Nanotubule-Based Molecular-Filtration Membranes

See allHide authors and affiliations

Science  24 Oct 1997:
Vol. 278, Issue 5338, pp. 655-658
DOI: 10.1126/science.278.5338.655


Polymeric membranes that contain a collection of monodisperse gold nanotubules, with inside diameters of molecular dimensions (less than 1 nanometer), were used in a simple membrane-permeation experiment to cleanly separate small molecules on the basis of molecular size. For example, when such a membrane was presented with an aqueous feed solution containing pyridine (molecular weight 79) and quinine (molecular weight 324), only the smaller pyridine molecule was transported through the nanotubules and into a receiver solution on the other side of the membrane.

Membrane-based chemical separations are potentially more economical and easier to implement than competing separations technologies, but membranes with higher transport selectivities are required (1-5). Chemical features of the membrane, and of the molecule to be selectively transported, that can be exploited to enhance transport selectivity include charge, chemical interactions, and molecular size. The ideal approach for implementing molecular size–based transport selectivity would be to design a synthetic membrane that has a collection of monodisperse nanopores, of molecular dimensions, that span the complete thickness of the membrane. If this could be accomplished, one could envision general “molecular filters” that could be used in a simple membrane-permeation experiment to cleanly separate small molecules on the basis of size. We describe such membranes here.

This idea of using membranes to filter molecules on the basis of size is not without precedent. Dialysis is used routinely to separate low–molecular weight species from macromolecules (6). In addition, nanofiltration membranes are known for certain small-molecule separations (such as water purification) (7), but such membranes typically combine both size and chemical transport selectivity and are particularly designed for the separation involved. And, although zeolites have molecule-sized pores, they are not typically used for membrane-based chemical separations (8). Hence, in spite of the importance of the concept, synthetic membranes that contain a collection of monodisperse, molecule-sized pores that can be used as molecular filters to separate small molecules on the basis of size are currently not available.

The membranes described here were prepared from polycarbonate filters (Poretics) that contain monodisperse, cylindrical pores 30 nm in diameter (9, 10). Martin and co-workers have shown (9,11, 12) that an electroless plating procedure can be used to deposit an Au nanotubule within each pore (Fig.1A). The inside diameter (ID) of these Au nanotubules can be varied by varying the plating time [see figure 1 in (11)]. At sufficiently long plating times, Au nanotubules with IDs of molecular dimensions (<1 nm) are obtained (11,13).

Figure 1

Schematic illustrations of the shapes of the Au nanotubules that we obtained by doing the electroless Au plating (9) at (A) pH = 10 and (B) pH = 12. The higher pH causes bottleneck tubules. The tubules plated at the lower pH also have some of this bottleneck character [see (11)]. Hence, the depictions in both (A) and (B) are approximate and serve to illustrate the conceptual differences between the two types of nanotubules investigated here.

We have since discovered that the shape of the Au nanotubule can be changed by varying the rate of the plating reaction. When high plating rates were used (14), Au was preferentially deposited on the faces of the membrane, and nanotubules with bottlenecks at both ends but a larger ID in the middle (Fig. 1B) were obtained. Such bottleneck tubules are a form of ultrathin film composite membrane (3,7) and should provide high permeate flux without sacrificing transport selectivity. This is so because selectivity should be determined by permeation through the bottleneck, but overall flux is determined by permeation in the larger ID tubule that spans the membrane (see below).

We show here that membranes containing such bottleneck nanotubules can be used to filter molecules on the basis of size. We mounted the nanotubule membrane in a U-tube permeation cell such that the membrane separated a feed solution from a permeate solution (15). The feed solution was equimolar in two compounds of different molecule size. We call these the “smaller” and the “larger” molecules; three “smaller molecule–larger molecule pairs” were investigated here (Fig. 2, A to C). The permeate solution, initially just pure water, was periodically assayed for the presence of both the smaller and the larger molecules. In all three cases (Fig. 2), easily measurable quantities of the smaller molecule were obtained in the permeate solution, but the larger molecule was completely undetectable.

Figure 2

Chemical structures and approximate relative sizes of the three pairs of molecules studied here: (A) methyl viologen chloride and Ru tris(2,2′-bypiridine) chloride, (B) anilinium chloride and rhodamine B chloride, and (C) pyridine and quinine. The charges on the molecules of each pair are the same. Because the permeate solution was initially just pure water, the cations [(A) and (B)] come across the membrane with their charge-balancing anions.

We began our studies, however, with simpler single-molecule permeation experiments (16). These experiments were done in the same U-tube permeation cell (15), but the feed solution contained only one of the molecules of the smaller molecule–larger molecule pair. The flux of this molecule across the nanotubule membrane was determined, and then in a separate experiment the flux of the other molecule of the pair was measured (17). The objective was to explore the effect of nanotubule ID on transport rate and selectivity. Nanotubules with the more conventional shape (Fig. 1A) were used for these preliminary studies.

Results of such single-molecule permeation experiments, based on the use of the methyl viologen (MV2+) Ru(II)tris(2,2′-bipyridine) [Ru(bpy)/3 2+] pair (Fig. 2A) and membranes with four different nanotubule IDs (13), are shown in Fig. 3, A to D. The slopes of these permeation curves define the fluxes of MV2+ and Ru(bpy)3 2+ across the membrane. One can obtain a permeation selectivity coefficient (αi) (18) by dividing the MV2+flux by the Ru(bpy)3 2+ flux.

Figure 3

Amounts of moles of MV2+ and Ru(bpy)3 2+ transported versus time. Membranes contained nanotubules as depicted in Fig. 1A with diameters (13) of (A) 5.5 nm, (B) 3.2 nm, and (C) 2.0 nm. The nanotubule diameter in (D) was too small to measure by the gas-permeation method (diameter < 0.6 nm). Only MV2+ was transported through this membrane. The data in (D) show more noise than the data in (A), (B), and (C) because the flux is lower.

The data in Fig. 3 may be summarized as follows: (i) Even for the nanotubules with the largest ID investigated (5.5 nm), αiwas substantially greater than the ratio of the diffusion coefficients (D) for these molecules in aqueous solution [αi = 50; ratio of diffusion coefficients = 1.5 (19)]. Hence, size-based sieving occurred in these large ID (>>molecular dimensions) nanotubules (20). (ii) As the nanotubule ID decreased, the fluxes for both molecules decreased; however, the flux of the larger Ru(bpy)3 2+decreased more rapidly. As a result, αi increased with decreasing nanotubule ID (21). Values for the nanotubule membranes with IDs of 5.5, 3.2, and 2.0 nm are αi = 50, 88, and 172, respectively. (iii) The nanotubule membrane with the smallest ID (Fig. 3D) showed a measurable flux for MV2+, but the larger Ru(bpy)3 2+ could not be detected in the permeate solution, even after a 2-week permeation time.

Although the results of these experiments were encouraging in terms of selectivity, the fluxes were low. We expected that higher fluxes could be obtained from the bottleneck nanotubule membranes (Fig. 1B). A simple single-molecule permeation experiment was done to demonstrate the bottleneck character of these nanotubules. The MV2+flux in a typical bottleneck membrane is 19 nmol hour−1cm−2. When the surface Au layer containing the bottleneck (Fig. 1B) was removed (12), the flux increased by one order of magnitude to 180 nmol hour−1 cm−2. When the second surface Au layer was removed, the flux again increased by an order of magnitude. That these very thin surface Au layers (150 nm versus 16 μm for the polycarbonate membrane) can have such a dramatic effect on the flux clearly shows that flux-limiting constrictions are present in the surface layers (Fig. 1B).

It is also important to prove that these bottleneck membranes can have the same selectivity but higher flux than membranes containing the conventional nanotubules (Fig. 1A). To demonstrate this point, we compared the rate and selectivity of transport across a conventional and a bottleneck membrane (22). Both membranes were able to cleanly separate MV2+ from Ru(bpy)3 2+ in the two-molecule permeation experiment (see below). Hence, these membranes showed comparable, excellent selectivity. However, as expected, the flux of MV2+ across the bottleneck nanotubule membrane was dramatically higher than that of the conventional nanotubule membrane (14 versus 0.07 nmol hour−1 cm−2).

We now turn to the more interesting case of having both molecules of a pair in the feed solution together. These experiments were done only on bottleneck nanotubule membranes that showed αi = ∞ (22). Typical results, for the pyridine-quinine pair, are shown in Fig. 4. The absorption spectra for 0.5 mM quinine and for 0.5 mM pyridine are shown in Fig. 4A. Pyridine shows a characteristic peak at ∼252 nm. Quinine shows a much more intense band centered at 225 nm and two other bands at ∼280 and ∼330 nm.

Figure 4

Absorption spectra for (A) 0.5 mM pyridine and 0.5 mM quinine, (B) the pyridine-quinine feed solution (both molecules 0.25 mM), and (C) permeate after transport for 72 hours through a bottleneck nanotubule membrane (22).

The absorption spectrum for the feed solution used in the permeation experiment is shown in Fig. 4B. Although both molecules are present in solution at the same concentration, the higher absorbance of the quinine nearly swamps out the 252-nm peak of the pyridine. The absorption spectrum of the permeate solution after 72 hours is shown in Fig. 4C. In spite of the higher absorbance of the quinine (larger molecule), only the peak for the pyridine (smaller molecule) is seen in this spectrum. The very intense quinine band centered at 225 nm is absent. To our ability to make the measurement, this bottleneck nanotubule membrane has filtered these two molecules on the basis of molecular size (Fig. 4C).

To verify this point, we used a much more sensitive analytical method, fluorescence (23), to search for traces of quinine in the permeate solution. The magnitude of the absorbance in Fig. 4C indicates that the pyridine concentration in the permeate is 7 × 10−5 M. With fluorescence analysis (23), it is possible to detect 5 × 10−9 M quinine in the presence of 7 × 10−5 M pyridine. However, no quinine fluorescence could be detected from the permeate solution. These data show that, to our (now much more sensitive) ability to make the measurement, this membrane has cleanly separated these two molecules and that if any quinine is present in the permeate solution, its concentration is less than 5 × 10−9 M.

These analytical data can be used to calculate a minimal selectivity coefficient, αmin. Because the concentration of the smaller molecule in the permeate solution was 7 × 10−5 M and the concentration of the larger molecule (if present at all) must be less than 5 × 10−9 M, the minimal selectivity coefficient for the pyridine-quinine pair is αmin = 15,000. Minimal selectivity coefficients obtained in this way (24) for the other pairs are shown in Table1. In all three cases, the larger molecule was undetectable in the permeate solution.

Table 1

Minimal selectivity coefficients for the three different small molecule–large molecule pairs.

View this table:

Nishizawa et al. have demonstrated that these Au nanotubule membranes can show charge-based transport selectivity (9), and we have shown here that these membranes can also have molecular size–based selectivity. It seems likely that chemical transport selectivity can also be introduced. Hence, these nanotubule membranes hold promise for the development of highly selective membranes for chemical separations.

  • * To whom correspondence should be addressed at crmartin{at}


View Abstract

Navigate This Article