Sol-gel synthesis of α-Al2O3 with enhanced porosity via dicarboxylic acid templating

One of the major routes to synthesize macroporous α-Al2O3 is the sol-gel process in presence of templates. Templates include polymers as well as carboxylic acids, such as citric acid. By careful choice of the template, pore diameters can be adjusted between 110 nm and several µm. We report the successful establishment of plain short-chain dicarboxylic acids (DCA) as porogenes in the sol-gel synthesis of macroporous α-Al2O3. By this extension of the recently developed synthesis route, a very precise control of pore diameters is achieved, in addition to enhanced macropore volumes in α-Al2O3. The formation mechanism thereof is closely related to the one postulated for citric acid, as thermal analyses show. However, since branching in the DCA-linked alumina nuclei is not possible, close monomodal pore width distributions are attained, which are accompanied by enhanced pore volumes. This is a significant improvement in terms of controlled enhanced porosity in the synthesis of macroporous α-Al2O3.

characterization of macroporous α-Al 2 o 3 . The obtained granular material was characterized by mercury intrusion, SEM, nitrogen sorption, and XRD. Furthermore, thermal analyses (TG and DTA) were conducted on as-synthesized, dried samples.
Mercury intrusion was performed on a Pascal 440 porosimeter by ThermoScientific/Porotec with pressures ranging from 0.2 mbar to 4000 bar. Mercury surface tension was assumed to be 0.484 N/m, its contact angle was set to 141.3°. Samples were outgassed at 0.2 mbar for 10 minutes at ambient temperature prior to filling the dilatometer with mercury.
Scanning electron microscopy (SEM) images were obtained using a Leo Gemini 1530 by Zeiss. Samples were fixated on a carbon foil and vapor coated with a gold film. Accelerating voltage was 10 kV. Secondary electrons were collected by an Everhart-Thornley detector.
Nitrogen sorption was carried out on an ASAP 2000, Micromeritics. Prior to examination, the samples were dried, outgassed, and activated at 300 °C under vacuum. Determination of the specific surface area (A BET ) was conducted using the linearized form of the BET equation in the range of 0.05 ≤ p/p 0 ≤ 0.30.
For thermal analyses, 25 mg of non-calcined sample were mixed with 25 mg of pure corundum and placed in a corundum crucible in a Netzsch STA 409 TG/DTA device. The continuous air flow rate was 75 mL/min. The heating rate was set to 10 K/min, starting from room temperature up to 1250 °C. All samples were dried to mass constancy at 120 °C prior to thermal analyses.
Complete transformation to the α-Al 2 O 3 phase and phase purity were confirmed by X-ray diffraction on a D8 Discover by Bruker, using a Vantec500-2D detector.

Results and Discussion
Depending on preparation conditions and thermal history, alumina can be generated in different modifications. The thermodynamically stable, intrinsically almost non-porous one is α-Al 2 O 3 , which is obtained from dried gels by calcination at 1200 °C for 6 h. We confirmed complete transformation to α-Al 2 O 3 for all synthesized samples by XRD to ensure that none of the generated porosity is due to remaining fractions of transition alumina, which is often the case even in patented processes 17 . An exemplary diffractogram of sample C3 is shown in Fig. 1.
By adding dicarboxylic acids to the standard sol-gel assay, porosity can be enhanced significantly. The additive-free sample Ref0 exhibits pores of 116 nm width, accumulating to a pore volume V p of 0.12 cm 3 /g. Only marginal deviations thereof are observed when 10 mol-% of adipic acid (sample C6, ϕ Al = 10) are added. However, these values can be augmented up to d p = 838 nm, and V p = 1.18 cm 3 /g, by adding the same equivalent of oxalic acid (sample C2#1, cf. Table 2). Porosity figures for malonic acid (C3), succinic acid (C4), and glutaric acid (C5) steadily decrease from these peak values down to the ones obtained for C6 and Ref0, as listed in Table 2. The virtual absence of porosity enhancement for C6 can partly be attributed to the poor solubility of adipic acid in water and ethanol. (Dissolution of the additive in the reaction mixture took several hours. We hence refrained from pursuing assays with dicarboxylic acids C7 and higher homologues.) More importantly, however, the general pore widening mechanism by DCA addition in the sol-gel process needs to be elucidated to explain the observed microstructural characteristics.
Clearly the enhancing effect on porosity is strongest for the shortest DCA, oxalic acid. In our recently published article on citric acid-assisted synthesis of highly porous α-Al 2 O 3 , we provided evidence for the formation of citrate-Al(III)-oligomers, eventually leading to phase separation and thereby an increased porosity 16 . For DCAs, the same underlying mechanism can be postulated, which is illustrated by SEM images given in Fig. 2.
Comparative thermal analyses of samples C3, CA68, and the additive-free reference Ref0 support this interpretation, as they reveal complete incorporation of malonic acid into the alumina network. No discrete mass loss for the DCA is visible in the TG curve in Fig. 3. However, the total weight loss amounts to 44.1 % for sample Ref0, while C3 presents a weight loss of 48.3 %, hence including the 0.34 g of malonic acid. Correspondingly, the weight loss amounts to 54.0 % for sample CA68, containing 0.68 g of citric acid. (The difference to pure Al(OH) 3 as starting material, which would result in a weight loss of only 35 %, can be explained by physisorbed excess solvent for all three samples.) Moreover, just like previously observed on sample CA68, the endothermic phase transition peak at 535 °C is missing in the DTA curve of sample C3. This indicates the strong structure directing effect of malonic acid, preventing the formation of a transitional γ-phase, and provides evidence of the incorporation of malonic acid into the alumina network on a molecular level. (For a more detailed discussion of the mechanism, please refer to 16 ). www.nature.com/scientificreports www.nature.com/scientificreports/ Addition of adipic acid then leads to almost no perceptible change in microstructure and porosity, since adipate-Al(III)-oligomers with their aliphatic C4-units are poorly soluble in the strongly polar medium and thus will not grow extensively to form larger particles. A shorter DCA chain facilitates the solubility of forming dicarboxylate-Al(III)-oligomers in a polar medium like the employed reaction mixture. The oligomers hence grow to form larger particles, resulting in a larger volume fraction of nanoparticle-free solvent, which in turn not only yields larger pore diameters but also an increased pore volume. This effect can be seen by comparing the SEM images of C3 and C2#1. Finally, when increasing the DCA concentration, phase separation occurs just like for citric acid-assisted samples. To demonstrate this effect, an additional sample C2#2 was prepared with twice the amount of oxalic acid as in sample C2#1, leading to ϕ Al = 5. The excess oxalic acid then causes the growth of larger primary alumina particles and consequently an eventual enthalpy-driven phase separation, in analogy to the one observed for citric acid 16 . Juxtaposition of samples CA68 and C2#2 in Fig. 2 shows the almost identical microstructural habitus. Further increase of the added amount of oxalic acid is not possible due to its water solubility of only 90-100 g/L for the non-hydrated form. It would be feasible for malonic acid since its solubility in water is 15 times higher 18 . However, due to the additional carbon atom in the chain, the generation of larger pores like in sample C2#2 would require the addition of vast amounts of malonic acid.
Despite the closely related mechanism, there is one major difference between the branched tricarboxylic citric acid and the linear dicarboxylic acids presented in this article. The latter ones offer no branching opportunities via the organic linkers, as illustrated in Fig. 4. Particles hence grow more uniformly, resulting in a more homogeneous alumina network, as shown in Figs. 2 and 5 (close-ups). The mercury intrusion histogram of sample C3 depicts a very narrow monomodal pore width distribution, with ≈ 80 % of the pores exhibiting the modal diameter. Citric acid is a branched molecule and thus can link more alumina nuclei, as schematically shown in Fig. 4. CA27 primary particles hence grow to about the same size as for C3 with only 60 % of the carboxylic acid functional groups.   www.nature.com/scientificreports www.nature.com/scientificreports/ The effect of this unbranched linking in DCA is reflected in larger V p generated for relatively smaller d p with DCA. For instance, oxalic acid, added with ratio of ϕ Al = 10 (sample C2#1) yields a V p of 1.18 cm 3 /g, with a d p of only 838 nm, while for an identical pore volume, CA addition (sample CA68, also with ϕ Al = 10) yields a d p of 3730 nm (cf. Table 2). As a second example, comparison of samples C3 (ϕ Al = 10) and CA34 (ϕ Al = 20) reveals the impact on the resulting A BET . For a slightly smaller V p (0.75 vs. 0.79 cm 3 /g), sample C3 exhibits an A BET of 10 m 2 /g due to its smaller d p of inly 259 nm, while for sample CA34, larger pores (d p = 916 nm) lead to a reduced A BET of 6 m 2 /g (cf. Table 2).
Moreover, CA already induces phase separation at ϕ Al = 15. Thus, the generation of a controlled monomodal pore system is considerably more difficult than with dicarboxylic acids, since small deviations in the amount of additive entail a significant alteration of the resulting porosity. The use of DCA thus broadens the range of www.nature.com/scientificreports www.nature.com/scientificreports/ attainable pore diameters, with both monomodal pore size distributions and enhanced pore volumes, offering new opportunities in terms of their use as catalyst support material, for instance 4 .
Apart from the limited solubility of DCA with even carbon chain numbers and their tendency to esterify with solvent alcohols, there are no major disadvantages to the DCA route. On the contrary, the two major improvements using DCA as templates in comparison with citric acid consist in increased pore volumes and more uniform monomodal pore structures. Sample C3, for instance, is the one with the highest BET surface area of all DCA-templated α-Al 2 O 3 gels due to its drastically increased pore volume in combination with a still relatively small pore diameter, which is yet sufficiently large to withstand sintering at higher temperatures ( < 1350 °C). This renders DCA-templated α-Al 2 O 3 particularly apt for further high temperature applications in refractory ceramic foams 4 , which then may be used as catalyst supports in catalytic cracking, for instance 17,19 .   www.nature.com/scientificreports www.nature.com/scientificreports/ conclusion Dicarboxylic acids were employed for the first time as porogenes in the sol-gel synthesis of macroporous α-Al 2 O 3 . The corresponding mechanism likely proceeds via dicarboxylate-Al(III)-oligomers, which yield larger primary particles, and consequently result in larger pore widths. With increasing carbon chain length, this effect decreases and almost vanishes for adipic acid, which forms only small linked structures, since larger ones would be insoluble.
Best results in adjusting pore diameters and increasing pore volumes are obtained for oxalic and malonic acid, leading to very narrow monomodal pore structures with pore sizes in the range of 200 to 900 nm (for 10 mol-% of DCA). When increasing the amount of additive, phase separation can be induced, resulting in similar structures as the ones obtained for citric acid, with pore diameters going up to 2.5 µm. However, larger amounts of DCA are necessary to attain the same effect as with the branched citric acid, which can link more alumina nuclei with its additional carboxylic acid group. In summary, dicarboxylic acids enable a more precise control of monomodal pore size and the generation of higher pore volumes in macroporous α-Al 2 O 3 synthesized via the sol-gel process than the previously reported routes involving citric acid or PEO.

Data availability
The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.