Anal. Chem. 2010, 82, 2087–2094
Locally Functionalized Short-Range Ordered Nanoplasmonic Pores for Bioanalytical Sensing Magnus P. Jonsson,* Andreas B. Dahlin, Laurent Feuz, Sarunas Petronis, and Fredrik Ho¨o¨k* Department of Applied Physics, Chalmers University of Technology, SE-41296 Gothenburg, Sweden Nanoplasmonic sensors based on short-range ordered nanoholes in thin metal films and discrete metal nanoparticles are known to provide similar sensing performance. However, a perforated metal film is unique in the sense that the holes can be designed to penetrate through the substrate, thereby also fulfilling the role of nanofluidic channels. This paper presents a bioanalytical sensing concept based on short-range ordered nanoplasmonic pores (diameter 150 nm) penetrating through a thin (around 250 nm) multilayer membrane composed of gold and silicon nitride (SiN) that is supported on a Si wafer. Also, a fabrication scheme that enables parallel production of multiple (more than 50) separate sensor chips or more than 1000 separate nanoplasmonic membranes on a single wafer is presented. Together with the localization of the sensitivity to within such short-range ordered nanoholes, the structure provides a twodimensional nanofluidic network, sized in the order of 100 × 100 µm2, with nanoplasmon active regions localized to each individual nanochannel. A material-specific surface-modification scheme was developed to promote specific binding of target molecules on the optically active gold regions only, while suppressing nonspecific adsorption on SiN. Using this protocol, and by monitoring the temporal variation in the plasmon resonance of the structure, we demonstrate flow-through nanoplasmonic sensing of specific biorecognition reactions with a signalto-noise ratio of around 50 at a temporal resolution below 190 ms. With flow, the uptake was demonstrated to be at least 1 order of magnitude faster than under stagnant conditions, while still keeping the sample consumption at a minimum. Bioanalytical sensor devices have emerged as increasingly important tools in medical diagnostics and in the development of new drugs. They are also essential for both environmental monitoring and food safety. In common sensing methods the target biomolecules are labeled with, for example, fluorophores, which in combination with array-based read-out can provide parallel identification of both DNA and protein targets.1 However, in many situations it is neither straightforward nor desirable to label the target molecules. In this context, label-free surface sensitive concepts, such as surface plasmon resonance (SPR),2 * To whom correspondence should be addressed. E-mail: magnus.p.jonsson@ chalmers.se (M.P.J.),
[email protected] (F.H.). (1) Sanders, G. H. W.; Manz, A. TrAC, Trends Anal. Chem. 2000, 19, 364–378. (2) Homola, J.; Yee, S. S.; Gauglitz, G. Sens. Actuators, B 1999, 54, 3–15. 10.1021/ac902925e 2010 American Chemical Society Published on Web 02/03/2010
quartz crystal microbalance (QCM),3 and electrochemical/impedance-based techniques,4 have emerged as attractive alternatives. With the receptor molecules immobilized on the sensor surface, binding of target biomolecules can be transduced to a detectable signal via changes in the optical, mechanical, and/or electrical properties of such devices. Besides avoiding the need of molecular labels, these concepts provide real-time recording of binding/ unbinding reactions and thereby information about binding kinetics and reaction pathways. SPR, which has been commercially available for nearly 20 years, is today the dominant transducer principle used for labelfree bioanalytical sensing. The method is based on optical excitation of collective charge oscillations called surface plasmons that propagate on a flat metal surface. The resonance condition for exciting surface plasmons is sensitive to changes in refractive index (RI) near the metal interface. Biomolecular binding/ unbinding reactions near the surface can therefore be followed in real-time and in a label-free format by monitoring the temporal variation in the surface plasmon resonance.5 However, to excite surface plasmons on a flat metal film, both their energy and momentum must match that of the incident light. Therefore, relatively complicated optical configurations are required, including a prism in optical contact with the substrate.6 Alternatively, light can be coupled to surface plasmons without prisms and even at normal incidence using a grating on the metal surface.7,8 More recently it was demonstrated that periodic arrays of nanoscale holes can provide the missing momentum that is required for exciting surface plasmons without prisms.9 Both of these concepts have been successfully used for RI-based plasmonic biosensing10,11 and, in the case of periodic arrays of nanoscale holes, also for flow-through sensing applications.12 Thin metal films perforated with nanoholes distributed in a short-range order (no periodicity) also exhibit distinct RI-sensitive plasmonic resonances that can (3) Janshoff, A.; Galla, H. J.; Steinem, C. Angew. Chem., Int. Ed. 2000, 39, 4004–4032. (4) Grieshaber, D.; MacKenzie, R.; Vo ¨ro ¨s, J.; Reimhult, E. Sensors 2008, 8, 1400–1458. (5) Liedberg, B.; Nylander, C.; Lundstrom, I. Sens. Actuators 1983, 4, 299–304. (6) Kretschmann, E.; Raether, H. Z. Naturforsch., A: Astrophys., Phys. Phys. Chem. 1968, A 23, 2135–2136. (7) Ritchie, R. H.; Arakawa, E. T.; Cowan, J. J.; Hamm, R. N. Phys. Rev. Lett. 1968, 21, 1530–1533. (8) Wood, R. W. Philos. Mag. 1902, 4, 396–402. (9) Ebbesen, T. W.; Lezec, H. J.; Ghaemi, H. F.; Thio, T.; Wolff, P. A. Nature 1998, 391, 667–669. (10) Brolo, A. G.; Gordon, R.; Leathem, B.; Kavanagh, K. L. Langmuir 2004, 20, 4813–4815. (11) Dostalek, J.; Homola, J. Sens. Actuators, B 2008, 129, 303–310. (12) Eftekhari, F.; Escobedo, C.; Ferreira, J.; Duan, X.; Girotto, E. M.; Brolo, A. G.; Gordon, R.; Sinton, D. Anal. Chem. 2009, 81, 4308–4311.
Analytical Chemistry, Vol. 82, No. 5, March 1, 2010
2087
be excited at normal incidence.13 These thin films transmit a significant fraction of light in the visible range (also without the holes), and the plasmon resonance appears as a peak in the extinction spectrum. This is in contrast to the phenomenon of enhanced transmission in perforated optically thick metal films, for which dips in the extinction spectrum are typically observed (peaks in the transmission spectrum).14 The decay length of the dominant plasmonic field associated with short-range ordered nanoholes has been shown to be around 1 order of magnitude lower (typically tens of nanometers)15,16 compared with that of SPR based on gratings, prisms or periodic nanohole arrays (typically hundreds of nanometers).2,10,11 In fact, the average field penetration depth of short-range ordered nanoholes is similar to that of localized surface plasmons associated with metal nanoparticles,17 and the two concepts provide very similar sensing capabilities.13,18 The sensitivity in peak position to changes in bulk RI is generally higher for SPR sensors based on prisms, gratings, or periodic hole arrays compared with nanoplasmonic sensors based on short-range ordered nanoholes or nanoparticles.2,10,19 However, the short decay length associated with the latter two makes a thin molecular layer occupy a larger fraction of the plasmonic field. This, in turn, makes the sensitivity to changes in interfacial RI comparable.18 Furthermore, while similar signals are induced upon adsorption of a biomolecular layer, the lower bulk RI sensitivity of nanoplasmonic sensors makes them less sensitive to variations in temperature and other potentially disturbing fluctuations of the bulk solution outside the adsorbed layer. Combined with recent development of read-out techniques from sub 100 µm2 nanoplasmonic areas with a performance equally good or better than that obtained using imaging SPR,20 this makes nanoplasmonics ideally suited for array-based labelfree screening.21 For detection of low abundant target molecules, a common limiting factor is the rate of binding, which, in turn, is highly dependent on the liquid handling. The reason is that upon biomolecular binding on a surface, the local concentration of target molecules at the interfacial region becomes depleted. Hence, for mass transport limited reactions, which is normally the case when measuring under stagnant conditions, the rate of binding will be determined by diffusion across an indefinitely growing depletion zone.22 By instead flowing the target solution parallel to the sensor surface, the flow will compete with the growth of the depletion zone until steady state is reached. The extension of the depletion zone decreases with increasing flow rates, which, in turn, leads (13) Prikulis, J.; Hanarp, P.; Olofsson, L.; Sutherland, D.; Kall, M. Nano Lett. 2004, 4, 1003–1007. (14) Genet, C.; Ebbesen, T. W. Nature 2007, 445, 39–46. (15) Jonsson, M. P.; Jo ¨nsson, P.; Ho ¨o ¨k, F. Anal. Chem. 2008, 80, 7988–7995. (16) Rindzevicius, T.; Alaverdyan, Y.; Dahlin, A.; Ho¨o ¨k, F.; Sutherland, D. S.; Ka¨ll, M. Nano Lett. 2005, 5, 2335–2339. (17) Haes, A. J.; Zou, S. L.; Schatz, G. C.; Van Duyne, R. P. J. Phys. Chem. B 2004, 108, 109–116. (18) Jonsson, M. P.; Dahlin, A. B.; Jo¨nsson, P.; Ho ¨o ¨k, F. Biointerphases 2008, 3, FD30–FD40. (19) Jonsson, M. P.; Jo ¨nsson, P.; Dahlin, A. B.; Ho ¨o ¨k, F. Nano Lett. 2007, 7, 3462–3468. (20) Dahlin, A. B.; Chen, S.; Jonsson, M. P.; Gunnarsson, L.; Kall, M.; Hook, F. Anal. Chem. 2009, 81, 6572–6580. (21) Endo, T.; Kerman, K.; Nagatani, N.; Hiepa, H. M.; Kim, D.-K.; Yonezawa, Y.; Nakano, K.; Tamiya, E. Anal. Chem. 2006, 78, 6465–6475. (22) Squires, T. M.; Messinger, R. J.; Manalis, S. R. Nat. Biotechnol. 2008, 26, 417–426.
2088
Analytical Chemistry, Vol. 82, No. 5, March 1, 2010
to increased binding rates until the reaction finally becomes limited by the reaction itself. However, the improved binding rates generally occur at the expense of capturing a smaller fraction of the target molecules that flow over the sensor surface.22 This drawback may be overcome by using sufficiently small (nanofluidic) channels, since for such systems the depletion zone may be thin but still comparable to the dimensions of the channel, thus enabling both high binding rates and high capture efficiency.23 Besides the liquid handling, the limit of detection in terms of concentration of target molecules in a sample solution is determined by the lowest surface coverage that can be detected, which in turn depends directly on the signal-to-noise ratio of the sensor. However, readout from single nanoscale devices is generally connected with high noise. Hence, to utilize the potential advantages of nanofluidics for sensing applications, while preserving a high signal-to-noise ratio, an attractive solution is a network of multiple nanoscale channels that can be probed in parallel and where each nanochannel contains a sensing element. In this paper we present such a concept based on nanoplasmonic pores that penetrate an optically thin multilayer membrane of gold and silicon nitride (SiN, see schematic illustration in Figure 1). The structure provides a two-dimensional parallel network of fluidic nanochannels with liquid access on both sides. To localize the sensitivity to changes in interfacial RI (induced by biomolecular binding reactions) to the inside of the pores, they have been arranged in a short-range order.24,25 In addition, a material-specific surface modification scheme for gold and SiN was developed and utilized to enable the investigation of specific biological recognition of target biomolecules that flow through the pores. MATERIALS AND METHODS Materials and Chemicals. If nothing else is stated, materials were purchased from Sigma-Aldrich, Germany. Si wafers were purchased from Si-Mat, Germany. PEG grafted to poly-L-lysine (PLL-g-PEG) was purchased from SuSoS AG, Switzerland. Thiol modified poly(ethylene glycol) (thiolPEG) and biotinylated thiolPEG (thiolPEGbiotin) were bought from Rapp Polymere GmbH, Germany. Fabrication of Nanoplasmonic Pores. A detailed description of the fabrication protocol is presented in the Supporting Information. In brief, a 65 nm thick gold film perforated with 150 nm in diameter short-range ordered nanoholes was fabricated on a SiN coated (200 nm) 2 in. silicon (Si) wafer using conventional colloidal lithography (Figure 1i-iv).26 Colloidal lithography was chosen both because it distributes the pores in a short-range order and also because it is a simple, rapid, and low-cost method that can be used to produce nanostructures on large areas. The nanoholes were subsequently coated by an additional 200 nm SiN layer (v), which protects the nanohole structure during subsequent wet etching. In addition, either this or the first SiN film was used to increase the mechanical stability of the final structure. Open squares determining the size of the membranes were subse(23) Schoch, R. B.; Cheow, L. F.; Han, J. Nano Lett. 2007, 7, 3895–3900. (24) Dahlin, A. B.; Jonsson, M. P.; Ho ¨o ¨k, F. Adv. Mater. 2008, 20, 1436–1442. (25) Rindzevicius, T.; Alaverdyan, Y.; Sepulveda, B.; Pakizeh, T.; Ka¨ll, M.; Hillenbrand, R.; Aizpurua, J.; de Abajo, F. J. G. J. Phys. Chem. C 2007, 111, 1207–1212. (26) Hanarp, P.; Sutherland, D. S.; Gold, J.; Kasemo, B. Colloids Surf., A 2003, 214, 23–36.
Figure 1. Schematic image illustrating the fabrication of sensor chips with membranes penetrated by short-range ordered nanoplasmonic pores. First, a perforated thin metal film is produced by colloidal lithography on a Si wafer coated with SiN (i-iv), as shown at the top. The nanostructure is then coated with a second SiN layer (v), which will act as protection during wet etching of Si. UV-lithography is used to define areas where to remove Si (vi), which is performed in the next step using wet etching in TMAH (vii). In the last step the nanoplasmonic holes are transformed into pores by etching the SiN with RIE. The gold itself acts as an etch mask in this step. The RIE can be performed either from the front side (viii a) or the backside (viii b) of the wafer, where the latter enables the gold to be accessible only in the regions of the membranes. Note that the schematic illustration is not to scale.
quently defined in a photosensitive protection layer (ProTEK PSB23, Brewer Science, USA) on the backside of the wafer by UVlithography (vi). Around the squares, thin slits were made to define the dimensions of the sensor chips (around 5 × 5 mm2). Silicon in all open areas could then be removed by wet etching in tetramethyl-ammonium-hydroxide (TMAH), while areas coated by the protection layer were not attacked (vii). The metal nanohole structure on the frontside of the wafer was protected by the top SiN coating. Wafer-through etching was completed after around 13 h in TMAH, and the SiN film initially deposited directly on the Si wafer acted as etch stop. This resulted in squares of thin free-hanging SiN/gold nanohole/SiN membranes supported by the surrounding Si wafer. Because Si was removed also in the slits around the squares, the wafer could be easily divided into approximately 50 sensor chips with conventional tweezers. The final step included the critical opening of the nanoholes into plasmonic pores, for which reactive ion etching (RIE) was chosen. RIE was applied either from the frontside (viii a) or the backside (viii b) of the wafer. The main difference between these approaches is that when etching is applied from the backside, the pores are opened in the membrane area only. Hence, the gold region becomes accessible solely in the membrane region. For both approaches the gold film itself was used as etch mask during RIE. Optical Measurements. Extinction spectra of the samples were acquired using a conventional microscope equipped with a 100 W quartz tungsten halogen light source (Newport, USA) and a back-thinned 2D CCD spectrometer (QE65000, OceanOptics, USA). The spectrometer was controlled with a custom designed LabView program (LabVIEW 7.1, National Instruments, USA). First, a dark spectrum (Idark(λ)) was recorded without illumination. This
was followed by recording a reference spectrum (Iref(λ)), and, finally, the sample was placed in focus and the extinction spectrum could be acquired and displayed according to
(
E(λ) ) 100 × 1 -
Isample(λ) - Idark(λ) Iref(λ) - Idark(λ)
)
where Isample(λ) is the intensity acquired through the sample. This thus gives the extinction in percentage of light that is not transmitted through the sample. For sensing experiments, a water immersion objective with 63 times magnification was used, where the immersion droplet was used as one of the two liquid compartments in contact with the nanoplasmonic pores. A reference spectrum was taken before measurements using the same objective and a water droplet on a microscope slide. Spectra were acquired from an around 3 × 16 µm2 sized area (corresponding to around 320 pores), fitted to a polynomial, and the centroid (center of mass) of the peak was calculated and plotted in realtime using a custom designed LabVIEW program. In this setup, a further reduction of the measurement area leads to long-term drifts and additional noise due to mechanical vibrations and insufficient light capturing.20 Quartz Crystal Microbalance Measurements. QCM measurements were performed using an E4 instrument from Q-Sense AB, Sweden. The QCM crystals were also provided by Q-Sense. To imitate the SiN surface of the nanoplasmonic structure, a SiO2coated QCM crystal was coated with the same type of SiN as used for the sensor chips. RESULTS AND DISCUSSION Particular efforts were devoted to the development of a parallel fabrication scheme, which will significantly facilitate the realization Analytical Chemistry, Vol. 82, No. 5, March 1, 2010
2089
Figure 2. (A) SEM images taken at an angle of 30 degrees (from normal incidence) of a sensor chip made by RIE from the frontside. The images were taken from the backside of the membrane showing that the pores are etched through the membrane. (B) SEM images taken at an angle of 30 degrees of a sensor chip made by RIE from the backside of the wafer. These images were taken from the frontside and show pores that are etched through. Note that the gold nanoholes were discernible also through the SiN coating. This was observed also for sensor chips that were not etched with RIE from any side, as shown in the inset to the right in (B).
of practical applications of the concept. Not only were the nanopores made all at the same time but also the method enabled multiple sensor chips (currently more than 50) to be fabricated in parallel. As described above, this was made possible by defining nanoholes in gold on a whole SiN coated wafer using colloidal lithography. This was followed by conventional UV-lithography and etching steps to define small sensor chips with thin membranes in gold and SiN. Finally RIE was used to create the pores through the thin membranes (see schematic illustration of the fabrication scheme and the resulting device in Figure 1). The successful fabrication of nanopores was verified by scanning electron microscopy (SEM). The structure shown in Figure 2A was etched with RIE for 2 min and 30 s from the frontside. The images were taken from the backside (opposite side from the RIE) and clearly demonstrate that the pores were etched all the way through the SiN membrane. The sensor chip shown in Figure 2B was instead etched from the backside for 2 min and 30 s, again demonstrating through-going nanopores when imaged from the frontside. In Figure 2B it is also clear that the pores were not etched outside the thin membrane area. A comparison between Figure 2A,B indicates that the SiN membrane is thinner for devices etched with RIE from the backside, attributed to the fact that the top SiN layer is etched around 70 nm during the 13 h long Si etch in TMAH.27 From the SEM images it is also apparent that the pores are slightly conical, with an angle of around 25° (where 0° corresponds to cylindrical pores). However, although this is likely to influence the detailed optical properties in terms of absolute peak position, RI sensitivity, etc., the degree of tapering is not expected to significantly affect the qualitative sensing (27) Yan, G. Z.; Chan, P. C. H.; Hsing, I. M.; Sharma, R. K.; Sin, J. K. O.; Wang, Y. Y. Sens. Actuators, A 2001, 89, 135–141.
2090
Analytical Chemistry, Vol. 82, No. 5, March 1, 2010
Figure 3. (A) Spectra in air for a typical sensor chip before and after 2 min and 30 s RIE from the frontside. (B) Spectra for a typical sensor chip before and after 2 min and 30 s RIE from the backside. (C) Optical images taken from the backside of a wafer with an array of 1225 membranes. The scale bars are 2 and 1 µm for the large image and the inset, respectively. (D) Histogram together with a Gaussian fit showing the variation in nanoplasmonic peak position along one of the diagonals of the wafer in (C).
capability of the nanoplasmonic structure and may even be an advantage in terms of guiding molecules into the pores. Characterization of the Optical Properties of the Nanoplasmonic Pores. The plasmonic properties of the sensor devices were investigated by probing sub 100 µm2 regions of the perforated membranes by microextinction spectroscopy.20 Spectra of a typical sensor chip before and after 2 min and 30 s RIE from the frontside are shown in Figure 3A. A small peak is observed already before RIE and becomes significantly pro-
nounced after completed RIE. It is worthwhile to note that the peak position is red-shifted after opening the pores. This might at first sight appear surprising considering the fact that removal of SiN and the corresponding decrease in the average RI around the nanostructure is expected to result in a significant blue shift (in the order 200 nm assuming a bulk RI sensitivity of 200 nm per refractive index unit, RIU). However, because the gold film is used as etch mask, the thickness of the gold film will decrease during RIE. This, in turn, is expected to result in a corresponding red-shift of the plasmon resonance.28 The parameters for the RIE process were optimized for high selectivity between SiN and gold (19:1), with etch rates of around 230 nm/min and 12 nm/min, respectively. The gold film was exposed to RIE for around 116 s (the time left after removing the top coating of 130 nm SiN) and was thereby etched approximately 23 nm. For nanoplasmonic holes of similar dimensions, this decrease in gold thickness is expected to give a red-shift in the plasmon resonance of around 230 nm.28 The expected shift in the plasmon resonance originating from removing the SiN and thinning of the gold film is thus of a few tens of nanometers to the red part of the spectrum, which is in agreement with the results shown in Figure 3A. Figure 3B shows similar results for a sensor chip etched from the backside for 2 min and 30 s, with the exception that the peak position is not red-shifted as in Figure 3A but instead slightly blueshifted. In this case, a 200 nm SiN layer must be removed before the gold is exposed to RIE, which means that the gold will be etched approximately 20 nm instead of 23 nm. Hence, the decrease in gold thickness is smaller than for the sensor chip etched from the frontside, which gives a correspondingly smaller red-shift in the plasmon resonance. It should be noted that these are rough estimates, and differences in sensitivity between the structures might also play a role, as shown below. Apart from enabling many sensor chips with single membranes to be fabricated in parallel, the fabrication scheme also provides the possibility to create over 1000 individual nanoplasmonic membranes simultaneously on one sensor chip. The fabrication method was evaluated with respect to success rate and variation in optical properties by producing an array of 35 × 35 nanoplasmonic membranes on a single wafer (see Figure 3C). Out of in total 1225 membranes made simultaneously, only around ten (
