Self-Assembly of Trioxyethylene-Encapsulated Gold Nanoclusters

Oct 28, 2004 - Arthur W. Snow , Edward E. Foos , Melissa M. Coble , Glenn G. Jernigan , Mario G. Ancona. The Analyst ... Brian T. Holmes , Arthur W. S...
0 downloads 0 Views 205KB Size
Langmuir 2004, 20, 10657-10662

10657

Self-Assembly of Trioxyethylene-Encapsulated Gold Nanoclusters under Aqueous Conditions Edward E. Foos,* Jason Congdon, Arthur W. Snow, and Mario G. Ancona Naval Research Laboratory, Washington, D.C. 20375 Received February 3, 2004. In Final Form: September 7, 2004 The aqueous self-assembly of methyl-terminated tri(oxyethylene)thiol-encapsulated gold nanoclusters of varying core size is demonstrated on micrometer scale Au/SiO2 interdigital electrodes. This self-assembly process consists of alternate exposures of the substrate to solutions of either an R,ω-dithiol or the gold nanoclusters, resulting in the deposition of these materials onto the electrode surface. A comparison of the procedure in both H2O and CHCl3 solvents shows that the assembly, as monitored by the electrical conductivity of the device, occurs more rapidly in the H2O system. This observation is complimented by XPS and UV/Vis measurements, which show that (1) the increased current is due to an increased amount of gold deposited on the surface under aqueous conditions and (2) the thiol exchange reaction occurs more rapidly in H2O in comparison to CHCl3.

Introduction The intriguing electronic properties of gold nanoclusters have generated substantial interest in both their study and potential applications.1 A real challenge remains, however, in exploiting these properties by coupling nanoscale structures to macroscale electronics. It stands to reason that preliminary steps forward in this area might be made by combining nanoscale structures with lithographically patterned surfaces and devices through selfassembly. We have been pursuing the use of such architectures in the areas of chemical sensors2 and nanocluster-based electron transport,3 and these studies have thus far focused primarily on the use of alkanethiolencapsulated gold nanoclusters due to their ease of preparation and manipulation. Recently, we have reported4 the preparation of a methyl-terminated tri(oxyethylene)thiol-encapsulated cluster which retains many of the desirable properties of the alkane counterpart, yet is soluble in water. Compatibility with an aqueous environment is necessary if the assembly of the clusters is to be accomplished using biological macromolecules, as several studies have shown.5 Thus, the objective of the work described herein is to demonstrate the self-assembly of these gold nanoclusters on micrometer-scale electrodes using an all-aqueous solvent system. Several related studies6 have examined the stepwise assembly of citrate-stabilized aqueous gold colloids; however, there are significant differences between these systems and the ethylene oxide-based materials utilized in this work. Most importantly, the citrate colloids are stabilized in solution by an ionically charged surface and cannot be stored as a solid and redissolved after removal (1) See the following and the references contained therein: (a) Brust, M.; Kiely, C. J. Colloids Surf., A 2002, 202, 175. (b) Scho¨n, G.; Simon, U. Colloid Polym. Sci. 1995, 273, 202. (c) Scho¨n, G.; Simon, U. Colloid Polym. Sci. 1995, 273, 101. (2) Wohltjen, H.; Snow, A. W. Anal. Chem. 1998, 70, 2856. (3) (a) Snow, A. W.; Ancona, M. G.; Kruppa, W.; Jernigan, G. G.; Foos, E. E.; Park, D. J. Mater. Chem. 2002, 12, 1222. (b) Ancona, M. G.; Kruppa, W.; Rendell, R. W.; Snow, A. W.; Park, D.; Boos, J. B. Phys. Rev. B 2001, 64, 033408. (4) Foos, E. E.; Snow, A. W.; Twigg, M. E.; Ancona, M. G. Chem Mater. 2002, 14, 2401. (5) (a) Loweth, C. J.; Caldwell, W. B.; Peng, X.; Alivisatos, A. P.; Schultz, P. G. Angew. Chem., Int. Ed. 1999, 38, 1808. (b) Mirkin, C. A.; Letsinger, R. L.; Mucic, R. C.; Storhoff, J. J. Nature 1996, 382, 607. (c) Alivisatos, A. P.; Johnsson, K. P.; Peng, X.; Wilson, T. E.; Loweth, C. J.; Bruchez, M. P.; Schultz, P. G. Nature 1996, 382, 609.

of the solvent. The tri(oxyethylene)-encapsulated clusters are charge neutral and contain a layer of surface-bound thiol molecules which protect the clusters from agglomeration, allowing them to be isolated, redissolved, and handled like a chemical compound. Previous studies have shown these clusters to be amenable to thiol exchange reactions,4 and such reactions are the basis for the selfassembly protocol under investigation. While these exchange reactions have been studied for alkane-encapsulated gold clusters in organic solvents under a variety of conditions, solubility requirements have severely limited their study under aqueous conditions, and we are aware of only one specific system where aqueous exchange has been examined.7 The methyl-terminated tri(oxyethylene)thiol-encapsulated clusters studied herein are water soluble, and, while suitable for aqueous exchange experiments, it is reasonable to expect significant differences in the rates and conditions of thiol exchange under these conditions. Thus, we first test the self-assembly of these clusters using polar organic solvents under conditions similar to those studied previously and then proceed to an all-aqueous system. The comparison of these two different chemical environments will provide useful information on the self-assembly behavior of gold nanoclusters in general, data which is needed before the construction of more complex structures can be attempted. Experimental Section General Considerations. All solvents and reagents were purchased from commercial sources and used as received. H2O was triply distilled in a quartz distillation apparatus. The gold nanoclusters used in the self-assembly experiments are encapsulated with the thiol CH3(OCH2CH2)3SH and are abbreviated as AuEO3(1:1), AuEO3(3:1), and AuEO3(5:1), where EO3 refers to the number of -CH2CH2O- units in the thiol shell and the Au:thiol molar ratio used in the original synthesis is indicated in parentheses. This ratio determines the cluster core size, with (1:1) yielding the smallest clusters and (5:1) yielding the largest. Average calculated core sizes are 1.8 nm for (1:1), 2.4 nm for (6) (a) Shipway, A. N.; Katz, E.; Willner, I. ChemPhysChem. 2000, 1, 18. (b) Musick, M. D.; Keating, C. D.; Lyon, L. A.; Botsko, S. L.; Pena, D. J.; Holliway, W. D.; McEvoy, T. M.; Richardson, J. N.; Natan, M. J. Chem. Mater. 2000, 12, 2869. (c) Musick, M. D.; Pena, D. J.; Botsko, S. L.; McEvoy, T. M.; Richardson, J. N.; Natan, M. J. Langmuir 1999, 15, 844. (d) Musick, M. D.; Keating, C. D.; Keefe, M. H.; Natan, M. J. Chem. Mater. 1997, 9, 1499. (7) Templeton, A. C.; Cliffel, D. E.; Murray, R. W. J. Am. Chem. Soc. 1999, 121, 7081.

10.1021/la0497043 CCC: $27.50 © 2004 American Chemical Society Published on Web 10/28/2004

10658

Langmuir, Vol. 20, No. 24, 2004

Scheme 1. Illustration of the Assembly Procedure. (1) Exposure, (2) Rinse/Dry, and (3) Measurement.

(3:1), and 3.6 nm for (5:1). Full details of the synthesis, purification, and characterization of these clusters have been reported elsewhere.4,8 The dithiol HS(CH2CH2O)2(CH2)2SH was prepared according to literature procedures.9 Current measurements of the self-assembly deposition onto the electrodes were made using a Keithley Model 617 Electrometer at a constant bias of 50 mV.10 XPS measurements were recorded on a VG 220iXL with a monochromatic Al KR source, using a pass energy of 20 eV. UV/vis measurements were made on a ThermoSpectronic Unicam UV 500 instrument, using 1-cm-path-length quartz cells. Cluster Self-Assembly Procedure. The electrodes used in these experiments consist of an interdigital gold electrode patterned on an SiO2 surface, fabricated at NRL. The dimensions of the electrode are as follows: 50 finger pairs with 15-µm spacing between the gold fingers, 4800-µm overlap length, and a thickness of 1500 Å. Immediately prior to deposition, the electrodes were cleaned by immersion in CHCl3 while sonicating for 2 min. Once dry, the substrate was transferred to a 90 °C solution of 1:1:3 30% H2O2:NH4OH:H2O (by volume) for 10 min. The substrate was then rinsed twice in distilled H2O and dried using heated air. It was immediately placed in a 1.0% w/w solution of HS(CH2CH2O)2(CH2)2SH in CHCl3 for 15 min in order to functionalize the gold electrode surface. The substrate was then immersed twice in CHCl3 and immediately transferred to a 5.0% w/w solution of HS(CH2)3Si(OCH3)3 in heptane for 15 min to functionalize the SiO2 surface. The substrate was rinsed twice in heptane and then placed back into the dithiol solution for 15 min to displace residual mercaptosilane which may have bonded to the gold electrode surface. The substrate was again rinsed twice with CHCl3 and finally placed in a 0.5% w/w solution of the gold nanocluster in CHCl3 for 15 min to allow binding of the clusters to the surface linking molecules. The substrate was rinsed twice with CHCl3, dried, and the current measured using a pair of custom designed pressure clips attached to the electrometer. This measurement procedure is illustrated in Scheme 1. For subsequent depositions, the substrate was immersed in the dithiol or cluster solution for 2 min, alternating between the two, rinsed twice in CHCl3, and the current measured prior to placing it in the next solution. Typically, eight cycles were performed, where “cycle” is defined as an exposure to both the dithiol and cluster solutions with a measurement taken after each. In experiments where H2O was used as the solvent, all procedures were performed as described above, substituting H2O for CHCl3. The dithiol HS(CH2CH2O)2(CH2)2SH was found to have a solubility of 0.2 mg/mL in water, based upon turbidity onset in a 300-nm scattering versus concentration experiment using 1 cm quartz cells. At the 1.0% w/w concentration used in the above procedure, agitation formed stable milky white suspensions which were suitable for use. XPS Measurements. SiO2 substrates were used for the XPS measurements, and depositions were conducted in a manner similar to that described above. The substrates were first immersed in a 90 °C solution of 1:1:3 30% H2O2:NH4OH:H2O (by volume) for 10 min. The substrate was then rinsed twice in distilled H2O and dried using heated air. It was immediately placed in a 5.0% w/w solution of HS(CH2)3Si(OCH3)3 in heptane for 15 min. This was followed by rinsing twice with heptane and immersion in the 0.5% w/w solution of the gold nanocluster in (8) Foos, E. E.; Snow, A. W.; Twigg, M. E. J. Cluster Sci. 2002, 13, 543. (9) Snow, A. W.; Foos, E. E. Synthesis 2003, 509. (10) Snow, A. W.; Wohltjen, H. Chem. Mater. 1998, 10, 947.

Foos et al. Scheme 2. Self-assembly Process Used in the Construction of Gold Nanocluster Electrode Devices

CHCl3 for 15 min. The substrate was then rinsed twice with CHCl3. Once the depositions were completed, samples were stored in a desiccator until placed in the load-lock of the instrument. This procedure was varied for certain control experiments by (1) skipping the HS(CH2)3Si(OCH3)3 deposition step or (2) preparing the cluster solution in H2O instead of CHCl3. UV/Vis Cluster/Dithiol Precipitation Study. Solutions of approximately 0.01 wt% were prepared of both the gold nanocluster and the dithiol HS(CH2CH2O)2(CH2)2SH in CHCl3. The solutions were combined and thoroughly mixed, and an aliquot transferred to a quartz cuvette. An initial spectrum was taken immediately at t ) 0, followed by spectra every 45 min for the first 40 h and then spectra every 1.5 h for the remainder of the experiment, which was terminated at t ) 72 h. An identical procedure was used for the second study, except that H2O was used to prepare the solutions in the place of CHCl3. Control experiments were performed using gold nanocluster solutions prepared in CHCl3 or H2O at identical concentrations to those described above, excluding the dithiol.

Results and Discussion The stepwise self-assembly approach used in these experiments has been described in detail elsewhere.3a Briefly, the surface of interest, which usually contains regions of both Au and SiO2, is functionalized sequentially with an R,ω-alkanedithiol and mercaptopropyl trimethoxysilane. In an idealized picture, one end of the dithiol binds to the gold, while the silane attaches to the SiO2, producing a surface containing exposed thiol groups which are available for binding. Exposure of this surface to a solution of the gold nanoclusters results in a thiol-exchange reaction between the surface-bound thiols and the alkanethiols attached to the gold core, resulting in binding of the nanoclusters to the surface. It should be noted that with this procedure, a single deposition is not expected to produce a single monolayer of clusters on the surface, but rather some fraction of this maximum, and this picture is consistent with our electrical measurements (vide infra) and previous XPS studies.3a Subsequent exposure of the sample to a dithiol solution repeats this exchange process, binding dithiols to the surface-bound clusters and providing an anchoring point for the next cluster exposure. By the repetition of these steps, gold nanoclusters can be built up on the surface stepwise almost indefinitely. This process is illustrated in Scheme 2. When the substrate used is insulating, the cluster buildup can be followed by monitoring the conductivity between appropriately configured electrodes. Figure 1 shows typical data from an alkane-capped cluster/alkane dithiol self-assembly experiment performed in CHCl3 as

Trioxyethylene-Encapsulated Gold Nanoclusters

Figure 1. Self-assembly data for the hexanethiol-encapsulated AuC6(1:1) cluster using octanedithiol and CHCl3 as the solvent.

described above. As the number of cluster layers increases, the current recorded from the electrode rises as expected. The other interesting feature in the figure is the decrease in conductivity seen after dithiol exposurespresumably caused by some clusters being removed from the surface by the dithiol and into solution. It is unlikely that this feature is due to any process other than cluster-removal since extended exposure of the device to a monothiol solution results in complete loss of current (vide infra). Following a cluster exposure in the next step of the experiment, the current recovers and increases slightly, consistent with the picture of additional nanoclusters being added. Our experiments have also shown a critical dependence of this assembly process on the presence of both the dithiol and the mercapto-silane surface-linking agents, as well as on the length of the dithiol chain.3a Control experiments conducted in the absence of each of these linking agents hinders current buildup, as does use of a dithiol which is either too short to span the alkanethiol shells and connect adjacent cores, or too long to allow electron tunneling between the gold cores. As shown by previous studies of the ethylene oxideencapsulated gold nanoclusters,4 the stabilizing thiol in this system exhibits a tradeoff between solubility and conductivity over a very narrow size range. If the ethylene oxide chain is too short, then the cluster is insoluble in H2O. As the length of this ligand chain increases, however, the conductivity is diminished, although it remains higher than that of the alkane-capped clusters when the two are compared on the basis of chain length. Thus, the AuEO3 cluster is a good compromise between solubility and conductivity and is the focus of the self-assembly studies described below. Conductivity is also related to core size, and we have found that the larger-core EO3-encapsulated clusters retain their water solubility and were also examined in this study. As previously stated, in selfassembly studies of the hexanethiol-encapsulated gold nanocluster system, the length of the dithiol chain was found to be extremely important and octanedithiol was found to give the highest overall current after multiple cycles.3a For consistency, here we have also chosen to use a dithiol containing an eight-atom chain length, HS(CH2CH2O)2(CH2)2SH. The substitution of two oxygen atoms would presumably provide compatibility with both aqueous solvents and the ethylene oxide shell of the nanoclusters.

Langmuir, Vol. 20, No. 24, 2004 10659

Figure 2. Self-assembly data for the AuEO3(1:1) cluster series. The solvents are listed in the order of dithiol/cluster.

Figure 3. Self-assembly control experiments for AuEO3(1:1).

Figure 2 shows the results of self-assembly experiments for the AuEO3(1:1) cluster using HS(CH2CH2O)2(CH2)2SH as the linker molecule. Three different sets of conditions were utilized in order to better explore the effect of aqueous and polar organic solvents on this system. First, the experiment was performed with both the cluster and dithiol solutions prepared using CHCl3 as the solvent. This was done to provide a direct comparison between this system and the alkanethiol-based system examined previously. For the second experiment, the dithiol was maintained in CHCl3 while the cluster solution was prepared in H2O. The final set of conditions utilized both cluster and dithiol solutions prepared using H2O. The most striking feature of Figure 2 is the significantly higher current that was obtained from the device when H2O was used as a solvent for both the dithiol and the cluster. This trend was a feature common to all the cluster systems studied (vide infra). Several control experiments were performed on this system as well to gauge the impact of the mercapto-silane and dithiol as surface- and cluster-linking agents. The data are shown in Figure 3. When the mercapto-silane is removed from the system, the current build-up on the device is hindered during the early cycles but quickly returns to a level comparable to what is seen in the normal experiment. This indicates that the silane coupling agent plays a role

10660

Langmuir, Vol. 20, No. 24, 2004

Foos et al.

Table 1. AuEO3(1:1) H2O/H2O Results for Self-assembly onto Interdigital Electrodes experiment

current after 8 cycles (nA)

1 2 3 4 5

2.8 3.7 4.0 9.0 7.4

in the formation of the initial cluster bonds to the surface, but once enough clusters have been deposited to provide anchor sites for the dithiol, the dithiol plays a much more important role. This is substantiated by removing the dithiol from the system. As seen in the plot, dithiol removal results in significantly retarded current build-up. The current which does eventually develop in this experiment could be the result of successive nonspecific attachments of the clusters to the SiO2 surface and/or each other. In an attempt to probe the reproducibility of this system, the results of multiple AuEO3(1:1) self-assembly experiments are tabulated in Table 1. Some variation in the results can be expected due to factors such as the slight difference in the gold core size of nanoclusters obtained from different preparative batches or variations in the surface of the substrates, although such differences were minimized whenever possible. The large currents seen relative to CHCl3 when H2O is used as the solvent also indicate that the amount of water present in the CHCl3 may be a factor in these experiments; however, no attempt to control the dryness of the solvents was made. Generally, current values above 1 nA are necessary for a stable reading with minimal drift. As indicated in Table 1, AuEO3(1:1) cluster experiments in the all-aqueous system give final current values from 1 to 10 nA, which is sufficient for use in MIME sensor experiments. Similarly, the larger clusters generally give final currents in the range of tens of nanoamps for AuEO3(3:1) or hundreds of nanoamps for AuEO3(5:1). Once formed, the self-assembled devices are stable and in most cases provide current readings in line with the data even after many months of storage. Exposure of the device to a solution of octanethiol in CHCl3 overnight results in a loss of all current, consistent with cluster removal from the surface through a thiol-exchange reaction.3a This experiment also implies that the ethylene oxide clusters are not depositing on the surface through an irreversible agglomeration process since the material deposited on the surface in such a scenario would likely not be susceptible to such exchange reactions. Figure 4 shows the data for the self-assembly of the AuEO3(3:1) cluster under the same conditions as those described above. The trends are similar, although the currents are higher by roughly a factor of 10. This is due to the larger gold core size of the cluster, and this increased conductivity has been observed previously.10 The data for the H2O/H2O system also seems slightly better behaved than what is seen in Figure 2, with an almost linear increase in the conductivity of the nanocluster deposition step observed. Similarly, Figure 5 shows the data from the AuEO3(5:1) system. A further increase in the current over the AuEO3(3:1) system is seen, although the magnitude of the change (approximately a factor of 4) is not as large as that observed between the (1:1) and (3:1) clusters. Also of note is the fact that the current increases consistently during the early cycles of the deposition, including the dithiol steps when a current decrease is expected. It is possible that this is due to a large amount of nonspecific deposition of the gold nanoclusters during

Figure 4. Self-assembly data for the AuEO3(3:1) cluster series. The solvents are listed in the order of dithiol/cluster.

Figure 5. Self-assembly data for the AuEO3(5:1) cluster series. The solvents are listed in the order of dithiol/cluster.

the cluster deposition step. Subsequent exposure to the dithiol solution binds these clusters into positions more amenable to electron transport, leading to the observed increase. The most noticeable feature of the data presented in Figures 2-5 is the large increase in overall current which is observed for experiments where H2O is used as the solvent for both the dithiol and cluster solutions. A possible explanation for this observation is that the water may interact more favorably than CHCl3 with the ethylene oxide structure through hydrogen-bonding interactions, causing the shell of the cluster to swell. This would allow the dithiol to penetrate the shell and bind with the cluster core more easily than is possible in the all-CHCl3 system. The fact that the mixed CHCl3/H2O system lies between these two extremes is consistent with this argument. To confirm that the increased current observed in the H2O/H2O system is due to the increased deposition of gold clusters onto the surface of the device, we have utilized XPS measurements to examine SiO2 surfaces exposed to the nanoclusters in CHCl3 and H2O solutions. The results of these measurements are presented in Table 2 as atomic ratios of C:Si and Au:Si. The SiO2 surfaces in Samples 1 and 3 were functionalized with mercaptopropyl tri-

Trioxyethylene-Encapsulated Gold Nanoclusters

Langmuir, Vol. 20, No. 24, 2004 10661

Figure 6. UV/vis data for the AuEO3/HS(CH2CH2O)2(CH2)2SH precipitation study: (a) reaction conducted in CHCl3; (b) reaction conducted in H2O. Spectra were recorded every 45 min for the first 40 h then every 1.5 h until t ) 72 h. Table 2. XPS Data for the Deposition of AuEO3 on SiO2 in CHCl3 and H2O sample

deposition

C:Si

Au:Si

1 2 3 4

HS(CH2)3Si(OCH3)3; AuEO3/CHCl3 AuEO3/CHCl3 HS(CH2)3Si(OCH3)3; AuEO3/H2O AuEO3/H2O

0.44 0.45 2.2 0.43

0.04 0.05 0.60 0.06

methoxysilane prior to exposure to the nanocluster solution and represent the deposition taking place on the SiO2 portions of the interdigital electrodes. In Samples 2 and 4, the silane was not used and these samples give some indication of the amount of nonspecific binding which may occur on an untreated surface. Clearly, the gold deposition is greatly enhanced when H2O is used as the solvent in the presence of a thiol-functionalized surface, consistent with the large current increases observed in this system. This suggests that the exchange chemistry is proceeding at a faster rate in H2O, which leads to an increased number of clusters attaching to the surface after the same amount of exposure to the substrate. To further probe this thiol exchange, we have also examined the solution reaction of the dithiol HS(CH2CH2O)2(CH2)2SH with the ethylene oxide-encapsulated gold clusters in both H2O and CHCl3. As this reaction proceeds, the clusters are expected to precipitate as they form large linked aggregates by exchange reactions with the dithiol. If monitored by UV/vis spectroscopy, as this precipitation occurs, the absorbance of the clusters in the supernatant decreases. The data from these experiments are shown in Figure 6. It is clear that the absorbance at 518 nm, where the surface plasmon of the gold cluster occurs, drops off much more rapidly in the H2O experiment than when CHCl3 is used as the solvent. The absorbance falls to a value approximately 80% of the original after 3 h in H2O, while the corresponding absorbance drop in CHCl3 occurs

Figure 7. Plot showing the change in absorbance at 518 nm during the first 14 h of the experiment shown in Figure 6.

after 72 h. In addition, there is a noticeable redshift in the H2O data within the first 0.75-1.5 h of the reaction, with an analogous shift in the CHCl3 data present after approximately 5 h. These changes are highlighted in Figure 7, which shows the absorbance change at 518 nm for the two solvents during the first 14 h of the experiment. Control experiments conducted over the same time period on cluster solutions containing no dithiol showed an approximately 3% difference in the absorbance intensity, with no redshifting of the 518-nm surface plasmon peak. While somewhat qualitative, this experiment nonetheless supports the hypothesis that thiol exchange occurs more rapidly in H2O in this system.

10662

Langmuir, Vol. 20, No. 24, 2004

In conclusion, we have demonstrated the self-assembly of methyl-terminated tri(oxyethylene)thiol-encapsulated gold nanoclusters onto lithographically patterned micrometer-scale electronic devices in aqueous solutions. An interesting and potentially useful result of these studies is the enhanced cluster deposition that is observed when H2O is used as the solvent in place of CHCl3, and this has been confirmed through both conductivity and spectroscopic measurements. These studies demonstrate that the aqueous self-assembly of these nanoclusters proceeds in an analogous manner to the assembly of alkane-encapsulated nanoclusters in polar organic solvents. Currently,

Foos et al.

we are working to exploit this result in more-complex systems,11 as well as understand the reasons for the increase in cluster deposition under these conditions. Acknowledgment. The Office of Naval Research (ONR) is gratefully acknowledged for financial support of this work. LA0497043 (11) (a) Jhaveri, S. D.; Foos, E. E.; Lowy, D. A.; Chang, E. L.; Snow, A. W.; Ancona, M. G. Nano Lett. 2004, 4, 737. (b) Jhaveri, S.; Foos, E. E.; Ancona, M. G.; Snow, A. W.; Twigg, M. E.; Chang, E.; Pilobello, K.; Goldman, E.; Lowy, D. Mater. Res. Soc. Symp. Proc. 2003, 735, 159.