Atomic layer deposition of dielectric Y2O3 thin films from a homoleptic yttrium formamidinate precursor and water

We report the application of tris(N,N′-diisopropyl-formamidinato)yttrium(iii) [Y(DPfAMD)3] as a promising precursor in a water-assisted thermal atomic layer deposition (ALD) process for the fabrication of high quality Y2O3 thin films in a wide temperature range of 150 °C to 325 °C. This precursor exhibits distinct advantages such as improved chemical and thermal stability over the existing Y2O3 ALD precursors including the homoleptic and closely related yttrium tris-amidinate [Y(DPAMD)3] and tris-guanidinate [Y(DPDMG)3], leading to excellent thin film characteristics. Smooth, homogeneous, and polycrystalline (fcc) Y2O3 thin films were deposited at 300 °C with a growth rate of 1.36 Å per cycle. At this temperature, contamination levels of C and N were under the detectable limits of nuclear reaction analysis (NRA), while X-ray photoelectron spectroscopy (XPS) measurements confirmed the high purity and stoichiometry of the thin films. From the electrical characterization of metal–insulator–semiconductor (MIS) devices, a permittivity of 13.9 at 1 MHz could be obtained, while the electric breakdown field is in the range of 4.2 and 6.1 MV cm−1. Furthermore, an interface trap density of 1.25 × 1011 cm−2 and low leakage current density around 10−7 A cm−2 at 2 MV cm−1 are determined, which satisfies the requirements of gate oxides for complementary metal-oxide-semiconductor (CMOS) based applications.


Introduction
Yttrium(III) oxide (Y 2 O 3 ) thin lms play an important and versatile role in various applications which arise from the valuable intrinsic material properties: the high relative permittivity (e r ¼ 14-18) and a large direct band gap (E g ¼ 5.5-5.8 eV) render this material useful as a high-k material for its implementation as a potential gate dielectric in metal oxide semiconductor eld effect transistors (MOSFETs). 1,2 Additionally it features an intrinsic hydrophobic surface owing to its special electronic structure among the other rare-earth oxides, 3 which together with its high chemical resistivity, 4 mechanical strength and melting point render Y 2 O 3 a useful candidate for application as protective coating even in very harsh environments. 5,6 Moreover, a high refractive index of n ¼ 2.1 enables the application of Y 2 O 3 as a waveguide in solid state lasers. 7,8 For all these applications, it is desirable that the coating is thin, while retaining its valuable intrinsic properties. Vapour phase depositions of thin layers of Y 2 O 3 on complex substrates, such as three-dimensional or sensitive surfaces is conveniently realized by atomic layer deposition (ALD). This technique enables the growth of a variety of materials with high degree of homogeneity, good compositional control and conformality due to controlled layer-by-layer growth. 9 Such a growth is initiated by saturative adsorption and reaction of the employed precursor on the surface of the substrate. The chemistry of the precursor and co-reactants have a signicant inuence on growth rate (growth-per-cycle, GPC), composition, structure and morphology of the thin lm. 10 In an ideal case, the growth rate within the so-called ALD window is independent of the deposition temperature, which in reality however is not a necessity to obtain high quality thin lms. 11 It should be noted that not only the growth rate, but also the chemical and physical quality of the resulting thin lms mainly dictate how broad or narrow the ALD window can be considered and thus in which range a high thin lm quality can be retained. In general, a broad ALD window can be achieved if the volatility and reactivity of the precursor is sufficient to prevent condensation and ensure chemisorption on the substrate at lower deposition temperatures, while the thermal stability and adsorption strength of the precursor prevents decomposition or desorption processes at higher temperatures. A rational choice of the ligands is a crucial step in the successful development of precursors to avoid thermal decomposition and ensure a high reactivity and volatility. For instance, incremental changes within the chemical backbone of the ligands can have a considerable inuence on the physicochemical properties of the precursors, ALD process parameters and quality of the resulting thin lms. In the past, different ALD precursors for the deposition of Y 2 O 3 have been employed. Especially the precursors based on the yttrium cyclopentadienyls [Y(Cp) 3 ] clearly demonstrate how small changes in the substitution pattern of the ligand can inuence the ALD relevant properties of the complexes, which has a strong inuence on their behaviour in the respective ALD processes (Fig. 1). 12,13 Exemplarily for water-assisted ALD processes, the cyclopentadienyl based complexes, [Y( H Cp) 3 ] and [Y( Et Cp) 3 ] show a similar growth rate in the range of 1.5-1.7Å per cycle, whereas the reported ALD process window is higher for [Y( H Cp) 3 ] with a maximum deposition temperature of 400 C. On the contrary, the volatility of [Y( Et Cp) 3 14,15 It should be noted that comparison of parameters in different ALD processes might not be directly possible if the processes are not optimized in the same type of reactor with similar geometries, ow rates, temperature gradients and many other ALD conditions. 16 This is directly apparent when comparing water assisted ALD processes using the heteroleptic yttrium isopropyl-cyclopentadienyl amidinate [Y( iPr Cp) 2 (DPAMD)] precursor which was employed in three different reactors: the growth per cycle (GPC) of 0.4-1.3Å, ALD-Windows (175-200 C; 200-350 C; 350-450 C), bubbler temperatures (120-150 C) and composition (C,N < 0.5 at% -C z 3.7 at%) vary signicantly, making a thoughtful and direct comparison to other precursors nearly impossible. [17][18][19][20] Overall, there are different precursors reported for water-assisted ALD of Y 2 O 3 (Fig. 1), which share the same drawback of low volatilities and thus necessitates high precursor evaporation temperatures that might limit their applicability in low-temperature ALD processes.
Thus, it is not only necessary to identify and develop new precursors and processes for the ALD of Y 2 O 3 lms but also equally important to optimize the processes employing different precursors in the same reactor setup to check reproducibility and conveniently compare the behavior of precursors and processes. This will enhance the understanding of precursor chemistry and process characteristics suiting the targeted applications. As shown earlier in studies by Rouf et al. 21 on the ALD of InN and Kim et al. 22 on In 2 O 3 , the change of the endocyclic substituents on the respective amidinate and guanidinate backbones (e.g. -H, -Me and -NMe 2 ) of the homoleptic complexes revealed a superior performance of the formamidinate derivative (-H) complexes in the corresponding processes. Herein, we report on a new water-assisted ALD process with the homoleptic precursor tris(N,N 0 -diisopropylformamidinato)yttrium(III) [Y(DPfAMD) 3 ] for the formation of Y 2 O 3 thin lms, which features distinct advantages such as higher volatility and favorable processing characteristics compared to the other known yttrium precursors of the same family. For [Y(DPfAMD) 3 ], the typical ALD characteristics were veried, the resulting thin lms thoroughly analyzed and nally applied in a metal-insulator-semiconductor (MIS) capacitor stack to investigate the electrical properties. Additionally, in this study, the newly developed ALD process employing [Y(DPfAMD) 3 ] is directly compared to a process employing the guanidinate [Y(DPDMG) 3 ] (Scheme 1) in the same reactor and under similar process conditions, 15 which clearly underlines the superior characteristics of the formamidinate backbone for the ALD of Y 2 O 3 thin lms.

Experimental section
The synthesis and handling of all reagents and compounds was carried out utilizing standard Schlenk protocols using Ar as an inert gas to prevent contact with ambient air and moisture. The precursors were handled and stored inside a MBraun 300 Glovebox system and the solvents were dried by a MBraun solvent purication system (SPS) and stored under inert gas atmosphere. All commercially available reagents were used without further purication. The two reported precursors [Y(DPDMG) 3 ] and [Y(DPAMD) 3 ] were synthesized according to literature known procedures by Milanov et al. and de Rouffignac et al., 14,23 while [Y(DPfAMD) 3 ] is commercially available. Electronic ionization mass spectra (EI-MS) were recorded by a Varian MAT spectrometer with direct sample injection. Thermogravimetric analysis (TG) was carried out with a Seiko Exstar TG/DTA 6500SII under a nitrogen ow (300 ml min À1 ) and a heating rate of 5 K min À1 using approx. 10 mg of each compound. The vapor pressure of the compounds was determined using stepped isothermal TGA. This approach is based on a study by Kunte et al. 24 For the depositions of Y 2 O 3 thin lms, 2 00 p-type Si(100) substrates with native oxide (SiO x , z2 nm) were used. The ALD experiments were carried out using a ASM Microchemistry F-120 reactor. The temperature of the precursor (200 mg for each deposition) was kept at 95 C for depositions carried out from 100-275 C and 98 C for depositions at 300-325 C with an active ow of 300 sccm N 2 . The water reservoir was always held at room temperature. Optimized pulse-purge sequence of the ALD process for the determination of the ALD window was 5 s of precursor pulse, 60 s of precursor purge, 5 s of water pulse and 30 s of water purge. The nal optimized sequence at a deposition temperature of 300 C is 5 s of precursor pulse, 10 s of precursor purge, 1 s of water pulse and 30 s of water purge. The thickness of the thin lms was determined via spectral reectance using a spectrometer F20 from Filmetrics. Grazing incidence X-ray diffraction (GI-XRD) was carried out using a PANalytical X'pert pro diffractometer using full 2 00 substrates. Thin lm density and the critical angle was derived via X-ray reectometry (XRR; Bruker D8 Discover XRD) with Cu-K a radiation (1.5418 A) in a Q-2Q locked coupled mode, while 2Q was increased from 0.1 to 3 with a step size of 0.01. Rutherford backscattering spectrometry (RBS) analysis and nuclear reaction analysis (NRA) were performed at the RUBION, Central Unit for Ion Beams and Radionuclides at the Ruhr University Bochum. For RBS, a 2.0 MeV 4 He + ion beam with an intensity of 20-40 nA was directed to a sample with an angle of 7 . The scattered particles were detected by a solid-state detector at 160 . NRA was performed to obtain the concentration of elements with a low atomic number like C, N and O. The concentration was obtained aer an induced nuclear reaction of the light elements by a 1.0 MeV deuteron beam and detection of the emitted protons at an angle of 135 . A 6 mm Ni foil was used to shield the detector from scattered deuterons. The beam penetrates the whole thin lm and is stopped in the sample substrate. The soware suite SIMNRA was used to determine the concentration of the elements in the thin lm, by using the data obtained by the RBS and NRA measurements. 25 X-ray photoelectron spectroscopy (XPS) was carried out in a PHI 5000 instrument. The X-ray source was operated at 10 kV and 24.6 W using Al Ka (1486.6 eV) radiation with a 45 electron take-off angle. The kinetic energy of electrons was analyzed with a spherical Leybold EA-10/100 analyzer using a pass energy of 18 eV. Aer measurements for the as introduced sample were completed, the surface was subjected to Ar + sputtering (1 min, 2 kV (2 Â 2)). The samples were analyzed by a combination of survey scans and core level scans for peaks of interest.
Step widths were adjusted to 0.5 eV for each survey scan and 0.05 eV for the core level scans. All binding energies of yttrium Y 3d, oxygen O 1s and other core levels were referenced to the Fermi edge position. The analysis chamber pressure was maintained at <10 À7 mbar. The deconvolution analysis was completed with a Shirley background processing and Gaussian functions using UniFit 2017 soware. The topography of the Y 2 O 3 lms was characterized by means of atomic force microscopy (AFM, Digital Instruments, Nanoscope V). Electrical characterization was carried out on metal-insulator-semiconductor (MIS) capacitors. For this, 20 nm thick Y 2 O 3 lm was deposited at T ¼ 300 C on a p + -type Si(111) substrate. 70 nm thick Pt gate electrodes were e-beam evaporated onto the Y 2 O 3 lms through a shadow mask. The diameter of the gate electrodes amounts to 70 mm. The capacitance-voltage (C-V) characteristics were measured using an Agilent E42821 A LCR meter. For the current-voltage measurements of the MIS structures a semiconductor parameter analyzer (HP Agilent 4156B) was used.

Results and discussion
Comparative mass spectrometry studies The most stable fragment [Y(DPfAMD) 2 ] + is then further split into smaller fragments, however the loss of another complete formamidinate ligand to [Y(DPfAMD)] + at m/z ¼ 216.0 was not observed. Also, a peak for the lone formamidinate [DPfAMD] + ligand was not detected at the expected m/z ¼ 143.2, which indicates that it directly splits into smaller and more stable fragments ( Table 1). The structurally related amidinate [Y(DPAMD) 3 ] exhibits a similar fragmentation behavior with an observable

Comparative thermogravimetric studies
To evaluate the behavior of the precursors under thermal exposure, thermogravimetric (TG) measurements were performed (Fig. 2). While all complexes feature a single-step evaporation, [Y(DPfAMD) 3 ] shows an onset of volatilization at 151 C (derived via tangents) which is signicantly lower than [Y(DPAMD) 3 ] at 197 C and [Y(DPDMG) 3 ] at 209 C, respectively.
The higher volatility for [Y(DPfAMD) 3 ] is accompanied by a low residual mass (3%) which indicates high thermal stability within the temperature range of evaporation. As expected, the amidinate [Y(DPAMD) 3 ] has a similar residual weight of 2%, while the guanidinate [Y(DPDMG) 3 ] exhibits a value of 8%. In terms of stability, a trend similar to the ndings in the EI-MS studies can be derived for the compound family: the thermal stability at elevated temperatures might be higher for

Y 2 O 3 ALD process development with [Y(DPfAMD) 3 ] and water
To evaluate the performance of [Y(DPfAMD) 3 ] in a thermal water-assisted ALD process for the formation of Y 2 O 3 thin lms, a thoughtful process development had to be carried out to nd optimal process parameters like ALD window and surface saturation characteristics. A saturation behavior of [Y(DPfAMD) 3 ] on 2 00 Si(100) substrates was reached aer 5 s of precursor pulse and 10 s precursor purge at a substrate temperature of T s ¼ 300 C with a GPC of 1.35Å (as shown in Fig. 3c and d). For these parameters, the lm thickness linearly increases with a slope of 0.136 nm per cycle according to a linear t (R 2 ¼ 0.9999) (Fig. 3b). Interestingly, a comparable saturation behavior is achieved for [Y(DPDMG) 3 ] which saturates aer 4 s of precursor exposure at T s ¼ 225 C and a GPC of 1.10Å but with a substantially higher precursor evaporation temperature of T p ¼ 130 C which is the same for [Y(DPAMD) 3 ]. The lower precursor evaporation temperature of T p ¼ 95 C for [Y(DPfAMD) 3 ] to reach saturation of the surface, underlines the higher volatility of this precursor in agreement with the TGA experiments. It is interesting to note that, [Y(DPfAMD) 3 ] can be  3 ] presumably started to decompose above 275 C, which led to higher GPCs and higher impurity levels. 14,15 The GPC remains nearly constant at 1.35Å from 225 C to 325 C, while below 225 C an increase in the GPC to 1.7Å and a decrease in lm homogeneity from 225 C to 200 C is apparent (Fig. 3a). Even though an increase in the GPC at lower temperatures normally can be attributed to condensation of the precursor, the saturation of the precursor at 200 C is also observed aer a 5 s precursor pulse, but a higher precursor purge time (30 s minimum) is needed for obtaining saturation without notable physisorption component (Fig. 3d). At a substrate temperature of 300 C, a precursor purge of 10 s is sufficient to remove excess precursor and by-products which might indicate faster surface kinetics at higher temperatures. Additionally, this might be explained by a shi from chemisorption to physisorption of the precursor on the surface below 225 C, which would explain the higher GPC and yet saturation behavior of the precursor. 31 The GPC steadily increases to 2.3Å below 200 C and below 150 C the GPC drops considerably and reaches a minimum value of 0.51Å at 100 C where the temperature is not effectual to reach sufficient adsorption of the precursor on the substrate surface. To obtain an initial insight into the properties of the Y 2 O 3 lms deposited within the ALD window, X-ray reectivity (XRR) curves were recorded and the critical angle Q c was extracted as a measure for the thin lm density. An increase of Q c with increasing temperature from Q c ¼ 0.45 at 100 C to Q c ¼ 0.59 at 225 C could be observed which remains constant at Q c ¼ 0.59 up to a deposition temperature of 325 C (Fig. 3a). This further supports the proposition of an ALD window with dense Y 2 O 3 lms reaching from 225 C to 325 C, whereas below this temperature range the measured critical XRR angle indicates a lower density. This, besides other factors, might be caused by a higher level of impurities and increased O/Y stoichiometry of the lms which indicates signicant hydroxyl (-OH) incorporation as shown later in the section "Compositional analysis". Overall, ALD with [Y(DPfAMD) 3 ] enables the usage of lower precursor evaporation temperatures and higher deposition temperatures when compared to processes with [Y(DPDMG) 3 ] and [Y(DPAMD) 3 ].

Thin lm characterization
To thoroughly identify the quality and properties of the deposited Y 2 O 3 thin lms and its properties using [Y(DPfAMD) 3 ] in terms of crystallinity, morphology and composition, complementary analyses were carried out and the results are subsequently discussed. If not stated otherwise, the thin lms were deposited at 300 C with the optimized parameters discussed before.
Thin lm crystallinity. To gather information on the crystallinity of the Y 2 O 3 thin lms deposited at 300 C, grazingincidence X-ray diffraction (GI-XRD) was carried out. The analysis of a 43 nm Y 2 O 3 lm on Si(100) revealed polycrystallinity with face-centred cubic packing (fcc) whereby the (222), (400), and (622) reections being strongly pronounced (Fig. 4).
Thin lm density and morphology. XRR measurements on a 43 nm Y 2 O 3 lm deposited at 300 C delivered typical Kiessig fringes and a critical angle at Q c ¼ 0.59 (Fig. 5, right), 32 which correlates to a Y 2 O 3 lm density of 4.85 g cm À3 . The derived density is close to the crystalline bulk density of Y 2 O 3 (5.03 g cm À3 ), 33 and the deviation might be caused by crystal grains from the pronounced polycrystalline nature of the thin lm, while additionally hydroxyl (-OH) incorporation as an intrinsic feature of the water-assisted ALD process could play a role here as shown later by XPS analysis. 34 Moreover, the obtained density of 4.70 g cm À3 for lms deposited at 200 C is still considerably higher than those obtained for the ALD process with [Y(DPDMG) 3 ] (4.24 g cm À3 ) at 200 C. From the slope of the XRR fringes a low roughness of r ¼ 0.64 nm is derived which is typical for ALD growth and close to the roughness of the uncoated Si(100) substrate with r ¼ 0.3 nm.
To conrm the low roughness and assess the overall morphology of Y 2 O 3 lms (thickness of 20 nm) grown at 300 C, AFM measurements were conducted (Fig. 5, le). A smooth  surface with an RMS roughness of r ¼ 0.63 nm (1 mm Â 1 mm) is observed. The Y 2 O 3 lms (20 nm) obtained by using [Y(DPDMG) 3 ] at T s ¼ 200 C feature a slightly lower but not signicantly deviating RMS roughness of r ¼ 0.55 nm, which might be explained by a lower degree of polycrystallinity as usually observed at lower T s in ALD. Interestingly, AFM measurements on 20 nm Y 2 O 3 lms grown at 200 C revealed a RMS roughness of r ¼ 0.60 nm that is only slightly higher than the comparable process, but lower than the value obtained at 300 C. Although the morphological and crystalline nature of the Y 2 O 3 conrm the characteristics typically seen for ALD processes, a thoughtful compositional analysis is still necessary and is discussed in the following section.

Compositional analysis
RBS and NRA. The purity in terms of thin lm stoichiometry and contamination levels on the atomic scale are accessible using RBS and NRA (Table S1 †). Accordingly, for Y 2 O 3 lms deposited at 300 C, compositional analysis via RBS revealed a O/Y ratio of 1.7, with 63 at% oxygen and 37 at% yttrium, while impurities such as nitrogen and carbon were near the detectable limits of NRA with <1 at%. The O/Y ratio of 1.7 is slightly higher than the expected ratio of 1.5 for Y 2 O 3 at this temperature, which might be explained by hydroxyl (-OH) incorporation as proposed previously for our water-assisted ALD process with 15 This was the case for lms deposited with [Y(DPAMD) 3 ] as reported by Rouffignac et al., where an O/Y stoichiometry of 2.0 for the as-deposited lms at 270 C and a stoichiometry of 1.7 was found aer in situ capping with Al 2 O 3 to prevent the adsorption of hydroxyls from the ambient atmosphere. 14 Although the O/Y ratio remains constant at 2.0 from 200 C to 250 C for our new process with [Y(DPfAMD) 3 ], it lowers to 1.6 at 325 C and rises to 2.3 at 150 C. This furthermore indicates that the deposition temperature plays a crucial role for the stoichiometry of the Y 2 O 3 lms and suggests, that a higher temperature might be more effective in removing the excess of adsorbed -OH species from the near surface of the lm during deposition. Depositing thicker layers (470 nm) of Y 2 O 3 at 300 C leads to an ideal value of 1.5. It must be considered that the longer exposure to higher temperatures during longer depositions due to the increased cycle amount might induce annealing effects and can consequently be responsible for the observed O/Y ratio. Furthermore, Y 2 O 3 is known to alter its surface upon exposure to the ambient with moisture and undergo additional hydroxylation which is more prevalent and causes an increased O/Y value for thinner lms as investigated by RBS where the whole depth of the sample is penetrated by the beam. 35 From these results it is apparent that a ratio closer to that of the bulk of Y 2 O 3 could be achieved using [Y(DPfAMD) 3 ] which was not the case for as-deposited lms using either [Y(DPDMG) 3 ] or [Y(DPAMD) 3 ] at lower temperatures. Moreover, the new process employing [Y(DPfAMD) 3 ] reduced the contamination levels with C and N below the detectable limits of NRA, which was not achieved to the same extent with [Y(DPDMG) 3 ] (C: 2-5 at% and N: 2-3 at% at T s ¼ 225 C).
XPS. The nature of chemical species within Y 2 O 3 thin lms deposited with optimized process parameters at 300 C for a representative 40 nm thick lm was investigated by XPS analysis. Hereby, the sample was exposed to the ambient as little as possible to prevent alteration of its surface through interactions with moisture and carbon dioxide which has been found to cause signicant surface hydroxylation as discussed earlier. 36,37 The survey spectra recorded for the as introduced and sputtered surfaces (ESI, Fig. S3 †) revealed the presence of all signals expected for yttrium and oxygen. While a weak signal originating from adventitious carbon was seen for the as introduced surface, nitrogen related signals were not found. The composition of the lm prior to sputter treatment and thereaer is given in Table 2, while Fig. 6 contains scans of the O 1s, Y 3d, C 1s and Y 3s core level regions.
Owing to the short exposure time to the ambient, the amount of adventitious carbon on the untreated surface was found to be rather low with around 4.8 at%. Oxygen was found to be predominant with 58.4 at% followed by yttrium with 36.8 at%. This resulted in a O/Y ratio of 1.59 which is well within the expected range, supporting the assumption that the water assisted ALD process leaves a partially hydroxylated surface aer the process and is moreover in accordance with the results obtained by NRA. Aer sputtering the carbon contamination level decreased to 2.5 at% while the O/Y ratio drops down to 1.18 as a consequence of preferential oxygen sputtering. 38 The oxygen core level regions presented in Fig. 6a for the as deposited and in Fig. 6e for the sputtered surface comprises of signals with two well distinguishable components of which the ones with a lower binding energy of 529.1 eV each are by far predominant and assigned to Y-O lattice oxygen. 15,39 For the as deposited surface, the second component is found at a higher binding energy of 531.3 eV and resembles Y-OH hydroxyls, 36,39 while this contribution is reduced and slightly shied to 531.5 eV aer sputtering. Thus, the presence of a minor amount of Y-OH species on the surface and in the bulk is conrmed and can expectedly be considered as an intrinsic feature of the ALD process. More precisely, the share of the Y-OH component for the as-deposited surface on the total integral of the O 1s core level amounted 24.5% and was reduced to 11.6% aer sputtering. In rst approximation, the low contribution of hydroxyls is concomitant with rather low hydrogen inclusion in the lm, Fig. 5 Left: AFM image of a 20 nm Y 2 O 3 thin film deposited at 300 C on Si(100). Right: XRR curve from the latter thin film with the simulated curve and obtained thin film parameters such as thickness, roughness, and density. but a precise determination of the hydrogen content is not possible by XPS. Interestingly, the share of the Y-OH component in the overall integral in the bulk is signicantly lower for this new water assisted ALD process with [Y(DPfAMD) 3 ] than for our prior process with [Y(DPDMG) 3 ]. We priorly proposed that it was the steric demand of the guanidinate backbone that was responsible of the high content of Y-OH in the lm as it could shield some hydroxyl groups during the process so that they could not react with incoming precursor molecules. 15 Moreover, for the water assisted ALD process with [Y(DPAMD) 3 ] it was reported that directly aer deposition at 275 C and capping with Al 2 O 3 to prevent further incorporation of -OH species from air, a signicant amount of hydroxyl species are remaining in the bulk of the lms. 14 The formamidinate skeleton on the contrary has been demonstrated to be less sterically demanding and more exible in the ALD of In 2 O 3 thin lms by the closely related precursors [In(DPfAMD) 3 ] and could thus facilitate inclusion of less Y-OH in the lm, which could be conrmed in the parent study. 22 Congruously, less inclusion of hydrogen in the Y 2 O 3 lms is enabled by utilization of [Y(DPfAMD) 3 ] compared to its congeners.
The Y 3d core level spectra of the as deposited (Fig. 6b) and sputtered (Fig. 6f) lm surface did not contrast the insights gained from the O 1s core region. In both cases the Y 3d region showed a signal feature with well resolved spin-orbital components. The signals themselves comprised two contributions described as doublets. In accordance with reported literature, 15,40 tting was performed with a xed energy separation of 2.05 eV between 3d 5/2 and 3d 3/2 components and intensity ratios between 3d 3/2 and 3d 5/2 of about 0.7. Binding energies for the Y 3d 5/2 component of Y-O are typically found in a range from 156.4-156.8 eV while the Y 3d 5/2 component of Y-OH species have been reported to be found between 157.3-158.0 eV. 28,41,42 In this study the Y 3d 5/2 component for Y-O is located at 156.4 eV and the one for Y-OH species at 157.3 eV for the as deposited surface. Expectedly, latter species only possesses a minor share in the overall integral and it is observed that this share is decreased even more aer sputtering while the positions remain roughly the same with 156.5 eV and 157.6 eV respectively. The C 1s core level spectra before and aer sputtering ( Fig. 6c and g) solely contain one species at 284.8 eV that originates from C-H type impurities. Contributions from carbonates were not seen and consequently not considered to be a factor in either the O 1s or Y 3d signal tting. Lastly, the overlapping N 1s/Y3d core level regions ( Fig. 6d and h) did not provide evidence for the presence of nitrogen in the lm, which  3 ] and water, while nearly stoichiometric lms can be obtained at a deposition temperature of 300 C without further post-treatment such as annealing or capping of the lms.

Functional properties
Optical characterization. As Y 2 O 3 nds its application in optoelectronic devices and is used as a high-k dielectric material in MOSFETs, an analysis and estimation of the allowed direct band gap energy (E g ) is of high interest. A 20 nm Y 2 O 3 lm was deposited at T s ¼ 300 C on fused silica substrates and was subjected to UV/Vis measurements (Fig. 7). The Y 2 O 3 thin lm presents a low absorption of <10% from wavelengths of 800 nm to 300 nm, aer which the absorption increases to <40% at 200 nm (Fig. 7a).
The Tauc plot represents an option to obtain the optical band gap energy for allowed direct band gaps from the (ahn) 2 term in dependence of the energy of the corresponding transmitting light beam (Fig. 7b). An extrapolation of the linear regime of the curve to the X-axis gives the direct optical band gap energy. For this thin Y 2 O 3 lm we derived a direct optical band gap energy of E g ¼ 5.56 eV, which is not only in accordance with band gaps derived from lms of processes with [Y(DPDMG) 3 ] (T s ¼ 225 C, 20 nm), but also with literature reported optical band gaps (E g ¼ 5.5-5.8 eV). 1 Electrical characterization. To demonstrate that the Y 2 O 3 thin lms grown at 300 C have the potential to be utilized in microelectronic devices, the C-V and I-V characteristics were examined in the form of a metal-insulator-semiconductor (MIS) structure with a Pt top electrode.
Typical C-V and G-V characteristics (f ¼ 1 MHz) of a Y 2 O 3 MIS device is depicted in Fig. 8. The permittivity is derived from the maximum capacitance in the accumulation regime at negative bias voltages. Considering a native 2 nm layer of SiO x on top of the p + -Si substrate, we estimate a permittivity of 13.9 with a standard deviation of s ¼ 0.  ¼ 14). 14,15,17 The hysteretic behavior in a C-V curve of a MIS device arises due to the presence of mobile charges and charge injection into the gate oxide from the semiconductor. Mobile charges would give rise to a hysteresis in the clockwise direction and the charge injection case would give rise to a counterclockwise hysteresis for a sweep from positive to negative gate voltage in a p-type substrate. Fig. 8 clearly shows the shi of the at band voltage towards more positive gate voltages during the reverse sweep from negative to positive gate bias exhibiting a clockwise hysteretic behavior for the Y 2 O 3 layers. Such a shi in the C-V curve could plausibly arise from the presence of negative charges accumulating at the interface. 43 In our case, the presence of -OH species accumulated at the interfacial layer could contribute to the hysteretic behavior and is consistent with the RBS and XPS results, which indicate a higher oxygen content compared to the ideal stoichiometry of Y 2 O 3 . The V FB derived for several devices from the 1/C 2 vs. V plot was found to be in the range of À2.93 V # V FB # À2.46 V at f ¼ 1 MHz and indicate the non-negligible amount of negative xed charges within the lm. It is known that such instabilities in at band voltage can be removed by subjecting the layers to forming gas treatment, which could be a possible route for further improvement of the electrical characteristics. The frequency dependent C-V characteristic of a typical capacitor device is depicted in Fig. 9, where the frequency was reduced starting from f ¼ 1 MHz to f ¼ 1 kHz. With decreasing measurement frequency, the capacitance slightly increases, and the at band voltage shis from V FB (f ¼ 1 The reduced capacitance at higher frequencies can be attributed to interface states, which cannot follow a high-frequency eld and only contribute to the overall capacitance at lower frequencies. Flatband voltage-shi and a reduced hysteresis for consecutive measurements are observed. Garvatin et al. ascribe such a behaviour in HfO 2 MIS devices to deep traps for electrons. 44 Due to discharging of the deep traps the at band voltage shis  to more positive values. In our case we can attribute the atband voltage shi to the trapping of mobile charge carriers from Si substrate resulting in a band bending of the dielectric layer. These charged trap states cannot be discharged in the subsequent C-V sweeps leading to a clearly reduced hysteresis and a more positive at band voltage. The interface trap density D it can be extracted using the C-V characteristic. A more sensitive approach is the conductance method proposed by Nicollian and Goetzberger. 45 The simpli-ed equivalent circuit in Fig. S5 † of a MIS capacitor consists of an oxide capacitance (C ox ), a substrate capacitance (C S ), an interface trap capacitance (C it ) and resistance (R it ). The parallel branch in Fig. S5a † can be converted into a frequency dependent capacitance C p in parallel with a frequency conductance G p (Fig. S5b †), where C P and G P are given by eqn (1) (in ESI, S6 †). C it denotes the interface trap capacitance C it ¼ q 2 D it , u ¼ 2pf the angular frequency and s it ¼ R it C it the interface-trap lifetime. The interface trap density D it is obtained by using the relation in terms of the maximum conductance G p,max in eqn (2) (in ESI, S6 †). Fig. 8 shows the measured conductance at 1 MHz for an Y 2 O 3 MIS device. From the maximum conductance the interface trap density the D it was extracted to be 1.25 Â10 11 cm À2 (eqn (3), ESI †). It should be noted that it is rare to nd values of the interface trap densities of ALD deposited Y 2 O 3 lms. For comparison we came across an interface trap density of 1.3 Â 10 12 cm À1 eV À1 determined by Lee et al. 17 It is one order magnitude higher compared to our value and exhibits superior quality of the interfaces in the present work. Finally, current density as a function of the electric eld (J-E) was measured for several devices to determine the leakage current and the dielectric breakdown (Fig. S4 †). All devices show a high breakdown eld between 4.2 and 6.1 MV cm À1 , which satises the requirements of CMOS gate oxide requirements. The values comply with breakdown elds determined for ALD grown Y 2 O 3 based on other precursors. 13,14 The leakage current density is around 10 À7 A cm À2 at 2 MV cm À1 (Fig. S4 †) and thus, in accordance with the lower leakage currents found in processes with [Y(DPDMG) 3 ].

Conclusions
In summary, we report a new and promising water-assisted Y 2 O 3 ALD process employing the precursor [Y(DPfAMD) 3 ] that features distinct advantages over its analogous [Y(DPAMD) 3 ] and [Y(DPDMG) 3 ] complexes. A low 1 torr vapor pressure temperature of 156 C combined with a high thermal stability compared to the other precursors render [Y(DPfAMD) 3 ] exceptionally useful for water assisted ALD processes for the formation of Y 2 O 3 thin lms applying low precursor evaporation temperatures. The new process gives rise to a broad ALD window ranging from 225 C to 300 C, while the obtained lms are polycrystalline, smooth and of very good compositional quality at a substrate temperature of 300 C. The origin of the enhanced ALD window and improved compositional quality especially in terms of thin lm stoichiometry of the new process compared to the already reported processes with [Y(DPDMG) 3 ] and [Y(DPAMD) 3 ] should be further studied with in situ diagnostics and theoretical studies, while the precursor [Y(DPfAMD) 3 ] with the corresponding new process on its own broadens the applicability of Y 2 O 3 ALD. The high permittivity of Y 2 O 3 lms deposited by [Y(DPfAMD) 3 ] obtained from an application in MIS capacitors is one of the highest reported in literature for ALD grown Y 2 O 3 lms and the low interface trap density of 1.25 Â 10 11 cm À2 and low leakage current around 10 À7 A cm À2 at 2 MV cm À1 underlines the high interface quality of the lms obtained from the new ALD process developed in this study. These experiments set an exciting starting point for further in-depth comparative studies on how the ligand sphere of the employed precursor can drastically alter its physicochemical properties which in turn inuences the ALD process parameters and thus the applicability of the process. This will further enhance the understanding and relation between precursor chemistry and ALD process performance which can be applied to other material systems.