Activity and Durability Patterns of 45 Binary Noble Metal Alloy Nanoparticle Variants for Commercial Diesel Exhaust Aftertreatment

Commercial diesel oxidation catalysis mainly uses monometallic and bimetallic Pt−Pd‐based catalysts, but alloying with different elements has rarely been done systematically under industrial testing conditions. 45 binary alloys made of platinum and a selected 1st and 2nd‐row transition metal, platinum group element, or coin metal were synthesized via a scalable laser synthesis method. Then, catalytic performance and durability were evaluated for one diesel oxidation and two ammonia‐slip environments. The results show the highest activity when the adsorption enthalpy of molecular oxygen of the alloy was similar to the value of Pt. Furthermore, the durability of the alloy catalysts was found to increase with the melting point of the 2nd element Pt was alloyed with, even at molar fractions. Our results further indicate beneficial synergies beyond the binary systems underlining the possibility of further improvements by considering ternary or multinary alloys, which are accessible via laser synthesis.


Introduction
[3][4] Apart from other strategies, car manufacturers use catalytic abatement of waste gas to adhere to regulations. [5]For automotive use, active and durable catalysts are needed to cover a large temperature range (RT-~700 °C) and perform in changing engine conditions with varying gas feed compositions. [6]Commercial diesel oxidation catalysts historically used noble metal nanoparticles (Pt and/or Pd). [6,7]With increasing noble metal prices, there is interest in reducing Platinum Group Elements (PGEs) usage.This can be achieved by alloying with different PGEs or cheaper noble or base metals.This strategy also promises to improve catalytic activity.For example, considering Montemore et al.'s work, it can be estimated that alloying of Pt (or Pd) with Ag, Au or Cu might improve CO-Oxidation capabilities. [8]Looking further at the literature, a lot of work has been put into abatement of NO x [9,10]   and soot emissions, [11] but previous studies on alloy catalysts either only comprised a very limited set of alloys [12][13][14][15][16] or were set far apart from automotive applications (e. g., electrochemical studies or fuel cells [17][18][19][20][21][22][23] ).Also, interesting work has been done in the field of three-way catalysts, [24,25] where multinary alloys have proven their good activity and durability.This points to the need for a systematic study, that focuses on alloyed nanoparticles in diesel oxidation catalysis.Since extrapolating from simple gas mixtures on the lab scale is difficult due to the complexity of exhaust gas catalysis, this study aims to evaluate catalysts close to real application.This is done by screening supported alloy nanoparticle catalysts while preparing them according to industry standards and testing in an industrial setting.Testing will be done in three reaction gas mixtures, one mixture representing a Diesel Oxidation Catalysis (DOC) environment (Feed=CO, C 3 H 6 , O 2 , H 2 O, CO 2 in N 2 ) and two mixtures representing an Ammonia Slip Catalysis (ASC) environment where either NO is present or not (ASC; Feed=NH 3 , (NO), O 2 , H 2 O in N 2 ).Due to the high noble metal content of commercial catalysts for DOC and ASC, this is where alloying can show the most benefit.
With these environments in mind, the alloys are chosen.[28] Typically alloys from Pt and Pd are used, [28] in part because of the excellent activity and durability, that come from Pd inhibiting the Pt particle growth. [29]For this reason, also PtÀ Pd is included in this study.Rh is chosen because it is typical for three-way catalysts [28] and has an activation energy similar to Pt for CO-oxidation, [30] making it a good candidate for high activity.Its high melting point should also aid in making the alloy durable.An alloy of Pt with either oxidation, as the oxides of the base metals belong to the active catalysts. [35]Especially Fe is expected to favour N 2 in the reaction (reducing NO x emissions) as its high activation energy for N 2dissociation [36] prevents N 2 , once formed, from splitting and thus removes nitrogen from the oxidation reaction.PtÀ Cu is also expected to show good CO-Oxidation properties due to the Sabatier principle. [8]Ruthenium is chosen as a promising candidate for CO-Oxidation and N 2 -formation. [36]Generally, its high activity is attributed to a thin oxide layer, which is inactivated through oxide layer growth and water adsorption. [37,38]This layer growth and water adsorption are hypothesized to be inhibited by ennobling Ru with Pt.Ruthenium is also expected to improve durability due to its high melting point, as are alloys from Pt and Nb or Mo.In addition to binary Pt alloys, some other combinations will be studied.PdÀ Cu alloys are chosen, as the Sabatier principle suggests an improvement in CO-oxidation. [34]Furthermore, AuÀ Ir and AuÀ Rh will be tested.Ir and Rh are expected to be moderated by the addition of gold, allowing for an increase in selectivity by an O-spillover from Rh or Ir to Au. [8,39] The precise composition of the alloy is determined mainly by the underlying bulk phase diagram of the mixture with a distinction into two main categories.For all alloys that were expected to show a high CO-oxidation activity (Pt + Pd, Ru, Cu, Au, Ag, Rh and Pd + Cu), a large mixing interval is chosen (10-99 at% Pt) with Pt + Rh, Au + Ir and Au + Rh as exceptions.Due to cost reasons the maximum proportion of Rh and Ir was kept at 50 at% for these alloys.For all alloys where an ennobling by platinum was hypothesized and where durability was the main focus, the minimum proportion of Pt was kept to 50 at% (Pt + Nb, Mn, Mo, Fe).Detailed compositions can be found in S1.
In total this screening study comprises 13 alloy systems with overall 45 different compositions.For this highly diverse sample setup, a high-throughput alloy nanoparticle synthesis method is needed.Rather than develop a different wet-chemical synthesis strategy for every alloy system, synthesis by laser ablation in liquids was chosen.][48][49][50] Also, this method is able to provide high amounts of nanoparticles in the gram scale, [12,51,52] which is needed to fulfil the demands of industrial catalyst testing.For example, Au-Pt alloy nanoparticle productivity has been demonstrated to robustly yield 8.3 g/h, [52] equivalent to 830 g supported catalyst at 1 wt% loading.
It will be shown that laser ablation synthesis is able to provide sufficient amounts of the 45 different alloy nanoparticle compositional variants, as a basis for a demanding industrial screening study for waste gas abatement of diesel combustion engines.

Result evaluation
For an evaluation of the catalysts, various single-point analyses of the spectra are used, which will be briefly presented below and in Figure 1.

DOC -CO-and HC-Oxidation
Since the main task of the diesel oxidation catalyst is to completely oxidize carbonaceous exhaust gases, the DOC test first determines the light-up temperature, at which a turnover (of CO or propene) of 50 % is achieved (LU 50 ) (cf. Figure 1 a) and b)).The lower the measured temperature, the better the catalyst is to be classified.In some samples, when temperatures rise, CO is formed due to the oxidation of propene.As a result, there is sometimes more than one crossing point at 50 % turnover.For samples where this is the case, the area is indicated by an error bar.

DOC -Nitric-oxide formation
In addition to the oxidation of the carbon-containing exhaust gas components, the diesel oxidation catalyst stage is intended to help provide NO and NO 2 in a 1 : 1 stoichiometric proportion to enable a fast-SCR reaction downstream in the motor vehicle.Since this cannot be guaranteed over the entire temperature range for thermodynamic reasons, the efficiency of the catalyst is mainly measured by how much NO 2 is formed at 250 °C.At this temperature, the equilibrium is still on the side of NO 2 , so it can be expected that a lot is formed in the presence of a good catalyst.The measured NO 2 concentrations are normalized for application to the NO input concentration (eq.( 1)).According to this equation, the percentage of NO 2 -content at 250 °C is evaluated (cf. Figure 1 c)).
As a side reaction, nitrous oxide can still form in the oxidation catalyst.As an undesirable greenhouse gas, the amount is also measured.The catalysts are then compared based on the average amount of N 2 O formed.For this purpose, the amount of nitrous oxide formed is first integrated over the entire temperature range and finally weighted with the temperature interval (eq.( 2)) (cf. Figure 1 d)).

ASC
The ammonia slip catalyst needs to oxidize surplus ammonia to nitrogen and water.This needs to happen as fast as possible, with as few nitrous oxide by-products as possible.The result evaluation, therefore, centres around the ammonia conversion, which is quantified, just as with COÀ and HC-conversion, as the LU 50 temperature (cf. Figure 1 a and b)).In the measurement NO 2 , NO, and N 2 O are also quantified and evaluated as the mean quantity in the complete temperature interval, as demonstrated for N 2 O in the DOC measurement (cf. Figure 1  d)).Additionally, the amount of formed nitrogen can be obtained.This is also evaluated as an average amount.

Particle characterisation
The particles were analysed as described in the methods section.To illustrate this process Figure 2 shows an exemplary analysis and summarizes particle size distribution and composition data.
The obtained particle size distributions (cf. Figure 2 (a)) show that all samples synthesized by the nanosecond laser at 532 nm wavelength have the peak particle diameter (distribu- tion modus) between 3.4 and 5 nm and at 1064 nm wavelength at about 2 nm diameter, whereas the samples synthesized by the picosecond laser show a modus between 1.8 and 3.8 nm.Overall, the particle size distributions are comparable, with some outliers towards larger particle sizes.The overview of the particle distribution modi can be found in Figure 3 and exemplary STEM images as well as all detailed particle distribution plots with additional information are given in sections S2.1 and S2.2 of the supplementary information.
The sample composition was checked via digestion in ICP-OES and TEM-EDX (cf. Figure 2 (b, c)) and S2.3 of the SI.While the TEM-EDX line scans mostly confirm the alloying of materials, the overall compositions differ from the results of ICP-OES in many cases.This difference is attributed to the low sample size in TEM-EDX of mainly 12 scans.For this reason, the data from ICP-OES is used for evaluations where sample composition is relevant.Additionally, the catalyst loading could be obtained from ICP-OES measurements.The nominal loading of the samples could be achieved to within 0.1 wt.% except for four samples (PtÀ Ru and AuÀ Ir).The large observed divergence, in this case, may be attributed to sedimentation issues that prevented accurate measurement of the catalyst loading.

Catalysis
The tested alloys have been chosen, to show two different properties.Firstly, the catalytic activity should be evaluated.Secondly, the metals were chosen to show a trend in catalyst ageing resistance.The catalyst stability was hypothesized to correlate to the melting point of the alloy-metal (i.e., refractory metals are better than non-refractory metals when only catalyst stability is considered).

DOC
Regarding the overall activity, the conventional PtÀ Pd system outperforms all other catalyst alloys.This system is closely followed by PtÀ Cu, which is especially good in the fresh state and shows remarkably little dependency on the amount of copper added.With alloys comprising 90 % copper still performing as well as alloys containing only 10 %.As a noble-metal combination, Pt x Rh 100-x also shows high activity for x � 90.Even though rhodium should show good oxidation activity on its own, alloys with high rhodium content drastically decrease in activity.Other noteworthy alloys are PtÀ Mo and PdÀ Cu.Both mainly because of their durability and for PdÀ Cu also the low dependency on mixing ratio.As is to be expected in a screening study, some alloys did not show appreciable activity at all.These alloys mainly contain gold (AuÀ Ir, AuÀ Rh, PtÀ Au) and the PtÀ Ag system also shows signs of inactivity, though not as strong as the aforementioned three systems.The observed trends with the CO oxidation, continue, though less obvious, in the evaluation of HC oxidation and NO 2 and N 2 O formation.The respective overviews are plotted in S5.2.
To give a better overview of the CO oxidation trends, the activity of the catalysts is evaluated regarding their affinity towards the reactants.The statistical significance of the obtained results is verified by using an ANOVA test.From DFT calculations it is known that the CO oxidation on metal nanoparticles will show a volcano like dependency when plotted over the CO adsorption enthalpy or the O adsorption enthalpy.The peak activity is calculated and found to occur for materials that have about À 1 eV adsorption enthalpy for CO as well as for atomic oxygen. [34]s obtaining the exact enthalpy for all alloys is beyond the scope of this study, the comparison is done based on alloy metal (i.e., Pt + X) and an extensive list of sources can be found in S4.Comparing our data to the literature we find that the dependence on CO adsorption enthalpy only shows a slight scaling relation and that also only for aged samples (cf.S5.3.1).This can be explained by observing the environmental conditions.Apart from other constituents, the catalyst is exposed to an abundance of oxygen.On Pt(111) and for CO:O 2 ratios < 10 % the CO mass transfer to the catalyst is the rate-limiting step. [53]This same rule applies also to platinum group metals. [54]ince the CO:O 2 ratio in this study equals 0.4 %, the limited CO mass transfer should explain the minor observed scaling relation.
When plotting the activity over the O adsorption enthalpy a volcano-like dependency can be found for the fresh samples (Figure 4).The plot using this adsorption enthalpy as the x-axis has several benefits because the activated oxygen is important for all occurring reactions in the DOC environment (i.e., CO oxidation, propene oxidation and NO oxidation [28,55] ).It may therefore show the effect of the combined gas mixture.Therefore, the LU 50 of the CO oxidation is analysed with respect to the adsorption enthalpy of atomic oxygen.
The observed volcano shape suggests an optimum close to the adsorption enthalpy of platinum This dependence disappears for aged samples that contain metals with an absolute affinity to oxygen higher than platinum.The optimum lies 1 eV lower than what would be expected for the CO oxidation from the literature. [34]As this study is working with a complex gas mixture this behaviour can be explained.Our gas mixture also contains NO, water, and propene.Propene, for example, is known to inhibit oxygen adsorption on Pt [28] and also the other gasses can be expected to compete with oxygen for surface occupancy.Since the availability of activated oxygen is important for all oxidation reactions, [28] the optimum should shift in the direction of higher oxygen affinity as observed here.
The high oxygen affinity could also explain, why the scaling relation for the aged samples and materials of a high affinity disappears.A strong bond to oxygen might lead to surface segregation of the base metal.Subsequent oxidation might then lead to slight deactivation of the catalysts.Counterintuitively, the opposite may also happen.It was recently shown, that PdO segregates from PtÀ Pd particles upon ageing under oxidative conditions.The formed janus-partices then enable a self-healing property, explaining the high durability of this alloy. [56]The role of phase-segregation upon oxidation and the formation of mobile oxide species is difficult to quantify but may play a significant role in the correlation to the heats of adsorption.Further in-depth study of the segregation and reconstruction behaviour and possibly measurement of the adsorption enthalpy of these platinum-nanoalloys and quantification of oxide formation especially after aging will help pinpoint the cause of the observed behaviour.Looking at the other sub-reactions (propene-oxidation, NO 2 and N 2 O formation, cf.S5.3), the correlation for metals with a lower affinity towards oxygen than platinum holds strong.Generally, it is obvious in those reactions that silver and gold make the catalyst performance worse.Also, no strong correlation between oxygen affinity and light-up/formation can be found for alloys with a higher oxygen affinity.This suggests, that with those alloys the reaction pathway is not strongly influenced by the adsorption enthalpy of molecular oxygen.
With respect to the durability of the catalysts, another overview plot is generated.In this, the difference in light-up temperature between the fresh and the aged samples (650 °C) is plotted.When alloying platinum with different metals a change in ageing behaviour is suspected.Alloying with metals of lower melting point lowers the overall alloy melting point and should allow for less inhibited material movement on the support surface during catalysis.If platinum is alloyed with metals of higher melting point this behaviour is expected to inhibit the material movement and thus the ageing due to particle growth and subsequent loss of active surface area.When plotting the difference in LU 50 in the CO-Oxidation reaction in dependence on the melting point of the alloy a clear trend is visible (cf. Figure 5).Figure 5 a) shows that the correlation is strongest when the melting point of the metal (Pt + X) with which the alloy is formed is used as the x-axis.If the melting point of the alloy is taken from the bulk phase diagram (Figure 5 b)), the correlation is smaller.This is to be expected, as many alloys in this study contain a high amount of platinum and thus the alloy melting point is situated close to that of platinum.As is in line with the hypothesis, alloys of low melting point (e.g., containing gold, silver or copper) belong to the catalysts that age the strongest, while catalysts with alloy metals of high melting point (e.g., niobium, molybdenum or ruthenium) show the highest durability.The analysis was repeated for the reactions of propene-oxidation and nitric-oxide formations and the same correlations were found.A correlation towards the alloy metal melting point can be found for all reactions except N 2 O-Formation (cf.S5.4).Variations in the durability within the alloy compositions might also correlate to oxidation and the formation of volatile oxides that increase the durability of some combinations (like for PtÀ Pd reported by Porter et al. [56] ) or decrease the durability for others, where no synergistic behaviour exists.
In summary, it may be concluded that when designing a catalyst for durability, the type of element used for alloying can be more important than the amount used.Even small amounts of a high melting point metal can improve resistance against aging.

ASC
In the ammonia oxidation catalysis, well-performing alloy catalysts are mixtures of Pt with Pd, Nb, Ru, Fe, and Mn containing a medium to high proportion of Pt.PtÀ Cu and PtÀ Mo show good activity over the whole tested mixing range.Those two systems also show exceptionally good ageing behaviour.Well-performing systems show NH 3 -LU 50 temperatures of ca.150-180 °C (with NO present) and 190-205 °C (without NO present).The alloys PtÀ Ru, PtÀ Rh, and PtÀ Ag containing a low amount of platinum are less active, while the systems AuÀ Rh, AuÀ Ir, AuÀ Pt and PdÀ Cu are inactive and durable.(Cf.S6.2 and S6.5).
In order to extract patterns, it seems worthwhile to also plot the ASC data depending on an adsorption enthalpy.From literature, it is known that the oxidation of ammonia begins with the adsorption of ammonia and oxygen on acidic and redox sites respectively.This is followed by the abstraction of hydrogen by active O species or OH groups formed either by O 2 activation on redox sites or O vacancies on metal oxides.Generally, the reaction pathway is divided into three distinct mechanisms, the imide-, hydrazine-and i-SCR-mechanism. [57]rom these three, the imide-mechanism is the most prevalent for Pt nanoparticles supported on alumina and needs the O 2 activation as an important step for NH 3 dissociation. [58]It is however notable, that on one catalyst surface, more than one mechanism can occur.Which reaction pathway takes place is dependent, for example, on catalyst support material and reaction temperature. [57]onsidering Pt(100), the rate of ammonia oxidation is mainly dependent on NO desorption. [59]As this study focuses on alloys comprising mainly Pt alloyed with different metals, this would make it prudent, to plot the obtained data over the NO adsorption enthalpy of the respective metal.
When evaluating the LU 50 of the NH 3 oxidation with and without NO present in the feed gas stream, it quickly becomes obvious that no clear correlation towards the NO adsorption enthalpy can be drawn (cf.S6).Just as with the DOC reaction, the common rate-limiting step was not obvious in our study.The complex gas mixtures and the industrial test setting apparently change the common mode of operation, underlining the importance of testing the catalysts in these real-world environments.
Our test regime showed that in the case of ammonia oxidation also the adsorption enthalpy of oxygen plays an important role.Especially, when considering side reactions towards nitric oxides, two major patterns emerge.All systems produce nitric oxides in varying amounts, but either the reaction towards NO or towards NO 2 and N 2 is favoured.
This correlation can be seen best in Figure 6.The negative ANOVA test of the linear fit towards oxygen adsorptions shows that there are two big groups.Pt-alloys with Mo, Ru, Nb, Fe, and Mn form little NO when compared to Cu, Rh, and Ag with the PtÀ Pd alloy being an exception.The positive correlation in the right arch of the aged samples is mainly attributable to the inactiveness of the PtÀ Au alloys and disappears when the worst catalysts are excluded.This leads to the hypothesis that when the adsorption enthalpy for molecular oxygen is equal to or higher than that of rhodium, NO is a favoured reaction product when considering the alloys with Pt.
This analysis has been performed for all reaction pathways, similar to the DOC section.All data is plotted in S6.3 and S6.6.The grouping shows up whether NO is present in the feed gas or not and is also visible when considering the other reaction products like NO 2 , N 2 O, and N, with N 2 O being the least obvious due to a large signal-to-noise ratio in the data.First, this allows the conclusion that the tested catalysts are all robust towards the NO concentration and can be expected to catalyse the ammonia oxidation selectively and independently from NO concentration.
Looking at the literature the grouping behaviour may be loosely attributed to the underlying reaction mechanisms.From Lan et al. we can conclude that the hydrazine mechanism is not relevant to our experiments since it requires low oxygen partial pressures that don't reflect our experiment. [57]The remaining three mechanisms are the imide-, i-SCR-, and N 2 -mechanism.
The i-SCR and N 2 -mechanism explain the observed intermediates for reactions over silver-based catalysts, so for catalysts that have a low adsorption enthalpy for oxygen.Since in the N 2 mechanism no NO is detected, but our catalysts produce it, this mechanism may also be excluded.The observed by-products of the i-SCR mechanism are the same as for the imide mechanism: N 2 , NO, and N 2 O. Considering that NO is mainly found as an intermediate in the imide mechanism, catalysts with a high affinity for oxygen and thus high surface coverage may have a high possibility to over-oxidise the NO to NO 2 .Catalysts with a low affinity toward oxygen would have a lower probability for this and thus mainly yield NO. Summarizing, this means that catalysts with a low oxygen affinity will favour NO as a by-product, while catalysts with a higher affinity will favour NO 2 .This fits the picture we observe, where the border appears to be the adsorption enthalpy of rhodium (cf.eq. ( 3)): It is therefore possible that a platinum alloy may exhibit two different ammonia oxidation mechanisms dependent on the alloy material.
Finally, the durability of the alloy catalysts was also evaluated.While the correlation between ageing and melting point is higher when only the melting point of the alloy element is considered, no significant correlation could be found for either the melting point of the alloy element or of the alloy (cf.also to S6.4 and S6.7).In contrast to the DOC measurement, the ASC ageing of the catalysts does not seem to correlate strongly to the alloy element or composition.This may be due to the influence of particle oxidation and the formation of volatile oxide species, which may, for the ASC environment, equalize the durability shown by the catalysts.

Conclusions
This study aims to broaden the composition-related knowledge regarding waste gas abatement by synthesizing 45 binary alloys of platinum with other metals.[62] The catalysts were tested in an industrial environment with one DOC and two ASC gas mixtures, and two aging regimes.This allowed for identification of composition-activity and composition-durability correlations.
When comparing the tested material systems individually our results show that alloys of Pt with Cu, Mo, or Pd show particularly promising activity correlations in DOC.Despite them being not the most active (especially in aged states), the alloy systems PtÀ Cu and PtÀ Mo are also very notable, as they show a hardly any dependence between activity and alloy composition.They share this behaviour with PdÀ Cu.
In the case of ASC, the most favourable compositions are alloys of Pt with Nb, Ru, Fe, or Mn.All these compositions show a high selectivity for N 2 and have high activity.One very promising alloy constituent was shown to be Cu.This metal showed a high activity whereby the activity of PtÀ Cu exceeds that of PdÀ Cu.Furthermore, alloys from Pd or Pt with Cu show particularly low dependence of the activity and selectivity on composition, making copper a very cost-effective catalyst additive.Pt-Mn and PtÀ Fe alloys showed good activities in the DOC-and the ASC environment with Fe also exhibiting a high selectivity for N 2 .Alloys of Pt and Ru exhibited a high selectivity for N 2 , although high amounts of Pt are required to also provide a good activity (DOC: > 80 at%; ASC: > 50 at%).The ennobling of Ru by Pt seems to work over the whole tested mixing range, with the PtÀ Ru-catalysts exhibiting very good durability in DOC as well as ASC.
Apart from the individual comparison, the data was used to develop composition-activity-and composition-durability-correlations for DOC and ASC.When correlating the catalytic activity with the adsorption enthalpy of molecular oxygen, a volcano trend appears for DOC.Compared to DFT studies, the observed oxygen adsorption enthalpy at maximum activity is shifted by 1 eV towards stronger oxygen binding.This behaviour could be attributed to the more complex gas mixtures present in this study.For the aged alloy catalysts, there is no statistically significant correlation between activity and oxygen adsorption enthalpy when E O;ads . This observation may be linked to element segregation and/or passivation due to oxidation of the catalyst surface, but formation of volatile oxides may also play a role.Regarding the durability in the DOC environment, the catalysts show a statistically significant positive correlation between the change in activity and melting temperature for both the alloy and the element platinum was mixed with.Hereby, the highest durability is observed for elements and alloys with the highest melting point.The data allowed for the conclusion, that the type of element used for alloying can be more important than the amount.For ASC reactions, no correlation was found between ammonia oxidation and NO adsorption enthalpy, contradicting literature, and highlighting the importance of testing catalysts in environments similar to waste gas abatement.In turn, a strong link was found between oxygen adsorption enthalpy and reaction products.Alloy catalysts with oxygen adsorption enthalpy below Rhodium favoured N 2 and NO 2 , while Pt alloys with lower oxygen adsorption enthalpy favoured NO and N 2 O.This was attributed to the imide-mechanism and i-SCR-mechanism.PtÀ Pd was the only exception, favouring N 2 and NO 2 .
In summary, this study shows that the industrial testing conditions are important to the catalyst screening, as some of the observed performance trends were not expected from literature (especially for ASC).It was shown that alloy catalysts have a high potential to reduce the need for noble metals (i.e., PdÀ Cu, PtÀ Cu) but the durability needs to be further optimized.In this regard, improved durability was observed when alloying platinum with refractory metals where particularly low contents were already beneficial.With this in mind, ternary alloys (e. g., PtÀ CuÀ Mo or PtÀ CuÀ Nb) appear as a sensible next step for further studies.With laser synthesis being a mature and wellestablished synthesis method, multinary alloy catalysts can be obtained similar to this study. [50,63,64]Multinary alloys consisting of Pt and (a combination of) Cu, Mo, Nb, or Mn to tune activity and durability appear possible.Such complex high-entropy alloys have found recent attention in three-way catalysis [24,25] (λ = 1) and multinary alloy catalysts designed with the knowledge from this study may be interesting to apply for λ > 1 environments.

Colloid synthesis
The tested catalysts were synthesised from independently produced mixed-metal targets.These targets were pressed from base metal powder and subsequently sintered.Additional detail on the target production process can be found in the supplementary information (cf.S1.1).
The mixed-metal targets were used to synthesise colloids via pulsed laser ablation.Noble metal nanoparticles were synthesised with water as the medium, and base metal combinations were ablated using propylene carbonate as the medium.The latter was selected as it is a technical liquid that has a very low vapour pressure allowing safe handling and is thermostable which would allow recycling via vapour condensation.The ablation was done using two different laser systems.Generally, a nanosecond pulsed laser system (Edgewave IS160-1-T) with 532 nm wavelength and 5 kHz repetition frequency and 50 W of power was used to ablate the water-based samples (Max.ablation rate: ~488 mg/h, i. e., ~11 mg/Wh).Ablation in water was assisted with adding micromolar amounts of additives (KOH, Sodium citrate or Ammonium citrate).The Figure 7. Experimental design of this study.After the decision on the systems (Pt + X, green; Au + X, blue), the elemental powders were mixed, pressed, and sintered under appropriate atmosphere to obtain alloyed targets.Targets were ablated in a liquid with a pulsed laser.The colloids were supported on alumina powder and prepared for catalysis measurement.The exact procedure differs for samples produced in water (a) and propylene carbonate (b).Performance testing was done with automotive gas mixture and test equipment.organic solvent-based samples were synthesised using a picosecond pulsed laser system (Amphos Poly 500).This system was used at 1030 nm wavelength, 5 MHz repetition frequency and 467 W of laser power before duty cycling (Max.ablation rate: ~1100 mg/h, i. e., ~14 mg/Wh).More details are described in the supplementary information, section S1.2.

Colloid immobilization
After obtaining the colloids the nanoparticles needed to be immobilized on a support material.For this purpose, pure aluminium oxide was chosen, so that no promoters would interfere with the catalytic reaction.The aluminium oxide powder was either wet-milled in an acid environment (pH 4, for water-based samples) or jet-milled to attain a highly disperse powder.
The water-based alumina was pH-adjusted and mixed with the dispersed colloids by pumping the colloid slowly into the alumina slurry and a pH-driven adsorption strategy was employed to adsorb the nanoparticles on the support surface.Sedimentation was then used to concentrate the samples to approx.150 g/l for use in dip casting.
For the propylene carbonate-based samples a pH-driven adsorption strategy could not be used.Therefore, the colloids were mixed with the support powder, which was dispersed by strong ultrasonication.While stirring vigorously under a vacuum, the propylene carbonate was gently distilled off.This causes diffusion-driven supporting, resulting in dry catalyst powder (Figure 7).A more in-depth description of the support-ing routes can be found in the supplementary information (cf.S1.3).

Particle analysis
Prior to catalytic testing, all samples were characterized in detail for particle size (STEM, JEOL JEM-2200 FS), composition (EDX (JEOL JEM-2200 FS), ICP-OES (Varian 725ES, microwave pressure digestion)) and loading (ICP-OES).Information on the composition was used to correct the measured data for deviations from the nominal composition values.
Particle size measurements were carried out using STEM images of diluted and dried washcoat.After being supported, further growth of the particles is inhibited, while the colloids might undergo ripening.As there will be a time delay between supporting and measuring the colloids, measuring the particle distribution via TEM should represent the samples best.With little exception, for all samples, at least 500 particles were manually counted, and the distribution was fitted.From this, the specific surface area and peak particle diameter were calculated.
Alloying of the particles was verified using the same device, measuring EDX line scans and EDX mappings.For most samples, about 12 EDX measurements were performed.Compositions found this way don't always match the compositions found by digesting the samples and measuring ICP-OES, but generally show homogeneous alloying.The discrepancy in the composition is attributed to the small sample size and high difficulty in analysing sub-10-nm particles.Data presented in this study, which contains the composition data, always shows the data obtained by ICP-OES.Both size-and elemental composition data can be found in the supplementary information S2.

Preparation of catalytic measurements
All water-based samples weighed in at 150 g of washcoat each.Due to limitations in the solvent process, samples produced this way were limited in mass and comprised 10 g.This in turn means that two different approaches need to be utilized to study the catalytic properties.All water-based samples were processed as slurry, that was dip-coated on cordierite drill cores.These coated drill cores were dried and calcined.After this  preparation, all samples were catalytically tested in one DOC and two ASC-gas mixtures.Testing was carried out with fresh samples and samples that were aged to two different extents.All catalysts were subjected to ageing by exposing the samples to a water-saturated atmosphere (10 vol % H 2 O, 10 vol % O 2 in N 2 balance) for 16 hours at 650 °C (16/650).A smaller set of catalysts was also subjected to ageing in the same atmosphere at 800 °C (16/800).Since the observed activity trends are similar for both ageing protocols and most catalysts, all data presented the discussion relates to the more complete set of aged samples at 650 °C.In addition, all catalytic data can be found in the supplementary information.
All solvent-based samples underwent the same thermal cycling as the water-based samples.After being thoroughly mixed they were pressed into a pill, which was subsequently broken and sieved to obtain catalyst particles of about 500 μm diameter.150 mg of these particles were then packed into a tube furnace where they were aged and tested with an identical procedure as the water-based samples.

Catalyst testing procedures
The gas mixtures were chosen according to automotive industry standards.The DOC-gas mixture was comprised of CO, C 3 H 6 , NO, O 2 , and H 2 O in an N 2 carrier gas flow.One ASC gas mixture was comprised of NH 3 , O 2 and H 2 O in an N 2 carrier gas flow, while a second mixture additionally contained NO.The detailed composition can be found in S5.1 and S6.1.
After the preparation of the samples, catalytic testing was carried out according to the schematic illustrated in Figure 8.All samples were first characterized in the DOC gas stream and after that, the ASC testing ensued, first without NO and then with NO present in the gas stream.A slight deviation from this plan was made for the Pt-Pd and PtÀ Cu systems (for details confer S1.4.).
The acquisition of measurement data was carried out using an FTIR (FT-IR MKS Multigas 2030).Zero-point calibration was carried out once a day.Before the measurements, the concentrations of the gas mixtures are verified using the same system by a bypass flow setup.The analysed gases are summarized in Table 1.Information on the concentration of the product gases is then used to calculate the extent of conversion or formation.
As the water-based samples were tested as cordierite cores and the solvent-based samples as powder, a crossover measurement was carried out.Select water-based samples were characterized in the powder reactor after being prepared like the solvent-based samples.The results showed, that while the absolute values sometimes differ from the measurement of the cordierite cores, the order in which they are classified, does not (for details confer S3).Therefore, the measurements are comparable between water-based and solvent-based samples.

Figure 1 .
Figure 1.Exemplary results evaluation of a PtÀ Pd sample.From the CO and HC conversion curves (a, b) the temperature of 50 % conversion (LU 50 ) is read out.The NO 2 formation (c) is evaluated at 250 °C and the N 2 O formation (d) is evaluated as mean formation.

Figure 2 .
Figure 2. Exemplary sample analysis (on Pt 50 Rh 50 ).Particles were counted via TEM (a), single particle composition was verified on several EDX line and areascans (b) and bulk composition was checked with ICP-OES (c).

Figure 3 .
Figure 3. Summary of the particle size distribution modi of all samples sorted by synthesis laser.

Figure 4 .
Figure 4. LU 50 of laser-generated alloy catalyst (Pt + X) in the CO oxidation reaction.In fresh state (a) a volcano-like dependence from the adsorption energy of atomic oxygen is visible.The dependence in the left arc disappears for aged samples (b).

Figure 5 .
Figure 5. Difference between LU 50 (fresh minus aged) of laser-generated alloy catalyst in the CO oxidation reaction.Comparison to melting point of alloy element (a) and to melting point of the mixture taken from the underlying bulk phase diagrams (b).

Figure 6 .
Figure 6.Average NO concentration of laser-generated catalyst in the ammonia oxidation reaction.In fresh (a) and aged state (b) a volcano-like dependence from the adsorption energy of atomic oxygen is visible for E O < E O,Rh .

Table 1 .
Measured gases and evaluated values for the performed analyses.