Ultrasensitive detection of toxic cations through ... - Semantic Scholar
ARTICLES PUBLISHED ONLINE: 9 SEPTEMBER 2012 | DOI: 10.1038/NMAT3406
Ultrasensitive detection of toxic cations through changes in the tunnelling current across films of striped nanoparticles Eun Seon Cho1,2† , Jiwon Kim3† , Baudilio Tejerina3 , Thomas M. Hermans3 , Hao Jiang4 , Hideyuki Nakanishi3 , Miao Yu2 , Alexander Z. Patashinski3 , Sharon C. Glotzer4 , Francesco Stellacci1,2 * and Bartosz A. Grzybowski3 * Although multiple methods have been developed to detect metal cations, only a few offer sensitivities below 1 pM, and many require complicated procedures and sophisticated equipment. Here, we describe a class of simple solid-state sensors for the ultrasensitive detection of heavy-metal cations (notably, an unprecedented attomolar limit for the detection of CH3 Hg+ in both standardized solutions and environmental samples) through changes in the tunnelling current across films of nanoparticles (NPs) protected with striped monolayers of organic ligands. The sensors are also highly selective because of the ligand–shell organization of the NPs. On binding of metal cations, the electronic structure of the molecular bridges between proximal NPs changes, the tunnelling current increases and highly conductive paths ultimately percolate the entire film. The nanoscale heterogeneity of the structure of the film broadens the range of the cation-binding constants, which leads to wide sensitivity ranges (remarkably, over 18 orders of magnitude in CH3 Hg+ concentration).
M
any cations are toxic, posing serious consequences for the environment and for human health. As an example, organomercury species (for instance, methylmercury (ii); refs 1,2) are extremely dangerous and cause neurological disorders such as Minamata disease3,4 , and cadmium poisoning results in kidney and/or liver failure, bone softening and other ill effects (for example, Itai-itai disease5,6 ). Therefore, sensitive and selective detection of toxic metal cations is of paramount importance in assessing water pollution and in the elimination of potentially serious health and environmental hazards7–9 . There are multiple approaches to detect these ions, including atomic absorption spectroscopy10,11 and inductively coupled plasma spectroscopy12,13 (ICP), a host of optical methods (colorimetric14,15 or fluorescence-based assays16,17 and systems based on surface plasmon resonance18,19 or surface-enhanced Raman scattering20,21 ), and electrochemical methods22,23 . Frequently, however, these methods require a sophisticated chemistry to incorporate host materials, such as macrocyclic rings (crown ethers, calixarenes or porphyrins), metal-ion-recognizable proteins, nucleic acids and polymers. Moreover, the sensitivity of these methods is insufficient24–27 in several situations of public health concern (for example, accumulation of low picomolar concentrations of CH3 Hg+ and Cd2+ in fish28,29 ). Furthermore, host–guest chemistries are not available for all cations, with CH3 Hg+ being a prominent example2 . Here we show that gold nanoparticles (Au NPs) coated with binary mixtures of n-hexanethiol (HT) and alkanethiols terminated with n = 1, 2 or 3 ethylene glycol (EG) units have a striped ligand structure (Fig. 1a,b) that generates supramolecular
pockets able to selectively trap ions, and that on such trapping the electron-conduction mechanism through films of these NPs changes markedly30 . We implement these findings in a solid-state sensor of unmatched sensitivity: HT/EG3 NPs sense CH3 Hg+ selectively and with a detection limit of ∼1 aM, HT/EG2 NPs sense ∼1 pM concentrations of Cd2+ , and HT/EG1 NPs are ∼1 nM sensors for Zn2+ . Au NPs (d = 4.7 ± 0.9 nm) were synthesized as described previously31,32 (see Methods) and were functionalized with alternating stripes (Fig. 1a) of HT and EGn thiols (HS–(CH2 )6 – (OCH2 CH2 )n –OH, ultrapure grade, ProChimia). These striped NPs were examined by scanning tunnelling microscopy (STM): the width of both HT and EG2 stripes was ∼1.3 ± 0.1 nm (Fig. 1b). For control experiments, we prepared NPs coated with disordered HT/EGn mixed self-assembled monolayers (SAMs; random NPs (ref. 33)), but with the same relative content of the two thiols as in the striped NPs (see Supplementary Section S1), and also NPs covered with one-component SAMs of EGn thiols. All types of NPs were dispersed in methanol with a concentration of 2 mM of Au atoms. To fabricate the sensors, Au electrodes (30 nm thick, 5 mm long) were sputter-coated onto a glass substrate using a shadow mask; the gap between the two electrodes was typically w = 50 µm wide and l = 5 mm long (Fig. 1c). Approximately 3 µl of the Au NP solution was drop-cast onto the patterned glass and dried for two days in vacuum conditions to give NP films about H ∼ 150 nm thick (determined by profilometry for each sample); in this way, the number of NPs between the electrodes was of the order of 1011 (Fig. 1d). To prevent re-dissolution of the film on exposure
1 Department
of Materials Science and Engineering, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139, USA, 2 Institute of Materials, Ecole Polytechnique Fédérale de Lausanne (EPFL), 1015, Switzerland, 3 Department of Chemical and Biological Engineering and Department of Chemistry, Northwestern University, 2145 Sheridan Road, Evanston, Illinois 60208, USA, 4 Department of Chemical Engineering and Department of Materials Science and Engineering, University of Michigan, Ann Arbor, Michigan 48109-2136, USA. † These authors contributed equally to this work. *e-mail: [email protected]; [email protected]. NATURE MATERIALS | ADVANCE ONLINE PUBLICATION | www.nature.com/naturematerials
© 2012 Macmillan Publishers Limited. All rights reserved.
1
NATURE MATERIALS DOI: 10.1038/NMAT3406
ARTICLES a
b
HT
=
S
EG
=
S
)nOH
(O
d
c A
m m
5 mm
Striped nanoparticle 1,14-tetradecanedithiol
Au electrode
Au electrode 30 nm
50 µm
5
A
1 mm
Glass slide
Glass slide
e
400 4 200
j (A m¬2)
j (A m¬2)
2 0 ¬2
¬20
¬200
Before Zn2+ After Zn2+
¬4 ¬10
0 E (kV m¬1)
10
0
20
¬400 ¬20
Before CH3Hg+ After CH3Hg+ ¬10
0
10
20
E (kV m¬1)
Figure 1 | Experimental set-up and typical j–E plots. a,b, An idealized scheme of a striped nanoparticle (a) and the corresponding STM image (here, for HT/EG2 NPs) obtained by a Veeco multimode scanning probe microscope on a vibration damping table (in an acoustic chamber, at room temperature, and in air; b). Scale bars, 10 nm (main image) and 5 nm (inset). c,d, The scheme and the dimensions of the device (c) and the side view of the NP film (d). e, Ohmic current density versus applied field, j–E, and dependencies for a film of striped HT/EG3 Au NPs before exposure to cations (black circles) and after exposure to cations (coloured circles). In the graph on the left, the conductance of the films is virtually unchanged after immersion in a 1 mM solution of Zn2+ (green circles). In the graph on the right, immersion of the same film in a 1 mM solution of CH3 Hg+ (red circles) results in a marked change in the conductance (note the difference in the j ranges in the two plots).
to solvent, the film was reinforced by crosslinking the NPs with 1,14-tetradecanedithiol (dithiol concentration: 5 mM in toluene, crosslinking time: 40 min; for details see refs 34–36). Subsequently, the films were washed with an excess of toluene, dried under a stream of Ar, and then placed in vacuum conditions for two days. After drying the solvent, the Au electrodes were connected to a high-precision Keithley 6517 electrometer housed in a home-made Faraday cage (to minimize the effects of static electricity). The films were subsequently immersed in 2 ml of solutions of K+ , Na+ , Zn2+ , Cd2+ , Cs+ , CH3 Hg+ , Co2+ , Hg2+ , Pb2+ , Cu2+ , Ni2+ , Tl+ , Ca2+ , Ag+ and CH3 Zn+ chloride salts in ultrapure water (resistivity: 18.2 M cm) with concentrations ranging from 1 mM down to 1 zM (zeptomolar, 10−21 M; note, the nature of the anions of the salts used, for example, chlorides versus bromides, had no effect on the results described below). The films were kept immersed in these solutions for t ∼ 120 s. This time was sufficient for the ions to penetrate the full depth of the film, H —using a conservative estimate of the diffusion coefficient of ions/small molecules in a nanoporous material, D √ ∼ 10−14 m2 s−1 (ref. 37), the diffusion depth can be estimated as Dt ∼ 1 µm, which is greater than H . Consequently, the conductivities of the films discussed below did not change for longer immersion times. After immersion, the films were washed with copious amounts of water to remove excess salt and were then thoroughly dried under a stream of nitrogen. All 2
subsequent measurements were performed at room temperature (T ∼ 300±2 K), under dry air in a hermetic Faraday cage, and in the presence of desiccant (phosphorous pentoxide, P2 O5 ; the presence of the desiccant, however, had no noticeable effect on the results). The current density–applied field (j–E) dependencies for all types of films/cations were recorded for fields up to 20 kV m−1 (higher fields cause irreversible changes in the NP films and coalescence of the neighbouring nanoparticles30 ). In this range, the j–E dependencies were ohmic (Fig. 1e) and the conductivities of the films, σ = j/E, were constant for a given film type or condition (see Methods and Supplementary Section S2 for typical j–V (t ) curves at various concentrations of CH3 Hg+ ). Before the immersion into the salt solutions, the conductivities of the striped EGn particle films were σbefore ∼2 × 10−4 −1 m−1 , which is close to the values previously recorded for films of Au NPs covered with thiolate SAMs of comparable thickness30 . The conductivities increased (to σafter ) when the films were exposed to the cations (Fig. 1e). The degree of this increase depended crucially on the nature of the SAM coating the NPs, and on the nature and the concentration of the cations (Fig. 2). In all cases, the sensitivity of the sensors was defined as the concentration of cations in solution at which the ratio of the conductivities of the NP film, χ = σafter /σbefore , was greater than unity at the 99% confidence level (for statistical analysis, see Supplementary Section S3).
NATURE MATERIALS | ADVANCE ONLINE PUBLICATION | www.nature.com/naturematerials © 2012 Macmillan Publishers Limited. All rights reserved.
NATURE MATERIALS DOI: 10.1038/NMAT3406
ARTICLES
NP types Au NPs covered with HT/EG3 SAMs
b CH3Hg+ Cs+ CH3Hg+ control Cs+
10
100
10
Hg2+
Cs+ CH3Hg+
Na+
CH3Zn+
Ag+
K+
Cd2+ Cd2+
Hg2+
Ni2+
Ag+
Na+
Zn2+
(0.1 mM) CH3Hg+ (0.1 mM, random HT/EG3) Cs+ (Random HT/EG3)
Pb2+ Ag+ Ca2+
(0.1 mM) K+
CH3Zn+
Ni2+
Co2+
Cu2+ Pb2+
CH3Zn+ CH3Hg+
Tl+
Cu2+
Pb2+ Cs+
Ni2+
Tl+ Ca2+ Hg2+
Cd2+ Zn2+
10
Cu2+
d
Cd2+ K+ Cd2+ control K+ control
1 mM
χ = σ /σ after before
10
10¬12
10¬8
10¬4
Concentration (M)
10
f
10 1 mM
χ = σ /σ after before
χ = σ /σ after before
Zn2+ Na+ Zn2+ control Na+ control
(Random HT/EG2)
10¬16
(0.1 mM)
10¬20
CH3Hg+
10¬24
(Random HT/EG2) K+
1
1
1
1 10¬16
10¬12
10¬8
Concentration (M)
10¬4
Na+
10¬20
(Random HT/EG1)
10¬24
Zn2+ (Random HT/EG1)
χ = σ /σ after before
c
Cd2+
10¬4
K+ Na+
10¬8
Co2+
10¬12
Concentration (M)
Co2+
10¬16
Zn2+
10¬20
Cs+
10¬24
Ca2+
1
1
e
Au NPs covered with HT/EG1 SAMs
1 mM
Tl+
χ = σ /σ after before
100
χ = σ /σ after before
a
Au NPs covered with HT/EG2 SAMs
Figure 2 | Sensitivity (left column) and selectivity (right column) of cation sensing by different types of Au NPs decorated by HT/EGn SAMs. These measures are quantified by the change in the conductivities of the films on cation exposure, χ = σafter /σbefore . Filled symbols correspond to nanoparticles covered with striped HT/EGn SAMs; open symbols correspond to control experiments in which the nanoparticles were covered with disordered/non-striped SAMs of the same composition. Standard deviations are for at least 12 independent experiments for each condition. The arrows in the left column point to the detection limit/sensitivity (defined in the main text; see also Supplementary Section S3). a,b, Au NPs covered with HT/EG3 SAMs detect CH3 Hg+ selectively with a detection limit of ∼1 aM. c,d, HT/EG2 Au NPs detect Cd2+ and K+ with a detection limit of ∼1 pM for both cations. e,f, HT/EG1 Au NPs detect Zn2+ and also Na+ (detection limits: ∼1 nM and ∼1 µM, respectively). Note that the plots in a, c and e are double logarithmic and the conductivity increases nonlinearly with cation concentration.
Specifically, the HT/EG3 striped Au NPs detected methyl mercury, CH3 Hg+ , with an approximately attomolar (10−18 M) sensitivity limit (corresponding to only ∼600 CH3 Hg+ cations in 1 ml of the surrounding solution; Fig. 2a). As the concentration of methyl mercury increased, χ increased nonlinearly with [CH3 Hg+ ] and reached a value of ∼75 at [CH3 Hg+ ] = 0.1 mM. Moreover, these sensing characteristics were selective for methyl mercury (Fig. 2b). For Co2+ , Pb2+ , Cu2+ , Ni2+ , Tl+ , Ca2+ , Ag+ , CH3 Zn+ , Cd2+ , K+ , Na+ , Cs+ and Zn2+ , the conductivity did not change
perceptibly (χ ≈ 1) over the entire range of concentrations tested; for Hg2+ and Cs+ it changed only slightly (χ < 2.9 and χ < 2.21, respectively). Furthermore, for Cs+ , the detection limit was only ∼1 µM—that is, 12 orders of magnitude above that of CH3 Hg+ . Films of Au NPs coated with striped SAMs of HT/EG2 detected Cd2+ with χ = 24.7 at 1 mM and a detection limit ∼1 pM (Fig. 2c,d). These NPs, however, also detected K+ with χ = 10.2 at 1 mM and with ∼1 pM detection limit and, to a lesser degree, Ag+
NATURE MATERIALS | ADVANCE ONLINE PUBLICATION | www.nature.com/naturematerials
© 2012 Macmillan Publishers Limited. All rights reserved.
3
NATURE MATERIALS DOI: 10.1038/NMAT3406
ARTICLES Table 1 | Film sensitivity and energies in cation complexation. χnon-binding
χmixture
χ Cs = 1.600 ± 0.512
χ Na = 0.888 ± 0.165
χ Cs
HT/EG2
2+ χ Cd
= 2.859 ± 0.858
χ Zn
HT/EG1
χ Zn
= 2.337 ± 0.716
χ Cd
NP types
χbinding
HT/EG3
+
2+
+
+ /Na+
= 1.671 ± 0.389
2+
= 1.170 ± 0.182
2+ /Zn2+
= 2.373 ± 1.238
2+
χ Cd
= 1.105 ± 0.125
χ Zn
2+ /Cd2+
= 2.406 ± 0.383
Sensitivity (that is, χ = σafter /σbefore ) in binary mixtures comprising cations that bind (denoted in bold) to the NPs and cations that do not bind (not bold). No significant differences between χ binding and χ mixture were observed. Standard deviations are for at least six independent experiments for each condition.
with χ = 2.86. On the other hand, they did not bind Co2+ , Pb2+ , Cu2+ , Ni2+ , Tl+ , Ca2+ , CH3 Zn+ , Na+ , Zn2+ , Cs+ and CH3 Hg+ . Finally, Au NPs coated with striped SAMs of HT/EG1 were least selective and detected Zn2+ , Ag+ and Na+ (Fig. 2e,f). In this case, the values of χ were rather moderate (for example, at 1 mM, 6.51 for Zn2+ , 5.00 for Ag+ and 2.72 for Na+ ) although the sensitivity towards Zn2+ remained relatively high (for example, detection limit ∼1 nM versus ∼1 µM for Na+ ). We note that Au NPs coated with random/non-striped HT/EGn SAMs were significantly less sensitive and in most cases less selective in detecting the highest-χ cations (but not some lower-χ cations, see open symbols in Fig. 2). Also, particles coated only with pure EGn SAMs exhibited even lower values of χ (for details, see Supplementary Section S4). We emphasize that all χ values discussed above reflect the effects of cation binding and not any water and/or drying effects. This was verified in a series of experiments in which the films were washed in cation-free, deionized water and then dried—in this case, no statistically significant changes in the conductivities of the films were observed after either one or multiple wash/dry cycles (for example, for HT/EG2 Au NP films, χ = σafter /σbefore = 1.011±0.029 after one cycle and 1.013 ± 0.031 after ten cycles). Also, when the cations were first captured into the films (such that conductivities increased) but were then washed off (by placing the films in 80 ◦ C deionized water for 2–3 h), the conductivities returned to their native values before cation capture (see Supplementary Section S7). We observe that the films owe their robustness during all of these experiments (see Supplementary Section S6) to the crosslinking of the NPs by dithiols; films in which the NPs were not crosslinked simply dissolved when washed with or immersed in water. The changes in the conductivity of the film result from the capture of the cations by the ligands on the NPs. This was directly verified by means of the ICP and X-ray photoelectron spectroscopy analyses of NPs before and after ion capture (see Supplementary Section S5 for these and other techniques used). In addition, Fourier transform infrared spectroscopy studies were performed that evidenced spectral changes on cation binding. Of particular importance here were the changes in the characteristic vibrational modes of the ethylene glycol units at 1001–1003 cm−1 and at 827 cm−1 . Remarkably, these changes were reversible in the sense that the original spectra were retrieved when the cations were washed off the NPs during immersion in 80 ◦ C water (see above). These results also indicate that the capture of the cations is mediated by coordination-type interactions with the EG units, as should indeed be expected on the basis of common chemical knowledge. It is also worth noting that cations for which χ = σafter /σbefore is close to unity do not cause significant changes in the NP/NP–film spectra as discussed in the Supplementary Section S5. Interestingly, a qualitatively similar result is obtained from solution experiments where the conductivities of NP suspensions are monitored during titration with different salt solutions. For those salts that are characterized by low χ in our solid-state (that is, NP–film) measurements, the conductance of NP solutions increases linearly 4
with the amount of salt added; for those, however, that exhibit higher χ in NP films, the solution conductance initially remains constant, indicating that the added cations are trapped on the NPs (and thus do not contribute to the conductance of the solution; see Supplementary Section S5.5). An essential feature of any chemical sensor is its selectivity, which in our case needs to be determined on exposure of NP films to cation mixtures. We first consider binary mixtures comprising cations that the NP films bind and sense with statistical significance (that is, giving conductance ratios χ = σafter /σbefore > 1 at 99% confidence levels) and cations that do not bind strongly and do not change conductivities at such confidence levels. Such pairs correspond to most of the possible combinations among the cations we tested (see Fig. 2). The results for some representative pairs are summarized in Table 1. Specifically, for the HT/EG3 Au NP films, the presence of weakly binding Na+ does not significantly change the signal coming from stronger binding Cs+ . Similarly, for HT/EG2 Au NP films, the presence of Zn2+ does not strongly affect the signal due to Cd2+ , and for HT/EG1 Au NP films the conductance on exposure to the mixture of Zn2+ and Cd2+ is close to the value characterizing the former cation. Such data confirm the selectivity of the films towards cations giving larger values of χ . Moreover, the signals from the NP–film sensors are not affected by the changes in the ionic strength of the solution due to the non-binding, low-χ cations (see Supplementary Fig. S9a). Of course, the selectivity decreases when the films are exposed to mixtures of cations characterized by similar χ > 1 values (Supplementary Section S12). This problem is the least pronounced for HT/EG3 Au NP films (see Fig. 2 and Supplementary Section S8), which are therefore suitable for deployment as environmental CH3 Hg+ sensors. This is illustrated in Fig. 3, where we used our sensors to measure the CH3 Hg+ content in environmental samples directly relevant to the issues of public health—tap water, lake water and fish (see Methods for detailed preparation and analysis of environmental samples). Importantly, the results we obtained for the lake water could be directly compared to the values reported in the literature for the same lake (Lake Michigan) sample, and the CH3 Hg+ content in fish could be compared to the value determined by the United States Geological Survey (USGS). For the lake water, our measurements gave a 5.9 fM concentration of CH3 Hg+ (range within one s.d. 0.35–43.3 fM) versus 5–210 fM reported in ref. 38. For the fish, we measured a concentration of 3.81 pM (range within one s.d.: 1.62–8.19 pM), in excellent agreement with the 3.58 pM value provided to us by the USGS after our experiments were completed. A reassuring result—at least for the citizens of Evanston—is that the content of CH3 Hg+ in the tap water of this city is very low (27.7 aM, range within one s.d.: 4.5–138 aM). We make two further comments about all of these results. First, when the environmental samples were spiked with known concentrations of CH3 Hg+ , the readings of our films fell onto the calibration curve obtained using the concentration standards (that is, solutions of CH3 Hg+ in pure, deionized water)—this means that there is no background interference from components other than CH3 Hg+ in the environmental samples. Second, we emphasize that our
NATURE MATERIALS | ADVANCE ONLINE PUBLICATION | www.nature.com/naturematerials © 2012 Macmillan Publishers Limited. All rights reserved.
NATURE MATERIALS DOI: 10.1038/NMAT3406
Table 2 | Film sensitivity and energies in cation complexation.
CH3 Hg+ Lake Michigan ( from ref. 38) Lake Michigan spiked Tap water Fish ( from USGS) Fish blank spiked
100
χ (¬)
ARTICLES
10
NP/cation types Free Zn2+ Cd2+
(This work)
CH3 Hg+
(This work)
1 10¬23
} (reference samples) 10¬20
10¬17 10¬14 10¬11 10¬8 Concentration CH3 Hg+ (M)
10¬5
results were achieved in a technically straightforward manner (by simply dipping the film in the sample solution), in contrast to other methods that require tedious sample pretreatment, including conversion of methylmercury to ethylmethylmercury and concentration on a dedicated column before analysis by gas chromatography combined with ICP mass spectrometry2,39 . We now turn our attention to the fundamental aspects of the relationship between the binding of cations into the film and the observed current changes. There are two interrelated components that need to be considered: cation-facilitated electron transport through the molecular bridge between proximal NPs, and formation of the conductive paths through the film, which ultimately give rise to the high sensitivity of the cation detection. To understand the electronic structure of the bridge, we performed quantum-mechanical calculations for a fragment of NP surface comprising two nearby EGn stripes and an HT stripe between them (Fig. 4a). Calculations were performed for both the cationfree HT/EGn systems, as well as a system in which different metal cations were placed either within the EGn stripes (denoted intra in Table 2) or at the edges of these stripes (inter). In all cases, the geometry of the system was optimized by a semiempirical quantum-mechanical method NDDO-PM6 (ref. 40; with the MOPAC2007 (ref. 41) software) at the Non-equilibrium Energy Research Center supercomputer centre. The calculations indicate that the experimentally observed selectivities do not correlate with the calculated energies of the complexation of the cations by various HT/EGn Au NPs or with the heats of formation of these complexes, nor with the energies of the highest occupied molecular orbital (HOMO) or the lowest unoccupied molecular orbital (LUMO). On the other hand, the observed cation selectivities can be related to the HOMO/LUMO gap, 1E (Table 2). This result suggests that both HOMO and LUMO orbitals are involved in electron transport. Although this is not a typical scenario in describing
HT/EG2
HT/EG3
9.160 1.834 1.398 4.395 3.928 6.045 6.154
9.143 2.408 1.976 3.969 3.676 6.392 6.129
9.142 3.098 3.031 4.088 4.243 6.272 6.045
Calculated energies of the HOMO/LUMO gaps (in eV) for different cations and EGn thiols. ‘Interband’ corresponds to the situation where the cations are bound at the edge of an EGn stripe; ‘intraband’ corresponds to the cation being placed within one EGn band. Entries in bold correspond to minimal gap energies for a given type of EGn and generally agree with/reflect the cation-binding selectivity observed in experiments.
10¬2
Figure 3 | Selectivity of the films on exposure to cation mixtures and environmental samples. The χ values measured for the environmental samples (blue star: water from Lake Michigan, purple diamond: Evanston tap water, green triangle: mosquito fish collected by USGS from Everglades National Park). Red symbols correspond to the measurements of standard solutions (that is, of known CH3 Hg+ concentration in deionized water) and the dashed calibration curve is the second order polynomial fit. Open symbols correspond to the environmental samples ‘spiked’ with known amounts of CH3 Hg+ . Pale green and pale blue regions correspond to the one-standard-deviation ranges in our measurements (three different films, at least six measurements per film) of, respectively, fish and lake water samples. Dark green horizontal bar corresponds to the range of CH3 Hg+ content in the mosquito fish determined by USGS, and the dark blue horizontal bar corresponds to the one-standard-deviation range of CH3 Hg+ concentration in Lake Michigan determined in ref. 38. For further experimental detail, see Methods.
Intra Inter Intra Inter Intra Inter
HT/EG1
conduction through a molecular bridge42 , similar models have been used previously to describe transport through disordered media (as our films are)43 . In this context, we observe that in our system the HOMO orbital (and other high-energy occupied orbitals) is always located close to the NP surface (Fig. 4b), whereas the LUMO orbital (and unoccupied orbitals close to it) is mostly centred on the EGn /cation fragments (Fig. 4c). Therefore, for the electrons to travel through the SAM, they must be injected from the HOMO into the LUMO. Importantly, the energy required for this injection is substantially diminished on cation binding, and the smallest values of 1E generally reflect the detection specificity observed in experiments: for Zn2+ , the smallest 1E corresponds to the binding to the HT/EG1 Au NPs; Cd2+ has the smallest 1E when binding to HT/EG2 Au NPs; for CH3 Hg + , 1E is the least when binding inter-stripe to HT/EG3 Au NPs (although, in this case, calculations give the same value for HT/EG1 particles; Table 2). We note that for any given HT/EGn /cation system, the 1E gaps (and the absolute binding energies) are typically lower when cations coordinate with ethylene glycols at the edges of the EGn stripes than when these cations are fully buried within the EGn stripes. This observation suggests that the free space provided by the HT stripes is important for the operation of our sensors (as also corroborated by molecular dynamics simulations illustrated in Fig. 4d and Supplementary Section S9). The decrease in 1E on cation binding increases the electronic conductance of a NP-SAM//SAM-NP molecular bridge, σc ∝ 1/2 (e2 /h)(e−2(m1E) L/h )2 , where h is Planck’s constant, m is the rest mass of an electron and L is the thickness of the NP ligand shell44 . Importantly, as multiple cations bind into the NP film, the local changes over individual bridges can translate into a global change in the conductance of the film. This process can be described by combining σc with the cation-binding kinetics and Bruggeman’s effective medium theory45,46 (EMT) accounting for the formation of conductive paths percolating the NP film (Fig. 5a,b). The binding of cations to the EGn ligands as a function of concentration [C]0 can be described in terms of the Sips isotherm47 : θ = (Keq [C]0 )a /[1 + (Keq [C]0 )a ], where θ is the fraction of occupied sites, Keq is the equilibrium association constant and a is the heterogeneity index. The Sips isotherm is a composite of the familiar Langmuir isotherm (θLM = KLM [C]0 /[1 + KLM [C]0 ]) and the Freundlich isotherm (θF = KF [C]1/n 0 ), commonly used for low-concentration binding in heterogeneous systems. The Sips isotherm approaches Langmuir’s formula for a close to unity and Freundlich’s equation at low concentration, and has been used to describe ligand binding to powders48 and aggregated nanoparticle media49 . The heterogeneity index (a) determines the width of the Gaussian-like distribution of Keq , which arises from steric effects, conformational changes and differences in the chemical microenvironment50 . Using the
NATURE MATERIALS | ADVANCE ONLINE PUBLICATION | www.nature.com/naturematerials
© 2012 Macmillan Publishers Limited. All rights reserved.
5
NATURE MATERIALS DOI: 10.1038/NMAT3406
ARTICLES a
b
c HT EGn
EGn
d
Figure 4 | Rationalizing the conductance through the molecular bridge between √ the√striped NPs. a, Top view of the modelled system comprising 15 HT and 4 EGn thiols arranged in alternating stripes on the Au(111) surface with a ( 3 × 3) R30◦ lattice distribution of the HT/ EGn ligands. b,c, HOMO (b) and LUMO (c) orbitals calculated for the HT/EG3 Au NPs capturing CH3 Hg+ . Note that the HOMO is closer to the surface of the nanoparticle (localized on the lone pairs of the sulphur atoms), whereas the LUMO is centred on the EGn /cation complexes (corresponding to the anti-bonding σ ∗ (O–M) of the EGn /cation complexes). These simulations are performed for nanoparticles in vacuum, which is appropriate for the experimental measurements being performed in dry films. d, A snapshot from the molecular dynamics simulations showing a typical ion-trapping scenario where the cation (green ball) is trapped near the edge of an EGn stripe (for further details, see Supplementary Section S9).
expressions for σc and θ, the EMT formula describing the change in the conductance of the film on cation binding becomes (see Supplementary Section S11): ! a 1 z Keq [C]0 a − 1 σc,n 2 1 + Keq [C]0 a ! ! Keq [C]0 1 a − 1 + z 1− 2 1 + Keq [C]0 " ! a 1 z Keq [C]0 a − 1 σc,n + 2 1 + Keq [C]0
1 χS-EMT = (z − 2)
+
(
1/2 a ! !!2 Keq [C]0 1 a − 1 z 1− + 2(z − 2)σc,n (1) 2 1 + Keq [C]0
where σc is normalized by dividing by the conductance of the film before binding cations (σc,n = σc /σbefore ), and z is the number of high-conductance bonds any NP can make to its neighbours. Figure 5c illustrates that with the experimentally measured (by isothermal titration calorimetry, see Supplementary Section S10) values of Keq , this model reproduces the essential features of the nonlinear cation concentration versus conductivity dependencies recorded in experiments. As expected, the fitted values of a and z for 6
both Cd2+ and K+ are nearly equal (for Cd2+ : a = 0.42 ± 0.32 and z = 2.1 ± 0.5; for K+ : a = 0.42 ± 0.55 and z = 1.9 ± 1.0), because they reflect the intrinsic nanostructure of the Au NP films. The values of z ∼ 2 suggest that there are relatively few possibilities to form high-conductance connections between neighbouring NPs in the film (z = 6 would signify that such connections can be made between all nearest-neighbour NPs in a randomly packed film)46 . More importantly, the values of a < 1 confirm that there is significant heterogeneity of the cation binding to the NP film (unlike in isothermal titration calorimetry solution studies for which a = 1). This result agrees with previous works on the nanoenvironment47 , and it endows our sensor with a remarkably wide range of sensitivity. If the cation binding properties in the films were the same as those observed in solution (a = 1), the sensitivity curves would look like the dashed curve in Fig. 5c. As seen, this curve transitions sharply between constant χmin and χmax values over a relatively narrow range of concentrations (∼4 orders of magnitude). In contrast, with the heterogeneity effects being operative, the ranges over which χ values change are much wider, ∼9 orders of magnitude for Cd2+ and K+ and even more (∼18 orders) for CH3 Hg+ . Although the broadening of binding energies is generally considered a disadvantage that complicates affinity studies50 , it is in fact the key—and unique— principle underlying the remarkable range of sensing of our NP devices. In other words, the packing of chemically identical NPs into a nanostructured thin film gives rise to adsorption sites with not a single but many different energies (and thus Keq
NATURE MATERIALS | ADVANCE ONLINE PUBLICATION | www.nature.com/naturematerials © 2012 Macmillan Publishers Limited. All rights reserved.
NATURE MATERIALS DOI: 10.1038/NMAT3406 a
ARTICLES c 100
Cd2+ (
fit S-EMT)
(
fit S-EMT)
K2+
χ (¬)
Langmuir
10
b
1 10–17
10–13
10–9 10–5 Concentration (M)
10–1
Figure 5 | Percolation of Au NP films on cation binding. a, Schematic representation of a film of striped Au NPs, between two Au electrodes. b, On binding of cations (blue spheres) conductive paths (blue line) connecting the two electrodes percolate the film and connect the two electrodes. c, Sensitivity (that is, χ = σafter /σbefore ) of HT/EG2 Au NP films on binding of K+ and Cd2+ from solutions of different concentrations. Symbols correspond to the experimental data (the error bars are based on 12 independent measurements for each condition). The solid lines are theoretical fits based on the model described in the main text (equation (1)). The values of Keq parameters were determined calorimetrically (see Supplementary Section S10) and were Cd2+ = 2.75 ± 0.31 × 106 M−1 for Cd2+ and K K+ = 1.5 ± 0.12 × 106 M−1 for K+ . As a reference, the Langmuir isotherm (a = 1) for Cd2+ is also included Keq eq (dashed black line).
values). The sites with high Keq values are the first ones to be occupied at very low concentration, whereas the low Keq sites saturate at much higher concentrations. This effect grants the NP films with both a low onset and, at the sametime, a very broad range of detection. Overall, the high sensitivity of our solid-state sensors derives from a subtle interplay between cation/striped-NP macromolecular recognition, electron transport over the NP/cation/NP bridges, and the formation of conductive paths through heterogeneous NP films. Together, these phenomena provide a new principle for solid-state sensing and enable inexpensive, sensitive and portable environmental sensors. With further tailoring—both in terms of ligand chemistries and stripe dimensions—it should be possible to extend the applicability of our nanostructured materials to other sensing systems whose properties derive from macromolecular recognition on NP surfaces.
Methods Synthesis of striped Au NPs. Striped Au NPs covered by 1-hexanethiol (Sigma-Aldrich 739286) and 1-hexanethiol terminated in n = 1–3 units of ethylene glycol (EGn , ProChimia) (n = 1, ProChimia TH 001-m6.n1-1; n = 2, ProChimia TH 001-m6.n2-1; n = 3, ProChimia TH 001-m6.n3-1, respectively) in 1:3 ratio were synthesized by a modified Stucky method51 . Briefly, 0.45 mmol of AuPPh3 Cl was mixed with 0.45 mmol of thiol mixture in 40 ml of chloroform/dimethylformamide co-solvent, and 5 mmol of tert-butylamine-borane complex was added. The reactant was heated to 55 ◦ C and stirred for 1 h. After cooling to room temperature, 200 ml of chloroform was added to precipitate Au NPs, and the product was centrifuged and washed with chloroform. The ligand shell structure was determined using STM as described previously (see Fig. 1b in the main text)32,51–54 . Further characterization is included in Supplementary Section S5. Film conductivities. The film conductivities of the HT/EGn Au NP films before capturing cations, σbefore , were 1.9 × 10−4 ± 1.6 × 10−4 −1 m−1 , 1.4 × 10−4 ± 8.7 × 10−5 −1 m−1 and 2.7 × 10−4 ± 1.3 × 10−4 −1 m−1 for n = 1, n = 2 and n = 3, respectively. These values were calculated as σbefore = I /wt × 1/V × l = j/V × l, where I and j are current (A) and current density (A m−2 ), respectively, V is applied voltage (V), and w, t and l are the width, thickness and gap of the Au electrode, respectively (w = 5 mm, t = 30 nm and l = 50 µm). The conductivity values and their s.d. are based on at least 12 different material samples for each type of a film. We note that these conductivity values are comparable to those measured for films of Au NPs covered with thiols of similar thickness30,55 . Preparation and analysis of environmental samples. To prevent any external contamination of the samples, all containers (glass vials, VWR) were cleaned
in 4 M aqueous HCl solution at 65 ◦ C (for 12 h) and rinsed three times with copious amounts of deionized water (resistivity: 18.2 M cm). All experiments were performed in a LabConco biosafety class II laminar flow cabinet with a high-efficiency particulate arresting (HEPA) 99.99% filter in a HEPA-filtered room, similar to class 100 cleanroom conditions. Water from Lake Michigan was sampled at Northwestern University (Evanston Campus) private beach into a clean polypropylene tube (VWR). Tap water was sampled from a water fountain in the Technological Institute building at Northwestern University. Fish samples were obtained from the USGS, Mercury Research Laboratory (http://wi.water.usgs.gov/mercury-lab/index.html). To prepare fish samples for analysis, around 10 mg of lyophilized mosquito fish from Everglades National Park collected by the USGS was exactly weighed and digested using 10 ml of 5 M HNO3 for 8 h at 65 ◦ C and then neutralized using 5 M aqueous KOH solution (note: the pH was checked by a pH meter to confirm that the solution was neutralized—our Au NP film is unstable in very acidic solutions). The solution was then diluted 1,000 times to bring the background salt (KNO3 ) concentration down to the millimolar range. For spiking the blank fish and lake water samples, 10% (v/v) of 10−12 M,10−8 M and 10−4 M of CH3 Hg+ solution were added.
Received 15 September 2011; accepted 20 July 2012; published online 9 September 2012
References 1. Boening, D. W. Ecological effects, transport, and fate of mercury: A general review. Chemosphere 40, 1335–1351 (2000). 2. Clarkson, T. W. The toxicology of mercury. Crit. Rev. Clin. Lab. Sci. 34, 369–403 (1997). 3. Matsumoto, H., Koya, G. & Takeuchi, T. Fetal minamata disease: A neuropathological study of two cases of intrauterine intoxication by a methyl mercury compound. J. Neuropathol. Exp. Neurol. 24, 563–574 (1965). 4. Harada, M. Minamata disease—methylmercury poisoning in Japan caused by environmental pollution. Crit. Rev. Toxicol. 25, 1–24 (1995). 5. Inaba, T. et al. Estimation of cumulative cadmium intake causing Itai-itai disease. Toxicol. Lett. 159, 192–201 (2005). 6. Nomiyama, K. Recent progress and perspectives in cadmium health-effects studies. Sci. Total Environ. 14, 199–232 (1980). 7. Nriagu, J. O. & Pacyna, J. M. Quantitative assessment of worldwide contamination of air, water and soils by trace-metals. Nature 333, 134–139 (1988). 8. Lin, S. H. & Juang, R. S. Heavy metal removal from water by sorption using surfactant-modified montmorillonite. J. Hazard. Mater. 92, 315–326 (2002). 9. Jensen, S. & Jernelov, A. Biological methylation of mercury in aquatic organisms. Nature 223, 753–754 (1969). 10. Narin, I., Soylak, M., Elci, L. & Dogan, M. Determination of trace metal ions by AAS in natural water samples after preconcentration of pyrocatechol violet complexes on an activated carbon column. Talanta 52, 1041–1046 (2000). 11. Ghaedi, M., Ahmadi, F. & Shokrollahi, A. Simultaneous preconcentration and determination of copper, nickel, cobalt and lead ions content by flame atomic absorption spectrometry. J. Hazard. Mater. 142, 272–278 (2007).
NATURE MATERIALS | ADVANCE ONLINE PUBLICATION | www.nature.com/naturematerials
© 2012 Macmillan Publishers Limited. All rights reserved.
7
NATURE MATERIALS DOI: 10.1038/NMAT3406
ARTICLES 12. Caroli, S., Forte, G., Iamiceli, A. L. & Galoppi, B. Determination of essential and potentially toxic trace elements in honey by inductively coupled plasma-based techniques. Talanta 50, 327–336 (1999). 13. Demuth, N. & Heumann, K. G. Validation of methylmercury determinations in aquatic systems by alkyl derivatization methods for GC analysis using ICP-IDMS. Anal. Chem. 73, 4020–4027 (2001). 14. Lee, J. S., Han, M. S. & Mirkin, C. A. Colorimetric detection of mercuric ion (Hg2+ ) in aqueous media using DNA-functionalized gold nanoparticles. Angew. Chem. Int. Ed. 46, 4093–4096 (2007). 15. Huang, C. C. & Chang, H. T. Parameters for selective colorimetric sensing of mercury(II) in aqueous solutions using mercaptopropionic acid-modified gold nanoparticles. Chem. Commun. 1215–1217 (2007). 16. Chiang, C. K., Huang, C. C., Liu, C. W. & Chang, H. T. Oligonucleotide-based fluorescence probe for sensitive and selective detection of mercury (II) in aqueous solution. Anal. Chem. 80, 3716–3721 (2008). 17. Chang, H. Y., Hsiung, T. M., Huang, Y. F. & Huang, C. C. Using rhodamine 6G-modified gold nanoparticles to detect organic mercury species in highly saline solutions. Environ. Sci. Technol. 45, 1534–1539 (2011). 18. Kang, T., Hong, S. R., Moon, J., Oh, S. & Yi, J. Fabrication of reusable sensor for detection of Cu2+ in an aqueous solution using a self-assembled monolayer with surface plasmon resonance spectroscopy. Chem. Commun. 3721–3723 (2005). 19. Li, D., Wieckowska, A. & Willner, I. Optical analysis of Hg2+ ions by oligonucleotide-gold-nanoparticle hybrids and DNA-based machines. Angew. Chem. Int. Ed. 47, 3927–3931 (2008). 20. Wang, G. Q., Wang, Y. Q., Chen, L. X. & Choo, J. Nanomaterial-assisted aptamers for optical sensing. Biosens. Bioelectron. 25, 1859–1868 (2010). 21. Yin, J. et al. SERS-active nanoparticles for sensitive and selective detection of cadmium ion (Cd2+ ). Chem. Mater. 23, 4756–4764 (2011). 22. Lin, Z. Z., Li, X. H. & Kraatz, H. B. Impedimetric immobilized DNA-based sensor for simultaneous detection of Pb2+ , Ag+ , and Hg2+ . Anal. Chem. 83, 6896–6901 (2011). 23. Zhang, Z., Yu, K., Bai, D. & Zhu, Z. Q. Synthesis and electrochemical sensing toward heavy metals of bunch-like bismuth nanostructures. Nanoscale Res. Lett. 5, 398–402 (2010). 24. Thompson, R. B., Maliwal, B. P., Feliccia, V. L., Fierke, C. A. & McCall, K. Determination of picomolar concentrations of metal lens using fluorescence anisotropy: Biosensing with a ‘reagentless’ enzyme transducer. Anal. Chem. 70, 4717–4723 (1998). 25. Forzani, E. S., Zhang, H. Q., Chen, W. & Tao, N. J. Detection of heavy metal ions in drinking water using a high-resolution differential surface plasmon resonance sensor. Environ. Sci. Technol. 39, 1257–1262 (2005). 26. Kumar, M. & Zhang, P. Highly sensitive and selective label-free optical detection of mercuric ions using photon upconverting nanoparticles. Biosens. Bioelectron. 25, 2431–2435 (2010). 27. Song, H. D. et al. Picomolar selective detection of mercuric ion (Hg2+ ) using a functionalized single plasmonic gold nanoparticle. Nanotechnology 21, 145501 (2010). 28. Hinkle, P. M., Kinsella, P. A. & Osterhoudt, K. C. Cadmium uptake and toxicity via voltage-sensitive calcium channels. J. Biol. Chem. 262, 16333–16337 (1987). 29. Rashed, M. N. Cadmium and lead levels in fish (Tilapia nilotica) tissues as biological indicator for lake water pollution. Environ. Monit. Assess. 68, 75–89 (2001). 30. Nakanishi, H. et al. Photoconductance and inverse photoconductance in films of functionalized metal nanoparticles. Nature 460, 371–375 (2009). 31. DeVries, G. A. et al. Divalent metal nanoparticles. Science 315, 358–361 (2007). 32. Jackson, A. M., Myerson, J. W. & Stellacci, F. Spontaneous assembly of subnanometre-ordered domains in the ligand shell of monolayer-protected nanoparticles. Nature Mater. 3, 330–336 (2004). 33. Witt, D., Klajn, R., Barski, P. & Grzybowski, B. A. Applications properties and synthesis of omega-functionalized n-alkanethiols and disulfides—the building blocks of self-assembled monolayers. Curr. Org. Chem. 8, 1763–1797 (2004). 34. Kowalczyk, B., Apodaca, M. M., Nakanishi, H., Smoukov, S. K. & Grzybowski, B. A. Lift-Off and micropatterning of mono- and multilayer nanoparticle films. Small 5, 1970–1973 (2009). 35. Klajn, R., Bishop, K. J. M. & Grzybowski, B. A. Light-controlled self-assembly of reversible and irreversible nanoparticle suprastructures. Proc. Natl Acad. Sci. USA 104, 10305–10309 (2007). 36. Kowalczyk, B., Lagzi, I. & Grzybowski, B. A. Nanoarmoured droplets of different shapes formed by interfacial self-assembly and crosslinking of metal nanoparticles. Nanoscale 2, 2366–2369 (2010). 37. Han, S. B. et al. Chromatography in a single metal-organic framework (MOF) crystal. J. Am. Chem. Soc. 132, 16358–16361 (2010).
8
38. Sullivan, K. A. & Mason, R. P. The concentration and distribution of mercury in Lake Michigan. Sci. Total Environ. 213, 213–228 (1998). 39. Jackson, B., Taylor, V., Baker, R. A. & Miller, E. Low-level mercury speciation in fresh waters by isotope dilution GC-ICP-MS. Environ. Sci. Technol. 43, 2463–2469 (2009). 40. Pople, J. & Beveridge, D. Approximate Molecular Orbital Theory (McGraw-Hill, 1970). 41. Stewart, J. J. P. Stewart Computational Chemistry (Colorado Springs, 2007). 42. Nitzan, A. Chemical Dynamics in Condensed Phases : Relaxation, Transfer and Reactions in Condensed Molecular Systems Ch. 16 (Oxford, 2006). 43. Burlakov, V. M. et al. Discrete hopping model of exciton transport in disordered media. Phys. Rev. B 72, 075206 (2005). 44. Holmlin, R. E. et al. Electron transport through thin organic films in metal-insulator-metal junctions based on self-assembled monolayers. J. Am. Chem. Soc. 123, 5075–5085 (2001). 45. Landauer, R. The electrical resistance of binary metallic mixtures. J. Appl. Phys. 23, 779–784 (1952). 46. Kirkpatr, S. Classical transport in disordered media—scaling and effective-medium theories. Phys. Rev. Lett. 27, 1722–1725 (1971). 47. Sips, R. On the structure of a catalyst surface. J. Chem. Phys. 16, 490–495 (1948). 48. Kumar, K. V. & Porkodi, K. Relation between some two- and three-parameter isotherm models for the sorption of methylene blue onto lemon peel. J. Hazard. Mater. 138, 633–635 (2006). 49. Hristovski, K., Baumgardner, A. & Westerhoff, P. Selecting metal oxide nanomaterials for arsenic removal in fixed bed columns: From nanopowders to aggregated nanoparticle media. J. Hazard. Mater. 147, 265–274 (2007). 50. Vijayendran, R. A. & Leckband, D. E. A quantitative assessment of heterogeneity for surface-immobilized proteins. Anal. Chem. 73, 471–480 (2001). 51. Verma, A. et al. Surface-structure-regulated cell-membrane penetration by monolayer-protected nanoparticles. Nature Mater. 7, 588–595 (2008). 52. Jackson, A. M., Hu, Y., Silva, P. J. & Stellacci, F. From homoligand- to mixed-ligand-monolayer-protected metal nanoparticles: A scanning tunneling microscopy investigation. J. Am. Chem. Soc. 128, 11135–11149 (2006). 53. Kuna, J. J. et al. The effect of nanometre-scale structure on interfacial energy. Nature Mater. 8, 837–842 (2009). 54. Centrone, A. et al. The role of nanostructure in the wetting behavior of mixed-monolayer-protected metal nanoparticles. Proc. Natl Acad. Sci. USA 105, 9886–9891 (2008). 55. Zabet-Khosousi, A. & Dhirani, A. A. Charge transport in nanoparticle assemblies. Chem. Rev. 108, 4072–4124 (2008).
Acknowledgements This work was supported by the Non-equilibrium Energy Research Center, which is an Energy Frontier Research Center funded by the US Department of Energy, Office of Science, Office of Basic Energy Sciences under grant number DE-SC0000989. E.S.C. and F.S. acknowledge the support of ENI within the MIT Energy initiative for their work. E.S.C. is supported by a Samsung Scholarship from the Samsung Foundation of Culture. H.J., S.C.G. and F.S. acknowledge support from the Defense Threat Reduction Agency under Grant No. HDTRA1-09-1-0012. T.M.H. is financially supported by the Human Frontier Science Program. We acknowledge L. Meda for her X-ray photoelectron spectroscopy measurements, and R. Borrelli and P. Cesti for helpful discussions. We also thank the USGS for providing fish samples and for helpful discussions.
Author contributions E.S.C. synthesized the striped nanoparticles, performed all of their characterization before and after binding to ions, and helped with some of the solid-state work.; J.K. made and characterized all sensors, performed cation capture experiments, measured the conductivities of all sensors and analysed data; B.T. developed quantum-mechanical models and ran quantum-mechanical calculations; T.M.H. developed and validated the binding/percolation models; H.J. ran and analysed MD simulations; S.C.G. analysed and supervised the MD simulations; H.N. performed initial experiments with NP films; M.Y. performed the STM characterization of the particles; A.Z.P. helped with the theory of percolation phenomena; F.S. and B.A.G. conceived the ideas and supervised the research.
Additional information Supplementary information is available in the online version of the paper. Reprints and permissions information is available online at www.nature.com/reprints. Correspondence and requests for materials should be addressed to F.S. or B.A.G.
Competing financial interests The authors declare no competing financial interests.
NATURE MATERIALS | ADVANCE ONLINE PUBLICATION | www.nature.com/naturematerials © 2012 Macmillan Publishers Limited. All rights reserved.
Ultrasensitive detection of toxic cations through changes in the tunnelling current across films of striped nanoparticles Eun Seon Cho1,2† , Jiwon Kim3† , Baudilio Tejerina3 , Thomas M. Hermans3 , Hao Jiang4 , Hideyuki Nakanishi3 , Miao Yu2 , Alexander Z. Patashinski3 , Sharon C. Glotzer4 , Francesco Stellacci1,2 * and Bartosz A. Grzybowski3 * Although multiple methods have been developed to detect metal cations, only a few offer sensitivities below 1 pM, and many require complicated procedures and sophisticated equipment. Here, we describe a class of simple solid-state sensors for the ultrasensitive detection of heavy-metal cations (notably, an unprecedented attomolar limit for the detection of CH3 Hg+ in both standardized solutions and environmental samples) through changes in the tunnelling current across films of nanoparticles (NPs) protected with striped monolayers of organic ligands. The sensors are also highly selective because of the ligand–shell organization of the NPs. On binding of metal cations, the electronic structure of the molecular bridges between proximal NPs changes, the tunnelling current increases and highly conductive paths ultimately percolate the entire film. The nanoscale heterogeneity of the structure of the film broadens the range of the cation-binding constants, which leads to wide sensitivity ranges (remarkably, over 18 orders of magnitude in CH3 Hg+ concentration).
M
any cations are toxic, posing serious consequences for the environment and for human health. As an example, organomercury species (for instance, methylmercury (ii); refs 1,2) are extremely dangerous and cause neurological disorders such as Minamata disease3,4 , and cadmium poisoning results in kidney and/or liver failure, bone softening and other ill effects (for example, Itai-itai disease5,6 ). Therefore, sensitive and selective detection of toxic metal cations is of paramount importance in assessing water pollution and in the elimination of potentially serious health and environmental hazards7–9 . There are multiple approaches to detect these ions, including atomic absorption spectroscopy10,11 and inductively coupled plasma spectroscopy12,13 (ICP), a host of optical methods (colorimetric14,15 or fluorescence-based assays16,17 and systems based on surface plasmon resonance18,19 or surface-enhanced Raman scattering20,21 ), and electrochemical methods22,23 . Frequently, however, these methods require a sophisticated chemistry to incorporate host materials, such as macrocyclic rings (crown ethers, calixarenes or porphyrins), metal-ion-recognizable proteins, nucleic acids and polymers. Moreover, the sensitivity of these methods is insufficient24–27 in several situations of public health concern (for example, accumulation of low picomolar concentrations of CH3 Hg+ and Cd2+ in fish28,29 ). Furthermore, host–guest chemistries are not available for all cations, with CH3 Hg+ being a prominent example2 . Here we show that gold nanoparticles (Au NPs) coated with binary mixtures of n-hexanethiol (HT) and alkanethiols terminated with n = 1, 2 or 3 ethylene glycol (EG) units have a striped ligand structure (Fig. 1a,b) that generates supramolecular
pockets able to selectively trap ions, and that on such trapping the electron-conduction mechanism through films of these NPs changes markedly30 . We implement these findings in a solid-state sensor of unmatched sensitivity: HT/EG3 NPs sense CH3 Hg+ selectively and with a detection limit of ∼1 aM, HT/EG2 NPs sense ∼1 pM concentrations of Cd2+ , and HT/EG1 NPs are ∼1 nM sensors for Zn2+ . Au NPs (d = 4.7 ± 0.9 nm) were synthesized as described previously31,32 (see Methods) and were functionalized with alternating stripes (Fig. 1a) of HT and EGn thiols (HS–(CH2 )6 – (OCH2 CH2 )n –OH, ultrapure grade, ProChimia). These striped NPs were examined by scanning tunnelling microscopy (STM): the width of both HT and EG2 stripes was ∼1.3 ± 0.1 nm (Fig. 1b). For control experiments, we prepared NPs coated with disordered HT/EGn mixed self-assembled monolayers (SAMs; random NPs (ref. 33)), but with the same relative content of the two thiols as in the striped NPs (see Supplementary Section S1), and also NPs covered with one-component SAMs of EGn thiols. All types of NPs were dispersed in methanol with a concentration of 2 mM of Au atoms. To fabricate the sensors, Au electrodes (30 nm thick, 5 mm long) were sputter-coated onto a glass substrate using a shadow mask; the gap between the two electrodes was typically w = 50 µm wide and l = 5 mm long (Fig. 1c). Approximately 3 µl of the Au NP solution was drop-cast onto the patterned glass and dried for two days in vacuum conditions to give NP films about H ∼ 150 nm thick (determined by profilometry for each sample); in this way, the number of NPs between the electrodes was of the order of 1011 (Fig. 1d). To prevent re-dissolution of the film on exposure
1 Department
of Materials Science and Engineering, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139, USA, 2 Institute of Materials, Ecole Polytechnique Fédérale de Lausanne (EPFL), 1015, Switzerland, 3 Department of Chemical and Biological Engineering and Department of Chemistry, Northwestern University, 2145 Sheridan Road, Evanston, Illinois 60208, USA, 4 Department of Chemical Engineering and Department of Materials Science and Engineering, University of Michigan, Ann Arbor, Michigan 48109-2136, USA. † These authors contributed equally to this work. *e-mail: [email protected]; [email protected]. NATURE MATERIALS | ADVANCE ONLINE PUBLICATION | www.nature.com/naturematerials
© 2012 Macmillan Publishers Limited. All rights reserved.
1
NATURE MATERIALS DOI: 10.1038/NMAT3406
ARTICLES a
b
HT
=
S
EG
=
S
)nOH
(O
d
c A
m m
5 mm
Striped nanoparticle 1,14-tetradecanedithiol
Au electrode
Au electrode 30 nm
50 µm
5
A
1 mm
Glass slide
Glass slide
e
400 4 200
j (A m¬2)
j (A m¬2)
2 0 ¬2
¬20
¬200
Before Zn2+ After Zn2+
¬4 ¬10
0 E (kV m¬1)
10
0
20
¬400 ¬20
Before CH3Hg+ After CH3Hg+ ¬10
0
10
20
E (kV m¬1)
Figure 1 | Experimental set-up and typical j–E plots. a,b, An idealized scheme of a striped nanoparticle (a) and the corresponding STM image (here, for HT/EG2 NPs) obtained by a Veeco multimode scanning probe microscope on a vibration damping table (in an acoustic chamber, at room temperature, and in air; b). Scale bars, 10 nm (main image) and 5 nm (inset). c,d, The scheme and the dimensions of the device (c) and the side view of the NP film (d). e, Ohmic current density versus applied field, j–E, and dependencies for a film of striped HT/EG3 Au NPs before exposure to cations (black circles) and after exposure to cations (coloured circles). In the graph on the left, the conductance of the films is virtually unchanged after immersion in a 1 mM solution of Zn2+ (green circles). In the graph on the right, immersion of the same film in a 1 mM solution of CH3 Hg+ (red circles) results in a marked change in the conductance (note the difference in the j ranges in the two plots).
to solvent, the film was reinforced by crosslinking the NPs with 1,14-tetradecanedithiol (dithiol concentration: 5 mM in toluene, crosslinking time: 40 min; for details see refs 34–36). Subsequently, the films were washed with an excess of toluene, dried under a stream of Ar, and then placed in vacuum conditions for two days. After drying the solvent, the Au electrodes were connected to a high-precision Keithley 6517 electrometer housed in a home-made Faraday cage (to minimize the effects of static electricity). The films were subsequently immersed in 2 ml of solutions of K+ , Na+ , Zn2+ , Cd2+ , Cs+ , CH3 Hg+ , Co2+ , Hg2+ , Pb2+ , Cu2+ , Ni2+ , Tl+ , Ca2+ , Ag+ and CH3 Zn+ chloride salts in ultrapure water (resistivity: 18.2 M cm) with concentrations ranging from 1 mM down to 1 zM (zeptomolar, 10−21 M; note, the nature of the anions of the salts used, for example, chlorides versus bromides, had no effect on the results described below). The films were kept immersed in these solutions for t ∼ 120 s. This time was sufficient for the ions to penetrate the full depth of the film, H —using a conservative estimate of the diffusion coefficient of ions/small molecules in a nanoporous material, D √ ∼ 10−14 m2 s−1 (ref. 37), the diffusion depth can be estimated as Dt ∼ 1 µm, which is greater than H . Consequently, the conductivities of the films discussed below did not change for longer immersion times. After immersion, the films were washed with copious amounts of water to remove excess salt and were then thoroughly dried under a stream of nitrogen. All 2
subsequent measurements were performed at room temperature (T ∼ 300±2 K), under dry air in a hermetic Faraday cage, and in the presence of desiccant (phosphorous pentoxide, P2 O5 ; the presence of the desiccant, however, had no noticeable effect on the results). The current density–applied field (j–E) dependencies for all types of films/cations were recorded for fields up to 20 kV m−1 (higher fields cause irreversible changes in the NP films and coalescence of the neighbouring nanoparticles30 ). In this range, the j–E dependencies were ohmic (Fig. 1e) and the conductivities of the films, σ = j/E, were constant for a given film type or condition (see Methods and Supplementary Section S2 for typical j–V (t ) curves at various concentrations of CH3 Hg+ ). Before the immersion into the salt solutions, the conductivities of the striped EGn particle films were σbefore ∼2 × 10−4 −1 m−1 , which is close to the values previously recorded for films of Au NPs covered with thiolate SAMs of comparable thickness30 . The conductivities increased (to σafter ) when the films were exposed to the cations (Fig. 1e). The degree of this increase depended crucially on the nature of the SAM coating the NPs, and on the nature and the concentration of the cations (Fig. 2). In all cases, the sensitivity of the sensors was defined as the concentration of cations in solution at which the ratio of the conductivities of the NP film, χ = σafter /σbefore , was greater than unity at the 99% confidence level (for statistical analysis, see Supplementary Section S3).
NATURE MATERIALS | ADVANCE ONLINE PUBLICATION | www.nature.com/naturematerials © 2012 Macmillan Publishers Limited. All rights reserved.
NATURE MATERIALS DOI: 10.1038/NMAT3406
ARTICLES
NP types Au NPs covered with HT/EG3 SAMs
b CH3Hg+ Cs+ CH3Hg+ control Cs+
10
100
10
Hg2+
Cs+ CH3Hg+
Na+
CH3Zn+
Ag+
K+
Cd2+ Cd2+
Hg2+
Ni2+
Ag+
Na+
Zn2+
(0.1 mM) CH3Hg+ (0.1 mM, random HT/EG3) Cs+ (Random HT/EG3)
Pb2+ Ag+ Ca2+
(0.1 mM) K+
CH3Zn+
Ni2+
Co2+
Cu2+ Pb2+
CH3Zn+ CH3Hg+
Tl+
Cu2+
Pb2+ Cs+
Ni2+
Tl+ Ca2+ Hg2+
Cd2+ Zn2+
10
Cu2+
d
Cd2+ K+ Cd2+ control K+ control
1 mM
χ = σ /σ after before
10
10¬12
10¬8
10¬4
Concentration (M)
10
f
10 1 mM
χ = σ /σ after before
χ = σ /σ after before
Zn2+ Na+ Zn2+ control Na+ control
(Random HT/EG2)
10¬16
(0.1 mM)
10¬20
CH3Hg+
10¬24
(Random HT/EG2) K+
1
1
1
1 10¬16
10¬12
10¬8
Concentration (M)
10¬4
Na+
10¬20
(Random HT/EG1)
10¬24
Zn2+ (Random HT/EG1)
χ = σ /σ after before
c
Cd2+
10¬4
K+ Na+
10¬8
Co2+
10¬12
Concentration (M)
Co2+
10¬16
Zn2+
10¬20
Cs+
10¬24
Ca2+
1
1
e
Au NPs covered with HT/EG1 SAMs
1 mM
Tl+
χ = σ /σ after before
100
χ = σ /σ after before
a
Au NPs covered with HT/EG2 SAMs
Figure 2 | Sensitivity (left column) and selectivity (right column) of cation sensing by different types of Au NPs decorated by HT/EGn SAMs. These measures are quantified by the change in the conductivities of the films on cation exposure, χ = σafter /σbefore . Filled symbols correspond to nanoparticles covered with striped HT/EGn SAMs; open symbols correspond to control experiments in which the nanoparticles were covered with disordered/non-striped SAMs of the same composition. Standard deviations are for at least 12 independent experiments for each condition. The arrows in the left column point to the detection limit/sensitivity (defined in the main text; see also Supplementary Section S3). a,b, Au NPs covered with HT/EG3 SAMs detect CH3 Hg+ selectively with a detection limit of ∼1 aM. c,d, HT/EG2 Au NPs detect Cd2+ and K+ with a detection limit of ∼1 pM for both cations. e,f, HT/EG1 Au NPs detect Zn2+ and also Na+ (detection limits: ∼1 nM and ∼1 µM, respectively). Note that the plots in a, c and e are double logarithmic and the conductivity increases nonlinearly with cation concentration.
Specifically, the HT/EG3 striped Au NPs detected methyl mercury, CH3 Hg+ , with an approximately attomolar (10−18 M) sensitivity limit (corresponding to only ∼600 CH3 Hg+ cations in 1 ml of the surrounding solution; Fig. 2a). As the concentration of methyl mercury increased, χ increased nonlinearly with [CH3 Hg+ ] and reached a value of ∼75 at [CH3 Hg+ ] = 0.1 mM. Moreover, these sensing characteristics were selective for methyl mercury (Fig. 2b). For Co2+ , Pb2+ , Cu2+ , Ni2+ , Tl+ , Ca2+ , Ag+ , CH3 Zn+ , Cd2+ , K+ , Na+ , Cs+ and Zn2+ , the conductivity did not change
perceptibly (χ ≈ 1) over the entire range of concentrations tested; for Hg2+ and Cs+ it changed only slightly (χ < 2.9 and χ < 2.21, respectively). Furthermore, for Cs+ , the detection limit was only ∼1 µM—that is, 12 orders of magnitude above that of CH3 Hg+ . Films of Au NPs coated with striped SAMs of HT/EG2 detected Cd2+ with χ = 24.7 at 1 mM and a detection limit ∼1 pM (Fig. 2c,d). These NPs, however, also detected K+ with χ = 10.2 at 1 mM and with ∼1 pM detection limit and, to a lesser degree, Ag+
NATURE MATERIALS | ADVANCE ONLINE PUBLICATION | www.nature.com/naturematerials
© 2012 Macmillan Publishers Limited. All rights reserved.
3
NATURE MATERIALS DOI: 10.1038/NMAT3406
ARTICLES Table 1 | Film sensitivity and energies in cation complexation. χnon-binding
χmixture
χ Cs = 1.600 ± 0.512
χ Na = 0.888 ± 0.165
χ Cs
HT/EG2
2+ χ Cd
= 2.859 ± 0.858
χ Zn
HT/EG1
χ Zn
= 2.337 ± 0.716
χ Cd
NP types
χbinding
HT/EG3
+
2+
+
+ /Na+
= 1.671 ± 0.389
2+
= 1.170 ± 0.182
2+ /Zn2+
= 2.373 ± 1.238
2+
χ Cd
= 1.105 ± 0.125
χ Zn
2+ /Cd2+
= 2.406 ± 0.383
Sensitivity (that is, χ = σafter /σbefore ) in binary mixtures comprising cations that bind (denoted in bold) to the NPs and cations that do not bind (not bold). No significant differences between χ binding and χ mixture were observed. Standard deviations are for at least six independent experiments for each condition.
with χ = 2.86. On the other hand, they did not bind Co2+ , Pb2+ , Cu2+ , Ni2+ , Tl+ , Ca2+ , CH3 Zn+ , Na+ , Zn2+ , Cs+ and CH3 Hg+ . Finally, Au NPs coated with striped SAMs of HT/EG1 were least selective and detected Zn2+ , Ag+ and Na+ (Fig. 2e,f). In this case, the values of χ were rather moderate (for example, at 1 mM, 6.51 for Zn2+ , 5.00 for Ag+ and 2.72 for Na+ ) although the sensitivity towards Zn2+ remained relatively high (for example, detection limit ∼1 nM versus ∼1 µM for Na+ ). We note that Au NPs coated with random/non-striped HT/EGn SAMs were significantly less sensitive and in most cases less selective in detecting the highest-χ cations (but not some lower-χ cations, see open symbols in Fig. 2). Also, particles coated only with pure EGn SAMs exhibited even lower values of χ (for details, see Supplementary Section S4). We emphasize that all χ values discussed above reflect the effects of cation binding and not any water and/or drying effects. This was verified in a series of experiments in which the films were washed in cation-free, deionized water and then dried—in this case, no statistically significant changes in the conductivities of the films were observed after either one or multiple wash/dry cycles (for example, for HT/EG2 Au NP films, χ = σafter /σbefore = 1.011±0.029 after one cycle and 1.013 ± 0.031 after ten cycles). Also, when the cations were first captured into the films (such that conductivities increased) but were then washed off (by placing the films in 80 ◦ C deionized water for 2–3 h), the conductivities returned to their native values before cation capture (see Supplementary Section S7). We observe that the films owe their robustness during all of these experiments (see Supplementary Section S6) to the crosslinking of the NPs by dithiols; films in which the NPs were not crosslinked simply dissolved when washed with or immersed in water. The changes in the conductivity of the film result from the capture of the cations by the ligands on the NPs. This was directly verified by means of the ICP and X-ray photoelectron spectroscopy analyses of NPs before and after ion capture (see Supplementary Section S5 for these and other techniques used). In addition, Fourier transform infrared spectroscopy studies were performed that evidenced spectral changes on cation binding. Of particular importance here were the changes in the characteristic vibrational modes of the ethylene glycol units at 1001–1003 cm−1 and at 827 cm−1 . Remarkably, these changes were reversible in the sense that the original spectra were retrieved when the cations were washed off the NPs during immersion in 80 ◦ C water (see above). These results also indicate that the capture of the cations is mediated by coordination-type interactions with the EG units, as should indeed be expected on the basis of common chemical knowledge. It is also worth noting that cations for which χ = σafter /σbefore is close to unity do not cause significant changes in the NP/NP–film spectra as discussed in the Supplementary Section S5. Interestingly, a qualitatively similar result is obtained from solution experiments where the conductivities of NP suspensions are monitored during titration with different salt solutions. For those salts that are characterized by low χ in our solid-state (that is, NP–film) measurements, the conductance of NP solutions increases linearly 4
with the amount of salt added; for those, however, that exhibit higher χ in NP films, the solution conductance initially remains constant, indicating that the added cations are trapped on the NPs (and thus do not contribute to the conductance of the solution; see Supplementary Section S5.5). An essential feature of any chemical sensor is its selectivity, which in our case needs to be determined on exposure of NP films to cation mixtures. We first consider binary mixtures comprising cations that the NP films bind and sense with statistical significance (that is, giving conductance ratios χ = σafter /σbefore > 1 at 99% confidence levels) and cations that do not bind strongly and do not change conductivities at such confidence levels. Such pairs correspond to most of the possible combinations among the cations we tested (see Fig. 2). The results for some representative pairs are summarized in Table 1. Specifically, for the HT/EG3 Au NP films, the presence of weakly binding Na+ does not significantly change the signal coming from stronger binding Cs+ . Similarly, for HT/EG2 Au NP films, the presence of Zn2+ does not strongly affect the signal due to Cd2+ , and for HT/EG1 Au NP films the conductance on exposure to the mixture of Zn2+ and Cd2+ is close to the value characterizing the former cation. Such data confirm the selectivity of the films towards cations giving larger values of χ . Moreover, the signals from the NP–film sensors are not affected by the changes in the ionic strength of the solution due to the non-binding, low-χ cations (see Supplementary Fig. S9a). Of course, the selectivity decreases when the films are exposed to mixtures of cations characterized by similar χ > 1 values (Supplementary Section S12). This problem is the least pronounced for HT/EG3 Au NP films (see Fig. 2 and Supplementary Section S8), which are therefore suitable for deployment as environmental CH3 Hg+ sensors. This is illustrated in Fig. 3, where we used our sensors to measure the CH3 Hg+ content in environmental samples directly relevant to the issues of public health—tap water, lake water and fish (see Methods for detailed preparation and analysis of environmental samples). Importantly, the results we obtained for the lake water could be directly compared to the values reported in the literature for the same lake (Lake Michigan) sample, and the CH3 Hg+ content in fish could be compared to the value determined by the United States Geological Survey (USGS). For the lake water, our measurements gave a 5.9 fM concentration of CH3 Hg+ (range within one s.d. 0.35–43.3 fM) versus 5–210 fM reported in ref. 38. For the fish, we measured a concentration of 3.81 pM (range within one s.d.: 1.62–8.19 pM), in excellent agreement with the 3.58 pM value provided to us by the USGS after our experiments were completed. A reassuring result—at least for the citizens of Evanston—is that the content of CH3 Hg+ in the tap water of this city is very low (27.7 aM, range within one s.d.: 4.5–138 aM). We make two further comments about all of these results. First, when the environmental samples were spiked with known concentrations of CH3 Hg+ , the readings of our films fell onto the calibration curve obtained using the concentration standards (that is, solutions of CH3 Hg+ in pure, deionized water)—this means that there is no background interference from components other than CH3 Hg+ in the environmental samples. Second, we emphasize that our
NATURE MATERIALS | ADVANCE ONLINE PUBLICATION | www.nature.com/naturematerials © 2012 Macmillan Publishers Limited. All rights reserved.
NATURE MATERIALS DOI: 10.1038/NMAT3406
Table 2 | Film sensitivity and energies in cation complexation.
CH3 Hg+ Lake Michigan ( from ref. 38) Lake Michigan spiked Tap water Fish ( from USGS) Fish blank spiked
100
χ (¬)
ARTICLES
10
NP/cation types Free Zn2+ Cd2+
(This work)
CH3 Hg+
(This work)
1 10¬23
} (reference samples) 10¬20
10¬17 10¬14 10¬11 10¬8 Concentration CH3 Hg+ (M)
10¬5
results were achieved in a technically straightforward manner (by simply dipping the film in the sample solution), in contrast to other methods that require tedious sample pretreatment, including conversion of methylmercury to ethylmethylmercury and concentration on a dedicated column before analysis by gas chromatography combined with ICP mass spectrometry2,39 . We now turn our attention to the fundamental aspects of the relationship between the binding of cations into the film and the observed current changes. There are two interrelated components that need to be considered: cation-facilitated electron transport through the molecular bridge between proximal NPs, and formation of the conductive paths through the film, which ultimately give rise to the high sensitivity of the cation detection. To understand the electronic structure of the bridge, we performed quantum-mechanical calculations for a fragment of NP surface comprising two nearby EGn stripes and an HT stripe between them (Fig. 4a). Calculations were performed for both the cationfree HT/EGn systems, as well as a system in which different metal cations were placed either within the EGn stripes (denoted intra in Table 2) or at the edges of these stripes (inter). In all cases, the geometry of the system was optimized by a semiempirical quantum-mechanical method NDDO-PM6 (ref. 40; with the MOPAC2007 (ref. 41) software) at the Non-equilibrium Energy Research Center supercomputer centre. The calculations indicate that the experimentally observed selectivities do not correlate with the calculated energies of the complexation of the cations by various HT/EGn Au NPs or with the heats of formation of these complexes, nor with the energies of the highest occupied molecular orbital (HOMO) or the lowest unoccupied molecular orbital (LUMO). On the other hand, the observed cation selectivities can be related to the HOMO/LUMO gap, 1E (Table 2). This result suggests that both HOMO and LUMO orbitals are involved in electron transport. Although this is not a typical scenario in describing
HT/EG2
HT/EG3
9.160 1.834 1.398 4.395 3.928 6.045 6.154
9.143 2.408 1.976 3.969 3.676 6.392 6.129
9.142 3.098 3.031 4.088 4.243 6.272 6.045
Calculated energies of the HOMO/LUMO gaps (in eV) for different cations and EGn thiols. ‘Interband’ corresponds to the situation where the cations are bound at the edge of an EGn stripe; ‘intraband’ corresponds to the cation being placed within one EGn band. Entries in bold correspond to minimal gap energies for a given type of EGn and generally agree with/reflect the cation-binding selectivity observed in experiments.
10¬2
Figure 3 | Selectivity of the films on exposure to cation mixtures and environmental samples. The χ values measured for the environmental samples (blue star: water from Lake Michigan, purple diamond: Evanston tap water, green triangle: mosquito fish collected by USGS from Everglades National Park). Red symbols correspond to the measurements of standard solutions (that is, of known CH3 Hg+ concentration in deionized water) and the dashed calibration curve is the second order polynomial fit. Open symbols correspond to the environmental samples ‘spiked’ with known amounts of CH3 Hg+ . Pale green and pale blue regions correspond to the one-standard-deviation ranges in our measurements (three different films, at least six measurements per film) of, respectively, fish and lake water samples. Dark green horizontal bar corresponds to the range of CH3 Hg+ content in the mosquito fish determined by USGS, and the dark blue horizontal bar corresponds to the one-standard-deviation range of CH3 Hg+ concentration in Lake Michigan determined in ref. 38. For further experimental detail, see Methods.
Intra Inter Intra Inter Intra Inter
HT/EG1
conduction through a molecular bridge42 , similar models have been used previously to describe transport through disordered media (as our films are)43 . In this context, we observe that in our system the HOMO orbital (and other high-energy occupied orbitals) is always located close to the NP surface (Fig. 4b), whereas the LUMO orbital (and unoccupied orbitals close to it) is mostly centred on the EGn /cation fragments (Fig. 4c). Therefore, for the electrons to travel through the SAM, they must be injected from the HOMO into the LUMO. Importantly, the energy required for this injection is substantially diminished on cation binding, and the smallest values of 1E generally reflect the detection specificity observed in experiments: for Zn2+ , the smallest 1E corresponds to the binding to the HT/EG1 Au NPs; Cd2+ has the smallest 1E when binding to HT/EG2 Au NPs; for CH3 Hg + , 1E is the least when binding inter-stripe to HT/EG3 Au NPs (although, in this case, calculations give the same value for HT/EG1 particles; Table 2). We note that for any given HT/EGn /cation system, the 1E gaps (and the absolute binding energies) are typically lower when cations coordinate with ethylene glycols at the edges of the EGn stripes than when these cations are fully buried within the EGn stripes. This observation suggests that the free space provided by the HT stripes is important for the operation of our sensors (as also corroborated by molecular dynamics simulations illustrated in Fig. 4d and Supplementary Section S9). The decrease in 1E on cation binding increases the electronic conductance of a NP-SAM//SAM-NP molecular bridge, σc ∝ 1/2 (e2 /h)(e−2(m1E) L/h )2 , where h is Planck’s constant, m is the rest mass of an electron and L is the thickness of the NP ligand shell44 . Importantly, as multiple cations bind into the NP film, the local changes over individual bridges can translate into a global change in the conductance of the film. This process can be described by combining σc with the cation-binding kinetics and Bruggeman’s effective medium theory45,46 (EMT) accounting for the formation of conductive paths percolating the NP film (Fig. 5a,b). The binding of cations to the EGn ligands as a function of concentration [C]0 can be described in terms of the Sips isotherm47 : θ = (Keq [C]0 )a /[1 + (Keq [C]0 )a ], where θ is the fraction of occupied sites, Keq is the equilibrium association constant and a is the heterogeneity index. The Sips isotherm is a composite of the familiar Langmuir isotherm (θLM = KLM [C]0 /[1 + KLM [C]0 ]) and the Freundlich isotherm (θF = KF [C]1/n 0 ), commonly used for low-concentration binding in heterogeneous systems. The Sips isotherm approaches Langmuir’s formula for a close to unity and Freundlich’s equation at low concentration, and has been used to describe ligand binding to powders48 and aggregated nanoparticle media49 . The heterogeneity index (a) determines the width of the Gaussian-like distribution of Keq , which arises from steric effects, conformational changes and differences in the chemical microenvironment50 . Using the
NATURE MATERIALS | ADVANCE ONLINE PUBLICATION | www.nature.com/naturematerials
© 2012 Macmillan Publishers Limited. All rights reserved.
5
NATURE MATERIALS DOI: 10.1038/NMAT3406
ARTICLES a
b
c HT EGn
EGn
d
Figure 4 | Rationalizing the conductance through the molecular bridge between √ the√striped NPs. a, Top view of the modelled system comprising 15 HT and 4 EGn thiols arranged in alternating stripes on the Au(111) surface with a ( 3 × 3) R30◦ lattice distribution of the HT/ EGn ligands. b,c, HOMO (b) and LUMO (c) orbitals calculated for the HT/EG3 Au NPs capturing CH3 Hg+ . Note that the HOMO is closer to the surface of the nanoparticle (localized on the lone pairs of the sulphur atoms), whereas the LUMO is centred on the EGn /cation complexes (corresponding to the anti-bonding σ ∗ (O–M) of the EGn /cation complexes). These simulations are performed for nanoparticles in vacuum, which is appropriate for the experimental measurements being performed in dry films. d, A snapshot from the molecular dynamics simulations showing a typical ion-trapping scenario where the cation (green ball) is trapped near the edge of an EGn stripe (for further details, see Supplementary Section S9).
expressions for σc and θ, the EMT formula describing the change in the conductance of the film on cation binding becomes (see Supplementary Section S11): ! a 1 z Keq [C]0 a − 1 σc,n 2 1 + Keq [C]0 a ! ! Keq [C]0 1 a − 1 + z 1− 2 1 + Keq [C]0 " ! a 1 z Keq [C]0 a − 1 σc,n + 2 1 + Keq [C]0
1 χS-EMT = (z − 2)
+
(
1/2 a ! !!2 Keq [C]0 1 a − 1 z 1− + 2(z − 2)σc,n (1) 2 1 + Keq [C]0
where σc is normalized by dividing by the conductance of the film before binding cations (σc,n = σc /σbefore ), and z is the number of high-conductance bonds any NP can make to its neighbours. Figure 5c illustrates that with the experimentally measured (by isothermal titration calorimetry, see Supplementary Section S10) values of Keq , this model reproduces the essential features of the nonlinear cation concentration versus conductivity dependencies recorded in experiments. As expected, the fitted values of a and z for 6
both Cd2+ and K+ are nearly equal (for Cd2+ : a = 0.42 ± 0.32 and z = 2.1 ± 0.5; for K+ : a = 0.42 ± 0.55 and z = 1.9 ± 1.0), because they reflect the intrinsic nanostructure of the Au NP films. The values of z ∼ 2 suggest that there are relatively few possibilities to form high-conductance connections between neighbouring NPs in the film (z = 6 would signify that such connections can be made between all nearest-neighbour NPs in a randomly packed film)46 . More importantly, the values of a < 1 confirm that there is significant heterogeneity of the cation binding to the NP film (unlike in isothermal titration calorimetry solution studies for which a = 1). This result agrees with previous works on the nanoenvironment47 , and it endows our sensor with a remarkably wide range of sensitivity. If the cation binding properties in the films were the same as those observed in solution (a = 1), the sensitivity curves would look like the dashed curve in Fig. 5c. As seen, this curve transitions sharply between constant χmin and χmax values over a relatively narrow range of concentrations (∼4 orders of magnitude). In contrast, with the heterogeneity effects being operative, the ranges over which χ values change are much wider, ∼9 orders of magnitude for Cd2+ and K+ and even more (∼18 orders) for CH3 Hg+ . Although the broadening of binding energies is generally considered a disadvantage that complicates affinity studies50 , it is in fact the key—and unique— principle underlying the remarkable range of sensing of our NP devices. In other words, the packing of chemically identical NPs into a nanostructured thin film gives rise to adsorption sites with not a single but many different energies (and thus Keq
NATURE MATERIALS | ADVANCE ONLINE PUBLICATION | www.nature.com/naturematerials © 2012 Macmillan Publishers Limited. All rights reserved.
NATURE MATERIALS DOI: 10.1038/NMAT3406 a
ARTICLES c 100
Cd2+ (
fit S-EMT)
(
fit S-EMT)
K2+
χ (¬)
Langmuir
10
b
1 10–17
10–13
10–9 10–5 Concentration (M)
10–1
Figure 5 | Percolation of Au NP films on cation binding. a, Schematic representation of a film of striped Au NPs, between two Au electrodes. b, On binding of cations (blue spheres) conductive paths (blue line) connecting the two electrodes percolate the film and connect the two electrodes. c, Sensitivity (that is, χ = σafter /σbefore ) of HT/EG2 Au NP films on binding of K+ and Cd2+ from solutions of different concentrations. Symbols correspond to the experimental data (the error bars are based on 12 independent measurements for each condition). The solid lines are theoretical fits based on the model described in the main text (equation (1)). The values of Keq parameters were determined calorimetrically (see Supplementary Section S10) and were Cd2+ = 2.75 ± 0.31 × 106 M−1 for Cd2+ and K K+ = 1.5 ± 0.12 × 106 M−1 for K+ . As a reference, the Langmuir isotherm (a = 1) for Cd2+ is also included Keq eq (dashed black line).
values). The sites with high Keq values are the first ones to be occupied at very low concentration, whereas the low Keq sites saturate at much higher concentrations. This effect grants the NP films with both a low onset and, at the sametime, a very broad range of detection. Overall, the high sensitivity of our solid-state sensors derives from a subtle interplay between cation/striped-NP macromolecular recognition, electron transport over the NP/cation/NP bridges, and the formation of conductive paths through heterogeneous NP films. Together, these phenomena provide a new principle for solid-state sensing and enable inexpensive, sensitive and portable environmental sensors. With further tailoring—both in terms of ligand chemistries and stripe dimensions—it should be possible to extend the applicability of our nanostructured materials to other sensing systems whose properties derive from macromolecular recognition on NP surfaces.
Methods Synthesis of striped Au NPs. Striped Au NPs covered by 1-hexanethiol (Sigma-Aldrich 739286) and 1-hexanethiol terminated in n = 1–3 units of ethylene glycol (EGn , ProChimia) (n = 1, ProChimia TH 001-m6.n1-1; n = 2, ProChimia TH 001-m6.n2-1; n = 3, ProChimia TH 001-m6.n3-1, respectively) in 1:3 ratio were synthesized by a modified Stucky method51 . Briefly, 0.45 mmol of AuPPh3 Cl was mixed with 0.45 mmol of thiol mixture in 40 ml of chloroform/dimethylformamide co-solvent, and 5 mmol of tert-butylamine-borane complex was added. The reactant was heated to 55 ◦ C and stirred for 1 h. After cooling to room temperature, 200 ml of chloroform was added to precipitate Au NPs, and the product was centrifuged and washed with chloroform. The ligand shell structure was determined using STM as described previously (see Fig. 1b in the main text)32,51–54 . Further characterization is included in Supplementary Section S5. Film conductivities. The film conductivities of the HT/EGn Au NP films before capturing cations, σbefore , were 1.9 × 10−4 ± 1.6 × 10−4 −1 m−1 , 1.4 × 10−4 ± 8.7 × 10−5 −1 m−1 and 2.7 × 10−4 ± 1.3 × 10−4 −1 m−1 for n = 1, n = 2 and n = 3, respectively. These values were calculated as σbefore = I /wt × 1/V × l = j/V × l, where I and j are current (A) and current density (A m−2 ), respectively, V is applied voltage (V), and w, t and l are the width, thickness and gap of the Au electrode, respectively (w = 5 mm, t = 30 nm and l = 50 µm). The conductivity values and their s.d. are based on at least 12 different material samples for each type of a film. We note that these conductivity values are comparable to those measured for films of Au NPs covered with thiols of similar thickness30,55 . Preparation and analysis of environmental samples. To prevent any external contamination of the samples, all containers (glass vials, VWR) were cleaned
in 4 M aqueous HCl solution at 65 ◦ C (for 12 h) and rinsed three times with copious amounts of deionized water (resistivity: 18.2 M cm). All experiments were performed in a LabConco biosafety class II laminar flow cabinet with a high-efficiency particulate arresting (HEPA) 99.99% filter in a HEPA-filtered room, similar to class 100 cleanroom conditions. Water from Lake Michigan was sampled at Northwestern University (Evanston Campus) private beach into a clean polypropylene tube (VWR). Tap water was sampled from a water fountain in the Technological Institute building at Northwestern University. Fish samples were obtained from the USGS, Mercury Research Laboratory (http://wi.water.usgs.gov/mercury-lab/index.html). To prepare fish samples for analysis, around 10 mg of lyophilized mosquito fish from Everglades National Park collected by the USGS was exactly weighed and digested using 10 ml of 5 M HNO3 for 8 h at 65 ◦ C and then neutralized using 5 M aqueous KOH solution (note: the pH was checked by a pH meter to confirm that the solution was neutralized—our Au NP film is unstable in very acidic solutions). The solution was then diluted 1,000 times to bring the background salt (KNO3 ) concentration down to the millimolar range. For spiking the blank fish and lake water samples, 10% (v/v) of 10−12 M,10−8 M and 10−4 M of CH3 Hg+ solution were added.
Received 15 September 2011; accepted 20 July 2012; published online 9 September 2012
References 1. Boening, D. W. Ecological effects, transport, and fate of mercury: A general review. Chemosphere 40, 1335–1351 (2000). 2. Clarkson, T. W. The toxicology of mercury. Crit. Rev. Clin. Lab. Sci. 34, 369–403 (1997). 3. Matsumoto, H., Koya, G. & Takeuchi, T. Fetal minamata disease: A neuropathological study of two cases of intrauterine intoxication by a methyl mercury compound. J. Neuropathol. Exp. Neurol. 24, 563–574 (1965). 4. Harada, M. Minamata disease—methylmercury poisoning in Japan caused by environmental pollution. Crit. Rev. Toxicol. 25, 1–24 (1995). 5. Inaba, T. et al. Estimation of cumulative cadmium intake causing Itai-itai disease. Toxicol. Lett. 159, 192–201 (2005). 6. Nomiyama, K. Recent progress and perspectives in cadmium health-effects studies. Sci. Total Environ. 14, 199–232 (1980). 7. Nriagu, J. O. & Pacyna, J. M. Quantitative assessment of worldwide contamination of air, water and soils by trace-metals. Nature 333, 134–139 (1988). 8. Lin, S. H. & Juang, R. S. Heavy metal removal from water by sorption using surfactant-modified montmorillonite. J. Hazard. Mater. 92, 315–326 (2002). 9. Jensen, S. & Jernelov, A. Biological methylation of mercury in aquatic organisms. Nature 223, 753–754 (1969). 10. Narin, I., Soylak, M., Elci, L. & Dogan, M. Determination of trace metal ions by AAS in natural water samples after preconcentration of pyrocatechol violet complexes on an activated carbon column. Talanta 52, 1041–1046 (2000). 11. Ghaedi, M., Ahmadi, F. & Shokrollahi, A. Simultaneous preconcentration and determination of copper, nickel, cobalt and lead ions content by flame atomic absorption spectrometry. J. Hazard. Mater. 142, 272–278 (2007).
NATURE MATERIALS | ADVANCE ONLINE PUBLICATION | www.nature.com/naturematerials
© 2012 Macmillan Publishers Limited. All rights reserved.
7
NATURE MATERIALS DOI: 10.1038/NMAT3406
ARTICLES 12. Caroli, S., Forte, G., Iamiceli, A. L. & Galoppi, B. Determination of essential and potentially toxic trace elements in honey by inductively coupled plasma-based techniques. Talanta 50, 327–336 (1999). 13. Demuth, N. & Heumann, K. G. Validation of methylmercury determinations in aquatic systems by alkyl derivatization methods for GC analysis using ICP-IDMS. Anal. Chem. 73, 4020–4027 (2001). 14. Lee, J. S., Han, M. S. & Mirkin, C. A. Colorimetric detection of mercuric ion (Hg2+ ) in aqueous media using DNA-functionalized gold nanoparticles. Angew. Chem. Int. Ed. 46, 4093–4096 (2007). 15. Huang, C. C. & Chang, H. T. Parameters for selective colorimetric sensing of mercury(II) in aqueous solutions using mercaptopropionic acid-modified gold nanoparticles. Chem. Commun. 1215–1217 (2007). 16. Chiang, C. K., Huang, C. C., Liu, C. W. & Chang, H. T. Oligonucleotide-based fluorescence probe for sensitive and selective detection of mercury (II) in aqueous solution. Anal. Chem. 80, 3716–3721 (2008). 17. Chang, H. Y., Hsiung, T. M., Huang, Y. F. & Huang, C. C. Using rhodamine 6G-modified gold nanoparticles to detect organic mercury species in highly saline solutions. Environ. Sci. Technol. 45, 1534–1539 (2011). 18. Kang, T., Hong, S. R., Moon, J., Oh, S. & Yi, J. Fabrication of reusable sensor for detection of Cu2+ in an aqueous solution using a self-assembled monolayer with surface plasmon resonance spectroscopy. Chem. Commun. 3721–3723 (2005). 19. Li, D., Wieckowska, A. & Willner, I. Optical analysis of Hg2+ ions by oligonucleotide-gold-nanoparticle hybrids and DNA-based machines. Angew. Chem. Int. Ed. 47, 3927–3931 (2008). 20. Wang, G. Q., Wang, Y. Q., Chen, L. X. & Choo, J. Nanomaterial-assisted aptamers for optical sensing. Biosens. Bioelectron. 25, 1859–1868 (2010). 21. Yin, J. et al. SERS-active nanoparticles for sensitive and selective detection of cadmium ion (Cd2+ ). Chem. Mater. 23, 4756–4764 (2011). 22. Lin, Z. Z., Li, X. H. & Kraatz, H. B. Impedimetric immobilized DNA-based sensor for simultaneous detection of Pb2+ , Ag+ , and Hg2+ . Anal. Chem. 83, 6896–6901 (2011). 23. Zhang, Z., Yu, K., Bai, D. & Zhu, Z. Q. Synthesis and electrochemical sensing toward heavy metals of bunch-like bismuth nanostructures. Nanoscale Res. Lett. 5, 398–402 (2010). 24. Thompson, R. B., Maliwal, B. P., Feliccia, V. L., Fierke, C. A. & McCall, K. Determination of picomolar concentrations of metal lens using fluorescence anisotropy: Biosensing with a ‘reagentless’ enzyme transducer. Anal. Chem. 70, 4717–4723 (1998). 25. Forzani, E. S., Zhang, H. Q., Chen, W. & Tao, N. J. Detection of heavy metal ions in drinking water using a high-resolution differential surface plasmon resonance sensor. Environ. Sci. Technol. 39, 1257–1262 (2005). 26. Kumar, M. & Zhang, P. Highly sensitive and selective label-free optical detection of mercuric ions using photon upconverting nanoparticles. Biosens. Bioelectron. 25, 2431–2435 (2010). 27. Song, H. D. et al. Picomolar selective detection of mercuric ion (Hg2+ ) using a functionalized single plasmonic gold nanoparticle. Nanotechnology 21, 145501 (2010). 28. Hinkle, P. M., Kinsella, P. A. & Osterhoudt, K. C. Cadmium uptake and toxicity via voltage-sensitive calcium channels. J. Biol. Chem. 262, 16333–16337 (1987). 29. Rashed, M. N. Cadmium and lead levels in fish (Tilapia nilotica) tissues as biological indicator for lake water pollution. Environ. Monit. Assess. 68, 75–89 (2001). 30. Nakanishi, H. et al. Photoconductance and inverse photoconductance in films of functionalized metal nanoparticles. Nature 460, 371–375 (2009). 31. DeVries, G. A. et al. Divalent metal nanoparticles. Science 315, 358–361 (2007). 32. Jackson, A. M., Myerson, J. W. & Stellacci, F. Spontaneous assembly of subnanometre-ordered domains in the ligand shell of monolayer-protected nanoparticles. Nature Mater. 3, 330–336 (2004). 33. Witt, D., Klajn, R., Barski, P. & Grzybowski, B. A. Applications properties and synthesis of omega-functionalized n-alkanethiols and disulfides—the building blocks of self-assembled monolayers. Curr. Org. Chem. 8, 1763–1797 (2004). 34. Kowalczyk, B., Apodaca, M. M., Nakanishi, H., Smoukov, S. K. & Grzybowski, B. A. Lift-Off and micropatterning of mono- and multilayer nanoparticle films. Small 5, 1970–1973 (2009). 35. Klajn, R., Bishop, K. J. M. & Grzybowski, B. A. Light-controlled self-assembly of reversible and irreversible nanoparticle suprastructures. Proc. Natl Acad. Sci. USA 104, 10305–10309 (2007). 36. Kowalczyk, B., Lagzi, I. & Grzybowski, B. A. Nanoarmoured droplets of different shapes formed by interfacial self-assembly and crosslinking of metal nanoparticles. Nanoscale 2, 2366–2369 (2010). 37. Han, S. B. et al. Chromatography in a single metal-organic framework (MOF) crystal. J. Am. Chem. Soc. 132, 16358–16361 (2010).
8
38. Sullivan, K. A. & Mason, R. P. The concentration and distribution of mercury in Lake Michigan. Sci. Total Environ. 213, 213–228 (1998). 39. Jackson, B., Taylor, V., Baker, R. A. & Miller, E. Low-level mercury speciation in fresh waters by isotope dilution GC-ICP-MS. Environ. Sci. Technol. 43, 2463–2469 (2009). 40. Pople, J. & Beveridge, D. Approximate Molecular Orbital Theory (McGraw-Hill, 1970). 41. Stewart, J. J. P. Stewart Computational Chemistry (Colorado Springs, 2007). 42. Nitzan, A. Chemical Dynamics in Condensed Phases : Relaxation, Transfer and Reactions in Condensed Molecular Systems Ch. 16 (Oxford, 2006). 43. Burlakov, V. M. et al. Discrete hopping model of exciton transport in disordered media. Phys. Rev. B 72, 075206 (2005). 44. Holmlin, R. E. et al. Electron transport through thin organic films in metal-insulator-metal junctions based on self-assembled monolayers. J. Am. Chem. Soc. 123, 5075–5085 (2001). 45. Landauer, R. The electrical resistance of binary metallic mixtures. J. Appl. Phys. 23, 779–784 (1952). 46. Kirkpatr, S. Classical transport in disordered media—scaling and effective-medium theories. Phys. Rev. Lett. 27, 1722–1725 (1971). 47. Sips, R. On the structure of a catalyst surface. J. Chem. Phys. 16, 490–495 (1948). 48. Kumar, K. V. & Porkodi, K. Relation between some two- and three-parameter isotherm models for the sorption of methylene blue onto lemon peel. J. Hazard. Mater. 138, 633–635 (2006). 49. Hristovski, K., Baumgardner, A. & Westerhoff, P. Selecting metal oxide nanomaterials for arsenic removal in fixed bed columns: From nanopowders to aggregated nanoparticle media. J. Hazard. Mater. 147, 265–274 (2007). 50. Vijayendran, R. A. & Leckband, D. E. A quantitative assessment of heterogeneity for surface-immobilized proteins. Anal. Chem. 73, 471–480 (2001). 51. Verma, A. et al. Surface-structure-regulated cell-membrane penetration by monolayer-protected nanoparticles. Nature Mater. 7, 588–595 (2008). 52. Jackson, A. M., Hu, Y., Silva, P. J. & Stellacci, F. From homoligand- to mixed-ligand-monolayer-protected metal nanoparticles: A scanning tunneling microscopy investigation. J. Am. Chem. Soc. 128, 11135–11149 (2006). 53. Kuna, J. J. et al. The effect of nanometre-scale structure on interfacial energy. Nature Mater. 8, 837–842 (2009). 54. Centrone, A. et al. The role of nanostructure in the wetting behavior of mixed-monolayer-protected metal nanoparticles. Proc. Natl Acad. Sci. USA 105, 9886–9891 (2008). 55. Zabet-Khosousi, A. & Dhirani, A. A. Charge transport in nanoparticle assemblies. Chem. Rev. 108, 4072–4124 (2008).
Acknowledgements This work was supported by the Non-equilibrium Energy Research Center, which is an Energy Frontier Research Center funded by the US Department of Energy, Office of Science, Office of Basic Energy Sciences under grant number DE-SC0000989. E.S.C. and F.S. acknowledge the support of ENI within the MIT Energy initiative for their work. E.S.C. is supported by a Samsung Scholarship from the Samsung Foundation of Culture. H.J., S.C.G. and F.S. acknowledge support from the Defense Threat Reduction Agency under Grant No. HDTRA1-09-1-0012. T.M.H. is financially supported by the Human Frontier Science Program. We acknowledge L. Meda for her X-ray photoelectron spectroscopy measurements, and R. Borrelli and P. Cesti for helpful discussions. We also thank the USGS for providing fish samples and for helpful discussions.
Author contributions E.S.C. synthesized the striped nanoparticles, performed all of their characterization before and after binding to ions, and helped with some of the solid-state work.; J.K. made and characterized all sensors, performed cation capture experiments, measured the conductivities of all sensors and analysed data; B.T. developed quantum-mechanical models and ran quantum-mechanical calculations; T.M.H. developed and validated the binding/percolation models; H.J. ran and analysed MD simulations; S.C.G. analysed and supervised the MD simulations; H.N. performed initial experiments with NP films; M.Y. performed the STM characterization of the particles; A.Z.P. helped with the theory of percolation phenomena; F.S. and B.A.G. conceived the ideas and supervised the research.
Additional information Supplementary information is available in the online version of the paper. Reprints and permissions information is available online at www.nature.com/reprints. Correspondence and requests for materials should be addressed to F.S. or B.A.G.
Competing financial interests The authors declare no competing financial interests.
NATURE MATERIALS | ADVANCE ONLINE PUBLICATION | www.nature.com/naturematerials © 2012 Macmillan Publishers Limited. All rights reserved.
