Review: Monoclinic zirconia, its surface sites and their interaction with carbon monoxide

This review concerns monoclinic zirconia, its surface sites and their probing with carbon monoxide. The surface sites and their modiﬁcations using thermal treatments with vacuum or reactive gases are also included. In this work, we present information on the nature and manipulation of hydroxyl species and their quantities on the surface, the different types of cationic sites where CO is adsorbed linearly and their energetics, as well as the surface sites and dynamics of formate formation. We also compare the surface concentrations of the different surface species to better understand the extent and nature of the interactions. Finally, we discuss some of the remaining open questions and how to approach them.


Introduction
Zirconium oxide has gained interest both as a catalyst support and a catalyst on its own, mostly due to its weak acid and basic sites 1 and stability under oxidizing and reducing atmospheres 2 .It has been an interesting catalyst material especially for biomass-related reactions that are actively investigated, as the future is bright for non-fossil fuels and chemicals.Biomass-related reaction networks usually include carbon oxides, as both carbon and oxygen are largely abundant in the starting material.
Monoclinic zirconia (also known as baddeleyite) is an oxide typically covered with hydroxyl species, similar to many metal oxides 3 used as catalysts and catalyst support materials.The structure of monoclinic zirconia provides a more versatile surface than the other polymorphs (cubic, tetragonal) due to a less symmetrical lattice 4 .The surface sites on monoclinic zirconia include hydroxyls 5 , oxygen vacancies 6 , coordinatively unsaturated (c.u.s.) Zr-O pairs 7 , and Lewis acid sites (Zr 3+ , Zr 4+ ) 8 .The hydroxyl groups on the surface can be manipulated by thermal treatment in vacuum or in different atmospheres 9 , which are often necessary for catalytic applications, e.g., in methanol synthesis 10 .Additionally, monoclinic zirconia has been suggested as a support for water-gas shift reaction catalysts with gold, platinum and copper [11][12][13] , and for reforming with platinum, nickel, cobalt and copper 14,15 .
Since 2006, monoclinic zirconia has been prepared also in nanoshapes, including nanorods 16,17 and nanosheets 16 .Using the nanoshapes in catalysis might be beneficial due to their well-defined surface sites; thus their selectivity might be more easily linked to the exposed surfaces than those of traditional catalysts in polycrystalline form.
Carbon monoxide interacts with monoclinic zirconia both on the clean surface with few or no hydroxyls 18 , as well as on the hydroxylated surface 19,20 .The surface hydroxyl groups seem to play an important role in the interaction of CO with zirconia: as a site for forming formates 5,7,[20][21][22][23] and bicarbonates 22,[24][25][26] but also inhibiting the formation of adsorbed linear CO species 27 .The main adsorbed CO species are linearly adsorbed CO at room temperature and below and formate species above 100 • C, the first one desorbing reversibly 27 and the second one decomposing both reversibly back to gas-phase CO and irreversibly to CO 2 and H 2 20,21,28 .
The interaction of carbon monoxide with zirconium oxide has been studied actively since the 1970s 5,18,21,23,29,30 .Due to differences in zirconia materials and their pretreatments, experimental setups and conditions, interpreting and comparing the obtained results is not straightforward.Most of the studies on the interaction with CO have been carried out using infrared spectroscopy 5,18,19,[22][23][24] .However, studies using other techniques, such as temperature-programmed methods 20,21,28,30 , calorimetry 27,31,32 , and gravimetry 33,34 , as well as theoretical studies 22,28 have been published.
The focus of this work is in the monoclinic polymorph of zirconia instead of the whole spectrum of zirconia materials: doped (including sulphated), tetragonal and cubic zirconia, to name the most significant zirconias excluded from this review.This choice is made for simplicity and clarity, as monoclinic and tetragonal zirconias differ in, e.g., surface hydroxyl species 7,25 and acidity/basicity 7,24,30 , further demonstrated in their interactions with CO 7,25,30 and in their catalytic activity, e.g., in water-gas shift 11,12 .More information can also be found in the reviews by Dyrek et al. 35 and Hadjiivanov 36 .Most of the pre-1993 works included in this paper were reviewed 20 years ago by Nawrocki et al. 37 in the context of chromatography; this review is providing an update

Catalysis Science & Technology Accepted Manuscript
with work reported since 1993, from the perspective of catalysis.
2 Preparation, structure, and surfaces

Crystal structure and surfaces
The unit cell of monoclinic zirconia is face-centered (spacegroup C 5 2h ), consisting of 4 Zr atoms and 8 O atoms 72 .In the bulk all Zr atoms are seven-fold coordinated to three threefold coordinated oxygen atoms and four four-fold coordinated oxygen atoms 73 .Based on theory, the coordination numbers on the surface differ, e.g., on the (111) surface some of the zirconium atoms are only six-fold coordinated and some of the oxygen atoms are three-or two-fold coordinated instead of four-or three-fold coordination 74 .On the (111) surface, some zirconium atoms can be even five-fold coordinated 74 .The experimentally determined lattice parameters are a = 5.17  4 .Warble reported (110), (100) and (111) surfaces to be exhibited in transmission electron microscopy (TEM) 76 .The (111) 39,40 , (001) 39,40 and (011) 40 planes have been reported based on high-resolution TEM (HRTEM) images.Theoretically, the most stable surface for monoclinic zirconia is the (111) surface 4 .The difference between the theoretically most stable surfaces and the experimentally observed surfaces might be due to inaccuracy in either method, and it is not clear whether the particle size or, e.g., the degree of surface hydration play a role.In nanocrystals, the (011) surface seems most abundant in TEM, while also (111), (111) and (001) surfaces have been observed 77 .The ideal (111), (111), (001) and (011) surfaces are shown in Figure 1.The atom positions are those of the bulk structure assuming no reconstruction takes place.Jung et al. report that the crystallographic structure of zirconia, instead of crystallite size or calcination temperature, is the most significant factor for determining the nature and density of the surface sites suitable for CO 2 and NH 3 adsorption, i.e., the basic and acidic surface sites 41 .Monoclinic zirconia has a higher CO adsorption capacity than tetragonal ZrO 2 , which is attributed to its higher Lewis acidity and basicity 30 .The tetragonal structure is more symmetrical than the monoclinic one 4 , leading to a smaller number of different surfaces and thus possibly also to less versatile surface site types.
Thermal treatment tends to increase the crystal size of monoclinic zirconia.Increasing the calcination temperature by 300 −500 • C can lead to a particle size of three or even four times the original 40,64 .Auger electron spectroscopy (AES) 79 shows that the surface O/Zr ratio decreases from the stoichiometric 2 down to 1.1 with evacuation at 427 • C and 727 • C, respectively.The surface reduction is very fast, and extending the time to one hour in ultra-high vacuum (UHV) does not affect the surface 79 .With H 2 the surface seems unreducible even at 900 Torr and 900 • C while atomic hydrogen reduces the surface already at 750 • C and H 2 pressure of 5 µTorr prior to atomization 80 .The ineffectiveness of H 2 compared to UHV in surface reduction is rather unexpected, yet atomic hydrogen

Catalysis Science & Technology Accepted Manuscript
seems to be a better reductant than molecular hydrogen.Based on the literature, the mechanism for oxygen removal remains unclear; is it removed as O 2 or perhaps through dehydroxylation via multicoordinated OH groups, as shown in Eq. 1: The presence of Zr 3+ sites has been proposed based on electron paramagnetic resonance (EPR) 8,46,51,57 .The Zr surface cations have been probed with N 2 O, it decomposes on Zr 3+ already at room temperature but also at higher temperatures, and reversibly adsorbs on Zr 4+ sites 6 .Vacuum treatment at above 600 • C is sufficient to initiate formation of oxygen vacancies 81 and reduction of cations from Zr 4+ to Zr 3+ 6,81 .Hydrogen treatment at 700 • C also transforms some of the Zr 4+ sites to Zr 3+ sites 81 , but with hydrogen treatment at 600 • C, the amount of Zr 3+ sites does not increase 9 in agreement with the low reducibility with H 2 mentioned previously.If the sample has been vacuum-activated at 400 • C, hydration quenches the defect centers (Zr 3+ ) and the defects have to be re-created with evacuation 8 .If the sample has been vacuum-activated at 800 • C, there are two options: (1) the defect centers are quenched but they are easier to re-create on a sintered surface, or (2) the defect centers are not quenched, rather coordinatively saturated with water and dehydration restores their coordinatively unsaturated state 8 .Based on EPR, the most exposed Zr 3+ are transformed to Zr 4+ when in contact with water (200 • C, 18 torr H 2 O) but the others are coordinatively saturated with water and remaining Zr 3+ 8 .
Syzgantseva et al. have calculated the formation energies for oxygen vacancies, and based on those the zirconia (111) surface is less reducible than the (101) surface 82 , in line with the stability observation by Christensen and Carter 4 .When comparing to other oxides, the oxygen vacancy formation energy on the zirconia surface is 820 − 880 kJ/mol and in the bulk ca.860 kJ/mol, whereas those for titania, which is considered a reducible oxide, are 530 − 580 kJ/mol on the surface and 670 kJ/mol in the bulk 82 .Syzgantseva et al. have also predicted the conditions necessary to create oxygen vacancies with surface hydrogenation followed by desorption of water (temperature above 927 • C and H 2 O/H 2 pressure ratio below 10 −5 ) 82 .They conclude that water desorption takes place already at milder conditions, thus the simplifications in their computations required for, e.g., surface models might cause the discrepancy between theory and experiments 82 .
The oxygen mobility, dissociation and recombination on zirconia among other oxide materials were probed by Martin and Duprez 66 .Isotopic oxygen exchange experiments between 16 O 2 and 18 O 2 showed surface oxygen exchange at 380−780 • C 66 .The maximum O-exchange rate was at 530 • C and the exchange rate was further expedited in the presence of Rh or Pt on the surface 66 .The exchange is attributed to c.u.s.Zr 3+ centers created by vacuum thermal treatment and their ability to dissociate molecular oxygen 66 .The number of exchanged oxygens on ZrO 2 was found to be greater than the theoretical number of surface oxygen species implying that bulk oxygen atoms participated in the exchange 66 .
3 Hydroxyl species and the interaction of water and hydrogen on monoclinic zirconia Dissociative adsorption of water on monoclinic zirconia is exothermic, occurring already at room temperature 31,69 .Therefore the surfaces of zirconia are hydroxylated under ambient and in most reactive atmospheres.In this section, different types of OH groups are discussed, followed by methods of manipulating the hydroxyl species on the surface of zirconia.

Nature, density and probing of hydroxyl species
Typical hydroxyl sites reported on monoclinic zirconia include terminal OH groups (also known as monocoordinated OH) and tribridged OH groups, first assigned by Tsyganenko and Filimonov as a part of a wider study on different oxides including cerium, hafnium, magnesium, nickel, cobalt and several other oxides 3,71 .Schematic drawings of the hydroxyl groups are presented in Fig. 2. The latter species has been assigned either tribridged 7,25,70,71 , bibridged 29,53,61,63 or simply presumed as multicoordinated OH groups 18,19,23,59 .In this work, the term multicoordinated is generally used to refer to this species.The terminal and multicoordinated hydroxyl species are usually observed in IR spectroscopy at 3780 − 3760 cm −1 and 3690 − 3650 cm −1 , respectively 7,19,50,61,71 , as shown in Table 2 and Figure 3. Jacob et al. assign the bands to OH groups related to trigonally coordinated O 2− anions (ca.3774 cm −1 ) and to tetrahedrally coordinated O 2− anions (ca.3668 cm −1 ) 65 , however, without further evidence.Yamaguchi et al. suggested that the bands at 3780 cm −1 and 3680 cm −1 would be isolated hydroxyls (as opposed to hydrogen-bonded ones) of bridged and terminal type adapting the interpretation for rutile, a type of TiO 2 59 .The occurrence of a band at 3740 − 3720 cm −1 is typically interpreted as a sign of tetragonal phase and its bibridged hydroxyl.Yet the band of terminal OH (typically at 3780 − 3760 cm −1 ) shifts towards lower

Catalysis Science & Technology Accepted Manuscript
wavenumbers (even down to 3680 cm −1 ) with increasing degree of hydration due to hydrogen bonding 18 and at that point the multicoordinated OH band is even below 3600 cm −1 40 , as suggested already by Tret'yakov et al. 44 .Thus interpreting the hydroxyl band position must be done carefully, considering the possibility of hydrogen bonding at higher degrees of hydration.Tsyganenko and Filimonov also report an IR band at 3380 cm −1 , assigned to hydrogen-bonded hydroxyl groups 3,71 .Unresolved broad bands at 3600 − 2800 cm −1 are assigned to hydrogen-bonded polynuclear water species 61 based on the IR spectra of hydrogen bonding of water 83 .The hydrogen bonding is suggested to occur between molecularly adsorbed water and the hydroxyl groups 18,33,42,53 .The OH wavenumber and thermal stability depend on the crystalline phase of zirconia 9 ; thus impurities in the crystalline phase can lead to misinterpretations.Terminal hydroxyls are bound directly to a single cation at oxygen lattice sites whereas the multicoordinated hydroxyls are located at low-index faces 61 .The former are able to accommodate a water molecule to hydrogen-bond on an edge or a corner site of the oxide particle while the latter are incapable of hydrogen bonding due to steric factors 61 .The assignment of multicoordinated hydroxyls to low-index faces 61 and the observed decrease in the relative amount of terminal OH species with vacuum treatment at 500 − 1000 • C 65 are in line with the observation that the regularity of sintered crystals is higher 39 .
Based on IR spectroscopy, the tribridged OH species (at 3690 − 3681 cm −1 and at 3670 − 3660 cm −1 ) on zirconia are suggested to be actually two species 7,40 , and signs of multiple tribridged species are also seen both at high 18,28 and low 18 degrees of hydration.The OH species differ slightly in their acidic character 40 and their behavior with calcination at 600 − 900 • C suggests that the low-frequency band (3670 − 3660 cm −1 ) should be assigned to tribridged species at crystallographic defects selectively annealed during sintering 40 .However, in our recent investigations for reduced zirconia with no visible H-bonding effect, both terminal and multicoordinated bands shift down in wavenumber with increasing temperature 28 .
The nature of the multicoordinated OH species has also been considered at the atomic level.The ZrO 2 (111) surface is suggested to have too small a distance between the c.u.s.O 2− and the c.u.s.Zr 4+ site (only 2 Å) 40 , that is not enough space to dissociate a water molecule to form terminal and bibridged OH groups.The corresponding distance between Zr 4+ and tribridged oxygen is ≈ 2.3 Å, allowing formation of a terminal and a tribridged hydroxyl via dissociation 40 .Based on theory, both bibridged and tribridged species exist 67,74 , Korhonen et al. report that at low coverages (θ = 0.25 ML), there are only bibridged hydroxyls on the (111) and (101) surfaces 67 .Bibridged hydroxyls are also claimed to form following dissociative adsorption on the (001) surface until a full coverage (θ = 1 ML) is obtained 85 .
The hydroxyl IR vibrations have been determined computationally 67,74,84 as shown in Table 2.It demonstrates the large variation of calculated frequencies which span, e.g., over 200

Catalysis Science & Technology Accepted Manuscript
wavenumbers for tribridged OH species.Diversity of calculated frequencies for adsorbed OH is not unique to ZrO 2 but also reported on CeO 2 86 .The overall discrepancy between calculated and experimental frequencies can follow from three different origins.First, frequencies are typically calculated using a so-called harmonic approximation, which specifically leaves out possible anharmonic effects leading, in part, to errors in calculated frequencies.Secondly, discrepancies can also originate from the inability of density functional theory (DFT) to reliably describe the electronic structure of a given system.Thirdly, for polycrystalline oxide nanostructures, the discrepancy of calculated frequencies compared to the experimental ones can stem from the fact that an active surface site cannot experimentally be identified definitely.This may lead to an inaccurate computational adsorption site model, which differs from the real adsorption site.Moreover, the application of a cluster model to simulate ZrO 2 surfaces can impact on the calculated frequencies.
Hydroxyl densities can be estimated based on the amount of water removed from zirconia 34 , water adsorbed on zirconia 27,69 , or on quantification via 1 H MAS NMR 68 , and these estimates are presented in Table 3.In the table, the scaled values represent OH density regardless whether it has been measured directly or by water adsorption (each water molecule is assumed to form two surface hydroxyls).
The observed amount of adsorbed water decreases with increasing temperature, as expected.Based on an estimated number of Zr atoms on the ZrO 2 surface (ca.12 µmol/m 2 ) and water adsorption leading to formation of two OH groups, one on Zr cation and one on a surface oxygen, the estimated hydroxyl densities seem reasonable in magnitude.Nawrocki et al. report a theoretical maximum OH concentration of ca. 25 µmol/m 2 based on the average surface Zr concentration of ca.12.2 µmol/m 2 37 .The amount of induced hydroxyls (20.2 µmol/m 2 desorbed above 200 • C) 34 seems to be in agreement with the estimated total OH capacity, also Piskorz et al. 77 have reported similar theoretical OH site densities.
The energetics of water adsorption have been investigated both with an experimental and a theoretical approach.Piskorz et al. have studied the effect of surface hydration on the stability of the crystal planes using DFT calculations 77 .A hydroxylated surface is favored over the clean surface on very small crystallites (< 20 Å), whereas hydration does not enhance the stability of the surface on crystallites between 500 Å and 2000 Å, and the authors claim that the transformation from clean to hydrated surface is attenuated 77 .However, this is not in line with experimental observations suggesting the hydroxylated surface to be the prevalent one.Based on microcalorimetry, the integral enthalpy of adsorption for a half-layer coverage of water (3.65 µmol/m 2 ) is −142 kJ/mol on monoclinic ZrO 2 nanoparticles (crystal size 100 − 500 Å, specific surface area 1.6 − 27.2 m 2 /g) 31 .This is in agreement with the values measured for powder zirconias (1.0 − 1.6 m 2 /g) 32 giving a range from −110 kJ/mol to −170 kJ/mol.Theoretical adsorption energy values typically range from −80 kJ/mol to −190 kJ/mol on (001), ( 111), (111), and (101) 67,69,85 .We have recently reported dissociative adsorption energies for the first adsorbing water molecule ranging from −106 . . .− 119 kJ/mol up to −297 kJ/mol, for flat, stepped and corner adsorption sites 28 .The calculated values demonstrate the structure sensitivity of dissociative adsorption of water, which is clearly most favorable on a c.u.s.site such as a corner site suggested by our DFT calculations 28 .

Manipulation of hydroxyl species
The intensity ratio of terminal and multicoordinated hydroxyls varies according to temperature, used atmosphere and pretreatment.Hydroxyl species can be added to the surface with water or hydrogen treatments and removed using heat together with inert gas or evacuation.The initial state of zirconia can usually be restored with evacuation or flushing with inert at the same temperature as before rehydration 28,53 .
Sample hydration is typically carried out by adsorbing water vapor, either by letting the sample adsorb moisture from air (virgin material in 18 ), equilibrating in a closed vessel with water vapor in nitrogen (to avoid CO 2 adsorption) 34 , or by feeding water vapor to the sample 18,23,28,58,65 .The rehydrated sample can be used as such or after further dehydration with vacuum and/or elevated temperature (e.g. 18,65).Rehydration has also been carried out by exposing the sample to hydrogen at room temperature 53 or by contacting the zirconia sample with water-saturated hydrogen for 10 min at 50 • C 42 .Unfortunately, rehydration is often described vaguely, omitting time, water vapor concentration or temperature, all relevant in controlling the degree of hydroxylation.
Commonly used methods for dehydration are vacuuming 20,24,33,70 or flushing 67 at elevated temperatures.Undissociated water adsorbed at room temperature is completely removed in vacuum by 127 • C 27 and after evacuation at 200 • C only two distinct OH bands at ca. 3775 cm −1 and 3665 cm −1 remain 40 .Köck et al. pretreated the zirconia in air at 900 • C to completely dehydroxylate the sample, and only very weak OH bands remained 62 .Evacuation at 500 • C is reported to be insufficient for complete dehydroxylation 20 , but already at 550 • C spectra with no trace of surface hydroxyls after evacuation are shown 53 .The conditions necessary for total OH removal are at ca. 550 − 750 • C in vacuo 18,44,65 , in agreement with the enthalpies for water adsorption discussed earlier in this paper in Section 3.1.As most pretreatments and processes do not reach these conditions, the presence of these OH groups on the zirconia surface in process conditions is practically inevitable, especially in biomass-based processes, where water is present.The partial dehydration method 18,44,53,65 can be used to vary the concentration of several types of OH groups on zirconia.Depending whether the amount of adsorbed water results from dehydration or rehydration, the surface species distribution might be different as demonstrated by Bolis et al. 27 by heating the sample in a closed vessel and analyzing the OH distribution on the sample before and after.Based on their findings, dehydration is systematic even in terms of the surface effects whereas rehydration is more blotchy as the water collides and dissociates on the surface as hydrogen-bonded pairs 27 .Thus allowing the system to approach equilibrium leads to a more even surface distribution.All in all, the method chosen to adjust the amount of hydroxyls influences their distribution on the zirconia surface.

Catalysis Science & Technology Accepted Manuscript
Cerrato et al. suggest based on theory that water dissociates forming a tribridged OH at a tricoordinated c.u.s.oxygen and a terminal OH on a c.u.s.cation, leaving a c.u.s.monocoordinated oxygen in the same sphere unsaturated 40 .The presence of the suggested c.u.s.monocoordinated oxygen is in agreement with CO 2 adsorption experiments on a fully hydrated surface, yielding monodentate and bidentate carbonate species requiring the presence of basic, c.u.s.oxygen ions 40 .
Based on theory, dissociative adsorption for the first adsorbed water molecule of a unit cell (θ = 0 . . .0.25 ML) has been reported on (111) 67,77 , (101) 67 , and (111) 77 surfaces.The second H 2 O molecule adsorbs molecularly on (111) 67,77 and (111) 77 , and dissociatively on (101) 67 surfaces.The additional H 2 O molecules adsorb molecularly 67,77 .For both (111) and (101) surfaces already the first water molecule forming two hydroxyls is hydrogen-bonded 67 .Iskandarova et al. have reported both dissociative and molecular adsorption enthalpies resulting in coverages of 0.5 ML and 1 ML on a (001) surface, the dissociative adsorption is favored by 45 − 75 kJ/mol in both cases 85 .Our recent investigations suggest that at θ = 0.25 ML, there is hydrogen bonding between the hydroxyl groups on the (111) surface but not on the hydroxylated (212) edge and corner sites 28 .Cerrato et al. point out that the hydrogen bonding at high hydroxyl coverages only takes place between OH species and coordinated undissociated water whereas hydrogen bonding between OH pairs is unlikely 40 .
Morterra et al. hypothesize that if all surface oxygens on a (111) surface are transformed into hydroxyls to maintain electrical neutrality, and then dehydration takes place via desorbing terminal hydroxyls and hydrogen atoms from bridging OH groups (shown in Fig. 4), only bridging oxygens are left behind, and thus highly uncoordinated Zr 4+ sites are achieved 39 .The intensity of the terminal hydroxyl species decreases more than that of the multicoordinated OH when the zirconia sample is thermoevacuated at 500 − 600 • C after calcination at 600 • C 70 .Dehydration in vacuum at 500 − 1000 • C followed by hydration caused the relative amount of terminal OH to decrease significantly compared to multicoordinated OH, whereas oxygen treatment (500 − 750 • C) before hydration had the opposite effect 65 .The decrease in terminal hydroxyls with high-temperature vacuum treatment is assigned to increasing amount of tetragonal zirconia, as the tetragonal phase is stabilized by oxygen vacancies 65 .
One approach to hydroxyl studies is to replace hydrogen with its heavier isotope deuterium using D 2 or D 2 O, or to replace 16 O with 18 O, all of which are easily observable in both IR and mass spectrometry.OD groups, known as deuteroxyls, have been investigated by many groups 19,21,24,42,44,56,59,61,87 while oxygen-labeling studies are more scarce 66, 69,81 .
Observed deuteroxyl IR wavenumbers are collected in Table 4. Erkelens et al. report that the ratios of the frequen- cies between the OH bands (3732 cm −1 , 3660 cm −1 , and 3584 cm −1 ) and the OD bands (2758 cm −1 , 2702 cm −1 , and 2651 cm −1 ) are ca.1.36, in agreement with the expected isotopic substitution 56 .Similar results with additional OH band at 3738−3727 cm −1 (OD at 2757−2748 cm −1 ) were reported also by Guglielminotti 29 .A spectrum showing the changes in OH/OD groups is shown in Fig. 5.  Repeated treatments with, e.g., 10 torr of D 2 O vapor 44 at room temperature almost completely replace hydrogen with deuterium in both terminal and multicoordinated OH, resulting in corresponding OD groups 19,24,44,56,59,61 .The terminal species seem to exchange more easily than the multicoordinated species 24,59 .
Ignatchenko et al. propose that the hydrogen-deuterium exchange mechanism would proceed via hydroxyl/deuteroxyl exchange so that a deuterated water adsorbs adjacent to an existing terminal hydroxyl species and a hydrogen bond is formed between the D 2 O and the OH species 69 .Then the hydrogen-bonding deuterium atom and the original OH group desorb as HDO species, leaving the OD species on the surface.The mechanism is shown in Fig. 6.
Labeling the oxygen of water (H 18 2 O) reveals that the oxygens of hydroxyls are thoroughly exchanged by 400 • C, yet at 200 • C the findings of Ignatchenko et al. 69 disagree with the suggested mechanism in Fig. 6.It appears that already at 200 • C the terminal hydroxyls can be exchanged but for the multicoordinated ones higher temperatures (closer to 400 • C) are required 69 , and the presented mechanism should be modified to also apply for the multicoordinated hydroxyls.Based on these findings, it seems that the hydrogen scrambling follows the suggested mechanism 69 (see Fig. 6).The terminal OH species are completely exchanged via the normal (de)hydroxylation mechanism involving also multicoordinated OH groups and labeled H 18 2 O as deduced from the applied temperature range (200 − 400 • C).The exchange of multicoordinated hydroxyls requires a new mechanism hypothesis and we propose that multi-OH groups are removed as water, leaving behind one oxygen atom and an oxygen vacancy (see Eq. 1).This mechanism may be predominating only at high temperatures when terminal OH species have already been desorbed.
In addition to water or its deuterated counterpart, also H 2 or D 2 can be used to create surface OH or OD groups on monoclinic zirconia.He and Ekerdt have suggested that gas-phase hydrogen is able to replenish OH groups 21 , the hydroxyl IR bands emerge at 200−600 • C 9,48,53 for both terminal and multicoordinated hydroxyl species 53 .
Hydroxyl formation mechanism by molecular hydrogen can be either homolytic, resulting in two hydroxyl species and two electrons 9 , or heterolytic, resulting in IR-inactive Zr-H (H + type) species and a hydroxyl (OH) species 49 .At temperatures above 100 • C, large amounts of hydroxyl species are formed likely due to homolytic dissociation of hydrogen 48 , whereas hydrogen contact at room temperature induces heterolytic dissociation 48 .Heterolytic dissociative adsorption of H 2 at room  82 .The proposed mechanism is presented in Fig. 7. Addition of gas-phase oxygen to the Zr-H hydrides, produced by gas-phase H 2 at room temperature, increases the OH intensity at 3668 cm −1 (generally considered as multicoordinated OH), creates the 3774 cm −1 OH band (terminal OH) and decreases the intensity of the Zr-H species at 1565 cm −1 81 .Substituting regular oxygen ( 16 O 2 ) with isotopically labeled oxygen ( 18 O 2 ) does not affect the position of the OH bands (expected shift 11 cm −1 ), thus the OH formation seems to occur on lattice oxygen rather than the gas-phase originating oxygen species 81 .Fig. 7 The hydrogen dissociation mechanism proposed by Syzgantseva et al. 82 .Reproduced from ref. 82 with permission.Copyright 2012, American Chemical Society.
Treatment with deuterium gas at 200 • C (1h, 18 torr) and 250 • C (1h, 250 torr), is sufficient to exchange virtually all hydrogen of hydroxyl groups to deuterium 56 .He and Ekerdt report that the deuterium in OD groups is replaced by hydrogen already at 200 • C with hydrogen dissociating on the zirconia surface 19 .At 150 • C in 488 kPa of D 2 , half of the hydrogen in surface OH species are changed to deuterium within 30 seconds 42 .The H/D exchange takes place already at 100 • C with D 2 in the gas phase, however, at 200 • C the exchange rate increases considerably 26 .The activation energies of the H/D exchange reaction with D 2 are similar for both terminal and multicoordinated hydroxyls, and they seem to increase with the progress of the reaction 26 .This is interpreted to be due to the overall exchange (migration and replacement of atomic hydrogen by deuterium) being limited by D migration on the surface, subject to heterogeneity of potential barriers to various sites 26 .
Merle-Méjean et al. have found that on an air-calcined zirconia the hydroxyl species are H/D exchanged in contact with D 2 (507 • C, 100 hPa) so quickly that it gives reason to believe the OH species are on the surface only 63 .Conversely, on the steam-calcined zirconia there are some hydroxyls exchanged to deuterium-containing species so slowly (during several hours), if at all, that they must be elsewhere in the oxide, likely in the bulk 63 .
The presence of formate species is suggested to decrease the number of available sites for H 2 /D 2 dissociation as well as partially block the path for surface transport of H or D atoms 42 .If there are formates on the surface, the H/D exchange between OH and OD species at 150 • C is 4 − 36 times slower depending on the formate coverage (0.3 or 0.8 times the maximum coverage) 42 .The overall extent of the H/D exchange of multicoordinated OH to OD is limited to 9% with formate and 2% with methoxy species as compared to normal ZrO 2 surface 26 .

Interaction with CO
Upon contact with monoclinic zirconia, CO tends to form several surface species: at low temperatures up to ca. 100 • C the preferred species is linearly adsorbed CO, at higher temperatures the dominating surface species is formate.In addition to these, also carbonate and carboxylate-type species have been observed.

Linear CO species, formation and stability
CO adsorption at room temperature leads to the formation of linear CO species on cationic sites of the zirconia surface (see Fig. 8), the corresponding bands in IR spectra are located at ca. 2200 − 2170 cm −1 5,18,48,50 .Spectra of the linear CO species as a function of CO pressure and with two differently prepared samples are shown in Fig. 9.The presence of a weak band at 2112 cm −1 is reported after C adsorption at room temperature 48 .With adsorption below room temperature, e.g., at −173 . . .−195 • C, the range of adsorbed linear CO species extends to ca. 2200 − 2140 cm −1 at varying CO pressures (from 10 −4 to 40 torr) 39,58 .The IR bands at ca. 2200 − 2190 cm −1 are interpreted as CO adsorbed on Zr 4+ 6,27,29 and the band at ca. 2120 − 2110 cm −1 is assigned to CO on Zr 3+ surface ions 6,7,9,29,81 , whereas ESR (electron spin resonance spectroscopy) results are interpreted so that Zr 3+ surface ions do not interact with CO at room temperature 47 .The appearance of IR bands assigned to adsorbed CO at higher wavenumbers than gas-phase CO (at 2202 cm −1 ) is attributed to polarization of the CO molecule on the surface 48 .At room temperature the OH groups are not modified during CO adsorption 54 .
The (CO) H and (CO) L species are suggested to be on two types of Lewis-acidic centers 18 , both types assigned as Zr 4+ ions 27 .For CO chemisorption, these centers are suggested to be caused by differences in crystallography and/or coordinative configurations as the (CO) L intensity increases while the (CO) H intensity declines with increasing activation temperature as a result of the beginning sintering process 18 .The (CO) L sites are therefore assigned to flat sites whereas the (CO) H sites are thought to be on rougher (high-index) planes or structural defects: steps, kinks, or corners 27 .Sintering the surface indeed causes a sharp relative decline on (CO) H intensity (at ca.2190 cm −1 ) in IR, and the sintered surface seems to have more extended and regular flat surface sites based on HRTEM images 39 .
Linear CO is reversibly adsorbed on the surface at room temperature as removing the CO from gas phase results in the disappearance of its IR band 20,27,54,55 .CO adsorption at room temperature at constant CO pressure shows constant intensity against time on stream if measured by the band at ca. 2192 cm −1 20,54 , indicating unactivated adsorption.Jung and Bell present an interesting scheme (see Fig. 10) relating linearly adsorbed CO and its interactions with the zirconia surface 25 .In the scheme they show two differently coordinated adsorbed CO molecules with bicarbonate and bidentate carbonate as transformation intermediates: in the former case the Zr 4+ cation will have a lower Lewis acidity in the vicinity of an OH group, leading to a lower displacement value of the IR wavenumber compared to the gas-phase CO IR band than with the bidentate carbonate intermediate in a c.u.s.oxygen environment 25 .Increasing the CO partial pressure increases the adsorbed CO band intensity 20,52 .The intensity ratio for the two adsorbed CO species favors (CO) H at low coverages and (CO) L at high CO pressures 18,27 .An increase in the CO partial pressure shifts the IR band position down from ca. 2195 cm −1 to 2188 cm −1 27,50,52,54 , and the overall surface area of the band indicates adsorption according to Langmuir's adsorption model with increasing CO partial pressure 54 .
Increasing the adsorption temperature shifts the main band at ca. 2190 cm −1 towards higher wavenumbers 18,55 .This shift is attributed to inductive effects, as the charge-release mechanism of the adsorbed CO is affected by those, as well as the influence of other surface species (e.g.OH) on the adsorbed CO 18 .The temperature range with detectable linear CO bands extends typically to ca. 100 − 150 • C 48,54,55 , but it has been reported even at 250 • C 25 at 2184.9 cm −1 (CO pressure not The reported linear CO coverages are scaled to µmol/m 2 and collected in Table 5.The amount of adsorbed CO depends on the adsorption temperature, the coverage at −173 • C is significantly higher than the coverage at room temperature.As can be seen in Table 5, dehydroxylation increases the linear CO adsorption capacity 20 , as dehydroxylated surfaces have a higher number of bare zirconium cations.Increasing activation temperature results in increasing monolayer capacities for both (CO) L and (CO) H 27 , as expected due to lower hydroxyl coverage with increasing activation temperature.It has been estimated that 50% of the dehydroxylated sites can adsorb CO reversibly 20 .The capacity for the (CO) H species is significantly lower than for the (CO) L species, the latter almost fourfold compared to the former 27 .
Dulaurent and Bianchi have assumed Langmuir adsorption, calculated adsorption coefficients from IR data and used them with statistical thermodynamics to extract heats of adsorption, and their results range from 55 kJ/mol to 42 kJ/mol (at zero and saturation coverage, respectively) 55 .Molar heat of adsorption determined with microcalorimetry is reported to be 65 − 73 ± 2 kJ/mol for (CO) H and 44 − 50 ± 2 kJ/mol for (CO) L 27 , and for Lewis-acidic sites at vanishing coverages ca.60 kJ/mol.Based on theory, the adsorption energy of linearly adsorbed CO was determined to be −45 kJ/mol 28 .To give an idea of the strength of the CO adsorption on the Zr 3+ sites, CO on Zr 4+ (at ca.2200 cm −1 ) can be removed by evacuation at room temperature and CO on Zr 3+ (at ca.2110 cm −1 ) is slightly more strongly bound to the zirconia surface, yet also possible to evacuate at room temperature 29 .The observed IR band intensities of the (CO) H and (CO) L species with increasing CO pressures are in line with their heats of adsorption: the (CO) H with a higher heat of adsorption has a higher intensity at low pressures and and vice versa at higher pressures 27 .
As shown in Table 5, the linear CO adsorption capacity on dehydroxylated surfaces is higher than on hydroxylated surfaces.The hydroxyl species has an adverse effect on CO adsorption as linear CO 20,27,28 , completely suppressing CO adsorption at room temperature already at a surface concentration of 2.4 µmol/m 2 H 2 O 27 , corresponding to a 20% surface coverage.The more strongly adsorbed linear CO species, (CO) H , seems to be suppressed more than the more weakly adsorbed species when the sample is changed from a dehydroxylated one to one with a low OH coverage 27 , suggesting that the site for (CO) H is the preferred site for hydroxyl formation.Four irreversibly held water molecules are required to eliminate one acidic site based on adsorption capacity experiments at varying degrees of hydration 27 .This 4:1 ratio between water and CO suggests that adsorption sites for linearly adsorbed CO represent only a minority of the sites available for water adsorption.This division is also reflected in the ad-sorbed amounts of water or OH groups (Table 3) and linearly adsorbed CO (Table 5).
In addition to increasing the adsorption capacities, dehydroxylation seems to shift the bands of the adsorbed CO species up in wavenumber, once full dehydroxylation has been carried out by evacuation at 597 • C, the bands only decrease in intensity: especially the (CO) H band intensity decreases as is expected due to the sintering process first affecting the minority sites 27 .Morterra et al. indicate that on a highly hydrated surface, local interactions among hydroxyls exceed the adsorbate-adsorbate interactions caused by CO, i.e., the ordered CO oscillator network is interfered with the hydroxyls present 39 .
Morterra et al. have looked at CO adsorption to cationic Zr sites after rehydration and they report that the CO species adsorbing at 2145 cm −1 (assigned to CO adsorbed to Zr 4+ centers via a σ bond) are quickly suppressed with water, but the lower wavenumber band tends to downshift from 2112 cm −1 to 2102 cm −1 (proposed to be c.u.s.cationic center Zr n+ , where n < 4) and increase in intensity 8 .The overall surface coverage of charge-releasing CO species inductively affect also the position of the adsorbed CO band 8 .
Even though hydroxylation decreases the linear CO adsorption capacity 27,58 , surface hydroxyls are an important surface site for CO adsorption as formate species 5,21 .Linearly adsorbed CO intensities during room-temperature adsorption have been compared before and after CO adsorption at elevated temperature 54,55 .Dulaurent and Bianchi report that after CO adsorption at 85 • C or 152 • C, cooling to 27 • C and another CO adsorption, the absorbance of the linear CO band is reported to decrease by 12% or 35%, respectively 55 .Mugniery et al. show spectra where formate preadsorption at 300 • C shifts the linear CO bands from 2192 cm −1 to 2177 cm −1 54 .When these bands are compared to the spectra of Morterra et al. 38 we note that the (CO) H species is suppressed by formates, linking the formate and the (CO) H to the same Zr 4+ surface site.As mentioned earlier, when investigating the CO pressure effect on the band, (CO) L is the preferred species at high coverages.However, it is not clear whether that applies also to the increasing formate coverage (or coverage of any species) or if the (CO) H site is occupied or otherwise hindered due to the adsorbed formate species.

Formates, formation and decomposition
Formate species consists of a HCOO − unit connected to a surface zirconium cation from oxygen atom(s).Two different surface configurations have been proposed for the formate species: a bidentate formate 20,22,30,42 and a monodentate formate 22,42 shown in Fig. 11.Main IR bands of formate species on monoclinic zirconia are observed typically (see Fig.   1379 cm −1 and ca.1365 cm −1 5, 19,22,23,88 , listed with their originating vibrations as well as thoretical IR bands in Table 6.Unlike for the linearly adsorbed CO species, formate formation (see Eq. 2) is an activated process, and its rate expression is shown in Eq. 3.
Due to the activated formation process, at low temperatures (e.g.T < 200 • C) the adsorption time affects the amount of formate formed, whereas at high temperatures the system quickly reaches equilibrium, however, the equilibrium coverage is also temperature-dependent.The activated nature of the process is demonstrated with increasing intensity of the formate IR bands at different temperatures (25 − 350 • C) with adsorption times ranging from 30 min up to 18 h 19,42,54 .Formate formation requires rearrangement of at least three bonds, the cleavage of the O-H bond, and the formation of the O-C and the C-H bonds 28 .The theory-based estimate for the activation energy of formate formation is 154 kJ/mol 28 , which is in agreement with the experimental observations, yet no experimental value for the activation energy has been reported.
Overall the temperature range of formate observations is wide, ranging from ca. 85 • C up to 550 • C 7,20,24,25,28,48,55 .The formate intensity maximum is at ca. 300 − 400 • C 23,28 , depending on the pretreatment and measurement conditions, with increasing intensities reported at lower temperatures 48 .All in all, the formate coverage depends on the adsorption conditions (temperature, CO pressure) and the contact time with CO.With increasing temperature (240 − 400 • C) the formate coverage decreases 23 while formate intensity increases with increasing CO concentration in the gas phase 23 .Silver et al. have adsorbed CO on pure ZrO 2 at 500 • C and they only discovered formates on the surface, no (bi)carbonates or adsorbed CO species 45 .Combination modes (ν as (COO)+ν s (COO)and ν s (COO)+δ (CH)) for theory-based vibrations were obtained as a sum of the corresponding frequencies.b The band not originally assigned as formate band rather as methoxide.c The author states that the "absorption in the 3000 cm −1 region is very weak", a band near 2900 cm −1 is discernable in the spectrum.

Catalysis Science & Technology Accepted Manuscript
At 250 • C the monodentate formate has a maximum at 1589 cm −1 and bidentate formate has maxima at 1568 cm −1 , 1388 cm −1 and 1371 cm −1 42 .The authors also suggest that the formation of monodentate formate is intensified with increasing time in contact with CO, in agreement with activated formation process, and that the shift from bidentate to monodentate formates is due to repulsion among the bidentate formates at high surface concentrations 42 as shown with increasing time-on-stream (CO at 250 • C) at 1568 cm −1 separating into two bands at 1589 cm −1 and 1556 cm −1 .
According to Bianchi et al., the formate species would be probably a bidentate as the difference between the observed bands at 1567 cm −1 (v as (OCO)) and 1367 cm −1 (v s (OCO)) is 200 cm −1 20 , and the same deduction has been used by Ma et al. for formate bands at 1570 cm −1 and 1361 cm −1 7 , both assigning the species based on the band separation.For carbonates, the typical difference should be ca.100 cm −1 in the monodentate case and 300 cm −1 in the bidentate case 89 .Bianchi et al. also state that another formate species might cause the band at 1382 cm −1 ; however, it should also have a doublet band near 1570 cm −1 , close to the one of the bidentate formate 20 .
Korhonen et al. have suggested a reaction scheme where the bidentate formate formation proceeds via an activated mono-dentate complex 22 .They state that based on DFT calculations the formate is likely in a bidentate configuration as the monodentate is unlikely to be stable 22 .Our investigations indicate a bridging bidentate formate configuration as the most stable geometry on all the tested surfaces 28 .
Formate forms on a surface hydroxyl species.Pozdnyakov and Filimonov stated already in 1972 that the formate is formed due to CO reacting with the surface hydroxyls 5 .Yamaguchi et al. in 1978 have shown formation of formate and disappearance of terminal OH and multicoordinated OH bands following the adsorption of deuterated acetone-d 6 , the terminal OH being more reactive toward formates than the multicoordinated one 59 .Amount of formate formed is dependent on the surface hydroxyls, a decrease in formate formation shown by Jackson and Ekerdt by removing water from CO/H 2 feed 10 and by Bianchi et al. by dehydroxylating the surface 20 .He and Ekerdt suggested that formate formation proceeds via gasphase CO and surface OH group 21 .
Formate formation has been reported at low temperatures (25 • C and 160 • C) on terminal OH (IR band at 3770 cm −1 ) 20 .In addition to the terminal OH site, formate formation on multicoordinated OH (band at 3680 cm −1 ) has been reported at higher temperatures (250 − 350 • C) 7,20,42 ; however, some experimental 23 and theoretical 22 results do not support the par-

Catalysis Science & Technology Accepted Manuscript
ticipation of the multicoordinated OH species.Jung and Bell suggested that the primary route for formate formation is via gas-phase CO and an OH group, after 9 hours at 250 • C and 162 kPa CO, all terminal hydroxyl species are consumed to formate formation as well as 38% of the bridged hydroxyl species 42 .Based on their spectral evidence 42 , it seems that the consumption of terminal OH species is faster than that of multicoordinated OH species, yet whether all bridged hydroxyls can be consumed is unclear based on the evidence.Jackson and Ekerdt suggested that formate formation in methanol synthesis involves an oxygen vacancy and an adjacent bridged hydroxyl site so that there is a terminal CO intermediate, the scheme is shown in Fig. 13 10 ., and in our investigations suggesting that the dehydration is two separate reactions: first reversible formate decomposition to CO resuming surface hydroxyls and then dehydroxylation to produce H 2 O, as the dehydroxylation process is observed at a similar temperature range also without CO present in the gas phase 28 .A lower limit estimate for the activation energy for the dehydrogenation reaction is its reaction energy at 178 − 363 kJ/mol based on theory 28 .The typical temperature range for formate de-composition is above 300 • C 28 , a desorption maximum has been reported at 410 • C 20 .The activated formate formation (increasing uptake rate up to 300 • C) and the formate decomposition pathways are demonstrated in Figure 14, where the zirconia sample is linearly heated from 100 • C to 550 • C in the presence of 2% CO, the y-axis corresponds to release/uptake from the sample.
Fig. 14 Temperature-programmed surface reaction (TPSR) in the presence of CO on reduced ZrO 2 .The y-axis corresponds to release (+) or uptake (-) from the sample.Reproduced from ref. 28 with permission from the PCCP Owner Societies.
Bianchi et al. have reported amounts of CO, CO 2 and H 2 that have adsorbed/desorbed during temperature-programmed desorption (TPD) after CO adsorption 20 .Not all carbon species are recovered, suggesting that the surface is not empty of formate (ca.10% of the CO adsorbed at 350 • C unaccounted for) by the end of the desorption process with T max at 410 • C 20 .Based on the observed CO 2 /H ratio the surface species is claimed to be formate 20 .The decomposition routes of formate have been reported to be either completely reversible decomposition resulting in restoring the terminal hydroxyls 23 or that only about 20 . . .40% of formate decompose forming hydrogen while the rest is decomposed reversibly 20,28 .For ceria-based catalysts, the presence of coadsorbed water on the catalyst significantly increased the decomposition rate of formate to CO 2 and H 2 90,91 and the same likely applies to zirconia as well.Also Korhonen et al. report observations of CO 2 as formate decomposition product above 500 • C 22 .The presence of an active metal (e.g.platinum) increases the rate of formate decomposition to CO 2 and H 2 23 , presumably by associating hydrogen (analogous to water-gas shift reaction), enabling a reasonable temperature window instead of 500 • C or more.
The adsorption temperature has a significant effect on the

Catalysis Science & Technology Accepted Manuscript
amount of CO desorbed from the surface as CO 2 during TPD, yet the overall profile of the CO 2 desorption curve remains qualitatively similar 30  If CO adsorption at 85 − 250 • C is followed by cooling down to 25 • C, the intensity of the linear CO band at 2190 cm −1 is smaller (by 12% with T ads at 85 • C, 35% at 152 • C) with increasing preadsorption temperature compared to room temperature preadsorption 20,55 .This is assigned to the formation of formate species at cationic Zr sites 54 which are thought to be the sites where CO adsorbs linearly 6,18,27,29 .When combined with our observations that formate formation at 100 • C in the presence of CO in the gas phase is accompanied by decreasing linear CO intensity (see Figure 15) 28 , it is suggested that site competition takes place and that linear CO facilitates formate formation compared to gas-phase CO.Assuming a bidentate formate species formed on a terminal OH and bound to a Zr cation (see Fig. 11), the cation site necessary for linear CO adsorption is blocked by the formate.
Measured adsorbed or desorbed CO amounts reported in literature are collected in Table 7. Increasing adsorbed/desorbed amounts of CO are reported with increasing adsorption temperature up to 350 • C 20,30 .Both Bianchi et al. 20 and Pokrovski et al. 30 have applied a similar TPD method, where adsorption is carried out at elevated temperature (T ads in the table) followed by cooling to room temperature, and thereafter temperature-programmed heating begins.The desorbed amount of CO x reported by Bianchi et al. 20 (0.12 µmol/m 2 ) are significantly lower than those reported by Pokrovski et al. 30 (0.35/1.34 µmol/m 2 ) after adsorption at 250 • C, while the specific surface areas are ca.200 m 2 /g and 19/110 m 2 /g, respectively.The values reported in our recent work 28 seem to be larger than those by others; this might be explained with a different experimental procedure, where weakly bound CO is not removed from the surface prior to temperatureprogramming.
As mentioned previously, the terminal OH group is the active species concerning formate or bicarbonate formation on the zirconia surface after gas-phase adsorption of CO at ele-vated temperature (240 − 400 • C) 23 .The activity of the terminal OH group has been further investigated using isotopelabeled experiments with D 2 O and D 2 .When deuterated formates were formed via CO adsorption to OD (deuteroxyl) species, they could be transformed back to HCOO with contact to hydrogen at 200 • C 19 .H/D isotope exchange of the surface formates with gas-phase D 2 seems to be possible, yet slow and competing with formate decomposition already around 300 • C 26 .Only 2 − 3% of the surface formate species are exchanged to DCOO species at 150 • C in 488 kPa D 2 42 .The necessity of gas-phase D 2 for formate scrambling was demonstrated as the H/D exchange did not proceed between formates and surface deuteroxyls 26 .However, as formates can form on surface deuteroxyls achieved by surface treatment with D 2 O 19 , it is implied that once formed, formates are stable and do not scramble with each other via cleavage of the O support -C bond.

Other species formed during interaction with CO
In addition to linear CO and formates also other species have been observed during CO adsorption.These species reveal the diversity of the interaction between CO and monoclinic zirconia although the number of reported observations remains low.The observed species include bidentate carbonates 48 , probably also carboxylate species as the band at 1416 cm −1 and its symmetric counterpart at 1560 cm −1 have been confirmed via difference spectra 20 .Monodentate carbonate (1469 cm −1 ) and ion carbonate bands (1303 cm −1 and 1442 cm −1 ) were observed after CO adsorption at 350 • C 7 .Similarly, ionic carbonate and carboxylate species have been suggested, their bands disappear when CO is removed from the gas phase at 400 • C while the formate species remain intact 54 .When comparing CO adsorption at 350 • C on hydroxylated and dehydroxylated samples, the latter shows less intense formate bands but more intense bands at 1440 cm −1 and 1416 cm −1 as well as new bands at 1540 cm −1 and 1317 cm −1 , suggesting carbonates present on the surface and perhaps also carboxylate species 20 .Also bidentate carbonates have been reported form during high-temperature adsorption (above 250 • C) of CO 48 .Ma et al. have suggested that bicarbonate and carbonate species could be formed on ZrO 2 from CO via carboxylate surface species 7 .He and Ekerdt have suggested that CO is adsorbed on the metal oxide oxygen forming a [COO] intermediate and then reacting further to carbonate or formate 21 .All these species require the participation of one or two surface oxygen atoms.
CO adsorption followed by carbonate formation and CO 2 desorption leads to surface reduction as oxygen is removed from the surface.During temperature-programmed surface reaction (TPSR) in CO, CO 2 amounts detected correspond to 10 − 14% of surface oxygen atoms depending on pretreat- During static adsorption of CO at 200 • C and above, CO 2 formation was observed in the IR spectra, originating either from formate decomposition or from CO oxidation via lattice oxygen species 62 .Ionic carbonate was observed during CO adsorption at 150 • C and above on rehydrated zirconia, and it was released as CO 2 instead of CO 20 .Silver et al. reported a small bicarbonate desorption peak during temperature-programmed heating in CO (25 − 620 • C), however, no gas-phase CO 2 or IR bands of carbonates or bicarbonates were observed during CO exposure at 500 • C 45 .These observations suggest that formation of bicarbonates takes place below 500 • C, as expected based on the knowledge of bicarbonate species (for more information, see 28 and references therein).

Future perspectives
As both monoclinic zirconia and its interaction with CO have been investigated for more than 40 years, in some regions knowledge is still lacking.The surface configuration is one of the remaining questions, as the most stable surfaces according to density functional theory are different from those assumed based on HRTEM, yet the amount of independent experimental observations on clearly monoclinic samples remain few.The stability of the surface structure in the reaction conditions and during the reaction should be investigated carefully.Also the IR designation of terminal and multicoordinated hydroxyls (or any of the other interpretations) remains without irresputable evidence, even though the multi-oxide studies by Tsyganenko and Filimonov 3,71 were very thorough in providing comparable information of several oxides with presumably different OH groups due to their crystal structure.Especially in the light of theoretical calculations suggesting mostly terminal and bibridged hydroxyl species, further confirmation of the assignment would provide more clarity.The Zr cations (Zr 3+ , Zr 4+ ) have also caused some confusion: the basis for the IR assignment is unclear, yet the trivalent species have  been successfully probed with N 2 O 6 .The EPR assignment of Zr 3+ species has also been under discussion 8,46,47,51,57 .Further information on surface vacancies and also surface defects, e.g., at the surface boundaries could give a more thorough look at the surface interaction with hydroxyl and CO species.Also tailoring the properties of monoclinic zirconia by using promoters and dopants without changing its crystalline phase could elucidate the interaction.Advanced methods similar to Raman in the case of ceria 93 might also provide surprisingly rich information.

Catalysis Science & Technology Accepted Manuscript
With the adsorbed CO species the knowledge on especially the formate species deserves more investigation.The surface configuration of the species (monodentate, chelating or bridging bidentate) is unclear based on experimental observations, although bidentate species has been speculated and theoretical calculations suggest both chelating and bridging bidentate species.Also the enthalpy of formation for formate species has only been estimated theoretically, an experimental confirmation for it, e.g., with microcalorimetry, would be welcome.Some clarity on the kinetics of formate formation on terminal and multicoordinated hydroxyls or even the extent of the reaction on both types of hydroxyls could provide new insights related to catalysis.Operando-style experiments with combined surface and gas-phase quantification would provide valuable input on all of the surface species with CO and hydroxyls.Observing the surface species in the same setup under vacuum might enlighten the mechanism of surface reduction via oxy-gen removal and why evacuation is a more efficient reduction method than hydrogen treatment 80 .
To improve the modeling of carbon oxides, a better understanding of an active surface site is highly important to be able to set up more accurate computational surface models to better describe the complexity of the ZrO 2 support/catalyst.This would not only impact on calculation of adsorption energies but would also influence on calculated frequencies.As long as the nature of an active site is not known exactly, a systematic approach, where, e.g., different surface models are investigated side by side, is a natural choice to obtain information of adsorption characteristics.To discover the active surface sites and to further improve the selectivity of catalysts, nanoshaped supports and catalysts with their surface regularity have proven an interesting alternative.The better the control over the surface sites, the better the catalyst selectivity.However, the control is only to be reached through monocrystalline nanoshapes as in polycrystalline shapes the surfaces are not controlled.
Growing monoclinic zirconia in the shape of nanofibres and nanorods has become gained some attention during the last 15 years, yet most of the shapes are polycrystalline.The preparation methods differ from traditional wet chemistry to prepare monoclinic zirconia powders.Nanorods have been prepared hydrothermally 16,17 in an autoclave, resulting in nanorods of various sizes, the diameter in general some tens of nanometers and the length a few hundred nanometers, and the length-

Catalysis Science & Technology Accepted Manuscript
diameter ratio ranging from ca. 5 up until 50 or even more.
Exposed faces of the nanoshapes should be characterized with advanced electron microscopy techniques (e.g.aberration-corrected TEM as in 94 ) to determine the orientation of the exposed surfaces.Boucher et al. 95 have tested different shapes of metal oxides with gold catalysts for steam reforming of methanol and water-gas shift, and they conclude that the different shapes show somewhat different activities.Li and Shen 96 discuss widely the oxide shape effects in nanocatalysis, and they mention some of the unknown issues with metal deposition on oxide nanoshapes: whether the metal atoms are located on a single type of surface only.They also bring up that the nanoshapes might not be stable under the reaction conditions, this might also affect the metal-support interface, often considered to be the active site 96 .

Conclusions
Monoclinic zirconia surface has three kinds of coordinatively unsaturated cationic sites, two types of Zr 4+ and one type of Zr 3+ , coordinatively unsaturated Zr 4+ -O 2− pairs, oxygen vacancies, terminal hydroxyls, and two types of multicoordinated hydroxyl species.The ratios of these sites can be modified with pretreatments by removing or adding oxygen and hydrogen to the surface through applying heat, vacuum or reactive atmospheres.The cationic sites are responsible for the linearly adsorbed CO species while formates are suggested to form preferably on a site where an unshielded zirconium ion is paired with a terminal hydroxyl species, assuming a bidentate formate.This would provide an explanation for the submonolayer quantities of formate on zirconia.The concentrations of active sites for linear CO and formate formation are of similar magnitude, corresponding to ca. 5% of a monolayer or less, whereas the amount of hydroxyl species on the surface is roughly tenfold.The formates as well as hydroxyl species prefer the defect type of Zr 4+ sites.The specific roles for Zr 3+ and the terrace-type Zr 4+ remain unclear.Other open questions include confirming the nature of the multicoordinated hydroxyls, the surface configuration of the formate species, and the energetics of formate formation.Nanoshapes might be a valuable tool in exploring the surface aspects related to formates and hydroxyls.

Fig. 1
Fig. 1 Ideal surfaces of monoclinic ZrO 2 , top view, Zr atoms are grey and O atoms red.The atoms further in the lattice are shaded lighter than those on the surface.The surfaces were visualized using VESTA software 78 .

Fig. 5
Fig. 5 Difference spectra of OH/OD species before and after 30 min in 100 torr D 2 at 150 • C. Reproduced from ref.87.

8 | 1 - 19 Catalysis
Science & Technology Accepted Manuscript temperature seems to require unhydroxylated c.u.s.Zr sites, as pretreatment above 600 • C is necessary to remove adsorbed water from the zirconia surface 26,49 .Even though Bianchi et al. have observed increasing hydroxyl species intensities in IR during H 2 treatment, they report no adsorption or desorption of hydrogen at 25 −400 • C, and no water evolution during H 2 -TPR from 25 • C up to 700 • C 53 .Assuming no redox process concerning the Zr cations, a hydrogen desorption mechanism is postulated for Zr-H and Zr-OH sites leading to Zr and Zr-O sites, as hydrogen adsorbed at 550 • C is desorbed from m-ZrO 2 at 600 • C 9 .According to Syzgantseva et al., hydrogen dissociates on Zr 3+ with a neighboring oxygen vacancy (v o ), leading to formation of Zr-H hydrides and the transformation of Zr cations into Zr 4+ species

Fig. 10
Fig. 10 Interactions of linearly adsorbed CO with the ZrO 2 surface.Reproduced from ref. 25 with permission.Copyright 2000, Elsevier.
reported).Ma et al. observed a band at 2109 cm −1 during CO adsorption at 350 • C, linked to CO adsorption on c.u.s.Zr3+ 7  .

58a
Assuming full adsorption on the surface from fed known amount of CO.IR band in the presence of CO gas at 800 Pa is thrice as intensive as the one with 2.7µmol/m 2 .b Saturation pressure, 70 torr for samples activated at 397 • C and 597 • C, 20 torr for sample activated at 797 • C. c Estimated to be 30% higher than adsorption capacity at 130 torr.

Fig. 13
Fig.13Proposed formate formation mechanism as a part of methanol synthesis over ZrO 2 .Adapted from ref.10 with permission.Copyright 1986, Elsevier.

Formate
decomposition has been proposed to take place by two different pathways: dehydrogenation producing CO 2 and H 2 , and dehydration releasing CO and H 2 O, to follow the naming of He and Ekerdt 21 .Similar decomposition pathways have been proposed by Bianchi et al. suggesting the release of CO and restoring the OH groups . CO desorption was clearly observable only after adsorption at 200 • C or 250 • C. At 250 • C there are formates on the surface according to IR 30 , thus the CO desorption is likely due to decomposition of surface formate species and the CO 2 originates from formates or, especially after low-temperature adsorption of CO, (bi)carbonate species.The temperature of maximum desorption was in the range of 330 • C for CO 2 and 330 − 430 • C for CO desorbing after CO adsorption.Köck et al. have adsorbed CO on ZrO 2 from room temperature up until 600 • C, yet their pretreatment (annealing at 900 • C and thereafter oxidation at 600 • C) of the sample has quenched most of the surface hydroxyls, leaving formate formation negligible and thus supporting the formate formation mechanism based on surface hydroxyl species 62 .

a
Department of Biotechnology and Chemical Technology, Aalto University School of Chemical Technology, P.O.Box 16100, 00076 Aalto, Finland.Email: sonja.kouva@aalto.fib Department of Chemistry, Nanoscience Center, University of Jyväskylä, P.O.Box 35, 40014 Jyväskylä, Finland c Faculty of Science & Technology, University of Twente, P.O.Box 217, 7500 AE Enschede, The Netherlands.

Table 1
Origin, preparation conditions and surface area of the zirconia samples a a Tsyganenko et al.

Table 3
Hydroxyl densities reported in literature

Table 5
Linear CO coverages reported in literature

Table 7
Adsorbed or desorbed CO amounts reported in literature Depending on sample preparation temperature (600 − 900 • C). b Depending on pretreatment (hydration, reduction, reduction and hydration).CO contact first 90 min at 100 • C, then continuing during heating up to 550 • C.