Adjusting the catalytic properties of cobalt ferrite nanoparticles by pulsed laser fragmentation in water with defined energy dose

Highly active, structurally disordered CoFe2O4/CoO electrocatalysts are synthesized by pulsed laser fragmentation in liquid (PLFL) of a commercial CoFe2O4 powder dispersed in water. A partial transformation of the CoFe2O4 educt to CoO is observed and proposed to be a thermal decomposition process induced by the picosecond pulsed laser irradiation. The overpotential in the OER in aqueous alkaline media at 10 mA cm−2 is reduced by 23% compared to the educt down to 0.32 V with a Tafel slope of 71 mV dec−1. Importantly, the catalytic activity is systematically adjustable by the number of PLFL treatment cycles. The occurrence of thermal melting and decomposition during one PLFL cycle is verified by modelling the laser beam energy distribution within the irradiated colloid volume and comparing the by single particles absorbed part to threshold energies. Thermal decomposition leads to a massive reduction in particle size and crystal transformations towards crystalline CoO and amorphous CoFe2O4. Subsequently, thermal melting forms multi-phase spherical and network-like particles. Additionally, Fe-based layered double hydroxides at higher process cycle repetitions emerge as a byproduct. The results show that PLFL is a promising method that allows modification of the structural order in oxides and thus access to catalytically interesting materials.

The TEM micrographs illustrate how the morphology of particles changes within the PLFL process ( Supplementary Fig. S1). Whereas the educt particles have a crystal-like shape, the product particles are spherical. In the product after the tenth passage, we cannot find any crystal-like particles, but we can find them in that after the fifth passage. Conclusively, a repetition of somewhat between five and ten passages is needed to induce a transformation of each educt to product particles. We expect that this is due to a not complete illumination of the colloid which is caused by optical refraction of the laser light at the border of the liquid jet.
However, the morphology change does not stop after the tenth PLFL passage. The amount of very small particles is high in the product after the tenth passage, and later on at higher passage numbers, those particles seem to melt together and fuse with bigger particles. We can nicely see the result of this effect in the micrograph of the product after the 50 th passage. In addition, sheet structures are visible in micrographs of the product after the 25 th passage (near the right border) and after the 50 th passage (top left corner). We will regard their formation in detail later in this supporting information.
From the histograms in Supplementary Fig. S1, we can clearly see that the surface contribution of very small particles increases drastically up to the product after the tenth passage. The broad size distribution of the educt tightens, but we can still find some spheres with more than 100 nm in diameter in all products, expect in that after the 50 th passages, but this is possibly due to the method of measuring sizes by a few TEM micrographs per sample.
The formation of the spheres we can relate to melting processes of agglomerates, which explain the perfect spherical shape.
We show an additional elemental mapping by energy dispersive X-ray spectroscopy (EDX) of a large section of a TEM grid coated with the product after the tenth passage of PLF in Supplementary Fig. S2. It contains hundreds of bigger spheres that exhibit a homogeneous distribution of Co, Fe and O. By investigation of the ratio of Co to Fe of single particles, we find occasional deviations in stoichiometry of around 1:1 compared to the expected ratio of 1:2. This affects approximately one of 100 investigated particles. Considering the high amount of crystalline CoO (around 60 vol.-%) in that sample, it is highly unlikely to have no other Fe phase present in the particles. We will discuss the role of Fe in the decomposition of CoFe 2 O 4 in detail in the sections of PXRD and the formation of LDH. Figure S1: HR-TEM micrographs and particle size histograms, volume and surface weighted, for the educt and products after first, fifth, tenth, 25 th and 50 th passage. Molten network structures as well as LDH sheets (mainly in products after 25 th and 50 th passage) are not taken into account for histograms. Figure S2: Dark field (DF) micrograph via TEM of the product after the tenth passage and EDX-mappings of Co, Fe and O.

Surface area by BET method
To determine the effect of the strong morphology change observed by the TEM analysis on the potentially available surface for catalytic interactions of the samples, we measure the specific surface area by the established method of Brunauer, Emmett and Teller (BET) for nitrogen adsorption nS1 . Beside the educt, we choose two products of PLFL, after the first and tenth passage, for the BET analysis. The results are 32, 30 and 42 m²/g, respectively. The BET surface areas do not show a clear dependence on the number of applied PLFL passages. At least for the product after the first passage of PLFL, we would have expected a BET surface area slightly higher than that of the educt according to the particle size distribution from the TEM analysis. A possible reason for the obtained result could be agglomeration of the particles during drying, which should be much stronger for the laser-treated particles due to the decomposition of ligands on the educt particle surfaces. The latter increase in the BET surface area for the sample after the tenth passage could be explained by the strong size reduction caused by PLFL up to that passage. However, the BET surface area does not correlate to the change of the electrochemical properties of the samples in this case.

Powder X-ray diffraction (PXRD)
We investigate the transition of crystalline phases during PLFL processing by PXRD, as described in the main text and shown in Fig. 4. A partly transition of CoFe 2 O 4 to CoO directly in the first passage is obvious. We calculate the lattice strain for the CoO-related refraction peaks to check whether they could be also connected to FeO. If we assume the refraction to originate from a CoO lattice, the strain of the lattice parameter a would be -0.28 % compared to a bulk CoO S2 . In comparison to a FeO bulk, we would get a strain of -1.64 % S3 . Both strains are compressions, whereas a CoO lattice is more likely due to the by six times lower compression. However, both relative strains a quite low and an exclusion of FeO or Fe x Co 1-x O as crystalline phase only by PXRD data is difficult. In addition, the lattice parameter a of the CoFe 2 O 4 phase has a strain of 0.08 % S4 . (We averaged the strain for all phases over all clearly assignable refractions and of all products without seeing a trend to the PLFL passage number.) The CoO phase reaches a saturated volume amount (relative to CoFe 2 O 4 + CoO) of 0.61 after the tenth passage.
In Supplementary Fig. S3a, we present the raw diffractograms of proceeded PXRD measurements, not background corrected and measured without monochromator. The background contains information about fluorescent species that are excited by the used Cu Kα radiation of the diffractometer, which is the case for Co and Fe, and x-ray diffraction at noncrystalline species S5 . Thus, the high backgrounds of all our measurements are not surprising because of the fluorescence active Fe and Co, but interestingly the slope of the background changes significantly already after the first passage of PLFL and changes back after the fifth passage. In products after higher passage numbers, the slope is not varying that much again.
The sample amounts and measuring time are comparable for all measurements. This is not turning the background into a source of quantitative information, but a by the half higher background intensity at high 2 θ values in the product after the fifth passage compared to the educt and product after the first passage is not negligible. Thus, we have probably two different varying background contributions, one affecting the intensity at small and one that of high 2 θ values. For further investigation, we centrifuge a product after the tenth PLFL passage and measure the dried sediment ( Supplementary Fig. S3b). (We chose the product after the tenth passage because of the reached saturation in investigated extinction data (Fig. 1b/c).) The background falls down drastically in the range of small 2 θ values but stays unchanged at high ones. Conclusively, species affecting the intensity at small angles are removed during centrifugation and those affecting the intensity at higher angles remain. The centrifugation parameters are suitable for selectively removing the low density LDH and ion complexes from the sample. The success is confirmed by UV-vis extinction investigation ( Supplementary Fig. S3b) and TEM ( Supplementary Fig. S5). The supernatant is not containing any particulate matter but shows a sharp absorption shoulder in the UV range indicating the presence of Fe and/or Co species. Conclusively, we assign the background contribution at small angles to LDH precursors. At higher angles, we expect the small particle fraction, which is amorphous, to mainly contribute to the background. Otherwise, we also investigate a product after the tenth passage of PLFL in acetone by PXRD ( Supplementary   Fig. S3b). The diffraction pattern looks similar to the one of the sample fragmented in water.
A formation of LDH and its precursors in acetone is unlikely since the solubility of Fe and Co is strongly limited compared to water. However, the amount of CoO in the sample is comparable to the sample produced in water. This result is controversial and requires further studies for clarification.
We further investigate the effect of the applied PLFL process on powders of CoO Figure S3: Raw PXRD patterns of CoFe 2 O 4 PLFL samples after different passages of PLFL (a), comparison of a product after tenth passage before and after centrifugation (dried sediment) as well as of a product after tenth passage of PLFL in acetone (b), as well as of CoO (c) and Fe 2 O 3 (d) before and after tenth passage, respectively. The inlet of (b) shows the UV-vis extinction spectra of the supernatant of the centrifuged sample.

Formation of layered double hydroxide (LDH)
As mentioned in the main manuscript and before in this supporting information, we want to discuss the formation of observed LDH species more in detail here. Supplementary Figure S2 shows TEM micrographs of different LDH sheets found in products after 25 th and 50 th passage of PLFL. The sheets clearly grow with number of passages and have dimensions of more than 200 nm in the product after the 50 th passage. The contrast is also strengthening. Thus, we can assume that the number of layers is also increasing. We can find single LDH sheets at first in the product after the tenth passage, and it is possible to remove them from the colloid by centrifugation, as already mentioned. Of one LDH sheet of the supernatant, we proceed an EDX element mapping, shown in Supplementary Fig. S3. Furthermore, this figure shows the atomic lattice of one LDH sheet we found in the product after the 25 th passage of PLFL. This LDH sheet mainly builds up on Fe and has only a weak cobalt Co signal in the range of the noise. The lattice is clearly hexagonal with the lattice parameter a = (2.9 ± 0.1) Å. This is in good agreement with common Fe-based LDH compounds S7 . A partial reduction of Fe, needed for LDH formation, can be induced by the laser irradiation as recently shown by Ishikawa et al. S8 . Figure S4: TEM micrographs of products after 25 th (left) and 50 th (right) passage containing representative LDH sheets. Figure S5: HR-TEM micrograph of a LDH sheet found in the product after 25 th passage (left) and EDX element mapping of an LDH sheet found in the supernatant of a centrifuged product after the tenth passage (right). The inlays of the micrograph on the left show the real and reciprocal lattice. The lattice is clearly of hexagonal type with a lattice parameter a = b = (2.9 ± 0.1) Å. Mappings of the kα emission of Co, Fe and O on the right exhibit a homogeneous distribution of Fe in the LDH sheet while the signal of Co is not significant and that of O already quite high in the background.

Supplementary
The decomposition of CoFe 2 O 4 to CoO releases Fe that converts at least partly into the observed LDH phase. We expect that Fe species solve into the surrounding water during the decomposition and are present as oxyhydroxide species at the pH of 8.6, we measure for the colloids. Laser energy that couples into the liquid triggers the formation and growth of the LDH sheets. Synthesis routes for LDH are manifold and can be proceeded at different pH values starting at various precursors S9 . Hunter et al. published a laser synthesis of LDH comparable to our expected formation mechanism S10 . They irradiated an alkaline metal salt solution containing micron scaled grains of another metal with a laser to synthesize bimetallic LDH. In the product after the 50 th passage, the LDH phase shows up in the recorded PXRD diffractogram (Fig. 4). The two main refraction peaks are related to [003] and [006] plane distances S11 . Based on the refraction angles, we calculate the interplane distances and subsequently the lattice parameter c as 22.87 Å. It is best fitting to a Fe-based LDH with incorporated carbonate anions S7 . In products after lower investigated passage numbers than the 50 th , the volume amount of LDH is apparently lower than the determination limit of the diffractometer.

Electrochemical analysis
Beside the correlation of electrochemical and material properties presented in the manuscript, we also proceed additional experiments to obtain further information about the mechanism of the activity improvement. We already presented those experiments shortly in the XRD part of the supplementary information. The first one is the centrifugation of a product after the tenth passage of PLFL to remove LDH from the sample, the second one is the PLFL of a CoO dispersion for also ten passages, and the third experiment is the PLFL of the CoFe 2 O 4 educt in acetone instead of water for ten passages to prevent the formation of LDH. We need to note here that all these experiments have their limitations in comparability to the main experiments. The centrifugation leads to agglomeration and aggregation of the nanoparticles.
Because of the quite low density of CoFe 2 O 4 and the very small particle size, high g-forces need to be applied. For the other two experiments, we need to mention that changing the to be fragmented material or the liquid environment will strongly affect the chemical processes during laser irradiation. In the supplementary XRD section, for example, we show that the

Pulsed laser fragmentation in liquid (PLFL)
At first, we want to present the important parameters of the used laser system. We use an Ekspla Atlantic 1064/532 for all experiments. The laser's beam quality k is 0.77, the pulse length t lp is given as 10 ps, the applied repetition rate R is 100 kHz and the wavelength λ of the used second harmonic 532 nm. In addition, the laser raw beam with a diameter d b at 1/e² of 2 mm is focused at its vertical axis by use of a cylindrical lens with a focal length f of 100 mm. All laser parameters, given in this section, are data of the distributors and have not been measured by us. We measure the laser power with a thermopile sensor as 7.2 W.
The process of pulsed laser fragmentation in a thin liquid jet has some advantages compared to other methods published by the community. One advantage is the possibility of calculating the average number of interactions of a particle or volume unit with the laser beam in one passage of PLFL and based on this the specific energy intake of the particulate matter. In a first step, we calculate the outflow velocity v o of the liquid jet on the colloid volume V c , the circular capillarity cross-section area A j , that is expected to be the cross-section area of the liquid jet also, and the outflow time to of the specific colloid volume (Equation S1).
Furthermore, the number of pulses N p illuminating a specific outflowing volume fraction during one PLFL passage is determined by the height of the laser spot h s at the liquid jet on its vertical cross-section and R (Equation S2). Caused by the waist of the jet, h s is varying from a value of 48.4 µm (Equation S3) to the focus diameter d f in dependence to the distance to the focus on the axis of irradiation x i . For the calculation of h s the Rayleigh length l R is needed (Equation S4). We perform the calculation of d f by using Equation S5 and it results in 44.0 µm in our case. Further focusing due to refraction at the jet border is ignored here, because of the small incident angle of 0.4 °. We approximate the value for h s to 46 µm in further calculations.
The results for v c and N p are 500 mm s -1 and 9.2, respectively.
For the calculation of the specific energy intake of each particle during one laser pulse, we determine the irradiated colloid volume V ic first. We simplify V ic to a cylinder with A j as base and h s as height. Actually, V ic should be regarded as the intersection of the circular jet and the elliptic beam cylinder volume. However, the beam ellipsis is just slightly bended in the area of beam penetration of the liquid jet because of the big differing lengths of both half axes of the beam (23 µm and 1000 µm) and therefore regarded as a straight line. The volume concentration c V of CoFe 2 O 4 in water is 0.0094 vol.-% and the measured power difference between an irradiated water jet and the colloid jet is 1.0, 0.8 and 0.6 W for first, second and all other performed passages of PLFL, respectively. A decrease of the power can be triggered by a partly decomposition of solid particle matter to solved ions and/or a less strong shielding effect caused by the decreasing average particle diameter d P . At given R and V c of 6.1 × 107 µm³, the specific energy dose of the colloid is 164, 131 and 98 mJ cm -3 per laser pulse of first, second and each following passage, respectively. After extrapolation to whole passages, by multiplication with 9.2, we receive 1.51, 1.21 and 0.90 J cm -3 . These values relate to the total colloid volume. A relation to the total particle volume is easily done by multiplying the colloid volume with non-percentage c V before dividing the energy amount, because the measured power is already corrected in terms of water absorption due to the differential measuring procedure. Related to the total particle volume, we receive 16.06, 12.87 and 9.57 kJ cm -3 for first, second and each following passage of PLFL or 1.74, 1.40 and 1.04 kJ cm -3 for a single laser pulse during first, second and each following passage.
The energy intake is related to all solid matter in an irradiated colloid volume. To distinguish the absorbed energy of single particles in dependence to their size and position in the liquid jet during the irradiation, we model the energy density distribution within the irradiated jet volume. Our model is based on a raytracing approach and considers one single laser pulse without time resolving. Firstly, we make some assumptions and simplifications to reduce the effort of calculation. The resolution of the model is set to one ray per 5 µm in the horizontal plane. The resolution of the horizontal direction of irradiation is also set to 5 µm and the vertical axis, in liquid jet flow direction, is not resolved. We only show the 5 µm thick center plane of the Gaussian beam in vertical direction. We expect the laser beam profile to be perfectly Gaussian-like in TEM00 mode and the liquid jet to be a perfect cylinder with static and smooth water-air boundary. A distribution of laser's wavelength is ignored as well as light reflection at the inner boundary of the jet. Furthermore, we apply all described experimental settings to the model.
Due to the Gaussian-like distribution, each ray carries another energy. Each ray interacts with the boundary of the liquid jet and the surrounding air, according to Fresnel equation and its form in special case of equated phase factors known as Snell's law. We assume commonly known parameters of air and water here. Within the liquid jet, the single rays spread linear and their energy decreases due to extinction by the colloid according to Beer's law. We extract the colloid's extinction coefficient from UV-vis extinction data at 532 nm. The extinction coefficient is not constant for all passages as shown in Supplementary Fig. S7a. This is of course due to changes in morphology and chemical composition of the colloid. We choose a coefficient of 2.4 × 10 4 cm -1 like it is a god average for first passages of PLFL. The contribution of water to the coefficient is with 3.5 × 10 -4 cm -1 negligible low S12 .
Supplementary Figure S7: Change of extinction coefficient of products of PLFL in dependence to their passage number (a), comparison of absorbed and threshold energy versus particle diameter for different energy density regimes in the liquid jet ( conclusively fusion due magnetic agglomeration, is the predominant mechanism for all particle diameters. The diagram in Supplementary Fig. S7c is supplementary presenting the different cross-sections, geometrical and optical, versus the particle diameter. It clearly shows that absorption is dominating scattering up to a diameter 60 nm. At higher diameters, the contribution of scattering to the extinction is predominant. However, we can conclude that a thermal decomposition of full particles is unlikely and only taking place for particles of higher diameters. Thermal melting of entire particles, on the other hand, is very probable for bigger particles and can also affect the smallest observed particles. Our model results are limited and need further refinements to fit the real conditions during the laser irradiation of a CoFe 2 O 4 nanocolloid. However, the model already predicts the existence of different thermal processes, which products can be find by our applied analytics. We show an overview of the different products generated from the CoFe 2 O 4 educt during PLFL in water in Supplementary Fig. S8. At the top level, we start with the educt particles in the mid which can be partly (to the left) or fully (to the right) melted depending on the absorbed laser energy. Moreover, bigger educt particles completely decompose during high laser energy absorption. Products of this decomposition process (in the mid of the second level) are ultra-small, amorphous particles and Fe oxyhydroxide species. We expect the ultrasmall particles to have a shift in stoichiometry away from CoFe 2 O 4 towards a Co-richness since fully (to the right) melted products contain crystalline CoO. Partly melted (to the left) species at this level are amorphous like the ultra-small particles. At the bottom level, we find a byproduct of the decomposition which most likely forms during laser irradiation of solved Fe oxyhydroxides.
Supplementary Figure S8: Schematic cartoon of laser-induced product generation from CoFe 2 O 4 educt during PLFL. CoFe 2 O 4 species (top level) include the educt and laser-melted species. Laser-decomposition results in ultra-small, amorphous particles and solved oxyhydroxides (middle level). Further laser irradiation of decomposition products leads to crystalline CoFe 2 O 4 /CoO particles and amorphous network structures due to melting and Fe-LDH sheets.