Effects of processing on the stability of molybdenum oxide ultra-thin films

The effects of wet chemical processing conventionally employed in device fabrication standards are systematically studied on molybdenum oxide (MoOx) ultra-thin films. We have combined x-ray photoelectron spectroscopy (XPS), angle resolved XPS and x-ray reflectivity techniques to provide deep insights into the changes in composition, structure and electronic states upon treatment of films with different initial stoichiometry prepared by reactive sputtering. Our results show significant reduction effects associated with the development of gap states in MoOx, as well as changes in the composition, density and structure of the films, systematically correlated with the initial oxidation state of Mo.


INTRODUCTION
Molybdenum oxide (MoO x ) is a transition metal oxide showing extraordinary and versatile electrical, structural, chemical and optical properties, which depend on the oxidation state of Mo, on the degree of crystallinity, on the sample morphology and on environmental conditions.This material system, particularly in the form of thin and ultra-thin films, finds applications in a variety of technologically relevant fields, including catalysis [1,2], gas sensors [3,4], optically switchable coatings [5,6], building-blocks for high-energy density solid-state microbatteries [7,8], smart windows technology [9,10], flexible supercapacitors [11], thin film transistors (TFTs) [12] and organic electronics [13][14][15][16][17][18][19][20][21][22][23].Owing to its high work function -reported to reach 6.9 eV [13] -and to the layered structure of α-MoO 3 , MoO x is currently also employed as a 2D material beyond graphene and as efficient hole contact on 2D transition metal dichalcogenides for p-type field effect transistors (p-FETs) [24][25][26].In view of a reliable device performance, the control over the chemical and physical properties of the MoO x system is mandatory.It has been recently reported [15,[27][28][29][30][31][32][33] that in MoO x with x < 3, oxygen vacancies originated from partially populated d -states, give rise to occupied energy states within the forbidden gap -reported to be ∼ 3.0 eV at room temperature [13] -becoming bands above a critical concentration and driving the Fermi level close to the conduction band.The oxygen vacancies concentration, and consequently the averaged oxidation state of Mo, is a key parameter directly affecting the properties of the MoO x system.When fully oxidized, i.e. for MoO 3 with the corresponding formal oxidation state 6+, MoO x is a standard closed d 0 oxide, transparent and devoided of oxygen vacancies.In this case, there are no occupied states within the wide band gap and the material follows an insulating behaviour.By reducing the metal cations, that is, by increasing the amount of oxygen vacancies, MoO 3 forms a series of stable and metastable suboxides (MoO x , 2 < x < 3), reported e.g. by Magnéli [34] and Kihlborg [35], showing semiconducting behaviour and a gradually increasing opacity.These sub-stoichiometric oxides present a complex x-ray photoelectron spectroscopy (XPS) spectrum, resulting from the convolution of three different Mo valence states, namely +6, +5 and +4.Moreover, MoO 2 , with formal oxidation state +4 and with the 4d 2 electron configuration, shows metallic conductivity and the highest opacity of the whole series.Lately, the effects of air exposure on MoO x have attracted considerable attention [36][37][38].For instance, recent developments in the field of organic solar cells have shown the critical impact of air exposure on the electronic structure of the MoO x surface and, consequelty, on the device efficiency and lifetime [13,[39][40][41][42][43][44][45][46].Surface hydration or hydroxylation are suggested as possible mechanisms for reduction of cations in the MoO x system and, thus, for electronic structure changes.In particular, the hydroxylation mechanism is supported by recent density functional theory (DFT) calculations [47].Moreover, several reports from the field of organic electronics [13,39,48,49] point at gap states formed not only by oxygen vacancies and cation reduction, but also by organic adsorbates at the interface between MoO x and the organic layers.
Process engineering for devices fabrication generally involves standardized steps in which thin layers are treated with various photo-resists and solvents, but systematic studies on the effects of processing on the properties of MoO x films are still wanted.In this work, we report on the effects of processing on the stability of ultra-thin MoO x layers treated with conventional photo-resist (Shipley S1818) and solvents.The in-depth study is carried out on a series of ∼ 10 nm thick amorphous MoO x layers fabricated by means of reactive sputtering and presenting different stoichiometry, spanning -over the series -from a metallic to a fully oxidized phase and including, in particular, Mo, MoO 2+ǫ , MoO 3−ǫ , and MoO 3 .

EXPERIMENTAL PROCEDURE
A combination of XPS, angle resolved XPS (ARXPS) and x-ray reflectivity (XRR) techniques is employed to determine the charge state of the cations, the valence band spectra and the structural arrangement of the samples as-grown and upon treatment/exposure.A list of the as-grown samples under study is provided in Table II of the Appendix, while the labeling followed to identify the samples series according to the exposure/processing, is given in Table I.
The as-grown virgin samples, labeled as A v , are stored under cleanroom conditions until they are stabilized after chemisorption of oxygen and moisture on the surface.Once stabilized, that is, once the samples show stable XPS spectra measured a week apart, these reference samples A ref are coated with photo-resist which is then removed following the different standard procedures considered here below and summarized in Table I.When treated according to a conventional cleaning protocol of 15 minutes in acetone ultrasonic bath followed by 15 minutes in 2-propanol, and ending with 1 minute rinse in DI water, the previ-  Å in a Seifert XRD 3003 PTS-HR is employed and the GenX reflectivity fitting package [50] supports the data analysis.Details on growth parameters and XRR measurements are provided in the Appendix.

RESULTS AND DISCUSSION
In Fig. 1 the high resolution Mo 3d core level XPS spectra for the samples: (i) A v asgrown and (ii) A ref upon stabilization in cleanroom conditions are reported.For each valence state, the core level is split into the 3d 5/2 and 3d 3/2 doublets, separated by 3.1 eV, and with binding energies assigned according to established reference values [51].A significant surface oxidation is observed for all the samples exposed to atmosphere.As expected, the oxidation effect is more pronounced in virgin samples with a more metallic character.For all the initial stoichiometries, the oxidation process is not confined to the uppermost atomic layers, but interests the whole probing depth.
The reduction effect reported previously [13,40] and due to air exposure is not found in this work, likely due the fact that the samples considered in literature were not completely stabilized, while we have ensured and confirmed by XPS analysis, that the A ref films studied here are fully stabile.In fact, it was theoretically demonstrated by Butler et al. [47], that essential properties of the MoO x system, like e.g. the ionization potential, critically depend on the exposure to moisture until stabilization.
By considering the samples B 0 , i.e. the A ref upon conventional treatment involving DI water, it is inferred from Fig. 2(a) that in the sub-stoichiometric samples MoO 2+ǫ and MoO 3−ǫ , the Mo 3d emission does not diminish significantly, but shifts to lower BE, pointing to a significant chemical reduction enhanced for the bulk angle, that is, for deep atomic layers.
After four days of atmospheric exposure, i.e.B 0 →B 1 , further oxidation leads to a lowering of the intensity of the Mo 4+ emission, while the one of Mo 6+ is augmented, recovering a spectrum resembling the one obtained from the virgin samples A v before stabilization.
Upon processing without DI water, i.Its intensity is correlated with the presence of Mo 6+ and Mo 5+ oxidation states in the Mo 3d core level spectra in Fig. 2(a).As evidenced in Figs.3(b-d), in the case of the oxide films, the valence band region shows the conventional transition metal oxide two band structure [31] resulting in: (i) one peak mainly due to O 2p orbitals centered at ∼6 eV, and which identifies the valence band maximum (VBM) at ∼3 eV, (ii) and a second peak emerging between the VBM and the E F .In the case of the intermediate oxides, this latter band of gap states originates from oxygen vacancies partially filling the empty Mo d levels.
For the as-grown samples before processing, there is a systematic correlation between the Mo oxidation state as in Fig. 2 and the relative intensity of the emission related to the gap states and of the one from the O 2p levels, as evidenced in Figs.3(a-d).The width of the emission from the gap states is assigned to the presence of different types of vacancies [53], whose complex geometry is beyond the resolution of the XPS system.
Processing involving DI water promotes substantially the development of gap states.
This effect is paticularly pronounced for the less oxidized samples MoO 2+ǫ , as summarized in Fig. 3 In all cases, the correlation between the reduction process emerging from the data shown in Fig. 2 and the evolution of the gap states is preserved, hence we attribute the enhancement of the gap state density to the presence of MoO x reduced states, in accordance with previous reports by other authors [13,40,53,54].Moreover, we can state that not only surface states are affected, but also deeper layers within the material, as deduced from the ARXPS measurements in Fig. 2.
From XRR measurements -whose details are provided in the Appendix -the variation of density and thickness due to exposure/processing has been obtained for the investigated MoO x layers and it is summarized in the panels of Fig. 4. Generally, an increment of density is reflected in a lowering of the thickness and vice versa, as inferred from Figs. 4(a,c,d).
In the case of the MoO 2+ǫ layers of Fig. 4  oxidation state for all the samples, and causes the complete dissolution of MoO 3 after 1 min. of exposure.The high etching selectivity between Mo and MoO 3 opens wide perspectives for structuring of MoO x by means of simple DI water chemical etching [55,56].The reduction effect, not confined only to the surface layers, is correlated with the development of gap states -likely due to chemisorbed species -which lead to an electron transfer to the transition metal oxide system [13,40].Independently of the processing protocol, reduction and gap state intensities are stable after four days of air exposure.We have found that fully metallic and fully oxidized phases have the highest stability against the various cleaning protocols.Our findings indicate that the optimization of the protocols for processing MoO x ultra-thin films are crucial for the engineering of band states, fundamental for e.g.charge transport applications [48,57].

Av 0 1 1
Virgin samples.Shortly exposed to air after growth.A ref Reference samples.Stored under cleanroom condition until stabilization and used for processing.B Samples measured 1h after exposure to conventional cleaning protocol by ultrasonic treatment in acetone, 2-propanol and DI water sequentially.B Samples measured 4 days after esposure to conventional cleaning protocol by ultrasonic tratment in acetone, 2-propanol and DI water sequentially.C 0 Samples measured 1h after exposure to conventional cleaning protocol by ultrasonic treatment in acetone, 2-propanol sequentially.C Samples measured 4 days after exposure to conventional cleaning protocol by ultrasonic treatment in acetone, 2-propanol sequentially.

Figure 1 .
Figure 1.Comparison between high resolution Mo 3d XPS spectra for the samples as-grown A v and upon stabilization in cleanroom atmosphere A ref .

Figure 2 .
Figure 2. High resolution Mo 3d ARXPS intensity maps as a function of depth and overall XPS intensity for (a) Mo, MoO 2+ǫ and MoO 3−ǫ , and (b) MoO 3 before and after processing.
Fig.3(a), and characterized by a broad overall band with a maximum intensity at 2.0 eV below the Fermi level[52] E F .The broad emission centered at ∼ 6.0 eV is assigned to an O 2p photoemission signal of adsorbed oxygen, with contributions from various oxygen species.
(b).When MoO 2+ǫ is treated in the absence of DI water (A ref →C 0 ), the intensity of the peak related to the gap states assumes a value intermediate between the one for A v and the one for A ref , and it keeps stable (C 0 →C 1 ).As shown in Fig. 3(d), for the fully oxidized MoO 3 , the intensity of the initial band states is minimized upon stabilization (A v →A v ) and does not significantly change after processing without DI water (A ref →C 0 , C 1 ).
(b) after processing in the absence of DI water (A ref →C 0 ), the density remains close to the one of the reference sample, while the thickness is reduced from 17.5 nm to 11 nm.Further analysis is required, in order to understand this anomaly.The MoO 3 films of Fig. 4(d), while being the most stable -as confirmed by XPS measurements -upon stabilization in cleanroom atmosphere and upon treatment with processes non including DI water, are completely dissociated after 1 min. of exposure to DI water.An intermediate case, is represented by the sub-stoichiometric samples MoO 3−ǫ in Fig. 4(c), which -upon processing with DI water (A ref →B 0 ) -looses significantly in thickness, pointing at an enhanced solubility of the Mo 6+ phase in water.In summary, reactively sputtered MoO x ultra-thin film layers with different oxidation states (Mo, MoO 2+ǫ , MoO 3−ǫ , and MoO 3 ) have been exposed to different conventional cleaning protocols and investigated by XPS, ARXPS and XRR.Water strongly reduces the Mo

Figure 4 .
Figure 4. Film density and thickness as deduced from XRR measurements for: (a) Mo, (b) MoO 2+ǫ , (c) MoO 3−ǫ , and (d) MoO 3 as-grown, stabilized in cleanroom atmosphere and upon processing.The solid bands are guide for the eyes and mark the values for the as-grown (A v ) and stabilized (A ref ) samples.

Figure 5 .
Figure 5. Experimental data and fit of the XRR as a function of the grazing incident angle before and after processing for all the samples under study, e.i., Mo, MoO 2+x , MoO 3−x , and MoO 3 .

Table I .
Labeling of the samples series according to the exposure/processing.
ously stabilized A ref samples are referred to as samples B. When subjected to the previous procedure, but avoiding the final rinse in DI water, the samples are labelled as series C. Subscripts 0 and 1 (B 0 , B 1 , C 0 , and C 1 ) correspond to samples measured one hour and four days after processing, respectively.The series of amorphous MoO x ultra-thin films are reactively sputtered in DC mode on glass substrates (Corning Eagle XG) at room temperature.The high resolution XPS and the conventional C 1s peak at 284.8 eV.For the XRR measurements, a wavelength of 1.54