Letter pubs.acs.org/NanoLett
Electron Conductive and Proton Permeable Vertically Aligned Carbon Nanotube Membranes Gregory A. Pilgrim,† Joanne W. Leadbetter,† Fen Qiu,† Anni J. Siitonen,† Steven M. Pilgrim,§ and Todd D. Krauss*,†,‡ †
Department of Chemistry and ‡The Institute of Optics, University of Rochester, Rochester, New York 14620, United States Inamori School of Engineering, Alfred University, Alfred, New York 14802, United States
§
S Supporting Information *
ABSTRACT: We report the fabrication of membranes hundreds of micrometers thick that demonstrate efficient electron conduction and proton transport through vertically aligned arrays of multiwalled carbon nanotubes (NTs) impregnated by epoxy. Electrical transport was Ohmic with a conductivity of 495 mS cm−1. Protons traversed the membrane through the NT bore with a current of 5.84 × 10−6 A. Good electron and proton transport, chemical robustness, and simple fabrication suggest NT membranes have potential in artificial photosynthesis applications.
KEYWORDS: Vertically aligned carbon nanotubes, membrane, hydrogen production, ionic conduction, electrical conduction, water splitting
M
the NT length through the bore,11,12 especially when driven by an electrical potential.13 The ability to vary the NT diameter and functionalize the ends also means ionic transport through NTs could be controlled based on size exclusion14 and/or chemoselectivity.15,16 Use of nanostructures as the basis for a membrane has the benefit of localizing transport properties into chemically distinct regions. These regions, as a result of the high surface area to volume ratio of NTs, are attractive attachment points for molecules that can perform other chemistries, such as catalysis.17−20 Here we present a facile synthesis of freestanding, vertically aligned carbon nanotube (VANT) membranes, characterize their electron conductivity and proton transport, and demonstrate their successful use in an integrated system with colloidal CdSe quantum dots to generate photocurrents. By synthesizing long NTs in a vertically aligned array, their excellent electron transport properties can direct electrons over micrometer distances. Vertical alignment also facilitates proton transport over long distances through the bore. Although the bulk electronic21 and proton transport properties15,22 of carbon NT membranes have been previously explored in separate applications, it is extremely rare to have them both integrated into the same system. Furthermore, the membranes described here provide for and allow the demonstration of the attachment of photoreceptive systems that make it particularly compelling
any potential uses exist for a membrane that can simultaneously transport protons and electrons under aqueous conditions. For example, during photosynthesis biological systems, notably photosystems I and II,1,2 use coupled transport of electrons and protons to move components for reduced chemical fuels through membranes while maintaining charge balance. Analogous inorganic systems supporting transport of protons and electrons are of significant interest for solar fuel technologies, i.e., the generation of hydrogen gas by photocatalyzed water splitting during artificial photosynthesis. Among the many technical requirements of such a system as presently conceived is a membrane capable of transporting both photogenerated electrons and the ionic components of the desired chemical fuel, i.e., protons. Although single materials that transport both protons and electrons have been reported, including tungsten oxides3,4 and various ceramics,5−7 they typically require operating temperatures in excess of 300 °C and are not compatible with aqueous hydrogen production systems. Nanostructured membranes that incorporate a mixture of electron-conducting and protonconducting materials provide the necessary functionality; however, their fabrication can be complex and expensive. Carbon nanotubes (NTs) are an attractive alternative for a solar fuel membrane material as they are electrically conductive and provide an avenue for proton transport as well. It is wellknown that the extended π structure of carbon nanotubes allows ballistic electron transport along their sidewalls8,9 and that pristine NTs have outstanding electron conductivity.10 Additionally, positive and negative ions in solution can traverse © 2014 American Chemical Society
Received: October 3, 2013 Revised: January 19, 2014 Published: February 24, 2014 1728
dx.doi.org/10.1021/nl403696y | Nano Lett. 2014, 14, 1728−1733
Nano Letters
Letter
in solar energy applications. Additionally, this work demonstrates that proton and electron conducting membranes can be produced with significant complexity reductions compared to other nanostructured membranes, while maintaining efficient charge transport performance, thus providing new potential capabilities for solar fuel technologies. The general scheme for membrane fabrication is shown in Figure 1 and includes three basic steps: NT growth, epoxy coating, and NT exposure. Vertically aligned nanotube arrays (VANTs) were grown via chemical vapor deposition on a crystalline silicon substrate in a 2.5 cm diameter single opening 23 tube furnace using variations of published methods.24−26 Using electron beam evaporation, 10 nm of alumina support followed by 12 nm of iron catalyst were deposited onto a silicon wafer topped by 500 nm of thermal oxide. A square section of silicon wafer roughly 2.5 cm2 was removed by etching with a diamond scribe followed by cleaving. The sample was placed in a quartz boat, introduced into a tube furnace, and heated to 800 °C under an argon atmosphere flowing at 595 sccm. Once at temperature, the catalyst was annealed under flowing hydrogen gas (13 sccm) for 15 min, after which hydrogen flow was ceased and argon flow resumed. Ethylene gas was added (545 sccm) for another 15 min to grow the NTs, which were subsequently cooled to room temperature under an argon atmosphere (further details in the Supporting Information). An electron microscope image of the vertically aligned NTs is shown in Figure 2A. The vertically aligned nanotube arrays grown via this process are multiwalled, 100−150 μm tall, with 15−20 nm outer diameters, 5−10 nm inner diameters, and 15−20 walls. Overall, 16 VANT arrays were fabricated with similar characteristics, and six of those arrays were used for the proton and electron crossing experiments. As-grown carbon nanotube arrays are fragile and are unable to form a freestanding and robust membrane by themselves. Thus, a second material was required to both fill the interstitial space between the NTs and to give the membrane rigidity. The filler material needed to have several important qualities: it must be resistant to oxidation and chemically inert to organic and polar solvents, while also being inexpensive, and easy to manufacture, ideally amenable to solution-based processing. Although Si3N4 has been used to seal the voids between vertically aligned NTs to form a rigid membrane,27 the Si3N4 deposition process is extremely expensive and impractical for the long NTs in this study. Polystyrene and polydimethylsiloxane (PDMS) were also evaluated as possible filler materials but were found to not fully penetrate the NT array. We chose to form freestanding membranes by controlled impregnation of the VANTs with a commercially available, two-part epoxy, Epotek 301. The epoxy was spin-coated at 1000 rpm for 15 s onto the arrays where it filled the interstitial spaces between the nanotubes28 (see Supporting Information). In order to conduct electrons, the NT ends must be exposed and in order to transport protons and the end-caps must be removed, as illustrated in Figure 1C−E. To expose the NT tips above the epoxy, a few drops of isopropyl alcohol were spun at 800 rpm for 10 s onto a still-wet epoxy-coated VANT array. The cross-linking agent (1,4-butanediol diglycidyl ether) is more soluble in alcohols than the resin (1,6-diamino-2,2,4(2,4,4)-trimethylhexane). Thus, the addition of the alcohol dissolved some of the cross-linker to a depth of 15−25 μm, allowing it to be spun off. After curing at room temperature overnight, those areas lacking cross-linker were uncoated by
Figure 1. General scheme for fabrication of a VANT membrane: (A) nanotube array growth on a silicon wafer (green) alumina support (blue) and iron catalyst (red); (B) epoxy coating (gray); (C) tip exposure; (D) removal of wafer and catalyst; (E) tube opening via ozone treatment. Components are not shown to scale.
epoxy as shown in Figure 2B. The silicon substrate was dissolved following epoxy impregnation by immersion in a hot 8 M KOH solution for approximately 2 h, depicted in Figure 1D. Removal of the substrate left behind the iron catalyst and alumina support, which block the NT bore and prevent proton 1729
dx.doi.org/10.1021/nl403696y | Nano Lett. 2014, 14, 1728−1733
Nano Letters
Letter
vertically aligned carbon nanotube columns with a 400 μm pitch instead of full arrays also have much lower conductivities, approximately 1/1000th of the values reported here. The loss in conductivity is again a result of the membrane having significant surface area and volume composed of nonelectrically conductive material. Proton transport through the membrane was characterized by monitoring the absorption spectrum of bromophenol blue, a common pH indicator,33−35 in a specially designed glassware system (Figure S5). The membrane was placed between two vessels: one vessel contained 0.1 M HCl and the other an equal volume of 4 μM bromophenol blue indicator solution. During the filling of the apparatus the membrane was carefully monitored for leaks, which would manifest as liquid seeping through the membrane from a filled chamber to the other, as yet unfilled, chamber. Indicator concentration was chosen to account for the weaker absorption of the protonated indicator species. Acid and indicator volumes were kept equal to eliminate any pressure driven proton transport. As protons cross through the membrane from the low pH to the high pH side, the concentrations of protonated and deprotonated indicator species changed, which was quantified by a Beer’s law analysis of the absorption spectrum. The VANT membrane, with an area of 0.38 cm2 exposed to each solution, was presoaked in deionized water for 48 h before the crossing experiment in order to allow water to infiltrate the nanotube bores and establish a conduit for proton transport. While it is clear both from this experiment and from the literature11,27 that water can be transported though carbon nanotubes, we see no evidence of, or motivation for, water flow through the NTs after the water initially enters the NT bores. Water levels remain constant on both sides of the membrane throughout proton crossing experiment. A 0.8 V potential difference was placed across the apparatus, which served to drive positively charged hydrogen ions across the membrane, preclude crossing by negatively charged chlorine anions, and supply electrons to balance charge across the membrane. Hydrogen ions have been shown to travel in liquid water via a “hopping” Grotthuss mechanism.36 Water itself, being neutral, is not driven across the membrane by the applied voltage. The spectral changes associated with protons crossing the membrane are shown in Figure 3. The peak at 593 nm (absorption by the deprotonated indicator) decreases as the peak at 449 nm (absorption by the protonated species) increases proportionally, indicating that protons were crossing through the NT membrane. Proton population on the indicator side increases linearly with time (Figure 3 inset) and corresponds to an average proton current of 5.84 × 10−6 A. The proton current in the VANT membranes is an order of magnitude greater than those measured in unaligned, epoxycoated NTs,16 even when correcting for differences in applied voltage and membrane geometries. Membranes composed of unaligned NTs are fabricated by mixing NTs, in powder form, into an epoxy, curing the epoxy, and using a microtome to cut the resulting block into strips. Membranes fabricated in this way have NTs in no well-defined orientation with respect to the membrane planes. We propose that the significant improvement here may be due to the greater NT density and orientation control gained by using aligned NT arrays as a membrane basis compared to a disordered NT/epoxy mixture. By way of control NT membranes were also evaluated without ozone opening and without removal of the iron and alumina catalysts. In both cases no proton transport was observed.
Figure 2. Images of the VANT membrane. (A) Scanning electron microscope image of vertically aligned NTs. The scale bar is 10 μm, and the NTs are approximately 150 μm tall. (B) Nanotube arrays following epoxy impregnation and isopropyl alcohol treatment. Scale bar is 10 μm, and approximately 20 μm of NT length is exposed above the bulk. Inset: photograph of the final, self-supporting membrane approximately 2.5 cm2.
transport. Both catalyst and support were removed by mechanical abrasion with 400 grit alumina sandpaper to visual inspection, followed by a 5 min immersion in concentrated hydrochloric acid (11 M). In this manner the NTs on the bottom of the membrane were exposed as shown in Figure 1D. Lastly, the membrane was subjected to ozone treatment (Figure 1E) to remove any fullerene caps on the top surface.29 The percentage of NTs opened was not able to be quantified; however, fullerene caps are etched quickly and preferentially over NT walls.29 Further, exposed SWNTs are completely destroyed by ozone treatment in about 7 min.30 Thus, the speed at which ozone destroys fullerenes and NTs leads us to believe 10 min of ozone treatment is more than enough to open all MWNTs in the membrane. The ozone opening process also introduces carboxylate groups31 onto the nanotubes, preferentially at the NT ends. The end result is a freestanding membrane containing open bored carbon nanotubes transversing the bulk and with exposed ends protruding from each surface (Figure 2 inset). Assuming 100% of the NTs are open, the resulting areal density is 3 × 1010 per cm2. Note that the morphology of the NT membrane before and after ozone treatment was unchanged as measured with SEM. Electrical conduction and proton transport properties were evaluated separately. Resistance measurements were taken using a Hewlett-Packard 4192A Impedance Analyzer. Samples with surface areas approximately 0.063 cm2 were coated with 200 nm of gold contact via thermal evaporation before resistances were measured.17 The device resistance was 0.497 Ω. The corresponding conductivity was measured to be 495 ± 12 mS cm−1 for a typical sample, which represents the average of 10 measurements at three separate probe positions on the sample surface. The reported standard deviation is reflective of those 30 measurements. This conductivity is ∼20 times higher than values for a previously reported silicon nanowire array/ Nafion/PEDOT-PSS17 membranes and ∼20 times lower than the best electron-conducting polymer (i.e., PEDOT:sPPO) films.32 This improvement in resistivity compared to silicon nanowire array/Nafion/PEDOT-PSS membrane is reasonable because the silicon nanowire array has a 7 μm pitch between conductive silicon nanorods. The pitch is nonconductive space, which lowers the number of potential electron pathways per area. In comparison, our VANT membranes have interstitial spaces on the order of 10 nm, making for a much higher density of electron pathways. Other membranes20 incorporating 1730
dx.doi.org/10.1021/nl403696y | Nano Lett. 2014, 14, 1728−1733
Nano Letters
Letter
membranes thinner than 123 μm because they were too fragile for iron/alumina removal and/or mounting in the proton crossing apparatus. We also measured proton current through a 125 μm Nafion film in an identical experiment to that used to evaluate the VANT membranes. Proton crossing, as depicted in Figure 4, was approximately 2 times faster in the Nafion film than in the VANT membrane. Untreated Nafion is reported to have a proton conductivity of 78 mS cm−1,37 similar to that reported for silicon/Nafion/ PEDOT-PSS membranes,17 roughly 4 times that of reported PEDOT:sPSS membranes,32 and 9 times that of a mixed VANT column/Nafion membrane despite serving Nafion as the proton transport material.20 Additionally, because proton transport in the VANT membrane only occurs through the carbon NT bore, the active area for proton transport in the VANT membrane is less than the Nafion film. However, a direct comparison to published membranes and a proton conductivity value are inappropriate for the VANT membrane reported here. NTs do not conduct protons but rather can provide a means for their transport. As a consequence of this lack of conductivity (and therefore resistivity), VANT membranes do not contribute resistance to proton flow beyond what is termed the electrolyte resistance (Relec) as discussed in the literature.17 A membrane such as the silicon/Nafion/ PEDOT-PSS membrane17 or a PEDOT:sPSS membrane32 or Nafion itself20 has a characteristic device resistance (Rmem) caused by the proton transport material (Nafion, sPSS) in addition to the electrolyte resistance, thus decreasing the net ionic current through the material (Rcell = Rmem + Relec).17 Increasing NT density would increase both proton transport and electron conductivity by providing more pathways per membrane area. Furthermore, narrowing the nanotube bore diameter may yield improvement in proton transport properties up to a certain minimum diameter.38 While narrower nanotubes have been grown with similar methods to ours, the well-documented difficulties39 in getting NT arrays with a
Figure 3. Temporal evolution of absorption spectra for bromophenol blue indicator during course of proton transport experiments. The peak at 593 nm corresponds to nonprotonated indicator while the peak at 449 nm corresponds to protonated indicator. Note the isosobestic point at around 500 nm. Proton crossing time goes from 0 to 225 min (black to purple) in 15 min intervals. Inset depicts change in concentration of protonated (red line) and nonprotonated (blue line) indicator.
These controls prove proton transport though the NT bore rather than though, for example, cracks in the membrane or unfilled interstitial spaces in the NT array. Furthermore, no cracks or unfilled interstitial spaces were observed in examination of the membranes with SEM. We also evaluated proton transport for thicker VANT membranes, 150 μm thick instead of 123 μm (Figure 4), and saw a 4-fold decrease in ionic current. For thin NT membranes, it was proposed that the ionic current depends inverse linearly on VANT array thickness (ref 16). Thus, it is unclear why such a small relative change in thickness would have such a large change on the transport. One possible explanation is that if some fraction of the NTs are not very straight for the taller arrays, then they would have a much longer path for proton crossing, and thus a simple inverse linear relationship may not apply here. We were unable to obtain proton crossing data on
Figure 4. Comparison of the disappearance of bromophenol blue indicator peak at 593 nm corresponding to a decrease in nonprotonated indicator species. Rate of disappearance, proportional to rate of proton crossing, is roughly twice as large in Nafion (blue, squares) as in 123 μm NT membranes (red, circles). A 4-fold decrease in proton crossing rate was observed for the 150 μm thick membranes (black, triangles). 1731
dx.doi.org/10.1021/nl403696y | Nano Lett. 2014, 14, 1728−1733
Nano Letters
Letter
Figure 5. Peak photocurrent generated in VANT membrane upon irradiation with 440 nm light (blue). Troughs occur when the light source was physically blocked, preventing irradiation. Red signal is the same experiment performed as a control on a VANT membrane without quantum dots on the surface.
required; this yields a significant savings in complexity as well as an increase in electron and proton conductivity. Fourth, the VANT membranes provide large surface area and clear points for the attachment of light absorbing or catalytic molecules like those involved in water splitting. Other heterogeneous systems, such as disordered polymer systems, make attachment of molecules to one polymer (and therefore one system− electronic or ionic) difficult as polymers are intermixed. Catalyst molecules intended to interact with the electronic system could only be attached to areas of the membrane where PEDOT:PSS was exposed on the surface. As fabricated, these membranes leave the nanotube π system available not only for electron transport but also for potential chemical modification, making them adaptable to use as sieves, sensors, electrodes, or other applications.18,19,43,44 Mixed polymer systems without nanoscaffolds, while displaying excellent conductive properties, have limited functionality in this regard due to their low surface roughness.32 In summary, we have reported here a simple, two-component membrane based on vertically aligned carbon nanotube arrays and commercially available epoxy that is capable of conducting electrons and provides a means for the transport of protons and electrons. Additionally, these membranes display other desirable characteristics including low cost, excellent durability (both chemical and physical), macroscale size, and the potential to incorporate other chemistries. This collection of properties is unique among materials and results directly from the nature of carbon nanotubes. On the other hand, while VANTs are promising for potential artificial photosynthesis applications, more studies are required to see if they indeed can function in that application. For example, the work function of multiwalled carbon nanotubes has been reported45,46 to be −4.3 eV, comparable to graphite. This value is low enough to reduce protons to hydrogen in water only at extremely low pH and only with a catalyst that has a low (
