High-Throughput Correlative Electrochemistry−Microscopy at a Transmission Electron Microscopy (TEM) Grid Electrode

.

Over the past three decades, science has been impacted massively by the revolution in nanotechnology. For example, nanoparticles (NPs) have found a plethora of technical applications, including in (electro)catalysis, sensing, spectroscopy, and (bio)medicine. [1][2] NPs are known to possess strongly structuredependent reactivity, meaning that at the single-NP level there can be large functional differences between even superficially similar NPs due to minute variations in size, surface faceting, defects etc. [2][3] For this reason, there is a need for techniques capable of producing single NPs and screening their properties at the single-entity level. [4][5][6] Scanning electrochemical cell microscopy (SECCM) 7 is a powerful tool in single-entity studies, in which the meniscus cell protruding from an electrolyte-filled micropipet (or nanopipet) probe is used to electrochemically interrogate a single or small population of supported NPs within an ensemble. 4,[8][9][10][11] For example, in a recent study, SECCM was deployed as a highthroughput screening method to probe the heterogeneous response of individual LiMn2O4 particles (a Li-ion battery cathode material), revealing a diverse library of responses within a family of superficially similar single-entities. 12 The meniscus cell configuration of SECCM also enables local decoration (i.e., micro-or nano-fabrication) at electrodes surfaces, e.g., with metal NPs, [13][14] polymer nanostructures 15 and graphene microwires. 16 Here, we present a platform whereby PtNP "microensembles" are locally electrodeposited under a series of different (tunable) conditions and subsequently characterized ex-situ with aberration-corrected high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM), allowing the relationship between applied potential (Eapp) and Pt deposit morphology (i.e., NP size, shape and density) to be studied in a detailed high-throughput manner.
Pt plays an essential role as a catalyst in many industrial applications and green energy technology (e.g., autocatalysts and fuel cells). Due its high cost and scarcity, considerable research efforts have focused on developing catalysts with high massspecific activities (i.e., low Pt loadings), usually achieved by engineering Pt-based nanomaterials with optimal surface-specific activities and high surface-area-to-volume ratios. 17 Among a range of synthetic approaches, 2 electrodeposition is attractive for producing PtNP electrocatalysts, as it generates particles that grow directly on the conductive support to ensure good electrical contact; can be performed with simple equipment under (near) ambient conditions; is scalable and; is relatively lowcost. Tuning of experimental parameters (e.g., Eapp) offers some control over NP morphology, 18 however, achieving a narrow size distribution is still a challenge, 19 at least partly due to an incomplete mechanistic (atomistic) understanding of the early stages of electrochemical nucleation and growth. 20 Classical models of nucleation and growth describe immobile nuclei that need to reach a certain critical size to become stable on a low-energy support, which then grow radially by direct addition of atoms through electroreduction (i.e., Volmer−Weber 3D island growth). 21 Recently, through the use of various techniques, including HAADF-STEM, 20, 22-24 ultramicroelectrodes 25 and SECCM 13-14 a more complex electrochemical aggregative growth mechanism has been proposed to explain the earliest stages of nucleation and growth, 20 where the formation, aggregation and coalescence of discrete NPs are initially the dominant processes driving the formation of stable structures during electrodeposition. The details of this generalized model are still under discussion, 26 but it has been applied to a number of metal systems, [13][14]24 including Pt, 23 which undergoes a relatively complex electrodeposition process due to sluggish nucleation/growth kinetics and competing electrochemical processes (e.g., hydrogen evolution reaction, HER). Here, we investigate this complex system to demonstrate the great versatility and strength of a high-throughput electrochemistry−microscopy platform, which enables Pt deposition to be probed with singleatom sensitivity for a range of different conditions in each single experiment, providing a wealth of experimental data.

EXPERIMENTAL SECTION
Details on the chemical reagents and electrode materials; electrochemical measurements; STEM imaging and; data processing and analysis procedures, are included in the Supporting Information (SI), Section S1. SECCM was performed on a home-built scanning electrochemical probe microscopy workstation, 27-28 detailed in Section S1 and shown schematically in Figure 1a. During operation, a micropipet (or nanopipet) probe filled with electrolyte solution and equipped with a quasi-reference counter electrode (QRCE), was used to perform local electrochemistry on a carbon-coated Au TEM grid (CCTG) working electrode. To facilitate the straightforward use of environmental control (vide infra), and subsequent microscopy analysis, probe positioning was achieved through in-situ topographical mapping, as shown in Figure 1b.
After topographical mapping, electrochemistry (e.g., electrodeposition) was performed on the free-standing carbon areas of the CCTG. The micropipet probe was of diameter 1.6 µm (Figure S1) and filled with a solution of K2[PtCl4] (Pt precursor salt, 1 mM) and HCl (supporting electrolyte, 100 mM). A Pt wire inserted into the back of the micropipet served as a QRCE, meaning all potentials are referenced to the Pt(II)/Pt(0) process (measured to have a stable potential of 0.55 ± 0.01 V vs. Ag/AgCl, 3.4 M KCl) and can thus be directly related to overpotential (ƞ). Oxygen (O2) was excluded from the SECCM droplet cell through the use of an argon-purged environmental cell, the designs of which are included in the SI, Section S2.
Following electrochemistry, the CCTG was "marked" by pushing the micropipet tip through the carbon film in adjacent unused squares (M1 -M3, Figure 1b

RESULTS AND DISCUSSION
Local electrochemistry on a TEM grid electrode. Prior to use as a support electrode for electrodeposition, the performance of the CCTG as a general electrode was benchmarked by investigating two rapid, mechanistically simple outer-sphere electron transfer processes: [Ru(NH3)6] 2+/3+ and FcDM 0/+ (where FcDM = 1,1′-ferrocenedimethanol), 29 shown in Figure 1d. Due to the conical geometry of the pipet probes used in SECCM, masstransport is predominantly governed by (quasi-)radial diffusion, allowing (near) steady-state conditions to be established on the ms timescale (scan rate, υ < 1 V s −1 ). 27,30 Thus, the voltammograms in Figure 1d are sigmoidal-shaped, with a mass-transport limited current (ilim) that is ca. 10% of that expected at a conventional inlaid disc electrode of the same diameter as the used nanopipet probe (ca. 150 nm). 27 The steady-state voltammograms measured on the CCTG (black curves) are indistinguishable from those from a conventional metal (Au) electrode (red curves), and satisfy Tomes criterion of reversibility (i.e., |E3/4 -E1/4| ≈ 56 mV, where E3/4 and E1/4 are the quarter-and three quarter-wave potentials, respectively). 29 This confirms that the free-standing carbon areas of the CCTGs are able to support rapid electron-transfer within the Eapp range explored during Pt electrodeposition [i.e., ca. −0.05 to −0.95 V vs. Pt(II)/Pt(0)]. Pt electrodeposition: cyclic voltammetry. Pt electrodeposition on a CCTG was initially investigated using cyclic voltammetry in the SECCM configuration. Shown in Figure 2a, the cyclic voltammograms (CVs) exhibit a non-classical shape expected for nucleation and growth on a low-energy substrate (i.e., CCTG), with no discernible reduction processes observed until Eapp ≈ −0.6 V vs. Pt(II)/Pt(0), where the onset of HER occurs. Noted above, the mass-transport limited current density (jlim) for a diffusion-controlled process in SECCM is ca. 10% of that for the same-sized microdisk electrode (≈2 µm), which for a 2e − process corresponds to ca. −3 mA cm −2 (ilim = −60 pA, detailed in the SI, Section S3). Evidently, no such process is observed, however the occurrence of the HER at Eapp ≈ 0 V vs. RHE (reversible hydrogen electrode) and the increase in jHER with cycling are both indicators that Pt is in fact deposited. Thus, it is clear that the Pt(II)/Pt(0) process is not mass-transport controlled, but rather kinetically-limited on the CCTG support. 20 As a point of comparison, the Pt(II)/Pt(0) process was further investigated in a conventional 3-electrode format on a glassy carbon (GC) macroelectrode, which is a widely used low energy carbon material in bulk electrochemistry and electrodeposition experiments. 29 Shown in Figure 2b, a similar pattern of reactivity is observed at the GC substrate, with relatively small j values initially measured (cycle 1) that grow continuously with potential cycling. Interestingly, j increases with cycling at a much faster rate on GC compared to CCTG suggesting that the kinetics of Pt nucleation is slower on the latter support, giving rise to a significant time lag during electrodeposition. Thus, comparatively more Pt is deposited during potential cycling on GC, resulting in a larger active (Pt) surface area and the appearance of the additional processes of hydrogen adsorption/desorption (symmetrical peaks at ca. −0.49 and −0.58 V) and Pt deposition onto Pt 20, 22-23 (reductive peak at ca. −0.67 V). Note that these processes also occur during potential cycling on CCTGs in the SECCM configuration (Figure 2a), but do not give rise to discernible reduction currents in the CV due to the comparatively small deposited surface area of Pt on this timescale (i.e., the processes cannot be distinguished from the nonfaradaic baseline current, analogous to cycles 1 -5 on GC, Figure 2b). Overall, the CCTG and GC supports behave analogously as supports for Pt electrodeposition, albeit with a significant time lag (i.e., slower kinetics) at the former compared to the latter. Pt electrodeposition: chronoamperometry. Pt electrodeposition was further explored using chronoamperometry, where a series of depositions were carried out on the CCTG electrode by varying Eapp from 0 to −0.9 V vs. Pt(II)/Pt(0) in 0.1 V increments. It is worth reemphasizing that this approach is: (i) highthroughput, i.e., a range of deposition parameters (Eapp and/or deposition time) can be explored in a single experiment; (ii) high-resolution, i.e., the use of a CCTG support enables complementary, co-located structural analysis through STEM with single-atom sensitivity 24 and; (iii) statistically sound, i.e., each individual NP within the microensemble can be counted and characterized. Importantly, although it has been exclusively applied to study electrodeposition herein, this approach would be applicable to any class of material that can be studied by (S)TEM (e.g., supported NPs, thin films etc.).
Chronoamperograms (CAs) obtained in the SECCM configuration on a CCTG are shown in Figure 3a. In general, |j|: increases with decreasing Eapp (increasing ƞ); decays towards zero at Eapp ≥ −0.6 V and; increases with time at Eapp < −0.6 V. The increasing |j| at Eapp < −0.6 V is attributed to the HER [the contribution of Pt(II)/Pt(0) to j is negligible, see Figure 2], which plateaus at longer times, shown in Figure 3b, implying that the active surface area of the deposited Pt initially grows before reaching a stable value. In other words, as the HER is predominantly surface-controlled at Eapp ≤ −0.9 V, j directly indicates the Pt surface area, which reaches a constant value after ca. 4 seconds. Shown in Figure 3c, the charge passed during reduction (Q) increases with decreasing Eapp (increasing ƞ), consistent with Figure 3b and c. In the absence of the competing HER (i.e., Eapp ≥ −0.6 V), Q is a direct measure of the amount of Pt deposited (Figure 3e, inset), which exhibits a relatively weak dependence on Eapp (i.e., Q only increased ca. 18-fold from ƞ = 0.1 to 0.6 V). Note that repeat experiments carried out in adjacent squares confirm that the Pt electrodeposition process was consistent across the CCTG surface, shown in the SI, Section S4.
Again, as a point of comparison, analogous experiments were carried out in a conventional 3-electrode format on a GC macroelectrode, shown in Figure 3d. The aforementioned time lag during Pt electrodeposition at CCTGs compared to GC is clear when comparing Figure 3a and d, where at Eapp = −0.8 and −0.9 V, |j| is ca. one order-of-magnitude lower at the former compared to the latter. At longer times, the CAs obtained at GC (Figure 3e) and CCTGs (Figure 3b) display very similar characteristics (i.e., morphology and j), indicating that while the initial nucleation process(es) are more sluggish on the CCTGs, the growth processes are largely unaffected by the identity of the low energy carbon support.
Classical 3D island growth models of electrodeposition predict a steady increase in the Pt surface area with time, contrary to the results presented in Figure 3b and e. One possible explanation for the constant active surface area would be that Pt electrodeposition completely ceases after ca. 4 seconds. While adsorbed H (Hads) has been shown to the limit growth of Pt on Pt (termed self-terminating growth), 31 this cannot explain inhibition of Pt nucleation on the CCTG (or GC) support itself, and further is inconsistent with our observation of 3D Pt nanostructures (vide infra). In addition, physical blockage of nucleation sites (on the CCTG) by H2 nanobubbles generated during the HER is also unlikely at pH > 0, due to fast gas-exchange across the gas-liquid interface in the SECCM configuration. 32 Another possible explanation is that the formation of Pt through nucleation and growth proceeds in parallel with aggregative growth processes 20, 23 that effectively annihilate Pt surface area. In this scenario, nucleation and growth processes would initially dominate at t < 4 s, before being eventually balanced by aggregative processes leading to coalescence, giving rise to a relatively constant active surface area at t > 4 s. Indeed, the existence of parallel classical nucleation/growth and aggregative growth pathways is supported by our STEM observations, below. Correlative STEM analysis. Following the electrodeposition experiments (Figure 3), the Pt "microensembles" on the CCTG substrate were imaged ex-situ with HAADF-STEM, which is a powerful tool for analyzing structures arising from nucleation and growth due to its large dynamic magnification range that enables visualization of the whole droplet footprint (commensurate with the size of the micropipet probe, shown in the SI, Section S5) and down to single atoms with sub-nanometer resolution. 24 We demonstrate this herein by considering the Eappdependent NP distributions (i.e., size, particle count and Pt mass) and NP morphologies. Note that only PtNP microensembles obtained at Eapp ≤ −0.6 V have been analyzed in detail (at Eapp > −0.6 V, the PtNPs were too small and/or too few to be resolved at a low magnification, Figure S7). Particle size analysis was performed on low magnification images (Figure 4a-d), with the resulting distributions given in Figure 4e (further details can be found in the SI, Section S6). For all Eapp, the histograms comprise bimodal distributions (also reflected in the N histograms, SI, Figure S9), with a sharp peak at very small NP sizes (<2 nm), and a broader wave for larger NP sizes, consistent with the literature. 20 The SECCM microensembles provide a wealth of information on how NP size distributions are affected by deposition parameters. In general, the percentage of small NPs (d ≤ 2 nm) decreases with increasing ƞ, while the percentage of mediumsized NPs (2 < d < 8 nm) increases. In addition, the percentage of large NPs (d > 10 nm) decreases with increasing driving force, as does the maximum diameter of the NPs (ca. 13 and 21 nm at Eapp = −0.9 and −0.6 V, respectively). For the largest driving forces, Eapp = −0.8 and −0.9 V, the histograms are very similar, indicating that the deposition mechanism does not change within this potential range (consistent with the electrochemical data, Figure 3). Low magnification images also show that the particle density and total amount of Pt increase with Eapp = −0.6 V to −0.8 V, and less strongly from −0.8 V to −0.9 V. A particle count and estimation of the total platinum mass (the methodology is described in the SI, Section S6) confirm this trend, as summarized in Table S1. Alluded to above, a narrow (and ideally controllable) NP size distribution is preferable from a practical standpoint, 2 which, from this SECCM experiment is shown to be very sensitive to Eapp, with larger ƞ (and possibly hydrogen coevolution 33 ) giving rise to a more narrow size distributions (i.e., smaller size polydispersity). Other experimental parameters that could be systematically explored in this high-throughput configuration include the deposition time and/or deposition waveform (e.g., designer E−t waveforms 18 ).  Figure 5 shows high magnification images of representative PtNPs morphologies at 3 different potentials where parallel HER is and is not important, which we use to gain an atomistic view of electrochemical nucleation and growth. 20,22,24 Generally, the smallest clusters do not have a well-defined structure (e.g., cluster 1 in Figure 5a; note the inset FFT, which indicates that the cluster is amorphous). For clusters with diameters larger than ≈2 nm, the structure becomes monocrystalline (e.g., cluster 2 in Figure 5a, and the smallest cluster in Figure 5b), and above ≈4 nm, one or more grain boundaries can be identified within the same particle (Figure 5b-d). From this, we infer that clusters are initially formed by a direct addition mechanism (i.e., addition of single atoms) up to a critical size, and larger particles must be formed by the aggregation of these smaller clusters, in agreement with the generalized electrochemical aggregative growth mechanism. 20,23 It is also important to note that, independently of Eapp and the size of and distance to the nearest NP, single Pt atoms were seen consistently across the CCTG (indicated by red arrows in Figure 5). From the spatial distribution of these single atoms, it is unlikely that they were ejected from neighboring NPs due to electron beam exposure (explored in the SI, Section S7), but rather must originate from the Pt electrodeposition process. There is significant interest in stable single atoms for use in applications such as (electro)catalysis, 34 but the role of these entities in the early stages of electrochemical nucleation and growth is still debated, although recent studies have provided indirect (through voltammetry 25 ) or direct (through ex-situ STEM imaging 24 ) evidence for their existence. In any case, the coexistence of single atoms (Figure 5a-d) and amorphous (Figure 5a), monocrystalline (Figure 5a-b) and polycrystalline (Figure 5b-d) NPs (and aggregates of these), indicates that several early growth mechanisms occur simultaneously. 20

CONCLUSIONS
In conclusion, we have presented a high-throughput SECCM-STEM platform that allows structure-electrochemistry to be correlated with single-atom sensitivity. Applied to Pt electrodeposition on a CCTG as an exemplary system, in a single experiment, it was shown that PtNP density increases and size polydispersity decreases with increasing driving force (i.e., increasing ƞ). Furthermore, it was shown that structures ranging from single Pt atoms (imaged directly on the CCTG support) to amorphous, monocrystalline and polycrystalline PtNPs (and aggregates of these) coexist during the early stages of nucleation and growth, further supporting the existence of an electrochemical aggregative growth mechanism. 20 Beyond this system, the presented correlative electrochemistry−microscopy approach should be generally applicable to any class of (electro)material, for example battery materials 12 and electrocatalysts, 9 positioning this technology as a powerful tool for structure-property elucidation and the rational design of functional materials.