The nature of Cu bonding to natural organic matter

Abstract

Copper biogeochemistry is largely controlled by its bonding to natural organic matter (NOM) for reasons not well understood. Using XANES and EXAFS spectroscopy, along with supporting thermodynamic equilibrium calculations and structural and steric considerations, we show evidence at pH 4.5 and 5.5 for a five-membered Cu(malate)₂-like ring chelate at 100–300 ppm Cu concentration, and a six-membered Cu(malonate))_1–2-like ring chelate at higher concentration. A “structure fingerprint” is defined for the 5.0–7.0 Å⁻¹ EXAFS region which is indicative of the ring size and number (i.e., mono- vs. bis-chelate), and the distance and bonding of axial oxygens (O_ax) perpendicular to the chelate plane formed by the four equatorial oxygens (O_eq) at 1.94 Å. The stronger malate-type chelate is a C₄ dicarboxylate, and the weaker malonate-type chelate a C₃ dicarboxylate. The malate-type chelate owes its superior binding strength to an –OH for –H substitution on the α carbon, thus offering additional binding possibilities. The two new model structures are consistent with the majority of carboxyl groups being clustered and α-OH substitutions common in NOM, as shown by recent infrared and NMR studies. The high affinity of NOM for Cu(II) is explained by the abundance and geometrical fit of the two types of structures to the size of the equatorial plane of Cu(II). The weaker binding abilities of functionalized aromatic rings also is explained, as malate-type and malonate-type structures are present only on aliphatic chains. For example, salicylate is a monocarboxylate which forms an unfavorable six-membered chelate, because the OH substitution is in the β position. Similarly, phthalate is a dicarboxylate forming a highly strained seven-membered chelate.

Five-membered Cu(II) chelates can be anchored by a thiol α-SH substituent instead of an alcohol α-OH, as in thio-carboxylic acids. This type of chelate is seldom present in NOM, but forms rapidly when Cu(II) is photoreduced to Cu(I) at room temperature under the X-ray beam. When the sample is wet, exposure to the beam can reduce Cu(II) to Cu(0). Chelates with an α-amino substituent were not detected, suggesting that malate-like α-OH dicarboxylates are stronger ligands than amino acids at acidic pH, in agreement with the strong electronegativity of the COOH clusters. However, aminocarboxylate Cu(II) chelates may form after saturation of the strongest sites or at circumneutral pH, and could be observed in NOM fractions enriched in proteinaceous material. Overall, our results support the following propositions:

• The most stable Cu–NOM chelates at acidic pH are formed with closely-spaced carboxyl groups and hydroxyl donors in the α-position; oxalate-type ring chelates are not observed.

• Cu(II) bonds the four equatorial oxygens to the heuristic distance of 1.94 ± 0.01 Å, compared to 1.97 Å in water. This shortening increases the ligand field strength, and hence the covalency of the Cu–O_eq bond and stability of the chelate.

• The chelate is further stabilized by the bonding of axial oxygens with intra- or inter-molecular carboxyl groups.

• Steric hindrances in NOM are the main reason for the absence of Cu–Cu interactions, which otherwise are common in carboxylate coordination complexes.

Introduction

The biogeochemistry of copper is largely controlled by its interactions with natural organic matter (NOM), not so much because of NOM’s abundance and polyfunctional character, as because of its remarkable affinity towards Cu(II) relative to other divalent cations (McLaren et al., 1983, Hering and Morel, 1988, Town and Powell, 1993, Ramos et al., 1994, Benedetti et al., 1996, McBride et al., 1997, Leenheer et al., 1998, Kinniburgh et al., 1999, Kogut and Voelker, 2001, Santos-Echeandia et al., 2008). The strong binding ability of NOM for Cu(II) likely results from the excellent match in size between the cupric ion and one or several ligands. The copper ligands must be well-defined structurally and chemically, and also numerous because NOM has a high sorption capacity and selectivity for Cu(II) over a large concentration range (Gao et al., 1997, Covelo et al., 2004). Since carboxylate moieties comprise the large majority of reactive sites below pH 7 in NOM, and can form a vast number of synthetic coordination complexes with Cu(II) (Melnik et al., 1998a, Melnik et al., 1998b, Melnik et al., 1999), the strongest bonds are expected to be with carboxyl ligands (Sposito et al., 1979, Boyd et al., 1981). The predominance of oxygen ligands does not exclude other electron donors from being involved in Cu bonding, such as nitrogen as suggested by electron spin resonance (ESR) spectroscopy (Boyd et al., 1983, Senesi and Sposito, 1984, Senesi et al., 1985, Luster et al., 1996). Based on solution chemistry the most likely ligands are dicarboxylate (malonate, log K = 5.04), mixed alcohol-carboxylate (citrate, malate, log K = 3.70 and 3.63), and aminocarboxylate (glutamate, log K = 8.32) groups attached to aliphatic chains (Fig. 1, Table 1; Gregor et al., 1989a, Gregor et al., 1989b, Town and Powell, 1993, Croué et al., 2003). Aromatic dicarboxylate (phthalate, log K = 3.22), alcohol-carboxylate (salicylate, log K = 2.22 for pH 5), and furan-carboxylate (furanate, log K = 1.10) ligands are less likely, because they have weaker binding strengths than aliphatic ligands. This interpretation is reinforced by infrared spectroscopy and two-dimensional NMR which showed that carboxylate moieties with –COOH, –OH, or –OR substituents on the α carbon from aliphatic chains or alicyclic structures constitute the majority of carboxyl structures, with the salicylate and furan-carboxylate aromatic structures being less reactive (Deshmukh et al., 2007, Hay and Myneni, 2007). Therefore, malonate, citrate/malate, and amino acid structures most likely play the important roles in Cu(II) chelation. Although stability constants can be used to predict the a priori nature of Cu(II)–NOM complexes, they do not provide unequivocal information on the true bonding environment of the metal in natural systems. They are, however, extremely valuable for making educated choices on the relevance of model compounds used in structural studies.

Using XANES and EXAFS spectroscopy, and Cu–glutamate as the best-fit structural analog, Karlsson and Skyllberg (2006) showed that Cu(II) forms a five-membered chelate ring with one amino nitrogen (α-NH₂) or alcohol oxygen (α-OH) and one carboxylate oxygen from an α-substituted aliphatic carboxylic structure (Fig. 1). Copper is bridged to four O/N ligands (CuL2 complex) in a square-planar geometry at low metal concentration, and to two (CuL complex) at higher concentration. Today the glutamate-like model is the most detailed, but despite its own merit the topic is far from being exhausted. Here are some remaining open questions, which are addressed in this article:

• Is it possible to distinguish oxygen from nitrogen ligands, at least at low Cu concentration when heterogeneity is minimized?

• What are the limits of the glutamate model? Does it always provide an adequate fit to EXAFS data? Since NOM contains several well-defined types of carboxyl structures with specific protonation constants, the coordination mode of Cu(II) should vary with pH, metal to ligand ratio, and chemical composition of the NOM.

• At medium and high Cu concentration, can the data be modeled by a mixture of discrete ligands, and if so, how many are needed?

Answering these questions rests upon three capabilities (i) to detect small variations of the local structure about Cu atoms up to approximately 4 Å (i.e., second C atomic shell), and over two orders of magnitude in concentration (i.e., Cu/C ratio); (ii) to determine from a large dataset of multicomponent spectra the number of independent patterns (or “principal components”, PCs) that represent the number of identifiable Cu binding environments (i.e., species) present in the set of spectra; (iii) to build a spectral database of references of known identity and structure, which includes all distinct binding environments seen by EXAFS and plausibly present in NOM.

We recorded our data at liquid helium (LHe) temperature to enhance the signal from higher atomic shells by reducing the lability of organics, and by using a high-flux spectrometer at the European Synchrotron Radiation Facility (ESRF, Grenoble) equipped with a 30-element Ge detector for the study of highly diluted samples (concentration range 100–6500 mg/kg). The significant number of independent patterns accounting for 98–99% of variance in the dataset was determined with principal component analysis (PCA, also known as abstract factor analysis, Malinowski, 1991, Wasserman et al., 1999: Ressler et al., 2000, Manceau et al., 2002). Here, linear algebra can be used for quantification because the EXAFS response from a multispecies sample is the weighted-superposition of the response of each species present in a sample (Manceau et al., 1996). The PCs obtained from the PCA are not real spectra (i.e., they are abstract components), but the single species spectra which make up the multicomponent spectra in the dataset are linear combinations of PCs. Thus, the spectra of all unknown species contained in a sample can be identified from a database by target transformation, provided the unknown is present in the library of reference spectra. Care has been taken to include in the database all plausible Cu(II) species discussed in the literature and to not miss any major species. With such an extensive and representative database, target testing goes beyond the usual fingerprinting approach between known and unknown spectra, because the entire dataset is analyzed at one time in a statistically meaningful way for similarity to a specific structural reference.