Chelation Gradients for Investigation of Metal Ion Binding at Silica

Aug 15, 2014 - Stacy L. Stegall , Kayesh M. Ashraf , Julie R. Moye , Daniel A. Higgins , Maryanne M. Collinson. Journal of Chromatography A 2016 1446,...
0 downloads 0 Views 3MB Size
Download PDF
Article pubs.acs.org/Langmuir

Chelation Gradients for Investigation of Metal Ion Binding at Silica Surfaces Balamurali Kannan,†,§ Daniel A. Higgins,*,‡ and Maryanne M. Collinson*,† †

Department of Chemistry, Virginia Commonwealth University, Richmond, Virginia 23284-2006, United States Department of Chemistry, Kansas State University, Manhattan, Kansas 66506, United States

‡

ABSTRACT: Centimeter-long surface gradients in bi- and tridentate chelating agents have been formed via controlled rate infusion, and the coordination of Cu2+ and Zn2+ to these surfaces has been examined as a function of distance by X-ray photoelectron spectroscopy (XPS). 3-(Trimethoxysilylpropyl)ethylenediamine and 3-(trimethoxysilylpropyl)diethylenetriamine were used as precursor silanes to form the chelation gradients. When the gradients were exposed to a metal ion solution, a series of coordination complexes formed along the length of the substrate. For both chelating agents at the three different concentrations studied, the amine content gradually increased from top to bottom as expected for a surface chemical gradient. While the Cu 2p peak area had nearly the same profile as nitrogen, the Zn 2p peak area did not and exhibited a plateau along much of the gradient. The normalized nitrogen-to-metal peak area ratio (N/M) was found to be highly dependent on the type of ligand, its surface concentration, and the type of metal ion. For Cu2+, the N/M ratio ranged from 8 to 11 on the diamine gradient and was ∼4 on the triamine gradient, while for Zn2+, the N/M ratio was 4−8 on diamine and 5−7 on triamine gradients. The extent of protonation of amine groups was higher for the diamine gradients, which could lead to an increased N/M ratio. Both 1:1 and 1:2 ligand/metal complexes along with dinuclear complexes are proposed to form, with their relative amounts dependent on the ligand, ligand density, and metal ion. Collectively, the methods and results described herein represent a new approach to study metal ion binding and coordination on surfaces, which is especially important to the extraction, preconcentration, and separation of metal ions.

■

INTRODUCTION

complexes on a single sample, which can then be probed by Xray photoelectron spectroscopy (XPS). Gradient materials were prepared via controlled rate infusion, whereby a silica surface is exposed to a solution containing organoalkoxysilanes, in a time-dependent fashion.9,14 In this case, the organoalkoxysilanes contain ethylenediamine or diethylenetriamine functional groups. Silica surfaces grafted with such groups have been widely used to bind, extract, preconcentrate, and separate metal ions from solution by forming stable metal−amine complexes15−22 as represented by the following equations:

A gradient surface is one that exhibits a gradual variation in a chemical or physical property along its length, width, and/or height.1 As described in recent reviews,1−8 many different approaches have been used to form gradient materials. Historically, such materials have been used to drive and control transport and as high-throughput tools to study protein adsorption and crystallization, cell adhesion, charge interactions, and phase separation.1−8 More recently, they have been used to investigate reaction kinetics,9,10 surface basicity,9 and synergistic interactions10,11 and to separate mixtures of similar compounds.12,13 In this work, we describe a new application: utilization of gradient surfaces as a high-throughput approach to study the complex coordination chemistry between immobilized chelating ligands and metal ions in solution. Subtle changes in the distribution of ligands and their spatial density along the length of a substrate leads to a series of coordination © XXXX American Chemical Society

≡SiOH + L−Si(OCH3)3 → ≡SiOSi−L + CH3OH Received: May 29, 2014 Revised: July 23, 2014

A

dx.doi.org/10.1021/la502088k | Langmuir XXXX, XXX, XXX−XXX

Langmuir

Article

ZnSO4 for ∼3 h, followed by rinsing with water and blowing dry with nitrogen. Due to limited chemical stability of amine-functionalized silica in unbuffered solutions, it is possible for leaching of the immobilized amines to take place during this process.33 However, this does not impact the results described below, as they report on the coordination chemistry of metal−ligand complexes that remain on the silica support after thorough washing. X-ray Photoelectron Spectroscopic Measurements. X-ray photoelectron spectroscopy (XPS) was performed with a ThermoFisher ESCAlab 250 imaging X-ray photoelectron spectrometer [Al Kα (1486.68 eV), 500 μm spot size, 50 eV pass energy, 0.1 eV step size]. Samples were typically placed on top of a piece of conducting tape on a 5 × 2 cm sample holder. XPS spectra were acquired in regular fashion (typically every 1−2 mm) across the wafer, starting about ∼1 mm from the edge. Spectra were calibrated by taking the C 1s peak as occurring at 284.6 eV. For XPS data analysis and peak fitting, the Thermo Scientific Avantage XPS 4.40 program was used in conjunction with a Gaussian−Lorentzian function (70:30; except for Cu metal, which was 90:10), after Shirley background subtraction. Appropriate full width at half-maximum (fwhm) peak constraints were applied.34 XPS sensitivity factors (Wagner) were 0.42, 4.2, and 4.8 for N 1s, Cu 2p3/2, and Zn 2p3/2, respectively, and were used in the calculation of normalized peak area.

≡SiOSi−L + Mn + → ≡SiOSi−L(Mn +)

The chemistry that takes place in solution between a metal ion and ligand can be very different from that which takes place when one or the other is immobilized on a surface.23,24 The stoichiometry of the coordination complex that forms between the immobilized ligand(s) and a metal ion in solution will depend on the metal ion, the ligand, the surface ligand density, its conformational flexibility and accessibility, and the intermolecular interactions that take place between all functional groups on the surface. Understanding the chemistry that takes place on these surfaces is particularly important to the design of better adsorbants for solid-phase extraction, preconcentration, and separation of metal ions17−22,25−28 as well as in the design of silica-supported catalysts using immobilized metal−amine coordination complexes.29 Functionalized silica, in particular, has had a long history of use for the extraction and separation of metal ions, including those that are toxic.26 In all these applications, elucidation of the surface composition, stoichiometry, and structure of the immobilized coordination complex is very relevant to the uptake efficiency and ultimate material performance. One of the unique attributes of gradient surfaces is that they provide a means for optimizing metal ion binding to an immobilized ligand in a single sample. For example, using polymer blends containing poly(acrylic acid), Stucky and coworkers30 established the composition and conditions for the optimal interaction of a gradient with Fe and Ce. Most previous studies, however, have involved a detailed study of one surface prepared from one set of experimental conditions.22,23,31,32 Changing the ligand density requires preparation of an entirely new sample in such studies. In the present work, we describe a high-throughput approach to study coordination complexes bound to functionalized silica, and in particular we focus on how the distribution of ligands on gradient surfaces influences metal ion binding. Significant differences between the chelating groups (bidentate vs tridentate) and metal ions (Zn2+ vs Cu2+) were noted.

■

■

RESULTS AND DISCUSSION Gradient Formation. Chelation gradients were prepared via controlled rate infusion (CRI) with either a bidentate (diamine) or tridentate (triamine) modifier known to form stable complexes with most metal ions. In CRI, a base-layercoated silicon wafer is placed vertically in a large glass vial and an organoalkoxysilane solution is slowly infused into the vial at a controlled rate.14 The height of the solution in the vial gradually increases, thus exposing the substrate to a reactive silane in a time-variant manner. Aminoalkoxysilanes work well because they self-catalyze.35 By manipulating the concentration of silane and/or the infusion rate, the amount of ligand immobilized on the surface can be controlled and the manner at which its surface coverage changes with distance can be defined.14 By default, the amount of chelating ligand immobilized near the top of the substrate will be less than that at the bottom of the substrate. Upon exposure of this gradient to a metal ion solution, a series of coordination complexes will form along its length, with their coordination number dependent on ligand density and conformational flexibility, ligand type (in this case mono-, bi-, or tridentate), and metal ion (in this case Cu2+ and Zn2+). Rinsing with water ensures that only the most stable complexes will remain on the surface for subsequent interrogation by XPS. A simple diagram of this process is shown in Scheme 1. Cu2+−Amine Gradients. In Figure 1, N 1s XPS spectra as a function of distance along the length of the substrate are shown

EXPERIMENTAL SECTION

Reagents and Sample Preparation. (3-Trimethoxysilylpropyl)diethylenetriamine (denoted herein as triamine, 97%), was purchased from Gelest. N-(3-Trimethoxysilylpropyl)ethylenediamine (denoted herein as diamine, 97%), tetraethoxysilane (TEOS, 98%), dimethyldiethoxysilane (DMDEOS, 97%), and 3-aminopropyltriethoxysilane (APTEOS, 99%) were purchased from Acros Organics. All the silanes were used as received. Silicon wafers (University Wafer, B-doped, ⟨111⟩) were cut to the appropriate size of 1 × 2 cm and then cleaned with fresh concentrated H2SO4/H2O2 (70:30) for 10 min at 70 °C in a water bath (CAUTION: Piranha solutions are extremely dangerous and react violently with organic materials). After cleaning, a sol containing TEOS/DMDEOS (1:0.15 volume ratio) was spin-coated on the wafers at 4000 rpm, as previously described.14 The base-layercoated substrates were dried for 12 h in a desiccator overnight and then soaked in ethanol for 10−12 h for stabilization prior to gradient preparation. Surface amine gradients were made by infusing a freshly prepared aminosilane solution into a vial containing a vertically aligned baselayer-coated silicon wafer at a controlled rate via a syringe pump (NewEra, NE-1000). During infusion, the base-layer coated substrate was exposed to the aminosilane solution for a different time scale along the length of the substrate, thus producing a gradual variation in amine concentration from top to bottom. The aminosilane solution for infusion was prepared by mixing ethanol/silane/water in slightly different volume ratios as defined below. To prepare the metal−amine complex films, the materials were soaked in either 0.1 M CuSO4 or

Scheme 1. Simple Depiction of the Modification of a Planar Substrate with Ethylenediamine Ligands and Subsequent Metal Ion Chelationa

a

B

Not drawn to scale. dx.doi.org/10.1021/la502088k | Langmuir XXXX, XXX, XXX−XXX

Langmuir

Article

To more easily see how the N 1s and Cu 2p3/2 peaks change along the length of the substrate, the areas under the N 1s peak and the more intense Cu 2p3/2 peak were determined. Figure 2

Figure 2. (Upper panels) N 1s and (lower panels) Cu 2p3/2 XPS peak areas as a function of distance for (left panels) diamine and (right panels) triamine gradient surfaces prepared with three different concentrations of aminosilane in solution (L, low; M, medium; and H, high).

shows the profiles for the two gradients (diamine and triamine) prepared with different concentrations of the aminosilane in solution (low, medium, and high). The profile plot clearly shows that the variation in peak area for both N and Cu is gradual from the top to the bottom of the substrate. The shape of the N 1s gradient for both diamine and triamine is similar to those in prior work.9,14 The large positive intercept and the steep rise are due to the fast reactivity of aminosilane with surface silanol groups, while a second slower process leads to a more gradual increase and flattening off of the profile.14 As expected, the gradient profile depends on the concentration of aminosilane in solution. Higher concentrations of aminosilane (Cu-Di-H and Cu-Di-M) yielded a slightly steeper rise in nitrogen at the very top of the gradient and a higher level near the bottom. Both ligand density, defined by the quantity of nitrogen on the surface, and Cu−amine chelate density, defined by the quantity of copper, show a similar trend along the length of the substrate. A steep increase in signal is followed by a more gradual rise. A higher quantity of surface-bound nitrogen also gives rise to a corresponding larger amount of copper on the surface (compare Cu-Di-H with Cu-Di-L and Cu-Tri-H with Cu-Tri-L). Under the experimental conditions used to make the gradients (fast infusion rate, relatively low concentration of aminosilane in ethanol), saturation of the surface with ligand was not achieved. For comparison, soaking a base-layer-coated wafer in a relatively high concentration of aminosilane in ethanol for 4 h yields nitrogen peak areas of ∼36 000 and 39 000 CPS eV for diamine and triamine, respectively.9 This is in contrast to ∼11 000−13 000 and ∼16 000−21 000 CPS eV observed herein (see Figure 2). As can be seen in Figure 2, the areas under both the N 1s and Cu 2p3/2 peaks are significantly higher for the triamine gradient compared to the diamine gradient, in part because there are three N atoms in triamine and two in diamine. When normalized by the number of nitrogen atoms, similar values were obtained, with the exception of the first point near the top of the gradient. In previous work, differences in normalized amine area between the diamine and triamine gradients were

Figure 1. (Left panels) N 1s and (right panels) Cu 2p spectra acquired along the length of the substrate at ∼1.5 mm intervals for (upper panels) diamine- and (lower panels) triamine-modified gradient surfaces prepared with two different concentrations of aminosilane (L, low, and H, high). The arrow indicates gradient direction from low to high amine.

for diamine- and triamine-modified surfaces prepared at an infusion rate of 0.6 mL/min (infusion time ∼4 min), followed by exposure to Cu2+ for several hours. Three different concentrations of aminosilane (high, medium, and low) were used to obtain different gradients termed Cu-Di-H, Cu-Di-M, and Cu-Di-L and Cu-Tri-H, Cu-Tri-M, and Cu-Tri-L. The aminosilane solutions were prepared by mixing ethanol/silane/ water in volume ratios of 5:0.05:0.005 (high), 5:0.025:0.0025 (medium), and 5:0.01:0.001 (low). The binding energy for the N 1s peak is 399.9 ± 0.1 eV and a small shoulder due to protonated and H-bonded amines can be seen at 401.9 eV. This shoulder is larger for diamine compared to triamine, indicative of a greater extent of protonation/H bonding in the diamine gradients. As can be seen, the intensity of the N 1s peak increases from top to bottom as expected for a gradient in amine groups across the ∼2 cm length substrate. Figure 1 also shows corresponding Cu 2p XPS spectra for gradients Cu-Di-H and Cu-Di-L and for Cu-Tri-H and Cu-Tri-L. In the Cu 2p spectra, two major peaks are observed (Cu 2p3/2 and Cu 2p1/2) with binding energies of 933.3 ± 0.2 eV and 953.3 ± 0.2 eV, respectively. The doublet separation of 20.0 eV is in agreement with the spin orbit coupling of copper.36 A small peak at a binding energy of 944.2 ± 0.2 eV was also observed, which corresponds to the Cu 2p3/2 satellite peak and is representative of copper in the 2+ oxidation state.37 This satellite peak is most pronounced in the Cu−triamine gradients (Figure 1 bottom). The peak positions are invariant with distance and extent of modification along the gradient. Analogous to N 1s, the intensity of the Cu 2p peaks also increased from top to bottom, indicative of a corresponding gradient in Cu2+. C

dx.doi.org/10.1021/la502088k | Langmuir XXXX, XXX, XXX−XXX

Langmuir

Article

hypothesis, a gradient surface was prepared with a monodentate ligand (APTEOS) and the presence/absence of copper was probed via XPS. Figure 3E,F shows the N 1s and corresponding Cu 2p XPS spectra. As can be seen, the area under the N 1s peak gradually increases from top to bottom along the substrate, as expected for a gradient in aminosilane.14 In contrast, no copper was observed, indicative of the lack of a stable coordination complex between the monodentate amine and Cu2+ and/or between Cu2+ and surface silanol groups. This experiment confirms that chelation is an important requirement for complex formation on these gradient films. Previous work, however, has reported copper ion binding to high-surface-area aminopropyl-modified silica.20,28,38 The differences between these works and ours likely reflect differences in the amount of amine bound to the silica and the material format (nearmonolayer film vs high-surface-area silica powder). Zn2+−Amine Gradients. Analogous to those described for copper, gradients in Zn−amine coordination complexes were also formed on diamine and triamine gradient surfaces prepared via CRI. In this case, two different concentrations of amine (medium and low) were used to obtain gradients with different steepness and coverage. These are termed Zn-Di-M and Zn-DiL for diamine and Zn-Tri-M and Zn-Tri-L for triamine. The volume ratio of ethanol/silane/water was 5:0.01:0.001 for ZnDi-L/Zn-Tri-L gradients and 5:0.025:0.0025 for Zn-Di-M/ZnTri-M gradients. After formation, the gradients were exposed to 0.1 M zinc sulfate for ∼3 h, followed by rinsing and drying. XPS was again used to evaluate the presence and extent of modification by the chelating ligands and their coordination with the metal ion. Figure 4 shows the N 1s and Zn 2p spectra acquired along the length of these gradient surfaces. Once again, the N 1s peak is at 400 eV with a small shoulder at 401.8 eV, corresponding to protonated/H-bonded amines.14 This shoulder is also slightly larger for diamine compared to triamine, as also observed for Cu. In the Zn 2p spectra, peaks at binding energies 1022.3 and 1045.4 eV were observed, corresponding to Zn 2p3/2 and Zn 2p1/2, respectively. The doublet separation of 23.1 eV is in agreement with the spin orbit coupling of zinc.36 Integration of the area under the Zn 2p3/2 and N 1s peaks yields the gradient profile. Figure 5 shows nitrogen and zinc profiles for the diamine and triamine gradients. As expected, the profile plot for nitrogen is concentration-dependent and shows a gradual increase in area from the top to the bottom of the substrate, similar to that shown in Figure 2 for the Cu−amine gradients. In contrast, the Zn profile does not follow the same trend as that observed for nitrogen. Instead of showing a gradual increase in area with distance, the quantity of zinc on the surface initially increases and then appears to saturate ∼5 mm from the top of the gradient. Also, both Zn-Di-L and ZnDi-M have similar amounts of zinc on the surface even though Zn-Di-M has more surface-bound nitrogen relative to Zn-Di-L. The same holds true for the triamine gradients (Zn-Tri-L and Zn-Tri-M). This difference shows that the interaction of Zn2+ with the gradients in chelating groups is different than that observed for Cu2+. Metal−Amine Coordination along Gradients. One of the unique features of gradient surfaces is the ability to probe multiple chemical (or physical) environments with one single sample rather than having to fabricate multiple samples each with slightly different ligand density, chemistry, etc.39−41 In this work, the coordination environment surrounding the metal ion was probed as the number of chelating groups changed along

noted and attributed to steric factors, but the concentration of aminosilane in the solution was significantly (∼5−25 times) higher in that case.9 To verify that the Cu 2p peak observed in XPS was the result of its coordination with surface-bound chelating groups that varied along the length of the surface, several control experiments were undertaken. In one experiment, a baselayer-coated silicon wafer was soaked in CuSO4 solution and XPS was performed on the rinsed and dried sample. No copper was observed on this sample (data not shown). In the next set of experiments, two uniformly modified surfaces were prepared by soaking the base-layer-coated silicon wafer in a diamine solution (ethanol/silane volume ratio of 5:1) for 5 min. One of these was subsequently soaked in 0.1 M Cu2+ solution for ∼3 h and then rinsed with water, while the other was not exposed to copper ions. Figure 3 A−D shows XPS spectra for N 1s and Cu

Figure 3. (Left panels) N 1s and (right panels) Cu 2p XPS spectra acquired along the length of the substrate at ∼1.5 mm intervals for (A−D) uniformly modified diamine surface, either (A, B) exposed to Cu2+ followed by rinsing or (C, D) not exposed to Cu2+, and for (E, F) gradient surface prepared from APTEOS and further exposed to Cu2+ followed by rinsing. The arrow indicates the direction of the gradient from low to high amine.

2p on the two samples. In both controls, no change in the N 1s intensity (area) with distance was observed, consistent with that expected for a uniformly modified surface. For the sample exposed to Cu2+, Cu 2p peaks were evident in the XPS spectra and showed little change with distance along the length of the substrate (Figure 3B). No signals for copper were observed on the sample not exposed to the metal ion (Fig. 3D). The N 1s peak area is slightly smaller in Figure 3A compared to Figure 3C, which could be due to sample-to-sample variability, some leaching of the surface-bound ligands during immersion in the metal ion solution,33 and/or slightly reduced detection of nitrogen resulting from surface sensitivity of XPS to only the top few nanometers. A final control experiment was performed to further establish the chelating effect of the triamine and diamine gradients. The strength of a chelate will depend on the number of coordinating atoms. Multidentate ligands such as diamine and triamine are expected to form a significantly more stable complex with the metal ion compared to a monodentate ligand. To evaluate this D

dx.doi.org/10.1021/la502088k | Langmuir XXXX, XXX, XXX−XXX

Langmuir

Article

Thermo Scientific Avantage program, which takes into account the XPS sensitivity factor (Wagner), attenuation length, and energy compensation factor. The nitrogen/metal (N/M) ratio is not the coordination number of each metal−amine complex because of the presence of uncoordinated amine (e.g., protonated amine groups) on the surface. Because the aminosilane is immobilized on silica, the measured XPS Si 2p includes contributions from both the substrate and the immobilized silane. Thus, the N/Si ratio does not provide meaningful information and we did not consider it. Cu2+−Amine Gradients. Figure 6 shows the profile plots of the normalized peak area ratio (N 1s/Cu 2p3/2) for the Cu2+−

Figure 4. (Left panels) N 1s and (right panels) Zn 2p spectra acquired along the length of the substrate at ∼1.5 mm intervals for (upper panels) diamine and (lower panels) triamine gradient surfaces prepared with two different concentrations of aminosilane (L, low, and M, medium). The arrow indicates gradient direction from low to high amine.

Figure 6. Normalized peak area ratio (N 1s to Cu 2p3/2) for (upper panel) Cu2+−diamine gradients and (lower panel) Cu2+−triamine gradients. The profiles of the three gradients prepared with different concentrations of amine in solution are indicated (L, low; M, medium; and H, high).

diamine and Cu2+−triamine gradients. When the concentration of the silane in solution was low, the corresponding gradient in Cu−amine complexes gave rise to a N/Cu ratio that averaged around 10−11, whereas for the gradient prepared from the higher concentration of ligand, the ratio was lower, ∼8. The smaller N/Cu ratio indicates that more copper is bound to CuDi-H relative to Cu-Di-L. This can also be noted in Figure 2, where the area of the Cu 2p3/2 peak is significantly higher for Cu-Di-H compared to Cu-Di-L. If all the surface-bound ligands are chelated (albeit highly unlikely24), the N/Cu ratio should be either 2 (for a 1:1 mono complex), 4 (for a 1:2 bis complex), or 6 (for a 1:3 tris complex). Formation of the latter is discounted on the basis of further analysis of the XPS data and its small equilibrium constant, as described below. The value of 8 or 10 obtained indicates that there are numerous amine functionalities on the surface not coordinated to copper ions. Protonation of the amine group and H-bonding to neighboring amines or silanol groups,42−44 as well as reduced conformational flexibility24 and electrostatic repulsion, could diminish the binding of ethylenediamine to copper. The extent of protonation/hydrogen bonding can be evaluated by deconvoluting the N 1s XPS spectra into two peaks corresponding to the protonated (402 eV) and free amine (400 eV). When the area under the free amine peak at 400 eV is used to calculate the N/Cu ratio, the value drops to ∼6 for Cu-Di-H, ∼8 for CuDi-M, and ∼6−8 for Cu-Di-L. It can also be seen in Figure 6 that the N/Cu ratio changes along the length of the gradient, though it is more pronounced in Cu-Di-L and Cu-Di-M where the N/Cu ratio initially decreases from ∼11 to ∼9 and then increases back to ∼11.

Figure 5. (Upper panels) N 1s and (lower panels) Zn 2p3/2 XPS peak area as a function of distance for (left panels) diamine and (right panels) triamine gradient surfaces prepared with two different concentrations of aminosilane in solution (L, low, and M, medium).

the length of the gradient surface. At the very top of the gradient, the surface coverage of the immobilized ligand is at its lowest and it is possible that multiple ligands will not be physically close enough to form stable bis (1:2) or tris (1:3) metal complexes. At the bottom of the gradient, the surface concentration is much higher, and thus greater densities of closely spaced ligands could lead to formation of bis (or tris) metal−amine complexes. To better facilitate the comparison, the ratio of normalized area under the N 1s peaks to the Cu 2p3/2 or Zn 2p3/2 peaks was plotted as a function of distance along the substrate. The normalized area was calculated in the E

dx.doi.org/10.1021/la502088k | Langmuir XXXX, XXX, XXX−XXX

Langmuir

Article

When the nonprotonated amine is used to calculate the N/Cu ratio, the values are smaller by ∼3 but the trend is the same. The experiment was repeated twice, and both showed a similar result (i.e., a slight dip midway down the gradient). The level of error observed in this work is similar to that described in our prior XPS work, where the sample-to-sample reproducibility as well as uniformity within one sample were evaluated.9 In both cases, the standard deviation was