Temporally Distinguishing Photocatalytic Dynamics

Abstract: Light can induce both photothermal and nonthermal catalytic activity in plasmonic nanoparticles, but the extent, timescale, and efficacy of these two mechanisms remain unresolved. Here we introduce pump-pump photocatalysis, an adaptation of ultrafast excitation correlation spectroscopy in which incident laser pulses are split into two spatially and energetically equivalent pulses separated by a variable time delay. For a given reaction and catalyst, time-sensitive nonlinear enhancements in photocatalytic activity may be distinguished from time-insensitive responses as a function of time delay, pulse power, excitation wavelength, applied temperature, reactant pressure, and pulse asymmetry. In this way, the timescales of photothermal and nonthermal activity of any photocatalyst may be monitored on sub picosecond to nanosecond timescales, and the conditions for optimal chemical reactivity may be discovered. Here we report ultrafast measurements of the ammonia decomposition reaction using a Cu-Ru antenna-reactor photocatalyst. Photothermal contributions exhibit little sensitivity to pulse delay, while nonthermal contributions are most apparent at low excitation intensity, moderate temperatures, and sub-nanosecond timescales. Here, nonlinear, nonthermal mechanisms enhance H$_2$ production by a factor up to eleven compared to when pulses overlap. This technique may be used to provide unprecedented \textit{in operando} diagnostics and control of any photocatalyst for any chemical reaction. Most importantly, these measurements allow us to ascertain the optimal distribution of light to maximize photocatalytic activity.

• The paper introduces pump-pump photocatalysis, a time-resolved operando method that uses variable pulse delays and product-rate ratios to distinguish photothermal responses from nonthermal catalytic dynamics.
• The study finds that resonant excitation of Cu-Ru catalysts enhances H₂ production by up to 11× at a 1 ps delay, supporting hot-carrier-mediated H₂ desorption below the thermal desorption temperature.
• The results identify low intensity, moderate temperature, near-resonant excitation, and sub-nanosecond pulse spacing as conditions that maximize nonthermal effects while showing that photon timing can improve photocatalytic efficiency.

Pump-pump photocatalysis (P3C) is introduced as a time-resolved, in operando diagnostic that separates photothermal from nonthermal contributions to plasmonic catalysis by exploiting their characteristic timescales. The method splits each femtosecond laser pulse into two spatially and energetically equivalent pulses separated by a variable delay $\Delta t$, illuminates a working reactor, and records product formation versus delay via gas chromatography. A rate ratio $R_{product}(\Delta t) = r(\Delta t)/r(0)$ quantifies temporal nonlinearity: $R = 1$ indicates a linear (photothermal-dominated) response, while deviations from unity reveal pulse-to-pulse correlation arising from nonthermal mechanisms on sub-picosecond to nanosecond timescales.

Experimental system

The demonstration uses NH$_3$ decomposition over a Cu-Ru antenna-reactor photocatalyst — 7.5 nm Cu nanoparticles bearing dilute Ru reactor sites (97.5:2.5 atomic ratio), synthesized by co-precipitation and characterized by HAADF-STEM, XRD, XPS, and diffuse reflectance. Excitation comes from a wavelength-tunable Ti:Sapphire/SHG system producing ~150 fs pulses at a 12.5 ns repetition period, with intensities of 35–145 W/cm² and controllable external temperature ($T_{ex}$ = 10–200 °C), NH$_3$ partial pressure, and pulse asymmetry $f_1$. The reaction was chosen for its well-established kinetics and minimal side reactions.

Distinguishing thermal and nonthermal responses

The central control experiment contrasts excitation near resonance ($\lambda_{ex}$ = 450 nm, close to the optimal 475 nm) with off-resonant excitation (700 nm). Off-resonant illumination yields $R_{H_2}(\Delta t) = 1$ at all delays — a purely linear, photothermal response. Near-resonant excitation produces a nonlinear enhancement peaking between roughly 1 ps and 100 ps, reaching $R_{H_2} \approx 1.7$ (70% more H₂) despite nearly identical measured temperatures (375 °C vs. 370 °C). Because the effect appears only under resonant conditions at matched temperature, it cannot be attributed to heating alone.

The only observed case of suppression ($R_{product}(\Delta t) = r(\Delta t)/r(0)$0, time-independent) occurs when Ru is removed from the catalyst, consistent with Sabatier-principle arguments that Cu alone cannot retain reactant for the second pulse.

Power regimes and threshold behavior

Rate-versus-power measurements reveal two linear regimes, each described by $R_{product}(\Delta t) = r(\Delta t)/r(0)$1, where $R_{product}(\Delta t) = r(\Delta t)/r(0)$2 is a threshold power and $R_{product}(\Delta t) = r(\Delta t)/r(0)$3 an efficiency in photons per molecule. Three insights follow directly: substantial "precursor" photons are consumed merely preparing the catalyst; above threshold, converting photons are used at constant efficiency; and the power dependence remains linear even where the delay dependence is strongly nonlinear.

Delay (ps)	$R_{product}(\Delta t) = r(\Delta t)/r(0)$4	$R_{product}(\Delta t) = r(\Delta t)/r(0)$5 (mW)	$R_{product}(\Delta t) = r(\Delta t)/r(0)$6 (°C)	$R_{product}(\Delta t) = r(\Delta t)/r(0)$7	$R_{product}(\Delta t) = r(\Delta t)/r(0)$8 (mW)	$R_{product}(\Delta t) = r(\Delta t)/r(0)$9 (°C)
0	132	44	121	5	63	185
1	45	40	108	3	58	168
10	75	40	108	4	59	172
200	85	40	108	4	48	168

The low-power regime has lower thresholds but poor efficiency (45–132 photons per H₂); the high-power regime is more than ten times more efficient (3–5 photons per H₂) but requires higher onset temperature. Correlating $R = 1$0 with known Ru surface chemistry — dissociative NH₃ adsorption beginning near 127 °C and associative H₂ desorption near 177 °C — the authors assign the low-power regime to photothermal activation sufficient for N–H bond scission but insufficient for H₂ desorption, which poisons the surface. The transition to the high-power regime coincides with thermally enabled H₂ desorption, after which efficiency becomes independent of whether heat is supplied optically or externally.

Nonthermal hot-carrier desorption

The key mechanistic claim is that resonantly generated hot electrons desorb H₂ below the thermal desorption temperature, clearing poisoned Ru sites. Several observations support this:

• Thermal hypothesis ruled out: fitting the nonlinear rates with a constant 1.2 eV activation energy would require transient temperature rises of 10–30 °C, roughly ten times larger than recent estimates for this system, and the effect vanishes entirely for off-resonant excitation at identical $R = 1$1.
• Sharp 1 ps feature: at the lowest intensity (60 W/cm²), a narrow enhancement peak at $R = 1$2 ps reaches $R = 1$3 at ambient temperature and 11 for the chilled catalyst — the paper's most striking numerical result, attributed to the second pulse generating hot electrons precisely when those from the first pulse are most effective.
• Broad sub-nanosecond enhancement: a slower nonlinear response ($R = 1$4–6) persists across most sub-ns delays, including a feature near 200 ps when the lattice is already cooling; the authors tentatively assign this to diffusion of adsorbed NH₃ from the Cu antenna to cleared Ru sites.
• Reaction order: dark thermocatalysis shows zeroth-order dependence on $R = 1$5 (N₂ desorption-limited), whereas illuminated measurements show first-order behavior ($R = 1$6) at nearly all delays, confirming NH₃ dissociative adsorption as the rate-determining step. At $R = 1$7 ps in the ambient low-power regime, the order becomes second order — evidence of severe reactant starvation caused by highly efficient hot-carrier-mediated H₂ removal.
• Pulse asymmetry: enhancements are maximal for balanced pulses ($R = 1$8) and asymmetric about $R = 1$9 at $_3$0 ps, favoring a weaker first pulse and stronger second pulse. This ordering dependence confirms the sequence: first pulse dissociates NH₃, stronger second pulse drives nonthermal H₂ desorption.

Limitations and open questions

The mechanistic assignments rest on several assumptions. The transient-temperature analysis assumes a constant activation energy of 1.2 eV, and the 200 ps feature is assigned to antenna-to-reactor diffusion without direct spectroscopic confirmation. The sharp 1 ps feature is deferred to a separate manuscript, so its full mechanism is not established here. The technique measures average surface temperature $_3$1 rather than instantaneous local temperature, so quantitative separation of thermal transients remains indirect. Whether P3C's conclusions generalize beyond NH₃ decomposition on Cu-Ru antenna-reactors to other reactions and catalyst architectures is not demonstrated within this work.

Conclusion

This paper establishes pump-pump photocatalysis as a practical, broadly applicable method — requiring only a delay stage and variable-reflectivity mirror — for temporally resolving photothermal and nonthermal contributions in operating photocatalysts. Applied to NH₃ decomposition on Cu-Ru, it demonstrates that delaying the second of two identical pulses enhances H₂ production by up to a factor of eleven at fixed photon number and temperature, identifies hot-carrier-mediated H₂ desorption as the responsible mechanism in the low-power regime, and delineates the conditions (low intensity, moderate temperature, near-resonant excitation, sub-nanosecond delays) under which nonthermal effects dominate. The results also imply a practical design rule: the temporal distribution of incident photons is itself a controllable parameter for maximizing photocatalytic yield.