Liquid chromatography with pulsed electrochemical detection at gold

Liquid Chromatography with. Pulsed Electrochemical. Detection. Dennis C. Johnson and William R. LaCourse. Department of Chemistry. Iowa State Universi...
1 downloads 0 Views 9MB Size
INSTRUMEN TATION

liquid Chromatography with Pulsed Electrochemical Detection Dennis C. Johnson and William R. LaCOUrSe Department of Chemistry Iowa State University Ames. IA 5001 1 The development of detection methods for aliphatic compounds in LC has represented a challenge of major analytical proportion. Significant advances have occurred in the use of preinjection and postcolumn chemical derivatizations to produce photometrically and electrochemically active adducw ( I , 2). Historically, however, researchers haw agreed that aliphatic compounds are not amenable to amperometric detection ( 3 ) .We believe that the simplicity of sensitive direct detection in 1.C will always he preferred, whenever available. In this article, we will descrihe the theory and use of multistep potential waveforms ( E - t ) for direct, sensitive. and reproducible detection of dcohok, glycols. carbohydrates, alkanolamines, amino acids, and sulfur compounds at Au and Pt electrodes. Numerous aromatic compounds are detected easily by anodic reactions at a cunstant (de) applied potential at solid electrodes-for example, Au. Pt. and various forms of carbon ( 4 4 ) . Significant examples include phenol, halogenated phenols, aminophenols, catecholamines, and other metabolic amines. In contrast, however. the majority of aliphatic alcohols and amines is not consistently observed to be electroactive under conditions of amperometric detection at congtant (dc) applied potentials. We have concluded that this diiference in reactivity results from the lack of conjugated bonding in the anodic re0003-2700/90/0362-589A/$02.50/0

@ 1990 American Chemical Society

action mechanisms for aliphatic compounds. Free-radical products from oxidations of aromatic molecules can be stabilized by n-resonance; hence, the activation barrier for the reaction is decreased. A similar mechanism for stabilization of aliphatic free radicals is absent and, consequently, oxidation rates are normally very low, even though anodic reactivity is predicted based on thermodynamic data. The absence of a-bonding in aliphatic compounds also results in the absence of sensitivity for photometric (UV-vis) detection of these compounds. The activation barrier for oxidation of aliphatic compounds can be decreased at noble-metal electrodes (e.g., Au and Pt) with partially unsaturated surface d-orbitals that can adsorb and thereby stabilize free-radical intermediate oxidation products. However, a serious consequence of strong adsorption can be the fouling of these electrodes by accumulated detection products ( 5 4 . Hence, the historical consensus of nonreactivity for aliphatic compounds a t Au and Pt electrodes is the result, in many instances, of a high but relatively short-lived catalytic activity for these electrodes in the “clean” state. Regeneration d electrode activity Adsorbed hydrocarbons can be oxidatively desorbed quite efficiently from these electrodes by application of a large positive potential excursion, which causes the formation of surface oxide (5-9).The intermediate products in the oxide formation (AuOH and PtOH) are reactive in the mechanism of oxygen transfer from HzO to the oxidation products. However, the final stable oxides (AuO and PtO) are quite

and

inert and must be cathodically dissolved by a negative potential excursion to restore the native reactivity of the clean metal surfaces. Some organic compounds are not electroactive a t the oxide-free noble-metal surfaces but are adsorbed a t these surfaces. These compounds also can be oxidatively desorbed simultaneously with the oxide formation process to give useful anodic signals. The application of large potential excursions a t noble-metal electrodes has long been known to result in the preparation of reproducibly clean and reactive electrode surfaces. Virtually every publication on voltammetric data for noble-metal electrodes briefly describes a protocol for electrode pretreatment that was applied to maintain electrode activity and give reproducible data. Most commonly, pretreatment includes the application of repeated cyclic potential scans or alternated positive and negative potential pulses. Such electrochemical treatment for maintaining the highest activity of Pt electrodes was reported by

ANALYTICAL CHEMISTRY, VOL. 62, NO. 10, MAY 15. I990

589A

Hammett in 1924 (20) for the study of H2oxidation and by Armstrong et al. in 1934 for studies of 02 reduction ( 2 2 ) . More recently, Clark et al. (22) reported greater reproducibility for anodic oxidation of ethylene a t Pt when cleaning pulses were applied, and MacDonald and Duke (23)offered a similar report relating to the detection of p-aminophenol. Comparable benefita of pulsed potential cleaning have been obtained also for detection of inorganic species a t noble metals, as reported by Stulik and Hora (14) for the detection of Fe(II1) and Cu(I1) a t Pt. The increased sensitivity and reproducibility from theapplication of potential pulses a t carhon electrodes also have been claimed by several workers, including Fleet and Little (19,van Rooijen and Ewing et al. (23,Berger Poppe (X), (U),and Tenygl(29).

concepts of pulsed electrocatalylic delecth

Faradaic processes that benefit from interaction of the electrode surface within the reaction mechanism are described as “electrocatalytic.” Based on this discussion, two modes of anodic electrocatalytic detection are obvious at noble-metal electrodes in conjunction with the application of potential pulses to achieve anodic cleaning and cathodic reactivation of electrodes. Mode I:Direct detection at oxidefreesurfaces. There is very little or no concurrent formation of oxide in this detection mode. The baseline signal originates primarily from double-layer charging and decays quickly to a virtual zero value. Simple alcohols, glycols, so-called sugar alcohols, monoaaccharides, and oligosaccharides are detected by Mode I a t Au electrodes in alkaline solutione and a t Pt electrodes in alkaline and acidic solutions (20-30). Mode 11: Direct oxide-catalyzed detection. Oxidation of adsorbed analyte is the primary contributor to the analytical signal in this detection mode; however, simultaneous catalytic oxidation of analyte in the diffusion layer is not excluded. The background signal is large, originating from anodic formation of surface oxide. Aliphatic amines and amino acids (31-33) as well as numerous sulfur compounds (34-37) are detectad by Mode II a t Au and Pt electrodes in alkaline solutions. A consequence of the electrocatalytic basis of detection is that the amperometric response of various members within a clans of compounds is controlled primarily by the dependence of the catalytic surface state on the electrode potential rather than by the redox potentials (E”) of the reactanta. Hence, there is little hope for voltam580A

metric resolution of complex mixtures, and electrocatalytic detection will be moat useful when coupled to LC. Au electrodes have become the moat significant of the noble metals for pulsed electrochemical detection. Figure 1 illustrates various potential regions of response by Modes I and I1 for Au in 0.1 M NaOH superimposed on the residual i-E curve obtained during a cyclic potential scan with respect to a standard calomel electrode (SCE) reference. Surface oxide is formed during the positive scan for E > +0.2 V (Wave A), and solvent breakdown with 02 evolution occurs for E > -+0.6 V (Wave B). During the subsequent negative scan, 02 evolution and oxide formation cease for E < +0.6 V and cathodic dissolution of the surface oxide produces the cathodic peak in the region +0.2 V to -0.1 V (Wave C). If present, dissolved 0 2 (indicated by a dotted line) is cathodically detected at E < --0.1 V (Wave D). All aldehydes, including the so-

called reducing sugars, are anodically detected during a positive potential change a t the oxide-free surface in the region --0.6 V to +0.2 V (Mode I). Large anodic signals are obtained for alcohols, polyalcohols, and nonreducing sugars in the region --0.3 V to +0.2 V (Mode I) with much less intense signals for some compounds from +0.2 V to +0.6 V (Mode 11). Amines and sulfur-containing compounds, for which a nonbonded electron pair resides on the N- and S-atoms, are adsorbed a t oxide-free Au surfaces for E < --0.1 V and can be anodically detected by oxide-catalyzed reactions during the positive scan for E > -+0.1 V (Mode 11). Detections a t E > -+0.6 V are not recommended because of the evolution of 02. Detection strategies

Pulsed electrochemical detection. Amperometric detection based on the electrocatalytic mechanisms under discussion can be accomplished automati-

Flgure 1. Resldual voltammetric response (/-E) for an Au rotated disk electrode in 0.1 M NaOH. Cordltions: 200 rev min-1 rotation. 100 mV s-’ potential scan. Waves (I-€,: A. oxide twmation (positive scan): B, O2evolution; C. ox& reduction (negative scan): 0.01reduction. Anodk response (positive scan): I. aldehydes and redwlng suws: 11. aiwhols and mnredwlno s w s : 111. amines and amino aci68: IV. sulhn compomds.

ANALYTICAL CHEMISTRY, VOL. 62. NO. 10, MAY 15. 1990

lime

d

-

Fkure 2. Potential-tlme E-t) waveforms. Raepses: E,, ancdic delecum;€2, oxidative cieenlng: E3. cahodic reenluation. Wawfwms: (a) pulsed SmperDmetrlc detealon (PAD) with a Dhort M B n l sampling period (e.@. 16.7 m),(b) PAD with a long cwent imegratim period ([email protected] m),and (c) inlegated PAD with a long i n r n l l o n p l o d (e.g., 200

cally by the multistep potential waveform illustrated in Figure 2a (20-30). The detection potential (E1)is chosen to he appropriate for the desired detection mechanism (see Figure l), and the faradaic signal can he sampled during a short time (e.g., 16.7 nu) after a delay of t d near the end of the detection period ( t l ) . Typical values of tl are in the 100-400-ms range. Following the detection process, the electrode surface is oxidatively cleaned by a positive step toEa(tn= 50-200ms)andthencathodi d l y reactivated by the large negative step to E3 ( t 3 = 100-400 ms) prior to the next detection cycle. Typically, the waveforms are executed a t a frequency of -1-2 Hz, which generally is appropriate for detection in LC. Amperometric detection under the control of a multistep waveform commonly has been called pulsed amperometric detection (PAD). The origin of detection peaks in LC using PAD based on Mode I is illustrated in Figure 3a hy generic chronoamperometric (i-t) response curves following the step from Ea to El a t an oxide-free electrode (30).The residual response from double-layer charging (curve A) decays very quickly, and the baseline signal in LC-PAD is very small for t d > -50 ms. Curve B in Figure 3a represents the i-t response for the presence of analyte, and the arrow represents the corresponding signal expected for LC-PAD with the indicated value of t d in the waveform. The origins of detection peaks in LCPAD based on Mode I1 are illustrated

in Figure 3b (30).Here, the residual i-t response (curve A) corresponds to the formation of surface oxide, which decays much more slowly than the current from double-layer charging. Baseline signals for Mode I1 typically have a nonzero value, as illustrated in Figure 3b for two values of t d . The i-t response (curve B in Figure 3b) corresponds to the presence of adsorbed analyte, and m o w are shown for the two values of

t d to represent the corresponding LCPAD response. For small t d , “negative” peaks can be obtained because of initial inhibition of the oxide formation process by the adsorbed analyte. For larger values of l d , “positive” peaks are obtained when sufficient oxide has been produced to catalyze the anodic reaction of the adsorbate. If an intermediate value of t d is chosen, a detection peak might not be obtained. Choosing t d > -150 ms usually is sufficient to assure “positive” LC-PAD peaks based on Mode I1 detection. Sampling of electrode response in PAD.The signal-to-noise ratio (SiN) for measurements of transient amperometric signals is influenced by the instrumental strategy used for sampling the electrode current. A major noise component of the chronoamperometric signal is sinusoidal and correlated with the60-Hzlinefrequency. Henceammmon strategy for current sampling in PAD involves some form of signal averaging over the period of one 60-Hz oscillation (Le., 16.7 ms). Accordingly, there is no contribution to signal strength from the 60-Hz noise. Actually, the time integral of a 60-Hz sinusoidal noise signal is zero for integration over any integral number ( m ) of 16.7ms periods (38).Yet the analytical signal strength can be increased significantly form >> 1. As an example, if the analytical signal is a constant value throughout the period m .16.7 ms, the S/N will be increased by the factor m. Typically, m = 12 and the integration period ( r , ) is 200 ms. This longer sam-

Figure 3. ChronoamperomaVic response (1-t) following a potential step from to €7 in the PAD waveform to illustrate the origins of chromatographic baseline and peak signals In LC-PAD. Dstectim:(8) Mode I and (b) Mode ii CYW (A) Background response in ma absence of analfie and (6) response in h e presence of analfie Note mat delay a b r step m E, 18 indicated by b, LC baseilnes am indicated by the labsled dashed lines. and LGPAD peak signals are lndkaled by mow.

ANALYTICAL CHEMISTRY. VOL. 62, NO. 10, MAY 15, 1990

581A

pling period for PAD is illustrated by the waveform in Figure 2b. Because the signal output for an integrated amperometric response has units of coulombs, the corresponding technique was originally called pulsed coulometric detection (38). Integrated PAD with a potential sweep. As discussed for Figure 3b, a large baseline signal is encountered in LC-PAD for the oxide-catalyzed detections of amino acids and sulfur compounds (Mode 11). Furthermore, the large baseline current is frequently observed to drift to more anodic values, especially for new or freshly polished electrodes. This drift occurs because of slow growth in the true electrode surface area that is attributable to surface reconstruction caused by the oxide onoff cycles in the applied multistep waveforms. Baseline offset and drift can be significantly diminished by use of the waveform in Figure 2c (39). Here, the electrode current is integrated throughout a rapid cyclic scan of the detection potential (El) within a pulsed waveform. The potential scan proceeds into (positive scan) and back out of (negative scan) the region of the oxide-catalyzed reaction for detection by Mode 11. The anodic charge for oxide formed on the positive sweep during the detection period tends to be compensated by the corresponding cathodic charge (opposite polarity) for dissolution of the oxide on the negative sweep. Hence the “background” signal on the electronic integrator a t the end of the detection period can be virtually zero and is relatively unaffected by the gradual change of electrode area. The detection procedure based on the waveform in Figure 2c was originally called potential scan pulsed coulometric detection (39); however, we now prefer the name integrated pulsed amperometric detection (IPAD). Several requirements pertaining to the cyclic sweep of E1 in IPAD must be satisfied to achieve maximum success for applications to LC. First, the cyclic scan of El must begin and end at a value for which the electrode is free of surface oxide; E < --0.1 V vs. SCE for Au in 0.1 M NaOH (see Figure 1).Second, the value of El should not extend into the region for cathodic detection of dissolved 0 2 ; E < -0.1 V vs. SCE (see Figure 1) if 0 2 is present. Finally, the positive scan must not extend beyond the value for anodic solvent breakdown; -+0.6 V vs. SCE (see Figure 1). From the residual i-E curve for Au in Figure 1, it is clear that only a small potential region centered a t --0.1 V vs. SCE is appropriate to satisfy the first two criteria in 0.1 M NaOH con592A

taining dissolved 0 2 . The purging of solvents with He and use of 02-impermeable tubing can result in virtually 02-free conditions that greatly relax the constraints of these two criteria. The advantage of IPAD compared with PAD relates to minimization of baseline drift for oxide-catalyzed detections (Mode 11). Comparisons of IPAD with PAD for carbohydrates at the oxide-free surfaces (Mode I) have indicated virtually no significant difference in detectability for the two techniques, and we recommend continued use of PAD (ti = 200 ms) for carbohydrate and alcohol detection. An additional consideration in IPAD relates to the electrochemical reversibility of the detection reaction. Clearly, the anodic signal is expected to be a t a maximum when there is no cathodic contribution to the net current integral from reduction of the oxidation product during the negative portion of the cyclic scan of El. All detection processes pertinent to this article are irreversible; the oxidation products cannot be detected cathodically. However, even for a reversible redox system, there is sufficient loss of soluble oxidation products from the diffusion layer a t the electrode by convective-diffusional mass transport so that the cathodic charge from reduction of detection products will not be equivalent to the anodic charge from the detection process. Chromatographic applications Polyalcohols a n d carbohydrates (Mode I). All aldehydes, simple alcohols, glycols, polyalcohols, and carbohydrates can be detected by pulsed

ANALYTICAL CHEMISTRY, VOL. 62, NO. 10, MAY 15, 1990

H-C

electrochemical detection at Au and Pt electrodes in alkaline media (pH > -12). However, use of Au electrodes has the distinct advantage that detection can be achieved without simultaneous reduction of dissolved 0 2 . The maximum anodic signal for polyalcohols and carbohydrates in 0.1 M NaOH is obtained at E = +0.1 - +0.2 V vs. SCE, and a value in this range is chosen for E1 in the PAD waveform (Figures 2a and 2b). The primary detection reactions are given by Equations 1-3 (see box) for glucose (40). The first step is a very fast two-electron oxidation of the aldehyde group to the corresponding carboxylic acid (Equation 1).This is followed by a series of relatively fast steps resulting in cleavage of the C1-C2 bond, with production of HCOzH, followed by conversion of the Cz and C6 to the corresponding carboxylates (Equation 2). Additional sequential anodic cleavage of the remaining terminal carbons can occur according to Equation 3, but it occurs very slowly (40, 41). The observed number of electrons (nabs) for glucose detection at an Au electrode in a typical thin-layer LC detector is -10 eq mol-l. At rates of convection higher than those that exist in the typical thin-layer LC detector, nabs = -8 eq mol-’. The limit of detection for glucose by PAD (ti = 200 ms) is below 1ng in 0.1 M NaOH with a linear response over more than 4 decades in concentration. The rates of anodic mechanisms a t Au electrodes decline with increasing solution acidity, and virtually no response for alcohols is obtained for pH