Efficient removal of AlN from secondary aluminum dross using binary fluoride: a theoretical and experimental study

Abstract

Secondary aluminum dross (SAD) typically contains significant amounts of AlN, which is a major source of the toxic gases released from SAD. Converting AlN into more stable aluminum compounds is crucial for the safe disposal and resource recovery of SAD. In this study, the oxidative removal of AlN from SAD was investigated, and an efficient denitrification method is proposed. Theoretical and experimental results demonstrate that AlN oxidation forms a dense, ordered [Al–O] oxide layer on its surface, which hinders further oxidation. NaF, AlF₃, and Na₃AlF₆ were found to enhance SAD denitrification, and binary fluoride additives KF–Na₃AlF₆ provided further improvement. Roasting SAD at 800 °C for 60 min with 12 wt% KF–Na₃AlF₆ (6 wt% Na₃AlF₆ + 6 wt% KF) achieved 93% oxidation of AlN. The treated SAD showed low reactivity and released negligible gas in aqueous solution. The promoting effect of Na₃AlF₆-based additives is attributed to their ability to facilitate the transformation of the dense, well-ordered [Al–O] surface layer into β-alumina.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-43443-6.

Introduction

Secondary aluminum dross (SAD) is a hazardous by‑product generated during aluminum production, processing, and recycling. Global SAD production exceeded 6.7 million tons in 2023¹. Predominant management practices such as landfilling or storage^2,3 entail serious environmental and health risks. Under humid conditions, SAD releases toxic ions (e.g., F⁻, CN⁻, NH₄⁺)⁴, flammable gases (e.g., CH₄, H₂, NH₃)⁵, and high‑pH chlorate‑containing leachates⁶, leading to groundwater contamination and soil salinization. On the other hand, SAD contains substantial amounts of valuable secondary resources, including metallic aluminum (typically 3–15 wt%), α‑Al₂O₃, AlN, fluorides, chlorides, MgAl₂O₄, SiO₂, and Fe₂O₃⁷. Effective treatment could thus deliver both environmental and economic benefits.

A major challenge is the significant AlN content in SAD, which acts as an important aluminum resource but also constitutes a primary source of toxic gas emissions. Converting AlN into more stable aluminum compounds is therefore crucial for the safe disposal and resource recovery of SAD. Current AlN‑removal methods mainly fall into hydrometallurgical and pyrometallurgical routes, each with inherent limitations. Hydrometallurgical processes can recover valuable components from SAD^8–10; however, they risk releasing flammable and toxic gases^4–6 and generally suffer from long processing times, high costs, and the generation of large volumes of saline wastewater¹¹. Pyrometallurgical methods, in contrast, can remove harmful components without hazardous gas emissions, but their poor recovery of valuable elements restricts their use mainly to producing sintered refractory materials such as firebricks and ceramics¹⁰. As a result, a combined pyro‑hydrometallurgical approach is regarded as the most promising strategy for efficient SAD treatment¹¹.

AlN oxidation typically starts at 800–900 °C¹², yet complete removal often requires prolonged roasting at even higher temperatures^13,14. Additives such as NaOH^16–18, Na₂CO₃^19–21, CaO²⁰, CaCO₃^19,21, and cryolite¹⁵ have been used to enhance AlN oxidation and promote Al₂O₃ recovery. However, most studies have focused primarily on removal efficiency, while overlooking practical challenges such as high additive dosages^17–19, elevated roasting temperatures^11,20,21, and extended processing times^15,19,20. Besides energy costs and secondary pollution, these drawbacks can also lead to salt volatilization, which deposits on furnace walls, accelerates corrosion, and shortens equipment service life. Hence, developing methods that enable rapid AlN oxidation at lower temperatures is essential for achieving efficient and sustainable SAD treatment.

In this work, the oxidation mechanism of AlN was firstly investigated. Subsequently, the effects of various additives on the oxidation of AlN in SAD were evaluated. Key parameters including the dosage and composition of additives, roasting temperature, and roasting duration were optimized. Based on these findings, an efficient denitrification method for SAD is proposed. We hope these results will contribute to advancing the recycling and resource recovery of SAD.

Calculation and experiment

Calculation details

Surface reaction

The reaction of O₂ with the AlN (001) surface was investigated using the VASP 6.3.0^22,23. The Projector‑Augmented Wave (PAW) method²⁴ was employed, and the exchange‑correlation functional was described by the Perdew–Burke–Ernzerhof (PBE) formulation of the generalized gradient approximation (GGA)²⁵, supplemented with the D3(BJ) dispersion correction. Spin polarization was included in all calculations. After convergence tests, a plane‑wave energy cutoff of 450 eV and a 3 × 3 × 1 Γ‑centered k‑point mesh²⁶ were used for Brillouin‑zone sampling. Geometry optimizations were performed until the residual forces on all relaxed atoms were below 0.02 eV Å⁻¹.

The transition states (TS) for the surface reactions were located using the climbing‑image nudged elastic band (CI‑NEB) method^27,28 and verified by vibrational frequency analysis (a single imaginary frequency corresponding to the reaction coordinate).

A 3 × 2 × 3 supercell of the wurtzite AlN (001) surface (space group P6₃mc, a = 3.113 Å, b = 5.392 Å, c = 4.981 Å, α = β = γ = 90°) was constructed. A vacuum layer of 15 Å was added along the z-direction to avoid spurious interactions between periodic images. During the optimization, the bottom two atomic layers were fixed at their bulk positions, while the upper layers were allowed to relax fully.

Deep potential modeling

A machine‑learning interatomic potential (Deep Potential, DP) for the Al–N–O system was trained using a combination of CP2K 9.1²⁹, DeepMD‑kit 2.2.7³⁰, and DP‑GEN 0.12³¹. The initial training dataset was generated from three separate ab initiomolecular dynamics (AIMD) simulations performed with CP2K/Quickstep in the NVT ensemble. Each simulation used a surface model consisting of a six‑layer AlN (001) slab and 30 O₂ molecules (see Figure S1, SM) and was run for 15 ps at a different temperature: 700 °C, 1500 °C, and 2300 °C.

The DP‑GEN framework was applied iteratively to refine the training dataset. In each iteration, four independent DP models were trained from different initial weights (600,000 steps per training). Model deviations were then evaluated via LAMMPS simulations at 700 °C, 1500 °C, and 2300 °C. Configurations with force deviations between 0.25 and 0.35 eV Å⁻¹ (up to 300 per iteration) were selected and added to the training set. This iterative process continued until the Deep Potential Molecular Dynamics (DPMD) simulations at the three target temperatures converged with a 99% accuracy threshold.

To validate the final DP models, 500 configurations were randomly sampled from 200‑ps DPMD trajectories simulated at 700 °C, 1100 °C, 1500 °C, 1900 °C, and 2300 °C. The total energies and atomic forces predicted by the DP models were compared against reference DFT calculations. As shown in Fig. 1, the strong linear correlation between the DP and DFT results confirms the high fidelity of the trained models. Further details of the AIMD simulations and DP training procedure are provided in the Supplementary Material (SM).

Fig. 1.

​ Correlation between the total energy and atomic forces calculated by the DP and DFT.

Experiment

Material

SAD samples were collected from an aluminum recycling facility in Jiangxi, China, using the cone and quartering method. The phase and chemical characteristics of SAD were analyzed according to the procedures outlined in Sect. 2.3, and the elemental composition is summarized in Table 1. It is worth noting that the oxygen (O) and silicon (Si) contents were estimated from qualitative XRD analysis. Since the concentrations of some oxide phases are below the XRD detection limit (~ 5 wt%), the reported values for O and Si are likely underestimated. As shown in the XRD pattern (Figure S3, Supplementary Material), the main phases identified in SAD are Al₂O₃, metallic Al, AlN, MgAl₂O₄, and Si, along with minor amounts of fluoride and chloride phases. These findings are consistent with the phase composition and distribution of SAD reported in previous studies^11,15,21.

Table 1.

Elemental composition of SAD (wt%) ^a.

Element	Al	O	N	Si	Mg	Fe	Ca	Na	Cu	Ti	F	Cl	Zn	S	K
Content	43.35	25.73	5.89	4.21	1.79	1.81	1.22	1.40	0.70	0.38	1.28	0.92	0.34	0.21	0.20

^a O and Si contents were estimated by XRD analysis. Other elements were determined by ICP-OES. Elemental compositions are reported for the dried SAD.

Roasting procedure

Roasting experiments were performed in a GSL‑1700X tube furnace (Hefei Kogent, China). The furnace and a corundum crucible were preheated to the target temperature before each experiment. Both the SAD and the additives were ground to a particle size of < 100 mesh (0.15 mm). A 15 g portion of SAD powder was thoroughly mixed with a predetermined amount of additive. The mixture was then placed into the preheated corundum crucible, which was immediately returned to the tube furnace. After the designated roasting time, the crucible was removed and cooled in ambient air (25 °C). Therefore, the roasting duration reported in this work specifically refers to the time the samples were held at the target temperature. A gas flow of 1 L/min was maintained through the tube furnace during the entire process to ensure an oxidative atmosphere.

Leaching experiment

Leaching experiments were conducted in sealed glass vessels at ambient conditions. The roasted SAD (RSAD) samples were ground to a particle size of < 100 mesh (0.15 mm), consistent with the raw SAD preparation. For each experiment, 5 g of RSAD was added to a vessel containing 50 mL of leaching solution: deionized water, NaOH solution (pH 13), or HCl solution (pH 1). The vessel was sealed and stirred continuously for 24 h. Any gases generated during leaching were collected. After the leaching period, aliquots of the supernatant were taken, filtered through 0.02 μm Anodisc membrane filters, and subjected to elemental analysis. The solid residues were washed with the corresponding leaching solution, dried, and further characterized. For comparison, the leaching behavior of the raw SAD was also examined under identical conditions. All experiments were carried out in triplicate, and the mean values are reported.

Analysis

The phase composition of solid samples was characterized by X‑ray diffraction (XRD, X’Pert3, Malvern Panalytical, Netherlands). Microscopic morphology was examined using scanning electron microscopy (SEM, Nova Nano SEM 450, FEI, USA) and transmission electron microscopy (TEM, Tecnai G2 F20 S‑TWIN, FEI, USA). The AlN content in the solids was quantified via the Kjeldahl titration method³². The denitrification rate reported in this work was calculated from the AlN content before and after the reaction.

Fluoride content in the solids was determined with a PXSJ‑216 F ion‑selective electrode. Samples were first fused with NaOH at 650 °C, and fluoride ions were subsequently extracted with sodium citrate and potassium nitrate solutions. Chloride content was measured by fusing the sample with Na₂CO₃, dissolving the fusion product in water, and titrating with silver nitrate solution. In leachates, fluoride and chloride concentrations were also analyzed using the same ion‑selective electrode. Concentrations of metallic elements in both solid and liquid samples were measured by inductively coupled plasma optical emission spectrometry (ICP‑OES, SPECTRO BLUE, SPECTRO, Germany).

In‑situ XRD analysis was carried out on a Malvern Panalytical Empyrean Alpha 1 diffractometer to monitor the phase evolution of AlN (with additives) in air from 25 °C to 800 °C. A 50 mg sample was loaded into the sample holder at room temperature (25 °C) and heated to 800 °C at 10 °C/min. XRD patterns were collected at six fixed time‑temperature points: 0 min (25 °C), 35 min (200 °C), 70 min (400 °C), 105 min (600 °C), 140 min (800 °C), and 175 min (800 °C), with a scan rate of 4 °/min. During each scan, the temperature was held constant to ensure measurement accuracy.

Results and discussion

At elevated temperatures, AlN undergoes oxidation to form Al₂O₃, which is a critical reaction in the pyrometallurgical processing of SAD. Enhancing this oxidation process is essential to improve the thermal efficiency of SAD processing. This study initiates with an investigation into the mechanisms of AlN oxidation.

Mechanism of AlN oxidation

The adsorption and dissociation of O₂ on the AlN surface constitute the fundamental initial steps in its oxidation. To elucidate these processes, density functional theory (DFT) calculations were performed using a six-layer AlN (001) slab model. Due to the non-coplanar arrangement of N and Al atoms within the same layer, the slab exhibits two distinct surface terminations. On surface-a, where Al atoms are positioned closer to the bulk, O₂ adsorbs at Al-N hexagonal sites, forming three O-Al bonds with an adsorption energy of −72.78 kcal/mol (Fig. 2a). In contrast, on surface-b, where N atoms are nearer to the bulk, O₂ binds to adjacent Al-N pairs with a positive adsorption energy of 6.03 kcal/mol (Fig. 2d). These results indicate that O₂ adsorption is thermodynamically favorable on surface-a but unfavorable on surface-b.

Fig. 2.

​Adsorption and dissociation of O₂ on the AlN (001) surface.

The adsorbed O₂ subsequently dissociates into two surface-bound O atoms. On surface-a, the dissociation proceeds with an energy barrier of 7.57 kcal/mol and releases 67.14 kcal/mol (Fig. 2b, c). On surface-b, the barrier is slightly higher at 10.08 kcal/mol, with a significantly lower energy release of 4.14 kcal/mol (Fig. 2e, f). These results suggest that O₂ dissociation is both kinetically and thermodynamically feasible on both surfaces. However, given the unfavorable initial adsorption on surface-b, the oxidation of AlN is hypothesized to occur predominantly on surface-a.

To further elucidate the oxidation process of AlN under more dynamic conditions, Deep Potential Molecular Dynamics (DPMD) simulations were conducted at 700, 900, 1100, 1300, and 2300 °C (see Figure S4 in the Supplementary Material). For clarity, the results at 700, 1100, and 2300 °C are presented in Fig. 3. At 700 °C, O₂ preferentially adsorbs and dissociates on surface-a, consistent with the DFT energy barrier results. Throughout the 2000 ps simulation at this temperature, neither significant oxygen migration from the surface into the bulk nor the formation of gaseous products was observed. As the simulation temperature increased, Al-N bonds on surface-a began to break, leading to unsaturated Al atoms bonding with oxygen, while nitrogen atoms combined to form N₂ molecules. As the simulation progressed, the Al-O structure transformed into a well-ordered, dense [Al-O] layer. This layer remained structurally stable in subsequent simulations and effectively prevented further interaction between O₂ molecules and the underlying AlN bulk. The formation of this passivating [Al-O] layer was consistently observed at 900, 1100, and 1300 °C. Concurrently, oxidation on surface-b also increased gradually with temperature, albeit to a much lesser extent than on surface-a.

Fig. 3.

Snapshots from DPMD simulations of AlN oxidation at 700 °C, 1100 °C, and 2300 °C.

Due to the differences in spatial and temporal scales between simulations and real systems, the computational results cannot fully replicate the actual reaction processes. Nevertheless, they reveal distinct oxidation sensitivities between the two AlN surfaces. On surface-a, the thermodynamically favorable adsorption and dissociation of O₂ render it more susceptible to initial oxidation. However, the rapid formation of a dense [Al-O] layer subsequently inhibits further oxidation. In contrast, on surface-b, the thermodynamically unfavorable O₂ adsorption results in inherently slower oxidation kinetics. The absence of a similarly dense passivating layer allows oxidation on surface-b to increase gradually with temperature. Consequently, the overall oxidation of AlN is negligible at lower temperatures but becomes progressively more pronounced as the temperature rises. These findings support the hypothesis that the dense [Al-O] layer plays a crucial role in passivating AlN against continuous oxidation.

To test this hypothesis, DPMD simulations were conducted at 2100 °C, a temperature exceeding the melting point of α-Al₂O₃ (Fig. 3c). At this elevated temperature, a pliable, less-ordered [Al-O] layer formed on surface-a. This deformable structure allowed O₂ to permeate and react with the underlying AlN bulk, ultimately leading to its complete oxidation. The concurrent enhancement of oxidation on surface-b under these conditions further supports the conclusion that disrupting the dense, passivating [Al-O] layer is critical for promoting extensive AlN oxidation. While simply increasing the temperature can overcome this passivation, the associated energy cost presents a major practical limitation. Therefore, employing chemical additives capable of disrupting or modifying the [Al-O] oxide layer at lower temperatures offers a more promising and energy-efficient strategy for SAD treatment.

Effect of different additives on AlN oxidation

In this section, the effects of various additives on AlN oxidation were examined, including Na₃AlF₆, Na₂CO₃, CaCO₃, NaOH, CaO, and AlF₃. These additives were selected for their potential to interact with alumina. All AlN samples were roasted in air at 900 °C for 150 min.

The results (Fig. 4a) demonstrate the high thermal stability of AlN, with only 19% oxidation achieved after 150 min of roasting at 900 °C in the absence of additives. XRD analysis (Fig. 4b) confirms the formation of Al₂O₃ in the roasted pure AlN, indicating its partial oxidation. SEM‑EDS results (Fig. 4c) reveal a heterogeneous oxygen distribution on the sample surface, ranging from 0.9% to 11.9%. TEM analysis (Fig. 4d) further corroborates the formation of a 20 ~ 40 nm thick [Al–O] oxide layer on the AlN surface, which is consistent with the computational findings regarding surface passivation.

Fig. 4.

(a) AlN oxidation rates with different additives, (b) XRD patterns of the roasted solids, (c) SEM images of roasted AlN, and (d) TEM images of roasted AlN. (Conditions: 900 °C, 150 min).

Among the additives tested, Na₂CO₃, CaCO₃, NaOH, and CaO showed minimal promotion of AlN oxidation. The addition of KF increased the AlN oxidation rate to 63%. Notably, the fluoride additives NaF, AlF₃, and Na₃AlF₆ exhibited significantly higher efficacy, achieving oxidation rates of 78%, 86%, and 99%, respectively.

XRD analysis identified Al₂O₃ as the predominant phase in AlN samples roasted with NaF, AlF₃, and Na₃AlF₆. Traces of residual AlN were detected in samples treated with NaF and AlF₃, whereas AlN was completely absent in the sample treated with Na₃AlF₆. These observations are in agreement with nitrogen‑content analysis of the roasted products, confirming that fluoride additives—and particularly Na₃AlF₆ (cryolite)—are highly effective in promoting the oxidation of AlN.

Fluorides promote denitrification of SAD

It is indicated that NaF, AlF₃, and Na₃AlF₆ promote the oxidation of pure AlN. To investigate whether they exhibit a similar effect on the AlN within SAD, the time‑dependent denitrification of SAD was examined at 600, 700, 800, and 900 °C using 10 wt% of each fluoride additive.

The results (Fig. 5) show that all three fluorides significantly enhance SAD denitrification compared to the untreated sample. At 600 °C, NaF demonstrated higher effectiveness than AlF₃ and Na₃AlF₆, while Na₃AlF₆ performed better at temperatures ≥ 700 °C. Specifically, with 10 wt% Na₃AlF₆, 63% and 67% of the AlN in SAD was oxidized after 60 and 150 min at 700 °C, respectively. At 800 °C, the denitrification rate increased to 80% (60 min) and 83% (150 min). At 900 °C, only 21% of AlN remained after 30 min, and extending the roasting time to 150 min raised the oxidation level to 89%. For comparison, 150‑min roasting with NaF and AlF₃ at 800 °C yielded denitrification rates of 73% and 77%, respectively; the corresponding values at 900 °C were 80% and 82%.

Fig. 5.

​ Denitrification efficiency of SAD as a function of roasting time and temperature with different fluoride additives (10 wt%).

These findings confirm that NaF, AlF₃, and Na₃AlF₆ effectively promote AlN oxidation in SAD. Although increasing the roasting temperature and duration enhances denitrification, the majority of oxidation occurs within the first 60 min, with prolonged roasting providing only marginal improvement. Therefore, from both economic and practical standpoints, roasting at 800 °C for 60 min is identified as the optimal condition for SAD denitrification.

The effect of cryolite (Na₃AlF₆) dosage was also evaluated at 800 °C for 60 min (Figure S5, SM). The denitrification efficiency increased with cryolite addition, reaching 81% at 6 wt%. Further increases in dosage yielded only slight improvement (e.g., 84% at 18 wt%). Accordingly, a cryolite dosage in the range of 6–10 wt% is recommended as optimal.

Binary fluoride enhance denitrification of SAD

In aluminum electrolysis, NaF, KF, and AlF₃ are commonly used to tailor electrolyte properties, including alumina solubility and liquidus temperature. To optimize the promoting effect of cryolite‑based additives, we investigated binary fluoride mixtures, i.e., NaF–Na₃AlF₆, KF–Na₃AlF₆, and AlF₃–Na₃AlF₆, with varied ratios. These additives were synthesized by melting and homogenizing Na₃AlF₆ with NaF, AlF₃, or KF at predetermined compositions.

For all three binary systems, SAD denitrification efficiency increases with the cryolite content in the additive. However, no significant further improvement is observed once the cryolite content exceeds 6 wt%. In the NaF–Na₃AlF₆ system (Fig. 6a), AlN oxidation increases with higher NaF content, achieving a maximum denitrification of 84% with an additive containing 6–8 wt% NaF and 6 wt% cryolite. For the AlF₃–Na₃AlF₆ system (Fig. 6b), denitrification efficiency first rises and then declines with increasing AlF₃ content, peaking at 89% with a blend of 6 wt% AlF₃ and 6 wt% Na₃AlF₆. The KF–Na₃AlF₆ system (Fig. 6c) shows a steady improvement in denitrification with KF content, attaining the highest efficiency of 93% using 6 wt% KF combined with 6 wt% Na₃AlF₆. This represents the highest denitrification performance observed in the present study.

Fig. 6.

​ Relationship between the molecular ratio of the binary additive and AlN oxidation in SAD: (a) NaF–Na₃AlF₆, (b) AlF₃–Na₃AlF₆, (c) KF–Na₃AlF₆.

These findings demonstrate that cryolite‑based binary fluorides are more effective for SAD denitrification than pure cryolite. Employing the KF–Na₃AlF₆ system (total addition: 12 wt%) under optimized conditions (800 °C, 60 min) resulted in the oxidation of 93% of the AlN in SAD. Compared to previously reported methods—for instance, roasting at 750 °C with 18 wt% Na₃AlF₆ for 194 min¹⁵—the present approach substantially reduces the processing time without requiring a significant increase in temperature, thereby markedly improving the overall efficiency of SAD treatment.

Leaching characteristic of RSAD

To assess the environmental stability and resource‑recovery potential of RSAD, leaching experiments were conducted under acidic (HCl, pH 1), neutral (H₂O, pH 7.3), and alkaline (NaOH, pH 13) conditions. The RSAD sample was prepared by roasting SAD with 6 wt% Na₃AlF₆ + 6 wt% KF at 800 °C for 60 min.

XRD analysis (Fig. 7a) confirmed the disappearance of AlN and metallic Al phases in RSAD, with Al₂O₃ and MgO·Al₂O₃ remaining as the primary components. Compositional analysis (Table 2) showed negligible nitrogen content. The slight decrease in aluminum content in RSAD is attributed to the oxidation of AlN and metallic Al. F and Cl contents were measured at 1.8 wt% and 0.5 wt%, respectively. An estimated 66.8% of F and 27.1% of Cl originally present in SAD (including additives) volatilized during roasting, indicating flue gas needs to be further processed to mitigate secondary pollution. We recommend that SAD be processed in aluminum smelters, where existing flue‑gas treatment facilities can be utilized to minimize cost. In addition, the contents of Na and K increased due to the addition of Na₃AlF₆ and KF, while other components remained largely unchanged.

Fig. 7.

Composition and phase analysis of RSAD and leaching residues. (a) XRD analysis of leach residues in acidic (-A), neutral (-W), and alkaline (-B) solutions. (b) Composition comparison of SAD and RSAD.

Table 2.

Major components in the leachates of SAD and RSAD. ^a (unit: mg/L).

SAD-W	Na	Al	K	S	F	Cl	Si	Cu	Ca	Fe	Zn	Mg	Mn
	845.41	113.83	85.24	184.03	45.52	882.56	146.52	0.12	0.53	0.04	—	—	—
SAD-B	—	1543.31	79.56	289.04	63.79	885.12	157.94	0.93	0.19	0.12	0.12	—	—
SAD-A	974.62	3.32	95.52	26.09	5.88	—	102.02	—	548.82	—	—	38.03	0.12
RSAD-W	544.02	30.08	131.43	170.92	16.24	485.34	6.01	—	0.23	—	—	0.11	—
RSAD-B	—	2105.59	239.93	195.88	340.53	520.71	20.28	0.19	0.13	0.13	—	—	—
RSAD-A	1513.03	300.20	801.77	94.01	5.67	—	25.69	106.77	480.18	5.88	7.43	28.59	10.58

^a SAD and RSAD were leached in pH 1 HCl solution (-A), neutral water (-W), and pH 13 NaOH solution (-B) under conditions of 50 °C, solid-to-liquid ratio of 1:10, and duration of 24 h.

Leaching tests were performed at 50 °C with a solid‑to‑liquid ratio of 1:10 for 24 h. In SAD leachates, the hydrolysis of AlN caused a substantial pH rise and significant gas generation. Final pH values reached 9.6, 12.3, and 14.1 in acidic, neutral, and alkaline solutions, respectively, with corresponding gas generation of 30.92 mL/g, 105.71 mL/g, and 154.43 mL/g. In contrast, RSAD leachates showed minimal pH variation, with final values of 4.4 (HCl), 8.3 (H₂O), and 13.2 (NaOH), and negligible gas generation (< 1 mL/g). These results confirm that roasting effectively oxidized AlN and metallic Al, thereby drastically reducing the release of hazardous gases and associated environmental risks.

The main soluble components in SAD and RSAD leachates (Table 2) included Al, Na, K, S, Si, F, and Cl. Under acidic leaching, notable amounts of Cu, Ca, and Mg were also detected. In neutral solution, the concentrations of Al, Si, F, and Cl in RSAD leachate were significantly lower than those in SAD leachate: aluminum decreased from 113.83 mg/L to 30.1 mg/L, silicon from 146.52 mg/L to 6.01 mg/L, fluorine from 45.52 mg/L to 16.24 mg/L, and chlorine from 882.56 mg/L to 485.34 mg/L. XRD patterns (Fig. 7a) revealed that during water leaching, Al and AlN phases in SAD disappeared and Al(OH)₃ formed as a reaction product. The leaching residue of RSAD retained a phase composition similar to the original RSAD, with minor amounts of Al(OH)₃ and nepheline (Na₃KAl₄Si₄O₁₆) detected. Although the absence of AlN and metallic Al substantially reduces the processing challenges of RSAD, potential environmental risks from inorganic‑salt leaching (e.g., Na⁺ and Cl⁻) remain, especially in humid environments.

In alkaline solutions, the Al concentrations in both SAD and RSAD leachates were markedly higher than under neutral conditions, reaching 1543.31 mg/L and 2105.59 mg/L, respectively. The greater Al leaching from RSAD compared to SAD under alkaline conditions is likely due to the dissolution of non‑α‑phase Al₂O₃ formed during roasting. Although roasting improves overall aluminum recovery, a considerable portion remains in the RSAD leaching residues. XRD analysis identified Al₂O₃ as the dominant phase in these residues, presumably the refractory α‑Al₂O₃.

In acidic solutions, Al concentration in SAD leachate was negligible (3.32 mg/L) because most HCl was consumed by AlN hydrolysis. Conversely, RSAD leachate contained 300.20s mg/L Al. Minor metal oxides in RSAD dissolved under acidic conditions, leading to elevated concentrations of Cu, Ca, Mg, Zn, and Mn in both SAD and RSAD leachates. XRD showed minimal changes in SAD leaching residues, while the RSAD residue resembled the original RSAD with small amounts of Al(OH)₃ present.

Owing to its high aluminum content, soluble fluorine/chlorine species, and minimal gas evolution in aqueous media, RSAD is more appropriately classified as a recyclable resource from which valuable constituents can be recovered via hydrometallurgical routes. Compared with acid leaching, alkaline leaching achieves higher extraction efficiencies for Al, F, and Cl while simultaneously limiting the dissolution of other metallic elements, thus representing a more suitable post‑treatment strategy for RSAD. Nevertheless, a key challenge remains: the low reactivity of the Al₂O₃ present in RSAD. Activating this inert Al₂O₃ phase to further increase aluminum concentration in the leachate is essential for enabling effective SAD recycling. Future research should integrate physical and chemical intensification strategies to enhance aluminum leaching efficiency from RSAD.

Mechanism of cryolite-based additives promoting AlN oxidation

Both theoretical and experimental results indicate that the [Al–O] layer formed on the AlN surface during oxidation inhibits further reaction. However, the mechanism by which cryolite‑based additives promote this process remains unclear. To gain deeper insight, roasting experiments were performed on AlN, AlN + 3 wt% Na₃AlF₆, AlN + 6 wt% Na₃AlF₆, and AlN + 6 wt% Na₃AlF₆ + 6 wt% KF at 800 °C for 60 min, followed by SEM analysis of the roasted solids.

The roasted pure AlN exhibited a dense surface morphology (Fig. 8a). With the addition of 3 wt% Na₃AlF₆, minor surface cracks appeared (Fig. 8b). When 6 wt% Na₃AlF₆ (Fig. 8c) or 12 wt% binary fluoride (6 wt% Na₃AlF₆ + 6 wt% KF) (Fig. 8d) was added, a fragmented, layered morphology was observed. This suggests that Na₃AlF₆ or the binary KF–Na₃AlF₆ mixture facilitates the transformation of the surface oxide from a dense layer into a highly cracked structure during oxidation, thereby enhancing AlN oxidation.

Fig. 8.

SEM analysis of the roasted (a) AlN, (b) AlN + 3wt.% Na₃AlF₆, (c) AlN + 6wt.% Na₃AlF₆, (d) AlN + 6wt.% Na₃AlF₆ + 6wt.% KF at 800 °C for 60 min and (e) in-situ XRD analysis of AlN + 6wt.% Na₃AlF₆ + 6wt.% KF in 25–800 °C.

In‑situ XRD analysis of AlN + 12 wt% KF–Na₃AlF₆ (6 wt% Na₃AlF₆ + 6 wt% KF) over 25–800 °C (Fig. 8e) showed no significant phase change below 600 °C. At 800 °C, new phases corresponding to (Al₂O₃)₆·Na₂O and (Al₂O₃)₆·K₂O were detected. These compounds are ascribed to β‑Al₂O₃, which possesses a structure of alternating closely‑packed slabs and loosely‑packed layers³³, consistent with the SEM observations.

We conclude that Na₃AlF₆ promotes the conversion of AlN oxidation products into Na‑β‑Al₂O₃, rather than a dense α‑Al₂O₃ layer³⁵. The fragmented morphology of β‑Al₂O₃ allows O₂ to access unreacted AlN, thus accelerating oxidation. The introduction of KF leads to partial substitution of Na⁺ by K⁺ in the β‑Al₂O₃ structure. Because K⁺ has a larger ionic radius, K‑β‑Al₂O₃ exhibits expanded interlayer spacing and weaker interlayer interactions compared to Na‑β‑Al₂O₃. Consequently, the binary KF–Na₃AlF₆ mixture performs better than Na₃AlF₆ alone.

Experiments with AlN and Na₂CO₃ did not significantly enhance AlN oxidation, suggesting that β‑Al₂O₃ formation does not arise simply from the reaction between Na₂O and Al₂O₃. Previous studies indicate that β‑Al₂O₃ formation from Na₂O and Al₂O₃ typically requires higher temperatures³⁴. Therefore, we propose that the generation of β‑Al₂O₃ is likely linked to low‑temperature eutectic points in the NaF–KF–AlF₃ system, such as 36.3 mol% NaF, 62.7 mol% KF, 1.0 mol% AlF₃ (t = 711.2 °C) and 51.9 mol% NaF, 27.4 mol% KF, 20.7 mol% AlF₃ (t = 734.5 °C)³⁵. Similar eutectic points (t ≈ 741 °C) exist in the Na₃AlF₆–AlF₃ system³⁶. This aligns with the report by Chuikin et al.³⁷ that AlN‑based sintered materials are not corroded by cryolite melts near 700 °C.

To date, the precise mechanism by which cryolite‑based additives promote β‑Al₂O₃ formation still requires further investigation, for example, through large‑scale ab initio molecular‑dynamics simulations coupled with in‑situ characterization. Elucidating this mechanism is crucial for advancing the resource utilization of SAD.

Conclusions

Through a combined theoretical and experimental study on the oxidative removal of AlN from SAD, we propose an efficient denitrification method that enables high denitrification efficiency. The main conclusions are summarized as follows:

(1) Theoretical calculations and experimental results demonstrate that the oxidation of AlN leads to the formation of a dense, well‑ordered [Al–O] oxide layer on the surface, which inhibits further oxidation of AlN.

(2) NaF, AlF₃, and Na₃AlF₆ effectively promote SAD denitrification. Among them, Na₃AlF₆ exhibits particularly superior performance at temperatures above 700 °C. Roasting SAD at 800 °C with 6 wt% Na₃AlF₆ for 60 min achieves a denitrification efficiency of 81%.

(3) Binary fluoride additives, especially KF–Na₃AlF₆, significantly enhance SAD denitrification. At 800 °C, roasting SAD with 12 wt% KF–Na₃AlF₆ (6 wt% Na₃AlF₆ + 6 wt% KF) for 60 min results in the oxidation of 93% of the AlN.

(4) The RSAD shows markedly reduced gas‑generation reactivity; however, it still poses a risk of inorganic‑salt leaching. Aluminum leaching is more pronounced in alkaline solutions than in acidic media, although most of the aluminum remains in the residue.

(5) The mechanism by which Na₃AlF₆‑based additives promote AlN oxidation likely involves facilitating the transformation of the dense [Al–O] oxide layer into β‑alumina.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

**Tianshuang Li** : Conceptualization, Formal analysis, Investigation, Methodology, Software, Supervision, Visualization, Writing – original draft; **Zhaohui Guo** : Formal analysis, Project administration, Supervision, Validation; **Hong Qin** : Data curation, Formal analysis, Methodology, Resources, Software, Validation, Writing – review & editing; **Huimin Xie** : Data curation, Formal analysis, Software, Visualization, Writing – review & editing.

Data availability

The datasets used and/or analysed during the current study available from the corresponding author on reasonable request.

Declarations

Competing interests

The authors declare no competing interests.