was used as the working electrode, which enabled the use of a Fourier transform infrared (FTIR) spectrometer (Varian 670) with an ATR configuration to detect adsorbed species on the Pt film. Approximately, 1 cm² surface area of the Pt film was exposed to the solution. The solution flowed into the top compartment through an inlet capillary and then contacted the working electrode. After electrochemical oxidation, the products were transferred into the bottom compartment with the solution flowing through six small capillaries. In the bottom compartment, there was a porous Teflon membrane (with ∼100 µm thickness, a pore size of about 0.02 µm, and a porosity of 50%, Gore-Tex), which served as the solution and vacuum interface. When the solution contacted the porous Teflon membrane, the volatile species could diffuse through it and evaporate into the vacuum system and, subsequently, be detected by using a quadrupole mass spectrometer (HiQuad with QMA 430, Pfeiffer). The vacuum system was differentially pumped by two turbomolecular pumps (80 l/s, Pfeiffer), which were backed with a rotary vane pump (10 m³/h, Pfeiffer). The incoming volatile species could be immediately pumped out after being detected by using the mass spectrometer. The solution flow rate was about 10 µl/s. Methanol [>99.9%, high-performance liquid chromatography (HPLC)] was obtained from Honeywell Burdick & Jackson. Formic acid (88%, AR) was obtained from Mallinckrodt Chemicals. Perchloric acid (70%–72%, ACS, ISO, Reag. Ph Eur, EMSURE) was purchased from Merck. The electrolyte was prepared with Millipore water (18.2 MΩ cm). The experiments were performed at room temperature (20 °C).

The DEMS calibration for CO₂ was carried out by using formic acid oxidation, and the calibration constant for CO₂ at m/z = 44 [K*(44)] was obtained using the equation: K*(44) = 2 × I_MS/I_F, where I_MS and I_F are the mass spectrometric current or charge at m/z = 44 and Faradaic current or charge during formic acid oxidation, respectively. To calibrate the DEMS for methylformate, we flowed a freshly prepared methylformate aqueous solution (C₆₀ = 10 mM) and CO₂ saturated aqueous solution (C₄₄ = 37 mM) into the dual thin-layer flow cell at the same flow rate as that during methanol oxidation measurement, respectively, and measured mass spectrometric signals of m/z = 44 (I₄₄) and 60 (I₆₀), respectively. The calibration constant [K*(60)] for methylformate at m/z = 60 was obtained by using the equation: K*(60) = K*(44) × (I₆₀/C₆₀)/(I₄₄/C₄₄). The details for calculating current efficiencies of CO₂ and methylformate were described in a previous publication.^{3,5–7,9–11}

An EG&G PARC 273 potentiostat and 175 programmer function generator were used for electrochemical measurements. A homemade LabVIEW software and a National Instruments data acquisition card were used to collect data. A dry air system (Agilent) was used to purge the FTIR system and accessories to remove humidity and CO₂. The electrolyte was purged with ultra-high purity Ar (Airgas) for at least 30 min before measurements. A reversible hydrogen electrode (RHE, 0.1M HClO₄) was used as the reference electrode.

A Pt film was chemically deposited onto the Si prism through the following procedures. The flat rectangular surface of the Si prism was polished with 0.05 µm alumina on a polishing cloth (Buehler). Afterward, the Si prism was sonicated in Millipore water and acetone, respectively, to remove alumina and organic contamination. The rectangular face of the cleaned Si prism was dipped into a solution containing 0.5% HF and 1 mM PdCl₂ for 5 min to deposit Pd on it. The Pd deposit can improve the adhesion of the subsequently deposited Pt film. Subsequently, under infrared light, 0.5 ml of Pt electroless plating solution (0.01M K₂PtCl₄ + 0.33M NH₃ + 0.06M NH₂NH₂) was dropped onto the rectangular face of the Si prism to deposit Pt for 15 min. The thickness of the Pt film was estimated to be ∼90 nm by assuming that the whole Pt salt was fully reduced. A second type of Pt film was prepared by using the same procedure, except employing a different plating solution (0.01M K₂PtCl₆ + 0.67M NH₃ + 0.06M NH₂NH₂). Except for mentioning the type of Pt film, the Pt film refers to the first type of Pt film throughout the entire manuscript.

The Pt film was characterized with x-ray diffraction (XRD). XRD patterns of the Pt film were acquired with a Rigaku Ultima VI diffractometer with a Cu K_α source at a scan rate of 5°/min and an incremental step of 0.01°.

The morphology and thickness of the Pt film were investigated with atomic force microscopy (AFM). AFM images were obtained with an Asylum-MFP3D-Bio-AFM-SPM instrument using a tapping mode.

The morphology of the chemically deposited Pt films on the Si prism was investigated with AFM, and its two-dimensional (2D) and three-dimensional (3D) images are presented in Figs. 2(a) and 2(b), respectively. A rough Pt surface was formed during the chemical deposition procedure, and the Pt film was composed of interconnected ∼100 nm-size particles with some pores. This porous and rough surface can provide strong infrared signals of adsorbed species due to surface enhancement effects. The thickness of the Pt film, estimated via AFM, was about 70 nm, in good agreement with that estimated from the amount of Pt in the electroless plating solution (∼90 nm).

XRD patterns of the Pt film deposited on the Si prism are shown in Fig. 2(c). Except for the peaks at 2θ = 22.7° and 34.4°, which can been assigned to XRD peaks of Si oxides,²³ the XRD peaks matched quite well with the standard XRD data of fcc Pt (PDF card No. 00-004-0802). The domain size of the Pt film was estimated from Rietveld analysis to be ∼10 nm.

Figure 2(d) shows the cyclic voltammogram (CV) of a Pt film in 0.1M HClO₄ at 5 mV/s, which exhibits the typical hydrogen adsorption/desorption and oxygen adsorption/desorption features. The roughness of the first and second Pt films was estimated to be ∼8 and 5, respectively, assuming that the H adsorption charge on polycrystalline Pt is 210 µC/cm².

The potential dependent water adsorption infrared spectra of the first Pt film in 0.1M HClO₄ are presented in Fig. S1(I) with a potential interval of 50 mV. In the hydrogen region, water co-adsorbed with adsorbed H on Pt so that there were two types of adsorbed water, indicated by two O–H vibrational bands of H₂O [ν(O–H)] at 3250 cm⁻¹ (low frequency band) and 3600 cm⁻¹ (high frequency band), which could be assigned to water adsorbed on H-covered Pt and on H-free Pt sites, respectively. In the double layer region, only one ν(O–H) band of adsorbed water was observed at around 3500 cm⁻¹. In the oxygen region, the ν(O–H) band intensity of adsorbed water decreased, suggesting that water adsorption to Pt was suppressed due to the formation of Pt oxides. Water might still adsorb on the oxygen-covered Pt, but it was far from the Pt surface so that the water band intensity decreased. In addition, the effects of adsorbed water orientation cannot be ruled out. The HOH bending band [δ(H–O–H)] of adsorbed water was also observed at ∼1648 cm⁻¹ and was parallel with the ν(O–H) band. The electrochemical current and corresponding ν(O–H) band intensity of adsorbed water are plotted vs potential in Figs. S1(II-a) and S1(II-b), respectively. At potentials below 0.8 V, the band intensity of adsorbed water was somewhat constant, while it gradually decreased with increasing potential in the potential region above 0.8 V. When the potential was scanned back, the band intensity of adsorbed water recovered.

Figure 3(I-a) shows the CV of methanol oxidation on the Pt film in 0.1M methanol + 0.1M HClO₄ at a scan rate of 5 mV/s and a flow rate of 10 µl/s. The corresponding mass spectrometric cyclic voltammograms (MSCVs) of CO₂ at m/z = 44 and methylformate at m/z = 60 are shown in Figs. 3(I-b) and 3(I-c), respectively. Before the measurement, to clean the Pt film surface, the potential was stepped to 1.2 V for 5 s and then stepped back to 0.1 V. Methanol oxidation on the Pt film onset at ∼0.55 V, and the oxidation current reached its maximum at ∼0.85 V and then decreased due to the formation of Pt oxides. In the reverse scan, at potentials below 1.0 V, the oxidation current increased again due to the reduction of Pt oxides to form a fresh Pt surface, reached a maximum at ∼0.7 V, and then decreased again with decreasing potential. The formation of CO₂ and methylformate was generally parallel to the Faradaic current of methanol oxidation. However, the current efficiency for CO₂ generation in one potential cycle was ∼80 ± 10%, while the current efficiency for methylformate was only ∼0.1%, suggesting that other intermediates, such as formaldehyde and formic acid, were also likely generated. Unfortunately, formaldehyde hydrate and formic acid are not volatile enough and, thus, cannot be directly detected by DEMS. Formic acid was detected by using a PtPb electrode in the previous publication.¹¹ It was reported previously that methanol oxidation product distribution is affected by the electrode roughness/catalyst loading, diameter of electrode, electrolyte, concentration of methanol, and convection conditions.^5–7,9 The current efficiency of CO₂ generation increases as the Pt roughness and diameters increase, and the flow rate decreases, and weakly adsorbing or low concentrations of strongly adsorbing electrolyte are used since the intermediates have a higher probability of being further oxidized before being swept away.

Simultaneously recorded attenuated total reflection-surface enhanced infrared absorption (ATR-SEIRA) spectra are presented in Figs. 3(III) and 3(IV). Besides water adsorption bands [Figs. 3(III) and S2], infrared bands for adsorbed CO (CO_ad), formate (HCOO⁻_ad), and formic acid (HCOOH_ad) were also observed. Two ν(C≡O) bands at around 2066 and 1840 cm⁻¹ were assigned to linearly bonded CO (CO_ad,L) and bridge-bonded CO (CO_ad,B), respectively. CO_ad,B exhibited a relatively higher intensity at low potentials (in the hydrogen region) when compared to CO_ad,L. CO_ad,L and CO_ad,B can be dynamically interchanged, and their relative coverages are dependent on the coverage of CO and potentials.²⁴ The infrared band at 1323 cm⁻¹ was assigned to the symmetrically OCO stretching vibration of HCOO⁻_ad [s−ν(OCO)]. HCOO⁻ was bridge-bonded to Pt through two oxygen atoms to form bidentate formate.¹³ A weak infrared band at around 1438 cm⁻¹ could be assigned to the C–H bending vibration of HCOOH_ad [δ(C–H)], rather than HCOO⁻_ad, since it was not parallel to the HCOO⁻_ad band at 1323 cm⁻¹ and the C–H bending band of HCOO⁻_ad occurred at around 1405 cm⁻¹ (vide postea). The infrared band intensities of CO_ad,L, HCOOH_ad, HCOO⁻_ad, and H₂O_ad are plotted vs potential in Fig. 3(II). In the hydrogen region, the Pt surface was covered by adsorbed H, which suppressed methanol decomposition, so the CO coverage was very low, as indicated by the small CO_ad,L band. It should be noted that methanol might not decompose, to form adsorbed CO, at a potential of 0.1 V or that its decomposition is very slow. A small amount of CO_ad, observed at 0.1 V, was formed during the potential step from 1.2 to 0.1 V. Methanol decomposition to form adsorbed CO is a surface structure sensitive process and mainly takes place at Pt(110) facets or steps.²⁵ At potentials below 0.2 V, the Pt(110) sites are largely occupied with adsorbed H, and the high H coverage would inhibit the methanol decomposition. As the potential was scanned positively, the intensity of the CO_ad,L band increased, suggesting that the CO coverage increased. The CO coverage reached a maximum at ∼0.55 V and then decreased, due to the oxidation of adsorbed CO. The onset potential of CO_ad oxidation was consistent with the onset potential of methanol oxidation, implying that adsorbed CO poisoned the Pt surface and, thus, suppressed methanol oxidation at potentials below 0.55 V. On the reserve scan, the CO_ad,L intensity increased again, reached its maximum at ∼0.5 V, and then remained relatively constant with further decreasing potential. Adsorbed H cannot displace adsorbed CO, and thus, H adsorption was suppressed in the hydrogen region. In the hydrogen region, the intensity of the CO_ad,L band slightly decreased, in parallel with an increase in the CO_ad,B band intensity. This might be due to a dynamic interexchange between CO_ad,L and CO_ad,B. CO molecules prefer to adsorb at bridge sites at low potentials and atop sites at high potentials.^24,26 It is generally accepted that methanol oxidation proceeds through a dual pathway mechanism. In one pathway, methanol decomposes to form CO_ad, which is further oxidized to form CO₂. In another pathway, methanol is directly oxidized to form formaldehyde and formic acid, and the extent of further oxidation of these soluble intermediates depends on the convection of electrolyte, electrolyte type and concentration, concentration of methanol, electrode surface roughness/catalyst loading, and even electrode diameter.^5–7,9 As CO_ad started to be oxidized, formate began to adsorb on the Pt surface. Its coverage reached a maximum at around 0.7 V and then decreased with further increasing potential. We found that the maximum in the Faradaic current of methanol oxidation and the maximum formate intensity did not occur at the same potential. This is different from the findings of the Osawa et al.^12,16 In previous studies, Osawa et al. claimed that adsorbed formate was an active intermediate for methanol oxidation on Pt since they observed that the infrared intensity of adsorbed formate was proportional to the Faradaic current for methanol oxidation on a Pt film in 0.1M HClO₄ + 0.5M methanol. However, we did not observe this phenomenon in 0.1M HClO₄ + 0.1M methanol. The adsorbed formate and formic acid in the electrolyte might be in equilibrium, and adsorbed formate is likely a spectator species. In Osawa’s works, formic acid concentration (or the formation rate of formic acid) at the Pt electrode surface might be accidently parallel to Faradaic current of methanol oxidation so that the formate coverage might also accidently follow the Faradaic current. We found that in the dual thin-layer flow cell, the intensity of the formate band was much lower than that in the stationary cell during methanol oxidation due to a lowering of the HCOOH concentration at the electrode surface by convection, though the Faradaic current was similar. Our conclusions are consistent with the results from Behm and Cai et al.¹⁷ In the reverse scan, the infrared intensity of adsorbed formate increased again and reached its maximum at ∼0.85 V and then decreased with further decreasing potential. The infrared band of adsorbed formate disappeared when the CO coverage reached its maximum at ∼0.5 V. It seems that CO adsorption displaced adsorbed formate from the Pt surface.

Similar to formate, formic acid also began to adsorb on the Pt surface as CO_ad started to be oxidized. The band intensity of adsorbed formic acid reached a maximum at ∼0.85 V in the positive-going scan, consistent with the Faradaic current peak. In the reverse scan, the intensity of adsorbed formic acid decreased slowly with a decrease in potential and then dramatically at potentials below 0.55 V, and reached zero at potentials below 0.2 V.

If we analyze the CO_ad,L band closely, we found that it is actually composed of two peaks. The low frequency peak (LFP) can be assigned to CO adsorbed at (low-coordination) defect sites (such as steps, kinks, and vertex), while the high frequency peak (HFP) can be assigned to CO adsorbed at (high-coordination) terrace sites. This assignment can be justified by methanol oxidation on stepped Pt(111) surfaces. Shin and Korzeniewski found that two types of linearly bonded CO formed during methanol oxidation on a corrugated single crystal Pt(355) electrode by using FTIR spectroscopy and assigned them to CO adsorbed at terraces and step edges, respectively.²⁷ As the potential increased from 0.1 V, the LFP of CO_ad,L band was observed earlier than the HFP of the CO_ad,L band, suggesting that methanol is more easily decomposed to form CO_ad,L at defect sites [likely Pt(110) sites] than at terrace sites.⁴ This is consistent with the fact that Pt(110) steps in Pt(111) terraces significantly enhance methanol decomposition into adsorbed CO.²⁵ With increasing potential, the LFP intensity of the CO_ad,L band increased first. When the LFP of the CO_ad,L band reached its maximum, the HFP of the CO_ad,L band appeared and its intensity increased with a further increase in the potential. At potentials higher than 0.5 V, the HFP intensity of the CO_ad,L band decreased with increasing potential, as CO_ad started to be oxidized. The LFP intensity of the CO_ad,L band started to decrease with an increase in the potential, once the HFP of the CO_ad,L band disappeared. The kinetics of adsorbed CO oxidation is also a surface structure sensitive process, and the initiation site for CO oxidation is generally the lower part of defects or steps.²⁸ It should be noted that the inversion of the direction of the LFP, or bipolar LFP, was observed in the high potential region due to anomalous infrared adsorption on nanostructured Pt films, as has been discussed by several groups.^21,22,29,30 We observed that this anomalous infrared effect occurred only for the LFP rather than HFP of the CO_ad,L band, and H co-adsorption in the hydrogen region in the first positive-going scan mitigated the anomalous feature of the LFP of the CO_ad,L band. In the reverse scan, the LFP intensity of the CO_ad,L band increased with decreasing potential. Then, the HFP intensity of the CO_ad,L band started to increase as the LFP of the CO_ad,L band did not grow anymore.

The Stark tuning rate for the HFP of the CO_ad,L band was about 39 cm⁻¹/V, which is consistent with previously reported values.³¹ However, the LFP of the CO_ad,L band was almost unaffected by potentials. This might be due to the cancellation of different orientations of CO adsorbed on different defect sites.

In another experiment, we used a different electroless Pt plating recipe (0.01M K₂PtCl₆ + 0.67M NH₃ + 0.06M NH₂NH₂) to obtain a smoother Pt film (the second Pt film) and employed a three-electrode glass cell to study methanol oxidation on the Pt film in a stationary solution. The CV of methanol oxidation on the second Pt film and plots of integrated intensities of different infrared bands vs potential are presented in Fig. 4(I). Recorded ATR-SEIRA spectra are shown in Figs. 4(II)–4(V). In general, the ATR-SEIRA spectra of CO_ad,L, CO_ad,B, HCOO⁻_ad, and H₂O_ad are similar to those in Fig. 3. The major differences are that a single linearly bonded CO band was observed, the water band became weaker when compared to that for the first Pt film, and the band of adsorbed formic acid at ∼1438 cm⁻¹ was hard to be observed. However, we observed a broad infrared band at ∼2923 cm⁻¹ in the high potential region, which had a Stark tuning rate of ∼58 cm⁻¹/V and can be assigned to the stretching vibration of the C–H bond [ν(C–H)] in adsorbed formic acid [Fig. 4(IV)].³² Its integrated intensity is plotted vs potential in Fig. 4(I-f). This band started to appear, as the adsorbed CO was oxidized. Its intensity increased with a decrease in the water band intensity. In the first positive-going scan, this broad band can also be observed at potentials below 0.5 V and can be assigned to the ν(C–H) of adsorbed methanol. In addition, we also observed a very weak infrared band at ∼2500 cm⁻¹ in the low potential region with a Stark tuning rate of 56 cm⁻¹/V. Its integrated intensity is plotted vs potential in Fig. 4(I-g). The variation of this band was similar to that of adsorbed CO, so this adsorbed species could be the intermediate for methanol decomposition into adsorbed CO. Therefore, we temporally assign it to the C–H stretching vibration of adsorbed CHO_ad or CH_xOH_ad. In addition, a weak infrared band was also observed at around 1405 cm⁻¹ [Fig. 4(III)], which was parallel to the symmetrical OCO stretching band of HCOO⁻_ad and, thus, can be assigned to the C–H bending vibration of adsorbed formate. The intensity of the HOH bending band of adsorbed water at around 1630 cm⁻¹, relative to the bands of adsorbed CO, became smaller for the second smooth Pt film, when compared to the first rough Pt film. The intensity of the C–H bending band of adsorbed formic acid at around 1438 cm⁻¹ was also significantly diminished and hard to be observed, when compared to the first rough Pt film. It should be noted that in the dual thin-layer flow cell, we obtained similar infrared spectra, so the effects of convection are minimal. The differences in the spectra between these two Pt films are attributed to their different structures. Later, we will discuss the effects of Pt film structure on adsorbed CO bands, based on simulations.

Since formic acid is one of the intermediates generated during methanol oxidation on Pt and adsorbed formate was also proposed to be the active intermediate for formic acid oxidation in the direct pathway,^12,16 we revisited the oxidation of formic acid on Pt and studied the effects of formic acid concentration and Pb modification on the infrared bands of adsorbed CO and formate.

Figure 5 presents simultaneously recorded DEMS data and ATR-SEIRA spectra for formic acid oxidation on a Pt film in 0.1M HCOOH + 0.1M HClO₄. Before the measurement, the potential was stepped to 1.2 V for 5 s and then stepped back to 0.1 V to clean the Pt film surface. Figure 5(I-a) presents the CV of formic acid oxidation on a Pt film at a scan rate of 5 mV/s and an electrolyte flow rate of 10 µl/s, and the corresponding MSCV of CO₂ at m/z = 44 is shown in Fig. 5(I-b). Figures 5(III), 5(IV), and S3 show the infrared bands of adsorbed CO, formic acid, formate, and water, and their intensities are plotted vs potential in Fig. 5(II). The oxidation of formic acid onset at around 0.4 V, and the Faradaic current and mass spectrometric signals of CO₂ increased slowly with increasing potential until ∼0.65 V and then increased rapidly due to the oxidative removal of adsorbed CO [Fig. 5(II-a)]. After reaching a maximum at ∼0.82 V, the formic acid oxidation current gradually decreased with a further increase in potential due to the formation of Pt oxides and then increased slowly again as the potential was beyond 1.0 V. In the negative-going scan, the oxidation current increased again due to the reduction of Pt oxides at potentials below 0.95 V. The oxidation current reached a maximum plateau between 0.5 and 0.7 V and then decreased with a further decrease in potential. It should be noted that several spikes were observed in the CV during the negative-going scan and are ascribed to the negative differential resistance in electrolytes of low conductivity, as has been discussed by Krischer and Varela.³³ These spikes were not seen at high electrolyte concentration (such as 0.5M HClO₄) due to an increase in the conductivity of the electrolyte.

During formic acid oxidation, CO_ad,L, CO_ad,B, HCOO⁻_ad, and HCOOH_ad infrared bands were also observed. Similar to the case of methanol oxidation, an anomalous bipolar or inversion feature for the LFP of the CO_ad,L band was also observed at potentials higher than 0.4 V. In contrast, the bipolar or inversion feature for the LFP of the CO_ad,L band disappeared in the hydrogen region. This is likely due to the effect of co-adsorbed H. In contrast to methanol oxidation, more CO_ad, as indicated by the higher intensity of the CO_ad,L band, was observed in the hydrogen region for HCOOH oxidation on Pt, suggesting that HCOOH is more easily decomposed to form CO than methanol on a hydrogen-covered Pt surface. The CO coverage increased with increasing potential until 0.4 V and then decreased with a further increase in potential due to the oxidative removal of adsorbed CO. In the negative-going scan, no adsorbed CO was observed on Pt at potentials higher than 0.4 V. Adsorbed CO started to form at 0.4 V in the negative-going scan, and the linearly bonded CO band intensity reached a maximum at ∼0.2 V and then slightly decreased due to the interexchange between CO_ad,L and CO_ad,B. This observation suggests that HCOOH decomposition to form adsorbed CO might occur only on hydrogen-covered Pt surfaces in the hydrogen region. HCOOH might first be reduced to CHO_ad by adsorbed H and then decomposed into adsorbed CO.³⁴ This is supported by the fact that the infrared band of CHO_ad at around 2500 cm⁻¹ was also observed for the HCOOH oxidation on the second Pt film in the low potential regions [Fig. S4(V)]. HCOOH dehydration to form CO_ad and direct oxidation to CO₂ are also surface structure sensitive. HCOOH cannot be dehydrated to form CO_ad on an ideal Pt(111) surface, while HCOOH dehydration to form CO_ad takes place on Pt(110) and Pt(100) in the low potential region with a maximum rate around the potential of zero total charge.³⁵ Thus, on the Pt film, HCOOH decomposition to form CO should mainly occur at Pt(110) and Pt(100) facets or defects. Similar to the case of methanol oxidation, the LFP frequency of the CO_ad,L band was almost unaffected by the applied potential. The Stark tuning rate for the HFP of the CO_ad,L band was about 38 cm⁻¹/V.

From Fig. 5(II-b), one can observe that the formation of HCOO⁻_ad onsets at ∼0.35 V in the positive-going scan and parallels the onset of CO_ad oxidation. The intensity of the HCOO⁻_ad band reached its maximum at around 0.7 V, where the CO_ad,L band disappeared, and then decreased due to the inhibition of adsorbed oxygen-containing species. This, again, suggests that formate can adsorb on Pt only at CO-free sites. The maximum intensity of the HCOO⁻_ad band did not occur at the potential where the maximum of Faradaic current appeared. In the negative-going scan, both formate and water adsorption were significantly suppressed, and they onset at ∼0.5 V, which is much more negative than their onset potential of 1.05 V for methanol oxidation. Thus, the presence of HCOOH in solution significantly inhibited formate and water adsorption. The C–H bending band of HCOOH_ad occurred at around 1438 cm⁻¹, especially in the negative-going scan, suggesting that HCOOH is also adsorbed on the Pt film, and its adsorption suppressed formate and water adsorption. The maximum intensity of the formate band appeared at ∼0.4 V in the negative-going scan and then decreased with a decrease in potential due to the onset of CO_ad formation. The band intensity of adsorbed formate was also much smaller than that in the positive-going scan, while the Faradaic current was higher than that in the positive-going scan. Their profiles did not match each other. This is, again, at odds with the findings of Osawa et al.^13,16

As we lowered the concentration of HCOOH from 0.1 to 0.02M, we observed some interesting phenomena. The simultaneously recorded DEMS data and ATR-SEIRA spectra for HCOOH oxidation on the Pt film in 0.02M HCOOH + 0.1M HClO₄ are presented in Fig. 6. The intensity of the formate band [Fig. 6(II-b)] and the water band [Figs. 6(II-c) and S5] significantly increased, when compared to the case of 0.1M HCOOH solution [Figs. 5(II-c), 5(II-d), and S3], though the Faradaic current was smaller than that in 0.1M HCOOH solution. This again suggests that HCOOH also adsorbed on the Pt film and competed with formate, as well as water, for Pt sites. The C–H bending band of HCOOH_ad was significantly smaller in 0.02M HCOOH than that in 0.1M HCOOH, suggesting that the coverage of HCOOH_ad decreased with a decrease in HCOOH concentration. In contrast, the coverage of formate increased with a decrease in HCOOH concentration. Previously, we studied the formate oxidation on Pt in alkaline media and found that its oxidation current is much smaller than that in acidic acid,³⁶ and it might need to convert into HCOOH before being able to be oxidized. Therefore, we believe that formate is not an active intermediate but rather is an inactive adsorbed ion.

It has been reported for several decades that Pb modified Pt can significantly enhance HCOOH oxidation. However, the enhancement mechanism is still not fully understood.^2,7,15,37 Therefore, we also revisited the effect of Pb modification on HCOOH oxidation on Pt.

Figure 7 presents the simultaneously recorded DEMS data and ATR-SEIRA spectra for HCOOH oxidation on a Pt film in 1 × 10⁻⁵M Pd(NO₃)₂ + 0.02M HCOOH + 0.1M HClO₄. The infrared spectra of adsorbed water are shown in Fig. S6. Figure 7(I-a) shows the CV of HCOOH oxidation on a Pb modified Pt film (Pt/Pb). At low potentials, Pb can be underpotentially deposited on Pt. As a result, HCOOH oxidation on Pt/Pb onset at ∼0.15 V, over 250 mV more negative than that on pure Pt (Fig. S7), and the positive-going scan and the negative-going scan profiles were almost superimposable. The LFP of the CO_ad,L band was suppressed by Pb modification, while the HFP of the CO_ad,L band became sharper, and its intensity increased [Fig. 7(III)]. This suggests that Pb prefers to occupy the low-coordinated defect sites rather than the high-coordinated terrace sites. Pb was likely deposited in the lower part of defects or steps.^38,39 Partially positively charged Pb adatoms can affect neighboring CO adsorption and also the electric field at the electrode so that a much larger Stark tuning rate of 68 cm⁻¹/V was observed. The maximum band intensity of adsorbed formate was also suppressed by Pb modification, though the maximum Faradaic current slightly increased. In addition, the C–H bending band of adsorbed HCOOH was still able to be observed. Pb modification also enhanced the oxidation of adsorbed CO, as indicated by a decrease in the CO_ad,L band intensity at less positive potentials. Although terrace sites on the Pt film were significantly covered by adsorbed CO at potentials below 0.4 V, a large Faradaic current was still observed. This suggests that HCOOH oxidation on Pt/Pb mainly takes place at Pb modified low-coordinated defect sites through the direct pathway.

The third-body effect has been proposed to account for the enhancement of HCOOH oxidation on Pt/Pb.^2,7,15,37 HCOOH decomposition to form CO might need at least 2–3 neighboring Pt sites. Pb modification can reduce the density of ensembles with 2–3 neighboring Pt and, thus, suppress the formation of adsorbed CO. As mentioned before, adsorbed HCOOH might need adsorbed H to form CHO_ad, which is subsequently decomposed to form adsorbed CO, since CO formation takes place only in the hydrogen region.³⁴ This poisoning process can take place only at the low part of defects [likely Pt(110) and Pt(100) steps in Pt(111) terraces].^38,39 Pb modification can suppress H adsorption at the low part of these defects or steps, where Pb prefers to deposit, and thus mitigates the formation of adsorbed CO. As a result, more free Pt sites are available for HCOOH oxidation through the direct pathway. Besides the third-body effect, electronic effects might also play a role in HCOOH oxidation on Pt/Pb. Pb modification might negatively shift the point of zero charge (PZC) of Pt, and thus, Pt/Pb has more positive charge on its surface,³⁹ which could enhance C–H bond splitting and also lower the coverage of adsorbed H. The suppression of adsorbed H can mitigate the formation of adsorbed CO, and the enhancement of C–H bond splitting, which is the rate determining step, can also boost HCOOH oxidation kinetics through the direct pathway.

In contrast to a smooth metal surface, nanostructured metal thin films exhibit some unusual optical properties, such as surface enhanced Raman scattering (SERS) and surface enhanced infrared absorption (SEIRA).^21,30,40,41 SEIRA often leads to anomalous infrared spectra with bipolar or inverted bands rather than normal spectra. This is particularly the case for the spectra of CO adsorbed on nanostructured transition metal surfaces.^21,30,41 Fresnel equations and Bruggeman effective medium theory have been employed to simulate anomalous spectra of CO adsorbed on nanostructured Pt film or Pt nanoparticles.^21,29,30 McPherson et al. found two linearly bonded CO bands in the main oxidation region.²² In this work, we also observed two peaks for linearly bonded CO during methanol and formic oxidation on the Pt film deposited on a Si prism, and the LFP and HFP of CO_ad,L exhibited anomalous (bipolar or inversion) and normal features, respectively. To better understand this special phenomenon, we developed a simple model in which we assumed the Pt film layer to be represented by heterogeneously dispersed Pt nanoparticles with two types of domains: one type with densely packed small Pt nanoparticles (smooth domains, dominated by terrace sites) and the other with loosely packed Pt nanoparticle agglomerates (rough domains, dominated by defect sites) immersed in a mixture of water and CO (Fig. 8).

The dielectric constant of the mixed phase of water and CO is given by Eqs. (1)–(3), and the Lorentzian function was used to model the contribution of CO to ɛ_mix,^21,22

where ɛ_mix is the dielectric constant of the mixed phase of water and CO, and ɛ_r and ɛ_i denote the real and imaginary parts of ɛ_mix, respectively. ν is the infrared frequency (cm⁻¹), ν_o is the center frequency of the CO band (cm⁻¹), $nH2O$ is the refractive index of water in the frequency between 1900 and 2200 cm⁻¹ (1.316),²² and B and γ represent the intensity and the full width at half maximum of the CO band, respectively.²¹ 

The Bruggeman effective medium theory was used to calculate the effective dielectric constant (ɛ_eff) of the Pt film layer (including Pt nanoparticles and a mixture of water and CO),^21,22 which is given by the following equation:

where the dimension D equals to 3 and the volume fraction of Pt (f) varies from 0 to 1. f = 0 means the mixed phase of water and CO without Pt nanoparticles, f = 1 denotes a densely packed smooth Pt surface without the mixed phase of water and CO, and 0 < f < 1 refers to Pt nanoparticles immersed in the mixed phase of water and CO. It should be noted that in Eq. (4), the Pt particle size was assumed to be much smaller than the infrared wavelengths, spherical, identical, and evenly distributed in a dielectric medium, and nanoparticle size effects were not taken into account.

The refractive index of the Pt film layer was obtained with Eq. (5), and the reflection coefficients for s-polarized and p-polarized incident infrared light were calculated with Eqs. (6) and (7), respectively. The reflectance was obtained via Eq. (8). The surface of the Pt film was assumed to be heterogeneously and consisted of two different types of domains. As mentioned before, domain 1 had a relatively high density of Pt nanoparticles, while domain 2 possessed a relatively low density of Pt nanoparticles. The total reflectance (R_total) for two different types of domains is given by Eq. (9). The absorbance (A) of CO was finally obtained with Eq. (10),

where $ňeff$ is the refractive index of the Pt film layer, r_s and r_p are the reflection coefficients for s- and p-polarized infrared, respectively. The refractive index of Si (n_Si) is 3.42.⁴² θ_i is the incident angle of the infrared radiation. R_o is referred to the reflectance for the background spectrum without CO present.

Figure S8 shows the calculated infrared spectra of CO adsorbed on the Pt film with homogeneously dispersed Pt nanoparticles. Pt nanoparticles were homogeneously distributed in the mixed phase of CO and water, so only one peak for the CO_ad,L band was obtained. As the volume fraction of Pt (f), i.e., the concentration of Pt nanoparticles, increased from 0.01 to 0.2, the infrared band of CO_ad,L was in a normal direction, and its intensity increased due to surface enhancement effects. As f increased to about 0.26, the infrared band of CO_ad,L started to exhibit a broad bipolar shape. As f reached about 0.3, an inverted symmetric CO infrared band was observed. When f was between 0.34 and 0.39, the infrared band of CO_ad,L became bipolar again, but slender than that for f = 0.26. For f higher than 0.4, the infrared band of CO_ad,L became normal again, and its shape was dependent on the value of f. For example, at f = 0.42, a symmetric CO band was observed again.

Figure 9(a) presents a representative calculated infrared spectrum of CO adsorbed on a heterogeneous Pt film, which was assumed to consist of two different types of domains with the size being much larger than infrared wavelength. Domain 1 had a relatively large volume fraction of Pt (f₁ = 0.42), while domain 2 had a relatively small volume fraction of Pt (f₂ = 0.37). The frequencies of adsorbed CO on these two domains were 2070 and 2050 cm⁻¹, respectively. As a result, two separate peaks for the CO_ad,L band with different shapes were obtained. Figures 9(b) and 9(c) show the decoupled CO bands for domains 1 and 2, respectively. Figure 9(a) simulated quite well the linearly bonded CO on the Pt film during methanol and formic acid oxidation at ∼0.5 V in the positive-going scan [Figs. 3(III), 5(III), and 6(III)]. At ∼0.5 V in the positive-going scan, the maximum CO coverage was achieved so that both domains of the Pt film were occupied with CO. For domain 1, the volume fraction of Pt was large (e.g., f₁ = 0.42), so a normal CO band was observed [Fig. 9(b)], while for domain 2, the volume fraction of Pt was small (e.g., f₂ = 0.37), so a bipolar CO band developed [Fig. 9(c)]. The simulations suggest that the observed linearly bonded CO band might be composed of these two different components, and the size of these two types of domains might be close to or even larger than the infrared wavelength scale.

In the positive-going scan, at low potentials such as in the hydrogen region, adsorbed hydrogen suppressed the decomposition of methanol and formic acid to form CO, and thus, only small amount of CO was present at the Pt film, likely on the rough domains (domain 2). In this case, the smooth domains (domain 1) might be free of CO and the rough domains (domain 2) partially occupied by CO, which likely prefers to adsorb at less rough sites of the rough domains due to the repulsion of adsorbed hydrogen. These less rough areas of the whole rough domains with CO covered have a larger volume fraction of Pt than the whole rough domains, for example, f₂ ≥ 0.4 so that the simulated CO_ad,L band exhibited a normal asymmetric or symmetric shape (Fig. S8). This is consistent with the fact that a normal broad CO band was observed in the hydrogen region for positive-going scans [Figs. 3(III), 5(III), and 6(III)].

At potentials between 0.6 and 1.0 V, in the positive-going scan or in the negative-going scan for methanol oxidation on the Pt film, the CO coverage became smaller since the adsorbed CO was partially oxidized. In this case, we clearly observed the bipolar or even inversion feature for the CO_ad,L band [Fig. 3(III)]. Figure S9 presented the simulated CO_ad,L band at a CO coverage of 0.1 and 0.01 on the Pt film, respectively. In the simulations, as the CO coverage decreased from 1 [Fig. 9(a)] to 1/10 or 1/00 value (Fig. S9), in addition to a decrease in the intensity of both the peaks, the bipolar or inversion feature for the LFP of the CO_ad,L band also became more discernible. If domain 1 had a lower coverage of CO_ad than domain 2, i.e., CO adsorbed on domain 1 is more easily oxidized or is formed more slowly than that on domain 2, the HFP of the CO_ad,L band can be even significantly smaller or even disappear [Figs. S9(c) and S9(d)].

The simulations in Fig. S10 also suggested that if we change the incident angle, we might see a change in the double CO_ad,L bands with varying the incident angle since the CO_ad,L band intensity for different volume fractions of Pt varies with the incident angle disproportionally.

The CO_ad,L band appeared only in the hydrogen region in the negative-going scan for formic acid oxidation [Figs. 5(III) and 6(III)], and the LFP of the CO_ad,L band had a normal shape. In contrast, the LFP of the CO_ad,L band had a bipolar or inversion shape at potentials higher than 0.4 V in the positive-going scan. This might be due to the effect of co-adsorption of H atoms. As mentioned before, co-adsorbed H atoms might prefer to occupy the most rough areas of rough domains so that the less rough areas (with a larger volume fraction of Pt) might be involved in CO adsorption and give more contribution to the LFP of the CO_ad,L band. In our simulations, as f₂ increased from 0.3 to >0.4, the LFP of the CO_ad,L band could change from a bipolar/inversion shape to a normal shape (Fig. 10).

As the Pt film is modified by adsorbed Pb, some defect sites (most likely defect sites in the rough domains) of the Pt film will be blocked by Pb, so the CO coverage on the rough domains decreases. As a result, the LFP intensity of the CO_ad,L band was significantly reduced. The intensity decrease in the negative direction of the bipolar peak mitigated the negative effect on the HFP of the CO_ad,L band, and thus, the intensity of the HFP of the CO_ad,L band became higher than that without Pb modification (Fig. S11).

According to our simulations, the volume fraction of Pt in the catalyst layer can affect the shape of the observed CO band. Different preparation methods of the Pt film can affect the morphology of the Pt film and, thus, will significantly affect the adsorbed CO band. If we want to obtain a normal infrared band of CO_ad,L with the largest intensity, the volume fraction of Pt in the catalyst layer must be ∼0.42. Assuming that the amount of Pt deposit on the Si prism remains constant, if the Pt film is very smooth, the volume fraction of Pt will be larger than 0.4. As a result, the resistance of the Pt film will be low, and the normal infrared bands will be observed. If the Pt film is rough and thin, i.e., the volume fraction of Pt is between 0.25 and 0.4, the resistance of the Pt film will be large, and the infrared bands will become bipolar or inverse. If the Pt film is too thick, it will be easily detached from the Si prism during potential cycling. In contrast, if the Pt film is very thin, Pt nanoparticles are less connected so that the resistance of the Pt film is large. As a result, a large IR drop will significantly affect electrochemical reactions. Moreover, an inverted CO band will occur. When a different electroless platinum plating recipe (0.01M K₂PtCl₆ + 0.67M NH₃ + 0.06M NH₂NH₂) was used, we could see only one peak for the linearly bonded CO band [Figs. 4(II) and S4(II)]. Using this new recipe, as we increased the Pt film thickness from 50 to 70 nm, we observed that the linearly bound CO band changed from the inverse to the normal direction. With a further increase in the Pt film thickness to 90 nm, the intensity of the linearly bonded CO band further increased. It is suggested that the electroless Pt plating recipe and the thickness of the Pt film can affect the shape and intensity of CO bands. Therefore, there must be an optimum Pt film thickness and morphology for studying the CO adsorption band, and the search for it is still in progress. In addition, the infrared incident angle also must be considered.

Methanol and formic acid electro-oxidation on a Pt film in 0.1M HClO₄ has been studied by a combined spectroscopic system of DEMS and ATR-SEIRAS under well-defined flow conditions. The effects of HCOOH concentration and Pb modification on HCOOH oxidation at Pt have also been investigated. CO₂ and methylformate were detected as volatile products by DEMS for methanol oxidation on Pt, while linearly bonded CO, bridge-bonded CO, adsorbed formate, adsorbed formic acid, and adsorbed CHO were identified as adsorbed species on Pt by ATR-SEIRAS. Less than 100% of current efficiency of CO₂ for methanol oxidation on Pt suggested that formaldehyde and formic acid were also formed as non-volatile intermediates. As for formic acid oxidation on Pt, besides CO₂ as the final product, linearly bonded CO, bridge-bonded CO, adsorbed CHO, adsorbed formic acid, and adsorbed formate were also identified as adsorbed species. HCOOH can also adsorb on Pt, as indicated by an infrared band at around 1438 cm⁻¹ and/or 2923 cm⁻¹. An increase in the HCOOH concentration and Pb modification of Pt suppressed formate adsorption, while the Faradaic current of HCOOH oxidation increased. Our data did not support the viewpoint that adsorbed formate is an active intermediate for methanol and formic acid oxidation on Pt. Pb modification significantly enhanced HCOOH oxidation on Pt due to the third-body effect and electronic effects.

The thickness and morphology of the Pt film can affect the shape and peak direction of infrared bands. It was observed that the infrared band of CO_ad,L for the first Pt film was composed of two peaks, and the low frequency peak had a bipolar or inversion shape, while the high frequency one exhibited a normal peak. Our simulations suggest that the low frequency peak and high frequency peak might be ascribed to CO adsorbed on rough domains of the Pt film with dominant defect sites and smooth domains of the Pt film with dominant terrace sites, respectively.

See the supplementary material for infrared bands of adsorbed water, infrared spectra for HCOOH oxidation in 0.1M HCOOH + 0.1M HClO₄ on the second Pt film in the dual thin-layer flow cell, simulated CO_ad,L bands, and comparison of DEMS data for HCOOH oxidation on the Pt film in 0.02M HCOOH + 0.1M HClO₄ with and without 1 × 10⁻⁵M Pd(NO₃)₂.

This work was supported by the Air Force Office of Scientific Research (Award No. FA9550-18-1-0420). This work made use of the Cornell Center for Materials Research Shared Facilities, which are supported through the NSF MRSEC program (Grant No. DMR-1719875).

We have no conflicts of interest to disclose.

The data that support the findings of this study are available within this article and its supplementary material.