Keywords

5.1 Introduction

Over the course of the three funding periods of the Collaborative Research Center (CRC) 920, subproject A04 has tackled a diverse set of topics centered around analyzing materials which are important in the context of this CRC, i.e., metal melt filtration in order to produce zero defect materials. This is unsurprising, given that the subproject was active in a time span of more than a decade.

The first two funding periods, which were shaped by subproject member Dr. Lilit Amirkhanyan, focused on intermetallics. Intermetallic phases are interesting systems for designing construction materials due to their low weight, their pronounced deformation resistance at elevated temperatures, their extraordinary hardness, and their general resilience against wear, oxidation, and corrosion [1, 2]. Through filtration processes such as those investigated in the CRC 920, the ratio of the components within the intermetallic phase can be regulated and impurities can be removed [1]. Within subproject A04, density functional theory (DFT [3, 4]) was chosen to non-empirically predict thermodynamic properties of intermetallic systems related to melts, such as heat capacities over a wide temperature range [1, 2, 5].

Moreover, research was also conducted on interface reactions and interface energies during this first funding period. Hercynite (Al2FeO4) has a spinel structure and is among the oxidic compounds that form at the interface between alumina-based filters and iron-containing melts such as steel melts [6]. The deposited hercynite particles are believed to aid filtration efficiency by acting as anchor points for impurities in the melt [6]. Here, DFT was applied to study possible solid state reaction pathways leading to the formation of hercynite at the interface. Rutile (tetragonal TiO2), on the other hand, is used as a coating for alumina filters which are employed in the filtration of aluminum melts and melts of aluminum alloys, particularly to deal with oxidic inclusions [7]. For these melts, rutile-coated filters can achieve both satisfactory melt flow rates and a decent filtration efficiency [7], favorable properties which are based on chemical reactions of rutile with the components of the melts. In this case, the aim of DFT was to evaluate the energetics of rutile-containing interfaces that are established because of the aforementioned reactions [7].

In 2017, the first major cooperation between DFT and Raman spectroscopy within the CRC 920 took place [8]. Raman spectroscopy is a non-destructive method to measure the vibrations of solids and molecules via the inelastic scattering of light on matter. Rudolph et al. investigated the thermally induced phase transition from boehmite (γ-AlO(OH)) to corundum (α-Al2O3) via the metastable transition phases γ-Al2O3, δ-Al2O3, and θ-Al2O3. This investigation was relevant to the goal of metal melt filtration because γ-Al2O3 is considered a promising coating material for corundum-based filters due to its high surface area and its function as a catalyst [8].

A year later, in 2018, there was a clear shift in focus within subproject A04. While carbon-bonded filters had been studied with Raman spectroscopy as part of CRC 920 before in the work of Röder et al. [9], the publication by Himcinschi and coworkers marked the starting point for the investigation of those carbon-bonded filters that use environmentally friendly materials as binders, i.e., lactose and tannin [10]. Compared to the usual pitch binders and phenolic resins, lactose/tannin binders do not lead to the emission of toxic phenol or carcinogenic polycyclic aromatic hydrocarbons like benzo[a]pyrene during either the production or the operation of the respective carbon-bonded filters [10, 11]. These emissions are not only a risk to humans and the environment in general, but provide a source of conflict with current pollution and hazard regulations in the European Union, which also has economic repercussions. However, the thermomechanical stability of the filters based on lactose/tannin binders was found to be not always on the level of state-of-the-art conventional binders, which is why further investigation was necessary [10, 11].

With Dr. Himcinschi joining the subproject as a project leader and Simon Brehm as well as Jakob Kraus replacing Dr. Amirkhanyan as project members, Raman spectroscopy was now an integral tool for addressing the remaining workload. Ex-situ Raman measurements were employed to study both the binder materials lactose, tannin, and Carbores®P themselves as well as carbon-bonded filters using various mixtures of these substances as binders [12]. For all of these systems, the prominent D and G Raman peaks were used to confirm graphitization during heat treatment and also to estimate the size of the resulting graphitic carbon clusters, with DFT playing a supporting role for the identification of minor peaks [12]. Afterwards, lactose and common tannins or tannin building blocks were also investigated in-situ in order to clarify their pyrolysis products and pyrolysis temperature [13]. Again, DFT-calculated Raman spectra were used for substance identification, in this case for the aforementioned products. On the purely theoretical side of things, the combination of self-interaction correction models with solvation was studied in order to possibly gain insights into chemical reactions taking place before the curing step [14]. Furthermore, a reaction mechanism including transition states was proposed for the pyrolysis of gallic acid, an important component of gallotannins [15].

In addition to this work focused on carbon-bonded filters, subproject A04 contributed to a publication investigating the synthesis of magnesium aluminum oxynitride (MgAlON), which has been suggested as a coating material for ceramic filters applied to magnesium, aluminum and steel melts due its oxidation resistance and wetting behavior [16]. However, this publication is not discussed in detail within this chapter.

In the following, some of the computational details for DFT and the experimental details for Raman spectroscopy as applied in this subproject are briefly mentioned, followed by our major findings and a conclusion.

5.2 Methods

5.2.1 Density Functional Theory

For the density functional theory calculations, a variety of open-source codes were employed to calculate the electronic and vibrational properties of investigated systems. In order to treat the solid state species, two codes were utilized, namely the Wien2K code [17], an implementation of the linear augmented plane wave (LAPW [18, 19]) method, and the plane-wave Quantum ESPRESSO code [20,21,22], which was combined with projector augmented wave (PAW [23]) ab-initio pseudopotentials. In both cases, the calculations were performed with the help of generalized gradient approximation exchange–correlation functionals, primarily the functional by Perdew, Burke, and Ernzerhof (PBE [24]). For molecular systems, the PySCF [25, 26], PyFLOSIC [27], ERKALE [28,29,30,31], and ORCA [32, 33] codes were used in combination with a multitude of functionals and Gaussian-type orbital basis sets.

5.2.2 Raman Spectroscopy

For Raman measurements, two spectrometers of the company Horiba Jobin Yvon were used, one of which was a HR 800 spectrometer with a frequency-doubled Nd:YAG laser (532 nm), a He–Ne laser (633 nm), and a diode laser (785 nm) as excitation sources. The second spectrometer was a HR 800 UV spectrometer with a He-Cd laser (325 nm and 442 nm). Both spectrometers were equipped with a Peltier-cooled CCD detector.

All measurements were performed in backscattering geometry, i.e., the incident and scattered laser beams were focused and collected by the same objective. For this, × 40 and × 50 objectives were used.

The temperature dependent in-situ measurements were carried out utilizing a Linkam TS 1200 heating chamber. The heating chamber was filled with argon to ensure an oxygen-free atmosphere.

5.3 Results and Discussion

5.3.1 Thermodynamic Properties of Intermetallics

This contribution focused on binary Al–Fe and ternary Al–Fe-Si intermetallic phases. Binary phases under investigation included η-AlFe [5], ε-AlFe [2], and η’-AlFe [34], whereas τ-AlFeSi was a ternary phase of interest [1].

DFT in combination with the quasi-harmonic approximation [35] was applied to binary and ternary intermetallic phases to study their thermodynamic properties, with a focus on calculating accurate heat capacities. Here, Quantum ESPRESSO was used in combination with PAW pseudopotentials and PBE.

5.3.1.1 Binary Phases: η-AlFe, ε-AlFe, and Η’-AlFe

The isobaric heat capacities of η-AlFe (Al5Fe2) and ε-AlFe (Al8Fe5) were evaluated in [5] and [2], respectively. Moreover, the crystal structure of η’-AlFe (Al8Fe3) was studied in [34]. Up to temperatures of 460 K, the DFT-calculated heat capacity of η-AlFe was favorably compared to experimental data taken from Chi et al. [36] and measured via differential scanning calorimetry (DSC) [5]. Here, it is important to stress that DFT-calculated heat capacities are not dependent on any free parameters to be chosen by the user or to be fitted to experiment. In this sense, they are properties evaluated from first principles. As an example for the match, a comparison between the data by Chi et al. and DFT is presented in Fig. 5.1 for temperatures up to 270 K. For higher temperatures, the DFT heat capacity was found to be lower than experiment. This was explained by the DFT structure corresponding to the ordered η-AlFe phase, whereas the disordered phase dominates for higher temperatures and was thus measured in experiment [5]. Nevertheless, the DFT results were qualitatively superior to those yielded by the Neumann–Kopp rule, which predicted a nonphysical local maximum of the heat capacity just below 1000 K.

Fig. 5.1
A line graph of C p versus temperature in kelvin for experiment chi e t a l and D F T Q H A. An increasing trend is illustrated.

DFT-calculated isobaric heat capacity of η-AlFe (Al5Fe2) [5], compared to experimental data by Chi et al. [36]

Full size image

For ε-AlFe, [2] showed agreement between the DFT heat capacity and the heat capacity as predicted by the code of T. Zienert, in contrast to the Neumann–Kopp rule. In this case, only a very small temperature range was accessible to DSC measurements, and these measurements indicated a phase transition above 1250 K. In [34], the η’-AlFe equilibrium crystal structure as predicted by DFT came very close to the data gathered from X-ray diffraction and subsequent Rietveld refinement.

5.3.1.2 Ternary Phase: τ-AlFeSi

The isobaric heat capacity of τ-AlFeSi (Al3FeSi2) was investigated in [1], and the result is shown in Fig. 5.2. According to Fig. 5.2, DFT is accurate with respect to DSC experimental values previously presented in [1] in the measured range of 900–1050 K, as opposed to the Neumann–Kopp rule. Instead, the Neumann–Kopp rule predicts a local maximum of the heat capacity, as was the case for the binary intermetallic phases, and it only somewhat approximates the DFT values in the range of 300–600 K.

Fig. 5.2
A multi line graph of C p versus temperature in kelvin for experiment D S A, neumann kopp and D F T Q H A. An increasing trend is plotted for all the lines. A zoomed in version of a part of the graph is also highlighted.

DFT-calculated isobaric heat capacity of τ-AlFeSi (Al3FeSi2), compared to experimental data from DSC measurements [1] and the Neumann–Kopp rule. The inset increases the visibility of the mean values and error bars of the DSC measurements

Full size image

5.3.2 Interface Reactions and Interface Energies

DFT was applied to study the formation of hercynite at the α-Al2O3||Fe interface, using LAPW as implemented in the Wien2K code and the PBE exchange–correlation functional. Furthermore, the interface energies for several TiO2||α-Al2O3 and TiO2||MgTiO3 interfaces were evaluated using PAW pseudopotentials in combination with the Quantum ESPRESSO code and, again, the PBE functional.

5.3.2.1 Interface Reactions: The Formation of Hercynite

In [6], several model compounds that play a role in the interaction of an alumina-based filter with a steel melt and the resulting formation of hercynite (Al2FeO4) were examined: α-Al2O3, FeO, Al2FeO4, AlFe2O4, AlFeO3, AlFe, and Fe (bcc). From the total energies of these systems, reaction energies were calculated, which suggested that the hercynite formation is not based on a direct reaction between α-Al2O3 and Fe, but between α-Al2O3 and FeO, instead, as shown in Table 5.1, which is inspired by [6].

Table 5.1 Changes in total energy ΔE for several chemical reactions suspected of occurring at the interface of alumina filter and steel melt [6]. ΔE is given in eV
Full size table

5.3.2.2 Interface Energies: Rutile Coatings and Aluminum-Containing Melts

If rutile-coated alumina filters come into contact with aluminum melts, a TiO2||α-Al2O3 interface is established at the solid/liquid phase boundary [7]. Substituting the aluminum melt for the aluminum alloy AlSi7Mg0.6, however, leads to the generation of a TiO2||MgTiO3 interface [7]. For certain orientations, both the TiO2||α-Al2O3 and the TiO2||MgTiO3 interfaces were calculated to be energetically favorable, i.e., negative, as presented in [7]. Specifically, the most negative interface energies were found for those interfaces which were also observed in the experiments described in [7]. The negative interface energies were assumed to contribute to the satisfactory adhesion of α-Al2O3 and MgTiO3 to rutile coatings [7].

5.3.3 Raman Spectroscopic Characterization of Carbon-Bonded Filters

In the publications reviewed in this part of the chapter, carbon-bonded alumina filters as well as their binders were investigated with Raman spectroscopy.

First, the results of measurements on the pure binders Carbores®P, lactose, and tannin are presented [12]. Afterwards, the results of measurements on the filters are shown [9, 10]. In addition, Raman spectroscopic investigations were carried out on other carbon-bonded materials, and changes in the microstructure or the formation of graphitic structures during mechanical tests at 1500 °C could be demonstrated [37, 38]. However, these results are not discussed in detail here.

5.3.3.1 Binder Materials

Carbores®P is a product made from coal tar pitch developed and manufactured by the company RÜTGERS. It was developed as an alternative to the classic binders for refractory materials such as phenolic resins and simple pitches, which contain a high amount of phenol or carcinogenic polycyclic aromatic hydrocarbons. With the development of Carbores®P, the amount of harmful substances could be reduced drastically, but it is still above the lawful limit in the European Union at the time of writing [10].

Lactose (sum formula: C12H22O11), informally called milk sugar, is one of the most well-known carbohydrates [39]. Lactose consists of the two monosaccharides galactose and glucose, which are connected by a glycosidic bond. In combination with tannins, lactose showed great promise as a binder ingredient for carbon-bonded magnesia refractories [40].

Tannins is the collective term for a group of polyphenolic biomolecules which can be found in the wood, bark, and fruit of certain plants. Tannins are often distinguished into hydrolyzable and condensed tannins [41]. Our focus in this subproject is on hydrolyzable tannins, which themselves can be divided into gallotannins, like tannic acid, and ellagitannins. While gallotannins contain gallic acid as their central building block, ellagitannins have an ellagic acid molecule as a common feature [41]. Gallic acid (GA, sum formula: C7H6O5) can be described as a benzene ring with three neighboring hydroxyl groups opposite a carboxyl group [42]. GA pyrolysis leads to the formation of pyrogallol and carbon dioxide and occurs at 175–200 °C [43]. Tannic acid (TA, nominal sum formula: C76H52O46) is a polyphenol, composed of a glucose ring that is linked to five m-digallic acid units via ester bonds. Digallic acids and especially GA are major products of TA pyrolysis [44]. TA decomposes starting at temperatures around 190 °C [45]. Notably, the sum formula for TA is mostly nominal, as TA vendors usually offer substances that are actually mixtures of various polygalloyl-glucose compounds [44]. Ellagic acid (EA, sum formula: C14H6O8) is another polyphenol, consisting of two GA molecules esterified with each other and connected by an additional C–C bond. Among the pyrolysis products of EA is 2,2',3,3',4,4'-biphenylhexol, a compound that is equivalent to two pyrogallol molecules linked by the aforementioned C–C bond [46,47,48]. EA was reported to melt and decomposes in the range of 350–360 °C [49, 50].

5.3.3.2 Carbon-Bonded Filters

The investigated carbon-bonded alumina filters were produced from dry slip with a total solid content of 78 wt.%. The composition of the slip is presented in Table 5.2. The main part of the raw materials is alumina (Martoxid® MR70) at 66 wt.%, while the binders, Carbores®P, lactose, and tannin, make up 20 wt.% in total.

Table 5.2 Dry slip composition for carbon-bonded alumina filters
Full size table

For different samples, different ratios of the binders were applied. This ratio ranged from Carbores®P only to lactose/tannin only, with the lactose-tannin ratio always being 5 to 1. Besides functioning as binders, these substances also served as carbon sources. Carbon black and graphite were further carbon sources. The additives TiO2 and Al work as antioxidants [51], SiO2 and Contraspum® K 1012 work as antifoam agents, Castament® VP 95 L works as a dispersing agent, and ammonium ligninsulfonate functions as a temporary binder and wetting agent. In order to obtain a higher carbon yield, n-Si was added to some of the samples [52].

5.3.3.3 Investigation of the Binder Materials

The investigated samples of Carbores®P, lactose, and tannin were fired under reducing atmosphere generated by using a coke bed [12]. The normalized Raman spectra of Carbores®P for annealing temperatures from room temperature (RT) to 1400 °C are shown in Fig. 5.3a. The so-called D peaks at ca. 1350 cm−1 and the G peaks at 1600 cm−1 are clearly visible on the left side of the figure. The G peak originates from vibrations of sp2-hybridized carbon atoms, as found in graphite, while the D peak is exclusively seen in disordered carbon systems.

Fig. 5.3
Two line graphs. A. A line graph of Raman intensity versus Raman shift for different temperatures ranging from 100 degrees to 1400 degrees Celsius. A fluctuating trend is illustrated. B. A line graph of Raman intensity versus Raman shift for R T and 1400 degrees Celsius. Two big spikes for the 1400-degree Celsius line are labeled D and G.

a Raman spectra of Carbores®P [12]. The Carbores®P samples were heated up to 1400 °C under reducing atmosphere, with RT representing room temperature. The right part of the subfigure shows the spectral region where -OH vibrations occur, b Direct comparison of the spectra of Carbores®P annealed to RT and 1400 °C. The D peaks shows a clear narrowing and an increase with temperature compared to the G peak. The arrows indicating the FWHM are only guides for the eyes and not the actual fit parameters

Full size image

In the spectra from RT to 600 °C, the D peak exhibits shoulders at 1160, 1240, and 1440 cm−1 originating from trans-polyacetylene, a chain-like hydrocarbon with alternating single and double bonds [12]. In Fig. 5.3a, these shoulders are marked with arrows. The disappearance of the shoulders at higher temperatures indicates the pyrolysis of trans-polyacetylene.

Moreover, the spectra indicate that the OH groups split off at ca. 400 °C, since the associated band with a maximum at 3350 cm−1 disappears above this temperature (see the right part of Fig. 5.3a). From the background slope of the spectra, the hydrogen fraction of the system could be estimated [53]. According to this estimation, the hydrogen fraction decreased from more than 40% at RT to less than 20% at 800 °C. Figure 5.3b shows the D peak becoming narrower and more intense with increasing annealing temperature, indicating an increase in order for the carbon system. Thus, the investigated Carbores®P samples changed chemically and structurally with increasing annealing temperature.

Further information about these changes is provided by the intensity ratios of the D and G peaks ID/IG (see the upper part of Fig. 5.4) as well as the position of the G peak (given in [12]) as a function of annealing temperature. To determine these parameters, the D and G peaks were fitted with a Lorentz and a Breit–Wigner–Fano function, respectively. An example for the fit process is shown in the inset of Fig. 5.4 for a Carbores®P sample fired at 600 °C. The spectra of lactose and tannin were treated in a similar manner.

Fig. 5.4
A whisker plot of L d slash L a and cluster size versus temperature in Celsius. The whisker plot on the top and bottom plots Carbores P, lactose, and tannin values. In the bottom graph, the amorphous carbon, transition region, and nanocrystalline graphite regions are also labeled.

The intensity ratios of the D and G peaks as a function of annealing temperature. An example fit of the D and G peaks is shown in the top left picture for a Carbores®P sample fired at 600 °C. In the lower part, the calculated cluster sizes for Carbores®P, tannin, and lactose are shown. Between 800 °C and 1000 °C, a conversion of amorphous, hydrogen-rich carbon to nanocrystalline graphite occurs [12]

Full size image

For temperatures up to 600 °C, there is hardly any variation in the intensity ratios and peak positions, indicating that the binder systems undergo minimal chemical and structural changes. In the range of 600–1000 °C, the intensity ratio ID/IG increases, and the position of the G peak is shifted to higher wavenumbers, indicating a transformation from hydrogen-rich amorphous carbon to nanocrystalline graphite [12]. This is associated with a decrease in the content of sp3−hybridized carbon, hydrogen, and chain-like carbon compounds such as trans-polyacetylene.

During this process, the size of graphitic sp2 carbon ring clusters increased [54]. With the Tuinstra model, these cluster sizes could be estimated from the ID/IG ratios [55, 56]. The lower part of Fig. 5.4 shows the results; a significant increase in cluster size with annealing temperature can be seen [12]. These values are in excellent agreement with the XRD results determined by subproject A05 (compare Chap. 6, Fig. 6.4a).

5.3.3.4 Investigation of Filters with Carbores®P Binder

In the work of Röder et al., carbon-bonded alumina filter with Carbores®P binder fired at temperatures in the range of 800–1600 °C were studied with Raman spectroscopy [9]. The samples exhibited a granularly structured surface. It was possible to distinguish between flakes, which were identified as graphite, and a surrounding matrix part in the sample. The presence of graphite in the sample can be explained by graphite having been added as a raw material (compare Table 5.2). The Raman spectra measured at the matrix part were assigned to nanocrystalline carbon. Additionally, it was shown that the sp2 carbon clusters in the matrix part grew from around 2.3 nm at 800 °C to roughly 4.4 nm at 1600 °C [9].

5.3.3.5 Investigation of Filters with Lactose/Tannin Binder and the Influence of Added n-Si

Filters with environmentally friendly binders based on lactose/tannin were investigated using Raman spectroscopy for the first time in [10]. Figure 5.5 (left side) shows Raman spectroscopic measurements at RT for Al2O3/lactose/tannin samples that were tempered at 200, 300, 400, and 500 °C. Apparently, the photoluminescence decreases with increasing annealing temperature and the aforementioned ID/IG ratio, which can also be described as the ratio of the sp2/sp3 character of the carbon compounds, increases. As explained earlier, this increase in the ID/IG ratio is likely associated with a growth of the sp2 carbon clusters. A possible mechanism could be the release of some of the OH groups from tannin due to thermal energy, so that the smaller molecular fragments cross-link and form larger aromatic sp2 carbon clusters. This behavior would agree with the temperature development of the photoluminescence, which indicates a loss of hydrogen in the system [53]. In support of this, Raman spectra presented in [10] show a clear decrease in intensity for the sample that was tempered at 500 °C in the range of -OH vibrations at ca. 3400 cm−1. All the findings detailed here are in excellent agreement with previously reported research performed on the pure binders [12].

Fig. 5.5
2 line graphs. Left. A line graph of intensity versus Raman shift plots lines for 4 different temperatures for A L 2 O 3 plus lactose plus tannin. Right. A line graph of intensity versus Raman shift plots lines for 4 different temperatures for n s i. Both graphs plot an increasing trend.

Raman spectra of Al2O3/lactose/tannin samples as a function of annealing temperature [10]. Raman spectra of the same samples with added n-Si (right)

Full size image

In the right part of Fig. 5.5, the Raman spectra of the samples after addition of n-doped Si are shown. As can be seen, the photoluminescence intensity is lower for the samples mixed with n-Si, except for the sample annealed at 500 °C. Moreover, an increase of the ID/IG ratio with annealing temperature is observed in these spectra. Such an increase was found by Ferrari et al. to be ‘proportional to the number and clustering of rings’, indicating a higher degree of order in the amorphous carbon system [56].

For a better visualization of the influence of n-Si addition, the Raman spectra, normalized for the same intensity of the D and G peaks, of the coked samples (800 °C) with and without n-Si are shown in Fig. 5.6. In addition to the overall lower intensity, the addition of n-Si clearly modified the carbon bonding. One can see (marked by the red arrows) that n-Si caused a reduction of the intensity of the bands at ca. 1150 cm−1 and a slight increase of the band at ca. 1460 cm−1. These modes were assigned, respectively, to C–C vibrations and C = C vibrations in sp2 carbon chains in trans-polyacetylene [57, 58].

Fig. 5.6
A line graph of intensity versus Raman shift for A L 2 O 3 slash lactose slash tannin at 800 degrees Celsius. A fluctuating trend is illustrated.

Raman spectrum of a Al2O3/lactose/tannin sample coked at 800 °C (solid line) and the Raman spectrum for the sample with n-Si (dotted line + symbols) [10]. The added n-Si leads to a lower intensity of the band at ca. 1150 cm−1 and a higher one for the 1460 cm−1 band, here indicated by the red arrows

Full size image

In order to achieve a better understanding of the temperature dependence of the chemical processes that take place in the filters with environmentally friendly binders, real-time in-situ Raman measurements were performed as well. The results of these in-situ Raman measurements are discussed in the following.

5.3.4 Raman Spectroscopic In-Situ Study of the Pyrolysis of Tannins

Temperature dependent in-situ Raman measurements of selected representatives of tannins, i.e., GA, TA, and EA were performed to understand their pyrolysis [13]. In Fig. 5.7, spectra of GA and TA as well as their chemical structures are displayed. The investigated samples were placed in a crucible in a heating chamber under argon atmosphere to ensure the absence of oxygen, which avoids a possible oxidation of the sample and thereby enables pyrolysis. The used heating chamber is shown in the inset of Fig. 5.8.

Fig. 5.7
A line graph of Raman intensity versus Raman shift for pyrogallol, gallic acid, and tannic acid. All three lines plot a fluctuating trend with a few spikes. Each individual compound's molecular structure is illustrated above the lines.

Raman spectra of pyrogallol, GA, and TA measured at RT as well as their Lewis structural formulas [13]

Full size image
Fig. 5.8
A line graph of Raman intensity versus wavenumber plots the Raman spectra of GA at different temperatures ranging from R T to 700 degrees Celsius. A fluctuating trend is plotted for all lines. A photograph of a device is also shown on the top right corner of the graph.

In-situ Raman spectra of GA at different temperatures [13]. The used heating chamber is shown as an inset in the upper right corner

Full size image

Besides, the Raman spectra of GA from RT up to 700 °C are shown in Fig. 5.8. Except for a broadening of the peaks and a slight shift to lower wavenumbers due to temperature effects, no changes can be observed in the spectra up to 200 °C.

In the temperature range of 250–350 °C, a white substance condensed at the window of the heating chamber. Since the laser light could not penetrate the condensate and reach the actual sample, measurements at these temperatures were not possible. Instead, the focus of the laser beam was set to the condensate at the heating chamber window. The spectrum recorded for the condensate could be assigned to pyrogallol, whose spectrum and structure are shown in Fig. 5.7.

As previously mentioned in this report, pyrogallol is well known as a pyrolysis product of GA [42, 43]. It is formed by the elimination of CO2 from the GA molecule. Pyrogallol could also be identified as a pyrolysis product of TA. For EA, 2,2’,3,3’,4,4’-biphenylhexol has been reported as a pyrolysis product in literature [46,47,48], which is also the assumption made in this in-situ study. The pyrolysis product identification was supported by DFT calculations. Moreover, pyrolysis temperature ranges based on Raman measurements could be estimated for GA (225–350 °C), TA (200–325 °C), and EA (150–250 °C).

When the tannins were heated further, they exhibited the Raman spectra of amorphous carbon systems. For GA, these spectra are also shown in Fig. 5.8 for temperatures of 400 °C and above. The spectra contain the characteristic D and G peaks at around 1350 and 1600 cm−1, respectively.

5.3.5 Fermi-Löwdin Orbital Self-Interaction Correction and Solvation

The Fermi-Löwdin orbital self-interaction correction (FLO-SIC [59,60,61,62]), which aims to correct prominent failings of DFT, was implemented in the form of PyFLOSIC [27], an extension to the PySCF code. The development of PyFLOSIC made FLO-SIC calculations with a wide variety of basis sets, functionals, and numerical integration grids possible, thus simplifying comparisons to other quantum chemical codes. One such comparison before the development of PyFLOSIC was presented in [63]. The PyFLOSIC code is open-source and freely available on GitHub (https://github.com/pyflosic/pyflosic).

This code, in addition to PySCF and ERKALE, was applied to the AQUA20 test set in order to calculate ionization potentials (IPs) and room-temperature standard enthalpies of formation (ΔfH°(298.15 K)) for the gas phase and an aqueous solution as simulated with the ddCOSMO [64,65,66,67,68,69,70] continuum solvation model [14]. The AQUA20 test set includes carboxylic acids and carboxylate anions, systems that share functional groups with the gallotannins making up part of the environmentally friendly binders for carbon-bonded filters. Moreover, the aqueous solution case was investigated in addition to the gas phase because the binders take the form of a slurry before the curing step during filter production, and the tannins themselves are naturally formed in aqueous polymerization reactions [14].

In [14], the FLO-SIC results were compared to quantum chemical methods, uncorrected DFT using several different functionals, and real-valued self-interaction correction as implemented in the ERKALE code, the latter of which is called RSIC. The FLO-SIC and RSIC values were found to be in good agreement with each other across the board, and gas phase trends previously reported in literature were reproduced. For the aqueous solution, the mean errors (MEs) and mean absolute errors (MAEs) when compared to experiment are presented in Fig. 5.9 for the IPs [14].

Fig. 5.9
Two bar graphs. Top. A bar graph of M E versus 7 different quantum mechanical methods plots the A Q U A 20 test set and the A Q U A 20 neutral subset. Bottom. A bar graph of M A E versus 7 different quantum mechanical methods plots the A Q U A 20 test set and the A Q U A 20 neutral subset. The highest bars are plotted for C O S M O F L O S I C.

The MEs and MAEs for IPaq in eV, calculated using 7 different quantum mechanical methods combined with COSMO solvation, and applied to the AQUA20 test set [14]

Full size image

According to Fig. 5.9 and the data presented in [14], the DFT approaches COSMO-SCAN and COSMO-B3LYP achieve the closest overall agreement to experiment for the AQUA20 test set, along with COSMO-CCSD(T). For the aqueous standard enthalpy of formation, it was shown that most of the error is caused by the charged species in the test set [14]. Moreover, COSMO-FLO-SIC and COSMO-RSIC improve on the results of the underlying functional, i.e., LDA, for Δf,aqH°(298.15 K), however, the deviations from experiment actually increase when applying self-interaction correction to aqueous IPs (see Fig. 5.9).

5.3.6 Theoretical Investigation into Gallic Acid Pyrolysis

Thermodynamic and kinetic data on the first two steps of GA pyrolysis, i.e., a decarboxylation followed by a dehydrogenation, were obtained with the help of DFT and quantum chemistry [15], using the ORCA code. The investigated reactions can be written as GA → PG + CO2 and PG → OQ + H2, with PG representing pyrogallol and OQ representing 3-hydroxy-o-benzoquinone.

For the kinetics, transition states were identified with the help of the climbing image nudged elastic band method (CI-NEB) with subsequent transition state optimization as implemented in ORCA [71,72,73,74], employing the PBE functional. Both reactions were found to exhibit two transition states. One of these transition states is related to the rotation of OH groups, and the other one is related to the breaking and forming of bonds. The results of the CI-NEB calculation for the first GA pyrolysis reaction [15] are shown in Fig. 5.10, with Fig. 5.10a displaying the converged minimum energy path and Fig. 5.10b displaying selected images, including the optimized highest-energy transition state.

Fig. 5.10
Two illustrations. A. A line graph of energy versus displacement. The line first increases, then decreases after reaching the third point. B. 4 compounds are highlighted each for the points plotted in graph a.

a Converged CI-NEB minimum energy path of the first reaction GA → PG + CO2. The x axis shows the cumulative displacement between subsequent images, which is identified as the reaction coordinate. The dots represent the PBE energies of the images relative to the first image, and the line is an interpolating spline. Code for selected images: I–GA, II–rotational transition state, III–converged CI, IV–PG/CO2. This subfigure was created with the help of a Python script by Ásgeirsson [75], b Selected images, with III† representing the converged transition state for which III acted as a starting point. Color code: white–H atoms, black–C atoms, red–O atoms. The solid lines represent fully formed bonds, the dotted lines represent bonds in the process of being broken or formed. This figure was previously published in [15] and is reused with permission from the publisher

Full size image

After applying several DFT functionals and wavefunction methods in order to calculate standard enthalpies of reaction and standard Gibbs energies of reaction, the combination of the first and second pyrolysis reactions was judged to be endothermal, and it is predicted to change from endergonic to exergonic between 500 and 750 °C. The second reaction, the dehydrogenation of PG, was identified as the rate-determining step of GA pyrolysis, with reaction rate constants below 1 s−1 for temperatures below 1250 K [15].

5.3.7 Thermally Induced Formation of Boehmite

Raman spectroscopy was used to understand the structural changes that occur during the thermally induced transformation of boehmite (γ-AlO(OH)) via the metastable transition phases γ-Al2O3, δ-Al2O3, and θ-Al2O3 to corundum (α-Al2O3) [8]. Figures 5.11 and 5.12 show characteristic Raman spectra of samples at RT that were annealed in air for 20 h at different temperatures. Figure 5.11 shows the Al-O vibrations range, while Fig. 5.12 corresponds to the –OH vibrations measured at higher Raman shifts.

Fig. 5.11
An X R D graph of intensity versus wavenumber plots : Raman spectra of boehmite, gamma A L 2 O 3, delta A L 2 O 3, and alpha A L 2 O 3. A fluctuating trend is observed.

Raman spectra of boehmite (γ-AlO(OH)) in the Al-O vibration range as a function of annealing temperature. They show the transition from boehmite to corundum (α-Al2O3) with the formation of metastable transition phases γ-Al2O3, δ-Al2O3, and θ-Al2O3

Full size image
Fig. 5.12
A line graph of intensity versus Raman shift for different temperatures ranging from R T to 900 degrees Celsius. An increasing trend is illustrated for R T, 300, and 400 degrees Celsius.

Raman spectra of spectra of boehmite (γ-AlO(OH)) in the -OH vibration range as a function of annealing temperature

Full size image

Temperature Range RT – 400 °C

Up to a temperature of 400 °C, the Raman modes marked by triangles in Fig. 5.11 were detected for boehmite at 229 cm−1, 257 cm−1 (weak), 270 cm−1 (weak), 341 cm−1, 363 cm−1, 451 cm−1, 500 cm−1, 638 cm−1 (weak), 676 cm−1, and 732 cm−1, indicating that the boehmite structure is stable up to this temperature. The positions of the peaks agree with Doss et al. [76]. In the higher wavenumbers region (Fig. 5.12), the modes at 3083 cm−1 and 3227 cm−1 can be assigned to the -OH stretching vibration within the boehmite structure [77].

Temperature Range 450–650 °C

The Raman spectra measured for γ-Al2O3 show no peaks in the low wavenumber range (Fig. 5.11). The absence of Raman activity was explained in literature [78] by the distribution of the cation vacancies preferentially to the tetrahedral sites and of the Al3+ ions preferentially to the octahedrally coordinated sites in the regular spinel structure. In the higher wavenumbers region (Fig. 5.12), the samples annealed at 450 and 500 °C exhibit a large broad band at ca 3500 cm−1, which corresponds to hydrogen bonded -OH stretching vibrations, and this band is assigned to adsorbed water. A possible reason could be water resorption, as the samples were examined ex-situ. A weak band at ca. 3760 cm−1 can be attributed to a terminal -OH stretching vibration in Al–OH (isolated surface hydroxyl groups) [79]. Above 600 °C, the peaks corresponding to water vibrations (H-OH or isolated -OH) are drastically reduced in intensity for all spectra (Fig. 5.12).

Temperature Range 700–1400 °C

At 700 °C, modes of a new phase (peaks marked by + in Fig. 5.11) become visible in the Raman spectra. The positions of the peaks are in agreement with the resonance Raman data for δ-Al2O3 [80] and θ-Al2O3 [78, 81] as measured with UV excitation. In literature, there is no clear distinction between the Raman bands from δ-Al2O3 and θ-Al2O3 phases [82]. At 1000 °C, the α-Al2O3 starts to form. Above this temperature, the seven expected Raman modes (marked by * in Fig. 5.11) of corundum [83] are detected at: 378 cm−1 (Eg), 418 cm−1 (A1g), 431 cm−1 (Eg), 449 cm−1 (Eg), 577 cm−1 (Eg), 644 cm−1 (A1g), 751 cm−1 (Eg).

In addition to the results recovered by Raman spectroscopy, DFT calculations using PAW pseudopotentials, Quantum ESPRESSO and the PBE functional were carried out as well [8]. The calculated enthalpies suggested that a negative pressure would lead to the γ-Al2O3 phase being more stable than both θ-Al2O3 and α-Al2O3, effectively decelerating the phase transition from boehmite to corundum.

5.4 Conclusion

This summary gave an insight how density functional theory calculations and Raman spectroscopic measurements were used to investigate filter materials and thereby can help to improve them. It could be shown that DFT and Raman spectroscopy are valuable additions to traditional methods in materials science.

Over more than a decade, DFT was used to calculate and study a wide range of physical properties for different systems. For example, the heat capacity of the systems η-AlFe, ε-AlFe, and τ-AlFeSi were calculated. The results showed a better agreement with experimental data compared to heat capacities calculated by the Neumann–Kopp rule.

Furthermore, the reaction energies of the formation of hercynite (Al2FeO4) at the interface during the filtration of a steel melt with an Al2O3 based filter were calculated by DFT. The results showed that hercynite is probably not a product of the reaction of α-Al2O3 with Fe, but rather with FeO. Another property determined by DFT were the interface energies for rutile (TiO2) coatings in aluminum melts and AlSi7Mg0.6. For the interfaces TiO2|| α-Al2O3 and TiO2||MgTiO3, the most negative energies were calculated for the orientations observed in experiment.

A major research field of the subproject A04 was the investigation of carbon-bonded alumina filters. In several publications, these filters and the binders Carbores®P, lactose, and tannin were studied with Raman spectroscopy. Raman spectra for different binder compositions and different coking temperatures were recorded and analyzed. Especially the D and G peaks typical for Raman spectra from samples containing sp2-hybridized carbon were interpreted.

For the pure binders, a decrease of the hydrogen content with increasing coking temperature and an increase of the sp2 carbon clusters at temperatures above 600 °C were observed. Similar results were found for filters with Carbores®P binder, filters with environmentally friendly lactose/ tannin binder, and filters with a mixture of both. For the filters with added n-Si, the Raman spectra showed a modification of the carbon bonding.

For a better understanding of tannin pyrolysis, temperature dependent in-situ Raman measurement under argon atmosphere were performed. The tannin representatives gallic acid, ellagic acid, and tannic acid were studied this way. Raman spectra of the pyrolysis intermediates and end products of the examined substances were recorded, and pyrolysis temperature ranges were determined. Furthermore, the pyrolysis, i.e., the decarboxylation and dehydrogenation, of gallic acid was investigated with DFT, identifying transition states for both reactions.

Besides investigating carbon-bonded filters, the formation of corundum (α-Al2O3) from boehmite (γ-AlO(OH)) via γ-Al2O3, δ-Al2O3, and θ-Al2O3 was monitored in a temperature range from room temperature up to 1400 °C with Raman spectroscopy. Each phase showed a clearly distinguishable spectrum, so an identification of these phases as well as their corresponding temperature ranges was possible.