High-resolution neutron imaging: a new approach to characterize water in anodic aluminum oxides

During the growth of anodic Al oxide layers water incorporates in the film and therefore influences the intrinsic properties of the oxide formed. In this study, we propose a new approach, based on the use of high-resolution neutron imaging, to visualize and quantify the water content in porous Al oxides as a function of anodizing conditions. Water in these porous films is either incorporated directly in the oxide structure (structural) and/or fills the pores (morphological). This preliminary study demonstrates that the differences in water content of porous anodic Al oxide layers are strongly related to the oxide growth parameters but interestingly cannot be directly correlated to a specific change in the amorphous oxide structure or in the pore morphology. Due to the high sensitivity of high-resolution neutron imaging to small changes in the water content, we furthermore show that the morphological water content in Al oxides formed in sulfuric acid as well as in phosphoric acid is partially reversible upon heat treatment and immersion. High-resolution neutron imaging is also found to be highly sensitive to structural disorder and crystallographic orientations, allowing to identify different crystalline Al oxide samples based on their structural and morphological defect content. This offers new perspectives to study the effect of the hydrogen and/or water incorporation as well as oxide-related structural modifications on Al oxihydroxides in relation to growth parameters, stability, functionalization as well as properties tuning, highly relevant to surface protection and their use as templates.


Introduction
Porous anodizing is a process widely used in industry as a surface finish for aluminum substrates to provide corrosion protection, wear resistance and decorative appearance [1][2][3] . It has also recently attracted great interest as a simple and inexpensive method to produce nanostructured materials [4][5][6] . By varying the anodizing conditions, the structure of the formed porous anodic Al oxide can be tuned and subsequently functionalized, making it highly attractive for various industrial applications from nanoscale electronics and optoelectronics to catalysts and sensors 4,5,[7][8][9][10][11][12][13][14] .
The presence of water related species, such as proton or hydroxyl groups integrated in the oxide structure or water of hydration, influences the growth and properties of porous anodic Al oxide layers [47][48][49] . It is therefore essential to include a proper quantification of these species when discussing the nature and structure of the formed anodic Al oxides. So far, the water content in anodic Al oxide layers has been estimated using AC bridge measurement 47,48,50 , infrared spectroscopy (IR) 42 , secondary ion mass spectroscopy (SIMS) [51][52][53] , nuclear magnetic resonance (NMR) 54 , N 15 nuclear reaction analysis (NRA) 55,56 , elastic recoil detection (ERDA) 56 , reflectivity techniques 57 and quartz crystal microbalance 58 . Although these techniques provide specific information about the hydrogen distribution as well as its chemical state, most of them do not allow the determination of water distribution profile in a micrometer range field of view and therefore are not suitable to investigate thick oxide layers.
The present study explores the applicability of high-resolution neutron imaging to characterize porous anodic Al oxides. Neutron imaging is a non-destructive technique, that is -due to the high sensitivity of neutrons to hydrogen -very appropriate for assessing the distribution of hydrogen-containing compounds, such as water, in structural materials that are almost transparent to neutrons (e.g. those based on aluminum). Thus, neutron imaging has been successfully used in a variety of research areas to quantify the water content in natural and engineered porous materials, such as rocks, soils, concrete and fuel cells [59][60][61][62][63][64][65][66] . To our best knowledge, it has not yet been used to study anodic oxides due to its limited spatial resolution of several tenths of micrometers. To overcome this limitation, a high-resolution imaging setup, the neutron microscope detector, has been developed at Paul Scherrer Institute (PSI) in Switzerland 67 . This state-of-the-art setup provides an unprecedented spatial resolution below 5 µm at an effective pixel size of 1.3 µm [67][68][69][70] .
The purpose of this study is therefore to address the prospects of high-resolution neutron imaging to characterize porous anodic Al oxides. With this in mind, we have investigated the effect of substrate purity, anodizing electrolyte and anodizing time on the hydrogen integration and water content of the formed porous anodic Al oxides.
The findings provided by neutron imaging were correlated to the structural and morphological evolution of the oxide layers. As few fundamental neutron studies exist, the results about anodic oxides were further compared to the one obtained for other typical Al oxide samples, such as crystalline C-sapphire, sintered Al oxide, plasma sprayed Al oxide and an Al hydroxide, Al(OH)3, representing various other structures and defect/water distribution.

Samples
This study focuses on porous Al oxide layers obtained by galvanostatic anodizing of pure Al substrates. Two Al purity grades, pure Al (Al 99.9%, 0.5 mm thick, Novelis, Switzerland) and AW1050 (Al 99.5%, 1.1 mm thick, Novelis, Switzerland), were investigated. To meet the requirements of the neutron imaging setup, the Al substrates were cut as following: about 2.00 mm in depth (plate width placed parallel to the beam) -an actual depth of 2.05 ± 0.03 mm was measured -and 25 mm in height. The substrates were used in the as-received condition. Prior to anodizing, the samples were first ultrasonically cleaned for 5 min in acetone, then for 5 min in ethanol and finally dried with Ar.
Double-sided galvanostatic anodizing was carried out at a current density of 35 mA cm -2 and at room temperature in either 0.1 M phosphoric acid or 0.5 M sulfuric acid, both unstirred. A two electrode electrochemical cell was used, the sample acting as the working electrode and a Pt ring electrode as counter electrode. A Keithley 2400 SourceMeter SMU instrument (Tetronix) was used to perform the anodizing process.
The anodizing times for pure Al were, respectively, 2320 s, 6900 s and 9300 s in  Table 1 summarizes the anodizing conditions for the different samples and the effective thicknesses obtained.
Additionally, a 50 µm thick Al hydroxide layer was prepared by evaporating a saturated Al(OH)3 (gibbsite) solution on a pure Al substrate. The samples were cut to present a sample depth of 2.00 ± 0.05 mm in the beam direction.

Microstructure and structure characterization
The thickness and morphology of the different oxide layers were studied using a FEI Nova NanoSEM field emission gun scanning electron microscope (FEG-SEM). Crosssections were prepared using a Hitachi IM4000 Ar ion milling system. All samples were coated with a 5 nm conductive gold layer prior to SEM investigation.
Structural characterization of the anodic Al oxides and of the reference samples was performed via X-ray diffraction (XRD) in a Bruker D8 diffractometer in Bragg Brentano geometry using Cu Kα radiation and a Ni filter.

Neutron microscope setup and operation conditions
High-resolution thermal neutron imaging was performed using the neutron microscope at the Pulse OverLap Diffractometer (POLDI) beamline at the Swiss spallation neutron source (SINQ), PSI. Spectral information are reported in 71 . The samples were measured during two different neutron beamtimes. Anodized Al layers, C-sapphire and Al hydroxide samples were measured using the detector being equipped with a sCMOS camera with the resulting image pixel size of 2.6 µm. The sintered Al oxide, plasma sprayed Al oxide and commercial Al oxide membrane were measured using the detector being equipped with a CCD camera with the resulting image pixel size of 2.7 µm (for details, see supplementary information). C-sapphire has then been measured in both detector configurations for comparison purposes. The state-of-the-art neutron microscope setup was equipped with a 3.5 µm thick 157gadolinium oxysulfide scintillator and in both cases, the samples were placed at about 2.5 mm from the scintillator. The setup therefore provides an achievable spatial resolution of at least 9.6 µm 67,70,72,73 . Figure S1 shows the setup used for these experiments. After having been corrected for scattering and bias, the obtained image was used for quantitative analyses ( Figure S2). The detailed data processing is described in the supplementary information.

Quantifying the water content
Since the attenuation of radiation passing through matter follows the Beer-Lambert law, the transmission contrast can be expressed as: where is the linear attenuation coefficient of the material [cm -1 ] through , the sample depth in the beam direction equal to 2.05 ± 0.03 mm for all studied samples.
Regardless of its structural composition, it is generally assumed that porous anodic Al oxide mainly consists of hydrated Al oxide, referred as Al2O3 · n H2O. This water comprises the contribution of both the water/hydrogen incorporated in the oxide structure and the hydration water that remains in the pores. The contribution of the electrolyte anions adsorbed at the anodic Al oxide/electrolyte interface to the transmission contrast is neglected because both sulfur and phosphorus possess rather low neutron cross sections. The transmission contrast can consequently be expressed as follow: is the linear attenuation coefficient of water-free porous anodic Al oxide [cm -1 ], 2 is the free water linear attenuation coefficient [cm -1 ], is the sample depth in the beam direction equal to 2.05 ± 0.03 mm for all the studied samples and 2 is the equivalent water thickness [cm], which includes the contribution of the water/hydrogen incorporated in the oxide structure (structural water) and/or the hydration water that remains in the pores (morphological water). Note that Eq. 2 is an approximation of the real system, since the contribution of structural and morphological water cannot yet be directly distinguished with neutron imaging.
The overall water content of the anodic Al oxide layers, described by the equivalent water thickness 2 , can be deduced from Eq. 2: provided that the neutron cross-sections are known. Those can be calculated from the atomic ones but, in the case of a multi-element compound, the empirical and calculated values differ due to non-negligible additional scattering effects (Table S1).
We therefore need to rely on experimental referencing to determine the linear attenuation coefficient.
The free water linear attenuation coefficient at the POLDI beamline, 2 , has been previously measured and is equal to 3.78 cm -1 . Since finding a suitable reference sample for the porous anodic Al oxides proved to be challenging, the following approaches were taken to determine the cross-section, i.e. linear attenuation coefficient of "dry" porous anodic Al2O3. Anodized samples were placed in ultra-high vacuum (UHV -~ 10 -8 mbar) for 12 h in an attempt to desorb water from the oxide layers. On the other hand, selected anodized samples were heated for 1 h (H1), 3 h (H3) and 5 h (H5) at 150 °C on a hot plate. The temperature was chosen to induce as little change to the oxide as possible. The "dried" porous oxide was then immersed for 30 min in water to determine if the pores could be refilled with water. After each step, neutron imaging was performed in order to monitor the water content evolution in the porous anodic films. The highest contrast transmission value obtained, corresponding to "dry" Al2O3, was used as reference value to derive the linear attenuation coefficient of water free porous anodic Al2O3. The water content determined is therefore relative to this "dry" Al2O3 value.

Microstructure and structure of porous anodic Al oxide layers
In this study, one-step galvanostatic anodizing at a current density of 35 mA cm -2 was  Table 1.    Table S2).  Figure S5. This is also observed for the porous anodic Al oxide layers grown on AW1050 but interestingly, no self-ordering occurs for this substrate (see Figure 1c and S6). Different pore ordering does not seem to influence the pore size, which is found to be around 220-230 nm on both substrates (see Table S2). It is however worth mentioning that identical current density and similar potential evolution lead to the formation of an 85 µm-thick oxide layer on AW1050 compared to 130 µm on pure Al, which suggests a lower efficiency of the oxide layer growth on AW1050 substrates.
From the XRD analysis presented in Figure 1d, it can be deduced that all the porous Al anodic oxides formed are in an amorphous state. This statement is supported by the absence of sharp diffraction peaks and by the presence of broad reflections between 10 ° and 30 °. However, small crystallites (nanocrystals) inside the oxide could also result in such diffractograms, thus their presence cannot be completely excluded.
Interestingly, as shown in Figure  respectively (see Figure S8). This is a clear indication that Al anodized layers cannot be shows that the porous anodic Al oxide layers formed exhibit morphological and/or intrinsic structural differences depending on the anodizing conditions.

Neutron imaging -transmission contrast profiles
High-resolution neutron imaging was performed on the different samples to first evaluate their water content (based on hydrogen content detection) as a function of anodizing conditions. Using the contrast differences observed (see Figure 2), the thickness of the porous anodic Al oxide layers can be directly derived from the obtained transmission contrast images. The values found are in good agreement with the values retrieved from SEM cross section images, as reported in Table 1. The remaining thickness of the substrate is also easily accessible.
To allow a more detailed comparison between the different anodized samples, transmission profiles are extracted from the neutron radiographs, as indicated in  Consequently, the neutron data discussed in this paper exclude those regions.
Moreover, due to the extent of the edge enhancement effects, the minimum thickness (oxide and substrate) that can be investigated in this study is therefore about 50 µm, as shown in Figure 2d. The Al hydroxide sample has been prepared by evaporating a saturated Al(OH)3 solution on a pure Al substrate, which allows to obtain such a thick hydroxide layer (left side in Figure 2d) but the whole sample is covered by a thin inhomogeneous hydroxide layer, which accounts for the dark contrast at the air/sample interfaces. This particular sample preparation results in rather rough interfaces, which contribute to broaden the interface region.

Quantification of water in porous Al oxide layers
The contrast transmission profiles provide valuable qualitative information about the morphological and structural water content and distribution in porous anodic Al oxide layers. In a next step, Eq. 3 was used to quantify this water content. Since producing a water-free anodic Al oxide that can act as reference to determine the linear attenuation coefficient of "dry" porous anodic Al2O3 proved to be challenging, one pure Al P130 sample was stored for 12 h in ultra-high vacuum (pure Al P130 UHV) while another was heat-treated, in an attempt to desorb the water initially present in the pores and subsequently, obtain a water-free porous Al2O3. In this preliminary study, pure Al P130 was selected as a quasi-reference sample, due to its initial low morphological water content and similar oxide structure compared to pure Al P50 and 1050 P85 (Figure 3).
We are nevertheless aware that it remains a quasi-reference sample, since it presents structural differences with pure Al P230 and morphological differences with pure Al S60, which can lead to over-and underestimation of the water content in these two samples, respectively.
The heat treatment was performed at 150 °C on a hot plate for 1 h (pure Al P130 H1), 3 h (pure Al P130 H3) and 5 h (pure Al P130 H5). It was then immersed in water at room temperature for 30 min (pure Al P130 W) to determine if the pores could be refilled with water. Neutron imaging was performed after each step to monitor the evolution of the morphological water content in the porous anodic films. The contrast transmission profiles of these samples are displayed in Figure 4a. Since no changes were observed in the oxide structure ( Figure S8) or in pore morphology after heating, this increase can be attributed to morphological water desorption occurring in the pores during the thermal treatment. The highest contrast transmission value was obtained after 5 h of heating, with the transmission value reaching 0.964 compared to 0.959 for pure Al P130 near the air/oxide interface, as indicated by the dashed line in Figure 4a. This evolution is observed on both replicates, pure Al P130 R1 and R2. This value of 0.964 was assimilated to the transmission value of "dry" porous anodic and used to derive the linear attenuation coefficient, . It was found equal to 0.176 cm -1 , value that was used in Eq. 3 to quantify the water content of the porous anodic Al oxide layers. The deduced water content, described as the equivalent water thickness, 2 , is reported in Figure 4b. It is necessary to mention that the quantified water content is relative to this specific "dry" Al2O3 value. It would be interesting to determine if all the morphological water can be removed by heating the sample without changing the oxide structure or pore morphology of the porous anodic Al oxide but this was not possible during the neutron beamtime allocated and therefore remains outside the scope of this paper.
Heating pure Al P130 for 1 h or 3 h leads to an increase of the transmission value from 0.959 to 0.961 that is equal to the desorption of 6 µm of water at top part of the oxide layer whereas after 5 h of heating, a water thickness of about 13 µm can be removed, which corresponds to about 0.73 % of the total oxide sample depth of 2.05 ± 0.03 mm.
Interestingly, water desorption only occurs in the top first 70 µm of the oxide layer. No changes is detected in the profile near the oxide/Al substrate interface, suggesting water entrapment in the pores at the bottom part of the oxide, probably due to pore branching and subsequent blockage. This trend is observed for all heated samples, as shown in Figure 4. When pure Al P130 H5 is then immersed in water at room temperature for 30 min, the equivalent of 5.5 µm of water is adsorbed in the first 70 µm of the oxide layer, which shows that the pores can reversibly be filled with water. As before, the profile at the oxide/Al substrate interface remains unaffected, indicating that water cannot reach the pores at the bottom part of the oxide. Figure 4 clearly demonstrates the potential of high-resolution neutron imaging to study the evolution of the water content and its distribution in porous anodic Al oxides.
Small changes in the equivalent water thickness, up to a couple of µm, which can drastically influence the oxide properties and stability, can be easily detected with this technique while being highly challenging to observe using techniques involving photon or electron beams. The equivalent water thickness is expressed for the total sample thickness of 2.05 mm.
Note that the water content is relative to the "dry" porous anodic Al2O3 value. Figure 5 displays the overall water content of the porous anodic Al oxide layers as a function of anodizing conditions. Note that the equivalent water thickness determined is relative to the "dry" porous anodic Al2O3 value. At the air/oxide interface, pure Al P50 contains twice as much water as pure Al P130, as reported in Figure 5a. Deeper in the oxide, the equivalent water thickness shows an increase of 25 µm followed by a loss of 10 µm towards the oxide/Al substrate interface. This increase occurs in the first 25 µm of the oxide layer, which corresponds to the thickness of the top ordered layer ( Figure   S5). This observation suggests a higher water intake in ordered pores compared to disordered, branched pores. This surface component is hidden in pure Al P130 due to edge enhancement effects while not distinguishable in pure Al P230 profile probably due to imperfect parallel positioning of the sample compared to the beam. Regarding pure Al P230, the structural change in the oxide associated to the presence of boehmite-like structure instead of gibbsite-like structure results in a water content ten times higher than pure Al P130. The corresponding water amount represents 7 to 11 % of the total sample thickness. Due to the structural differences between pure Al P130, used as quasi-reference sample, and pure Al P230, the water content of this sample may be overestimated but remains anyway higher than in pure Al P130. Similarly, the water content of porous anodic Al oxide formed in sulfuric acid may be underestimated because of their morphological differences. Therefore, at equivalent oxide thickness, Figure 5b indicates that porous anodic Al oxides formed in sulfuric acid contain at least two and half times more water than the ones formed in phosphoric acid. Pure Al S60 was then heated for 3 h at 150 °C, resulting in desorption of morphological water equivalent to 8 µm within the top first 30 µm of the oxide layer, which is comparable to the amount removed from pure Al P130 H3. Moreover, once again, no changes is observed in the profile near the oxide/Al substrate interface, implying that the water remains trapped in the pores at the bottom part of the oxide. This suggests that despite the long-range pore order observed in porous anodic Al oxides formed in sulfuric acid, water desorption from the pores is also only possible in the top part of the oxide layer as will probably be reversible filling of the pores. When using AW 1050 as a substrate instead of pure Al, the water content of the respective porous anodic Al oxides is comparable near the air/oxide interface, as shown in Figure 5c, whereas close to the oxide/Al substrate interface, the oxide formed on AW1050 contains two times more water than pure Al P130.   The aluminum oxide grown by plasma spraying consists of metastable, cubic γ-Al2O3.

Neutron imaging of Al oxides and hydroxide reference samples
Its transmission value of 0.948 is slightly lower compared to the stable α-Al2O3 phase.
This lower transmission value can be attributed to the even more defective nature of this oxide layer, which contains larger structural defects, such as cracks and pores, defects inherent to the plasma spray process and that can integrate hydrogen contamination and/or water in them. These results show, on the other hand, that the changes in transmission value allow to differentiate crystalline Al2O3 based on their crystalline features. Interestingly, the two polycrystalline (with different structures) Al oxides have similar contrast transmission values compared to the porous anodic Al oxides investigated. The Al hydroxide reference sample has a transmission value of 0.816, which is, as expected, lower than the polycrystalline Al oxides. It is worth mentioning that the sample preparation preserved the gibbsite crystalline structure.
Since the porous anodic oxide layers investigated are amorphous, a commercial membrane, Anodisc 13, was included as a comparison. The membrane presents a high degree of pore order, suggesting a two-step anodizing formation process ( Figure S7). The investigations on Al oxides/hydroxide reference samples clearly show that the neutron imaging data cannot be solely interpreted based on theoretical cross-section values of pure materials and that both structural and morphological parameters play an important role. This further evidences that neutron imaging is very useful to identify small differences in oxide structure and defects but it also points out that the use of a quasi-reference like the one defined in this study is necessary for data quantification.

Discussion
These results demonstrate the applicability of high-resolution neutron imaging to study porous anodic Al oxide layers. In particular, due to the high sensitivity of neutrons to hydrogen while Al2O3 (single crystalline) and electrolyte species are almost neutron transparent, this technique is ideal to visualize and quantify small changes in the water present in porous films as a function of the anodizing conditions. This preliminary study confirms that porous anodic Al oxides contain water, either as water/hydrogen incorporated in the oxide structure or/and hydration water in the pores. However, the amount of morphological and structural water strongly depends on the anodizing conditions and intrinsically on the oxide formed. Galvanostatic anodizing of pure Al in 0.1 M phosphoric acid leads to the growth of oxides that present similar morphology and amorphous structure up to 6900 s of anodizing, which corresponds to a thickness of 130 µm. The oxide formed after 2320 s (50 µm) however contains two times more water than after 6900 s (130 µm) and present a different water distribution, notably with water enrichment in the top layer constituted of ordered pores. Pore ordering seems therefore to facilitate water incorporation in the pores.
Longer anodizing time, for instance 9300 s (230 µm), leads to the formation of an oxide ten times more hydrated than after 6900 s (130 µm). This increase in water content corresponds to a change in the oxide structure associated to the presence of boehmitelike phase instead of gibbsite-like structure. This lower transmission value does not reflect the structural difference in hydrogen content between boehmite and gibbsite but might be attributed to a higher degree of structural disorder as well as a different density of corundum-like and diaspore-like structures compared to thinner oxides. The presence of boehmite could also facilitate incorporation of water in the oxide structure during growth. On the other hand, despite different "amorphous" oxide structure and higher structural disorder, the porous films formed on AW1050 shows similar water content near the air/oxide interface than the ones formed on pure Al whereas notable changes can be observed near the oxide/Al substrate. It seems that anodizing on AW1050 leads to more water entrapment during pore growth at the bottom part of the layers, probably due to higher pore disorder and blockage. Furthermore, at equivalent thickness, the porous anodic Al oxides grown in 0.5 M sulfuric acid contains at least two and half times more water than films grown in 0.1 M phosphoric acid.
Although both oxides have similar "amorphous" XRD fingerprint, different morphologies are obtained with regards to the electrolyte used. Anodizing in sulfuric acid leads to the formation of long-range ordered pores with an average size of 20 nm whereas in phosphoric acid, branched pores with a pore diameter of 220 nm are formed. This long-range arrangement of 20 nm pores seems to favor water entrapment in the pores but the difference in the contrast transmission value might also be partly associated to the more defective nature of the oxides formed in sulfuric acid compared to oxides formed in phosphoric acid, as it is the case for crystalline Al oxides. These preliminary results show nevertheless that the different water contents are not necessarily correlated to a specific structural or morphological change but strongly depend on anodizing conditions and therefore, the oxide formed. In a next step, it would be interesting to explore to which extent the changes in structural and morphological water content relate to the intrinsic properties (mechanical, optical and electrical properties and chemical stability) of porous anodic Al oxides.
This study moreover shows the potential of high-resolution neutron imaging to study Al oxide structures from crystalline to amorphous. The technique proves to be sensitive to structural disorder and crystallographic orientations, allowing to differentiate crystalline Al2O3 based on their crystalline features. Notably, it is worth emphasizing that each of the Al oxides investigated, crystalline or amorphous, presents different characteristics and are not necessarily interchangeable. Crystalline Al2O3 therefore cannot be used to describe porous anodic Al oxides. Finding a suitable reference sample for porous anodic Al oxides is therefore challenging and requires compromises.
The approach taken in this study was the most pragmatic one. By heating the porous anodic Al oxides without inducing any structural changes in the oxide (detectable by XRD), we prove that water can desorb in the top part of the oxide layers, resulting in a "dryer" porous Al2O3 that can be used as quasi-reference to extract quantitative information. Immersing the sample in water results in reversible filling of the pores.
Interestingly, the bottom part of the oxide layers remains unaffected, which suggests that water is trapped near the oxide/Al substrate interface due to pore blockage.
Valuable information about the mobility of water within the porous anodic Al oxides can therefore be gained and used to optimize functionalization of the oxides.
The presence of edge enhancement effects at the interfaces however limit data quantification in a region of around 25 µm on both sides of the interface, which not only results in the loss of valuable information about the oxide water content at the interfaces but also indirectly requires a minimum sample thickness of 50 µm. The edge enhancement effects only became detectable in the last couple of years with the increase of spatial resolution in neutron imaging 76,77 . New methodological approaches are therefore being developed to minimize and even suppress them.
The use of high-resolution neutron imaging for the characterization of porous anodic Al oxides offers new perspectives to investigate the structural and morphological water content and distribution in relation to growth conditions, functionalization and stability of the oxides. The fact that for crystalline Al oxides the contrast transmission value can be correlated to the crystallite density and orientation further demonstrates promising application of high-resolution neutron imaging to study the effect of the alloying elements and intermetallic particles on fine variation of the oxide structure formed.
With the fast development and advances of high-resolution neutron imaging, higher lateral resolution will soon be achieved. Neutron imaging can then become a very adequate technique to monitor the structural changes occurring in the oxides during application, notably with regards to the formation of hydroxides that can be highly detrimental to surface protection.

Conclusions
For the first time, to our best knowledge, high-resolution neutron imaging was used to study porous anodic Al oxides. Due to the high sensitivity of neutrons to hydrogen, this technique is very appropriate for the assessment of the water content and distribution in porous anodic Al oxides as a function of anodizing conditions. While single crystal oriented Al2O3 is almost neutron transparent, we confirm that, as would be expected, porous anodic Al oxides contains structural and morphological water. Variations of overall water content in porous films are not necessarily correlated to a specific structural or morphological change but strongly depend on the anodizing parameters and intrinsically on the oxide formed. High-resolution neutron imaging proves to be highly sensitive to small changes in the water content, which would be highly challenging to characterize with other techniques. We evidence that morphological water desorption in the top part of the oxides can be promoted by heating the samples while subsequent immersion in water, leads to reversible filling of the pores. No changes is observed near the oxide/Al substrate interface, suggesting water entrapment due to pore blockage. This consequently motivates a proper quantification of the water content when discussing the nature and structure of anodic Al oxides in relation to their intrinsic properties and stability. The use of high-resolution neutron imaging is however not limited to porous anodic Al oxides. It shows promising application to characterize crystalline Al oxides, since the contrast transmission value can be associated to the presence of defects, both structural and morphological, in the oxides. Lateral resolution permitting, high-resolution neutron imaging can become the ideal technique to study the changes occurring in the oxides, in particular with regards to the formation and thermal stability of hydroxides, which can be highly detrimental to surface protection and template applications.
performed at the Swiss spallation neutron source SINQ, Paul Scherrer Institute, Switzerland.
This research was financed internally by Empa and did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Data availability
The raw data required to reproduce the findings associated to Al oxides and hydroxide reference samples (XRD and high-resolution neutron imaging) are available to download from 10.5281/zenodo.3944610. The raw/processed data required to reproduce the findings associated to porous anodic Al oxide layers cannot be shared at this time as the data also forms part of an ongoing study.

Dark current correction
A stack of dark current images was acquired at the beginning of the beamtime. The intensity value obtained was averaged and the mean value determined. This value was then subtracted from the intensity value of , + , and + to correct for the background noise.

Removing the bright outliers
Outliers -bright pixels due to scattered γ-rays -were removed.

Image normalization using BB correction
This has been performed using the "Extended Image Referencing" plugin in ImageJ, developed by the Neutron Imaging and Activation Group (NIAG) at PSI 1-3 . The resulting image was used for quantitative analyses. Figure S2: Summary of the data processing with the example of pure Al P130

Determining the linear attenuation coefficient
The attenuation of radiation passing through matter follows the Beer-Lambert law. The transmission contrast intensity T can be expressed as: where is the linear attenuation coefficient of the material [cm -1 ] through , the sample thickness in the beam direction equal to 2.05 ± 0.03 mm for all studied samples.
The linear attenuation coefficient for a material can be derived experimentally from (1) or it can be calculated using (2) and (3): in which, and are the absorption and scattering microscopic cross sections [cm 2 ], is the number density [cm -3 ], is the Avogadro number 6.022 10 -23 [mol -1 ], is the material density [g cm -3 ] and is the molecular weight [g mol -1 ]. Table S1 compares the values derived from Eq. 1, 2 and 3 with the values measured by neutron imaging, considering: = 2.7 g cm -3 , 2 = 1.0 g cm -3 , 2 3 , = 4.0 g cm -3 ,  Table S1 clearly shows that there is a good agreement between the contrast transmission calculated and measured for pure Al. However, as soon as multi-element compounds are considered, the two values differ due to non-negligible additional scattering effects.
The NEA Java-Based Nuclear Data Information System (JAVIS) database 5 , can also be used to derive the linear attenuation coefficient and subsequently, the expected transmission value. The transmission value calculated for pure Al at the POLDI beamline is 0.97, which is in good agreement with the measured value. The simulation was also performed for a polycrystalline α-Al2O3, using the crystal structure of the collection code 9770, taken from the online ICSD database. The transmission value obtained is 0.948, which is relatively closed from the measured value for the polycrystalline corundum. The NEA could however not be used to derive the transmission values for γ-Al2O3 (metastable phase) or Al(OH)3.
This shows that we need to rely on experimental referencing to determine the linear attenuation coefficient of Al oxides.     Despite the gold coating, the sample cross sections are charging ( Figures S3-S7) but this can hardly be avoided considering the thick Al oxide layers that are investigated. Table S2 reports the thickness and pore sized of the different porous anodic Al oxides investigated. Various reference Al oxides are also included as a comparison. The layer thickness was derived from the neutron images , considering that the size of a pixel corresponds to 2.7 µm, and from the SEM images of the cross-section of the obtained porous Al anodic oxide layers ( Figures S3a, S4a, b and c, S6a and S7d). Both values are in good agreement (see Table S2).

Microstructural characterization of the porous Al anodic oxide layers
The pore size could be roughly estimated from the SEM images. For pure Al S60 (pore size smaller than 100 nm), it was validated by N2 adsorption isotherms at 77 K are scattered, since the pore growth was not controlled in this study.