In-situ observation of plasmon-controlled photocatalytic dehydrogenation of individual palladium nanoparticles

Plasmonic nanoparticle catalysts offer improved light absorption and carrier transport compared to traditional photocatalysts. However, it remains unclear how plasmonic excitation affects multi-step reaction kinetics and promotes site-selectivity. Here, we visualize a plasmon-induced reaction at the sub-nanoparticle level in-situ and in real-time. Using an environmental transmission electron microscope combined with light excitation, we study the photocatalytic dehydrogenation of individual palladium nanocubes coupled to gold nanoparticles with sub-2 nanometer spatial resolution. We find that plasmons increase the rate of distinct reaction steps with unique time constants; enable reaction nucleation at specific sites closest to the electromagnetic hot spots; and appear to open a new reaction pathway that is not observed without illumination. These effects are explained by plasmon-mediated population of excited-state hybridized palladium-hydrogen orbitals. Our results help elucidate the role of plasmons in light-driven photochemical transformations, en-route to design of site-selective and product-specific photocatalysts.


Supplementary Note 1: The definition of 'induction time' and its relation to hydrogen desorption/diffusion from the Pd lattice
Our previous results 1-3 indicate that single-crystalline palladium nanoparticles (like the investigated nanocubes) do not sustain thermodynamic phase coexistence; the core of singlecrystalline nanoparticles will exclusively be either phase or phase in equilibrium. This result is evidenced by both EELS, allowing us to assign each region in the nanocube to a specific phase, as well as diffraction. Using both techniques, sharp transitions in pressure-composition isotherms are obtained. Also using both techniques, we have shown that both in the dark and upon illumination, the transformation reaction is initiated by the nucleation of the new phase at the corner of the cube. Naturally, there must be a vacant site in the palladium lattice to host the hydrogen atom while it diffuses out of the nanocube. However, to date, our method is not sensitive to the sub-surface, and we cannot rule out diffusion of hydrogen from the ~1 nm subsurface of the palladium cube. It is possible that hydrogen from the first few atomic monolayers of the Pd cube diffuses out before we are able to observe a change in the particle.
Thus, we included this uncertainty in the definition of the induction time by explicitly specifying that we refer to the observed nucleation of the new phase as the onset for the diffusion of hydrogen atoms and the beginning of the second reaction step. Nevertheless, the fact that our measurements allow us to distinguish between two subsequent steps with different wavelength dependences, (the induction time and reaction time) implies that each of these steps is governed by one primary mechanism, characterized by a different activation energy.

Supplementary Note 2: The role of Au nanodiscs
The presence of the Au nanodiscs next to the Pd nanocubes is important in two aspects: first, it significantly improves the visible-frequency absorption cross section of the Pd nanocube by inducing a "forced plasmon"; second, it breaks the symmetry of the geometry, compared to an isolated Pd nanocube, by forming a plasmonic hot-spot in the nanogap. This makes the corner of the Pd nanocube closest to the Au antenna more reactive and eventually leads to the observed site-selectivity. Nevertheless, we would expect that illuminating an isolated Pd nanocube in the UV, nearer to the plasmon resonance of the Pd itself, would also affect the kinetics and possibly the mechanism (for example, by illuminating at the corner versus edge mode of the cube). In that case however, no site-selectivity is expected to be seen and a much higher illumination power is 14 needed since the absorption cross section for the Pd itself is relatively poor. Such a study is beyond the scope of this paper and is the subject of future work.

Supplementary Note 3: Near-constant reaction time with wavelength
The diffusion of hydrogen in the lattice has a lower activation energy than the desorption of hydrogen from the surface. While the excitation of plasmons in the Au-Pd system overcomes the activation barrier for the diffusion even in 'off resonance' wavelengths, it is still insufficient to bring the system to a higher energetic state beyond the activation barrier for the desorption of hydrogen. Basically, the excitation of plasmons at different wavelengths modifies the 'effective' activation barrier for the desorption process while it completely removes the activation barrier for the hydrogen diffusion step. The difference between the two activation barriers with the one for diffusion being lower than that for the desorption is also found in bulk measurements of palladium hydrides. It is worth pointing out that lower laser intensities and further off-resonant wavelengths that we have tried using weren't sufficient enough to initiate the reaction by inducing the desorption of hydrogen from the surface and thus prevented the second reaction step from occurring.

Supplementary Note 4: Temperature increase upon illumination and its contribution to the observed results
In order to estimate the temperature increase upon illumination in the antenna-reactor system, we used COMSOL to simulate the time-dependent temperature evolution and its profile due to pulsed illumination. We consider pulsed illumination with peak power of 1500 W/cm 2 (corresponding to an average power of 5 mW of a 100 ps pulse at 700 nm with repetition rate of 78 MHz). We also consider a 30-nm thick silicon nitride membrane with a thermal conductivity of 2 W/mK 4 at our operating temperature (243 K). As seen in Supplementary Fig. 11, our calculations indicate a transient temperature increase of ~100 K for the Au nanodisc and ~400 K for the Pd nanocube. The contact with the cooled membrane helps bring the two nanoparticles to their steady state temperature of 240 K within ~ 200 -400 ps. Thus, the average temperature increase of the Au nanodisc is less than 1 K while for the Pd nanocube it is ~ 7 K. Except for localized heating, when illuminating large number of particles, it is also important to consider the collective heating of the sample in which case the temperature becomes uniform 5 . Under these conditions we cannot rule out heating effects as the reason for the enhanced kinetics, which are 15 likely to be convoluted with the effects emerging from hot carrier excitation. This assumption is further confirmed by power dependent measurements of the reaction rate which show nonlinear behavior of the induction rate, −1 (Supplementary Fig. 12). This nonlinear dependence can be related to thermal effects 6 as well as to multiple electronic excitations 7 .
However, heating by itself cannot account for the site selectivity unless a temperature gradient across the Pd nanocube is formed. Our COMSOL simulations take into account a uniform absorption cross section (1•10 -14 m 2 for the Au nanodisc and 2.5•10 -14 m 2 for the Pd nanodisc taken from Lumerical FDTD simulation as shown in Supplementary Figure 7) and in such case, as can be seen in Supplementary Fig. 11, the temperature is uniform across each nanoparticle.
Nevertheless, a transient local temperature increase can be formed on one side of the Pd nanocube due to an enhanced local absorption at the hotspot which will result in an increased hydrogen desorption at this site. Due to fast thermal equilibration in the Pd lattice, in the order of ~ 100 ps (thermal diffusivity 10 -5 m 2 s -1 ) 8 , we assume this is not the dominant process. Thus, we can conclude that while temperature increase might play a role in enhancing the reaction rate, it cannot account for the observed site selectivity.