Martin
Pumera
* and
Yuji
Miyahara
International Center for Materials Nanoarchitectonics, and Biomaterials Center, National Institute for Materials Science, 1-1 Namiki, Tsukuba, Ibaraki 305-0044, Japan. E-mail: pumera.martin@nims.go; Fax: (+)81-29-860-4714
First published on 28th August 2009
As carbon nanotubes become a part of everyday life, the paramount questions about their toxicity persist. While it has been proven that a significant part of the toxicity of carbon nanotubes is due to the redox activity of the residual metallic impurities present within them, there is still no response to the fundamental question of where the borderline is that would render the impurities ‘redox invisible’. Herein we investigate the electrochemical response of carbon nanotubes containing different amounts of impurities towards the reduction of an important biomarker , hydrogen peroxide, and the oxidation of an important impurity marker, hydrazine. We found that the borderline between being redox active/inactive for iron-based impurities lies in the middle-ppm range.
A whole spectrum of methods for purifying CNTs is available. However, they are often not very effective and CNTs containing 1% or less of residual metallic impurities are considered ‘high quality’.15 Such a large amount of impurities is not acceptable for release into the environment. For example, just 1 ton of CNTs with 1 wt% of metallic impurities would include 10 kg of toxic metallic nanoparticles. Given the fact that it is expected that millions of tons of CNTs will be produced worldwide,16 the presence of toxic metallic impurities within CNTs is a problem of paramount importance. Therefore, one might ask an important question—“what amount of metallic impurities in carbon nanotubes is small enough to not dominate their redox properties?”.
We investigated the involvement of metallic impurities in the redox properties of CNTs by electrochemistry. Before discussing the experiments any further, it is important to make the following remarks. The electrochemistry of CNTs resembles that of graphite—heterogeneous electron transfer is fast at the ends of the CNTs and at their defect sites (similar to the edge plane of graphite) while pristine CNT walls (similar to the basal plane of graphite) show very low rates of heterogeneous electron transfer.17–22 It has also been shown that residual metallic catalyst impurities in CNTs are, in some cases, responsible for the ‘exceptional electrocatalysis’ of CNTs. It has been demonstrated that iron-based impurities are responsible for the ‘electrocatalysis’ of hydrogen peroxide23,24 and hydrazine25 while copper nanoparticle impurities occluded within multi-wall carbon nanotubes (MWCNTs) are responsible for the ‘electrocatalysis’ of halothane.26 We have shown that multicomponent (Mo/Co/Fe) metallic impurities within CNTs catalyze the electrochemical oxidation of hydrazine.27 Recently, we have demonstrated that the issue is even more complex, and that even traces of iron within non-catalytic cobalt impurities can catalyze the reduction of hydrogen peroxide.4 Furthermore, we showed that the bi-component nickel and iron residual catalyst impurities within SWCNTs are responsible for the electrochemical oxidation of arginine, where Fe and Ni participate, and histidine, where only Fe participates.28
In this article, we investigated this paramount question “what amount of metallic impurities in carbon nanotubes is small enough to not dominate their redox properties?” using examples of the reduction of hydrogen peroxide and the oxidation of hydrazine on CNTs containing different amounts of residual metallic impurities. Hydrogen peroxide is one of the most important biomarkers and thus naturally one of the most important analytes for biosensing. In addition, electrochemical oxidation of hydrazine is very sensitive to small amounts of metallic impurities and therefore we used hydrazine for the confirmation of redox accessible impurities within CNTs.29,30 Because there is no available standard reference for CNT material,31,33 we chose five CNT samples with distinctively different (on the order of several magnitudes) amounts of impurities to determine the threshold below which impurities do not play a significant role in the redox properties of CNTs.
A transmission electron microscope equipped with an energy dispersive X-ray analyzer (TEM/EDX) was used to observe the structure of the CNTs and ascertain the presence of impurities. For example, Fig. 1 shows TEM micrographs of CNT-A, B, C, D, and E. As can be seen, the CNTs have profoundly different structures. This plays an important role in their electrochemistry, as we show later. CNT-A and B have diameters of approximately 5–10 nm and consist of a mixture of two-, three-, and four-walled nanotubes (generally called few-walled CNTs, FWCNTs) while CNT-C, and D have a herringbone structure with open ends, an outer diameter of 80–100 nm and contain an inner channel with a diameter of about 40–50 nm. CNT-E consists of hundreds of concentric graphene tubes. It is clear from the TEM micrographs of CNT-A, B, and C that these CNTs contain a significant number of embedded metallic impurities (dark contrast dots; see arrows in Fig. 1). In contrast, CNT-D and E contain no observable metallic impurities (note that the observable high-contrast parts of the walls of CNT-C, D and E result from electron diffraction by the CNT lattice and these are not related to metallic impurities).34TEM/EDX analysis of the impurities in samples CNT-A, B, and C confirmed that they contain Mo, Co, and Fe in the cases of CNT-A and B (see the selected TEM images and EDX spectra in Fig. 1) in different ratios. CNT-C contains only Fe nanoparticles. TEM/EDX analysis of the residual metallic nanoparticles within CNT-A, B, and C showed that the average composition of the metallic impurities is as follows. In CNT-A: Co 94.28; Mo 4.70; and Fe 0.52 atm%. (n = 28); in CNT-B: Co 71.53; Mo 25.68 ; and Fe 1.48 atm% (n = 8); and in CNT-C: Fe 100% (n = 9). It should be noted here that TEM/EDX is not a statistically representative method. It is used here as an additional analytical method. ICP-AES and magnetic susceptibility measurements provide statistically reliable values. CNT-D and E are virtually impurity-free and it was not possible to localize/analyze any impurities under TEM and TEM/EDX.
![]() | ||
Fig. 1 TEM images (left) and EDX spectra (right) of metallic impurities in ‘washed’ carbon nanotubes. (A) CNT-A, (B) CNT-B, (C) CNT-C, (D) CNT-D, and (E) CNT-E. Scale bar in TEM images is 10 nm (A, B) and 100 nm (C, D, and E). |
In order to establish the influence of metallic impurities on the electrochemical redox behavior of CNTs, we selected two important markers, hydrogen peroxide and hydrazine. Hydrogen peroxide is a highly important biomarker , which is prone to being electrocatalyzed on Fe particles but not on other metals.4Hydrazine is an important marker of the presence of impurities because a wide variety of metals is able to electrocatalyze its oxidation.29
We first started with the evaluation of the redox response of hydrogen peroxide (H2O2). Fig. 2 exhibits the voltammetric responses related to the electrochemical reduction of 5 mM H2O2 on CNT-A, CNT-B, CNT-C, CNT-D, CNT-E, and Fe3O4NP-modified BPPG electrodes. A corresponding voltammogram of the reduction of H2O2 on a bare BPPG electrode is also shown. The significant reduction of H2O2 on the CNT-A modified BPPG electrode (blue line) starts at −0.02 V and reaches a maximum around −0.70 V. The reduction of H2O2 on the CNT-B modified BPPG electrode (purple line) starts at −0.10 V and reaches a maximum at −0.48 V. The reduction of H2O2 on the CNT-C modified BPPG electrode (orange line) starts at −0.24 V and reaches a maximum at −0.49 V. These potentials are similar to those observed previously using MWCNTs with Fe(III) impurities4,23 and in agreement with the reduction of H2O2 on an Fe3O4NP-modified BPPG electrode (green line), which starts around −0.10 V. In contrast, the CNT-D and E modified electrodes (magenta and red lines, respectively) display negligible reduction current starting at around −0.45 V. The reduction of H2O2 on a bare BPPG electrode (black line) also starts at around a potential of −0.45 V. Thus, it is clear that the electrochemistry of CNT-D and E is dominated by edge-plane defects and ends of CNTs (similar to the electrochemistry of bare BPPG, which is governed by edge-like defect sites) and not by Fe-based impurities.
![]() | ||
Fig. 2 Cyclic voltammograms resulting from the electrochemical reduction of 5 mM H2O2 at BPPG electrodes modified with CNT-A (blue line), CNT-B (purple line), CNT-C (orange line), CNT-D (magenta line), CNT-E (red line), Fe3O4NPs (green line), and at bare BPPG (black line). Conditions: scan rate, 100 mV s−1; background electrolyte, 50 mM phosphate buffer, pH 7.4. |
Next, we investigated the influence of metallic impurities upon the oxidation of hydrazine because, as mentioned above, hydrazine (N2H4) is very sensitive to the presence of metallic impurities in general. Fig. 3 displays cyclic voltammograms corresponding to the electrochemical oxidation of 5 mM N2H4 at CNT-A, B, C, D, and E and Fe3O4NP-modified BPPG electrodes. We also show the corresponding control voltammogram of bare BPPG. The significant wave of the oxidation of N2H4 at CNT-A (blue line) starts around −0.02 V and reaches a maximum at +0.42 V. The oxidation of N2H4 at a CNT-B modified BPPG electrode (purple line) starts at +0.14 V while of a CNT-C modified BPPG electrode (orange line) at +0.18 V. The oxidation of N2H4 on an Fe3O4NP-modified electrode (green line) starts at +0.15 V. In contrast, the oxidation of N2H4 on a CNT-D modified BPPG electrode (magenta line) starts around +0.45 V while that on a CNT-E modified BPPG electrode (red line) displays negligible oxidation current for the oxidation of N2H4, starting slowly at +0.90 V. This is similar to the response of bare BPPG where negligible oxidation current is also observed around +0.90 V. The difference between the behaviors of CNT-D and E is attributed to their different structures, where the ‘herringbone’ CNT-D exhibits significantly larger numbers of edge-like sites on its surface, thus enhancing the heterogeneous electron transfer rates, while CNT-E has a tubular structure and its ‘edge-like’ defects are limited to nanotube ends and surface defects. The density of such defects in the case of CNT-E is lower than that in the case of CNT-D.35 This electrochemical behavior is analogous to the oxidation of N2H4 on BPPG and EPPG.27 Thus, we can conclude that the electrochemistry of CNT-D and E is dominated by the edge-like defects and ends of the CNTs and not by metallic impurities. The observed data are in good agreement with a previous report of the oxidation of N2H4 at MWCNTs containing iron-based impurities and at bare BPPG.25,27
![]() | ||
Fig. 3 Cyclic voltammograms resulting from the electrochemical oxidation of 5 mM N2H4 at BPPG electrodes modified with CNT-A (blue line), CNT-B (purple line), CNT-C (orange line), CNT-D (magenta line), CNT-E (red line), Fe3O4NPs (green line), and at bare BPPG (black line). Conditions: scan rate, 100 mV s−1; background electrolyte, 50 mM phosphate buffer, pH 7.4. |
The comparison of the response of the CNT-A, B, and C and Fe3O4NP-modified BPPG electrodes led us to the conclusion that iron oxide impurities (in the case of hydrogen peroxide) and metallic impurities (in the case of hydrazine) are responsible for the observed redox behavior of CNT-A, B, and C. The reason for this is that minute amounts of electroactive nanoparticles with small random array coverage might exhibit behavior similar to that of related macroelectrodes because of greatly overlapping diffusion layers, especially when interconnected with CNT nanowires.36–38 It should be noted that the average thickness of the diffusion layer (according to the conditions of our voltammetric experiments) is around 128 and 115 µm for hydrogen peroxide and hydrazine, respectively.4,27TEM images show that the distance between the residual metal catalyst nanoparticles, which perform as metallic nanoelectrodes, is on the order of magnitude from the tens to the hundreds of nanometres. Therefore, their diffusion layers overlap a great deal and the metal nanoparticles are responsible for the observed redox action of CNT-A, B and C.
In order to respond to our principal question “what amount of metallic impurities in carbon nanotubes is small enough to not dominate their redox properties?” the following items should be considered. First, we should emphasize that hydrogen peroxide could ‘see’ only iron-based impurities and none of the other metals present in the CNTs used in this study.4 In contrast, hydrazine is prone to electrocatalytic oxidation on any metal surface that is present in the CNT sample as an impurity. Therefore, the electrochemically visible content of impurities for the reduction of hydrogen peroxide is 450 ppm for CNT-A, 100 ppm for CNT-B, 14000 ppm for CNT-C, 10 ppm for CNT-D, and 3 ppm for CNT-E. The electrochemically noticeable content of impurities for hydrazine is 45550 ppm for CNT-A, 15100 ppm for CNT-B, 14000 ppm for CNT-C, 10 ppm for CNT-D, and 3 ppm for CNT-E. Interestingly, it is clear that even 100 ppm of Fe impurities within a CNT sample (such as in the case of CNT-B), which is significantly lower than the detection limits of the SEM/EDX or XPS methods and close to the limit of detection of ICP-AES,1,33 still dominate the redox properties of hydrogen peroxide. This is very disturbing for real-world applications of CNTs because the amount of metallic impurities is usually not estimated with such trace analysis methods. One could thus say that the borderline where iron-based impurities stop dominating the redox properties of CNTs lies somewhere between 100 and 10 ppm. However, the potential influence of the structure of the CNTs should also be considered. Both CNT-A and CNT-B are FWCNTs with up to four walls and a small diameter (about 5–10 nm). The CNT-C, D, and E samples contain large nanotubes with diameters of about 90 nm. CNT-C and D exhibit open ends and large inner channels with diameters of about 50 nm and wall thicknesses of about 25 nm. The mechanism of interaction of the molecules with biomarkers is thought to be via intercalation into the defects of the nanotubes or via metal release from the CNTs.2,3,6,7 The maximum proven depth of inclusion of metallic impurities within CNTs in which they still participate in the electrochemistry is 13 nm.2 If one limits the depth of such intercalation or release in the timescales presented here to 13 nm, then the amount of electrochemically accessible iron-based impurities in the CNT-C and D samples is 100% (because of the large inner channel). However, the electrochemically accessible amount of impurities in the CNT-E sample would be only about 46% of the total amount (assuming homogeneous distribution of the impurities in the nanotube in the form of ‘atomic doping’, as described in ref. 1). Thus, the electrochemically accessible concentration of impurities within the CNT-E sample might be only 1.38 ppm. Therefore, it can be established that the borderline for the amount being “small enough to not dominate the redox properties of CNTs” lies between 100 and 10 ppm for hydrogen peroxide and somewhere between 15100 and 10 ppm for hydrazine. Although the range of the borderline for hydrazine might seem too wide, one should note that from a biomedical point of view, the borderline for hydrogen peroxide is much more important.
This journal is © The Royal Society of Chemistry 2009 |