Basal Plane Pyrolytic Graphite Modified Electrodes: Comparison of

The Journal of Physical Chemistry C 2014 118 (39), 22635-22642 ..... Kshama Parate , Taylor M. Davis , Loreen R. Stromberg , Edward K.L. Chan , Joseph...
0 downloads 0 Views 292KB Size
Anal. Chem. 2004, 76, 2677-2682

Articles

Basal Plane Pyrolytic Graphite Modified Electrodes: Comparison of Carbon Nanotubes and Graphite Powder as Electrocatalysts Ryan R. Moore, Craig E. Banks, and Richard G. Compton*

Physical and Theoretical Chemistry Laboratory, Oxford University, South Parks Road, Oxford OX1 3QZ, U.K.

The oxidations of NADH, epinephrine, and norepinephrine are studied using carbon nanotube and graphite powder-modified basal plane pyrolytic graphite electrodes. Immobilization is achieved in two ways: first, via abrasive attachment of multiwall carbon nanotubes or graphite powder by gently rubbing the electrode surface on a fine quality paper supporting the desired material; second, via “film” modification from dispersing either graphite powder or nanotubes in acetonitrile and pipeting a small volume onto the electrode surface and allowing the solvent to volatilize. While electrocatalytic behavior of both types of nanotube-modified electrodes is shown, with enhanced currents and reduced peak-to-peak separations in the voltammetry in comparison with naked basal plane pyrolytic graphite, similar catalytic behavior is also seen at the graphite powder-modified electrodes. Caution is, therefore, suggested in assigning unique catalytic properties to carbon nanotubes. Carbon nanotubes have emerged as a new class of nanomaterials which are receiving considerable interest.1-4 There are two distinct classes of nanotubes: single-wall nanotubes (SWNTs) and multiwall nanotubes (MWNTs). SWNTs are made from a single graphite sheet rolled flawlessly, producing a tube diameter of 1-2 nm, whereas MWNTs are composed of concentric and closed graphite tubules having diameters varying from 2 to 50 nm with an interlayer distance of ∼0.34 nm.2-7 The open ends of CNTs * To whom correspondence should be addressed. Phone: 01865 275413. Fax: 01865 275410. E-mail: [email protected]. (1) Wang, J.; Deo, R. P.; Musameh, M. Electroanalysis 2003, 15, 1830-1834. (2) Iijima, S. Nature 1991, 354, 56-58. (3) Sherigara, B. S.; Kutner, W.; Souza, F. D. Electroanalysis 2003, 15, 753772. (4) Ajayan, P. M. Chem. Rev. 1999, 99, 1787-1799. (5) Luo, H.; Shi, Z.; Li, N.; Gu, Z.; Zhuang, Q. Anal. Chem 2001, 73, 915-920. (6) Wang, J.; Li, M.; Shi, Z.; Li, N.; Gu, Z. Microchem. J. 2002, 73, 325-323. 10.1021/ac040017q CCC: $27.50 Published on Web 04/20/2004

© 2004 American Chemical Society

have been likened to edge planes of highly orientated pyrolytic graphite (hopg), whereas the walls are suggested to have properties similar to those of basal planes of hopg electrodes.8,9 Nanotubes have attracted considerable attention in electrochemistry due to the reported electrocatalytic properties of carbon nanotube-modified electrodes.1,5,10-16 It is also suggested that nanotubes represent a much more well-defined chemical system with easier access to meaningful molecular-scale interrogation and benefits in applications requiring assembly. Wang and co-workers6 investigated the electrocatalytic behavior of single-wall carbon nanotube films on a gold electrode. They observed catalytic activity of the nanotubes toward oxidation of uric acid, dopamine, and norepinephrine.13 They later extended this using a glassy carbon electrode modified with single-wall carbon nanotubes, again observing high electrocatalytic activity for the oxidations of norepinephrine, dopamine, epinephrine, and ascorbic acid.13 Rubianes and Rivas17 were the first to evaluate the performance of carbon nanotube paste electrodes, which showed excellent electrocatalytic activity, as compared to that of a conventional carbon paste electrode (CPE), toward ascorbic acid, uric acid, (7) Dresselhaus, M. S.; Dresselhaus, G.; Saito, R. Carbon 1995, 33, 883-991. (8) Gooding, J. J.; Wibowo, R.; Liu, J.; Yang, W.; Losic, D.; Orbons, S.; Mearns, F. J.; Shapter, J. G.; Hibbert, D. B. J. Am. Chem. Soc. 2003, 125, 90069007. (9) Nugent, J. M.; Santhanam, K. S. V.; Rubio, A.; Ajayan, P. M. Nano Lett. 2001, 1, 87-91. (10) Wang, J.; Li, M.; Shi, Z.; Li, N.; Gu, Z. Electrochim. Acta 2001, 47, 651. (11) Wu, F. H.; Zhao, G.-C.; Wei, X.-W. Electrochem. Commun. 2002, 4, 690694. (12) Zhao, Y.-D.; Zhang, W.-D.; Chen, H.; Luo, Q.-M. Sens. Actuators, B 2003, 92, 279-285. (13) Wang, J.; Li, M.; Shi, Z.; Li, N.; Gu, Z. Electroanalysis 2002, 14, 225-230. (14) Wang, Z.; Wang, Y.; Luo, G. Electroanalysis 2003, 15, 1129-1133. (15) Musameh, M.; Wang, J.; Merkoci, A.; Lin, Y. Electrochem. Commun. 2002, 4, 743-746. (16) Wang, J.; Musameh, M.; Lin, Y. J. Am. Chem. Soc. 2003, 125, 2408-2409. (17) Rubianes, M. D.; Rivas, G. A. Electrochem. Commun. 2003, 5, 689-694.

Analytical Chemistry, Vol. 76, No. 10, May 15, 2004 2677

dopamine, 3,4-dihydroxyphenylacetic acid, and hydrogen peroxide.17 A dramatic decrease in the overvoltage of ascorbic and uric acids and hydrogen peroxide was observed, with improvement of the reversibility of the 3,4-dihydroxyphenylacetic acid and hydrogen peroxide voltammetry.17 This work was developed by Valentini et al.,18 who explored the electrochemical behavior of a SWNT paste electrode toward the reduction of ferricyanide and compared it with a carbon paste electrode (CPE). They observed broader peaks with higher peak-to-peak separations (209 mV) at the CPE compared to that seen at the SWNT paste electrode (90 mV).18 This comparison was then extended to a whole range of redox systems, such as hydrogen peroxide, NADH, and dopamine.18 The electrochemical oxidation of thymine at an R-cyclodextrin incorporated carbon nanotube film on a highly orientated pyrolytic graphite (hopg) electrode has been explored.14 The oxidation of thymine when compared with a bare hopg electrode and CNT film electrode was found to be electrocatalytic, leading to the development of a sensitive detector for thymine.14 Gooding et al.8 have reported an investigation of electrontransfer properties of microperoxidase MP-11 attached to the end of aligned SWCNTs, which are attached to a modified gold electrode.8 Essentially, the SWNTs are shortened and aligned normal to the electrode surface by self-assembly. Furthermore, it was shown that the shortened nanotubes effectively act as molecular wires: they did not add significant electrical resistance.8 Wang et al. have evaluated a carbon nanotube film-modified glassy carbon electrode for the detection of NADH,15 observing a substantial decrease in the overvoltage of the oxidation of NADH, as compared to ordinary carbon electrodes.15 Subsequently, they reported the electrochemical detection of phenolic compounds at a glassy carbon-modified nanotube film electrode,1 solubilization of carbon nanotubes by Nafion for the preparation of amperometric biosensors,16 nanotube screen-printed electrochemical sensors,19 enzyme-dispersed carbon-nanotube electrodes for glucose monitoring,20 and a reagentless amperometric alcohol biosensor using a carbon-nanotube-Teflon composite electrode.21 In this paper, we first study the model system of the aqueous ferri/ferrocyanide redox couple using carbon nanotube- and graphite powder-modified basal plane pyroltic graphite (bppg) electrodes. These electrodes are formed via one of two methods. The first is by abrasively attaching carbon nanotubes or graphite powder to the surface of the bppg electrodes by gently rubbing a polished bppg electrode on a fine quality filter paper containing either graphite powder or nanotubes. The second is by dispersing either graphite powder or nanotubes in acetonitrile and pipeting a small volume onto the polished bppg electrode and allowing the solvent to volatilize leaving a carbon “film” on the electrode surface. Next, we consider the response of these electrodes to norepinephrine, NADH, and epinephrine, which have all been deemed “electrocatalytic” with respect to carbon nanotubes. Surprisingly, similar catalytic behavior is found using the graphite powder instead of the nanotubes as the electrocatalyst! The need for caution in attributing special catalytic or electrocatalytic properties to nanotubes is emphasized. (18) Valentini, F.; Amine, A.; Orlanducci, S.; Terranova, M. L.; Palleschi, G. Anal. Chem. 2003, 75, 5413-5421. (19) Wang, J.; Musameh, M. Analyst 2004, 129, 1-2. (20) Wang, J.; Musameh, M. Analyst 2003, 128, 1382-1385. (21) Wang, J.; Musameh, M. Anal. Lett. 2003, 36, 2041-2048.

2678

Analytical Chemistry, Vol. 76, No. 10, May 15, 2004

EXPERIMENTAL SECTION Reagents and Equipment. All chemicals used were of analytical grade and were used as received without any further purification. These were β-nicotinamide adenine dinucleotide, reduced disodium salt hydrate (NADH, Aldrich, >98%), epinephrine (4-[1-hydroxy-2-(methylamino)ethyl]-1,2-benzenediol, Aldrich, >98%), norepinephrine ((R)-4-(2-amino-1-hydroxyethyl)-1,2-benzenediol, Aldrich, >98%), and potassium ferricyanide (99+%, Aldrich). All solutions were prepared with deionized water of resistivity not less than 18.2 MΩ cm (Vivendi water systems, U.K.). Graphite powder was obtained commercially from Aldrich (Product No. 28,286-3, 2-4-µm diameter, see below); the multiwall carbon nanotubes (>95% purity, 10-20-nM diameter, 5-20-µm length) prepared by chemical vapor deposition were obtained from NanoLab (Brighton, MA). Voltammetric measurements were carried out using a µ-Autolab II (ECO-Chemie, The Netherlands) potentiostat. All measurements were conducted using a three-electrode configuration. A basal plane pyrolytic graphite electrode (5-mm diameter, Le Carbone, Ltd. Sussex, U.K.) was used as the working electrode. The counter electrode was a bright platinum wire, with a saturated calomel electrode (Radiometer, Copenhagen, Denmark) completing the circuit. Preparation and Modification of the Electrodes. A basal plane pyrolytic graphite electrode was prepared for modification by renewing the electrode surface with cellophane tape.22 This procedure involves polishing an old bppg electrode surface on carborundum paper, pressing cellophane tape on the cleaned bppg surface and then removing the cellophane tape, along with several surface layers of graphite.22 This is repeated multiple times to achieve the final surface. Before use, the electrode is then cleaned in acetone to remove any adhesive. Carbon nanotubes or graphite powder were abrasively immobilized onto the bppg electrode by gently rubbing the electrode surface on a fine quality filter paper containing either graphite powder or carbon nanotubes. Alternatively, a film of graphite powder or carbon nanotubes was attached via dispersion of 0.18 g of graphite powder in 15 mL of acetonitrile or 0.018 g of carbon nanotubes dispersed in 4 mL of acetonitrile. These suspensions were agitated with 10 µL of a suspension of the required modifier pipeted onto the bppg electrode surface. After the solvent is allowed to volatilize a carbon “film” is left on the electrode surface. SEM images of abrasively attached MWCNT (Figure 1a,b) and graphite powder on a bppg electrode are shown (Figure 1c). Many MWCNT bundles can be seen (Figure 1a,b) with a diameter in the range 10-20 nm, while in Figure 1c, graphite powder, with a diameter of 2-4 µm, immobilized on the bppg electrode is shown. RESULTS AND DISCUSSION We first explore the response of bppg electrodes modified with carbon nanotubes or graphite powder toward the one-electron aqueous redox probe potassium ferricyanide. A carbon nanotube abrasively modified bppg electrode was prepared and immersed into a 1 mM aqueous solution of potassium ferricyanide (in 0.1 M KCl) in which the voltammetric response was investigated over (22) McCreery, R. L. In Electroanalytical Chemistry; Bard, A. J., Ed.; Marcel Dekker: New York, 1990; Vol. 17, p 304. (23) Wang, J.; Musameh, M. Anal. Chem. 2003, 75, 2075-2079

Figure 2. (a) Voltammetric responses of a carbon nanotube abrasively modified bppg electrode to 1 mM ferricyanide in 0.1 M KCl at different scan rates (inner to outer): 0.01, 0.025, 0.05, 0.075, 0.1, 0.125, 0.150, 0.175, and 0.2 V s-1. (b) Voltammetric responses of a graphite powder abrasively modified bppg electrode to 1 mM ferricyanide in 0.1 M KCl at different scan rates (inner to outer): 0.01, 0.025, 0.05, 0.075, 0.1, 0.125, 0.150, 0.175, and 0.2 V s-1.

Figure 1. (a,b) SEM images of carbon nanotubes abrasively attached to the surface of a basal plane pyrolytic graphite electrode. (c) A SEM image of graphite powder abrasively attached to the surface of a basal plane pyrolytic graphite electrode.

a scan range of 0.01-0.2 V s-1. The peak-to-peak separation is 146 mV at 100 mV s-1, with a formal potential of 0.17 ((0.01) V. As can be seen in Figure 2a, the peak-to-peak separation increases with increasing scan rates, suggesting quasireversible behavior. Next, a graphite powder abrasively modified bppg electrode was prepared. Again, the cyclic voltammograms were recorded over a scan range of 0.01-0.2 V s-1 in 1 mM ferricyanide (in 0.1 M KCl). As depicted in Figure 2b, quasireversibility is observed, with

a formal potential of 0.16 ((0.01) V and a peak-to-peak separation of 167 mV at 100 mV s-1, with an increasing separation with increasing scan rates. For comparison, a bare bppg electrode was investigated in the ferricyanide solution. A large peak-to-peak separation of 630 mV (at 100 mV s-1) was observed, which is consistent with literature reports.22 Comparison of the slow electron transfer at the bare bppg electrode with that of the graphite- or nanotube-modified electrode suggests that the carbon nanotubes and the graphite powder might have similar responses as electrocatalytic materials. It is this issue we next address. Oxidation of Norepinephrine. Wang and co-workers13 have investigated the behavior of the important neurotransmitter norepinephrine (NE) at a glassy carbon electrode modified with single wall carbon nanotubes.13 They observed that the oxidation peak of NE shifted to less positive potentials, as compared with that seen at a bare glassy carbon electrode, attributing it to the electrocatalytic activity of the carbon nanotubes toward NE oxidation.13 We next investigate nanotube- and Analytical Chemistry, Vol. 76, No. 10, May 15, 2004

2679

Figure 3. (a) Cyclic voltammograms of 1 mM norepinephrine in pH 5.75 acetate buffer for nanotube (curve B) and graphite powder abrasively attached bppg modified (curve A) electrodes. Curve C shows a bare bppg electrode. All voltammograms were recorded at a scan rate of 0.1 V s-1. (b) Cyclic voltammograms of 1 mM norepinephrine in pH 5.75 acetate buffer for nanotube- (curve A) and graphite powder film-modified (curve B) electrodes. Curve C shows a bare bppg electrode. All voltammograms were recorded at a scan rate of 0.1 V s-1.

graphite powder-modified electrodes for use in electroanalysis. First, a bare bppg electrode was prepared as described in the Experimental Section. This was then placed into a 1 mM aqueous solution of norepinephrine in pH 5.75 acetate buffer; the pH of the buffer was optimized by Wang et al., who found that the oxidation peak current was at its maximum and the peak-to-peak potential separation at its lowest at pH 5.75.13 The cyclic voltammogram of the bare bppg graphite electrode is shown in Figure 3a (scan C), where the oxidation and reduction peaks can be seen at +0.50 ((0.01) V and +0.04 ((0.01) V, respectively (vs SCE), with a large peak-to-peak separation (∆Epp > 450 mV). The formal potential for the oxidation of NE at this pH can be inferred as being ∼0.27 ((0.03) V, corresponding to the midpoint between the two peaks. Second, a bppg electrode was abrasively modified with carbon nanotubes, and the voltammetric response was explored in a 1 mM NE solution in pH 5.75 acetate buffer. As is 2680 Analytical Chemistry, Vol. 76, No. 10, May 15, 2004

shown in Figure 3a (curve B), oxidation and reduction waves are observed at +0.33 V and +0.16 V, respectively. Comparison of the bare and the modified electrodes reveals a reduction in the peak-to-peak separation in the latter case. The large decrease in the peak-to-peak separation, as compared to the bare electrode, reflects a measure in the rate of the electrochemical process, which also manifests itself in an increase in the peak heights as the process becomes electrochemically less irreversible. Next, an abrasively modified graphite powder electrode was prepared and explored in the NE solution. As shown in Figure 3a (curve A), the oxidation and reduction waves are observed at +0.32 V and +0.17 V, respectively. Thus, closely similar voltammetric signals are seen at both the nanotube and graphite powder versions of the abrasively modified electrodes. Next, a carbon nanotube film bppg modified electrode was prepared and explored in the NE solution. A well-defined oxidation wave of NE is observed at +0.30 ((0.01) V, and a reduction peak, at +0.18 ((0.01) V (vs SCE, Figure 3b, scan A). For comparison, a graphite powder film bppg modified electrode was prepared and voltammetrically examined in the NE solution. The peak potential for oxidation and reduction occur at +0.29 ((0.01) V and +0.17 ((0.01) V, respectively (Figure 3b, scan B), suggesting that the electrochemical reaction occurs with a similar rate constant on the graphite powder and nanotube film electrodes. Note that the difference in voltammetric shape between the film and the abrasively modified electrodes reflects the different amounts of material immobilized and the potentially partially “porous” nature of the film electrode, leading to “thin layer” behavior and the loss of the “diffusional tail” seen with the abrasively modified electrodes. Comparison of the nanotube-modified electrode with the graphite-modified electrode in both the abrasively and filmmodified cases suggests that there is no significant difference between the two for use as electrocatalysts in the electroanalytical determination of norepinephrine. Oxidation of NADH. Musameh et al. have reported on the low-potential stable electrocatalytic response of NADH using glassy carbon-modified electrodes.15 We next turn to investigating this system using bare, abrasively, and film-modified bppg electrodes. Figure 4A compares cyclic voltammograms for 1 mM NADH in pH 7.4 phosphate buffer recorded at 100 mV s-1, at an unmodified bppg electrode (curve C), at a MWCNT film bppg eleectrode (A), and at a graphite powder film-modified bppg electrode (B). For the unmodified bppg electrode, a broad oxidation peak is observed at +0.82 ((0.01) V (vs SCE). This value is almost identical to the peak potential of NADH reported by Musameh and co-workers under similar conditions.15 For the nanotube-modified bppg electrode, a substantial shift to less positive potentials (+0.31 V) is observed with amplification of the current signal. However, comparison with the graphite powdermodified electrode (Figure 3, curve B) reveals that the electrode modified with graphite powder exhibits a potential shift to 0.34 V vs SCE, also with an increased current signal. We note that Musameh et al.15 observed potential shifts of 490 mV for a MWNT modified electrode and of 460 mV for a SWNT modified glassy carbon electrode, deeming the substantial potential shift to arise from the electrocatalytic activity of carbon nanotubes. In our case,

Table 1. Literature Comparison of the “Catalytic” Uses of Nanotubes in Electrochemistry target

nanotubesa,b

phenolic compounds dopamine, epinephrine, ascorbic acid DOPACd nitric oxide cysteine norepinephrine dopamine, epinephrine, ascorbic acid thymine NADH H2O2

MW SW (functionalized) SW MW MW SW MWf MW/SW MW

substratec

electrocatalytic w.r.t.c

ref

GC GC GC GC Pte GC

bare GC bare GC bare GC bare GC bare GC, graphite modified Pt bare GC

1 5 10 11 12 13

HOPG GC GC

bare HOPGg bare GC bare GC

14 15 23

a MW, multiwall carbon nanotubes. b SW, single-wall carbon nanotubes. c GC; glassy carbon working electrode. d DOPAC, 3,4-dihydroxyphenylacetic acid. e Microelectrode. f Modified with R-cyclodextrin. g HOPG, highly orientated pyrolytic graphite electrode.

Figure 4. (a) Cyclic voltammograms of 1 mM NADH in pH 7.4 phosphate buffer for nanotube (curve A) and graphite powder film bppg modified (curve B) electrodes. Curve C shows the response of a bare bppg electrode. All voltammograms were recorded at a scan rate of 0.1 V s-1. (b) Cyclic voltammograms of 1 mM NADH in pH 7.4 phosphate buffer for nanotube (curve A) and graphite powder (curve B) abrasively attached bppg modified electrodes. Curve C shows the response of a bare bppg electrode. All voltammograms were recorded at a scan rate of 0.1 V s-1.

Figure 5. (a) Cyclic voltammograms of 5 mM epinephrine in pH 5 acetate buffer for nanotube (curve A) and graphite (curve B) powder film bppg modified electrodes. Curve C shows the response of a bare bppg electrode. All voltammograms were recorded at a scan rate of 0.1 V s-1. (b) Cyclic voltammograms of 5 mM epinephrine in pH 5 acetate buffer for nanotube (curve B) and graphite powder (curve A) abrasively attached bppg modified electrodes. Curve C shows the response of a bare bppg electrode. All voltammograms were recorded at a scan rate of 0.1 V s-1.

we observe a larger potential shift for the nanotube bppg modified electrode and one of a similar magnitude for the graphite powdermodified bppg electrode. We next compare this behavior with that of abrasively attached graphite powder and MWCNTs.

First, a carbon nanotube-modified electrode was prepared and studied by cyclic voltammetry in the NADH solution. Figure 4B (curve A) shows a potential shift relative to the bare electrode of 300 mV to a less positive potential (+0.52 ( 0.01 V vs SCE), with Analytical Chemistry, Vol. 76, No. 10, May 15, 2004

2681

a small increase in the current response. Next, an abrasively modified graphite powder electrode was prepared and examined in the NADH solution. A shift of the peak to a less positive potential of ca. +0.56 V vs SCE is observed (curve B). Comparison of the data collected for both carbon nanotubemodified electrodes and the graphite powder-modified electrodes shows that in both cases (film and abrasively modified), there is no significant advantage of using a nanotube-modified electrode for the clinical sensing of NADH. Oxidation of Epinephrine. A bare bppg electrode was first prepared as described in the Experimental Section. This was then immersed into a 5 mM solution of epinephrine in pH 5 acetate buffer, and the voltammetric response was recorded. As can be seen from Figure 5a (curve C), an irreversible oxidation is observed at +0.65 ((0.01) V (vs SCE). Next, the carbon nanotube film bppg electrode response was investigated. A shift of the oxidation wave was observed to +0.44 ((0.01) V (vs SCE), with an increase in the peak current, as compared to the response of the bare electrode (see Figure 5a, curve A). This ∼200 mV shift suggests significant electroactivity of the carbon nanotube film bppg modified electrode toward epinephrine oxidation. This is next contrasted to a graphite powder film bppg modified electrode. Figure 5a (curve B) shows a similar response is obtained, as compared to that of the nanotube film electrode, where the oxidation wave is seen at +0.48 ((0.01) V (vs SCE) with a slightly lower peak current than that seen for the nanotube case. Last, abrasively attached nanotubes and graphite powder were explored for the oxidation of epinephrine. As can be seen in Figure 5b, the nanotube-modified bppg electrode (curve B) exhibits only a small increase in current, with the oxidation peak occurring at

2682

Analytical Chemistry, Vol. 76, No. 10, May 15, 2004

a slightly less positive potential of +0.50 V (vs SCE). Comparison with the graphite bppg modified electrode (curve A) again shows an oxidation peak at +0.53 (( 0.01) V, which is very similar to the nanotube-modified electrode. CONCLUSIONS We have studied the oxidation of NADH, epinephrine, and norepinephrine using carbon nanotube- and graphite powdermodified basal plane pyrolytic graphite via abrasive attachment or “film” modification. Comparison of the response seen for the graphite powder and nanotube modified electrodes suggests that the latter are effective electrocatalysts, but with the former producing similar effects. We emphasize the need for caution in attributing electrocatalytic properties to nanotubes without conducting the appropriate control experiments. Table 1 summarizes literature reports of “electrocatalytic” effects seen in nanotubes and the electrode surface(s) in comparison with which they are deemed catalytic. The vast majority of studies employ glassy carbon; we suggest that in the future, nanotube catalytic activity be additionally compared against that of graphite powder. The latter is conveniently studied via immobilization on a bppg electrode. ACKNOWLEDGMENT C.E.B. thanks the EPSRC for support via a project studentship (Grant GR/R14392/01). Received for review January 23, 2004. Accepted March 11, 2004. AC040017Q