Spatially Coupled Catalytic Ignition of Co Oxidation on Pt: Mesoscopic versus Nano-scale

Spatial coupling during catalytic ignition of CO oxidation on μm-sized Pt(hkl) domains of a poly-crystalline Pt foil has been studied in situ by PEEM (photoemission electron microscopy) in the 10 À 5 mbar pressure range. The same reaction has been examined under similar conditions by FIM (field ion microscopy) on nm-sized Pt(hkl) facets of a Pt nanotip. Proper orthogonal decomposition (POD) of the digitized FIM images has been employed to analyze spatiotemporal dynamics of catalytic ignition. The results show the essential role of the sample size and of the morphology of the domain (facet) boundary in the spatial coupling in CO oxidation. Automobile engines emit a large amount of pollutants just after starting (the so called " cold-start problem "), i.e. before the catalyst in the catalytic converter reaches the " light off " temperature when the effectivity of the converter switches rapidly from a low to a high level (catalytic ignition). This effect, observed in exothermic catalytic reactions, is usually considered as a heat balance problem , with the critical ignition temperature being defined as a point where the reaction-generated heat exceeds heat dissipation [1]. In case of CO oxidation, catalytic ignition is an environmentally related topic: in order to fulfill today's stringent emission standards , sophisticated catalyst heating processes have recently been developed to quickly reach the critical temperature. These processes vary from lean air-to-fuel ratio operation, exhaust system combustion devices, secondary air injection into the exhaust, to electrically heated catalysts, etc. [2]. Alternatively, a reduced critical temperature may shorten the period between the engine start and reaction " light off ". To reveal the role of the critical temperature one has to consider that factually , catalytic ignition represents not solely a heat production problem, but is rather a convolution of reaction kinetics and heat generation, since the latter is determined by the reaction rate. The reaction rate results in turn directly from reaction kinetics. In case of model studies under high-vacuum ($ 10 À 6 to 10 À 4 mbar) conditions, where heat and mass transport do not play an observable role, catalytic ignition can be treated as pure kinetic phenomenon, i.e. as a kinetic transition from low-rate steady state to a high-rate steady state [3,4]. Such transitions in the CO oxidation reaction are well studied, however mainly under iso-thermal conditions by varying the CO/O 2 ratio [5–7]. Only few studies, e.g. for Pd [4,8] or …


Int r oduct i on
Autom obile en gin es em it a large am oun t of pollutan ts just after startin g (the so called "cold-start problem " ), i.e. before th e catalyst in the catalytic converter reaches the "ligh t off" tem perature w h en th e effectivity of th e converter sw itches rapidly from a low to a high level (catalytic ignition). Th is effect, observed in exoth erm ic catalytic reaction s, is usually considered as a heat balan ce problem , w ith th e critical ign ition tem perature bein g defined as a poin t w h ere th e reaction-generated heat exceeds h eat dissipation [1]. In case of CO oxidation , catalytic ign ition is an en viron m en tally related topic: in ord er to fulfill today's strin gen t em ission standards, sophisticated catalyst heating processes have recently been developed to quickly reach the critical tem perature. Th ese processes vary from lean air-to-fuel ratio operation, exh aust system com bustion devices, secon dary air in jection in to th e exh aust, to electrically heated catalysts, etc. [2].
Alternatively, a reduced critical tem perature m ay shorten the period betw een the engin e start an d reaction "ligh t off" . To reveal th e role of the critical tem perature one h as to con sider th at factually, catalytic ign ition represents not solely a heat production problem , but is rath er a convolu tion of reaction kin etics an d h eat generation, since the latter is determ in ed by the reaction rate. The reaction rate results in turn directly from reaction kinetics. In case of m odel studies under h igh -vacuum ( 10 6 to 10 4 m bar) conditions, w here heat and m ass tran sport do not play an observable role, catalytic ignition can be treated as pure kinetic phen om enon , i.e. as a kinetic tran sition from low -rate steady state to a high-rate steady state [3,4]. Such transitions in the CO oxidation reaction are w ell studied, how ever m ain ly un der isoth erm al con ditions by varyin g the CO/O 2 ratio [5][6][7]. Only few studies, e.g. for Pd [4,8] or for a Pt w ire [9], w ere perform ed at increasin g tem perature under isobaric con ditions, i.e. m odeling th e " cold-start" process. Moreover, n o studies h ave been perform ed to our know ledge w h ere differen t length scale system s are com pared w ith respect to isobaric kinetic transitions un der the sam e experim ental conditions.
Recently, w e h ave developed an experim en tal approach w h ere kinetic ph ase tran sitions in CO oxidation can be studied in situ in a spatially resolved w ay, e.g. on in dividual differen tly orien ted grain s of polycrystallin e m etal foil by an alysis of local PEEM in ten sities [10]. Th e idea of th is ap proach is based upon th e fact th at th e local PEEM in ten sity (i.e. th e ph otoem ission yield) is directly dependent (via th e local w ork fun ction) on local CO or oxygen coverage. In turn, the rate of CO 2 form ation depen ds directly on CO or oxygen coverage [11]. From these tw o findings the dependence of the local PEEM im age in tensity on the local reaction rate m ay be concluded, w hich allow s a spatially resolved m onitoring of kinetic phase transition s [10,12,13].
In the present contribution w e extend th e above approach to th e n an oscale, exploitin g th e fact th at th e local in ten sity of th e FIM im age obtained by O 2 þ ions depends on the oxygen coverage of th e Pt-tip surface [14,15]. In th is w ay, catalytically active an d in active steady states of th e catalyst surface can be distinguished an d kinetic transition s can be m onitored, as w as dem on strated in our previous isothermal studies [16]. Herein w e show the first FIM observation s of th e isobaric kin etic tran sition s, i.e. catalytic ign ition of CO oxidation on th e apex of a Pt-nan otip. Such an apex exhibits a h eterogeneous surface form ed by differently orien ted n an ofacets an d can thus serve as a suitable m odel for a catalytic particle of com parable dim ensions [17,18]. As a con tribu tion to bridgin g th e structural com plexity gap betw een single crystal studies an d m ore soph isticated m odel system s, it w ould be interesting to com pare the catalytic beh avior of individual μm -sized grain s of a polycrystallin e Pt foil (as m onitored by PEEM on a m esoscopic scale) w ith th at of identically orien ted facets of a Pt tip (as visualized by FIM/FEM on th e n anoscale). Below w e present such a com parison betw een [100]-, [110]-and [111]-orien ted grains and identically orien ted facets of a Pt nanotip focusing m ainly on spatial correlation in catalytic ignition of CO oxidation for differently oriented grains (facets). We use proper orth ogon al decom position (POD, also know n as a Karh u n en -Loeve decom position ) of our FIM video-data to prove the syn chron ization of catalytic ign ition on different facets of th e Pt n an otip.

Exper i m ent al
The experim en ts w ere perform ed in tw o different all-m etal UHV setups, as described in detail elsew here: (i) a PEEM/XPS setup consisting of tw o individual PEEM and XPS cham bers connected w ith each other by a sam ple transfer line [19], (ii) a FIM setup th at can be operated either in the traditional FIM m ode using Ne as the im aging gas (for tip preparation ) or oxygen , w h ich serves as reactant an d im agin g gas at the sam e tim e (for in situ field ion im agin g of th e CO oxidation reaction ) [20].
Th e PEEM/XPS set u p is equ ipp ed w it h a PEEM (St aib In strum en ts), a deuterium discharge UV lam p (photon energy 6.5 eV) for electron excitation , an MS (MKS In strum en ts), XPS-system (Phoibos 100 h em isph erical energy analyzer and XR 50 tw in anode X-ray source, SPECS), a h igh purity gas supply system (O 2 : 99.999%, CO: 99.97%) and sam ple preparation facilities for clean ing th e sam ple by argon ion sputtering and subsequent an nealing. Differential pum ping of the PEEM intensifier section by a separate turbom olecular pum p an d tw o in-line apertures along the photoelectron trajectory (diam eters 4 m m an d 0.3 m m ) allow to keep the pressure inside the PEEM below 10 7 m bar, w ith the local pressure of reactants at th e sam ple up to 10 4 m bar.
The PEEM cham ber is used as a flow reactor for CO oxidation on polycrystallin e Pt foil, th e PEEM im ages w ere record ed in situ by a high -speed CCD cam era (Ham am atsu). Magn ification w as calibrated by com parison w ith optical m icrographs of the sam e Pt foil. A PEEM im age, form ed by photoelectrons, represents the lateral distribution of the local w ork function across th e sam ple. This allow s for differen tiation betw een differen tly orien ted grain s an d betw een different adsorbates by correlation of local im age inten sities w ith th e kn ow n w ork fun ction values of th e correspon din g clean an d adsorbate-covered single crystals. Th e recorded PEEM video-files w ere related to MS-data obtained sim ultan eously by a quadrupole m ass spectrom eter (MKS) placed in th e vicin ity of th e sam ple.
Th e PEEM sam ple con sisted of a 10 12 m m 2 polish ed polycrystalline Pt foil of 0.2 m m thickn ess (Mateck, 99,99%) w hich w as flam e ann ealed in air and further clean ed in UHV by repeated cycles of sputterin g w ith Ar þ ion s at 1 keV at 30 0 K an d consecutive an nealin g to 973-1073 K for 30 m in. The cleanness of the sam ple w as XPS con trolled after each sin gle m easurem en t. Th e sam ple tem perature w as m easured by a Ni/NiCr th erm ocouple spot-w elded to th e back side of the sam ple. The general configuration of the experim ental setup and the schem e of the PEEM experim ent are sh ow n in Fig. 1a.
In a sim ilar w ay, the FIM ch am ber ( Fig. 1b) w as used as a flow reactor for CO oxidation on a Pt n an otip, the correspondin g UHV system con tains a tip assem bly, w h ich allow s operation in a controlled tem perature range of 78-900 K, a ch annel plate, a gassupply system for im aging both n oble (Ne) and reactive (O 2 , CO) gases. Addition ally to the conven tion al DC high-voltage supply a pulsed high-voltage supply w ith pulses of triangular sh ape (w ith varyin g len gth ( 4 150 n s), m agnitude up to 5 kV, repetition frequencies up to 100 kHz) could be used in order to reveal the role of th e applied field, as discussed below. The Pt nan otip w as cleaned by field evaporation at 77 K, and th e tem perature of the tip w as m easured by a Ni/NiCr therm ocouple spot-w elded to its sh an k. FIM im ages during the ongoin g CO oxidation reaction w ere recorded w ith the sam e cam era as in the PEEM experim ents. Fig. 1b illustrates the n anoscale (FIM) experim en ts.

PEEM studies
In the usual param eter ran ge of an autom otive converter for CO oxidation, th e catalytic converter system m ay exh ibit tw o stable steady states, nam ely, a state of low reactivity w ith a predom inantly CO covered surface (cold start), an d a h igh reactivity state w ith a m ostly oxygen -covered surface (optim al operatin g regim e). Varying the extern al param eters, kinetic tran sitions betw een these tw o states can be enforced, w h ere a hysteresis is  alw ays observed, if an external con trol param eter (e.g., CO partial pressure or tem perature) is varied back an d forth . This ch aracteristic beh avior is present also for m odel system s under h igh vacuum con dition s as a m an ifestation of th e in trin sic bistability of th e catalytic CO oxidation [6,7,21]. Such beh avior results from the Lan gm u ir-Hinshelw ood reaction m echanism due to asym m etric inhibition of th e dissociative oxygen adsorption by CO. Oxygen n eeds tw o adsorption sites per m olecule an d can th us h ardly adsorb on a den sely CO-covered (poisoned) surface, w h ereas CO, in turn , can easily adsorb on a surface precovered w ith oxygen. Th erefore, the recovery of a CO poisoned surface to the active state occurs at low er CO pressure th an w hat w as necessary to poison th e surfacea hysteresis is observed. This m eans that on e of tw o stable states of the system , w ith h igh and low reactivity, can be achieved at th e sam e external param eters depen din g on th e preh istory: th is is an attribu te of bistability. Fig. 2a illustrates such behavior in th e particular case of polycrystallin e Pt foil: a hysteresis in CO 2 production is observed at cyclic varying of th e CO pressure at constan t oxygen pressure of 1.3 10 5 m bar an d tem perature of 513 K: tran sition τ A from th e catalytically active (oxygen covered) to th e inactive (CO poisoned) state an d th e reverse transition τ B take place at clearly different CO pressures. The hysteresis curve appears to be sm ooth ened out in com parison w ith the know n sin gle crystal m easurem ents [6,7], especially th e tran sition τ A . Such sm ooth appearance of th e transitions τ A an d τ B upon CO pressure variation reflects, as discussed in detail below, the fact th at not all the grain s are poisoned sim ultaneously, but th e process occurs sequen tially.
As already m en tion ed in Section 1, kin etic transition from the CO poisoned state to th e active state can also be induced by sim ple tem perature increase at isobaric condition s: initial desorption of CO provides m ore and m ore free adsorption places for dissociative adsorption of oxygen, this leads to an avalanche-like increase of the oxygen coverage and thus of the CO 2 production rate. Such an isobaric kinetic transition leads under real conditions to increasing heat production , an d the reaction eventually becom es self-sustained (light-off) [4,8]. Fig. 2b sh ow s th e global ign ition beh avior of th e polycrystallin e Pt foil as m on itored by MS at con stan t p O2 ¼ 1.3 10 5 m bar and p CO ¼ 6.6 10 6 m bar: follow in g the tem perature ram p, the global CO 2 production sudden ly increases, the isobaric tran sition poin t τ B * from the low activity state to the h igh activity state in high-vacuum conditions corresponds (like in the presen t case) to the ignition point. Correspondingly, the reverse transition to the low activity state τ A * corresponds to the extinction point. The global extinction curve also appears to be sm oothen ed out, an d a distinct " global extin ction tem perature" can hardly be assign ed. As w ill be sh ow n below, th is " sm oothin g" is caused by the absen ce of syn ch ron ization of kin etic tran sition s for differen tly orien ted Pt (hkl) dom ain s.
Monitoring the reaction evolution by PEEM provides th e possibility of spatially resolved m easurem ents. Fig. 3a-d shows a sequen ce of PEEM video-im ages taken during CO oxidation on th e Pt foil at constant p CO ¼ 6.6 10 6 m bar and p O2 ¼ 1.3 10 5 m bar w h ile the tem perature is ram ped from 483 K (fram e a) to 568 K (fram e d). From local PEEM in ten sities read out from video-data, spatially-resolved ignition data can be extracted by analyzing local PEEM in ten sities from in dividual Pt(hkl) dom ain s (Fig. 3e). In analogy to the MS signal in the overall CO 2 reaction rate, jum ps in th e local PEEM intensity represent local kinetic transitions on individual grain s. Since the in active (CO covered) surface exhibits th e brigh t contrast (low w ork fun ction), th e ignition jum ps occur from high to low intensity an d correspon d to th e jum ps from low to h igh local CO 2 production rate.
As is clearly visible from Fig. 3e, local tran sition s do n ot occur sim ultan eously for the different orien tations but show a pron oun ced structure sen sitivity w ith clearly iden tified critical temperatures of 417 K for Pt(110), 423 K for Pt(100), and 432 K for Pt (111). That m eans that individual grains "ligh t-off" sequentially: first the [110]-oriented dom ains, then the [100]-oriented and then th e [111]-dom ain s. Th is exp lain s im m ediately th e sm ooth ed character of the global curves in Fig. 2 and em phasizes the m ain draw back of con ventional global m easurem en ts, nam ely, th at the m easured data are averaged over th e w h ole sam p le con sistin g typically of differen tly active region s.
The present experim ents show that individual grains of the polycrystallin e Pt foil beh ave in depen den t in catalytic ign ition , sim ilar as observed earlier for kinetic tran sition s induced by CO pressure variations [10,12,13]. This con clusion agrees w ith the observation th at propagatin g reaction fronts are con fined w ithin grain boundaries, again an alogous w ith th e observation during cyclic CO pressure variation s [12,13]. Such in dependent behavior of in dividual Pt(hkl) dom ain s allow s assertin g th e validity of presen t light-off data for single crystals w ith corresponding orientation: e.g. the kinetic phase diagram for the Pt(111) dom ain on the foil (m easured using th e PEEM approach ) coin cides w ell w ith th at of a Pt(111) single crystal as m easured by MS [10].

FIM studies
Sim ilarly as in PEEM, the contrast in FIM also depends on the local w ork fun ction . How ever, th e con trast m ech an ism in FIM is m uch m ore com plex: w h ereas in PEEM the w ork fun ction directly govern s the photoelectron yield [22,23], in FIM the w ork function in fluences the critical distance of field ion ization an d th us th e probability of O 2 þ ion form ation [24]. In addition , th e atom ic scale roughness of the surface m odifies the local electric field [25] and th us th e rate of ion form ation. Last but not th e least, a resonance field ion ization of oxygen also contributes to th e im age contrast [15], but in sum , the CO covered areas appear alw ays darker than th e oxygen covered regions, at least on th e platin um m etal surface. For a Pt su rface, th is m ean s t h at a FIM im age ap p ears as a "negative" of a PEEM im age, but th e active and inactive state in the CO oxidation can be still discrim in ated reliably. O 2 þ ion s w ere first used for in situ visualization of CO oxidation already in 1990s in order to study th e oscillatin g regim e of th is reaction [26]. In this w ork w e apply th is im agin g m ode to visualize th e catalytic ign ition for the first tim e. Fig. 4a shows a sequence of six consecutive FIM im ages of a [10 0]-orien ted Pt field em itter tip illustratin g th e tran sition from low to h igh activity upon lin ear in crease of th e tem perature. Th e crystallography of the sam ple is indicated by th e positions of the low Miller-in dex dom ain s (fram e 1 in Fig. 4a). In the first th ree fram es of Fig.4a the surface of the Pt nanotip is still covered w ith CO an d th u s in th e in active state w ith a correspon din gly low FIM im age brightness. Betw een fram e 3 and 4 ign ition occurs: the surface gets quickly covered w ith oxygen, visible as a sudden increase in brightness.
Kin etic transition starts apparen tly from the center of the tip and instantaneously (that is, w ithin the tim e span of tw o consecutive fram es) spreads to other parts of the surface in a concentric m anner. Fig. 4b show s the corresponding FIM in ten sity analysis, w h ere th e in tegrated in ten sity of the w hole im age as w ell as th e local in ten sities w ith in th e defined ROIs indicated in th e first fram e of Fig. 4a are show n . After a sudden rise of the FIM brigh tness (lightoff) th e intensity and size of the brigh t area on the sam ple contin ue to vary sligh tly in tim e. Th is can be attributed to fluctuations w hich are know n from experim en tal studies perform ed on sim ilar sam ples [27,28].
Both visual inspection and local in ten sity an alysis w ith in the ROIs placed on low -index planes (Fig. 4b) create th e im pression th at ign ition occurs in a spatially coherent w ay over the m ajority of the facets. To obtain a profoun d know ledge about the existence of th e coheren t m odes w e apply local in tensity analysis and proper orthogonal decom position (POD) to the FIM video-data. POD analysis proved to be effective in th e detection of coh erent spatiotem poral m odes, e.g. in hydrodynam ics [29,30]. On catalytic surfaces POD has been em ployed by Graham et al. to an alyze spatio-tem poral tem perature pattern [31]. We have applied this m eth od earlier for th e an alysis of reaction -in du ced fluctuations [32] or as a proof of th e spatial desynchronization of glycolytic w aves [33].
The idea of POD is based on the fact that a signal w(x,t) th at varies in space and tim e can be decom posed in to tim e-dependent am plitudes a n (t) and tim e-independent m odes b n (x) which form an orthogonal basis: i.e. they capture th e overall dyn am ics of a system properly, the dim en sion ality (that is, the num ber of eigenvectors) of th e KLbasis can be chosen m uch sm aller than in th e original w ith out losing im portan t inform ation. In th e present case, w e use th is property to obtain the m ain features of a com plex dataset like FIM video-data. We applied POD analysis to the sam e FIM video-sequen ce th at w as used for th e local FIM im age in ten sity analysis ( Fig. 4b). Th e obtained results are sum m arized in Fig. 5, where Fig. 5a presen ts a set of th e origin al FIM video-fram es (upper row ) in com parison w ith th e first POD m ode (low er row ). The ROI that w as ch osen for the POD analysis an d w hich includes all th ree low Miller-in dex dom ain s (10 0), (110) an d (111) is indicated in the first fram e. Fig. 5b show s th e evolution of the original FIM intensity integrated w ithin the POD-ROI in com parison w ith the recon struction perform ed using on ly th e first POD m ode. As can be seen from Fig. 5a,b, all m ain features of the inten sity plots (integrated w ith in the big POD-ROI in Fig. 5 or w ithin the sm all ROIs for individual orien tations in Fig. 4) are sufficien tly reproduced by th e first POD m ode w hich is characterized by a sharp jum p in FIM intensity at the point of the kinetic phase transition at about fram e 29 an d a sm all subsequent drop to a sligh tly low er level.
Th e im age in ten sity th en fluctuates around this level for th e rest of the im age sequen ce. Such a coin cidence m eans that the first eigenvector, w hich am oun ts 86% of th e total contribution, domin ates the w hole decom position .
The higher m odes contribute only little to the overall dynam ics of the system , indicating a high degree of spatial correlation. Since the first POD m ode represents a constant spatial picture w hich varies in accordance w ith Eq. (1) in tim e as a 1 (t), such a high contribution of the first m ode m eans that the differently regions of the tip surface (i.e. differently oriented dom ains) are synchronized, i.e. the kinetic phase transition occurs sim ultaneously for all surface orientations. This result is in strong contrast to the observations m ade by PEEM for the m icrom eter-sized dom ains w here a quasi-independent behavior of the individual dom ains was observed. The observed differences betw een PEEM and FIM observations shed light on the coupling m echanism in CO oxidation: since the degree of therm al coupling and of the local pressure variations (w hich could cause coupling through the gas phase) are sim ilar for the Pt foil and the Pt-tip, only differences in the diffusion coupling via surface CO supply m ay be responsible for the observed effects. In fact, the diffusion length for CO reach es μm-range under the present reaction conditions (10 5 pressure range, T4 300 K), [34,35]. The facet size (nm-range) on a Pt tip is significantly sm aller than the diffusion length of CO in the present tem perature range, thus diffusive coupling provides synchronous ignition of the reaction on all the facets in the field of view. In principle, the diffusion length of CO m ay even provide sufficient coupling for different mm -sized dom ains of the polycrystalline Pt-foil surface. How ever, CO diffusion between the dom ains of a polycrystalline foil is effectively hindered by grain boundaries w hich confine therefore the propagation of reaction fronts w ithin the individual dom ains and prevent thus the reactive spatial coupling between neighboring grains, at least under the present vacuum conditions. This suggestion is strongly supported by our recent experim ents, in w hich a polycrystalline Pd foil was intensively sputtered by Ar þ ions [36]. Such sputtering fills up the cracks betw een the grains by Pd, enabling CO diffusion across the boundaries and, as a result, the independency of the individual (hkl)-dom ains vanishes.

Concl usions
In situ visualization of catalytic ign ition of CO oxidation on Pt has been realized on tw o different length -scales: (i) using PEEM for in dividual grain s of a polycrystallin e Pt-foil (μm -scale) and (ii) by FIM for an apex of a Pt-nanotip (nm -scale). The results allow to sh ed ligh t to th e peculiarities of spatial couplin g in th is reaction : w hereas the ignition process occurs in dependently on individual Pt(hkl)-dom ains of the polycrystalline Pt sam ple, the differently oriented facets of the Pt-nanotip light-off in a synchronized w ay. To reveal the degree of spatial synchronization, proper orthogonal decom position (POD) w as applied to th e FIM video-files parallel to local in ten sity an alysis of th e FIM im ages. Th e first POD-m ode contains 86% of th e total con tribution, confirm ing thus the high degree of spatial correlation on the Pt-nanotip. Th e obtain ed results can be traced back to differences in the CO diffusion paths on th e polycrystallin e Pt-foil an d on th e Pt-n an otip surface: w h ereas the individual facets on th e tip-apex ch an ge over from one crystallograph ic orien tation to an oth er via stepped "transition regions" consisting solely on Pt-atom s, the individual grains on the Pt-foil surface are separated by grain boundaries consisting of cracks filled w ith im purities. Th ese cracks h in der effectively CO d iffusion an d thus the couplin g betw een different grains, w hereas on th e Ptnanotip CO can easily diffuse from one facet to another and synchronize th e catalytic ignition .