Electrochemical Atomic Layer Deposition by Surface Limited Redox Replacement of Pd Thin Films in One-Cell Conﬁguration Using Cu UPD Layers: Interrupting Mass-Transport Limited Growth

This work presents a study focused on optimization of the growth of Pd thin ﬁlms by surface limited redox replacement (SLRR) of an underpotentially deposited (UPD) Cu layer in a one-cell conﬁguration, and their thorough characterization. The growth was monitored by open-circuit chronopotentiometry and deposited ﬁlms were consequently characterized for roughness evolution by H UPD andCu UPD cyclicvoltammetry(CV)whiletheﬁlmthicknesswasevaluatedusinganodicstrippingvoltammetryandintegration of theaccumulated stripping charge.Thegrowthof smoothPd ﬁlmswas observedfor ∼ 15successively grownequivalent monolayers (MLs) of Pd followed by a rapid transition to 3D growth. A comparison with counterpart results of Pd SLRR deposition in a ﬂow-cell where no such growth-mode change occurs suggests that the transition seen in one-cell is likely associated with establishment of mass-transport limitations. Close examination of SLRR transients, in-situ STM results on the ﬂat grown layers, and SEM images of the surface after the transition occurs conﬁrm the likelihood of growth kinetics evolution. Finally, this one-cell speciﬁc shortcoming was addressed by the introduction of SLRR-cycle disruption along with forced convection, which resulted in the smooth growth of Pd ﬁlms on Au up to 26 Pd MLs. © The Author(s) 2018. Published by ECS. This is an open access article distributed under the terms of the Creative Commons Attribution 4.0 License (CC BY, http://creativecommons.org/licenses/by/4.0/), which permits unrestricted

The study of metal thin films on various substrates has been intensified in the last few decades due to fundamental and applied research needs. The growth of metal thin-films of varied thickness, on multiple substrates, under controlled conditions, for fundamental understanding and for various applications has driven thin-film research and development into a new perspective altogether. In comparison to the bulk counterparts, the ultra-thin metal films experience surface strain and unique electronic interactions with the substrates consequently exhibiting unique physical and chemical properties. 1,2 The effects of such interactions include improved catalytic activity, enhanced durability of catalysts and sometimes improved electronic and magnetic properties. In particular, palladium (Pd) has a very high affinity for hydrogen (H), it cannot only reversibly and selectively adsorb a monolayer of H but also absorbs large amounts of hydrogen, as high as 900 times its volume under ambient conditions of temperature and pressure. 3 With such properties, Pd thin films can be used for H storage, as H 2 sensors 4 or in the fuel cell industry as catalysts. 5,6 To commercialize fuel cell powered vehicles, cheaper Pd or its alloys are preferred as oxygen reduction catalyst relative to platinum (Pt), the preeminent catalyst material. 7,8 The growth of thin films on different substrates has been achieved via variety of approaches aiming at epitaxial and continuous deposits. Among these are physical vapor deposition, 9 molecular beam epitaxy, 10 chemical vapor deposition, 11 magnetron sputtering 12 and electrodeposition. 13,14 These techniques allow for controlled thickness, morphology and continuity of the epitaxially deposited thin films. The electrochemical approach offers a more economical and sustainable deposition route as it is operational under ambient conditions relative to the other techniques which require ultra-high vacuum conditions along with elevated temperatures that are very expensive to sustain. 14 In contrast, electrodeposition only requires oxygen evacuation for proper functioning. With electrodeposition, the application of a constant potential negative to the equilibrium potential of the metal ion/complex of interest leads to bulk deposition which oftentimes preferentially follows the Stranski-Krastanov transition, 15 * Electrochemical Society Member. z E-mail: dimitrov@binghamton.edu i.e. after initial deposition of a few 2D heteroepitexial film layers, the growth transitions to 3D growth. The electrochemical atomic layer deposition (E-ALD) has been achieved by employing surface limited redox replacement (SLRR) [16][17][18][19] technique to grow epitaxial and conformal thin layers of metals or metal alloys on different substrates. The SLRR protocol involves the underpotential deposition (UPD) of a monolayer of less noble element (Cu, Pb or H in the case of Pd thin layer deposition) to form a sacrificial layer on a conductive substrate such as polycrystalline Au poly electrode or glassy carbon electrode. This is followed by a redox replacement reaction at open circuit potential (OCP), in which ions of a more noble metal (Au, Pd, Pt, Rh, Ag, Cu, etc) of interest to the growth exchange with the UPD layer atoms. The completion of the SLRR cycle results in the deposition of a conformal layer of the more noble metal on the substrate surface. The thickness of the thin film of the growing metal depends on the number of the SLRR cycles completed. [16][17][18][19][20] Pd ultra-thin film deposition via an automated flow-cell system has been extensively studied by J. Stickney et al. 6,[21][22][23] This group's work, identified Cu as a better sacrificial element than Pb, because the use of Pb UPD layer resulted in the deposition of Pd thin films containing considerably more residual Pb impurities than Cu. 23 The Pb atoms are expected to be replaced during the OCP redox reactions but due to the high stability of the Pb-Pd alloy and the negative enthalpy of formation of the alloy, some Pb remains trapped in the growing Pd film. 23 While Cu did work better, one of the crucial issues encountered in Stickney's work was associated with the directional flow of the electrolyte in the flow cell. It was demonstrated that in the growth of the Pd thin films, a descending thickness gradient was developed from the ingress to the egress of the cell, a phenomenon attributed to the dominancy of the local-cell exchange mechanism due to limited solution exchange period versus the high redox replacement rate. 19,24 The proposed solution to the high rate of the redox exchange was to complex the Pd 2+ ions to slow the reaction and thus allow for exchange to occur evenly across the substrate. At a [EDTA]/[Pd(II)] ratio of 1, the thin films grew more uniformly with minimized 3D transition but at the expense of exchange efficiency. 23 For improved uniformity and higher exchange efficiency, the deposition conditions were optimized by changing the complexing agent, and with the [Cl − ]/[Pd(II)] ratio of 500, J3075 Stickney et al. managed to complete 75 SLRR cycles at an exchange efficiency of about 95%. 6 The high maintenance and relatively high operation cost, makes the automated flow cell system less cost effective than the one-cell approach that features simplicity of operation and the comfort of having all solution components for E-ALD via SLRR in a single solution vessel. 25 Although the one-cell approach addresses some drawbacks experienced in the automated flow-cell set-ups, it also has some limitations including, among others, the inevitable transitioning to diffusion-limited growth in the course of thicker film deposition. 26 Such a transitioning has been recently believed to facilitate systems growing in a quasi-2D mode to proceed eventually to a 3D deposition leading to growth of dendrites. 26,27 In this work, we present a study on the SLRR growth of Pd thin films in a one-cell configuration using Cu as a sacrificial metal. Along with detailed optimization effort assessing the impact of absolute and relative concentrations of Pd 2+ and Cl − ions on the deposition outcome (kinetics and morphology), of interest to this work has been the transitioning to dendritic growth and development of protocols to minimize its adverse effects on the Pd growth in general. More specifically, the growth of Pd is conducted and studied by potential pulse routines and open circuit chronopotentiometry (OCCP). The film roughness is assessed by H UPD and Cu UPD cyclic voltammetry (CV) experiments and the film thickness (deposition efficiency) was determined by anodic stripping voltammetry (ASV) of Pd films following all test and characterization routines. In situ STM experiments were also performed to monitor the growth process and to observe the transition from 2D to 3D dendritic growth. SEM results show the dendritic growth through the images of the surface morphology. XPS results also show that negligible Cu residues remain in the Pd thin films as grown on the Au substrate.

Experimental
Electrode preparation.-The working electrodes for all electrochemical and morphological characterization experiments in this work are polycrystalline Au poly discs (0.9999 purity) of 6 mm diameter and thickness of 2 mm. Their preparation involves mechanical polishing down to 1 μm using water-based, de-agglomerated alumina slurry (Buehler). After which, they are rinsed with water before being immersed in warm concentrated HNO 3 to remove any polishing residual particles, followed by thorough rinsing with Barnsted NanoPure water (18.2 M .cm). Next, the Au disks are electropolished via protocol described elsewhere. 28 The final step of preparation involves carefully annealing the crystals to red-hot in a propane torch for minimum 5 min before cooling rapidly in ultrapure nitrogen atmosphere to avoid surface oxidation. The process is terminated with a droplet of the Barnstead NanoPure water to prevent surface contamination. These electrodes are then mounted onto a conductive vacuum holder and immersed in the electrolyte via a hanging meniscus configuration 29 in a three-electrode cell for surface characterization and subsequently, for thin film growth. All electrochemical experiments were performed in three electrode cells using solutions made from ultra-high purity grade chemicals as received from the vendors and Barnsted NanoPure water. A saturated Mercury-Mercurous Sulfate Electrode (MSE) is used as reference electrode in most experiments unless stated otherwise. Also, a Pt wire serves as the counter electrode (CE) in all experiments. All potentials in the manuscript are presented versus MSE (except the Cu UPD and Pb UPD characterization experiments) and all current densities are normalized with respect to the geometric area of the electrode. The Single crystalline Au (111) (1.0 cm in diameter and 2 mm thickness, Monocrystal Company) used for morphology characterization in in-situ STM experiments were prepared following a procedure described in detail elsewhere. 28 Initial characterization of Au poly electrodes.-The Au electrodes are characterized using Pb UPD CV 26 to determine the quality of their surfaces before Pd deposition as well as to determine the electrochemical surface area (ECSA). The solution composition for Pb UPD experiments is given elsewhere. 26 The CV measurements were performed using Model AFCBP Bipontentiostat (Pine Instruments) interfaced with a PC through the PineChem software (Version 2.80). Pb wire was used as a pseudo-reference electrode (PRE).
Pd Ultra-thin film growth.-The formation of the ultra-thin layers of Pd on Au was performed via different number of SLRR events in a one cell configuration. The optimized deposition solution made of 0.3 mM PdCl 2 (Aldrich, 99.99%), 30 mM HCl (J.T. Baker, 36.5-38%), 3 mM CuSO 4 (Aldrich, 99.995%) and 0.1 M H 2 SO 4 (GFS Chemical, redistilled 95-98%) was purged with ultrapure N 2 for at least 30 min. In all experiments, purging of ultrapure N 2 continued throughout the experiments to minimize oxygen interference. Similarly to a procedure used for the SLRR deposition of Pt, 30 one SLRR cycle involves the formation of a conformal submonolayer or monolayer of Cu UPD by application of a potential pulse to a potential of −0.35 V (MSE) for 1s followed by a redox replacement step by the more noble metal -Pd, occurring at OCP until the potential reaches a cutoff value in the range 0.00 to −0.05 V. The cycles could either be run continuously, or be broken into successive groups of four runs, until the total number of intended cycles is reached. During the breaks, the working electrode is pulled out of the growing solution (without opening the cell) and the solution is shaken to homogenize the components before it is placed back in the hanging meniscus for further SLRR runs. The Pd thin-film growth protocols were administered and monitored using a potentiostat/galvanostat Princeton Applied Research (PAR) Model 273 coupled with Corrware Software. Data on the continuous SLRR transients was collected on a PC by analog-to-digital DrDAQ Data Logger controlled by PicoLog software. 26 Electrochemical testing and characterization.-The H UPD and Cu UPD CV curves were used for electrochemical characterization of the as-grown Pd thin films. The H UPD CV was conducted in 0.5 M H 2 SO 4 (GFS Chemical, highest purity grade, redistilled) within a potential range that minimizes H absorption i.e. 0.1 V to −0.67 V (vs MSE) at a sweep rate of 50 mV.s −1 . We conducted the Cu UPD characterization via CV in 3 mM CuSO 4 (J.T. Baker, 99.8%), 0.1M H 2 SO 4 solution using Cu wire as PRE at a sweep rate of 20 mV.s −1 . The characterization of these Pd thin films was done using the instruments and software used for Pb UPD characterization mentioned early.
Anodic stripping of Pd films.-When characterization of the thinfilms of palladium on gold is done, anodic stripping of the same films is done by ASV in a 0.1 M HCl (J.T. Baker, 36.5-38%) solution over a potential range −0.3 to 0.4 V against MSE as a reference electrode. The ASV sweep rate 3 mV.s −1 and was controlled by Model AFCBP Bipontentiostat (Pine Instruments) interfaced with a PC through the Aftermath Data Organizer software.

In-situ STM monitoring and characterization.-In-situ STM was
carried out utilizing an Agilent 4500 SPM microscope, a 2100 controller and PicoScan software (version 1.14) in an analogous solution to the optimized one as described in the deposition protocol section above but with ratio [Cl − ]/[Pd(II)] = 8. The potential was controlled by an Agilent 300S Pico Bipotentiostat, and a Pt ring and Pd wire were used as counter and reference electrodes respectively. The growing solution was also purged with ultra-high purity N 2 to minimized oxygen interference during the experiment. STM tips were fabricated by etching Pt 0.9 Ir 0.1 wire in 1.6 M CaCl 2 utilizing an AC voltage of 25-30 V, followed by insulation with Apiezon wax to minimize the tip exposure to solution thus minimizing leakage currents during the experiments.  survey. For surface cleaning and depth profiling, the as-grown Pd thin film was sputtered with Ar sputtering gun 2 min. SEM characterization.-Scanning Electron Microscopy (SEM, Zeiss Supra 55 VP) coupled with an in-lens detector at an accelerating voltage of 10 kV and a working distance of 2 mm was performed for morphological characterization of as-grown Pd films on Au electrode discs.

Results and Discussion
Initial Au electrode characterization.-The predominant crystallographic surface orientation of the Au substrate influences the structure of the growing Pd metal thin layer. To assess their quality, the Au electrodes used in this work are characterized by Pb UPD ( Figure 1A) and Cu UPD CV curves ( Figure 1B) in which both shapes and features are typical to polycrystalline Au with strongly dominating (111) surface orientation. Work on electrodes of the same or comparable quality was also reported in our previous publications. 26 Samples must exhibit the typical peaks that represent low-index Au facets of (111), (110) and (100) to indicate the readiness of the substrate for the deposition step.
Both Cu and Pb have been demonstrated to be effective sacrificial elements in the growth of epitaxial thin films of metals on a variety of conductive substrates. 16,18,23,27,31 As mentioned earlier, in the work of Stickney's group, 23 Cu was found to be a better sacrificial element relative to Pb for Pd thin film growth via SLRR. Therefore, growth via SLRR using Cu UPD was preferred to that of Pb UPD. In their work, using the flow cell system led to a known issue manifested by a differential Pd deposition along the cell width with more Pd depositing at the ingress region where Pd activity is highest in the beginning of each SLRR step. This resulted in higher rate of exchange at the ingress relative to the solution introduction time. 6 To solve this issue, the Pd ions were complexed EDTA to slow down the exchange reaction but as mentioned earlier the growth outcome was unsatisfactory even at the optimal ratio [Pd(II)/EDTA] = 1. 6 A more labile (weaker) complexing agent was needed to address both growth homogeneity and exchange efficiency concerns. The work of both Uosaki 14 34 Stickney's work showed that use of Pd(II) complexes with Cl − , resulting in a known negative shift of Pd reduction potential by ∼0.3 V due to the presence of less free Pd(II) ions, 35 leads to formation of high-quality deposits. 6 The reported optimized concentration ratio that resulted in best quality of Pd films deposited by SLRR in  Figures 2A and 2B, represent the H and Cu UPD characterization CVs on the as grown ultra-thin Pd layers on Au poly respectively. As can be seen, based on the Cu UPD analysis the roughness (preferential 3D growth) starts to develop between 4 and 7 SLRR cycles. This roughness evolution is seconded clearly by the adsorption portion of H UPD CV, from −0.5 V to −0.6 V. Further consideration of the CV curves negatively to -0.6 V is not informative as in that range H not only adsorbs on the Pd surface, but it also absorbs into the Pd lattice. The anodic stripping voltammograms (Fig. 2C) and the number of MLs deposited, determined by integration of the charge of ASVs, with respect to the number of SLRR cycles completed (Fig. 2D) also supports the roughness observed in the CVs. More specifically, the departure from linearity observed for the deposit's thickness in equivalent monolayers (eq. MLs) after 7 SLRR cycles is attributed to sudden roughening that could be a result of initiation of 3D growth. Subsequently, the 3D clusters could serve as nuclei of dendrites that would grow readily if the Pd concentration at the interface of working electrode is less than that of the bulk solution. 36,37 Thus, the growth process will be affected by the diffusion limitation resulting from hindered mass transport to the electrode surface with the increasing number of SLRR cycles.
The potential transition from quasi-2D growth to dendritic growth inevitably impacts the potential transient associated with SLRR deposition in one cell configuration. Figures 3A and 3B, present typical potential transients reflecting the potential changes in the course of each SLRR cycle at different Cl − /Pd 2+ concentration ratios. Apparently, the duration of one cycle is directly dependent on the strength of complexation between Pd(II) and Cl − which is affected indirectly by the ratio between these ions as they form a variety of complexed species, [PdCl x [H 2 O] 4-x ] 2-x thus making the Pd less available to the growth process (compare Figs. 3A and 3B). Also, if no mass transport limitations are impacting the deposition all cycles should have identical duration (Fig. 3A). Some difference could be expected only in the very beginning when the Au surface is still not completely covered by Pd (Fig. 3B). Overall, as also seen in Figure 3B, the potential transients during SLRR deposition in one-cell configuration could be used as an early indicator of likely growth mode transitioning manifested by prolonging of single-cycle duration. These observations could be cross-compared later with those from the CV and ASV experiments    and (as needed) ultimately validated by direct imaging, such as SEM and/or STM.
The optimization of the SLRR deposition routines in one cell continued with preserving the [Cl − ]/[Pd(II)] ratio of 8 but increasing by a factor of 2 the absolute concentrations of said ions. Unfortunately, no significant difference was registered during the characterization tests (not presented). In a marginal improvement the transition to roughening occurred between the 6 th and 10 th cycle instead of 4 th and 7 th in the experiment presented in Figure 2. Overall, the CV curves and the departure from linearity in the stripping tests confirm the transitioning from layer-by-layer to 3D growth at higher [Pd(II)] as well. A higher [Pd(II)] concentration could provide favorable conditions for longer quasi-2D growth. However, such concentration is inapplicable for SLRR in one-cell as it will promote a substantial contribution of electrodeposited Pd during the sacrificial layer formation step taking place under potential control. 25  38 suggested that there was virtually no roughness development even with 20 SLRR events however, a further quantification of the Pd from the anodic stripping curves revealed that such concentration ratio had a drawback of very low exchange efficiency, whereby the 20 SLRR cycles were found to result in the deposition of less than 3 MLs of Pd. As explained elsewhere in this work, the higher the concentration of the Cl − ions the stronger Pd-Cl − complexation, this ultimately reduces the exchange efficiency during the SLRR deposition process. This is explained by the high value of formation constant for the [PdCl x [H 2 O] 4-x ] 2-x complex, formed as a result of H 2 O -Cl − ligand exchange, β4 = 10 11.53 , which depicts a very stable complex. As a result of this stability the concentration of free Pd(II) ions is very low. 34 This condition in turn indirectly favors the side reactions such as ORR, which are suppressed due to N 2 purging, but cannot be completely excluded via this approach, over Pd reduction, thus further contributing to lowered deposition efficiency.

Growth dependence on the [Cl − ]/[Pd(II)] ratio.-Next
Further experimental designs for optimization of the conditions maintained the absolute concentration of Pd(II) ions fixed but changed the concentration ratios from [Cl − ]/[Pd(II)] = 500 to 100 and 170. Figure 4 shows the best results of accordingly optimized Pd growth by SLRR of Cu UPD layer (at [Cl-]/[Pd(II)] = 100) as demonstrated by CV characterization and subsequent anodic stripping of all accordingly grown Pd thin films. It is clearly seen from the results in Figure 4 that there was smooth growth up to the 12 th cycle and a substantial roughness evolution after 20 cycles of SLRR growth. The inference from the Figures 4C and 4D confirm the smooth growth of Pd thin films up to the 12 th SLRR cycle which is equivalent to 16 MLs of Pd deposited. The perfect linear trend of charge (equivalent number of monolayers) against the number of SLRR cycles up to the 12 th cycle (in Fig. 4D) acts as evidence of the same. Just like other reported results, the transition toward 3D, the dendritic growth is evident in Overall, the optimization study of Pd deposition by SLRR of Cu UPD along with results from earlier work 26 whereby Pd was grown by replacement of H UPD layers, both conducted in one cell system, suggested the persistent presence of a 2D to 3D transition at different growth thickness. This transition has been attributed to dendritic growth occurring upon establishment of mass-transport limitations at the growing interface. Then, as a result of the hindered transport of the [PdCl x [H 2 O] 4-x ] 2-x complex, the growing surface, a progressively formed concentration gradient (from zero on the electrode surface to the bulk concentration) serves to enhance the vertical growth of any 3D clusters nucleated in the initial deposition stages leading eventually to dendrite formation. 39,40 Therefore, we present, in the following part of our work some surface characterization and elemental analysis results that are aimed at assessing aspects of the Pd layer quality before and after the transition.

In-situ STM characterization (setting the limits of flat growth).-
The evolution of Pd film deposit morphology was explored as a function of the number of SLRR cycles completed via in-situ STM. This was done to gain insight into the deposition growth mode and morphological evolution that drives the transition to dendritic growth. These films were deposited with a [Cl − ]/[Pd(II)] = 8 ratio, 23 to ensure that roughness evolution could be observed over the time of deposition. Figure 5A illustrates the un-reconstructed Au (111) single crystal substrate. The morphology of the substrate exhibits large terraces separated by mono-atomic steps or step bunches. The terraces are also decorated with a small coverage of Au clusters that are a result of the lifting of the Au (111) surface reconstruction. In Figure 5B it can be seen that following the completion of first SLRR cycle the surface is uniformly covered by a minimum of one Pd ML, based on the deposition efficiency determined by the ASV, and evidenced by the loss of sharp step edge boundaries initially present on the substrate. Following the first SLRR cycle additional Pd clusters atop the pre-nucleated Pd underlayer are also present. These clusters are preferentially located at defect sites that are originally present on the Au (111) substrate, i.e. step edges and clusters that result from lifting of the Au (111) reconstruction. Similar preferential cluster nucleation and growth has been demonstrated in deposition systems in which the growing metal exhibits a high melting point, and consequently slow surface diffusion. 32 After two SLRR cycles, the clusters grow in size and merge where the size prohibits further lateral growth, at this time the step edges are completely decorated with Pd clusters. After two SLRR cycles there is also evidence of additional layer-by-layer deposit propagation on terraces external to the nucleated clusters. This additional growth is especially evident in branching extending from the largest island present in the upper right corner of Figure 1A, branching indicated by the arrow in Figure 5C. This additional layer-by-layer growth indicates a continued epitaxial relationship between the film and the substrate. Following five SLRR cycles ( Figure 5D), additional clusters nucleate on the substrate terraces and there is little exposed flat terrace remaining. Finally, following 10 SLRR cycles there are few recognizable features present from the initial substrate, and instead the surface is uniformly covered with large Pd clusters some of which are clearly thicker than the majority in the dense cluster network as seen in Figure 5E.
The cluster deposition and growth following five Pd SLRR cycles on Au is likely responsible for the roughness evolution that is seen in the UPD voltammetry. Similar cluster deposition and growth is seen in the deposition of Pt on Au via SLRR 30,41 and conventional methods. 32 However, in the deposition of Pt on Au a pseudomorphic overlayer never develops. The pseudomorphic overlayer present in this system is likely to be as a result of the strong interaction between Pd and Au, an interaction that ultimately promotes Pd UPD layer formation on Au and in alternative systems. Interestingly, D. Kolb et al. showed that following the conventional electrodeposition of 10 layers of Pd on Au a fractal deposit morphology is developed, a morphology quite different 32 from that seen in Figure 5. However, that work does not utilize a pre-nucleated Cu UPD wetting layer, which likely promotes more uniform growth in cluster form across the entire substrate. In addition, the Pd deposition overpotential applied by a nucleated Cu UPD layer in this work is significantly greater than that  Figure 6 suggest the presence of practically only trace amounts of Cu not exceeding 2-3 at% in the deposited by SLRR Pd film. The sample analysis shows also the minor presence of other elements which can be attributed to oxidation of Pd and adsorption of adventitious C based oxides (O1s and C1s, respectively), with trace amounts of Al from the mechanical polishing and Cl − from the growth solution. This level of incorporated Cu agrees very well with the results obtained by Stickney's group 23 for Pd growth by SLRR of Cu UPD in flow cell, demonstrating 0.9 and 1.7 at% for 15 and 25 cycle grown layers, respectively. The slightly higher limiting level of incorporated Cu in our work could be attributed to the thicker Pd layer (grown by 50 SLRR cycles) used in the XPS analysis. Given the trend of amount of incorporated Cu increase with the number of cycles in Stickney's work, one would not be surprised to see levelling of the Cu content at 2.5 to 3.0 at%. While many studied SLRR based growth approaches for other systems result in no sacrificial metal incorporation, the low level of Cu inclusion in the present case could somewhat be expected given the high bulk miscibility of Pd and Cu manifested by the formation of a single-phase alloy over the entire compositional range. 42 This generally suggests no statistically significant difference in the Cu incorporation in Pd layers deposited by one-cell and flow cell approaches. Also, the practically negligible presence of incorporated Cu atoms in the Pd film is largely insufficient to serve as source of lattice strain that could promote structural defects at low film thickness. Therefore, it is highly unlikely to believe that the observed transitioning to dendritic growth demonstrated clearly by the results presented in Figure 2 is a result of Cu incorporation. That is why in the next paragraph we present a closer look on the surface morphology of Pd layer grown by 50 SLRR cycles (after the roughening occurred, Fig. 2) and more detailed analysis of the entire SLRR deposition kinetics. While the image quality is insufficient to reveal a typical dendritic structure, one can certainly associate the substantially higher surface roughness measured by the analysis of the H and Cu UPD characterization results in Figures  4A and 4B with the presence of distinct 3D structure on top of the Pd film as well as some fine nanoporosity evidenced by the darkest contrast regions. In order to better understand the transition from a well-established quasi 2D deposition seen in the STM images in Figures 5 to 3D growth leading to the surface morphology seen in Figure 7A we present 25 cycles of the potential transient registered during the growth of the SEM characterized Pd film (Fig. 7B). A closer look at the transient suggests relatively steady cycle duration to about 11 to 13 of the successive SLRR steps. Then, the duration of the subsequent cycles persistently increases. Such continuous retardation of the redox exchange reaction could be due either to change in the interaction of the growing film with the sacrificial metal or the reduction of the growing-metal-ion concentration. However, no reason for interaction changes between sacrificial layer and growing metal film could be seen given the (virtually) contamination free Pd deposition  (see XPS results). Therefore, the most likely reason for transition to 3D growth remains to be a gradual alteration of the exchange reaction kinetics, transitioning from activation to mass-transport controlled. The likelihood of such kinetic transition was first brought up and demonstrated analytically by B. Rawlings et al. through a recently drafted model describing the kinetics of successive SLRR cycle in one cell configuration. 43 Under these circumstances, 3D clusters formed in the course of the cluster network densification seen already in Figure 5E could serve as nucleation centers of the yet-to-be-grown dendritic structure. As mentioned in the Introduction, transitioning from quasi 2D to dendritic growth has also been shown to occur in the SLRR and surfactant mediated growth of Au on Pt surfaces 27 as well as in the deposition of Pd on Au by SLRR of H UPD . 26 It is noteworthy that the deposition process in both SLRR systems exhibiting said transition was handled in one-cell configuration.

SEM characterization and Chrono
SLRR deposition with disruption of the growth process.-All results presented earlier in this work indisputably demonstrate that the SLRR growth of Pd on Au in a one-cell system is associated with transitioning from layer-by-layer to 3D, dendritic growth at a certain point of the deposition process, dictated by deposition conditions. Even more, the transition appears to be an inevitable phenomenon and its commencement time depends on the concentration of the growing metal. At the same time no evidence for such transition could be found in the experiments of Stickney et al. in the same system, but carried out in an automated flow-cell setup. 6 Most likely this is due to the absence of mass-transport limitations in the flow cell system whereby (unlike in the one-cell case), each SLRR cycle is initiated at maximum concentration of [PdCl x [H 2 O] 4-x ] 2-x complex in the near electrode vicinity. This inference is seconded by the significantly shorter time needed for a single SLRR cycle in flow cell compared to one-cell configuration under identical other growth conditions. Indeed, this is not surprising given the naturally existing identical starting conditions before each SLRR cycle in the flow-cell setup as opposed to the steadily decreasing growing metal concentration in the near electrode vicinity in one-cell configuration. 26 Having recognized the inherent phenomenon of transitioning to 3D growth in a one-cell environment, we decided to optimize the deposition process by disrupting the growth after several successive SLRR cycles accompanied by introducing electrolyte convection during the break. The convection of the electrolyte homogenized the solution and refreshed the concentration of the Pd ions in the interface on placing the Au electrode back in the hanging meniscus position. Then, the growth was resumed for another set of successive cycles until the next interruption. Figure 8 presents summary of the characterization of one such optimized deposition experiment performed by SLRR in a one-cell configuration. The growth in that case was administered in multiples of four successive SLRR cycles with a 30-second growth interruption accompanied with forced electrolyte convection (swirling). The CV characterization by both H UPD and Cu UPD reveals virtually no roughness development even after 24 SLRR cycles of growth. It is worth noting that this resulted in the deposition of continuous Pd film with a smooth morphology and thickness of 26 equivalent MLs. This is almost twice the thickness of the thickest smooth film (before the 2D-3D transition) presented earlier in Figure 3. While this result is comparable with results obtained in flow cell, 6 not even a single experiment carried out in conventional one-cell configuration mode (without interruption and convection) came even close to this quality of growth in our previous optimization attempts. This result is seconded by the stripping experiments ( Figures 8C and 8D) suggesting almost perfectly linear increase of the grown layer thickness of up to 26 equivalent MLs deposited by 24 SLRR cycles in one-cell configuration.

Conclusions
In conclusion, this work has demonstrated the use of one-cell setup for the growth of Pd thin films on polycrystalline Au electrodes using E-ALD by SLRR of Cu UPD sacrificial layer. Results reported elsewhere, utilizing an automated flow-cell to deposit Pd films on Au have been critically compared with ones obtained in this work. The Cu contamination in Pd films deposited in one-cell found to be up to 3 at% is at a comparable level to the reported in flow-cell counterparts. It has also been shown that there is an inevitable transition to 3D, likely dendritic growth in the one-cell approach. This transition has been attributed to establishment of mass-transport controlled deposition at some point of the process duration as evidenced by SLRR transients, STM and SEM images. The uninterrupted SLRR growth in a onecell was optimized to yield smooth and continuous films of thickness up to 15 equivalent MLs at Pd(II) ion concentration of 0.3 mM and [Cl − ]/[Pd(II)] ratios of between 100 and 170. A major improvement was achieved by implementing a modified SLRR protocol for deposition in one-cell with interruption of the successive SLRR cycles and applying forced electrolyte convection. While more labor intensive, this modification resulted in thicker Pd deposits (∼26 equivalent MLs) that were smooth and continuous.