Covalent Assembly and Micropatterning of Functionalized Multiwalled

Jun 5, 2008 - Université de Rennes 1 and CNRS UMR 6226 Sciences Chimiques de ... di Trieste, Piazzale Europa 1, 34127 Trieste, Italy, and CNR-INFM La...
23 downloads 0 Views 2MB Size
Download PDF
Langmuir 2008, 24, 6595-6602

6595

Covalent Assembly and Micropatterning of Functionalized Multiwalled Carbon Nanotubes to Monolayer-Modified Si(111) Surfaces Bruno Fabre,*,† Fanny Hauquier,† Cyril Herrier,† Giorgia Pastorin,‡ Wei Wu,‡ Alberto Bianco,*,‡ Maurizio Prato,*,§ Philippe Hapiot,† Dodzi Zigah,† Mauro Prasciolu,| and Lisa Vaccari⊥ UniVersite´ de Rennes 1 and CNRS UMR 6226 Sciences Chimiques de Rennes, Matie`re Condense´e et Syste`mes Electroactifs MaCSE, Campus de Beaulieu, 35042 Rennes Cedex, France, CNRS, Institut de Biologie Mole´culaire et Cellulaire, Laboratoire d’Immunologie et Chimie The´rapeutiques, 67000 Strasbourg, France, Dipartimento di Scienze Farmaceutiche, INSTM, unit of Trieste, UniVersita` di Trieste, Piazzale Europa 1, 34127 Trieste, Italy, and CNR-INFM Laboratorio Nazionale TASC and Sincrotrone Elettra, S.S.14 Km 163.5, Area Science Park 34012 BasoVizza - Trieste, Italy ReceiVed February 1, 2008. ReVised Manuscript ReceiVed April 4, 2008 Multiwalled carbon nanotubes (MWNTs) covalently bound to monocrystalline p-type Si(111) surfaces have been prepared by attaching soluble amine-functionalized MWNTs onto a preassembled undecanoic acid monolayer using carbodiimide coupling. SEM analysis of these functionalized surfaces shows that the bound MWNTs are parallel to the surface rather than perpendicular. The voltammetric and electrochemical impedance spectroscopy measurements reveal that the electron transfer at the MWNT-modified surface is faster than that observed at a MWNT-free alkyl monolayer. We have also demonstrated that it is possible to prepare MWNT micropatterns using this surface amidation reaction and a “reagentless” UV photolithography technique. Following this approach, MWNT patterns surrounded by n-dodecyl areas have been produced and the local electrochemical properties of these micropatterned surfaces have been examined by scanning electrochemical microscopy. In particular, it is demonstrated that the MWNT patterns allow a faster charge transfer which is consistent with the results obtained for the uniformly modified surfaces.

1. Introduction Carbon nanotubes (CNTs) are the focus of intense attention owing to their high active surface area and their unique electrical, mechanical, and thermal properties.1–3 The integration of these one-dimensional nano-objects with metallic or semiconducting substrates is thus a very promising approach for developing new nanoscale devices for future applications in electronics (e.g., memories,4,5 field-effect transistors,6,7 optoelectronics,8,9 highresolution scanning probe microscopies,10,11 and chemical/ biological sensors).12–18 However, the fabrication of CNT-based devices is severely hindered by the lack of simple and reliable * Corresponding author. E-mail: [email protected] (B.F.); a.bianco@ ibmc.u-strasbg.fr (A.B.); [email protected] (M.P.). † Universite´ de Rennes 1 and CNRS UMR 6226 Sciences Chimiques de Rennes. ‡ CNRS, Institut de Biologie Mole´culaire et Cellulaire. § Universita` di Trieste. | CNR-INFM Laboratorio Nazionale TASC. ⊥ Sincrotrone Elettra. (1) Dresselhaus, M. S.; Dresselhaus, G.; Eklund, P. C. Science of fullerenes and carbon nanotubes; Academic Press: New York, 1996. (2) Saito, R.; Dresselhaus, M. S.; Dresselhaus, G. Physical properties of carbon nanotubes; Imperial College Press: London, 1998. (3) Ajayan, P. M. Chem. ReV. 1999, 99, 1787–1799. (4) Rueckes, T.; Kim, K.; Joselevich, E.; Tseng, G. Y.; Cheung, C. L.; Lieber, C. M. Science 2000, 289, 94–97. (5) Mason, N.; Biercuk, M. J.; Marcus, C. M. Science 2004, 303, 655–658. (6) Tans, S. J.; Verschueren, A. R. M.; Dekker, C. Nature 1998, 393, 49–52. (7) Martel, R.; Schmidt, T.; Shea, H. R.; Hertel, T.; Avouris, P. Appl. Phys. Lett. 1998, 73, 2447–2449. (8) Sheeney-Haj-Ichia, L.; Basnar, B.; Willner, I. Angew. Chem., Int. Ed. 2005, 44, 78–83. (9) Guldi, D. M.; Rahman, G. M. A.; Prato, M.; Jux, N.; Qin, S.; Ford, W. Angew. Chem. Int. Ed 2005, 44, 2015–2018. (10) Wong, S. S.; Joselevich, E.; Wooley, A. T.; Cheung, C. L.; Lieber, C. M. Nature 1998, 394, 52–55. (11) Yang, Y.; Zhang, J.; Nan, X.; Liu, Z. J. Phys. Chem. B 2002, 106, 4139– 4144. (12) Huang, M.; Jiang, H.; Qu, X.; Xu, Z.; Wang, Y.; Dong, S. Chem. Commun. 2005, 5560–5562.

methods to deposit CNTs in a controlled fashion. In view of facilitating their solubility, their manipulation, and therefore their use, CNTs have been covalently derivatized at their sidewalls or ends with many organic groups. The end attachment of carboxylic functions that are formed by oxidation of CNTs using strong acids is the most employed method to enhance their dispersibility in organic solvents.18–23 The -COOH groups of such CNTs dispersed in an organic solvent (e.g., DMF) are further reacted with amino group endcapping organic monolayers linked to conducting substrates using a classical carbodiimide coupling. This type of surface chemistry has been successfully used for the covalent attachment of CNTs to gold8,15,18,22,24–26 and ITO12 (13) Wang, M.; Shen, Y.; Liu, Y.; Wang, T.; Zhao, F.; Liu, B.; Dong, S. J. Electroanal. Chem. 2005, 578, 121–127. (14) Tsai, Y.-C.; Li, S.-C.; Chen, J.-M. Langmuir 2005, 21, 3653–3658. (15) Patolsky, F.; Weizmann, Y.; Willner, I. Angew. Chem., Int. Ed. 2004, 43, 2113–2117. (16) He, P.; Dai, L. Chem. Commun. 2004, 348–349. (17) Wang, J.; Li, M.; Shi, Z.; Li, N.; Gu, Z. Anal. Chem. 2002, 74, 1993– 1997. (18) (a) 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, 9006– 9007. (b) Chou, A.; Bo¨cking, T.; Singh, N. K.; Gooding, J. J. Chem. Commun. 2005, 842. (c) Banks, C. E.; Davies, T. J.; Wildgoose, G. G.; Compton, R. G. Chem. Commun. 2005, 829. (d) Su, L.; Gao, F.; Mao, L. Anal. Chem. 2006, 78, 2651. (19) Liu, J.; Rinzler, A. G.; Dai, H. J.; Hafner, J. H.; Bradley, R. K.; Boul, P. J.; Lu, A.; Iverson, T.; Shelimov, K.; Huffman, C. B.; Rodriguez-Macias, F.; Shon, Y. S.; Lee, T. R.; Colbert, D. T.; Smalley, R. E. Science 1998, 280, 1253– 1256. (20) Liu, Z.; Shen, Z.; Zhu, T.; Hou, S.; Ying, L.; Shi, Z.; Gu, Z. Langmuir 2000, 16, 3569–3573. (21) Nan, X. L.; Gu, Z. N.; Liu, Z. F. J. Colloid Interface Sci. 2002, 245, 311–318. (22) Diao, P.; Liu, Z. J. Phys. Chem. B 2005, 109, 20906–20913. (23) Banerjee, S.; Hemraj-Benny, T.; Wong, S. S. AdV. Mater. 2005, 17, 17– 29. (24) Huang, X.-J.; Im, H.-S.; Yarimaga, O.; Kim, J.-H.; Jang, D.-Y.; Lee, D.-H.; Kim, H.-S.; Choi, Y.-K. J. Electroanal. Chem. 2006, 594, 27–34. (25) Profumo, A.; Fagnoni, M.; Merli, D.; Quartarone, E.; Protti, S.; Dondi, D.; Albini, A. Anal. Chem. 2006, 78, 4194–4199.

10.1021/la800358w CCC: $40.75  2008 American Chemical Society Published on Web 06/05/2008

6596 Langmuir, Vol. 24, No. 13, 2008

surfaces. In contrast, the anchoring of CNTs to silicon substrates is less developed. Several published procedures resulted in the controlled adsorption of CNTs on silicon surfaces27–31 and the covalent linking of oxidized CNTs to oxidized silicon surfaces.32–34 The covalent attachment of CNTs to unoxidized silicon surfaces has been achieved using either oligo(phenylene ethynylene) aryldiazonium salts35,36 or ω-functionalized alkyl monolayers37,38 as linker units. Such procedures resulted in Si-C linked assemblies with no intervening oxide layer. In a recent communication, we demonstrated the feasibility of preparing CNT-modified silicon surfaces using an amidation reaction between soluble amino sidewall functionalized multiwalled carbon nanotubes (MWNTs) and the activated ester headgroups of an alkyl monolayer bound to Si(111).38 The MWNTs were highly soluble in common organic solvents and aqueous solutions because of sidewall organic functionalization with ammonium-terminated triethylene glycol chains.39–41 Interestingly, this functionalization route preserves the length of CNTs. Our attachment procedure offers also several other advantages. First, the molecular films produced from the reaction of hydrogen-terminated silicon surfaces with ω-substituted 1-alkenes, used in this work, are usually well-ordered, densely packed, and robust monolayers,42,43 while the reaction of diarylazonium salts with surfaces may lead to monolayers and multilayers.44 Second, activated esters react with amines under very mild conditions to form the corresponding amides in high yields. Also interestingly, the surface coverage of CNTs can in principle be controlled by either diluting the acid-terminated chains with chemically inert n-alkyl chains of the starting monolayer or changing the amount of amino groups linked to MWNTs. In this paper, we present a detailed electrochemical characterization of MWNT-insulator-silicon assemblies using cyclic voltammetry and impedance spectroscopy. The MWNT covalent attachment is verified by both scanning electron microscopy (SEM) and atomic force microscopy (AFM). Moreover, with the (26) Wei, Z.; Kondratenko, M.; Dao, L. H.; Perepichka, D. F. J. Am. Chem. Soc. 2006, 128, 3134–3135. (27) Choi, K. H.; Bourgoin, J. P.; Auvray, S.; Esteve, D.; Duesberg, G. S.; Roth, S.; Burghard, M. Surf. Sci. 2000, 462, 195–202. (28) Hertel, T.; Martel, R.; Avouris, P. J. Phys. Chem. B 1998, 102, 910–915. (29) Burghard, M.; Duesberg, G.; Philipp, G.; Muster, J.; Roth, S. AdV. Mater. 1998, 10, 584–588. (30) Liu, J.; Casavant, M. J.; Cox, M.; Walters, D. A.; Boul, P.; Lu, W.; Rimberg, A. J.; Smith, K. A.; Colbert, D. T.; Smalley, R. E. Chem. Phys. Lett. 1999, 303, 125–129. (31) Im, J.; Kang, J.; Lee, M.; Kim, B.; Hong, S. J. Phys. Chem. B 2006, 110, 12839–12842. (32) Huang, X.-J.; Ryu, S.-W.; Im, H.-S.; Choi, Y.-K. Langmuir 2007, 23, 991–994. (33) Yu, J.; Shapter, J. G.; Johnston, M. R.; Quinton, J. S.; Gooding, J. J. Electrochim. Acta 2007, 52, 6206–6211. (34) Yu, J.; Shapter, J. G.; Quinton, J. S.; Johnston, M. R.; Beattie, D. A. Phys. Chem. Chem. Phys. 2007, 9, 510–520. (35) Flatt, A. K.; Chen, B.; Tour, J. M. J. Am. Chem. Soc. 2005, 127, 8918– 8919. (36) He, J.; Chen, B.; Flatt, A. K.; Stephenson, J. J.; Coyle, C. D.; Tour, J. M. Nat. Mater. 2006, 5, 63–68. (37) Yu, J.; Losic, D.; Marshall, M.; Bo¨cking, T.; Gooding, J. J.; Shapter, J. G. Soft Matter 2006, 2, 1081–1088. (38) Hauquier, F.; Pastorin, G.; Hapiot, P.; Prato, M.; Bianco, A.; Fabre, B. Chem. Commun. 2006, 4536–4538. (39) Tasis, D.; Tagmatarchis, N.; Bianco, A.; Prato, M. Chem. ReV. 2006, 106, 1105–1136. (40) (a) Georgakilas, V.; Tagmatarchis, N.; Pantarotto, D.; Bianco, A.; Briand, J.-P.; Prato, M. Chem. Commun. 2002, 3050–3051. (b) Bianco, A.; Prato, M. AdV. Mater. 2003, 15, 1765–1768. (c) Prato, M.; Kostarelos, K.; Bianco, A. Acc. Chem. Res. 2008, 41, 60–68. (41) Bianco, A.; Kostarelos, K.; Partidos, C. D.; Prato, M. Chem. Commun. 2005, 571–577. (42) Buriak, J. M. Chem. ReV. 2002, 102, 1271–1308. (43) Wayner, D. D. M.; Wolkow, R. A. J. Chem. Soc., Perkin Trans. 2002, 2, 23–34. (44) Pinson, J.; Podvorica, F. Chem. Soc. ReV. 2005, 34, 429–439.

Fabre et al.

aim of developing novel CNT-integrated devices, we also report a UV-directed photolithography method to fabricate MWNT micropatterns on silicon. Such patterned interfaces could be useful for the controlled cell adhesion and growth, but also for the development of DNA and protein electrochemical microarrays, and functionalized CNT-based field-effect transistors for the ultrasensitive detection of biological analytes.

2. Experimental Section 2.1. Reagents. Acetone (min. 99.8%, Carlo Erba), methanol (min. 99.9%, Carlo Erba), dichloromethane (anhydrous analytical grade, SDS), tetrahydrofuran (HPLC grade, SDS), DMF (puriss, Fluka, stored over 4 Å molecular sieves), ethanol (puriss, Merck), acetic acid (analytical grade, SDS), and diisopropylethylamine (DIEA) (99.5%, Acros) were used without further purification. Acetonitrile from SDS was distilled over calcium hydride before use. The chemicals used for cleaning and etching of silicon wafer pieces (30% H2O2, 96-97% H2SO4, 50% HF, and 40% NH4F solutions) were of semiconductor grade (Riedel-de-Hae¨n). Undecylenic acid (Acros, 99%) was passed through a neutral, activated alumina column to remove residual water and peroxides. 1-Dodecene (Fluka, purity >99%) was used without further purification. N-Hydroxysuccinimide (NHS) (Acros, 98+%) and 1-(3-dimethylaminopropyl)-3-ethylcarbodiimide (EDC) (Acros, 98+%) were used as received. DiaminoPEG1500 [O,O′-bis(3-aminopropyl)polyethylene glycol 1500] was obtained from Fluka. 2.2. Preparation of Soluble Ammonium Functionalized MWNTs. MWNTs 1 were prepared from commercially available MWNT (Nanostructured & Amorphous Materials Inc., Houston, TX; stock #: 1240XH) and Trt-NH-(CH2CH2O)2-CH2CH2-NH-CH2COOH (Trt, triphenylmethyl) following our recently reported procedure.38 The amount of free amino groups of 1 was calculated at 0.50 mmol per gram of MWNTs. MWNTs 2 were prepared using oxidized MWNTs subsequently reacted with diamino-PEG1500 (see Supporting Information for the details on the synthesis and characterization). The amount of free amino groups of 2 was calculated at 0.10 mmol per gram of MWNTs. 2.3. Covalent Attachment of the COOH-Terminated Monolayer to Silicon. A single side polished silicon(111) shard (1.5 × 1.5 cm2, 1-5 Ω cm, p-type, boron doped, thickness ) 525 ( 25 µm, Siltronix) was sonicated for 10 min successively in acetone, methanol, and ultrapure 18.2 MΩ cm water. It was then cleaned in 3:1 v/v concentrated H2SO4/30% H2O2 at 100 °C for 30 min, followed by copious rinsing with ultrapure water. Caution. The concentrated H2SO4:H2O2 (aq) piranha solution is very dangerous, particularly in contact with organic materials, and should be handled extremely carefully. The surface was etched with ppb grade 40% aqueous argondeaerated NH4F for 15 min to obtain atomically flat Si(111)-H.45 It was then dipped in argon-deaerated ultrapure water for some seconds, dried under an argon stream, and transferred immediately into a Pyrex Schlenk tube containing ca. 10 mL of neat undecylenic acid, previously deoxygenated at 100 °C for 2 h at least, then allowed to cool down to ca. 30-40 °C before introducing Si(111)-H. After further argon bubbling for 30 min, the Si(111)-H surface was irradiated in a Rayonet photochemical reactor (300 nm) for 3.5 h. The carboxylic acid-modified silicon surface was rinsed copiously with tetrahydrofuran and dichloromethane, then dipped in hot acetic acid for 2 × 20 min and dried under an argon stream. It has been recently demonstrated that the rinsing in hot acetic acid leaves the functionalized surface smooth and perfectly free of physisorbed contaminants.46 2.4. Covalent Attachment of MWNTs 1 and 2 to Si(111). The terminal COOH groups were activated with NHS by immersing the modified silicon surface for 3 h in a freshly prepared mixture of a deaerated solution of EDC at 0.2 M in DMF (5 mL) and a deaerated (45) Wade, C. P.; Chidsey, C. E. D. Appl. Phys. Lett. 1997, 71, 1679–1682. (46) Faucheux, A.; Gouget-Laemmel, A. C.; Henry de Villeneuve, C.; Boukherroub, R.; Ozanam, F.; Allongue, P.; Chazalviel, J.-N. Langmuir 2006, 22, 153–162.

MWNTs Bound to Monocrystalline p-Type Si(111)

Langmuir, Vol. 24, No. 13, 2008 6597

Scheme 1. Covalent Assembly of MWNTs to Si(111) Surfacesa

a Reagents and Conditions: (i) Neat undecylenic acid, 300 nm, 3.5 h; (ii) 0.1 M NHS + 0.2 M EDC in DMF, rt, 3 h; (iii) 1 mL DMF + ca. 1 mg of MWNTs 1 + 10 µL DIEA, rt, overnight.

solution of NHS at 0.1 M in DMF (5 mL). The mixture was gently purged by bubbling with argon. The surface was then rinsed with DMF, dried under an argon stream, and used immediately for amide formation. The covalent attachment of MWNTs on Si(111) was performed by immersing the NHS-activated silicon surface in a DMF solution (1 mL) containing ca. 1 mg of 1 or 2 neutralized with DIEA (10 µL, 57 µmol) which had been previously sonicated for 10 min and deaerated with argon for 20 min. The amidation reaction was performed at room temperature overnight under an argon stream. The MWNT-modified surface was rinsed copiously with DMF, ethanol, and dichloromethane and sonicated in dichloromethane for 5 min. 2.5. Micropatterning of MWNTs 2 Bound to Si(111) Surfaces. The Si(111)-H surface was irradiated through a photomask (20 µm open squares separated by 40 µm; the mask from Photronics, UK is chromium on quartz with an antiscratch coating) with UV light (Hg(Ar) pen lamp, Oriel Model 6035) for 30 min under ambient conditions. It was then transferred immediately into a Pyrex Schlenk tube containing ca. 10 mL of neat 1-dodecene, previously deoxygenated at 100 °C for 1 h at least, then allowed to cool down to ca. 30-40 °C before introducing Si(111)-H. After further argon bubbling for 30 min, the Si(111)-H surface was irradiated in a Rayonet photochemical reactor (300 nm) for 3 h. The n-dodecyl/silicon oxide micropatterned surface was rinsed copiously with dichloromethane, methanol, and again dichloromethane, and dried under an argon stream. The surface was then dipped into 1:1 v/v H2O/ethanol mixture containing 2% HF for 2 min, dried under an argon stream, and then reacted with neat undecylenic acid as detailed in section 2.3. The covalent attachment of MWNTs 2 to the patterned surface is described in section 2.4. 2.6. Surface Characterization. AFM images were recorded in intermittent contact mode with a PicoSPM II microscope from Molecular Imaging using n+-type silicon tips (ac mode, FM, 65-90 kHz resonance frequency) from ScienTec-Nanosensors. 2.7. Electrochemical Characterizations. Cyclic voltammetry and impedance spectroscopy measurements were performed with an Autolab electrochemical analyzer (PGSTAT 30 potentiostat/ galvanostat from Eco Chemie B.V.) equipped with the GPES and FRA softwares in a self-designed three-electrode Teflon cell. The working electrode, modified Si(111), was pressed against an opening in the cell bottom using a FETFE (Aldrich) O-ring seal. An ohmic contact was made on the previously polished rear side of the sample by applying a drop of an In-Ga eutectic (Alfa-Aesar, 99.99%). The electrochemically active area of the Si(111) surface (namely, 0.7 cm2) was estimated by measuring the charge under the voltammetric peak corresponding to the ferrocene oxidation on Si(111)-H and

comparing this value to that obtained with a 1 cm2-Pt electrode under the same conditions. The counter electrode was a platinum foil and aqueous KCl saturated calomel electrode (SCE) was used as the reference electrode. Potassium ferro- and ferricyanide were from Acros (reagents ACS) and were used at a concentration of 2 mM. All electrochemical measurements were carried out inside a homemade Faraday cage in the dark, at room temperature (20 ( 2 °C) and under a constant flow of argon. Solution resistance was compensated by positive feedback. For impedance spectroscopy measurements, the amplitude of the ac signal was 10 mV and the frequency was varied from 100 kHz to 0.1 Hz with a logarithmic distribution (50 frequencies). Impedance spectra were analyzed with the EQUIVCRT program of B. A. Boukamp using the nonlinear least-squares fit method. Scanning electrochemical microscopy (SECM) measurements were performed using the CHI900B instrument from CH-Instruments equipped with an adjustable stage for tilt correction. The electrochemical cell was the one furnished with the SECM and was used in a typical three-electrode configuration for unbiased experiments. The tip electrode was a 5 µm radius Pt disk ultramicroelectrode (UME, purchased from CH-Instruments) with a typical RG ) 5-10 (RG is the ratio of the total electrode radius including the glass insulator over the radius of the Pt disk).47 The reference electrode was an Ag/AgCl, aqueous KCl 3 M electrode. UME was characterized by cyclic voltammetry and by typical approach curves recorded on a Pt conductive substrate. Imaging was performed under constant current mode at a tip-substrate distance expressed as L ) d/a around 1 (d is the distance between the tip and the substrate and a is the radius of the UME). The ferrocene/ferrocenium couple (Fc/Fc+) was used as the redox mediator at a typical concentration of 2 mM in dry N,N-dimethylformamide DMF (stored over molecular sieves, from Fluka) containing tetrabutylammonium hexafluorophosphate Bu4NPF6 (electrochemical grade, Fluka) at 0.1 M as the supporting electrolyte. All SECM experiments were performed in the dark. The potential applied to the tip electrode was chosen as to be at the level of the diffusion plateau of the mediator.

3. Results and Discussion 3.1. Covalent Attachment of MWNTs to Si(111) Surfaces Modified with NHS-Terminated Monolayers. The covalent derivatization of Si(111) surfaces by MWNTs 1 is depicted in Scheme 1. In the first step, a Si-C linked organic monolayer (47) Bard, A. J.; Mirkin, M. V. Scanning electrochemical microscopy; Marcel Dekker: New York, 2001.

6598 Langmuir, Vol. 24, No. 13, 2008

Fabre et al.

Figure 2. Cyclic voltammograms of the (a) MWNT- and (b) undecanoic acid monolayer-modified Si(111) surfaces at 0.1 V s-1 in H2O + 0.5 M NaCl + 2 mM Fe(CN)63-/Fe(CN)64-. The trace in c is the background current of the MWNT-modified electrode without the redox couple. Figure 1. SEM image of the MWNT-bound Si(111) surface.

terminated by carboxyl groups is prepared from the photochemical reaction of hydrogen-terminated monocrystalline Si(111) (Si(111)H) with undecylenic acid.46,48–53 It has been demonstrated that this direct hydrosilylation route does not lead to appreciable reaction between the carboxyl groups and the surface provided that short UV irradiation times are used (typically, less than 4 h).54 After conversion of COOH headgroups of the monolayer to N-hydroxysuccinimidyl leaving moieties, the nanotubes are covalently linked to silicon using the amidation reaction with MWNTs 1. The scanning electron microscopy (SEM) and atomic force microscopy (AFM) analyses of these surfaces show a high density of tubes lying flat on the surface rather than perpendicular, consistent with the sidewall functionalization of the MWNTs by the amino groups (Figure 1 and Figures S3 and S4 in the Supporting Information). Furthermore, the mean attached MWNTs diameter estimated by SEM and AFM falls into the range 20-60 nm which matches the mean diameter of free 1 measured by transmission electron microscopy.38 Such a result indicates that the attached MWNTs remain predominantly as individual nanotubes. It must be pointed out that no MWNT deposition was observed using a silicon surface derivatized with a simple nonfunctionalized n-alkyl (e.g., decyl) monolayer. Moreover, in order to ascertain if a strong adsorption of carbon nanotubes on silicon could result from simple electrostatic interactions between ammoniumfunctionalized carbon nanotubes and carboxylate-terminated organic chains, a control experiment has been performed. Accordingly, the immersion of an undecanoic acid monolayermodified Si(111) surface in a DMF solution containing 1 followed by rinsing steps led to a surface covered by CNTs. However, sonication of this surface in an organic solvent (e.g., in CH2Cl2) for a few minutes was sufficient to remove most of adsorbed MWNTs, which was not the case when the organic monolayer was activated with NHS. Such a result provides strong evidence that the formation of a covalent amide linkage ensures robust CNT assemblies. 3.2. Electrochemical Characterization of the MWNT Functionalized Si(111) Surfaces. Figure 2 shows a typical cyclic voltammogram obtained at the MWNT-modified electrode in (48) Mitchell, S. A.; Ward, T. R.; Wayner, D. D. M.; Lopinski, G. P. J. Phys. Chem. B 2002, 106, 9873–9882. (49) Boukherroub, R.; Petit, A.; Loupy, A.; Chazalviel, J.-N.; Ozanam, F. J. Phys. Chem. B 2003, 107, 13459–13462. (50) Boukherroub, R.; Wojtyk, J. T. C.; Wayner, D. D. M.; Lockwood, D. J. J. Electrochem. Soc. 2002, 149, H59-H63. (51) Voicu, R.; Boukherroub, R.; Bartzoka, V.; Ward, T.; Wojtyk, J. T. C.; Wayner, D. D. M. Langmuir 2004, 20, 11713–11720. (52) Perring, M.; Dutta, S.; Arafat, S.; Mitchell, M.; Kenis, P. J. A.; Bowden, N. B. Langmuir 2005, 21, 10537–10544.

contact with an aqueous solution of 0.5 M NaCl and a redox couple composed of Fe(CN)63- and Fe(CN)64-. It must be noted that the formal potential of this system E°′ has been estimated at 0.2 V vs SCE from the average of anodic and cathodic peak potentials determined by cyclic voltammetry at a platinum electrode. Indeed, any stable cyclic voltammogram with reversible waves was not obtained at a Si(111)-H surface in deaerated neutral aqueous medium within the time scale of cyclic voltammetry (Figure S5 in Supporting Information). Under these conditions, the silicon surface was found to be rapidly oxidized. In comparison to a freshly prepared p-type Si(111)-H electrode (Figure S5) the MWNT-modified electrode displays smaller currents without the presence of oxidation and reduction peaks within the -0.6 to 0.8 V potential range (Figure 2). This demonstrates, as expected, that the charge transfer through the MWNT-monolayer assembly is reduced. It must be pointed out that the observed currents correspond solely to charge transfer to Fe(CN)63-/Fe(CN)64and not to the oxidation of the underlying silicon substrate or the double-layer charging, as proven by the negligible background current recorded in the absence of the redox couple. For both the hydrogen-terminated and the MWNT-modified Si(111) electrodes, the observed cathodic currents are much smaller than the anodic currents because the concentration of available charge carriers in silicon (namely, valence band holes) is not sufficient for the reduction of ferri- to ferrocyanide to occur at a significant rate in the dark. For covalently attached molecules, it has been reported that the hole tunneling through σ-bonded alkyl chains was more efficient than electron tunneling55,56 and as efficient as electron tunneling through π-conjugated systems.57–59 For the MWNT assembly, this tunneling process should be strongly limited because of the length of the organic chain between the silicon substrate and the MWNTs.60 Importantly, the voltammetric analysis of a pure COOH- or COOC2H5-terminated alkyl monolayer without attached MWNTs reveals a totally blocking behavior as proved by the absence of a significant current response (see curve b in Figure 2).61 Such a result indicates that the rate (53) Fabre, B.; Ababou-Girard, S.; Solal, F. J. Mater. Chem. 2005, 15, 2575– 2582. (54) Asanuma, H.; Lopinski, G. P.; Yu, H.-Z. Langmuir 2005, 21, 5013–5018. (55) Selzer, Y.; Salomon, A.; Cahen, D. J. Am. Chem. Soc. 2002, 124, 2886– 2887. (56) Cahen, D.; Hodes, G. AdV. Mater. 2002, 14, 789–798. (57) Paddon-Row, M. N.; Shephard, M. J.; Jordan, K. D. J. Phys. Chem. 1993, 97, 1743–1745. (58) Shephard, M. J.; Paddon-Row, M. N. Chem. Phys. Lett. 1999, 301, 281– 286. (59) Hsu, C.; Marcus, R. A. J. Chem. Phys. 1997, 106, 584–598. (60) From calculations of energy minimization using the semi-empirical PM3 method, the fully extended length of the linker between MWNTs and the silicon surface has been estimated at ca. 26 Å. (61) Lagrost, C.; Alcaraz, G.; Bergamini, J.-F.; Fabre, B.; Serbanescu, I. Chem. Commun. 2007, 1050–1052.

MWNTs Bound to Monocrystalline p-Type Si(111)

Langmuir, Vol. 24, No. 13, 2008 6599 Table 1. Characteristic Parameters of the Equivalent Electrical Circuit Shown in Scheme 2 and Evaluated from the Fitting of the Impedance Data in Figure 3a Rs/Ω

Rsc/Ω

Csc/F cm-2

Rm/Ω

Cm/F cm-2

26

46 × 103

9.7 × 10-6

240 × 103

1.0 ×10-6

a

Figure 3. (A) Nyquist and (B) Bode plots recorded at 0.20 V vs SCE (between 100 kHz and 0.1 Hz) on the MWNT-modified Si(111) surface. The solid lines correspond to the fitted curves assuming the equivalent electrical circuit shown in Scheme 2. Electrolytic solution: H2O + 0.5 M NaCl containing Fe(CN)63- and Fe(CN)64- at 2 mM each.

Figure 4. Capacitance vs frequency plot at 0.20 V vs SCE on the MWNTmodified Si(111) surface. Scheme 2. Equivalent Electrical Circuit Used in Fitting the Data for Figure 3a

a Rs, RMWNT, Rm, and Rsc represent the uncompensated solution, the attached MWNTs, the monolayer, and the space charge resistances respectively. CMWNT, Cm, and Csc are the MWNTs, the monolayer, and the space charge capacitances, respectively.

Uncertainty (10%.

surface in the presence of the Fe(CN)63-/Fe(CN)64- couple, measured at E ) E°′ ) 0.20 V in the frequency range 105-10-1 Hz. The data are presented as plots of the imaginary part (Z′′) vs the real part (Z′) of the complex impedance (Z ) Z′ + iZ′′) as a function of frequency f (Nyquist diagram) and of the magnitude (|Z|) and phase angle of the impedance vs f (Bode diagram). The Nyquist plot is characterized by a semicircle with a diameter of ca. 280 kΩ. This is ascribed to the slow charge transfer kinetics of the Fe(CN)63-/4- couple at the modified electrode.62 The variation of the Bode plot with frequency can be summarized as follows. At the highest frequencies, the total impedance is low (30-100 Ω) and the phase angle is between 0° and 20°. As the frequency decreases, the impedance increases linearly with decreasing f up to reach a limiting value higher than 3 × 105 Ω below 1 Hz. Meanwhile, the phase angle increases to a constant value higher than 80° in the range 5-(1.5 × 103) Hz and then decreases for f < 5 Hz to values approaching 0°. Since an ideal capacitor has a phase angle of 90°, it can be concluded that the MWNT-modified surface in contact with Fe(CN)63-/Fe(CN)64- is essentially capacitive for f ranging from 5 to 1.5 × 103 Hz and resistive for the lowest and highest frequencies. Within the range 5-(1.5 × 103) Hz, the total capacitance C extracted from the data using C ) -1/2πfZ′′ is close to 0.9 µF cm-2 (Figure 4). The impedance spectra can be reasonably fitted according to the equivalent electrical circuit shown in Scheme 2 in which the three components of the assembly are considered (i.e., semiconductor, monolayer, and MWNTs). The values of the different fitted parameters are gathered in Table 1. Note that the used fitting program gave numbered resistances and capacitances corresponding to the selected electrical circuit. The assignment of each of these impedances to the different parallel elements of the interface is discussed below. Moreover, four other electrical circuit models have been tested but the quality of the fits over the whole frequency range (105 to 10-1 Hz) was found to be lower than this based on three RC components in parallel (see Supporting Information). First, the uncompensated solution resistance of about 30 Ω is perfectly consistent with that theoretically expected.63 Second, since our measurements are performed at 0.2 V, the semiconductor space-charge region is in accumulation (i.e., excess of holes) and consequently the space-charge capacitance Csc is given by eq 1:64

Csc ) of charge transfer through the covering assembly can be significantly increased by attaching carbon nanotubes, in agreement with other electrochemical data obtained at CNT-modified electrodes.18 Further information on the charge-transfer properties of the MWNT assembly can be provided by electrochemical impedance spectroscopy (EIS) using the same electrolytic medium. This technique is commonly used in investigating the electron transfer processes at metallic or semiconductor electrodes modified by organic monolayers. It has the advantage of producing electrical readout signal that can be easily analyzed in electronic circuits. Figure 3 shows the impedance spectra of the MWNT-modified

(

q2εε0Nd 2kT

) ( 1⁄2

exp

q(E - Efb) 2kT

)

(1)

where ε is the relative permittivity of silicon (11.7), ε0 is the permittivity of free space, k is the Boltzmann constant, T is temperature, q is the electronic charge, Efb is the flatband potential (62) Bard, A. J.; Faulkner, L. R. Electrochemical Methods. Fundamentals and Applications, Wiley & Sons: New York, 1980. (63) (a) The solution resistance Rs between the sample and the reference electrode can be roughly estimated from the equivalent conductivity Λ of an aqueous solution containing 0.5 M NaCl (Λ ) 93.6 Ω-1 equiv-1 cm2),63b Rs ) Fl /A with F ) 1/Λc where l is the reference electrode-sample distance, A is the area of the electrode surface, F is the resistivity of the electrolyte, and c is the molar concentration of the electrolyte. So, Rs can be estimated at 30 Ω with l ) 1 cm and A ) 0.7 cm2 as that is the case for our electrochemical cell. (b) Robinson, R. A.; Stokes, R. H. Electrolyte Solutions, Butterworths Ltd: London, 1959.

6600 Langmuir, Vol. 24, No. 13, 2008

Fabre et al.

Scheme 3. Micropatterning of MWNTs Bound to Si(111) via UV Irradiation through a Photomaska

a Reagents and Conditions: (i) 1-dodecene, 300 nm, 3 h; (ii) HF 2% in H2O/ethanol 1/1 v/v, 2 min; (iii) neat undecylenic acid, 300 nm, 3.5 h; (iv) 0.1 M NHS + 0.2 M EDC, rt, 3 h; (v) 1 mL DMF + ca. 1 mg of MWNTs 2 (n ) 33) + 10 µL DIEA, rt, overnight.

(potential at which the Fermi energy of the semiconductor lies at the same energy as the solution redox potential), and Nd is the acceptor dopant density of the semiconductor. Nd and Efb can be estimated from the slope and the ordinate of a linear Mott-Schottky plot (Csc-2 vs E, eq 2) available under depletion conditions (i.e., removal of holes in the space charge region), e.g., over the potential range -0.05 to -0.50 V.

Csc-2 )

2 kT E - Efb 2 q qεε0NdA

(

)

(2)

cm-3

A dopant density of (4.5 ( 0.5) × and a flatband potential of 0.00 ( 0.02 V vs SCE are obtained for the sample. The value of Nd is consistent with the dopant density derived from the four-probe resistivity measurements and the calculated Efb is also in good agreement with other electrochemical data reported for HF-treated p-doped silicon.65 This indicates that there is no surface dipole at the silicon-alkyl interface, as expected from the weak polarization of the Si-C bond. So, from eq 2 and the values calculated for Nd and Efb, Csc can be roughly estimated at 10 ( 4 µF cm-2 for E ) 0.2 V. Although the uncertainty in the determination of this parameter is high, it compares reasonably well with the value calculated from the fitted data (Table 1). Assuming that the organic monolayer on silicon behaves as an ideal capacitor, the theoretical Cm per unit area is given by 1015

Cm )

εmε0 d

(3)

where εm is the effective dielectric constant of the monolayer and d is the thickness of the monolayer. d can be estimated at ca. 22 Å60 considering that the chains are tilted by 35° with respect to the surface normal.66,67 This yields an εm value of 2.5. This dielectric constant compares reasonably well with the accepted value for polyethylene (2.3)68 and the values usually determined for alkyl monolayers on n-type silicon (3.3 ( 0.6)69

(64) Zhang, X. G. Electrochemistry of silicon and its oxide; Kluwer Academic: New York, 2001; p 12. (65) Laser, D.; Bard, A. J. J. Phys. Chem. 1976, 80, 459–466. (66) Bansal, A.; Li, X.; Yi, S. I.; Weinberg, W. H.; Lewis, N. S. J. Phys. Chem. B 2001, 105, 10266–10277. (67) Wallart, X.; Henry de Villeneuve, C.; Allongue, P. J. Am. Chem. Soc. 2005, 127, 7871–7878. (68) Lanza, V. L.; Herrmann, D. B. J. Polym. Sci. 1958, 28, 622. (69) Yu, H.-Z.; Morin, S.; Wayner, D. D. M.; Allongue, P.; Henry de Villeneuve, C. J. Phys. Chem. B 2000, 104, 11157–11161.

and for alkanethiols on gold (2.6).70 This relatively small value of εm suggests that the molecular chains bound to the MWNTs are closely packed and are weakly permeable to the solvent and ions. The apparent heterogeneous rate constant of charge transfer kapp corresponding to the oxidation of Fe(CN)64- through the assembly can be determined from the monolayer resistance Rm (eq 4).62

Rm )

RT n F AkappC* 2 2

(4)

where n is the number of exchanged electrons (n ) 1), F is Faraday’s constant, A is the area of the electrode surface (0.7 cm2), and C* is the bulk concentration of Fe(CN)63- or Fe(CN)64-. From eq 4 and the Rm value (Table 1), kapp is found to be ca. 1 × 10-6 cm s-1 for the MWNT-modified surface. 3.3. Micropatterning of MWNTs 2 Covalently Bound to Reactive Monolayer-Modified Si(111). The surface amidation reaction has also been used to micrometer-scale pattern the reactive silicon surface with MWNTs. The strategy used is depicted in Scheme 3 and is based on the photochemical oxidation of Si(111)-H through an optical mask.71–73 Briefly, the Si(111)-H surface is irradiated at 254 nm through a photomask (20 µm squares spaced 40 µm apart) in ambient air for 30 min, converting the exposed area into silicon oxide. The unexposed area, which remains hydrogen-terminated, is then photochemically reacted with 1-dodecene to form a n-dodecyl monolayer. The silicon oxide is then converted into a newly hydrogen-terminated surface by exposure to HF and then reacted with undecylenic acid to provide a surface patterned with a reactive acid-terminated alkyl monolayer. Finally, MWNTs 2 are covalently bound to this patterned surface using the same strategy described for MWNTs 1 for the large-scale modification. Although both the nature and the length of the linker unit substituting MWNTs 2 are different from those of MWNTs 1, they have the same properties in terms of structure and chemical reactivity toward the monolayer-modified silicon surface. The produced patterns can be visualized by exposure to water vapor (70) Porter, M. D.; Bright, T. B.; Allara, D. L.; Chidsey, C. E. D. J. Am. Chem. Soc. 1987, 109, 3559–3568. (71) Wojtyk, J. T. C.; Tomietto, M.; Boukherroub, R.; Wayner, D. D. M. J. Am. Chem. Soc. 2001, 123, 1535–1536. (72) Mischki, T. K.; Donkers, R. L.; Eves, B. J.; Lopinski, G. P.; Wayner, D. D. M. Langmuir 2006, 22, 8359–8365. (73) Fabre, B.; Wayner, D. D. M. Langmuir 2003, 19, 7145–7146.

MWNTs Bound to Monocrystalline p-Type Si(111)

Langmuir, Vol. 24, No. 13, 2008 6601

Figure 5. CCD optical images of the (A) n-C12H25/SiO2, (B) n-C12H25/-C10H20COOH, and (C) n-C12H25/monolayer-MWNT patterned surfaces exposed to water vapor. The photomask contains 20 µm squares separated by 40 µm.

Figure 6. (A,B) SEM and (C) intermittent contact mode AFM images of the n-dodecyl/MWNT-monolayer patterned surface. The inset in B is taken at a higher magnification inside one MWNT-containing pattern.

after each chemical step. For the n-dodecyl/SiO2 micropatterned surface, the water droplets are expectedly centered on the hydrophilic oxide squares (Figure 5A). The patterns are kept after the photochemical attachment of the hydrophilic acidterminated chains into the squares (Figure 5B). In contrast, the water droplets are rod-shaped after the assembly of MWNTs, evidencing more extended hydrophilic regions. Such intriguing geometrical features are difficult to explain but may be related to much smaller differences in wetting properties between MWNT- and methyl-terminated chains, compared with methyl/ silicon oxide or methyl/acid systems. The presence of MWNT patterns is clearly revealed by SEM and AFM images (Figure 6). However, instead of obtaining MWNT squares with well-defined edges, the produced MWNT network consists of diamond-shaped features interconnected between them at the level of their corners (Figure 6A,B). Such microstructures are believed to result from the aggregation of MWNTs driven by electrostatic interactions between some remaining ammonium headgroups of the organic chains substituting the MWNTs and the negative charges of COOH groups on MWNTs, which are generated during the oxidation process of the nanotubes and which are not completely reacted with the diamino-PEG chains. Indeed, the pegylation of COOH is not

exhaustive, as calculated by the Kaiser test (see Supporting Information). It is also possible that the hydrophobic interactions between nanotubes can account for the network of interlinked nanotubes from different patterned areas. To characterize the local electrochemical properties of these MWNT-micropatterned Si(111) surfaces, SECM has been used. SECM is an in situ scanning probe microscopy (SPM) that has been devoted to the quantitative investigation of a wide range of processes occurring at interfaces.47 SECM can be used when the sample is not electrically connected and as a major advantage provides a localized examination of the redox properties of the surface down to submicrometer scale.47,74,75 More specifically for our study, there are only a few publications where electron transfer at semiconductor electrodes has been examined by (74) Bard, A. J.; Fan, F.-R. F.; Pierce, D. T.; Unwin, P. R.; Wipf, D. O.; Zhou, F. Science 1991, 254, 68–74. (75) Sun, P.; Laforge, F. O.; Mirkin, M. V. Phys. Chem. Chem. Phys. 2007, 9, 802–823. (76) Mandler, D.; Bard, A. J. Langmuir 1990, 6, 1489–1494. (77) Horrocks, B. R.; Mirkins, M. V.; Bard, A. J. J. Phys. Chem. 1994, 98, 9106–9114. (78) Haram, S. K.; Bard, A. J. J. Phys. Chem. B 2001, 105, 8192–8195. (79) Bozic, B.; Figgemeier, E. Chem. Commun. 2006, 2268–2270. (80) Ghilane, J.; Hauquier, F.; Fabre, B.; Hapiot, P. Anal. Chem. 2006, 78, 6019–6025.

6602 Langmuir, Vol. 24, No. 13, 2008

Fabre et al.

Figure 7. (A) SECM approach curves obtained at the (0) MWNT-containing areas and (9) MWNT-free n-C12H25 areas in DMF + 0.1 M Bu4NPF6 containing 2 mM ferrocene as the redox mediator. The UME tip was a 5 µm radius Pt disk (RG ) 5) and the dashed line is the theoretical SECM curve for a totally insulating substrate. (B) 200 × 200 µm2 SECM image of the MWNT-micropatterned Si(111) surface performed at L around 1.

SECM.76–80 The principle of the SECM is based on the electrochemical interactions of a redox species produced at a probe electrode (tip) and the substrate under investigation.47,74,75,81 In the simplest mode used here (feedback mode), the substrate electrode, namely, the MWNT-micropatterned Si(111) electrode, is not connected (unbiased configuration) and the probe electrode is a 5 µm radius (a) platinum disk ultramicroelectrode. After diffusion of the electrogenerated species to the sample, an electrochemical reaction is possible on a localized spot on the surface (the diffusion cone of the UME) where the initial form of the mediator can be regenerated, resulting in an enhancement of the current at the probe electrode. The process is then analyzed by recording the approach curves, i.e., the normalized current It ) it/iinf versus the normalized distance L ) d/a where it is the current at the tip electrode localized at a distance d from the substrate, iinf is the steady-state current when the tip is at an infinite distance from the substrate iinf ) 4nFDCa, with n the number of electrons transferred per species and D and C the diffusion coefficient and the initial concentration of the mediator, respectively. SECM approach curves are shown in Figure 7 in the presence of ferrocene when the UME is passed over either the n-C12H25 or MWNT-containing diamond-shaped regions. For the n-C12H25 region, negative feedback is obtained which confirms the blocking properties of the dodecyl chains in agreement with our previous SECM results.80 In the case of the MWNT-containing regions, the current values corresponding to the regeneration of the ferrocene form are higher in spite of a much longer insulating linker, which is consistent with the occurrence of charge transfer at the surface of the attached MWNTs. By positioning the UME at a distance L around 1 over the sample, the SECM image clearly shows the expected current regeneration contrast between the dodecyl and the MWNT regions (Figure 7B). The pink-colored areas indicate a faster charge transfer than that in the green-colored areas. It must be noticed that both the size and the shape of the more conducting areas are similar to those of the MWNT-containing regions evidenced by SEM. However, the SECM resolution being on the order of the UME radius (in our case, 5 µm), the attached MWNTs cannot be characterized individually by this technique, and consequently, the image shown in Figure 7B provides a global view of the charge-transfer properties of the scanned surface. (81) Fan, F.-R. F.; Kwak, J.; Bard, A. J. J. Am. Chem. Soc. 1996, 118, 9669– 9675.

4. Conclusions In this work, sidewall-functionalized MWNTs have been covalently bound parallel to a silicon surface via a self-assembled acid-terminated monolayer used as an organic molecular glue. Although the insulating linker between the MWNTs and the silicon surface is longer than 2 nm, it is demonstrated that the MWNT assemblies allow efficient electrical communication between the underlying silicon substrate and a redox probe in solution. Furthermore, the charge-transfer properties of these assemblies can be in principle tuned by diluting the preassembled acid monolayer with chemically inert decyl chains. Interestingly, the amidation reaction used for the covalent attachment of the MWNTs is mild and efficient, and consequently perfectly compatible with the “reagentless” micropatterning technique based on the photooxidation of Si(111)-H as the initial step. Future work will focus on the possibility of fabricating nanometersized CNT structures on silicon. Among the modern nanopatterning techniques, direct-write dip-pen nanolithography82–84 would enable the parallel alignment and the covalent attachment of CNTs over a large range of length scales while exhibiting control over feature size. This approach would be particularly attractive for developing novel CNT-integrating electronic devices, such as field-effect transistors and chemical/electrochemical sensors. Acknowledgment. We are grateful to the University of Trieste and MUR (PRIN 2006, prot. 2006034372 and Firb RBIN04HC3S) for financial support. This work has been supported by the CNRS and the Agence Nationale de la Recherche (grants ANR-05JCJC-0031-01, Strasbourg and ANR-06-BLAN-0296-02, Rennes). Supporting Information Available: Cyclic voltammograms at Si(111)-H, different tested electrical circuits for the fitting of the electrochemical impedance spectroscopy data of the MWNT-modified silicon surface, and details on the synthesis and TEM characterization of MWNTs 2. This material is available free of charge via the Internet at http://pubs.acs.org. LA800358W (82) Piner, R. D.; Zhu, J.; Xu, F.; Hong, S.; Mirkin, C. A. Science 1999, 283, 661–663. (83) Hong, S. H.; Mirkin, C. A. Science 2000, 288, 1808–1811. (84) Ginger, D. S.; Zhang, H.; Mirkin, C. A. Angew. Chem., Int. Ed. 2004, 43, 30–45.