Ind. Eng. Chem. Res. 2005, 44, 3229-3237
3229
Development of a Multidimensional High-Performance Liquid Chromatography (HPLC) Separation for Bayer Humic Substances Thelma-Jean Whelan Department of Chemistry, Materials and Forensic Science, University of Technology Sydney, PO Box 123, Broadway NSW, 2007, Australia
R. Andrew Shalliker* Nanoscale Organizational and Dynamics Group, University of Western Sydney, Locked Bag 1797, Penrith South DC NSW, 1797, Australia
Cameron McIntyre CSIRO Petroleum, PO Box 136, North Ryde NSW, 1670, Australia
Michael A. Wilson Deans Unit, College of Science, Technology and Environment, University of Western Sydney, Locked Bag 1797, Penrith South DC NSW, 1797, Australia
Multidimensional high-performance liquid chromatography has been used to analyze the structure of humic substances in Bayer liquor from an alumina separation plant. Separation has been achieved on a Phenomenex BioSep-S2000 size-exclusion column and a Phenomenex Synergi Polar-RP column. The results demonstrate a degree of resolution that previously could not be observed by more-conventional liquid chromatographic methods. In three fractions that were heart-cut from the BioSep-S2000 size-exclusion column in the first dimension, at least 18 components were resolved to baseline resolution in the second dimension, with many other peaks exhibiting less resolution. Analysis of these compounds by mass spectrometry showed that small molecules are present in fractions collected with relatively high apparent molecular weights, with the small molecules being held either in a supramolecular or micelle structure with larger molecules. Evidence is presented here that the latter is more likely. 1. Introduction Bayer humic substances are the sodium hydroxide (NaOH)-soluble geo-organic matter that accumulates during the Bayer process. The humic substances enter the Bayer process during the separation of alumina from iron oxide in bauxite ore via dissolution in concentrated NaOH (3.5-5 M) at high temperatures. Up to 3% of the organic matter associated with the bauxite ends up in the liquor, although some insolubles are removed with the red mud.1,2 The soluble organic species can accumulate in the process liquor as the caustic solution is recycled for the digestion of fresh bauxite after the precipitation of the aluminum hydroxide trihydrate.3 Bayer humic substances are important because of their adverse effect on the industrial scale production of alumina from bauxite. They interfere in the process by affecting the crystallization rates and precipitation of alumina4 and also the stability of sodium oxalate in solution.5 Bayer liquor organics range in molecular weight from 300 000 Da. The low-molecular-weight organic molecules (1200 Da have focused on the measurement of bulk * To whom correspondence should be addressed. E-mail address:
[email protected].
spectroscopic and chemical properties using techniques such as pyrolysis gas chromatography/mass spectroscopy (py GC/MS) and solid-state nuclear magnetic resonance (NMR) spectroscopy.5 Attempts to separate Bayer humic substances using liquid chromatography have been reported in the literature; however, these have had limited success. Most methods of analysis have focused on monitoring simple organic and inorganic ions such as oxalates, chlorides, and sulfates7,8 or have provided a quantitative analysis of the humics present in the Bayer liquor.9 We recently reported separations of Bayer humics using ion-paired reversed-phase high-performance liquid chromatography (HPLC).10 Analysis times of >10 h were required, and even then, limited resolution was achieved as the peak capacity (this is a term referring to the maximum number of peaks that can fit in the available separation space, separated with the minimum stated resolution11) of the separation system was vastly exceeded. This separation highlighted the complexity of the isolation problem and showed that one-dimensional reversedphase HPLC was inadequate to separate such a mixture. Even under the most favorable conditions, the statistical limitations on separating complex samples using a one-dimensional HPLC separation indicate that, to successfully separate such a sample, a multidimensional HPLC approach must be used.12-14 The intent of this study was to improve the separation process for Bayer humic substances by expanding the
10.1021/ie049046h CCC: $30.25 © 2005 American Chemical Society Published on Web 03/16/2005
3230
Ind. Eng. Chem. Res., Vol. 44, No. 9, 2005
peak capacity of the system, using multidimensional HPLC. In multidimensional chromatography, sample components are displaced along more than one axis of separation, as is the case in one-dimensional chromatography.15-17 In such a technique, the separation space is expanded by as much as the product of the peak capacities of each individual separation dimension. Ideally, each dimension of the separation would use a stationary phase that would generate orthogonal retention behavior for the solutes being studied and, as a result, would maximize the peak capacity of the system.17 In reality, at least some retention correlation does exist between the dimensions and this, to some extent, decreases the optimum resolution and peak capacity that could be achieved. In principle, to take full advantage of the expanded separation space afforded by multidimensional systems, the number of system dimensions should equal the number of definable sample attributes.18 That is, if two factors can be used to describe the sample (for example, molecular size and polarity), two separation steps that are designed to take advantage of each sample attribute should be used. Here, we present the results of a multidimensional technique for the isolation of constituents in Bayer humic samples where molecular size and polarity were used to define the separation mechanisms for each dimension. Separations have been performed in a heartcutting mode, using a column switching technique where a portion of the eluent is switched from the first dimension to the second dimension. Using two-dimensional chromatography, we were able to increase the peak capacity of the system, resulting in an increase in resolution. Some of the resolved materials were then further characterized by electrospray ionization (ESI) mass spectrometry. 2. Experimental Section 2.1. Preparation of Humic Substances for Analysis. Isolation of the humic substances from the Bayer process was performed using the method reported by Whelan et al.;10 however, for the sake of clarity, the method is summarized here. Specifically, an aliquot of Bayer liquor (200 mL) from the digestion of bauxite at 145-150 °C at the Kwinana Alcoa refinery in Western Australia was diluted at a volume ratio of ∼1:10 (v/v) with distilled water and acidified to pH 1.5 with a 1:1 (v/v) hydrochloric acid-water (HCl-H2O) mixture. The soluble fulvic acid fraction was decanted and then filtered through a glass microfiber paper (Whatman 15.0 cm GF/C) and filtered again through a sintered glass filter (porosity 4) to remove undissolved particles in the acidified Bayer liquor. The pH of the precipitated humic acid fraction was gradually increased to 4 with NaOH (1 M) to solubilize the humic acid. This solution and the fulvic acid solution were combined and adsorbed onto an XAD-7 resin, as described below. Extraction of the humic substances in the Bayer liquor was then conducted on a prewashed and acidified Amberlite XAD-7 resin column (6 cm × 2 cm, 200 g resin). The humic substances in the acidified Bayer liquor solution (previously described) (1 L) were adsorbed on the column by passing the solution through the XAD-7 column at a flow rate of ∼1 mL/min until the outlet liquor became yellow in color. The column was washed with 2 L of 0.1 M HCl, then with 1 L of doubly distilled deionized water until the pH of the outlet water was neutral. The column was eluted with a 0.1 M
potassium hydroxide (KOH) solution (450 mL) and the concentrated (brown-colored) humic substances were collected (400 mL). The protonated form of the Bayer humic substances was obtained by passing 200 mL of the alkaline humic substances solution through a prewashed (with doubly distilled deionized water) cationexchange resin (Amberlite 120, 60 cm × 2 cm). The humic substances were washed from the column with a one-column volume of doubly distilled deionized water (120 mL), to yield the protonated form (humic and fulvic acids). The procedure was repeated 20 times, using 200mL aliquots of the initial Bayer liquor solution, and the solutions combined to form 2.4 L of combined solution. This solution was filtered through a 0.45 µm Millipore glass fiber filter and 16 aliquots of 60 mL were removed for molecular weight separation. The rest of the material was freeze-dried and stored. The yield of organic material from the Bayer liquor was 11.6 g/L. The Bayer humic sample was dissolved in water/methanol (80/20, v/v) and ultrasonicated for 5 min. The solution was made up to a concentration of 6 mg/mL. 2.2. Chromatographic Analysis. The multidimensional reversed-phase HPLC system used for the separation of the Bayer humic sample was a Waters LC system that incorporated a 717plus autosampler, two 600-pumps and controllers, two 2487 dual wavelength detectors, and two six-port, two-position switching valves. Column switching was achieved using six-port, two-position switching valves that were fitted with micro-electric two-position valve actuators (Valco Instruments Co., Inc., Houston, TX). The Waters Millenium32 4.00 software controlled the switching valves. All mobile phases for this system were helium-degassed, and chromatograms were recorded at 280 nm. The first separation process used a Phenomenex BioSep-SEC-S2000 column (300 mm × 7.8 mm) running a mobile phase of sodium nitrate (NaNO3, 0.05 M) at a flow rate of 2.5 mL/min and was thermostated to 35 °C. The manufacturer reports the exclusion range of the BioSep-S2000 column to be 1000-300 000 Da. The interstitial column volume (Vi) and the total void volume (Vt) of the first dimension were determined by injecting blue dextran (2 mg/mL) and acetone, respectively. The second separation process used a Phenomenex Synergi Polar-RP column (150 mm × 4.6 mm) running a curved gradient (4) of formic acid (0.1%) and acetonitrile at a flow rate of 2.0 mL/min and was thermostated to 40 °C. The Synergi Polar-RP column is an ether-linked phenyl phase with polar endcapping. The Synergi phase was developed for separating extremely polar aromatic analytes or mixtures and is reported to improve the peak shape of both acids and bases and is stable in 100% aqueous conditions. The design of the multidimensional chromatographic system used for the separation of the Bayer humic sample incorporating two six-port, two-position switching valves is illustrated in Figure 1(panels a and b). The Bayer humic sample was injected (250 µL) into the first dimension, where the sample was separated according to the mechanism of the mobile phase and stationary phase of the chromatographic column chosen for the first dimension, principally size exclusion (Figure 1a). The peak eluting in the first dimension was cut at 200µL intervals across the entire band. In total, the sizeexclusion band was divided into 90 different sections. Each cut was subsequently transferred to an injection loop using the switching valves that were programmed
Ind. Eng. Chem. Res., Vol. 44, No. 9, 2005 3231
Figure 1. Schematic diagram of two-dimensional high-performance liquid chromatography (HPLC) column switching system: (a) system configuration for the separation of the Bayer humic sample on size-exclusion column and flushing of the sample loop onto the Synergi Polar-RP column; (b) system configuration for the “heart-cutting” of the elution band in the first dimension.
to periodically “heart-cut” the solute from the first dimension (Figure 1b).19 The injection loop was then back-flushed to load the sample onto the second dimension (Figure 1a), which was the Synergi Polar-RP column, where the sample components were further separated according to the retention mechanism of the second-dimension chromatographic column. The total analysis time (90 × 200 µL fractions) was on the order of 7 days of continuous operation. For identification purposes, the most intense bands in the second dimension of the multidimensional HPLC method were collected and further analyzed by mass spectrometry. The experiment was repeated at an analytical scale 25 times to collect enough of each sample for mass spectrometric analysis. They were then freeze-dried and made up to 50 mL in methanol/water (10/90, v/v) for mass spectrometric analysis using flow injection. 2.3. Mass Spectrometry Analysis. A VG-Quattro II triple quadrupole mass spectrometer (Micromass, Altrincham, UK), which was fitted with an electrospray source, was used to collect negative-ion ESI mass spectra, and Masslynx software was used for data acquisition and processing. Instrumental parameters, such as a capillary voltage of 2.5 kV, a source temperature of 80 °C, and a sampling cone of 30 V, were used. The collision gas was argon with a collision energy of 10 V and quadrupole resolution settings that would achieve unit mass resolution (50% full width at half maximum (fwhm)) were used. Spectra were recorded in MCA mode and were background-subtracted and smoothed. Samples were introduced to the source via flow injection of a mobile phase from a liquid chromatograph (Hewlett-Packard, model 1090 LC) at a flow rate of 10 mL/min. The mobile phase was methanol/water (50/50, v/v). Full-scan mass spectra of the isolated Bayer humic bands were recorded by flow injection of 10 mL of the sample and scanning over a range of m/z 100-700 for 30 s, using a scan time of 3 s. Product ion spectra
Figure 2. Separation of the Bayer humic sample on the BioSepS2000 size-exclusion column in the first dimension. (AU ) arbitrary units.) The shaded regions indicate 3 of the 90 fractions that were heart-cut from the first dimension at (a) 3.80 min, (b) 5.88 min, and (c) 7.08 min and loaded onto the second dimension for further separation.
3232
Ind. Eng. Chem. Res., Vol. 44, No. 9, 2005
Figure 3. Bayer humic fractions cut from the first dimension at (a) 3.80 min, (b) 5.88 min, and (c) 7.08 min and separated in the second dimension on the Synergi Polar-RP column. (AU ) arbitrary units.)
obtained under collision-induced decomposition (CID) conditions were also collected. Product ion spectra were recorded over the mass range of m/z 20-320 with a scan time of 1 s. Aliquots (10 mL) of the sample solution were flow-injected, and spectra were recorded for 1 min. The resolution of the first quadrupole was reduced slightly, to increase ion transmission while retaining unit mass resolution for both quadrupoles. 3. Results and Discussion 3.1. Development of a Multidimensional HPLC Separation. The first attribute chosen to describe the Bayer humic sample was molecular size, using a Phenomenex BioSep-S2000 size-exclusion column. Sizeexclusion separations are based on entropic exclusion principles, where the larger molecules elute first and the smaller molecules elute last. For these types of solutes, it is difficult to eliminate solute adsorption to the surface of the size-exclusion media completely, and, hence, secondary retention processes can occur. There are several reported problems associated with the separation of humic substances using size-exclusion chromatography (SEC). The most common problems occur because of possible charge interactions that could
produce attractive or repulsive effects between the matrix and the humic substances, leading to a possible overestimation or underestimation of the elution volume and sorption reactions that also result in an overestimation of the elution volume.20 To some extent, the addition of sodium nitrate (as used here) to the mobile phase limits these secondary adsorption effects; however, they can never be entirely eliminated.20-24 Nevertheless, primarily, we assume that the dominant retention process is size exclusion. The second separation process used a Phenomenex Synergi Polar-RP reversed-phase column, with the retention process being governed by the chemical interactions between the solute and the stationary phase. Consequently, two different elution mechanisms (at least in principle) function in both dimensions, and this leads to a substantial increase in the separation space and, hence, an increase in the peak capacity of the system. The chromatograph shown in Figure 2 illustrates the separation that was achieved in the first dimension on the BioSep-S2000 size-exclusion column. The retention time at which the volume of solvent was equivalent to the interstitial column volume (V0) was 2.20 min, and
Ind. Eng. Chem. Res., Vol. 44, No. 9, 2005 3233
Figure 4. Three-dimensional surface representation of the three fractions cut from the first dimension at 3.80, 5.88, and 7.08 min that were subsequently separated in the second dimension.
Figure 5. HPLC chromatogram of the fraction cut at 6.92 min in the first dimension that was subsequently separated in the second dimension. Bands at 15.25 min (peak 1), 17.30 min (peak 2), and 20.20 min (peak 3) were collected for further analysis by mass spectrometry. (AU ) arbitrary units.)
the retention time at which the volume of solvent was equivalent to the void volume (Vt) was 7.20 min. The exclusion limit is defined by V0, which is reported to be equivalent to a molecular weight range of 300 000 Da, whereas the inclusion limit is reported to be 1000 Da (according to the manufacturer’s specifications). In the chromatogram shown in Figure 2, there is a single broad band eluting virtually throughout the entire exclusion range of the column, with some secondary retention phenomena obviously apparent, as some components elute after Vt. The broad elution profile indicates that a substantial size distribution of the Bayer humic substances exists, with the majority of the sample eluting after ∼3.80 min. The shaded regions in Figure 2 indicate three fractions that were heart-cut from the first dimension at 3.80 min (region a), 5.88 min (region b), and 7.08 min
(region c) and loaded onto the second dimension for further separation. Although a total of 90 fractions were heart-cut to the second dimension, only 3 have been shown to illustrate the methodology. The chromatograms depicted in Figure 3a-c illustrate the elution of the three heart-cut Bayer humic fractions in the second dimension. Minitab software was used to develop a three-dimensional surface representation (Figure 4) of the three fractions cut from the first dimension that were separated in the second dimension and clearly illustrate the complexity of the fractions. Figure 3 shows that it is possible to resolve uniform band profiles that show the promise of being essentially pure individual components. In the three fractions that have been heart-cut from the first size-exclusion dimension, we have isolated at least 18 components resolved to baseline resolution in the second dimension, with
3234
Ind. Eng. Chem. Res., Vol. 44, No. 9, 2005
Figure 6. Negative-ion electrospray ionization (ESI) mass spectra of the three bands collected from the second dimension fraction cut at 6.92 min. Bands were collected at (a) 15.25, (b) 17.30, and (c) 20.20 min.
many other peaks exhibiting less resolution. Some of these have been characterized by mass spectrometry below. 3.2. Mass Spectrometry. The most intense bands in the second-dimension cut fractions were collected on an analytical scale for further analysis by mass spectrometry. The second-dimension chromatogram shown in Figure 5 is of the cut fraction taken at 6.92 min from the first dimension. The chromatogram displays three
major bands at 15.25 min (peak 1), 17.30 min (peak 2), and 20.20 min (peak 3). These three bands were collected for further analysis by ESI mass spectrometry. The negative-ion ESI mass spectra of the three bands in Figure 6a-c showed ions of low intensity at relatively low m/z values. The ESI spectra of band 1 displayed ions of greater intensity at m/z values of 137, 181, and 249 Da; ESI spectra of band 2 displayed ions of greater intensity at m/z values of 137 and 153 Da; and ESI
Ind. Eng. Chem. Res., Vol. 44, No. 9, 2005 3235
Figure 7. Product ion spectra of (a) m/z 137 Da from band 1 (15.25 min) and (b) m/z 167 Da from band 3 (20.20 min).
spectra of band 3 displayed ions of greater intensity at m/z values of 157, 161, and 167 Da. Product ion spectra obtained under collision-induced decomposition (CID) conditions were obtained for each of these m/z values, with several revealing structural information about the compounds that are present in each of the bands. Spectra for m/z 137 Da from band 1 and m/z 167 Da from band 3 shown in Figure 7 have fragmentation patterns that are characteristic of substituted aromatic compounds derived from lignin.25 The product ion spectrum of the peak at m/z 137 Da (Figure 7a) in band 1 is characteristic of hydroxy benzoic acid with m/z 93 Da, corresponding to a phenolate fragment and a loss of 44 Da from the molecular ion due to the loss of CO2 from a carboxylic acid group. The product ion spectrum of the peak at m/z 167 Da (Figure 7b) in band 3 is characteristic of a methylbenzene, because of the presence of the fragment ion at m/z 91 Da and the fragment ion at m/z 123 also corresponds to a loss of 44 Da and indicates the presence of a carboxylic acid group. Product ion spectra were also collected for band 2 for the predominate peaks at m/z 137 and 153 Da and contained similar characteristics. The product ion spectrum of the peak at m/z 137 Da confirmed the presence of hydroxy benzoic acid and the product ion spectrum of the peak at m/z 153 Da is characteristic of a methylbenzene, because of the presence of the fragment ion at m/z 91 Da. 3.3. Separations and Humic Substance Behavior. The contour plot in Figure 8 represents the separation in the second dimension of fractions cut between 3 and 8.00 min from the first dimension. Although 90 fractions were cut from the first dimension and separated in the second dimension, the contour plot only shows cuts taken at 3.32, 3.64, 3.80, 4.20, 4.68, 5.08, 5.48, 5.88, 6.28, 6.68, 7.08, 7.48, and 7.88 min. The contour plot illustrates the complexity of the Bayer humic substances, even after undergoing separation in the second dimension. Despite taking only 200-µL fractions for separation from the first dimension, the second-dimension fraction cuts are obviously still quite complex. To a large extent, the separation in the second dimension was based on polarity. As can be seen from Figure 8, the second-dimension separation of the 200µL fractions cut from the first dimension are still quite complex, with each fraction cut containing many compounds, and, in some cases, individual species can be
seen to be resolved. Analysis of the three bands collected in the second-dimension fraction cut at 6.92 min by mass spectrometry clearly indicated that they contained some compounds that have a simple structure. This begs the following question: why are low-molecular-weight compounds present in molecular-weight fractions that are of significantly larger molecular weight? Furthermore, are these low-molecular-weight compounds present as some type of structure that disaggregates when separated in the second dimension? The results presented here would suggest so. The results of the mass spectrometric analysis of the second-dimension bands collected from the fraction cut at 6.92 min from the first dimension (Figure 5) were significant, because they indicated the presence of lowmolecular-weight material. Although it is impossible to assign accurate molecular weights for humic substances based on the available calibration standards for SEC, which include proteins, polystyrenes, and polysaccharides, the cut at 6.92 min must be >1000 Da, because it was cut within the exclusion range of 1000-300 000 Da, as stated for the column. However, the mass spectrometric analysis showed them to have molecular weights of
