Laser-driven growth of structurally defined transition metal oxide nanocrystals on carbon nitride photoelectrodes in milliseconds

Fabrication of hybrid photoelectrodes on a subsecond timescale with low energy consumption and possessing high photocurrent densities remains a centerpiece for successful implementation of photoelectrocatalytic synthesis of fuels and value-added chemicals. Here, we introduce a laser-driven technology to print sensitizers with desired morphologies and layer thickness onto different substrates, such as glass, carbon, or carbon nitride (CN). The specially designed process uses a thin polymer reactor impregnated with transition metal salts, confining the growth of transition metal oxide (TMO) nanostructures on the interface in milliseconds, while their morphology can be tuned by the laser. Multiple nano-p-n junctions at the interface increase the electron/hole lifetime by efficient charge trapping. A hybrid copper oxide/CN photoanode with optimal architecture reaches 10 times higher photocurrents than the pristine CN photoanode. This technology provides a modular approach to build a library of TMO-based composite films, enabling the creation of materials for diverse applications.


Supplementary Figures
2 Supplementary Fig. 1 Characterization of carbon nitride films. S8 3 Supplementary Fig. 2 Preparation of donor slides and acceptor slides. S9 4 Supplementary Fig. 3 488 nm LTRAS machine setup showing the different components. S10 5 Supplementary Fig. 4 Corresponding laser parameters in the 488 nm LTRAS setup for the gradients. S11 6 Supplementary Fig. 5 Characterization of thin CN layer after transfer. S13 7 Supplementary Fig. 6 Survey, C 1s, and N 1s XPS spectra of CN and CuO/CN composite films. S14 8 Supplementary Fig. 7 Evidence for rapid (millisecond) synthesis process with LTRAS. S15 9 Supplementary Fig. 8 Thickness profile and 3D view of the spots with 0.132 mW/µm² laser power density and 40, 30, and 20 ms irradiation. S18 10 Supplementary Fig. 9 Proposed growth mechanism for laser transfer. S19 11 Supplementary Fig. 10 High-resolution TEM images of CuO and proposed mechanism for the laser-driven transfer synthesis. S20  Fig. 21b) and EDX mapping show that no CuO was formed in this process even though the effective laser power was more than 70 times higher than that in the LTRAS method (1.9 W @ 450 nm vs. 25 mW @ 405 nm). These results indicate that the wellestablished photodeposition method needs much more energy and/or time than the LTRAS method. Therefore, we switched to a UV light approach together with a longer irradiation time used as the counter electrode, reference electrode, and electrolyte respectively.

IPCE measurements
A series of fiber-coupled LEDs (SMA, ThorLabs) with wavelengths of 415 nm, 455 nm, 530 nm, 625 nm, and 850 nm were connected to the LED Driver (DC2200, ThorLabs) and used as light sources for the IPCE measurements. For each wavelength, we studied the dependence of current density versus intensity of the incident light (5 data points) ( Supplementary Fig. 27).
The current density depends linearly on the intensity of the incident light. The slope of linear fitting corresponds to the ratio between JPh and JLight.

Glucose detection by amperometry
The response of the pristine CN and the composite CuOF1/CN electrodes to glucose was evaluated using amperometry by successively adding small amounts of concentrated glucose solution. In detail, the phototelectrochemicial performance was assessed using a three-electrode We use continuous wave (cw) laser setups in the line scanning mode with speeds between 16.7 -60 µm/ms. Usually, the differentiation between pulsed and cw mode is around 10^-6 s, where ablation effects start to dominate, which is why we consider our approach cw, although we not only used line scanning, but also ms pulses for the experiment in Our analyses corroborate that the complete process/growth time is in the millisecond regime.
As described before, the local laser interaction is in the range of (sub-)milliseconds (0.33 -

ms). The energy is absorbed in the µm thick polyimide (PI) film. The polymer matrix with
the precursor is only about 1 µm thick, thus axial (in laser beam direction) heat diffusion can be neglected. The process is thus dominated by thermal gradients induced in the PI film.
In the main Figure 3a,b we have shown a typical thermal gradient induced by a local laser irradiation for a comparably long laser irradiation (20 -40 ms). The laser heating in axial direction can be assumed as nearly instantaneous. Yet, laterally, it will take some milliseconds to heat up the material.
Once the laser irradiation stops, the polymer quickly cools down again. This happens quite fast due to the heat capacity of the acceptor slide.
To prove this, we followed a spot irradiation with a high-speed camera ( Supplementary   Figure 7), showing that the polymer matrix, which is irradiated for 50 ms with a laser power of S16 25 mW (405 nm laser) cools down below the melting temperature of the polymer of 210 °C within less than 3 ms.
The optical appearance of a solid polymer film solidifying during the film preparation from solvent (the morphology is characterized with small dewetting holes and other drying effects) and the same film solidified (frozen) due to a temperature related phase transition can easily be distinguished.
The frozen organic film on the other hand is harder to distinguish from the liquid phase, since it rapidly freezes the molten morphology in place. If it would freeze more slowly, other characteristic morphologies can appear. [1] The polymer solidification is measured by analyzing the local brightness value inside the molten spot. 3 ms after the laser is turned off, there is an increase by 10 %, which only varied by 1 % during the next hundreds of ms (data only shown for 52 ms time point). Since the laser wavelength is filtered out in the optical microscope, we ruled out that it could be the laser (in fact, the brightness value increases after the laser is off, thus it cannot be the laser irradiation, which would cause a decrease of the grey value upon switching off).
Thus, we conclude that after a 50 ms long laser irradiation, the molten polymer spot requires only about 1 -3 ms to cool down below its melting temperature of around 210 °C after heating to about 400 -500 °C in the center.
This means, for a shorter irradiation time, the cooling time should be even shorter. Taking this into account, total process time (limited by the molten polymer) is in the range of about 1 -5 ms.
Once the polymer is "solid" again, the CuO cannot freely move and nucleate anymore.
Concluding, the LTRAS mechanism for the formation of the CuO is based on the phase transition of the polymer matrix. As long as the polymer is "frozen/solid" the CuO cannot form.
Once it melts, the CuO can form and nucleate. Since heterogeneous nucleation is always favored, this will happen at the acceptor surface interface. Additionally, the cooler interface will lower the energy barrier and favor nucleation. Considering that the process window is S17 defined by the polymer in its liquid state, the process needs a temperature above 210 °C. Thus, the process is very fast and in the millisecond regime. Supplementary Fig. 8 Thickness profile and 3D view of the spots with 0.132 mW/µm² effective laser power density and 40, 30, and 20 ms irradiation. Supplementary Fig. 9 Proposed growth mechanism, based on three steps: (1) precursor decomposition, (2) growth at the interface, and (3) solubility differences in the molten vs. solid polymer.
Supplementary Fig. 10 (a) High-resolution TEM images of CuO. The inset shows the diffraction pattern with selected area electron diffraction (SAED). (b) Proposed mechanism for the laser-driven transfer synthesis.
To reveal the mechanism of laser-driven synthesis, high-resolution transmission electron microscopy (TEM) and selected area electron diffraction (SAED) were performed for the LTRAS-obtained CuO. In these images, a highly ordered crystalline structures of CuO can be clearly observed. The fringe spacing is measured to be approximately 0.35 nm, which corresponds to the [010] lattice fringe of the monoclinic CuO. This result suggests that the preferred growth direction of the nanorods is [010]. Accordingly, we propose the following S21 mechanism: The synthesis starts from the decomposition of Cu nitrate, which follows the typical process steps of nucleation and growth. [2] First, during the growth step nanocrystals are formed. In the second stage, the obtained nanocrystals arrange into nanostructures. [3] Here, we propose the well-established theory of 'oriented attachment'. [4] Small crystallites attach to each other via their related crystal facets along the same directions, forming larger crystals. [