A. Starting material and sample preparation

PB (Sigma Aldrich) was used without further treatment as a starting material. Thermal analysis (TG/DSC) of PB was carried out using a thermal analyzer (STA 449 C Jupiter, Netzsch), which was coupled to mass spectrometer (QMS 403 Aëolos, Netzsch) to analyze the evolved gases (EGA) during the decomposition reaction.

PB powder (10 mg) was placed into an open alumina crucible. Then, it was heated up to 900 °C under argon atmosphere (gas flow 30 ml min⁻¹), with heating and cooling rates of 5 and −40 K min⁻¹, respectively. An additional set of PB samples were heated up to selected temperatures (190, 300, 370, 540, 680, and 790 °C). The resulting samples were labeled as PB190, PB300, PB370, PB540, PB680, PB790, and PB900, respectively.

B. Characterization techniques

XRD patterns were recorded at room temperature (RT) with a PANalytical X'Pert PRO MPD diffractometer in the Bragg-Brentano geometry, CoK α radiation (40 kV, 30 mA, λ = 0.1789 nm), equipped with an X'Celerator detector and programmable divergence and diffracted beam anti-scatter slits. Samples were placed on a zero-background Si slide, gently pressed and scanned with a step size of 0.017°, and 2θ scan range from 15° to 105°. Phase identification and Rietveld quantitative phase analysis were performed using PANalytical HighScore Plus software with PDF-4+ and ICSD databases.

Transmission ⁵⁷Fe Mössbauer spectroscopy was carried out in a constant acceleration mode using a ⁵⁷Co source in rhodium matrix at RT. The spectrometer was calibrated with an α-Fe foil. The spectra were folded and fitted by Lorentzian functions using the computer program CONFIT2000 (Žák and Jirásková, Reference Žák and Jirásková2006).

C. High-temperature in situ XRD

Thermal decomposition of PB was studied from the structural point of view by in situ XRD. Diffraction patterns were collected using a PANalytical X'Pert PRO MPD diffractometer (CoK α radiation) in the Bragg–Brentano geometry, equipped with an X'Celerator high-speed detector and an Anton Paar XRK-900 reaction chamber. The powder sample was placed directly on a glass ceramic Macor holder (14 mm diameter, 0.5 mm deep). Data collection was carried out under nitrogen atmosphere (pressure 1 bar, gas flow rate 20 ml min⁻¹), with a heating rate of 5 K min⁻¹ from RT up to 900 °C. The collection time of each XRD pattern was 10 min (0.12 s step⁻¹, step size of 0.017°, 2θ scan range from 15° to 105°).

A. Characterization of PB

The purity of commercial PB was checked by XRD and ⁵⁷Fe Mössbauer spectroscopy. Both techniques showed that PB powder consisted of iron(III) hexacyanoferrate (PDF 01-073-0689) and a minor phase (~3%) identified as the mineral Jarosite [PDF 01-076-0629, (K,Na)Fe₃(SO₄)₂(OH)₆] (Aparicio et al., Reference Aparicio, Machala and Marusak2012). In addition, energy dispersive spectroscopy (EDS) has been performed on the PB sample, confirming the presence of the elements Na, S, and K (Figure S1).

B. Decomposition of PB under argon atmosphere: TG/DSC analyses

Six differentiated steps can be resolved in the TG curve of PB from RT to 900 °C (Figure 1, Table I). In the same graph the DSC curve and the evolution of gases during the heating process are presented. Water release (m/z = 18) takes place from 55 to 300 °C (according to EGA), which corresponds well with the observed decrease in mass and a wide endothermic peak at 150 °C (steps I and II). Using this information, the amount of water molecules per unit cell was calculated to be five, a number lower than the theoretical value for PB structure. Because of the small amount of Jarosite impurity (~3%) in the PB sample, its contribution is neglected from calculation of water content.

Figure 1.

TGA, DSC, and EGA graphs of PB treated under argon atmosphere with a heating rate of 5 K min⁻¹. Roman numbers indicate the different steps of the decomposition.

Table I.

Mass loss estimated from TG curve for PB sample heated up to 900 °C in Ar.

The thermal decomposition of PB in argon atmosphere started above 350 °C (step III), with the release of cyanide groups (CN⁻, m/z = 26), forming thereafter hydrogen cyanide HCN (m/z = 27) and cyanogen gas (CN)₂ (m/z = 52), and a small amount of carbon dioxide (CO₂) (m/z = 44) according to the EGA results (Figure S2). It is reflected by the two small downward peaks in DSC (endo-effect) at 330 and 360 °C.

In contrast, the mass decrease in the other steps (IV–VI) is accompanied by the release of nitrogen (N₂, m/z = 28) and probably also a small amount of CO₂, during the formation of iron carbides above 400 °C (see the next section). Given that m/z could be assign either to N₂ or CO, one cannot rule out the presence of small amount of CO. In this particular case, the presence of the m/z = 14 (N), which is identified as the fragment of the molecular ion N₂, besides it have the same emission profile as m/z = 28 (Figure S2). From the previous discussion, it can thus be concluded that m/z = 28 is more likely to be assigned to N₂, and the CO contribution might be minimal.

C. Thermal behavior of PB under Ar atmosphere: ex situ XRD and Mössbauer spectroscopy

The phase composition of the PB samples thermally treated up to selected temperatures was variable, showing clearly not only the thermal decomposition of initial PB, but also the formation of iron carbides and their subsequent decomposition (Figure S3).

At 190 °C, a structurally unchanged PB and traces of Jarosite were observed, but at 300 °C PB becomes dehydrated, still preserving its crystalline character. Above 350 °C, PB decomposition has started; in sample PB370 were detected mainly non-transformed PB and an iron cyanide compound, additionally minor organic phases were identified (urea and malic acid). The formation of the aforementioned organic phases from PB (e.g. urea) has been reported earlier during thermal wet decomposition under inert atmosphere by Ruiz-Bermejo et al. (Reference Ruiz-Bermejo, Rogero, Menor-Salván, Osuna-Esteban, Martín-Gago and Veintemillas-Verdaguer2009).

The formed iron cyanide compound match the PDF card 01-075-023, belonging to a Cu(II)–Fe(II) cyanide compound with a cubic structure (space group $F\overline 4 3m$ ). It is known, that most of the ferrocyanide compounds with the structure M₂[Fe(CN)₆] crystallize in the cubic system, and can be indexed either with the space groups F432, $F\overline 4 3m$ , or $Fm\overline 3 m$ (Ratuszna et al., Reference Ratuszna, Juszczyk and Małecki1995), some examples of these cubic ferrocyanides are displayed in Figure S4 and the details are listed in Table S1. Other clue about the structure of this iron cyanide compound is given by the Mössbauer spectrum of the sample PB370 (Figure 2(a), Table II). In the spectrum, spectral components were identified (L1, D2) corresponding to undecomposed PB, the high value of quadrupole splitting (QS) can be related to strain in the PB lattice (Samain et al., Reference Samain, Grandjean, Long, Martinetto, Bordet and Strivay2013). Sub-spectra L2 (singlet) and D1 (doublet) correspond to Fe²⁺ cations with low-spin and high-spin states, respectively. The presence of the singlet with a small isomer shift (IS = −0.09 mm s⁻¹) close to zero and absence of a QS, is a clear indicator of a highly symmetric site typical for cubic structures with fcc arrangement, where the Fe²⁺ is six-coordinated. The doublet (IS = 1.12 mm s⁻¹, QS = 1.47 mm s⁻¹), represent a Fe²⁺ cation with high-spin state. Then,  both sub-spectral components can be assigned to the formed iron cyanide compound, where the values of the IS and QS are close to the ones measured in a pure sample of ferrous ferrocyanide (Hu and Jiang, Reference Hu and Jiang2011). It can be concluded that the formed compound is likely the iron(II) hexacyanoferrate(II) (Fe₂[Fe(CN)₆]), also known as Berlin white (BW). Finally, the provenance of the singlet (L3) with a broad linewidth will be explained below.

Figure 2.

Room-temperature Mössbauer spectra of samples PB370 (a) and PB540 (b). PB (gray and dark gray subspectra), BW (white doublet and dashed singlet), Fe₂C (black sextet).

Table II.

Mössbauer parameters obtained from the fit of spectra for samples PB370 and PB540 heated under argon atmosphere.

IS, isomer shift; QS, quadrupole splitting; W, experimental line width; B _hf, magnetic hyperfine field; HS, high spin; LS, low spin; S, sextet; D, doublet; L, singlet.

The formation of BW differs from the previous study, where a monoclinic iron cyanide structure was identified (Aparicio et al., Reference Aparicio, Machala and Marusak2012); however, it agrees with other study performed in argon atmosphere (Allen and Bonnette, Reference Allen and Bonnette1974). It is suggested that the reason of such disagreement might be the different heating rate of 10 vs. 5 K min⁻¹ in this study, as this can be compared with the diffraction patterns from PB samples taken after the dehydration steps at the beginning of the PB transformation (step III, at 400 and 380 °C for samples heated under argon with 10 and 5 K min⁻¹, respectively) in Figure S5, where the diffraction patterns are significantly different.

At 540 °C, all PB has been transformed into BW, and also simultaneously started the formation of iron carbides: hexagonal (Fe_2+x C, P6₃22) and orthorhombic (Fe₂C, Pbcn) forms. Minor phases oxydiacetic acid and urea has been also identified. From the Mössbauer spectrum (Figure 2(b), Table II) BW was identified (D2, L1), where the doublet has a higher value of QS, denoting a deformation of the octahedral site of ferrous ions. It was not possible to assign the doublet D1 (ferric ions most probably in octahedral coordination) to any compound identified in the XRD pattern. The singlet L2, with a very wide linewidth, has a similar isomer shift to singlet L3 in sample PB370. It could reflect the arrangement of iron cations prior to the formation of iron carbides, being in accordance with the poor crystallinity of iron carbides observed in the respective XRD pattern. The presence of a sextet with small subspectral area (7.3%) is ascribed to crystalline Fe₂C.

In sample PB680, it is observed that the formation of Eckstrom-Adcock carbide (Fe₇C₃, P6₃ mc) along with other iron carbides. At higher temperatures, the samples typically contain different iron carbides: in sample PB790 were identified Fe₇C₃, Hägg carbide (Fe₅C₂, C12/c1), cementite (Fe₃C, Pnma), and a small amount of K₄[Fe(CN)₆]. The formation of Fe₅C₂ was not observed at higher heating rates (Aparicio et al., Reference Aparicio, Machala and Marusak2012). In sample PB900 the predominant phase is cementite, additionally, diffraction lines from metallic iron (α-Fe, $Im\overline 3 m$ ) were also present.

D. Thermal behavior of PB under nitrogen atmosphere: in situ XRD

The 43 XRD patterns collected during the high temperature measurements were categorized into six groups, corresponding to the steps observed in TG (Figure 1). Representative XRD patterns from each group are shown in Figure 3. Quantitative phase analysis was performed for all diffraction patterns using the Rietveld method, considering only the known phases and excluding the minor phases in the XRD patterns. Unit-cell of PB was refined using the Herren model with $Fm\overline 3 m$ space group (Herren et al., Reference Herren, Fischer, Ludi and Hälg1980), and BW was refined using the Ratuszna and Juszczyk model with $F\overline 4 3m$ space group (Juszczyk et al., Reference Juszczyk, Johansson, Hanson, Ratuszna and Małecki1994; Ratuszna et al., Reference Ratuszna, Juszczyk and Małecki1995) for anhydrous forms of M ₂[Fe(CN)₆].

Figure 3.

Selected XRD patterns from in situ measurements of PB heated under N₂. The symbols represent the identified phases: Jarosite (J), Fe₄[Fe(CN)₆]₃ (), Fe₂[Fe(CN)₆] (), Fe_2+x C (), Fe₂C (), Fe₇C₃ (), Fe₂C₅ (), Fe₃C (), γ-Fe⁰ (), C (), Urea (+), Malic acid (#), 2-Naphthylamine-1-sulfonic acid (z), Na₂(SO₃) (w), K₃N (p), NaCN (y).

Below 300 °C, PB is not decomposed and the structure is preserved, but one can observe changes in the position of the main diffraction peaks in the XRD pattern (Figure S6). Up to 180 °C, reflections (200), (220), and (400) are shifted to the high-angle side with the increase in temperature, clearly indicating a negative thermal expansion (Matsuda et al., Reference Matsuda, Kim, Ohoyama and Moritimo2009). On the contrary, a positive thermal expansion is observed above 180 °C. The observed negative thermal expansion can be related to dehydration of PB, as a result of a progressive contraction of the unit cell with the release of water molecules (Figure 4).

Figure 4.

Temperature dependence of unit-cell parameters of PB showing negative thermal expansion. The dashed line is a guide to the eye.

Figure 5 summarizes the course of PB transformation into BW, further decomposed into iron carbides, leading to the final products cementite, iron, and carbon; results which are comparable with those obtained when PB was heated under argon.

Figure 5.

(a) Quantitative results of the Rietveld analysis of XRD patterns collected in situ during the thermal decomposition of PB under nitrogen atmosphere. (b) Close-up view of the central region of the graph marked with a rectangle in (a).

E. Overall mechanism of thermal decomposition of PB and relationship between iron-carbide phases

The first step in the thermal decomposition of PB is dehydration. In the TG graph (Figure 1) the mass loss between 35 and 189 °C corresponds to the release of 3.5 uncoordinated water molecules from the PB structure (Δm _th = −6.39%). Then, a slight decrease of mass up to 300 °C, corresponds to the release of 1.5 coordinated water molecules (Δm _th = −3.09%).

Between 300 and 420 °C, along with the release of (CN)⁻ groups from the dehydrated PB structure, PB is transformed into BW (Figure 5), according to the following reaction:

Figure 5 shows that orthorhombic Fe₂C and hexagonal Fe_2+x C (0 ≤ x ≤ 0.33) are the first iron carbides formed up to 580 °C, followed by a steady increase of the amount of Fe₂C between 580 and 660 °C, thus indicating a faster decomposition of BW. At 620 °C, hexagonal iron carbide Fe₇C₃ is formed. The process can be explained by the following equations:

Between 660 and 760 °C, the iron carbide Fe₅C₂ has been formed, and its weight percentage increases steeply at the same rate as Fe₇C_3. The weight percentage of both iron carbides increases (Fe₅C₂ and Fe₇C₃) at the same time that the weight percentage of Fe₂C decreases, suggesting a transformation of the primary iron carbide. Minor phases Fe₃C and C are also present. The possible relation between these carbides can be expressed by:

A decrease in Fe₇C₃ (−11 wt%) is observed between 760 and 780 °C. At the same time, the weight percentage of Fe₅C₂ and Fe₃C increased by 5 and 13%, respectively. The sharp rise in Fe₃C weight percentage might be explained by the equations:

Equation (7) has been proposed earlier by Iguchi et al. (Reference Iguchi, Kouda and Shibata2004) and Herranz et al. (Reference Herranz, Rojas, Pérez-Alonso, Ojeda, Terreros and Fierro2006). It is also known that formation of cementite occurs at temperatures higher than 480 °C and that Fe₅C₂ always precedes Fe₃C formation (Cohn and Hofer, Reference Cohn and Hofer1953).

Above 800 °C, the phase composition consists of Fe₃C, γ-Fe, and C. It is obvious from Figure 5 that the decrease in weight percentage of Fe₃C corresponds to a formation of γ-Fe and C (3 Fe₃C → 3 Fe + C), because of the instability of cementite at temperatures higher than 700 °C (Iguchi et al., Reference Iguchi, Kouda and Shibata2004; Chaira et al., Reference Chaira, Sangal and Mishra2007).

The observed transformations between carbides agree with the known fact that the stability of iron carbides increases when the carbon content decreases: Fe₂C < Fe_2.2−2.4C < Fe₅C₂ < Fe₃C (Le Caer et al., Reference Le Caer, Dubois, Pijolat, Perrichon and Brussiѐre1982; Jung and Thomson, Reference Jung and Thomson1992).

Taking into account the provided information by TG/EGA and XRD, one can summarize the overall decomposition mechanism of PB in inert atmosphere by Eq. (8). In this equation, nitrogen was included. Although its provenance is not well understood, it was released during the transformation of iron carbides (Figure 1): Fe₂C to Fe₅C₂ and Fe₇C₃ (642 °C), Fe₇C₃ to Fe₃C (763 °C), and Fe₃C to Fe (842 °C). Such as nitrogen release has been witnessed before in other studies during the thermal decomposition of insoluble PB (Allen and Bonnette, Reference Allen and Bonnette1974), soluble PB (Chamberlain and Greene, Reference Chamberlain and Greene1963), and other hexacyanoferrates (Gallagher and Prescott, Reference Gallagher and Prescott1970) in inert atmosphere or vacuum. In those cases, the nitrogen was detected above 400 °C in the case of the PB, and above 600 °C in the case of the other hexacyanoferrates. Additionally, as in the present study, the final decomposition products of the aforementioned compounds are Fe₃C, Fe, and C. These studies only mention the final products, but not the intermediate iron carbides, and they claim thermal decomposition follow the route: ferrocyanide → Fe(CN)₂ → Fe₃C, Fe, C. Other reasons supporting the evolution of nitrogen is the fact that the main products of PB decomposition are iron carbides and not iron nitrides as proved by EDS analysis (Figure S1). Other minor compounds containing nitrogen has been identified from XRD (Figure 2 and S1), but they also contain elements from the impurities of the precursor material (K, Na, and S) as proved from EDS analysis (Figure S1).

The theoretical mass loss is −53.74%, as calculated from Eq. (8), which slightly differs from the experimentally observed mass loss of −56.39%. The reason for the differences could be attributed to the minor impurities in the initial PB.