Understanding the thermal conductivity variations in nanoporous anodic aluminum oxide

Anodic porous alumina (AAO) templates, also known as nanoporous anodic alumina (NAA) platforms or membranes, are widely used as templates in nanoscience and nanotechnology. During fabrication, characterization, or device performance they are sometimes exposed to heat treatments. We have found that the thermal conductivity of the AAO is strongly dependent on the temperature at which the sample is exposed, showing a certain variation even at very low temperatures, such as 50o C. Because of that, we have performed a study of the thermal conductivity (k) variation for different AAO templates using 3ω-Scanning Thermal Microscopy (3ω-SThM) as a function of the annealing temperature. The AAOs studied in this work, where produced in the most common electrolytes sulfuric, oxalic, and phosphoric acids-, and heated from RT up to 1100oC. To explain those variations in an in-depth and enlightening way, structural characterizations of the different nanoporous aluminas have been performed. It is shown that even at low temperatures, below 100oC, the AAOs loose water, which explains the reduction of their thermal conductivity. The minimum value of thermal conductivity can be found for AAO samples prepared in sulfuric acid and heated at 100oC (0.78 ± 0.19 W/m·K), which corresponds to a 50% reduction from the original value. It has also been found that for AAO annealed over 950oC, the variations on the thermal conductivity are mainly related to phase transitions from amorphous to crystalline alumina and gas evolution of CO2, or SO2 and SO, depending on the electrolyte used during the anodization. This is also an indication that the counter ions are trapped inside the alumina template during the anodization process. And, their presence determines the Page 1 of 27 Nanoscale Advances N an os ca le A dv an ce s A cc ep te d M an us cr ip t O pe n A cc es s A rt ic le . P ub lis he d on 1 7 A ug us t 2 02 0. D ow nl oa de d on 9 /1 1/ 20 20 9 :4 7: 05 A M . T hi s ar tic le is li ce ns ed u nd er a C re at iv e C om m on s A ttr ib ut io n 3. 0 U np or te d L ic en ce . View Article Online DOI: 10.1039/D0NA00578A crystallization temperatures at which the different crystalline phases are formed. And so, on the variations in thermal conductivity measured. The thermal conductivity values of the AAO can reach values as high as 4.82 ± 0.36 W/m·K for AAO samples prepared in oxalic acid and heated up to 1300oC. We consider that the understanding of the changes in their thermal conductivity can explain the different values in the thermal conductivity of the AAO template found in the literature. And, draw some attention to the importance of the history on the AAO platforms that are going to be used or measured, since it may change the final thermal properties of the template or device.


INTRODUCTION
Nanoporous anodic aluminum oxide (AAO) membranes are very popular and widely used in nanoscience and nanotechnology. Due to its easy manufacturing and size tailoring, AAO membranes are an elegant nanostructuration approach. And they are recognized as an alternative to more sophisticated and expensive methods currently used in nanofabrication.
When the AAO membranes are used in many cases they are exposed to low-temperature heat treatments. Either during the nanostructure fabrication, when using techniques such as metal deposition [8], thermal evaporation, sputtering [16], melt infiltration [24][25][26], etc. Or during the post-processing, where fabricated nanowires or nanotubes must be heated to crystallize them. Or during their characterization, such as the fabrication of lithographically designed contacts where the resists have to be cured, or the AAOs are exposed to laser beams for long-times, etc. In these processes, the temperature of the AAOs can rise from tens to hundreds of degrees for different extents of time.
It is generally assumed that the AAO membranes do not change their chemical and physical properties when exposed to low temperatures. But in the literature, there is a wide range of reported values for the thermal conductivity of AAOs (see Table I), with values ranging from 0.5 to 1 W/m·K were measured by a steady-state method at room temperature (RT) for AAO membranes fabricated in sulfuric acid when varying the anodizing temperatures [27], or higher values, such as 1.3 ± 0.1 W/m·K at room temperature when these sulfuric membranes are measured by the 3ω-technique, the photothermal method [28] or laser flash [29]. In the case of oxalic anodized AAOs, values between 0.63 W/m·K and 1.33 W/m·K were measured by the steady-state technique at RT when the anodizing temperature varied [27]. The range for thermal conductivity at RT for phosphoric anodized AAOs goes from 1.32 W/m·K when measured by the photoacoustic technique [30] to 1.9 W/m·K when similar AAOs were measured by 3ω-SThM [31]. Sometimes these discrepancies in the values have been interpreted as compositional changes, or to different grain boundaries in the non-ordered porous nanostructures [37]. All these, without forgetting that when the AAO membrane is heated at high temperatures, it crystallizes, and the values of bulk α-Al 2 O 3 thermal conductivity at RT can be as high as 30 W/m·K [32]. Anodic oxidation in sulfuric, oxalic, and phosphoric acids. [30] To explain these discrepancies, in this work, a set of samples was fabricated using the three most used acid solutions (namely, sulfuric, oxalic, and phosphoric) and anodizing conditions. The changes in the thermal conductivity, structure, and crystallization are studied in AAO membranes as-obtained and after annealings up to 1300ºC. The thermal conductivity of the AAO membranes has been measured by 3ω-Scanning Thermal Microscopy (3ω-SThM), which allows us to characterize this property at the nanoscale.
Also, some of the values obtained were cross-checked with the photoacoustic method, obtaining similar thermal conductivity values within the experimental error for both techniques [42,43]. This in-depth study is crucial to draw some light on the variation of performed via a two-step anodization process, as it is described in [30]. In brief, it consists of a first anodization where an aluminum oxide layer is created on the surface. This is carried out for 24 hours to achieve the ordering of the pores, followed by the removal of this first anodic alumina film by chemical etching in phosphoric acid 6% wt., chromic oxide 1.8% wt. and deionized water. Then, a second anodization step is During this process, the alumina template could be heated at low temperatures because of the exothermic reaction that is taking place. So if this process is made or how it is made can explain the differences in the water content in the alumina at RT that we have observed in the literature when we compare our TGA/DSC loss from our aluminas with other from the literature like in [44]. This is also a clear source of variation on the actual values of the thermal conductivity of the alumina since some alumina templates will have more water inside than others. This would change the thermal conductivity value at room temperature and their variation from RT to 100º C.
Finally, the barrier layer was dissolved in 10%wt. H 3 PO 4 at 30º C, obtaining a nanoporous membrane. The phosphoric-AAOs were commercial membranes from the company Whatman®, with pores around 200 nm in diameter. All the measured AAO membranes presented a total thickness between 35 to 60 µm.

Annealing processes
The annealing process was performed in a tubular oven (Model ST115020 from Hobersal equipped with an external temperature controller). A second thermocouple was placed inside the oven with the samples to read the actual temperature at which the samples were being heated. The samples were introduced at room temperature and then heated with a heating rate of 10 ºC per minute up to the desired temperature (50, 100, 150, 200, 600, 950 and 1100 ºC), where they were kept for 1 hour, and then they were cooled down to ambient temperature. The samples heated up to 1300º C were made in different equipment, heated with at the same rate of 10º C per minute, and cooled down after the maximum temperature was reached. Given that it has been reported that for higher temperatures the phase transformation takes place in shorter times [45], we have also included them in our study. Finally, AAOs without any thermal treatment (which we will call room-temperature, RT, samples) were also included for comparison.

Characterization techniques
The characterization of the samples was carried out with a Scanning Electron Microscope (SEM) VERIOS 460 from FEI. Differential Scanning Calorimetry (DSC), Thermogravimetry (TGA), and mass spectroscopy were measured in a DSC/DTA/TGA module Q600 from TA Instruments, which can measure simultaneously DSC and TGA from ambient to 1500º C (from the SIdI-UAM service), with a nitrogen gas flow of 100 ml/minute. The measurements were performed in a dynamic regime while applying a thermal ramp of 10ºC/minute. The system has also a mass spectrometer that can give information on the gasses released by the sample. The crystalline structure of the AAOs has been studied with an XRD Philips X'Pert four circles diffractometer with CuK α wavelength. Density measurements were carried out via the Archimedes method, using an analytical balance XSE105DU from Mettler Toledo. Raman spectra were recorded with a confocal Micro-Raman spectrometer LabRAM 800 from Horiba Jobin Yvon with an excitation wavelength of 532 nm. The thermal conductivity of some membranes was also measured by the photoacoustic method for metrology porpoises. More detailed information can be found in reference [30]. This method is based on detecting the phase-shift between periodical heating of the sample with a pulsed laser and the acoustic waves which are generated due to this heating and propagated inside a specially designed photoacoustic cell.
Topographic images of the AAOs at room temperature were obtained with an Atomic Force Microscope (AFM) from Nanotec Electronica, operated in contact mode. To obtain the thermal images, the AFM was operated as scanning thermal microscopy (SThM) in 3ω mode, scanning the surface of the samples with a sharp temperaturesensing tip [46], which has to be previously calibrated (see Supporting Information for further details). From this calibration, we obtain a heat exchange radius of 250 nm. In brief, this mode consists of sending a modulated voltage at a frequency ω to the tip, As a result of these temperature changes, the resistance of the tip also changes, and those changes can be recorded thanks to a lab-made Wheatstone bridge. Analyzing the signal obtained at the third harmonic (3ω) of the original modulated voltage, it is possible to obtain the thermal profile of the studied sample, and, after applying an effective medium theory to take into account the contributions of the pores and the skeletal alumina (see Supporting Information for further details, and Figures S2 and S3), obtain the thermal conductivity coefficient [31,47]. The probes used in this 3ω-SThM setup were micro-fabricated Pd/Si 3 N 4 from Bruker®. Data acquisition was done using a lockin amplifier from Zurich Instruments. The data were obtained and processed using the WSxM software [48]. It is important to highlight that in the case of SThM measurements, there is no need for previous preparation of the sample or minimum area required. So the thermal conductivity of the samples is not affected by any posttreatment of the AAO membranes.

RESULTS AND DISCUSSION
The thermal conductivity was obtained at room temperature for the different AAOmembranes previously annealed, using the 3ω-Scanning Thermal Microscopy technique (3ω-SThM). In Figure 1 the obtained images (topography, 3ω, and 1ω) of oxalic-AAO treated at different temperatures are shown, as an example. The images were recorded for different AAOs as obtained and annealed at different temperatures up to 1300º C.
For each of the samples measured along with the topographical image, we obtain the 1ω signals, which allow us to extract qualitative information of the temperature, and the 3ω signals, which allow as to quantitative measure the thermal conductivity. These images can be obtained with high sensitivity thanks to a recently improved lab-made electrical circuit connected to the thermal tip, which gives also a better resolution for the first and third harmonic of the signal. The tip has been calibrated similarly to previous works [46,49] (see Supporting Information). As a result of the calibration, we have obtained the thermal exchange radius of our tip to be of 250 nm.
Just by looking at the images of Figure 1, it can be observed that the as-obtained membrane has a different temperature distribution (Figure 1.c) than the samples heated at 100º C. As it can be observed in Figure 1.f, the image presents some hot spots with a local change of their thermal conductivity. This is an indication that something is happening at the nanoscale inside the AAO membrane even at 100ºC. These observed variations are of importance to understand the disagreements in thermal conductivity values obtained in the literature when measuring similar AAO membrane. Furthermore, taking into account that each measuring technique may need different post-processing of the AAO membrane that can be associated with heating processes.
It is interesting to appreciate that although the thermal exchange radius is 250 nm for the 3ω, higher resolution can be achieved in the 1ω images, although it cannot be quantified. But it allows hinting that the thermal conductivity must be slightly different between the so-called inner and outer layers of the AAO membranes. The outer layer is the region next to the pore channel walls, which has more anion contamination from the electrolyte, while the inner layer is a higher purity alumina region (the approximate limit of both regions is marked with a dashed line in figures 1d and 1h). So it makes sense that they have different thermal conductivities since they have different chemical compositions. The fact that we can observe differences between both layers is interesting because it denotes the high resolution we can achieve in 1ω. It is also interesting to note in Figure 1 that by heating, the inner and the outer AAO regions behave differently. To better show this variation, the AAO between the pores is amplified in figures 1d, h, and l. Finally, in Figure 1i the images of the oxalic-AAO membrane annealed at 1100º C are shown. One interesting feature is that the pore size is bigger, which means that the AAO composition must be changing. It can also be observed that the thermal conductivity of the AAO membrane is strongly modified (see the scale in V of Figure 1b, f, and j). In the first low-temperature region (up ~200º C) a reduction on the thermal conductivity is observed and followed by an increase. This reduction is more noticeable for the samples produced in sulfuric acid solution (sulfuric-AAO). In these samples, a reduction in thermal conductivity is observed from 1.24±0. 12  This reduction of the thermal conductivity upon heating at such low temperatures can explain the variability of the thermal conductivity in the literature, as shown in Table I.
Since some of the measurement techniques need sample preparation, such as the deposition of a metal layer, which can heat the sample to those temperatures very easily.
A second region, between temperatures over 200 to 600 ºC, is characterized by the thermal conductivity of the AAO membranes to be around 2 W/m·K, with differences in the thermal conductivity between the different membranes that stay within the error limit of the measurement technique. To understand the process that is involved in the variation of the thermal conductivity at the different temperatures, we have performed TGA-DSC-mass spectroscopy in the three types of samples to analyze: the weight loss, if the process is endo-or exothermic, and which gas is released during each particular process (see Figure 3). This analysis has been coupled with X-ray diffraction analysis, (shown in Figure 4), and Raman Spectroscopy ( Figure 5) to understand if there is also a crystallization process associated with those temperatures. annealing, which is also related to the increase of the thermal conductivity seen in this region. This is an interesting result since the composition of the AAO templates is changing at very low temperatures, which can explain the variation in thermal conductivity observed in Figure 2 and on Table I.
In the middle-temperature range (from >200º C to 600º C in annealing) there is not an important weight variation in the different AAO samples as observed by TGA in Figure   3. The loss of trapped counter ions is produced at the same time that the crystallization of the alumina, as evidenced by the XRD (see Figure 4). In both cases, sulfuric-AAOs and oxalic-AAOs, the samples annealed at 950º C start to show diffraction maxima around 2θ 46º and 67º. So the membranes crystallize as γ-Al 2 O 3 or δ-Al 2 O 3 phase (it is difficult to distinguish between these two phases, given that most of their diffraction peaks are common, according to JCPDS-ICDD 00-029-1486 and JCPDS-ICDD 00-016-0394, respectively). This transition is marked with the letter "γ" in Figure 3. For oxalic-AAOs annealed at 1100º C, apart from these phases, some maxima corresponding with the α- important to highlight also that the degree of phase transformation at 1100º C is higher for the oxalic-AAOs than for the sulfuric-AAOs at 1300º C, as detected by XRD. Since the α-Al 2 O 3 phase has a higher thermal conductivity, as seen in Table I, this explains why the thermal conductivity is higher in the oxalic-AAO than in sulfuric-AAO at 1300ºC.
In the case of the phosphoric-AAO, there is not any important weight loss in that particular range. According to mass spectroscopy, there is not gas evolution either. This is an indication that phosphates trapped inside the alumina membrane do not thermally decompose and stay in the alumina membrane. By DSC, only two small exothermic processes have been observed around 880 and 970ºC, which are related to crystallization, but no other changes are observed above those values. These exothermic processes must be related to the crystallization of the phosphoric-AAO, which starts to be identifiable by XRD, around these temperatures, as shown in Figure 4. At 950ºC annealing temperature, the diffraction maxima observed can be associated with the γ-  The Raman spectra of the three sets of samples ( Figure 5) confirms both the different phase transitions observed by XRD and the gas evolution detected by TGA-DSC mass spectroscopy, which were comparable to those discussed in previous works on similar samples [44,50], and it adds some further details in the crystallization route followed.
For sulfuric-AAOs from RT to 600 ºC annealing temperatures, the Raman spectra show only vibrational modes corresponding to sulfur compounds, like those associated to SO 4 2-(located at 451 cm -1 (ν 2 ), 626 cm -1 (ν 4 ) and 984 cm -1 (ν 1 ) [51]). These cannot be found for samples annealed over 950 ºC, whereby TGA-DSC it has been seen that these trapped counter ions have been thermally decomposed at 980ºC, and thus, are not present over this temperature. Another vibrational mode, clearly visible from RT is that associated to HSO 4 -, around 1050 cm -1 (ν s ), which is maintained until the sample is annealed at 950 ºC, where it changes to a different HSO 4mode, namely a ν 3 vibration centered at 1100 cm -1 [52], and decreases in intensity. This coincides with the gas evolution of thermal decomposition of SO 2 and SO in this range, and the faint intensity of this vibrational mode at the highest temperatures reflects that there is still a residual presence of these defects in the structure, still detectable by Raman. This is also confirmed by mass spectroscopy, where a second gas evolution around 1175º C can be also observed. Although it has to be pointed out that the presence of electrolyte counterion related phases is not detectable by XRD. In the case of the crystallization route of the samples, for those above 950 ºC annealing temperature, other faint peaks associated with the α-Al 2 O 3 crystallographic phase and weak Raman lines, at around 252 and 268.6 cm -1 are found. These bands have been discussed in the literature [53,54] as a mark of the co-existence of θ -Al 2 O 3 or γ-Al 2 O 3 phases with α-Al 2 O 3 phase for AAO-membranes treated over 1000º C, which is what is obtained by XRD for these samples.
In the case of oxalic-AAOs, the Raman spectra for the samples under 600º C show important photoluminescence, which is usually associated with impurities or oxygen vacancies [55]. These impurities amount seem to be reduced when the sample is annealed at 950 ºC (which is over the temperature found in TGA-DSC for the first thermal decomposition of C 2 O 4 2to CO 2 , ~890ºC). This gas evolution can explain why the photoluminescence is much lower and peaks associated with aluminum oxalate, [51], centered at around 846, 981, 1110, and 1387 cm -1 appear weak and as broad bands in the spectra. For lower temperatures, these peaks are present in the sample, but the presence of the photoluminescence band, which is strong and broad, masks the Raman for the laser wavelength used in this study. Nevertheless, when the oxalic-AAO samples are measured with the appropriate wavelength (for instance, an Ultraviolet laser with 244 nm wavelength), as it was made in reference [51], the photoluminescence interference can be avoided and then, the oxalate impurities can be observed, even in the room temperature samples. This oxalate peaks disappear in the sample heated at 1100ºC, where no trace of aluminum oxalate could be identified in agreement with the information from mass spectroscopy. When the annealing temperature is increased at 1100ºC, some α-Al 2 O 3 phase vibrational modes start to appear. Those α-Al 2 O 3 phase or sapphire vibrational modes are located at 378 cm -1 (E g external), 418 cm -1 (A 1g ), 432 cm -1 (E g external), 451 cm -1 (E g internal), 580 cm -1 (E g internal), 645 cm -1 (A 1g , polarized zz) and 756 cm -1 (E g internal) as reported in [56]. In Figure 5 their positions are marked as grey dashed lines, and it can be seen that all of them are present and well defined for the sample annealed at 1100ºC.  [59] could be identified at those temperatures. It is worth to mention that other works dealing with phosphoric-AAOs studied in temperature have also shown the presence of AlPO 4 even at temperatures of 1450ºC [44]. So, it can be concluded that at those temperatures the trapped phosphates forms α- (with a value of 4-5% in water loss). Finally, the water loss is very low phosphoric-AAOs, with around 1-2% water loss at that temperature, which correlates with a less marked change in the thermal conductivity values for this range.
In the second region, between 200 to 600ºC, the thermal conductivity was stable around a value of around 2 W/m·K, and this corresponds to TGA-DSC-mass spectroscopy results showing no special gas evolutions, or crystallization occurring in that range, as observed by XRD and Raman spectra.  To identify if those gas evolution and crystallization processes have any effect on the morphology of the AAO membranes, SEM images of each of the AAO membranes were taken after the different temperature treatments and compiled in Figure 6. To quantize those changes, the porosity and pore diameter has been analyzed as a function of the annealing temperature (see Table S.II of the Supporting Information). In all cases, the SEM images show that the porous diameter does not change up to 950º C, indicating that the membranes can be useful in different applications in which temperature processes up to this value are involved. For annealings over 950º C, the porous structure of AAO membranes is slightly modified, as it has also been reported in previous works, for instance, oxalic-AAOs [60]. Their porosity increases to more than double its value for sulfuric-and oxalic-AAOs. And so does the pore diameter. This can be due to the two processes that were mentioned before for those two types of AAOs in that temperature range: gas evolution and crystallization. It is worth reminding that, for oxalic-AAOs, after 1100 ºC annealing, the α-Al 2 O 3 phase is obtained, which has a higher density than the others (as it can be also seen in the density plot, Figure S4 of the Supporting Information) In the case of phosphoric-AAOs, they do not show any clear morphological change in pore diameter. This can be explained by the fact that there is not gas evolution and the crystallization of the sample is barely starting. Nevertheless, it is important to say that phosphoric-AAOs become buckled or cracked when heated over 700º C. For the highest annealing temperatures, the SEM images of the phosphoric-AAO show the apparition of "whiskers" covering the surface, similar behaviour has been observed in previous works for the same type of AAOs [61]. This morphological change is seen above 950º C in this study, as can be observed in the SEM images ( Figure 6.f, and 6.c). These whiskers are amorphous in XRD, but we have been able to identify them by Raman Spectroscopy as AlPO 4 and Al 2 (PO 4 )(OH) 3 , see Figure 5. The formation of these phases can also be identified in DSC. And it is marked as AlPO 4 in Figure 3. This phase segregation at the top and bottom of the membrane generates a phosphorous gradient concentration though the pores and can explain why these membranes became more brittle than the other membranes. To take proper thermal conductivity measurements as shown in Figure 2, the samples were polished.

Conclusions
In this work, we have understood the influence of annealing treatments on the thermal conductivity of different AAO membranes prepared in the most common anodization acids: sulfuric, oxalic, and phosphoric. With the information gathered in this work, we can conclude some important tips for the scientific community using alumina membranes.
On the one hand, the thermal conductivity of the AAO membranes depends on the acid used in the anodization. This is due to the influence of counter-ions that get trapped inside the outer layer of the alumina during the anodization process. Therefore, each type of alumina presents a different thermal conductivity value because they are slightly different.
On the other hand, the thermal conductivity of the AAO membranes changes even at very low temperatures (50ºC). This is very important because depending on the thermal conductivity measurement technique, one may need to evaporate metal contacts, or transducers, or curing a photo-resist when making lithography process, or the samples must be heated by lasers, etc. which can generate some heating in the AAO membrane.
This heating can produce water evolution from the AAO membrane, which modifies their thermal conductivity. This is for example of importance when determining the thermal conductivity of the nanowires inside AAO membranes. To obtain an accurate value of the nanowires, the empty AAO membrane must be measured under the same conditions and post-treatments that the AAO with the nanowires, to get an accurate value of the nanowires, instead of using the values from the literature for untreated AAOs.
Another important conclusion is that the thermal conductivity of sulfuric-AAOs can be reduced to values around 0.7 W/m·K upon heating the membrane at 100ºC. That is a good result, for example, if the AAO membrane wants to be used in areas in which a low thermal conductivity is of importance, like in thermoelectrics.
It is also worth noting that the impurities are different depending on the electrolyte used in the anodization process and that they affect differently the crystallization of the AAO membranes. These impurities influence the temperature at which the transformation starts and also in the crystalline phases in which the AAO crystallizes. These, of course, have an influence on the value of thermal conductivity of the membranes after different thermal treatments. For example, over 1300ºC the total transformation to a stable hcp α-Al 2 O 3 phase happens for oxalic-AAOs, partially for sulfuric-AAOs, and it is not reached for phosphoric-AAOs.
Therefore, this work provides a better understanding of how the thermal conduction is changing in AAO membranes. This is key to understand the impact that the annealing temperature and the anodization conditions have on the physical, chemical, and structural properties of this oxide. It must be taken into account when applying this type of structure as self-supporting membranes or coatings. These findings could have a great impact on the engineering design of the different anodized oxide porous structures, having also a high relevance for the development of thermal transport models for these types of structures.
Finally, this understanding of the change of thermal conductivity in AAOs must be taken into account when using these membranes in real devices. If the membrane is heated up (even at very low temperatures) the thermal conductivity will change, modifying its final performance.