Stabilization of Fully Deprotonated Melaminate Anions (C₃N₆)^6– in M₃(C₃N₆) (M = Cd, Ca)

Abstract

Hydrogen-free melaminate salts M₃(C₃N₆) (M = Cd, Ca) were synthesized in laser-heated diamond anvil cells at 34–48 GPa and 2000–2500 K. Cd₃(C₃N₆) was synthesized via a direct reaction between the elements, while Ca₃(C₃N₆) was obtained following a rational chemical design approach from calcium carbodiimide, Ca­(NCN), which served as a single-source precursor. Both compounds contain the fully deprotonated melaminate anion (C₃N₆)^6–, representing a fundamental milestone in nitridocarbonate chemistry. The crystal structures of M₃(C₃N₆) were solved and refined using synchrotron single-crystal X-ray diffraction data and were fully corroborated by density functional theory calculations. Cd₃(C₃N₆) crystallizes in the acentric R3c and Ca₃(C₃N₆) in the centrosymmetric R3̅c space groups, and both compounds are recoverable to ambient conditions. Extending this design principle, our calculations indicate that Zn and Pb melaminates are thermodynamically accessible under similar conditions, highlighting the general stability of hydrogen-free nitridocarbonates of selected divalent metals.

Introduction

Molecular melamine (C₃H₆N₆) and its derivative compounds serve as fundamental building blocks in the synthesis of melamine-formaldehyde resins or flame-retardant materials. , Moreover, it is one of the main precursors for producing graphitic carbon nitride (g-C₃N₄), which is studied for its high-performance photocatalysis and optoelectronics applications. The physicochemical properties of g-C₃N₄ can be tuned by incorporating different metals. − For instance, lithium or potassium ions can act as structure-directing agents during the formation of the porous frameworks, influencing photocatalytic activity. , Despite a variety of synthetic routes to g-C₃N₄, the ideal composition C₃N₄ is almost never reached, but instead, one finds partially amorphous phases, often contaminated with hydrogen and oxygen; likewise, there is no convincing single-crystal structure model yet. A rational design route toward pure g-C₃N₄ is to start by synthesizing its hydrogen-free building blocks. The most straightforward metathetic approach to substituted hydrogen-free nitridocarbonates is well established for acidic precursors (e.g., HCN) but becomes increasingly more difficult for basic precursors containing larger numbers of hydrogen atoms.

For example, in the case of guanidine CNH­(NH₂)₂, usually only single or double deprotonation can be achieved, yielding compounds M­(CN₃H₄)₂ (M = Eu, Ba) , and MC­(NH)₃ (M = Ca, Sr, Eu, Yb). − The first examples of stabilizing the fully deprotonated guanidinate anion (CN₃)^5– were reported in compounds SbCN₃  and Ln₃O₂(CN₃) (Ln = La, Eu, Gd, Tb, Ho, Yb). In both cases, laser-heated diamond anvil cells (LHDACs) were used to synthesize and study the reaction products in situ, revealing the importance of hydrogen-free synthesis conditions. Stelzer et al. have produced deprotonated (CN₃)^5– stabilized in (Sr₉N_1.33(8))­(SrIn₃)­[CN₃] and Sr₄(Sr₆N)₂­[In₄]­[CN₃]₄ compounds without application of nitrogen pressure from a sodium flux.

Complete deprotonation of the next members of the nitridocarbonate series becomes increasingly challenging. In 1922, Franklin had reported the synthesis of melaminate salts KC₃N₆H₅·NH₃ and K₃C₃N₆H₃ in liquid ammonia from melamine and potassium amide, based on elemental analysis. The confirmation of single-deprotonated melaminates via single-crystal X-ray diffraction (scXRD) was reported by Görne et al. in 2021 with the potassium and rubidium compounds KC₃N₆H₅·NH₃ and RbC₃N₆H₅·¹/₂NH₃. Based on IR spectroscopy, they also obtained NaC₃N₆H₅·nNH₃ and K₃C₃N₆H₃, which was the proposed reaction product by Schnick et al. in 1995 as well. , The melaminate anion (C₃H₃N₆)^3– was reported in 2021 by Kallenbach et al. with the synthesis of a Metal–Organic Framework containing dehydrogenated melamine. The synthesis of tricopper­(I) melaminate Cu₃(C₃H₃N₆) was achieved by a solid-state reaction of CuCl with melamine under flowing argon at 275 °C. Recently, Bayat et al. observed the same reaction product by a reaction of CuCl and sodium hydrogen cyanamide Na­(HCN₂), forming Cu₃(C₃N₆H₃) and NaCl. Double-deprotonated melaminate was synthesized in 2023, also by Bayat et al., in a solid-state reaction of antimony­(III) chloride SbCl₃ and melamine, which led to the formation of SbCl­(C₃N₆H₄) and (C₃N₆H₇)­Cl. The summary of synthetic routes to substituted melaminate salts is given in Table S1.

Recently, Chen et al. proposed a synthetic route to the completely deprotonated melaminate salt of the composition WC₃N₆. The metathetic pathway involves a reaction between tungsten oxide (WO₃) and melamine (C₃H₆N₆), forming hydrogen-free tungsten melaminate (WC₃N₆) and gaseous water (H₂O). Calculations show that both predicted polymorphs of WC₃N₆ are indirect semiconductors, with electrical and optical properties, suitable for photocatalysis and optoelectronic devices. During the preparation of this manuscript, the high-pressure synthesis of hydrogen-free lead melaminates hP72-Pb₃(C₃N₆) and tP48-Pb₃(C₃N₆) has been reported.

Results and Discussion

Here, we present the stabilization of a fully deprotonated melaminate (C₃N₆)^6– anion in a series of salts of divalent metals M₃(C₃N₆) (M = Ca, Cd) in several independent experiments using various precursors (Table ). Cadmium melaminate Cd₃(C₃N₆) was synthesized in a LHDAC in two independent experiments in a pressure range 44–48 GPa (Table ). In experiment #1 the synthesis was performed directly from the elements. A piece of Cd was placed on a diamond culet, and the sample chamber was filled with nitrogen, which served both as a pressure-transmitting medium and as a reactant. The DAC was compressed to the target pressure, and then a focused Nd:YAG laser (λ = 1064 nm) was used to heat the Cd piece. The products were then studied by means of synchrotron X-ray diffraction at the ESRF (ID15b and ID27 beamlines). The Supporting Information Section B provides complete experimental details.

1. Summary of Experimental Conditions for the Synthesis of M₃(C₃N₆) (M = Cd, Ca).

experiment	reagents	pressure, GPa	beamline
#1	Cd + Cdia + N2	48	ESRF ID15b
#2	Cd + C6N4	44	ESRF ID27
#3	Ca(NCN)	34	DESY P02.2, ESRF ID27

The diffraction patterns of the heated samples revealed the presence of numerous single-crystalline grains in the sample chamber, and the data sets were analyzed using the well-established procedures for handling multigrain samples using Domain Auto Finder (DAFi) program. Based on this analysis, several most prominent grains could be indexed with the R-centered hexagonal unit cell with a, b = 11.535(1) Å and c = 5.189(2) Å for one of such grains at 48(1) GPa (see Figure S1 and Tables S2 and S4–S8 for full details). The structure solution and refinement revealed the chemical formula of the new compound as Cd₃(C₃N₆) crystallizing in space group R3c (Figure ). In addition to this compound, we also found cadmium diazenide CdN₂, which will be reported elsewhere. To enhance the synthetic strategy and avoid the formation of CdN₂, we used tetracyanoethylene (C₆N₄) in experiment #2 as a precursor of carbon and nitrogen for the synthesis of Cd₃(C₃N₆). In this experiment, C₆N₄ also served as a pressure-transmitting medium (Figure S2). Laser-heating at 44(1) GPa once again led to the formation of Cd₃(C₃N₆), as confirmed by scXRD as well as Raman spectroscopy (Figure ). The sample was decompressed stepwise, allowing the collection of scXRD data at each pressure point down to 30(1) GPa (Figure ). Powder XRD from Cd₃(C₃N₆) crystallites could be traced down to atmospheric pressures (Figure S3), indicating recoverability to ambient conditions.

1.

Powder X-ray diffraction and Raman spectroscopy data for Cd₃(C₃N₆). (a) Le Bail fit from multigrain Cd₃(C₃N₆) at 38(2) GPa compared with calculated peak positions for Cd₃(C₃N₆) and CdN₂. Structure refinements and phase identification were based on single-crystal data sets. (b) Raman spectrum of Cd₃(C₃N₆) at 44(1) GPa compared with calculated Raman spectrum of Cd₃(C₃N₆). Calculated Raman spectrum frequencies were scaled by a factor of 1.007, and intensities were normalized using the most intense peak as a reference.

2.

Compressional behavior of M₃(C₃N₆) (M = Cd, Ca). Shown are the experimental (black and red stars) and calculated (black and red dashed line) unit cell volumes of M₃(C₃N₆). The 3rd order Birch–Murnaghan equation of state was used to determine the bulk moduli of M₃(C₃N₆) based on the calculated data.

The crystal structure of Cd₃(C₃N₆) contains one crystallographically independent Cd, one C, and two N atoms all occupying Wyckoff sites 18e (see Tables S2 and S4–S8 for complete refinement details). The main structural feature of the compound is a melaminate anion (C₃N₆)^6– as shown in Figure . The anion is slightly distorted out-of-plane and there are three inequivalent, but similar within the standard uncertainties, C–N distances in the range 1.32–1.35 Å, demonstrating a significant degree of π-electron delocalization, which is consistent with protonated melamine itself. Melaminate anions have two resonance form types: one where the aromatic system in the ring persists and the nitrogen atoms of the amine groups are sp³ hybridized, and another where one or more of the amine nitrogen atoms donate a lone pair and become sp² hybridized to create a double bond and thus become part of the conjugated system. Note that the sp³ and sp² designators refer to the simplistic valence-bond model.

3.

Crystal structures of Cd₃(C₃N₆) at 38(2) GPa and Ca₃(C₃N₆) at 34.4(10) GPa. (a) A view along the crystallographic c-axis of Cd₃(C₃N₆). (b) A view along the crystallographic c-axis of Ca₃(C₃N₆). (c) Melaminate anion (C₃N₆)^6–. x, y, and z denote distinct bond distances presented in Table . (d) A view along the crystallographic b-axis of Cd₃(C₃N₆). (e) A view along the crystallographic b-axis of Ca₃(C₃N₆).

2. Experimental and Calculated C–N Bond Distances within the Melaminate Anions .

compound	C–N bond length (x), Å	C–N bond length (y), Å	C–N bond length (z), Å	pressure (GPa)
Cd3(C3N6)	1.388[a]	1.338[a]	1.380[a]	0.0001
	1.347[a]	1.318[a]	1.339[a]	38
	1.344(33)	1.322(21)	1.347(31)	38(2)
Ca3(C3N6)	1.396(6)	1.329(10)	1.397(8)	0.0001
	1.353[a]	1.302[a]	1.353[a]	35
	1.353(5)	1.303(3)	1.352(5)	34.4(10)

^aNotes: ^[a] marks calculated bond lengths. The geometry of the anion and labeling of the C–N bonds (x, y, z) are shown in Figure e.

The melaminate anion (C₃N₆)^6– perfectly satisfies the charge balance in Cd₃(C₃N₆), where Cd exhibits an oxidation state of +II. The empirical formula of cadmium melaminate, CdCN₂, corresponds to the empirical formula of cadmium carbodiimide. This stoichiometric match suggests that single-source carbodiimide precursors could be used for the production of melaminate salts. To prove this hypothesis, we used Ca­(NCN) as a single-source precursor for the synthesis of Ca₃(C₃N₆). In this experiment, phase-pure Ca­(NCN) was loaded in a DAC without any pressure-transmitting medium and laser-heated at 34.4(10) GPa using a CO₂ laser (λ = 10 600 nm) (Table , Figure S4). The diffraction patterns and additional Raman spectra were collected and analyzed in a manner similar to that for Cd₃(C₃N₆) (Figures and S5). Structure solution and refinement revealed the formation of calcium melaminate Ca₃(C₃N₆) crystallizing in the centrosymmetric space group R3̅c (see Tables S3 and S9, S10 for full details). The difference between the crystal structures of Ca₃(C₃N₆) and Cd₃(C₃N₆) is the rotation of melaminate groups with respect to each other in the neighboring layers (Figure ), as well as the significant degree of out-of-plane distortion of melaminate anions in the latter compound (Figure ). According to our density functional theory (DFT) calculations and experimental results, the relative rotation of melaminate anions in Cd₃(C₃N₆) correlates with both the out-of-plane distortion of the melaminate units and the applied pressure (Figure S6). It should be noted that we have performed structure solution and refinement in both space groups (R3̅c and R3c) for both Ca₃(C₃N₆) and Cd₃(C₃N₆) compounds. In the case of Cd₃(C₃N₆), it was not possible to obtain a reasonable structure refinement in a centrosymmetric R3̅c space group. In the case of Ca₃(C₃N₆), the refinements in R3c and R3̅c resulted in nearly identical agreement factors, which allows us to prefer the R3̅c model. This assignment is also supported by our DFT calculations and the crystallographic structure validation algorithms.

4.

Powder X-ray diffraction and Raman spectroscopy data. (a) Le Bail fit from multigrain Ca₃(C₃N₆) at 34.4(10) GPa compared with calculated peak positions for Ca₃(C₃N₆). Structure refinements and phase identification were based on single-crystal data sets. (b) Raman spectrum of Ca₃(C₃N₆) at 34.4(10) GPa compared with calculated Raman spectrum of Ca₃(C₃N₆). Calculated Raman spectrum frequencies were scaled by a factor of 1.008, and intensities were normalized by using the most intense peak as a reference.

5.

(a) Ring distortion of melaminate anions (C₃N₆)^6– in M₃(C₃N₆) (M = Cd, Ca). The distortion is indicated by the angle φ between the CNNN and C′N′N′N′ planes as shown in (b). At the highest-pressure point, the atom quartets defining the planes (CNNN and C′N′N′N′) become nonplanar themselves, which might indicate further structural distortion.

To get a deeper insight into the electronic structures of the synthesized compounds, we performed theoretical calculations within the framework of the plane-wave DFT which was then unitarily transformed to local orbitals (LOBSTER) to allow for chemical-bonding analysis. − Full details are provided in the Supporting Information. First, geometry-optimized crystal structures of the synthesized compounds are in good agreement with the experimental data, and our calculations properly reproduce the distortion of the melaminate groups in Cd₃(C₃N₆). The reason for that distortion is the covalency of Cd–N bonds, which imposes geometric constraints on the rings. Wiberg–Mayer bond orders for the Cd–N bonds range between 0.30 and 0.35, while for the far less covalent, essentially ionic Ca–N bonds, the numbers are smaller, 0.06–0.18. One can draw an analogy with azides: Cd­(N₃)₂ has a distorted N₃ ^– anion while it is less distorted in Ca­(N₃)₂. , Similar anionic distortions have also been observed in covalent carbodiimides and cyanamides, such as Pb­(NCN). There are significantly different anionic C–N distances (1.16 Å and 1.30 Å) in Pb­(NCN), mirroring the greater covalency of Pb–N bonds, while C–N distances are of equal length (1.22 Å) in ionic Ca­(NCN).

For detailed chemical-bonding analysis of Ca­(NCN) and Ca₃(C₃N₆) we recalculated the crystal orbital bonding index (COBI) of all C–N bonds. The energy-resolved COBI results are presented in Figure a,b. The Fermi level (ε_F) nicely separates bonding (below ε_F) and antibonding (above ε_F) levels for Ca­(NCN), but one recognizes a few tiny populated antibonding levels below ε_F in Ca₃(C₃N₆), which explains the necessity of applied pressure to stabilize this compound similar to a variety of high-pressure dinitrides. , The bond order can be quantified by the energy integral of COBI (ICOBI), which is also presented in Figure . In Ca­(NCN), the ICOBI sum reached 11.286 going back to six C–N bonds, so the bond order is 11.286 ÷ 6 = 1.88 ≈ 2, very close to a CN double bond. For Ca₃(C₃N₆) the C–N ICOBI values arrive at 4.236 (for three terminal C–N bonds) and 6.786 (for six in-ring C–N bonds), so the terminal C–N bonds come out stronger (bond order 1.412), and the in-ring C–N bonds are weaker (bond order 1.131), in approximate agreement with the idealized bond orders derived from valence-bond theory (1.44 and 1.28, respectively). The sum over all C–N bonds is 11.022, so covalency has decreased by a small (2%) amount compared to the prior 11.286 for Ca­(NCN). To compensate for that, the Löwdin charges (see Table S21) show that the cationic Ca charge also decreases upon pressure increase, while the cationic C/anionic N charges increase, so opposite effects are at play. Given the smaller volume, however, the calculated Madelung energies based on Löwdin charges and interatomic distances increase from Ca­(NCN) (−6.36 MJ mol^–1) to Ca₃(C₃N₆) (−6.73 MJ mol^–1), so the overall ionicity is significantly enhanced upon forming Ca melaminate (by ca. 6%), making Ca₃(C₃N₆) more salt-like compared to Ca­(NCN).

6.

Crystal orbital bond index (COBI) analysis of C–N bonds in (a) Ca­(NCN) and (b) Ca₃(C₃N₆) under a pressure of 35 GPa, with bonding interactions to the right, antibonding to the left. For easier comparison, the ICOBI of “C₃N₆” entity was calculated, and the ICOBI value for Ca­(NCN) was multiplied by 6 to account for the six CN double bonds in Ca­(NCN) whereas the values for each bond (designated by x, y, and z in Figure e, respectively) in the melaminate anion were weighted by its multiplicity. (c) The fragment molecular orbital diagram of the (C₃N₆)^6– entity embedded in solid-state Ca₃(C₃N₆). The MOs with energies below −12 eV or above 5 eV are omitted.

To further examine the intramolecular orbital (MO) interactions within the C₃N₆ entity, we generated the melaminate MO diagram from C/N atoms and their orbitals as fragments, presented in Figure c. The C₃N₆ entity in solid-state Ca₃(C₃N₆) features a singly degenerate highest occupied molecular orbital (HOMO) and a set of doubly degenerate lowest unoccupied molecular orbitals (LUMOs). The HOMO consists of N 2p _z orbitals (major contributor) and N 2p _x orbitals (minor contributor) located in the deprotonated amino groups of the C₃N₆ entity, whereas the LUMOs are formed by 2p _z orbitals from both carbon and nitrogen atoms. There is a bonding and antibonding character in the HOMO which further cross-validates the conclusion that external pressure must compete with also populating such C₃N₆ antibonding interactions. Although the DFT calculations reproduce the high-pressure structural data of Cd₃(C₃N₆) well (Figure ), a slight discrepancy remains between the ambient-pressure experimental data and the DFT predictions. The latter predicts an isostructural first order phase transition at about 2 GPa (Table S11). It is likely that the transition might be kinetically hindered, and moreover, we have observed significant decrease in crystal quality for Cd₃(C₃N₆) at 30 GPa, which prevented unambiguous single-crystal structure analysis below this pressure.

General crystal chemical knowledge allows us to hypothesize that melaminate salts of selected divalent metals could be thermodynamically stable and synthesizable under moderate pressure. To explore this, we performed structure optimizations for Zn- and Pb-melaminates based on the composition of M₃(C₃N₆) (M = Zn, Pb). For both metals we considered three possible structure models: hR72 from this study as well as hP72 and tP48 reported by Ranieri et al. for Pb₃(C₃N₆). For Zn₃(C₃N₆) an acentric R3c structure is the most thermodynamically stable, while the reported hP72-Pb₃(C₃N₆) and tP48-Pb₃(C₃N₆) structures are thermodynamically more stable than in hR72 polymorph of Pb₃(C₃N₆) featuring R3̅c space group symmetry. For the hR72 polymorphs, the calculations always started in the acentric R3c space group, which converged to R3̅c in the case of Pb melaminate. Calculated crystal structures and enthalpies of Zn- and Pb-melaminates are given in Tables S12–S20.

There are several strategies to direct the synthesis of hydrogen-free nitridocarbonates in a thermodynamically controlled regime in which the compounds can be produced directly from elements. The most critical factor is the choice of the countercation, stabilizing the C–N anions. For example, highly charged guanidinate anions (CN₃)^5– can be stabilized by Sb⁵⁺ in a simple calcite-type SbCN₃. Bigger cations like Bi³⁺/Bi⁵⁺ allow stabilization of polymerized C–N networks in the compound Bi₇C₁₀N₁₈(N_3(1–x)O_3x ). Pressure plays an equally important role; according to the pressure–coordination rule, carbon tends to increase its coordination number from three to four which, within valence bond theory, would be coined going from sp² to sp³ hybridization, e.g., upon compression in carbonates. − Binary carbon nitrides with condensed CN₄ tetrahedra can be synthesized at pressures above 72 GPa, while pressures between 90 and 111 GPa are required for the formation of lanthanoid polynitridocarbonates MCN₃ (M = La, Tb, Ce, Tb). The fully deprotonated form of the ortho-nitridocarbonate anion (CN₄)^8– remains to be discovered but is predicted with tetravalent cations in M₂(CN₄) (M = Ti, Zr, Hf). The selection of suitable single-source precursors could enable the synthesis of nitridocarbonates under milder, kinetically controlled conditions, avoiding the cleavage of C–N bonds in the precursors. A notable example is the synthesis of tricyanomelaminate hydrate Na₃C₆N₉·3H₂O, featuring the (C₆N₉)^3– anion by trimerization of sodium dicyanamide NaC₂N₃.

Conclusions

A series of recoverable melaminate salts, M₃(C₃N₆) (M = Cd, Ca), with fully deprotonated melaminate (C₃N₆)^6– was synthesized under high-pressure, high-temperature conditions. The fact that M₃(C₃N₆) (M = Cd, Ca) phases were obtained with different cations and from different precursors demonstrates the high stability of such compounds and anions. While the trimerization of carbodiimide to melaminate leads to a small weakening of the C–N bonds, the M₃(C₃N₆) phase mirrors an increasing Madelung field or, alternatively expressed, a more salt-like behavior upon pressure increase. Calculations also showed that the synthesis of melaminate salts might be possible with other divalent cations, e.g., zinc and lead. This opens synthetic ways to a series of inorganic nitridocarbonates with fully deprotonated melaminates by the appropriate choice of elements, even in large quantities when considering starting from single-source precursors, such as M­(NCN) (M = Cd, Ca, Zn, or Pb).

Glossary

Abbreviations

LHDAC

laser-heated diamond anvil cell

DFT

density functional theory.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/jacs.5c16752.

• Overview of synthetic routes to melaminate salts; synthesis of M₃(C₃N₆) (M = Cd, Ca); X-ray diffraction studies and programs; structure refinement details of M₃(C₃N₆) (M = Cd, Ca); microscope picture and 2D-Raman map of M₃(C₃N₆) (M = Cd, Ca); decompression by Raman of Ca₃(C₃N₆); calculations of M₃(C₃N₆) (M = Cd, Ca) (PDF)

‡.

P.L.J. and L.B. contributed equally to the final form of the submitted manuscript. All authors reviewed and approved the final form of the submitted manuscript.

M.B. acknowledges the support of Deutsche Forschungs­gemeinschaft (DFG Emmy-Noether Project BY112/2-1). E.B. acknowledges the support of Deutsche Forschungs­gemeinschaft (DFG Emmy-Noether Project BY101/2-1). E.B. and M.B. acknowledge the financial support of Johanna-Quandt Young Academy. M.B. acknowledges the support from the Loewe Start Professorship Program of the State of Hesse and the support from the Adolph-Christ Foundation. We acknowledge DESY (Hamburg, Germany), a member of the Helmholtz Association HGF, for the provision of experimental facilities.

The authors declare no competing financial interest.