Converting the residue from an infiltration water treatment plant (Wrocław, Poland) into a hematite red pigment—optimising the process with a Doehlert experimental matrix

Abstract

Referring to the concept of a circular economy and using optimal experimental design (OED), the research deals with optimising the conversion of industrial residue from the water de-ironing process into a red hematite pigment. The raw, chemically pretreated sludge was thermally converted into hematite under various conditions established in the optimised procedure. The products were thoroughly characterised using XRD, Raman spectroscopy, Diffuse reflectance spectroscopy, and SEM EDS to gain better insight into the process. Because of the unambiguous and quantitative definition of colour and its individual perception, resulting in difficulty in determining the appropriate system response in the optimisation procedure, products’ quantitative colour parameters, determined in the CIELab colour space, were compared to those obtained for the commercial red hematite pigment, and the most suitable ones (a*, C*, and L*) were chosen as system responses in the optimising procedure. As a result, the optimal ranges of the process parameters—temperature and holding time—that can ensure a pigment with the desired colour were established by OED methodology. Moreover, it was found that the key factor enabling effective conversion of the studied residue into red pigment is a proper chemical pretreatment in which the process temperature is decreased to below the hematite sintering temperature.

Supplementary Information

The online version contains supplementary material available at 10.1007/s11356-025-37275-6.

Introduction

An increasingly serious approach to the principles of the circular economy, in connection with the current issues in waste and raw material management and the gradual depletion of natural minerals, has intensified the research on converting existing waste streams into value-added products, emphasising process optimisation. Using industrial residues offers the possibility of a more rational waste management in various industries, in addition to reducing the consumption of natural resources and production costs.

One of the concepts to which much attention has been given is the conversion of iron-rich residues into red, hematite-based pigment. Every year, vast quantities of iron oxide pigments are used to produce ceramics, paint, and building materials, making this idea potentially attractive from a commercial perspective (Khanam et al. 2023). However, optimisation of the process using advanced statistical methods is an issue when it comes to pigment production, mainly because the ambiguous, subjective definition of colour creates difficulty in determining the appropriate product parameter to be used as a system response in the optimisation procedure. Consequently, few studies have been published that used advanced optimisation methods (Bogacki et al. 2004; Colpas-Ruiz et al. 2020). However, these methods are commonly used in the extraction processes of natural dyes, where process efficiency is the sole and sufficient system response, making the problem of a quantitative definition of colour irrelevant (Yolmeh et al. 2014; Prabhu et al. 2023; Rao et al. 2025). However, it should be emphasised that several systems exist that enable the quantitative determination of colour parameters in different colour spaces (CIELab, Mansell) (Broadbent 2010), making optimisation procedures possible, but still complex and requiring individual development and refinement.

Simultaneously, most of the research on synthesising Fe-based pigments from waste streams is focused on residues mainly from ceramic and steel industries, because of the beneficial high iron content and low level of contaminants (Das et al. 2007; Legodi and de Waal 2007; Prim et al. 2011; Colpas-Ruiz et al. 2020; Rahman et al. 2020; Dias et al. 2021; Khanam et al. 2023; Peng et al. 2024). Very few studies refer to the by-products generated during water or wastewater treatment, and others with higher organic content (Fiuza et al. 2018; Lu et al. 2020). However, what should be emphasised is that the residue from the infiltration water deironing process, less commonly used in water treatment plants, is composed mainly of iron oxyhydroxides, with only very small amounts of organic contaminants, making it a potentially beneficial raw material for pigment production (Jacukowicz-Sobala et al. 2015). What also distinguishes this kind of residue is its form—a water sludge composed mainly of iron oxyhydroxide microparticles, which are expected to be readily transformed into hematite red pigment. Another essential advantage of infiltration water is its more stable composition throughout the year, which results in a by-product with reproducible properties that are essential for potential commercial applications.

The nature of the water de-ironing process causes the main compounds of the by-product to be iron oxyhydroxides and iron hydroxides, which can be thermally converted into hematite in a simple process. In practice, however, even a low level of inorganic contaminants (Al, Mn, Si, P, and Ca) in the resulting sludge increases the conversion temperature above the hematite agglomeration and sintering temperature (Cornell and Schwertmann 2003), thus increasing the costs of this high-temperature process even more and making the resulting hematite pigment bluish or grey. Therefore, utilising such a residue for hematite pigment synthesis requires proper pretreatment to decrease the content of pollutants, in addition to optimisation of the process itself.

The main aims of the research were to (a) characterise the residue from an industrial infiltration water treatment plant (before and after pretreatment) in terms of its possible use for pigment synthesis; (b) gain insight into the mechanisms of converting waste iron oxyhydroxides into hematite; (c) determine various quantitative colour parameters of the products and identify the most appropriate one through comparison with a commercial hematite pigment; and (d) optimise the conversion process using optimal experimental design (OED) with a Doehlert matrix.

Materials and methods

Fe-containing residue

The residual sludge from the deironing/demanganisation of water was collected from the Na Grobli Water Treatment Plant in Wrocław, Poland (https://www.mpwik.wroc.pl/o-nas/wodociagi/). The raw sludge was rinsed several times with distilled water, dewatered under vacuum, dried at 30 °C to a constant mass, and stored in sealed containers for further experiments. Before drying, some of the sludge was also treated with 0.1 M HCl for 20 min (1:10 v/v), and then rinsed with distilled water to a neutral pH (Ociński et al. 2016). Such a simple acid treatment can change the raw material’s microstructure and surface properties by dissolving undesirable impurities. After sieving (0.25-mm sieve), the dried sludge (WTR) and the chemically treated sludge (CH-WTR) were characterised and thermally treated under various temperature and time conditions.

Thermal conversion of the residue

One-gram samples of WTR and CH-WTR were thermally converted into a red pigment by calcination in porcelain crucibles in a muffle furnace (PEM-2, Prodryn, Poland). The preliminary ranges of temperature (650–850 °C) and time (60–240 min)—two key variables influencing both the effectiveness and cost of converting iron oxyhydroxides into hematite (α-Fe₂O₃)—were defined by thermogravimetric analysis and preliminary studies, respectively. To find an optimal range of these variables and thus optimise the conversion process, optimal experimental design (OED) with a Doehlert matrix was used. This design has been chosen because it requires fewer experiments while ensuring better efficiency as compared to the Central Composite design (CCD) (Ferreira et al. 2004). A detailed description of this statistical method can be found elsewhere (Doehlert 1970). Each combination of temperature and time, except for the centre of the design, was duplicated, resulting in 12 experimental runs; three experiments were conducted at the centre of the experimental region, following the rules of the Doehlert experimental design (Fig. S1, Table S2). A detailed description of the design of the experiments (DoE) is presented in the Supplementary Materials. For planning experiments and calculating coefficients of polynomial models, NemrodW® software, version 2017 (Marseille, France), was used.

Analytical methods

The elemental composition of the dried sludge was determined by X-ray fluorescence using an Epsilon 3x XRF analyser (Malvern Panalytical, UK). The water treatment plant supplied the concentrations of Fe and Mn in the infiltration water before and after the de-ironing process.

The thermogravimetric experiments were carried out using a 209 F1 Libra thermobalance (Netzsch, Selb, Germany) at a temperature range of 25–1000 °C, under an air atmosphere (protective gas: nitrogen, 10 ml/min; purge gas: air, 30 ml/min). For each experiment, a sample of approximately 5 mg was loaded into an open alumina crucible (150 μL). The software program Netzsch Proteus Analysis 7.1.0 (Netzsch, Selb, Germany) was used to determine the mass change and derivative thermogravimetric (DTG) curve. The measurements for each sample were repeated at least twice.

Diffuse reflectance spectroscopy (DRS) was applied to determine the samples’ electronic spectra and colourimetric parameters. The analyses were conducted with a UV-VIS spectrophotometer equipped with a 75-mm integrating sphere (Specord 210, Analytik Jena, Jena, Germany), operated in diffuse reflectance mode. The spectra were collected in the range of 380–780 nm (resolution: 1 nm; integration time: 1 s; slit: 4 nm) with the Spectralon® white standard as the reference material. The colour parameters in the CIELab system (Commission Internationale de l'Eclairage) were determined with WinAspect software (Analytik Jena, Germany), a D65 illuminator and the standard observer 10° (Broadbent 2010). The values were compared with those determined for a commercial hematite pigment (PR102, Roman Szmal Art, Krakow, Poland).

Raman spectroscopy and X-ray powder diffraction analysis were employed to analyse the structural evolution of iron-containing phases in the WTR and CH-WTR sample series. The Raman measurements were performed using a LabRAM HR800 Raman spectrometer, equipped with an Ar-Kr laser as the excitation source. The laser’s power was set to 140 mW, ensuring a strong Raman signal while minimising potential thermal effects on the samples. Spectra were collected over a measurement range of 50–1800 cm^–1, covering both low- and high-wavenumber vibrational modes essential for identifying iron oxyhydroxides and oxides (Hanesch 2009; Cambier et al. 2014). The exposure time was set to 30 s per acquisition, optimising signal quality while preventing excessive noise. Powder X-ray diffraction patterns were conducted with a Bruker D2 PHASER diffractometer (Bruker AXS, Karlsruhe, Germany). The diffractometer, with Cu Kα1.2 radiation (1.5418 Å), was used in the angular range of 10° ≤ 2θ ≤ 60° (Δ2θ = 0.02°; t = 0.1 s/step). The X-ray tube was operated at 30 kV and 10 mA. The optical system comprised 2.5° Soller modules, a divergence slit of 0.2 mm, an anti-scatter slit of 0.5 mm, and a nickel filter. The detector was a LynxEye linear position-sensitive detector. Diffractograms were obtained in Bragg–Brentano (θ/2θ) horizontal geometry. All samples were studied in air at room temperature. The XRPD patterns were processed with the software program DIFFRAC.EVA v 3.2 (BRUKER AXS, Karlsruhe, Germany).

The morphology of the samples was analysed with a scanning electron microscope (Thermo Scientific, SCIOS 2) equipped with an energy-dispersive spectrometry (EDS) microanalyser. The hydrodynamic diameters of the samples were measured with the dynamic light scattering (DLS) technique. The analysis was performed using a Nano-ZS Zetasizer (Malvern Instruments, UK) with a detector position of 173° and the laser light set at λ = 633 nm. The measurements were conducted in a disposable translucent polyacrylic DTS-0012 cuvette (Malvern Instruments, Malvern, UK) filled with 1 ml of the sample. Each sample (1 mg) was diluted with deionised water (5 ml). The cuvette was placed in a temperature-controlled measurement cell. The test was run at 25 ± 0.1 °C, and the samples were measured at least 6 times. The results are expressed as the mean and standard deviation of three independent measurements. Malvern Zetasizer Software v7.10 was used to process the data.

Results and discussion

Source of the iron-containing sludge

The Na Grobli Water Treatment Plant is one of two main plants supplying drinking water to Wrocław (a provincial capital with 670,000 inhabitants), producing over 70,000 m³ of drinking water daily. The plant is distinguished by its technology: infiltration water is used as a feedstock, simplifying the treatment process and ensuring high-quality water. Before the primary process is carried out within the plant, the raw surface water is directed to a system of infiltration wells, where it percolates through the soil for several weeks, taking on the characteristics of groundwater. It is then directed to the water treatment plant by a ditch system and infiltration water pumping station (Fig. 1a). It is worth noting that such a system ensures a much more stable composition of the water throughout the year, which makes it possible to obtain by-product streams with reproducible properties, which is essential for potential industrial applications. Since the concentrations of Fe(II) and Mn(II) in the infiltration water exceed the permissible level, it is initially subjected to de-ironing/demanganisation before further treatment. This process is based on water aeration, leading to Fe²⁺ and Mn²⁺ oxidation, followed by filtration of the precipitated sediments on a MnO₂-coated sand filter. The concentrations of iron and manganese in the infiltration water before and after filtration (throughout the year) and the calculated iron content in the resulting residue are presented in Figs. 1b–d.

Fig. 1.

Simplified scheme of the infiltration water treatment system and photograph of the water treatment plant (a); annual changes in the concentration of Fe, Fe(II) (b) and Mn (c) in the infiltration water; and the calculated iron content in the by-product (d) (press materials and data supplied by MPWiK S.A., Na Grobli 19, 50–421 Wrocław, Poland)

Considering the conditions of the aeration process (fast oxidation and hydrolysis of Fe²⁺, neutral pH and ambient temperature), it can be assumed that amorphous hydrous ferric oxides (HFO) should account for most of the precipitate, most likely with some amount of 2-line ferrihydrite, lepidocrocite, and goethite (Cornell and Schwertmann 2003). This was further confirmed by Raman spectroscopy: the untreated samples’ spectra were dominated by FeOOH-related bands at 215–277 cm^–1 and 388–472 cm^–1, with broad, low-intensity features indicating poorly crystalline structures (Fig. 5). Both spectra exhibited characteristic peaks corresponding to goethite (α-FeOOH), ferrihydrite (Fe₅HO₈·4H₂O), and lepidocrocite (γ-FeOOH), confirming the presence of hydrated iron phases (Wang et al. 1998; Hanesch 2009). Although the composition of this by-product is more advantageous in terms of its further utilisation as an iron source (as compared to typical water treatment residues from coagulation processes), the main issue is the high content of manganese dioxide. It is well known that the presence of manganese, aluminium, calcium, and phosphates hinders the thermal transformation of amorphous or low-crystalline iron oxides into hematite, increasing the transformation temperature from 400 to 500 °C to over 800 °C (Cornell and Schwertmann 2003). The hematite particles are agglomerated and sintered at this temperature, leading the red colour to change into a more bluish, purple, or even greyish tone; additionally, the extra heating incurs higher costs. For these reasons, the raw sludge (WTR) was subjected to chemical pretreatment, substantially decreasing the manganese and calcium content and simultaneously increasing the relative iron content (Table 1). Both the raw (WTR) and chemically treated (CH-WTR) residues were used in further experiments on their thermal conversion into hematite-based pigments.

Fig. 5.

Raman spectra of the products obtained by thermal conversion of the CH-WTR sample under various conditions

Table 1.

Elemental composition of the raw (WTR) and chemically pretreated (CH-WTR) samples obtained through XRF quantitative analysis (wt.%)

	Al, %	Fe, %	Mn, %	P, %	Ca, %	Si, %
WTR	0.136	52.913	1.886	4.628	3.934	0.901
CH-WTR	0.097	60.822	0.614	4.668	0.246	0.842

Thermogravimetric studies

Thermogravimetric studies were conducted to determine the temperature range of the phase transformations of the iron oxides and oxyhydroxides into hematite in the WTR and CH-WTR samples. As shown in Fig. 2a, b, the shapes of the curves are characteristic of iron oxyhydroxides, and they show four typical steps of mass loss (TG/DTG).

Fig. 2.

TG/DTG and DTA curves of the chemically pre-treated (a) and the raw (b) residue from water deironing; the possible transformation pathways of the Fe(II) ions present in the infiltration water into hematite in the final product (c); the selected DTG peaks with assigned transitions (d)

The first loss of mass, observed below 200 °C, was related to the physically adsorbed surface water, the amount of which depends mainly on the pretreatment of the sample. This was followed by a dehydroxylation reaction (the loss of the water in the iron oxyhydroxides), which usually occurs below 280 °C (Müller et al. 2015). The partial substitution of iron with aluminium in the structure of the iron oxyhydroxides may cause a slight increase in the dehydroxylation reaction temperature, to about 300 °C (Cornell and Schwertmann 2003). Because some increase was observed for both samples (288.6 and 286.6 °C), it can be concluded that the aluminium in the sludge was structural, which influenced the thermal transformation of the samples. As seen in Table 1, the chemical pretreatment did not significantly lower the aluminium content in the sample. Its presence was also confirmed by the SEM-EDS analysis of the CH-WTR sample after thermal treatment at 750 °C (Fig. 6).

Fig. 6.

SEM images and EDS spectra of the product obtained from CH-WTR (750 °C, 2.5 h)

Special attention should be paid to the third weight loss, which was more visible for the untreated sample (WTR). This may be attributable to the phase transition of amorphous and poorly crystalline iron oxyhydroxides to maghemite, or the release of the organic carbon (Lu et al. 2020; Rahman et al. 2020). The much smaller peak observed for the CH-WTR sample may result from partially removing some impurities during its chemical pretreatment (Table 1). Aside from the rise in the conversion temperature, their presence can direct the process into a two-stage transformation of iron oxyhydroxides (through maghemite) instead of directly converting them to hematite (Fig. 2c). Based on the XRD and Raman analyses, which indicate the presence of maghemite in both product series obtained at lower temperatures, it seems most likely that both mechanisms were involved in hematite formation.

The last weight loss (TG/DTG) and the exothermic peak (DTA) visible above 700 °C were related to a final transformation into hematite—the most thermodynamically stable product in the reaction system (Fig. 2d) (Cornell and Schwertmann 2003). The key observation here is the distinctly lower temperature of this phase transformation for the sample after chemical treatment (717.3–717.4 °C), compared to the raw sample (754.2–760.8 °C). This confirms that impurities led to the higher hematite formation temperature, allowing the parameter range to be properly set in future experiments. It should also be noted that because of the dynamic nature of thermogravimetric studies, the phase transformations can occur even at a slightly lower temperature than was determined from the DTA curves.

Optimising the thermal treatment of the residue

To effectively optimise a process using the given experimental matrix, the system response must be correctly set, directly reflecting its efficiency. This is particularly difficult in the case of pigment production, as the unambiguous and quantitative definition of colour (which is not an intrinsic property of an object) may be an issue. Also, the hematite content may be misleading—although it is adopted by some authors as the system response (Colpas-Ruiz et al. 2020)—because the resulting pigment’s colour depends not only on the hematite content but also on the shape and size of the particles.

Meanwhile, a few systems can determine the quantitative colour parameters, allowing them to be positioned in a particular colour space (CIExyz, CIELab, CIELch, or Munsell). Thus, several measures were proposed based on the CIELab and CIELch systems, including the quantitative colour difference (ΔE^*_ab, ΔE^*₀₀, ΔC*_ab, DΔL*_ab, and Δh*_ab) (Table S3). Another measure that is potentially useful as a system response in red pigment synthesis is the redness index (RI), based on the CIE colour spaces (Barron and Torrent 1986) and often used to quickly determine the hematite content in soils and sands (De Souza Bahia et al. 2015; Cao et al. 2022). However, since colour is a complex, psychophysical phenomenon, the calculated values often do not coincide with the perceived colour differences (Broadbent 2010), requiring careful analysis and interpretation of the results.

As is well known, the crucial variables influencing the conversion of iron oxyhydroxides into red hematite pigment are temperature and holding time. Based on the thermogravimetric studies (Fig. 2a), the temperature range in this research was set at 650–850 °C. The preliminary experiments indicated less of an influence from holding time, so the range was established at 60–240 min. The design of the experiments, based on a Doehlert experimental matrix, is presented in Table 2.

Table 2.

Doehlert matrix (normalised (x_i) and effective (u_i) variables) and experimental responses for CH-WTR samples after thermal treatment

Exp	Normalized variables		Effective variables		System response
	x1	x2	u1 (T, °C)	u2 (t, min)	a*	Chroma, C*	Lightness, L*
1	1.0	0.000	850	150	11.506	11.967	23.480
2	−1.0	0.000	650	150	16.359	22.957	25.690
3	0.5	0.866	800	240	18.413	20.830	25.940
4	−0.5	−0.866	700	60	16.252	22.832	27.180
5	0.5	−0.866	800	60	20.013	22.856	26.570
6	−0.5	0.866	700	240	17.601	23.448	27.650
7	0.0	0.000	750	150	21.070	24.723	27.336
7’	0.0	0.000	750	150	21.234	24.477	26.605
7’’	0.0	0.000	750	150	21.582	24.933	26.946

Since the main goals of this research were to gain more insight into the process and to optimise the red pigment synthesis, all the products—as well as the commercial red pigment—were thoroughly characterised in terms of their structure and colour, and various parameters reflecting the effectiveness of the red pigment formation were calculated and discussed. Finally, the most suitable parameter was chosen as the system response, based on which the Doehlert optimisation procedure was further conducted.

Diffuse reflectance spectroscopy (UV-Vis)

Figure 3 presents the changes in the shape of spectra for the CH-WTR (a–b) and WTR (c–d) samples during thermal conversion into hematite and the spectra for the commercial hematite pigment. Tracking these changes provides good insight into the samples’ structural and physical transformations.

Fig. 3.

Diffuse reflectance spectra (DRS) and selected photographs of the CH-WTR (a–b) and WTR (c–d) samples after thermal treatment under various conditions

It is clearly visible that the shape of the hematite spectra, with the characteristic reflectance bands from 650 to 750 nm, which are related to intra-atomic transitions in the d-d orbitals of Fe³⁺ ions in octahedral sites (Fiuza et al. 2018), appears above 700 °C for the chemically pretreated sample (CH-WTR) and only above 800 °C for the raw sample (WTR). Essentially, the reflectance initially tends to increase with an increase in temperature and time, but after reaching the maximum, a rapid decrease is visible for both sample series. Since the hematite is the final product of the iron oxyhydroxides’ thermal transformations, and the spectra’s shapes do not undergo further changes, the drop in the reflectance can be explained by the agglomeration and sintering of the hematite particles (Fig. S3–S5). The XRD results also confirm this supposition, as seen in Fig. 4. Along with the increase in temperature and time, the XRD patterns of both samples (WTR and CH-WTR) became similar to the pattern of the commercial red hematite pigment, reaching the highest similarity at 850 °C. It is worth noting, however, that these transformations were observed at lower temperatures for the chemically pretreated samples. This is discussed in detail in the following paragraph, devoted to Raman spectroscopy. As seen from Fig. 3, in the case of the untreated sample (WTR), the shape of DRS spectra characteristic of hematite only forms above 800 °C and after a longer holding time (4 h), under which conditions the reflectance drops significantly. This is also evident in the Kubelka–Munk transformation plots (F(R) vs λ), as shown in Fig. S2, where a slight similarity in the shape of the spectra of the samples and the commercial pigment is only visible at 800 °C and 4 h. This leads to the key conclusion, essential from a practical point of view, that in the case of untreated samples, the conversion temperature required for hematite formation exceeds the sintering temperature. As a result, the pigment with the desired red shade cannot be obtained regardless of temperature or time. This is also clearly visible in the photographs of the samples.

Fig. 4.

XRD patterns of the commercial red pigment, the raw samples (wtr), the chemically treated samples (ch-wtr) and the thermally treated samples under various conditions

Raman spectroscopy

The Raman analysis of the raw (WTR) and thermally treated (CH-WTR) samples provides better insight into the progressive transformation of iron oxyhydroxides into hematite with increasing temperature and holding time (Fig. 5, Fig. S6). The absence of hematite (α-Fe₂O₃) in the raw samples is evidenced by the lack of its characteristic Raman peaks at 225 cm^–1, 295 cm^–1, 498 cm^–1, 612 cm^–1 and 1320 cm^–1 (Jubb and Allen 2010; Buzgar et al. 2013). Thermal treatment at 650 °C resulted in dehydration and the formation of maghemite (γ-Fe₂O₃), as confirmed by the peaks emerging at 288 cm^–1, 379 cm^–1, 500 cm^–1, and 716 cm^–1 (Jubb and Allen 2010). It is particularly important that the phase transition kinetics differ between the series, with the chemically pretreated samples exhibiting a more advanced transformation than the WTR ones at the same temperatures, likely due to enhanced precursor homogeneity resulting from the pretreatment (Cambier et al. 2014; Li and Hihara 2015). At 700 °C, hematite (α-Fe₂O₃) formation begins, evidenced by peak shifts to 405 cm^–1 and 503 cm^–1, alongside the gradual disappearance of FeOOH features (Hanesch 2009; Cambier et al. 2014). This transformation is further corroborated by the spectral evolution in the high-wavenumber region, where a broad maghemite-associated feature near 1400 cm^–1 shifts towards 1320 cm^–1, indicative of the two-phonon mode of hematite (Jubb and Allen 2010).

A clear temperature-dependent pattern of hematite formation was observed, with an increasing intensity and sharpness of its characteristic peaks as the temperature rose (Buzgar et al. 2013). At 750 °C, hematite became the dominant phase, as confirmed by the strong peaks at 406 cm^–1 and 498 cm^–1, with minor traces of maghemite remaining in the WTR samples but being nearly absent in the CH-WTR ones (Wang et al. 1998; Jubb and Allen 2010). By 800 °C, the FeOOH and maghemite phases were fully converted, with sharp hematite peaks at 407 cm^–1, 498 cm^–1, 612 cm^–1, and 1320 cm^–1, indicating high crystallinity (Cambier et al. 2014; Li and Hihara 2015). The final samples of CH-WTR and WTR, treated at 850 °C, exhibited fully developed hematite spectra, with no residual intermediate phases, confirming the complete structural transition into thermodynamically stable α-Fe₂O₃ (Jubb and Allen 2010; Buzgar et al. 2013).

The comparative analysis of the raw and chemically pretreated samples highlights the impact of chemical pretreatment on phase transition kinetics and structural homogeneity (Wang et al. 1998; Cambier et al. 2014). While both series followed the same thermal transformation pathway—from the FeOOH phases (goethite, ferrihydrite, and lepidocrocite), through maghemite at intermediate temperatures and finally to hematite—the pretreated samples exhibited a faster, more complete transition at lower temperatures (Hanesch 2009; Li and Hihara 2015). This difference is particularly evident in the 1400–1300 cm^–1 spectral region, where the maghemite-associated band in the WTR samples persisted longer before stabilising as the hematite two-phonon mode at 1320 cm^–1 (Jubb and Allen 2010). The hematite formation process is strongly temperature-dependent, with its characteristic peaks becoming more intense, better resolved, and shifting slightly as crystallinity improves. This confirms that hematite formation is a gradual, multi-step process, with full structural ordering achieved only at the highest temperatures (Jubb and Allen 2010). The enhanced phase transformation in the CH-WTR samples suggests that chemical pretreatment improves precursor purity and structural ordering, facilitating a more homogeneous and crystalline final hematite phase (Cambier et al. 2014; Li and Hihara 2015).

Morphology of the products

As seen from the SEM images (Figs S4, S5), the particles of all the materials, before and after thermal treatment, are agglomerated and show significant irregularity in shape and size. It is noteworthy that the particles in the untreated sample (WTR) show distinctly higher diversity and greater particle size distribution. Moreover, a further agglomeration can be observed along with the temperature increase for both the WTR and CH-WTR samples, so the particles’ characteristic shape and precise size are difficult to determine. In addition to the agglomeration, the particles are sintered above 800 °C, which is seen at a higher magnification of 60,000x. That matches well with the visible change in the sample colour, from red to purple-like in the temperature range of 750–850 °C.

At first view, the morphology of both sample series (WTR and CH-WTR) after thermal treatment looks similar. However, a more detailed analysis of the photos reveals their essential difference. For the CH-WTR samples, the thermal treatment at 750 °C and 800 °C formed many cubical-like particles, about 1 μm in size, that are evenly dispersed on the entire sample’s surface (Fig. 6).

These particles are not visible at higher temperatures in the products, likely because of the sintering and further agglomeration processes. The EDS analysis showed a higher Fe content in those particles and a lower content of the other metals, indicating that they may be pure hematite crystals. Similar cubical-like particles of hematite, ranging in size from tens to hundreds of nanometres, are identified in synthetic hematite obtained in various ways (Hoffmann et al. 2013; O’loughlin et al. 2021; Jin et al. 2022). On the contrary, very few such particles were found on the surface of the thermally treated WTR samples (without chemical pretreatment), regardless of the process temperature. It should be noted that this explains some of the differences observed for these two sample series in the Raman analysis, where the WTR samples seemed not only less crystalline, but also much less homogeneous.

Modelling of the process using a Doehlert experimental matrix

Based on the DRS spectra (Fig. 3), the quantitative characteristics of the colours in the CIELab space were determined for all the products obtained under various conditions specified in the experimental matrix (Table 3).

Table 3.

The values of CIELab/CIELCh colour parameters calculated based on DRS spectra (illuminator D65 and the standard observer 10°) for thermally treated WTR and CH-WTR samples; standard deviation values are given in parentheses

Sample	CIElab raw parameters
	L*	a*	b*	C*	h*
CH-WTR, 650 °C, 2.5 h	25.69 (± 0,24)	16.36 (± 0.14)	16.11 (± 0.26)	22.96 (± 0.29)	0.78 (± 0.004)
CH-WTR, 700 °C, 1.0 h	27.18 (± 0.39)	16.25 (± 0.16)	16.04 (± 0.07)	22.83 (± 0.17)	0.78 (± 0.003)
CH-WTR, 700 °C, 4.0 h	27.65 (± 0.46)	17.60 (± 0.28)	15.49 (± 0.27)	23.45 (± 0.03)	0.72 (± 0.017)
CH-WTR, 750 °C, 2.5 h	26.97 (± 0.52)	21.29 (± 0.41)	12.54 (± 0.07)	24.71 (± 0.31)	0.53 (± 0.011)
CH-WTR, 800 °C, 1.0 h	26.57 (± 0.29)	20.01 (± 0.03)	11.04 (± 0.14)	22.86 (± 0.10)	0.50 (± 0.005)
CH-WTR, 800 °C, 4.0 h	25.94 (± 0.20)	18.41 (± 0.06)	9.74 (± 0.46)	20.83 (± 0.27)	0.49 (± 0.018)
CH-WTR, 850 °C, 2.5 h	23.48 (± 0.60)	11.51 (± 0.93)	3.26 (± 0.86)	11.97 (± 1.12)	0.27 (± 0.048)
WTR, 650 °C, 2.5 h	25.35 (± 0.39)	18.75 (± 0.25)	17.32 (± 0.69)	25.53 (± 0.65)	0.75 (± 0.013)
WTR, 700 °C, 1.0 h	27.72 (± 1.45)	17.14 (± 0.43)	20.02 (± 1.51)	26.36 (± 1.43)	0.86 (± 0.025)
WTR, 700 °C, 4.0 h	28.06 (± 0.18)	17.43 (± 0.002)	19.22 (± 0.08)	25.94 (± 0.06)	0.83 (± 0.002)
WTR, 750 °C, 2.5 h	27.04 (± 0.19)	16.69 (± 0.14)	16.77 (± 0.01)	23.66 (± 0.11)	0.79 (± 0.004)
WTR, 800 °C, 1.0 h	25.73 (± 1.35)	16.25 (± 0.67)	12.92 (± 2.31)	20.78 (± 1.96)	0.67 (± 0.067)
WTR, 800 °C, 4.0 h	25.09 (± 0.75)	16.17 (± 0.59)	8.98 (± 0.71)	18.50 (± 0.87)	0.51 (± 0.018)
WTR, 850 °C, 2.5 h	22.52 (± 0.32)	8.67 (± 0.67)	1.90 (± 0.51)	8.88 (± 0.76)	0.21 (± 0.040)

The goal was to find the most suitable measure that would coincide with the perceived colour (Fig. 3), to be used as the system response in the optimisation procedure. As mentioned above, quantification of the similarity of the colours is the main issue when optimising a process using statistical tools. As seen in Fig. 7 and in Table S4, none of the calculated parameters defining colour differences (ΔE^*_ab, ΔE^*₀₀, ΔL*_ab, ΔC*_ab, and Δh*_ab) reflect the perceived colour impression of the samples. Despite the distinctly brown shade of the WTR samples, the calculated colour differences between them and the commercial red hematite pigment were often lower than those of the red-shaded CH-WTR samples. The redness index (RI) was also unsuitable since the values correlated strictly with the hematite content, regardless of the red shade of the samples. It is worth mentioning, though, that the values of the redness index were closest to that of the commercial red pigment for the samples obtained at 750 °C and above 800 °C for the CH-WTR and the WTR samples, respectively, corresponding well with Raman analyses.

Fig. 7.

Graphical representation of the calculated colour parameters (a* and RI) and colour differences (ΔE^*_ab, ΔE^*₀₀, Δh*_ab, DΔC*_ab, ΔL*_ab) between all the products and the commercial red hematite pigment

Nonetheless, one of the basic measures of the CIELab system, parameter a*, which expresses the red-green character of the colour, correlates very well with the perceived colours. The difference between the a* for the commercial red pigment and the product was lowest for the visually reddest sample (CH-WTR at 750 °C) (Table S4, Fig. 7). Thus, as this parameter was characterised by a clear functional dependence with a maximum at the point perceived as the “reddest,” it was selected as the suitable system response that would reflect the effectiveness of the conversion of the industrial residue into red hematite pigment. However, as chroma (C*) and lightness (L*) strongly affect perceived pigment quality, they were used as additional system responses, although they do not themselves reflect the efficiency of the analysed process. Table 2 presents the values of the experimental responses: a*, C*, and L*. The coefficients of the quadratic polynomial models were calculated from the results as follows:

$$a_{}^{\ast }=21.295-0.855x_1-0.073x_2-7.363x_1^2-1.846x_2^2-1.703x_1{x}_2$$

$$C_{}^{\ast }=24.711-4.096x_1-0.407x_2-7.249x_1^2-0.543x_2^2-{1.525x}_1{x}_2$$

$$L_{}^{\ast }=26.962-1.123x_1-0.046x_2-2.377x_1^2+0.623x_2^2-0.635x_1{x}_2$$

Figure 8 shows two- and three-dimensional graphs based on these models, depicting the influence of the temperature and holding time on the effectiveness of the conversion process. The curves on the contour plots represent the constant values of the mathematically estimated parameters: a*, C*, and L*, respectively. The detailed regression diagnostics for the quadratic model are presented in Tab. S5-S7 in the Supplementary Materials.

Fig. 8.

Contour and three-dimensional plots illustrating the influence of temperature and holding time on the efficiency of red pigment synthesis, expressed by parameter a*, chroma C*, and lightness L* values, based on the theoretical model calculated from the Doehlert experimental plan

Based primarily on the values of parameter a*, it is clearly seen that the optimal ranges of the process variables (T, t) are 721–770 °C and 100–200 min for the temperature and the holding time, respectively. The pigment closest in colour to the commercial product was obtained at 745 °C in 151 min, which agreed with the experimental results (Fig. 3). Considering the economics of the process, it can be concluded that the favourable range of the conditions is within the blue-marked area on the chart. Moreover, when analysing the influence of process parameters on chroma (C*) and lightness (L*) of the product, a small shift of the optimal area towards lower temperatures (725 ± 20 °C) is clearly visible, which seems to confirm the above assumption. From the plots, it is also clearly visible that the influence of process time on the values of parameters C* and L* is even smaller than in the case of parameter a*. However, it should be highlighted that decreasing the time from 4 to 1 h, at a temperature of 700 °C, resulted in a significant increase in the chroma difference (ΔC*_ab) between the obtained product and the commercial pigment (Table S4). It is also worth paying attention to the specific shape of the plot for the L* parameter (Fig. 8). It indicates that both shortening and extending the process time (compared to the optimal value) increases the lightness of the product and, thus, the difference ΔL*_ab (Tab. S4, Fig. 7). Nonetheless, it is evident that temperature has a crucial impact on pigment formation, far greater than holding time. Thus, considering the influence of time on the costs for the high-temperature processes, and the above-presented dependencies, it is worth mentioning that the product of the desired quality is likely to be obtained even at a shorter holding time, which, however, needs further and more detailed studies, preferably in a pilot-scale system.

Conclusions

The research proved the feasibility of thermally converting the industrial residue from an infiltration water treatment plant into hematite red pigment, highlighting the crucial role of chemically pretreating the raw material. An attempt was made to optimise the conversion process using RSM with a Doehlert design, indicating that the appropriate selection of a group of colour parameters as system responses allows for determining the optimal parameter range consistent with the perception of the product’s colour. Also, the changes in residue composition throughout the year were observed, which are crucial from a technological perspective. The key observations and conclusions arising from the study are listed below.

1. The similar composition and supply of the industrial residue over the course of a year, due to the nature of infiltration water treatment plants, indicates that the parameters of the resulting pigments can be reproducible, which is essential from a commercial point of view.

2. The presence of some impurities (e.g. Mn, Al, or Ca) in the raw residue shifts the phase change temperature of iron oxides into hematite above its sintering and agglomeration temperature, which prevents the direct synthesis of the red pigment. Thus, the proper chemical pretreatment was proven to be an essential step in order to decrease the process temperature. Simple chemical preprocessing of the raw residue resulted in the effective formation of a red pigment that was very close in colour to the commercial product.

3. The process optimisation, based on a Doehlert experimental matrix, defined the optimal range of the process parameters (holding time and temperature). Considering the much smaller impact of time on the effectiveness of the process as compared to temperature, and the ambiguity of quantitatively determining the colour, the process can be further refined from the point of view of its economics, preferably on a larger scale in a pilot plant.

4. The main established goals—demonstrate the feasibility of converting the given industrial residue into red hematite pigment, and to explain the necessity and role of chemical pretreatment—were achieved. However, in addition to the necessary further optimisation studies on a pilot scale, many aspects of the proposed process must be carefully analysed. These include economic analysis (residue vs. natural resources as the feedstock), environmental analysis (acid treatment of the raw residue and generated sewage treatment), and the synthesis of multiple product batches throughout the year to confirm the constant parameters and quality of the obtained pigment: colour, oil absorption, dispersion/fastness data in a model paint or cement matrix.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contribution

Daniel Ociński: conceptualisation, supervision, methodology, investigation, validation, formal analysis, visualisation, and writing—original draft. Michał Ozga: methodology, investigation, validation, visualisation, and writing—original draft. Igor Mucha: methodology, investigation, validation, and visualisation. All authors read and approved the final manuscript.

Funding

The authors declare that no funds, grants, or other support were received during the preparation of this manuscript.

Data availability

Raw data analyzed in this study (XRD, Raman, DRS spectra), thermogravimetric curves, and Doehlert design are now publicly available in the Mendeley Data repository: https://data.mendeley.com/datasets/kwnt82bhkf/1, 10.17632/kwnt82bhkf.1.

Declarations

Ethics approval

This is not applicable.

Consent to participate

This is not applicable.

Consent for publication

This is not applicable.

Competing interests

The authors declare no competing interests.