Rational Development of Guanidinate and Amidinate Based Cerium and Ytterbium Complexes as Atomic Layer Deposition Precursors: Synthesis, Modeling, and Application

Abstract Owing to the limited availability of suitable precursors for vapor phase deposition of rare‐earth containing thin‐film materials, new or improved precursors are sought after. In this study, we explored new precursors for atomic layer deposition (ALD) of cerium (Ce) and ytterbium (Yb) containing thin films. A series of homoleptic tris‐guanidinate and tris‐amidinate complexes of cerium (Ce) and ytterbium (Yb) were synthesized and thoroughly characterized. The C‐substituents on the N‐C‐N backbone (Me, NMe2, NEt2, where Me=methyl, Et=ethyl) and the N‐substituents from symmetrical iso‐propyl (iPr) to asymmetrical tertiary‐butyl (tBu) and Et were systematically varied to study the influence of the substituents on the physicochemical properties of the resulting compounds. Single crystal structures of [Ce(dpdmg)3] 1 and [Yb(dpdmg)3] 6 (dpdmg=N,N'‐diisopropyl‐2‐dimethylamido‐guanidinate) highlight a monomeric nature in the solid‐state with a distorted trigonal prismatic geometry. The thermogravimetric analysis shows that the complexes are volatile and emphasize that increasing asymmetry in the complexes lowers their melting points while reducing their thermal stability. Density functional theory (DFT) was used to study the reactivity of amidinates and guanidinates of Ce and Yb complexes towards oxygen (O2) and water (H2O). Signified by the DFT calculations, the guanidinates show an increased reactivity toward water compared to the amidinate complexes. Furthermore, the Ce complexes are more reactive compared to the Yb complexes, indicating even a reactivity towards oxygen potentially exploitable for ALD purposes. As a representative precursor, the highly reactive [Ce(dpdmg)3] 1 was used for proof‐of‐principle ALD depositions of CeO2 thin films using water as co‐reactant. The self‐limited ALD growth process could be confirmed at 160 °C with polycrystalline cubic CeO2 films formed on Si(100) substrates. This study confirms that moving towards nitrogen‐coordinated rare‐earth complexes bearing the guanidinate and amidinate ligands can indeed be very appealing in terms of new precursors for ALD of rare earth based materials.


Introduction
Rare earth (RE) metal containing materialsa re very interesting in different fields of applications ranging from opticalc oat-ings, [1] opticalw aveguides, [2] catalysis, [3] protective coatings, [4] fuel cells [5] to high-k materials [6] in the microelectronic industry. In particular,c erium oxide (CeO 2 )i se ncouraging for catalysis, [7] waters plitting, [8] solid oxide fuel cells, [9] protective coatings, [10] and is also considered as ap ossible high-k gate dielectric material in complementary metal-oxide-semiconductord evices. [11] Yb containing films are finding increasing attention in the advancement of semiconductor devices [12] with ytterbium-doped optical fibers being relevantf or high power laser applications. [13] Thus,r ecently the interesti nt he growth of high-quality RE-based materials [14] has been on the rise, particularly for conformal coatings with ap recise tunable thickness on complex architectures which can be obtained by atomic layer deposition( ALD). [15] ALD is ap owerful technique that uses as elf-limiting growth mechanism by employing pulses of ag aseous chemical metalorganic compound (precursor) and as uitable co-reactant for the desired thin film materialw hich are separated by inert gas purges to ensure layer-by-layer growth of conformal, uniform and pinhole-free films. [16] Due to the unique surface saturation caused by the chemical surface reactions, the ALD processi s strongly dependento nt he chemical properties of the employed metal precursor. [16a] Therefore, ALD precursors must fulfill several requirements.F irst, they should be reactive towards the substrate surface and the co-reactant. Secondly, they should be volatile to be brought into the gas phase and thermally stable for ap rolonged time at the chosen evaporation temperature. [17] Thirdly,t hey should at least be thermally stable on the time scale of an ALD cycle to prevent their decomposition and uncontrollabler eactions at given deposition temperatures. Naturally,l ong-term stabilitya tt heset emperatures is a practical feature. Additionally,aprecursor in the liquid state is advantageous as it can provideareproducible rate of vaporization more likely than as olid one. From ac hemistry point of view,t hesep roperties can be tuned by modifyingt he ligand of am etal-organic complex,t om eet the demands of the process. Typically,l igands [18] such as b-diketonates, [19] cyclopentadienyls, [20] alkoxides, [21] bis(trimethylsilyl)amides, [22] amidinates [23] and guanidinates [24] are used in the case of rare-earth metals.S ince the majority of available RE precursors do not satisfy one or more requirementsd escribed above, the deposition process can be affected negativelys ot hat the properties of the deposited films do not match the desired specifications. The number of reports on the suitable precursors for rare earth metals particularly for cerium and ytterbium, is limited in comparison to other metals.
Early-generation precursors such as RE-alkoxides have been demonstrated to exhibit poor volatility as they tend to oligomerize and therefore, could only be applied successfully in liquid injection delivery ALD systems [25] as shownf or [Ce(mmp) 4 ]( mmp = 1-methoxy-2-methyl-2-propanolate) in combination with water. [21,26] Contrasting this, RE-b-diketonates often require strong oxidizing agents such as ozone (O 3 )a sthe already existing REÀOb onds, whichc ontributet ot he thermal stability, exhibit al ow reactivity towards mild oxidizing agents. Furthermore, they require high volatilization temperatures of 140-170 8Ci nc ase of [Ce(thd) 3 ] [27] or [Yb(thd) 3 ], [28] (thd = 2,2,6,6-tetramethyl-3,5-heptanedionato), respectively which compromises on their applicability in the low temperature regime. RE-cyclopentadienyl (RE-Cp)complexes andtheir derivatives have demonstrated higher volatility in comparison to RE-b-diketonates while maintaining high thermal stability.T he homoleptic precursor [YbCp 3 ] [20] was reported for thermalA LD of Yb 2 O 3 with water as ac o-reactant and [CeCp 3 ] [29] was employed for CeO 2 togetherw ith O 2 plasma in ap lasma-enhanced ALD (PE-ALD) process. Contrasting the former,t he reactivity of the cerium derivate was insufficient to react with water and required as tronger oxidant. Addressing this crucial shortcoming, heteroleptic Ce complexes with Cp and amidinate ligands [30] were introduced as ALD precursors. They possess higher thermals tabilitya nd volatility than the homoleptic RE-Cp 3 complexes and are reactive towards water.F urthermore, the precursors are liquid at room temperature and can be evaporated at 145 8Ci nc ase of the [CeCp 2 (iPr-AMD)] [bisisopropylcyclopentadienyl-N,N'-diisopropylacetamidinate-cerium(III)] [31] However,t he preparationo ft hese precursors is an intricateprocess and generally delivers low yields.
Homoleptic RE-tris-amidinates and tris-guanidinates [32] containing Gd, Dy,E r, Yh ave been demonstrated to be promising for the ALD of rare-earth oxide (REO). In contrastt oo xygenbased ligands, the presenceo fs ix REÀNb onds makes them strongly oxyphilic, which promotes their high reactivity towards mild oxidizing agentsl ike water.M oreover,s teric, and electronic properties can be tuned by varying the steric bulk at the N-C-N backbone.I na ddition, the bidentatechelating effect of amidinates and guanidinates providest hermal stabilityt o the resulting RE complex. [24] Previously,R E-guanidinates have been successfully utilized for the growth of RE oxides such as Gd 2 O 3 , [33] Dy 2 O 3 , [33] Er 2 O 3 [34] and Y 2 O 3 [35] in water-assisted ALD. Furthermore, the reactivity of Er and Yb tris-guanidinates was provent ob es uitable enough for the deposition of inorganicorganic hybrid materials by atomic/molecular layer deposition (ALD/MLD). [36] Apart from the experimentals tudies, the correlation of their thermals tability and their reactivity with different co-reactants via theoretical calculations is advantageous to gain better insighti nto systematic precursor engineering, which to the best of our knowledge has not been carried out beforef or any rare earth precursors. The present study reports on all-nitrogenc oordinated RE complexes that are promising new Ce and Yb derivatives. Given that they are promising as highly reactive ALD precursors, ar ecent ALD study on the homoleptica midinate [Ce(N-iPr-AMD) 3 ]( tris(N,N'-diisopropylacetamidinato)cerium(III)) highlighted high evaporation temperatures of 170 8C. [37] In order to address this and to understand the effect of the substituents on the N-C-N backboneo fa midinates and guanidinatesi nt erms of the physicochemical properties such as evaporation behavior,aseries of Ce and Yb complexes was rationally designed. Herein, we report as ystematic study by varying the organic moieties attached to the Ca tom of the N-C-N backbone by Me, NMe 2 ,N Et 2 ,a nd by varying the N substituent from iPr to tBu and Et to investigate the influence on the volatility,s tability,a nd reactivity of af amily of complexes.
The resulting thin films were analyzed with respect to their crystallinity,composition, and optical properties.

Precursor synthesis and characterization
The synthesis of all the complexes 1-7 was achieved by as alt metathesis reaction of the anhydrous metal chloride MCl 3 (M = Ce, Yb) and three equivalents of the respective lithium (Li)s alts of the ligand [Li(NR 1 )(NR 2 )C(R 3 )] (R 1 = iPr, tBu;R 2 = iPr,E t; R 3 = Me, NMe 2 ,N Et 2 )a ss hown in Scheme1.
The lithium salts werep repared in situ by insertionr eaction of LiMe, LiNMe 2 ,L iNEt 2 into N,N'-diisopropylcarbodiimide or Ntert-butyl-N'-ethylcarbodiimidei nt etrahydrofuran (THF) for cerium complexes 1-5 and in diethyl ether (Et 2 O) for ytterbium complexes 6 and 7.T he purification of the products wasa chieved by recrystallization and/or sublimation.T he products were obtained in good yields (70-90 %). From the handling of the complexesd uring synthesis, some general statements can be derived. All the complexeshave good solubility in THF,pentane, hexane, benzene, Et 2 Oa nd can be applieda sp otential precursors forc hemical solution deposition thin film processes as well. The cerium complexes 1-5 are extremely sensitivet o air and moisture and immediately turn black when they are exposed to air and moisture;t hese observations are in agreement with those for other organo-cerium complexes. [39] All the complexesa re paramagnetic in nature( one unpaired electron in Ce 3 + and one in Yb 3 + ), necessitating large chemical shift ranges (ppm) for the measurement of proton and carbon nuclear magnetic resonance (NMR) spectra of the cerium complexes. For the ytterbium complexes, the recording of the spectra failed due to significant paramagnetic shifts. 1 HNMR was measured for compounds 1-5 in C 6 D 6 ,t he chemicals hift peaks were found to be broad in the range À10 ppm to 11 ppm and the intensity was significantly reduced. However, the integrationo ft he peaks matched with the number of hydrogen atoms present in the complexes (Figure 1). Ad etailed NMR study was done to relate the paramagnetic shifts caused due to the unpaired electron of Ce 3 + in the complexes. 13 CNMR was performed ( Figure S1 in the Supporting Information) and correlated with the 1 HNMR by 1 H- 13 Figure 1e)c an be found (peak A) at 7.52 ppm and 6.36 ppm, the CH 3 protons of the Et group (peak D) at À5.72 ppm and À6.16 ppm and the CH 3 protons of the tBu groups (peak C) at À2.83 ppm and À3.27 ppm, respectively. Noticeably,t he CH 3 protons of the iPr moieties are similar for the complexes 1, 3,a nd 4 (peak C) but behave very differently in our NMR studies:While the peak CinFigure 1f or complex 1 is very broad in the range of À10 ppm to 4ppm (Figure 1a), for complex 3 it is split into two peaks at 2.53 ppm and À6.47 ppm, (Figure 1c)a nd for complex 4 it is as ingle sharp peak at À3.99 ppm. (Figure 1d)T oi dentify the reason,t emperature dependent 1 HNMR was performed on complex 1,F igure S3. From the spectra,a tÀ50 8Ct wo clearp eaks at 4.33 ppm and 10.30 ppm for the CH 3 protons of the iPr group are observed. This can be explained by ah indered rotationo f the CH 3 protons, resulting in chemically different environments for the CH 3 groups and hence, splitting of the signal is observed. As the temperature increases, these two peaks are broadened and at 50 8C, both the peaks coalescet of orm one peak at 1.91 ppm. Upon further heating to 100 8C, it forms a sharp peak at 1.50 ppm. At higher temperatures, the rotation of the ligand is increased and hence the CH 3 moieties of the iPr groups rotate fast enough that the CH 3 protons become equivalent yielding one signal. Similarly,f or complex 3 even Scheme1.Generalreaction scheme for the synthesis of homoleptic rareearth guanidinate and amidinate complexes. higher sterich indrance caused by the NEt 2 group restricts the rotationf or the CH 3 groupso ft he iPr moieties and facilitates the appearance of two peaks (peak C) at 2.53 ppm and À6.47 ppm at room temperature. In complex 4,t he steric hindrance is the lowest due to the small Me group at the N-C-N backbonea nd hence the rotationi sn ot hindered, which leads to one signal at À3.99 ppm at room temperature. Furthermore, for complex 1 in Figure S3, the CH protons of the iPr group (peak A) and the CH 3 protons of C-NMe 2 group (peak B) are shifted upfield with increasing temperature because the binding strength of the ligand lowersd ue to the increased rotation and hence electron density increases at the ligand.
Electron-impact mass spectrometry (EI-MS) analysis was carried out to confirm the formationo ft he target compound, analyze the fragmentation pattern, and to get an insight into the structuralf eatures of the metal-organic complexes.T he EI-MS spectra for all the complexes are given in the Supporting Information (Figures S6-S12) and selected peaks with assigned fragments are listed in Table 1. For all the complexes, the respective molecular ionp eaks (ML 3 + )w ith expectedm ass to charge ratios (m/z)a nd considerable relative intensities of 14.8 %f or 1, 9.4 %f or 2,4 6.8 %f or 3,3 8.8 %f or 4,4 4.6 %f or 5,2 .0 %f or 6, 7.0 %f or 7 were found. Peaks at higher m/z ratios than the molecular ion peak were not observed under experimental conditions, whichs uggests that all the complexes can exist as monomers in the gas phase. For all the complexes, the fragments from cleavage of one ligand (ML 2 + ), or of two ligands (ML + )s pecies, as well as the ligand L + itself, were detected. Interestingly,f or all amidinate complexes,t he fragment with the highesti ntensity (100 %) is the ML 2 + fragment. The guanidinates seem to decompose into smaller fragments, indicated by the peak with 100 %i ntensityw hich is observed for fragments associated with the organic ligands and their decomposition fragments. As imilar fragmentation behavior is observed for the literature knownr are-earth tris guanidinates [24,40] andt risamidinates [40a, 41] complexes.
The molecular structures of 1 and 6 were determined by single-crystal X-ray diffraction( SC-XRD) and are depicted in Figure 2a,b while the crystallographic data is given in Ta ble S1. The crystalso fs uitable quality for measurement could not be obtained for other complexes. Complex 1 crystallizes in the monoclinic crystal system in the C2/c space group with 8m olecules per unit cell having ac alculated density of 1.299 gcm À3 . Complex 6 crystallizes in the triclinic crystal system in the P1 space group with two molecules per unit cell having ad ensity of 1.299 gcm À3 .B oth the complexes exist as monomers in the solid-state and are isostructural with six-fold coordination of nitrogent ot he lanthanide center.T hus, it is surroundedb y three bidentate h 2 -guanidinato ligandsw ith at rigonal planar structure. This structure is accounted to ad elocalized p-electron system as known from the literature [24] and indicated by a mean bond length of N b ÀC = 1.336(5) for 1 and 1.335(1) for 6,a nd N m ÀC = 1.400(3) for 1 and1 .399(4) for 6 (Table 2), that is shorter than at ypical CÀNs ingle bond length of 1.474 . [42] Here, the nitrogen atoms which constitute the N-C-N backbonea re labeled as N b (N1/N2/N5/N6/N8/N9)a nd the nitrogen atoms connected to methyl groups are labeled as N m (N3/N4/N7).
The coordination geometry of the complexes (as shown in Figure2ca nd d) can be described as distorted trigonal prismatic.T his is indicatedb yt he torsion angle N b -Q1-Q2-N b ,r anging between 15.388 and 17.908 for 1 and between 21.628 and 22.618 for 2 (Table 2), where Q1 is the centroid of the backplane of the prism spanned by N1, N5, N8, Q2 is the centroid of the frontp lane of the prism spanned by N2, N6, N9 and here, N b are the nitrogen atoms of the same guanidinate ligand that are coordinated to the metal center. Ideally,the torsion angle is 08 for at rigonal prismatic geometry and 608 for the octahedral geometry.T hus, the structure is distorted by a twisted offset of the two planest owarde ach other.  Table 2a nd can be correlatedt oo ther lanthanide guanidinate complexes, reportede arlier ( Figure 3). The previously reported isostructural complexes bearing identical guanidinate ligands [24,40] for rare earth metals show a relationship between their effective ionic radius (M 3 + ) [43] and specific geometrical parameters including the MÀNb ond length and the N-M-N bite angle of the guanidinate ligand.   The ionic radius along the series of lanthanides (La-Lu) decreasesdue to poor shielding of the 4f electrons known as lanthanidec ontraction. [44] The ionic radius for Ce 3 + and Yb 3 + is 1.01 and 0.868 ,r espectively. [43] As can be seen from Ta ble 2 the MÀNb ondl ength in compounds 1 and 6 range from 2.480(3) to 2.527(4) for 1 and from 2.327(2) to 2.341(3) for 6.T he different MÀNb ondl engths result in mean bond lengths of 2.500(1) for 1 and 2.335(6) for 6 which is in agreement with the trend depicted in Figure3of al ongerM À Nb ond for an increasing ionic radius. The N-M-N bite angles of the guanidinate ligands are ranging between 53.48(6)8 and 53.94(6)8 for 1 and 57.61(9)8 and5 7.78(9)8 for 6 which results in am ean bite angle of 53.78(6)8 for 1 and 57.71(9)8 for 6.T his observation is matching again the trend depicted in Figure 3, indicating as maller bite angle for larger rare-earth ion centers.
As the guanidinate ligand itself is quite rigid because of p system,t he trends in the MÀNb ond lengths and N-M-N bite angles resultsi nt he twist of the trigonal planeso ft he trigonal prismatic structure, expressed by the N b -Q1-Q2-N b torsion angle. This is larger for as maller rare earth ion. The mean values are 16.988 for 1 and 22.238 for 6 which match this trend.

Evaluation of thermal properties
To evaluatet he potentiala pplication of ac ompound as ap recursor for ALD applications, the study of the thermalp roperties is important and in this contextt he volatility,m elting point, and thermal stabilityw ere investigated.
With low melting points being generally desirable, they can also be used as af irst indicator for the extent of intermolecular interactions present in ac ompound. The meltingp ointo ft he metal-organic complexes was analyzed by differential scanning calorimetry (DSC, not shown), and the resultsa re summarized in Table S2. It is found that the meltingp oint for complex 1 is 104 8Ca nd for complex 2 the melting point is 88 8C. The difference can be explained by the asymmetry in the molecule because of which the crystal packing and hence the entropy of crystallization is affected. [45] Complex 3 has ah igher melting point of 134 8C. The lowest melting point of 50 8Ci so btained for complex 5 which can be ascribed to ah igh asymmetry in the molecule. The ytterbium complex 6 is melting at ah igher temperature of 110 8Ct han its isostructural cerium analog 1. No melting point was observed for the complexes 4 and 7.
Thermogravimetric analysis (TGA) was employed to study the evaporation behavior and stabilityo fc omplexes 1-7.T he weightl oss of the complexes as af unction of temperature in the range of 35 8CÀ400 8Ci ss hown in Figure 4, and their onset of volatilization[ 8C] and residual weight [%] are given in Ta ble S2.
As it can be seen for the Yb complex 7,t he initial mass loss is very low and the onset of volatilization, here defined as the temperature at which 1% weight loss occurs, is at 110.6 8C.
The main weight loss step is observed at ah igher temperature of 253.5 8C( here defined as the step temperature assessedb y the method of tangents [46] )a fter which ar esidual weighto f 5.2 %was observed. Thus, the weight loss can mostly be attrib-uted to the evaporation of the intact complex. Complex 6 shows an onset of volatilization at 155.3 8Cw hichi ss ignificantly higher than that of compound 7,a nd the step temperature is 253.3 8Cw hich is nearly the same temperature as for complex 7.T he residual mass of 6 is 20.8 %w hich is considerably higher than that of 7,y et lower than any expected decomposition product of Yb (nitride, carbide). Hence, it indicates that thermale vaporation overlaps with decomposition from which evaporation is the predominantp henomenoni nb oth Yb complexes. However,t he volatility of 7 is higher than that of 6 as indicated by al ower onset temperature of volatilization for 7 and al ower rest mass. The potentiala pplicability of the complexes 6 and 7 is indicated based on the TGA of the Yb complexes.
Observationsonthe Ce compounds appear to be partly contrasting. While the residual weights of all TG measurements for the complexes 1-5 weren ot negligible, they are still lower than possible Ce decomposition products (nitrides, carbides), indicating againt he coexistence of evaporation and decomposition under the applied experimental conditions upon increasedh eat exposure. Complex 4 shows ao ne-step weight loss with an onset of volatilization at 70.8 8C, and the step temperaturei s2 39.9 8C. For the asymmetricc omplex 5,t he onset temperature is slightly higherw ith7 6.4 8Ct han for 4 and a two-step weight loss is observed. The first step at at emperature of 218.5 8Ci st he major step which ends with ar est mass of 27.5 %. However,t he remaining substance, likelyadecompositionp roduct, undergoes further evaporation at as econd step at 277.8 8Cr esulting in af inal residual mass of 15.5 %. Hence, the asymmetry in complex 5 resultsi nl ower thermal stabilityc ompared to complex 4.F or complex 1,t he onset is 90.1 8Cw hich is higher than fort he homoleptic amidinates of cerium, but the step temperature of 239.6 8Ci ss imilar for 1 and 4.T his indicates that the complex 4 is more volatile than complex 1.A fter this step, furtherw eight loss can be observed beforei tb ecomesc onstant with ar esidual weight of 25.2 %indicating that the decomposition product is still volatile. The complexes 2 and 3 show onset temperatures of 76.4 8Ca nd 67.1 8C, respectively.T he step temperature for 2 is at 193.1 8C and for 3 at 197.3 8Ca nd is again accompanied by decomposi-tion. The residual masses obtained were 34.1 %a nd 33.4 %, respectively,i ndicating lower thermal stability for the complexes 2 and 3 when comparedt oc omplex 1.
This study exemplifies how systematic variation of substituents on the side chains and backbones of the employed amidinate and guanidinate ligandsc an be used to tune the thermal properties of the precursors. The thermal properties of complexes having the same molecular mass and an identicalc omposition, namely 1, 2,a nd 4, 5,d iffer noticeably.D ue to the large asymmetry in its structure, complex 5 has the lowest meltingo fa ll the complexes and complex 2 possesses al ower meltingp oint than 1.O nt he other hand, the higher asymmetry in the molecular structure can lead to lower thermal stability and it was additionally found that the thermals tability and volatility of the amidinates is highert han the one of the guanidinates.I nterestingly,t he cerium guanidinate 1 is meltinga ta lower temperature than the isostructural ytterbium complex 6. Besides, the cerium complexes 1-5 are found to be more volatile than ytterbium complexes 6 and 7 while exhibiting less thermalresilience.

DFT studies
To obtain an insight into the fundamental aspects of the chemistry of the compounds on the molecular level, DFT was used to modelt he atomic structures and to simulatet he reactivity of the complexes towards potential co-reactants. In the first set of calculations, the precursors [Ce(dpdmg) 3 ] 1,[ Ce-(dpamd) 3 ] 4,[ Yb(dpdmg) 3 ] 6 and[ Yb(dpamd) 3 ] 7 were modeled as isolated molecules in vacuum at zero Kelvin (K) and zero Giga Pascal( GPa), with the relaxed atomic structures shown in Figure S13. The geometries of the complexes were accurately reproduced by DFT calculations.T he MÀNb ond lengths and bite angles (N-M-N) of the optimized structures 1, 4, 6,a nd 7 are given in Table S3 and Ta ble S4 respectively and are consistent with the solid-state studies.
The bond dissociation energy defined as the energy for the removal of the first ligand,w as computed for all compounds. Based on bond dissociation energies, as showni nT able 3, it is anticipated that cerium guanidinate 1 will be more stable compared to cerium amidinate 4 under vacuum conditions. It is to be noted that, under thermal conditions other reactions or decomposition pathways can also occur;f or example, for guanidinates,t he carbodiimide deinsertion [47] can also take place which has not been taken into consideration for DFT experiments.F or the Yb precursors, the trend is the same, although the differencei sl ess significant. Taking into consideration the bond dissociation energies, 6 and 7 would have similars tability under vacuum conditions. Based on the bond dissociation energies and the overall trend described above,t he stabilitya nd potentialr eactivity of these precursors is not necessarily correlated to the bonding properties within their molecular structure. We would expect that precursorsw ith the shorter MÀNb onds would be less reactive;h owever,t his is not apparent from the data in Table S3 and Ta ble3.
Keeping this in mind, we expanded the model system of the precursors to include the interaction with one O 2 and one H 2 O molecule, respectively and investigated the reactivity again at zero Ka nd zero GPa.A nO 2 molecule was placed at 2.50 from the metal center in its gas phase geometry and was allowed to relax. Figure 5a-d showst he relaxed atomic structure of the precursorsa fter the incorporation of the O 2 molecule and demonstrates that the interaction with O 2 depends on the metal center. Ce promotes the breaking of the O=Ob ond in both precursors, which is typical for Ce 3 + species. [48] For the complexes 1 and 4 (Figure 5a,b) one oxygen atom insertsi nto the originalC e ÀNb ond creating new CeÀOa nd OÀNb onds while the second oxygen atom binds with the Ce centerf or both the complexes forming an oxo-ligand.
In contrast to cerium, ytterbium does not break the O=O bond, Figure5c,d. Instead, in the complexes 6 and 7,t he O 2 molecule forces its way to closep roximity to the metal center and forms at ricycle with Yb while one YbÀNb ond is cleaved, which again results in a7 -fold coordination sphere. One of the oxygen atoms forms an OÀNb ond with the non-metal coordinated N. The OÀOb ond length is found to be in the range of 1.45 to 1.46 which is characteristico faperoxide species. Ta ble S5 shows the MÀO, OÀN, and MÀNb onds in the presence of oxygen.
According to bond dissociation energies shown in Table 3, 1 would be more reactive compared to 4 with O 2 .F or the Yb precursors, the difference in bond dissociation energies is almostn egligible and slightly changed from the gasp hase precursor.T his suggests that the reactivity of the ytterbium containing precursors is little affected by the nature of the ligand regarding the interaction with O 2 . Figure 5e-h shows the optimizeds tructures of the precursors after the interaction with one H 2 Om olecule. When one H 2 Om olecule interacts with the precursors, it preferably binds to the central Ce and Yb atom and dissociates. The OH group of water binds to the M, andt he remaining Ha tom binds to nitrogen upon metal-nitrogen bond breakage.
Ta ble S6 shows the MÀOH and MÀNb onds in the presence of water.O nce the water molecule hasr eactedw ith the metal center, dissociation of the semi-protonated, solely one-fold bonded ligand was identified as ap referentiald issociation pathway.B ased on the computed bond dissociation energies, 1 is expectedt ob em ore reactive comparedt o4.a nd 6 would be more reactive compared to 7.T hus, the reactivity of all these precursors can be strongly influenced by the interaction with O 2 or H 2 Om olecules. This study shows that these precursors are potentialc andidates for ALD precursors. Interestingly, the bond dissociation energy for the cerium complexes 1 and 4 is found to be less in the presence of an oxygen molecule than in the vicinity of aH 2 Om olecule which suggestst hat the elemental O 2 couldb ea ni nteresting co-reagent forA LD with Ce complexes.

ALD of CeO 2 thin films
Based on the promising resultso btained from the thermal characterization of the precursors as well as the DFT studies, the next objective was to evaluate the precursors for ALD applications.A sarepresentative case, we chose [Ce(dpdmg) 3 ] 1 as it was found to be very reactive towards water based on our DFT calculations. Therea re very few reports on water assisted ALD, as highlighted in the introduction.T hus, such a study can widen the library of water assisted ALD processes for RE oxides. In this context, proof-of-principle ALD experiments on Si(100) substrates were performed with [Ce(dpdmg) 3 ] 1 using water as co-reactant.
To verify the self-limiting nature of the thin film growth, a saturation study of the precursor vaporized at 140 8Ca nd the co-reactant water maintained at room temperature was carried out for ad eposition temperature of 160 8C ( Figure 6a). The precursor pulse lengthw as varied from four to twelve seconds, while the other parameters werek ept constant with ap recursor purging time of 30 s, aw ater pulse length of 3s,and water purging time of 30 s. As seen in Figure 6a,t he precursor satu-rates after 8spulse with ac onstant GPC of 2.1 ,t hereby confirming as elf-limiting growth. Similarly,t he water purge length was varied from 15 st o4 5s,F igure 6a,w hile the other parameters were kept constant.T here was no change in the GPC after more than 30 so fw ater purge. The increased growth was observed below 30 sofw ater purge time, probablyd ue to ar eactiono fa dditional adsorbed water molecules on the surface with the precursor,w hich has also been observedi na similarp rocess for the deposition of Y 2 O 3 in the same reactor type. [35] The dependencyo ft he film thickness on the number of applied cycles was subsequently analyzed as shown in Figure 6b for the precursor 1 pulse/purge/waterp ulse/purges equenceo f8s/30s/3 s/30 s( illustrated in Figure S14). The obtained fit value R 2 of 0.99913 shows that for each cycle, the same amount of material is deposited and therefore, the thickness can be tuned precisely.T hese initial set of resultsi nt erms of validating ALD growth characteristics further confirm that the precursor is suitable for water assisted ALD. More detailed experiments varying the processp arameters have to be performed to optimize the new ALD process for CeO 2 .
The crystallinity of ALD grown thin films was assessed by grazing incidence X-ray diffraction (GIXRD). As exemplarily illustrated by the GIXRD pattern obtained for a4 2nmt hick film grown at 160 8C ( Figure 6d)t he as-deposited layers possess a polycrystalline nature matching the computed reflections of a refined cubic CeO 2 reference pattern (ICDD:0 4-016-4620). Xray reflectivity (XRR) patterns obtained for films with varying total number of cycles, i.e.,d ifferentt hicknesses, are shown in Figure6c. From the respective fits, an average thin film density of around 5.0 gcm À3 could be estimated based on the critical angle fitting.  Ta ble S5 and Table S6.
Seeking furthere vidence for the formation of the highvalent CeO 2 phase and to obtain insights into the chemical composition of the films, X-ray photoelectron spectroscopy (XPS) wasc onducted on a4 2nmt hick film deposited on Si(100) substrate. The Ce 3d core level spectrum of the as introduced film is shown in Figure 6e and represents the conditions of the surfacei namaximum depth of around 5nm. Contributionsofb oth Ce 4 + and Ce 3 + species at the expected binding energies (listed in the Supporting Information, Ta ble S7)t o the core level were identified as the formationo fo xygen vacancies and partial reduction of Ce 4 + to Ce 3 + species are known phenomena on ceria surfaces. [50] Following the method developed by Romeoe tal. [51] that is well described by Preisler et al., [49] fitting of all spin-orbital and splitting components and summation of peak areasa ssociated to Ce 3 + and Ce 4 + (see Ta ble 4) allowed to estimate the concentration of the first-men-tioned to be 24.8 %a nd of the latter-mentionedt ob e75.2 %. Herebyt he componentsd escribed as v 0 and u 0 as well as v" and u" represent Ce 3 + species while v, u, v'',u '' as well as v''' and u''' are associated with Ce 4 + .T he Ce 3 + /Ce 4 + ratio for the untreated thin film surfacew as found to be in good agreement with prior reports on ALD grown films. [52] In light of this, analysiso ft he O1 sc ore level (see Figure 6f) allowed to confirmt he off stoichiometry of the thin film surface. Next to the O 2À species associated to CeO 2 lattice oxygen at 530.0 eV [49,53] and adsorbed OH À species at 532.9 eV, [37] a minor contribution from O 2À species arising from Ce 2 O 3 [49,53] at 531.6 eV was found. The binding energies for all components were well within the range of positions prior reported for the respective species. In terms of the overall surface composition, determined for the as introduced surfacea nd after 60 so fA r + sputtering, the complete absence of nitrogen impurities was noteworthy.W hile the carbon concentration diminished from 30.5 at.% (contribution from adventitious carbon) to roughly 6.0 at.%, the Ce/O ratio decreased from 1.99 to 1.53Àac onsequenceo fA r + ion induced reduction. [50b] The results are summarized in Ta ble S7.
The bandgapo fa26 nm thick CeO 2 film deposited on quartz was estimated by the measured UV/Visible absorption spectrum in the range 200-800nm. As trong absorption peak is observed in the UV region at 304 nm Figure 7a,d ue to the charge-transfer transition from O(2p) to Ce(4f) orbitals in Figure 6. (a) The black data points representp recursor saturation studiesb yv arying [Ce(dpdmg) 3 ] 1 pulse length and the red data points representt he variation of GPC with water purge time (b) Thickness of the film vs. number of applied ALD cycles;both at the deposition temperature of 160 8Co nS i(100). (c) XRR patterns of films with varying total numbero fc ycles( black dotted line representst he simulated pattern)( d) GIXRD patterns at an incident angle of 0.58 of the film deposited on Si (the black dashedl ine represents the refined computational pattern with reference to cubic CeO 2 (ICDD:0 4-016-4620)) (e) XPS analysisoft he Ce 3d core levels pectrum of the as introduceds urface of a42nmthick CeO 2 film growno nS i(100). Experimental and fitting curves for all spin-orbital splitting's are given following the nomenclatureofP reisler et al. [49] (f) XPS analysiso ft he O1sc ore level spectrum for the same film. Experimental and fitting curves for all oxygenc omponents are given.  [54] The film shows high transparency as indicatedb y transmittance values of > 92.5 %i nt he range between 450 nm À800 nm. The Ta uc plot method was utilized for the direct and indirect opticalb andgapc alculations (see Figure 7b). The obtained direct allowedb andgapw as estimated to be 3.36 eV, and the indirect allowed bandgapw as 2.66 eV,w hich are consistent with those reported in the literature. [11,52] The preliminary data for the growth of CeO 2 thin films at mild substrate temperatures via ALD and the film characteristics shows that the initial resultsu sing [Ce(dpdmg) 3 ] 1 precursor and water as ac o-reactant are highlye ncouraging for ALD applications. The next step is to vary the ALD process parameters to develop ar eliable water assisted ALD process and then investigate the functional properties of the CeO 2 film. Particularly the facile conversion of cerium between Ce + 3 and Ce + 4 oxidation states and the tunable oxygen vacancies makes it interesting for solid oxide fuel cells [9] and catalytic activity for water splittinga pplications. [8] Similar studies will be performed with other analogous precursors of Ce and Yb to compare their efficiency for new ALD process development.

Conclusion
In the pursuit to identify new and improved precursors for ALD of Ce and Yb containing thin films, as ystematic approach was undertaken tuning the ligand moieties surrounding the metals namely cerium and ytterbium. As ar esult,aseries of cerium and ytterbium complexes [Ce(dpdmg [Yb(dpdmg) 3 ] 6,[ Yb(dpamd) 3 ] 7 were successfully synthesized in good yields. The complexes can be quantitativelys ublimed and exist as monomers in the gas phase. Noteworthy was the influence of the asymmetry in the molecule that could alter the meltingp oints and thermals tabilities of the different compounds investigated. DFT study was used to analyze in detail the atomistic structure and the reactivity of 1, 4, 6,a nd 7 in vacuum and in the presence of O 2 andH 2 Omolecules. Interestingly,i nt he presence of aH 2 Om olecule, the bond dissociation energy is lower for 1 and 6 than for 4 and 7,s uggestingt hat guanidinate compounds exhibit ah igher reactivity towards water,aw ell-established ALD co-reagent comparedt ot he structurally related amidinates. The presence of O 2 molecules had almost no effect on Yb complexes,o nt he contrary 1 was found to have higherr eactivity toward elemental O 2 than 4 suggesting that it could also be used as ap otentialp recursor for ALD with molecular oxygen. Based on the promising thermal properties in terms of volatility and thermal stabilitya s well as data inferred from the reactivity of the molecules towards water from DFT studies, these complexes certainly bear the potential to serve as new ALD precursors. Thus, proof of principle studies for water-assisted ALD was performed with [Ce(dpdmg) 3 ] 1,y ielding polycrystalline CeO 2 thin films on Si(100) substrates. The co-existence of Ce 3 + and Ce 4 + oxidation states in the films was evidenced from XPS analysis. UV/Vis analysiss howed the direct allowed and indirect allowed bandgaps and hence these films could find scope for potential opticala nd catalytic applications which will be the focus once the ALD process is optimized. Additionally,t hin CeO 2 layers can be investigated as dielectric layers for high-k applications. As an outlook, ac omparativeA LD investigation of the guanidinates vs. the amidinates for both Ce and Yb will be the focus of our future work. This study,w hich comprises of ar ational approachu ndertaken towards new precursor development for Ce and Yb, enlightens the powero fs ynthetic organometallic chemistry involving rare earths.I th as always been ac hallenge to develop monomeric, volatile and reactive precursors for rare-earths and hence the output of this study has substantially contributedt ot he expansion of the library of raree arth based precursors which to date has been particularly limited for Ce and Yb.

Experimental Section Precursor synthesis
The handling and syntheses of all air and moisture-sensitive compounds were carried out under argon atmosphere using standard Schlenk techniques. The solvents used were dried by as olvent purification system (MBraun SPS). Synthesis of tris(N,N'-diisopropyl-2-dimethylamido-guanidinato) cerium(III) [Ce(dpdmg) 3 ]1 :T he synthesis procedure was adopted based on the literature. [39b] . N,N'-diisopropylcarbodiimide (1.82 mL,11.76 mmol) was added dropwise to ac ooled solution of lithium dimethylamine (0.6 g, 11.76 mmol) in tetrahydrofuran (THF) (25 mL). The resulting solution was stirred for to form [Li-(NiPr) 2 CNMe 2 ]. In another flask THF was added to anhydrous CeCl 3 (0.946 g, 3.91 mmol) and stirred to form as uspension. [Li-(NiPr) 2 CNMe 2 ]w as added to the suspension and the mixture was refluxed at 60 8C. The resulting solution was cooled to room temperature (RT) and the solvent was removed, and the product was extracted in hexane while the precipitated LiCl was filtered off. After removal of the solvent under reduced pressure, the resulting yellow product was sublimed in vacuum at 120 8Ct hat yielded 2.09 go fayellow crystalline product. Crystals suitable for singlecrystal X-ray analysis were obtained by sublimation. Yield:8 2.07 %.  (NCH(CH 3 ) 2 ) 2 CNMe 2 ]. 13   14 mmol) suspension in THF (15 mL) and refluxed at 60 8C. Following the same work up as described above, 1g of yellow powder was obtained. Yield 71.64 %. 1

Synthesis of tris(N,N'-diisopropyl-acetamidinato) Ytterbium (III)
[Yb(dpamd) 3 ]7 :F ollowing the same procedure as for 6,[ Li-(NiPr) 2 CMe] was prepared by N,N'-diisopropylcarbodiimide (2.5 mL, 16.14 mmol) and 1.6 m LiMe in hexane (10.09 mL, 16.14 mmol) in diethyl ether (70 mL). [Li(NiPr) 2 CMe] was added to ac ooled solution of YbCl 3 (1.5 g, 5.38 mmol) in THF (100 mL) and stirred. Following the same work up as described above, al ight green product was obtained which was further purified by sublimation to obtain 2. 8  Precursor characterization 1 H, 13 Cn uclear magnetic resonance (NMR) and heteronuclear single quantum coherence (HSQC) spectra were measured on aBruker AV III 400 spectrometer and aB ruker AV III 300 spectrometer at 298 K. All signals were referenced to the residual proton signals of deuterated solvents and corrected to the TMS (tetramethylsilane) standard values. Te mperature dependent NMR was recorded on a Bruker Advance DPX 250 spectrometer.F or that, the compound was filled in ah eavy-walled NMR tube and dissolved in af reshly prepared degassed [D 8 ]toluene and sealed by melting. The NMR spectra received were further analyzed with the MestReNova software. [55] CHNS elemental analysis (EA) was performed using av ario Micro cube from Elementar Analysensysteme. Electron impact ionization mass spectrometry (EI-MS) was performed using aV GInstruments Autospec instrument at an ionization energy of 70 eV.F ourier-transform infrared (FTIR) spectroscopy was performed between 400-4000 cm À1 using aSpectrum Twoinstrument from PerkinElmer with an attenuated total reflectance (ATR) unit in an argon-filled glove box. Thermogravimetric analysis (TGA) was carried out by using aN etzsch STA4 09 PC/PG at ambient pressure (sample size % 15 mg in ar ound alumina crucible having diameter 6.15 mm. For TGA, ah eating ramp of 5 8Cmin À1 and nitrogen (AirLiquide, 99.998 %) flow rate of 90 sccm was used. Melting points were determined by simultaneous differential scanning calorimetry (DSC) (mW mg À1 ).
Crystals of suitable quality were selected from perfluoropolyether oil on amicroscope slide under an optical microscope with apolarized light source and was immediately mounted in al iquid nitrogen cooled gas stream of ad iffractometer. Ac rystal of [Ce(dpdmg) 3 ] 1 was measured on an Agilent Te chnologies Super-Nova diffractometer with an Atlas CCD detector and Cu Ka radiation from am icrofocus X-ray source with multilayer X-ray optics and [Yb(dpdmg) 3 ] 6 on an Oxford Diffraction Xcalibur2 diffractometer with aSapphire2 CCD and Mo Ka radiation. The diffraction data were processed with CrysAlisPro. [56] Empirical absorption correction was done using spherical harmonics, implemented in SCALE3 AB-SPACK scaling algorithm. The crystal structures were solved and refined by using SHELXL, [57] SHELXLe-2014 [58] and OLEX2. [59] Deposition Numbers 2023020 ([Ce(dpdmg) 3 ] 1)a nd 2023021 ([Yb(dpdmg) 3 ] 6)c ontain(s) the supplementary crystallographic data for this paper.T hese data are provided free of charge by the joint Cambridge Crystallographic Data Centre and Fachinformationszentrum Karlsruhe Access Structures service www.ccdc.cam. ac.uk/structures.

Computational method
The ground state electronic wave function of each molecule was calculated self-consistently within Kohn-Sham Density Functional Theory (DFT) using the TURBOMOLE suite of quantum chemical programs. [60] These calculations were performed by using the hybrid PBE0 functional, which incorporates 25 %e xact HF exchange, [61] and ap olarized split valance basis set, denoted def-SV(P). [62] An effective core potential is used for the Ce and Yb metal sites with 28 core electrons on both rare earths. Af ine integration grid (m3) was used and the SCF convergence criterion was set to 10 À6 Ha. Precursor atomic structures are prepared from the experimental cif files in Materials Studio 8.0 and exported in xyz format;a ll structures are freely available in aG itHub repository. [63] Convergence criteria for the geometry was set to 10 À3 Ha.
The energy needed to lose the first ligand is calculated using [Eq. (1)]: E L :C omputed total energy of one free ligand; E P :C omputed total energy of the precursor molecule; E PÀ1L :C omputed total energy of the precursor without one ligand For the example of [Ce(dpdmg) 3 ] 1 [Eq. (2)]: E Ligand ¼ðE dpdmg þE CeðdpdmgÞ 2 ÞÀE CeðdpdmgÞ 3 ð2Þ

Thin-film deposition
Cerium oxide thin films were deposited using tris(N,N'-diisopropyl-2-dimethylamido-guanidinato) cerium(III) [Ce(dpdmg) 3 ] 1 as the precursor and deionized water as the co-reactant. The synthesis of [Ce(dpdmg) 3 ] 1 was upscaled to large batches of ca. 10 gf or preliminary ALD experiments. All the depositions were carried out in a F-120 ASM Microchemistry flow-type ALD reactor on 2cm*2cm silicon and quartz substrates. The reactor is setup into eight zones to achieve ag radually increasing temperature profile from precursor zone to deposition zone. Nitrogen (99.999 %p urity) gas was implemented as ac arrier and purging gas at 300 sccm. The sublimation temperature for [Ce(dpdmg) 3 ] 1 was set 140 8C( zone 2) and H 2 Owas maintained at room temperature. The pulse purge sequence applied for thickness dependent studies (illustrated in the SI in Figure S14) at deposition temperature of 160 8C( zone 7) is precursor 1 pulse (8 s)/N 2 purge (30 s)/H 2 Op ulse (3 s)/N 2 purge (30 s).

Thin-film characterization
Film thickness was measured by X-ray reflectivity (XRR), and the film crystallinity was assessed by grazing incidence X-ray diffraction (GIXRD) using aP analytical XPert diffractometera nd Cu Ka source on silicon substrates. The GIXRD fitting was performed by using Reflex module in the Materials Studio 8.0 (BIOVIA Software Inc., USA). The background for GIXRD was calculated with aG aussian width of 0.01 and polynomial order of 2a nd subtracted. The reference GIXRD pattern was computed using cubic CeO 2 (ICDD:0 4-016-4620). The Rietveld refinement method with as mall degree of Zero shift and peak boarding was employed to achieve the fitting of GIXRD pattern; [64] aB raggBrentano function [65] was used for instrument geometry,T hompson-Cox-Hastings [66] for peak profile, and Finger-Cox-Jephcoat function [67] for Asymmetry correction. The X'Pert Reflectivity program v1.3 from PANalytical was utilized for the fitting of the XRR patterns. X-ray photoelectron spectroscopy (XPS) was carried out in aP HI 5000 instrument. The X-ray source was operated at 10 kV and 24.6 Wusing Al Ka (1486.6 eV) radiation with a4 5 8 electron takeoff angle. The kinetic energy of electrons was analyzed with as pherical Leybold EAÀ10/100 analyzer using a pass energy of 18 eV.T he samples were analyzed by core level scans for peaks of interest. The step width was adjusted to 0.05 eV for the core level scans. Spectra were recorded prior to and after sputter cleaning (1 min. 2kV2 2). All binding energies of cerium Ce 3d and oxygen O1sw ere referenced to adventitious carbon C1sa t2 84.8 eV.T he analysis chamber pressure was maintained at < 10 À8 mbar.T he deconvolution analysis was completed with a Shirley background processing and Gaussian functions using UniFit