REVIEW OF SCIENTIFIC INSTRUMENTS
VOLUME 70, NUMBER 12
DECEMBER 1999
High resolution surface plasmon resonance spectroscopy N. J. Tao,a) S. Boussaad, W. L. Huang, R. A. Arechabaleta, and J. D’Agnese Department of Physics, Florida International University, Miami, Florida 33199
共Received 30 June 1999; accepted for publication 31 August 1999兲 A method for detecting surface plasmon resonance with high resolution (⬃10⫺5 degrees or ⬃10⫺8 refractive index units兲 and fast response time 共1 s兲 is described. In the method, light is focused through a prism onto a metal film on which molecules to be detected are adsorbed. The total internal reflection of the incident light is collected with a bicell photodetector instead of a single cell or an array of photodetectors that are widely used in previous works. The ratio of the differential signal to the sum signal of the bicell photodetector provides an accurate measurement of shift in surface plasmon resonance angle caused by the adsorption of molecules onto the metal films or by conformational changes in the adsorbed molecules. Using the method, we have studied subtle conformational changes in redox protein, cytochrome c, due to an electron transfer reaction. © 1999 American Institute of Physics. 关S0034-6748共99兲02212-1兴
intensity change in the reflection due to SPR angular shift.6–8 A major advantage of this approach is that the response time is only limited by the photodetector and the associated electronics which can be as fast as nanoseconds.8 A drawback, however, is that the relationship between the intensity and the resonance angle is sensitively dependent on the angle at which the photodetector is fixed. Major limitations in the resolution of the method come from the intensity fluctuation in the laser and from thermal and mechanical drift in the setup. Another widely used ATR-based method is to replace the collimated incident light in the above setups with a convergent beam that covers a range of incident angles. The reflections from different incident angles are collected simultaneously with a linear diode array 共LDA兲 or charge coupled device 共CCD兲.9,10 This method involves no mechanical movements, but simultaneous detection of many channels 共e.g., 1024 in a typical LDA兲 slows down the response time. The typical angular resolution is 10⫺2 – 10⫺3 deg or 10⫺5 – 10⫺6 RIU. As in the method with a rotating prism, high angular resolution of this method requires a large distance between the prism and the photodetector. The above setups involve reflection intensity versus incident angle 共an angle-scan system兲; SPR has also been detected by modulating the wavelength of incident light.11,12 The wavelength modulation causes modulation in the reflection intensity which is monitored with a lock-in amplifier and provides an accurate measurement of the SPR dip position. Using an acousto-optic tunable filter 共AOTF兲, it was demonstrated that a wavelength change of 0.0005 nm,11 corresponding to 5⫻10⫺7 RIU at a wavelength of 630 nm,1 can be detected. When applied to DNA-SH adsorption on gold, the signal to noise ratio of the AOTF SPR is six times better than that achieved by an angle-scan system.12 We describe here a new SPR detection method that has allowed us to achieve an angular resolution of 10⫺5 deg 共or 10⫺8 RIU兲 and time response of 1 s. The method has several additional features which include simplicity, good lin-
I. INTRODUCTION
Surface plasmon resonance 共SPR兲 spectroscopy has emerged as a powerful technique in recent years for detection and analysis of chemical and biological substances in many research areas and industrial applications, such as surface science, biotechnology, medicine, environment, and drug and food monitoring.1,2 In all these applications, improving the resolution and time response of SPR detection is of vital importance and is the primary goal of the present work. The widely used methods for detecting SPR are based on attenuated total reflection 共ATR兲 in a prism on which a thin metal film is coated.3,4 In the methods, p-polarized light is incident upon the metal film and the total internal reflection of the incident light is detected. When the incident light reaches an appropriate angle the reflection decreases sharply to a minimum, corresponding to the excitation of surface plasmon waves in the film. The total internal reflection has been detected with a photodetector as a function of incident angle by rotating both the prism and the photodetector.5 The angular resolution achieved by this rotating prism approach is typically 10⫺2 – 10⫺3 deg 共degrees兲, limited by errors in the angular position and noise in the intensity of the reflected light. For comparing different SPR detection techniques, the SPR resolution is often described in terms of the smallest detectable change in the refractive index of an analyte 关refractive index units 共RIU兲兴. The above angular resolution corresponds to 10⫺5 – 10⫺6 RIU at a wavelength of 630 nm. For higher angular resolution, a large distance between the prism and the photodetector is required which makes the setup not only bulky but also more susceptible to mechanical noise and thermal drift. The response time is slow because of the mechanical movements in the setup. Mechanical movements can be avoided by fixing the photodetector at an angle near resonance and measuring the a兲
Author to whom correspondence should be addressed; electronic mail:
[email protected]
0034-6748/99/70(12)/4656/5/$15.00
4656
© 1999 American Institute of Physics
Rev. Sci. Instrum., Vol. 70, No. 12, December 1999
FIG. 1. 共a兲 Schematic diagram of the SPR setup. 1—Prism. 2—Supporting frame. 3—Glass slide. 4—Solution port. 5—N2 port. 6—Solution cell. 7—Glass window. 8—Adsorbed molecules to be detected. 9—Counterelectrode. 10—Solution port. 11—Reference electrode. 12— Bicell photodetector. 13—Lens. 14—Diode laser. 15—Metal film. 共b兲 Intensity profiles of the two cells before and after a shift in the SPR. The intensity of the two cells, A and B, is first balanced 共area under the solid line兲. A shift in the SPR results in an intensity imbalance of the two cells 共dashed lines兲 which is detected as differential signal of the photodetector.
earity, compactness, and immunity to ambient light. The method uses a convergent beam focused onto a thin metal film, but the total internal reflection is collected by a bicell photodetector instead of a CCD or a LDA 关Fig. 1共a兲; a detailed explanation is given in Sec. II兴. The reflected light falling on the two cells of the photodetector is first balanced so that the SPR dip is located near the center of the photodetector. A small shift in the dip position causes an imbalance in the two cells which is detected as a change in the differential signal of the bicell photodetector 关Fig. 1共b兲兴. Because the differential signal is linearly proportional to the shift in the SPR angle and can be easily amplified without saturation problem, it provides an accurate detection of SPR. We note that a bicell photodetector has been used in the atomic force microscope 共AFM兲 in which the deflection of a laser beam due to bending of the AFM cantilever is measured.13,14 In the present application, it is the intensity distribution due to a SPR angular shift rather than physical movement of the laser beam that is measured 关Fig. 1共b兲兴. II. EXPERIMENT
In the SPR setup 关Fig. 1共a兲兴, a BK7 plano-cylindrical lens 共Melles Griot兲 was used as a prism. The prism is close to but not exactly hemicylindrical. On the prism, a 50 nm thick gold film evaporated on a BK7 glass slide in ultrahigh
Plasmon resonance spectroscopy
4657
vacuum was placed with an index matching fluid. The gold film was annealed in a hydrogen flame briefly before each experiment in order to reduce surface contamination. A 5 mW diode laser 共⫽635 nm, Hitachi兲, driven with a homemade laser controller, was collimated and then focused by a 14 mm local-length lens through the prism onto the gold film. Light reflected from the gold film was detected with a bicell photodiode detector 共Hamamatsu Corp., model S272102兲 which was mounted on a precision translation stage. The photocurrents from the two cells (A and B) were converted to voltages with a homemade circuit. The circuit also calculated the differential, A⫺B, and the sum, A⫹B, signals which were then sent to a PC computer equipped with a 16-bit data acquisition board 共National Instrument兲. For fast kinetic studies, the differential and sum signals were sent to a 150 MHz digital oscilloscope 共Yokogawa, DL1520L兲. Before each measurement the prism was rotated so that there was a dark line located at the center of the laser beam. The dark line is due to the absorption of the light by the surface plasmon which occurs at the angle of resonance. The reflected light falling onto the two cells of the photodetector was then balanced by adjusting the photodetector position with the translation stage until A⫺B approached zero. Because of the high sensitivity of the method, drift in the A⫺B signal due to mechanical stress was clearly visible immediately after alignment but it settled down typically over a period of 15–30 min when all the screws were properly tightened. The ratio of the differential to sum signals, which is linearly proportional to the SPR angular shift, was obtained numerically by dividing A⫺B with A⫹B. On the gold film a Teflon sample cell was mounted to hold sample solutions. The cell has two ports for flowing sample solutions in and out, and a port for purging O2 out of the solutions with N2, which is necessary for many experiments. To control the electrochemical potential of the gold film electrode, Pt and Ag wires were used as counter- and quasireference electrodes, respectively. The quasireference electrode was calibrated against a Ag/AgCl reference electrode. The electrochemical potential of the gold film was controlled with an EG&G model 283 potentiostat. III. RESULTS AND DISCUSSION A. Linearity
The ratio of the differential to sum signals (A⫺B)/(A ⫹B) is linearly proportional to the SPR angular shift, ⌬, over a given angular range, as shown by both theoretical simulation 共Fig. 2兲 and experimental calibration 共Fig. 3兲. The simulation was performed in the framework of Fresnel optics using a matrix method.15 Three phases, corresponding to the BK7 prism, 50 nm thick gold film, and aqueous solution, were assumed. We have carried out the simulation including an additional phase, adsorbed molecular layer, and found no significant difference. The indexes of refraction for the prism and the solution were taken to be 1.515 and 1.33, respectively, and the dielectric constant for gold film was ⫺12 ⫹1.26i. The wavelength of light was 635 nm, and the incident beam had a convergent angle of 3°. As shown in Fig. 2, (A⫺B)/(A⫹B) is linear for ⌬ ⬍0.7 deg, which is large
4658
Tao et al.
Rev. Sci. Instrum., Vol. 70, No. 12, December 1999
FIG. 2. Theoretical simulation of (A⫺B)/(A⫹B) vs SPR dip position. See the text for details of the simulation.
enough for proteins, such as monolayer cytochrome c 共cyt. c) used in the present study. For larger proteins, the linear range can be increased by simply increasing the convergent angle of the incident beam. The linear relationship between (A⫺B)/(A⫹B) and ⌬ can be understood by the following analysis. Let us denote the reflectivity versus incident angle as R(⌬ ). For a small ⌬, R(⌬ )⫽R(0)⫹ 关 dR(0)/d⌬ 兴 ⌬ ⫹ 21 关 d 2 R(0)/d⌬ 2 兴 ⫻(⌬ ) 2 ⫹ . . . . The first term is small because the reflectivity at the resonant angle approaches zero, and the second term is zero because the reflectivity at the resonant angle is minimum. By keeping the first nonzero term, the quadratic term, one finds that (A⫺B)/(A⫹B)⫽6⌬ / 0 , where 0 is the convergence of the incident light. This relation tells us clearly that the angular resolution is inversely proportional to the angular range. For 0 ⫽3°, the slope of the (A⫺B)/ (A⫹B) versus ⌬ plot is 2, which is greater than 1.6 found from the numerical simulation. This discrepancy is due to ignoring higher order terms. We have compared the SPR angular shift of a gold film measured by the present setup with that by a conventional LDA SPR setup. The shift in the SPR angle was induced by varying the electrochemical potential of the gold film in phosphate buffer solution with respect to the reference electrode in the solution. It has been found that the potential changes the SPR angle by changing the electron density in the gold film.16 The SPR shift by the present setup was obtained from (A⫺B)/(A⫹B), and the shift by the LDA
FIG. 3. Experimental calibration of the SPR dip position of a gold film in a phosphate buffer at various potentials measured by the present method vs the dip position of the same sample measured with a conventional linear diode array setup, showing excellent agreement between the two methods.
FIG. 4. SPR resolution of the present setup. The noise is primarily due to thermal drift and mechanical vibration in the setup. The data were obtained with a MPA-coated gold film in 50 mM phosphate buffer.
method was determined by fitting the dip position at various potentials. As shown in Fig. 3, the SPR shifts by the two methods are in excellent agreement. B. Angular resolution
The SPR angle of a gold film in 0.1 M phosphate buffer over a period of ⬃15 min as determined from (A⫺B)/ (A⫹B) is plotted in Fig. 4. The root mean square 共rms兲 of the fluctuation is about 5⫻10⫺6 deg or 3⫻10⫺8 RIU. To the best of our knowledge, this angular resolution is an order of magnitude better than the reported resolution achieved with other angle-scan systems.17 Errors in the ATR-based SPR methods come mainly from three sources: laser intensity fluctuation, mechanical vibration and thermal drift, and noise in the photodetector and its electronics. The intensity fluctuation in a typical diode or HeNe laser is between 0.1% and 1% which is a serious source of errors in the methods based on detecting the ATR intensity. This problem is greatly reduced in the present setup because the common mode noise in the laser intensity is subtracted out in the differential signal. The measurement of the ratio of the differential to the sum signals further reduces errors due to the intensity fluctuation. The detection of the differential signal also makes the present setup largely immune to noise due to ambient background light. In fact, all the measurements reported in the article were done with the setup exposed to room light. These advantages are not shared by the methods using either a single cell or an array of photodetectors. Noise in the photodetector and its electronics is another source of errors in all the ATR-based methods. In our present setup, the noise in A⫺B due to the photodetector and electronics has a rms of ⬃1 uV, corresponding to an angular resolution of ⬃10⫺7 deg (10⫺9 RIU兲 for a 5 mW diode laser that gives a A⫹B of 3–5 V, so the resolution achieved here is clearly not limited by this noise. The last source of errors comes from mechanical vibrations and thermal drift in the systems and depends on the design of each setup. This is the dominant source of errors in the present setup. We have minimized this error by placing
Rev. Sci. Instrum., Vol. 70, No. 12, December 1999
Plasmon resonance spectroscopy
4659
the laser diode, prism, and bicell photodetector close together on an optical breadboard, which was placed on a vibration-isolated plate used for a scanning tunneling microscope. Further improvement can be made by incorporating all the components into a single piece made of low thermal expansion material and by controlling the temperature of the piece. This is easily achievable because the setup contains only three basic elements, the laser, the prism, and the bicell photodetector. We note that, to achieve a 10⫺5 deg angular resolution using a LDA or a CCD detector, the distance between the sample and the detector has to be large. For a LDA or CCD that has a pixel size of 5 m, a distance of 30 m would be needed! Noise due to thermal drift and mechanical vibrations over such a large distance is much harder to reduce. C. Response time
The response time of the present setup achieved is ⬃1 s, and is limited by the preamplified bandwidth of the homemade electronic circuit. The bicell photodetector itself has a response time of ⬃10 ns, so further improvement can be made. A SPR response time of ns has been reported in the study of laser ablation of organic absorbates.8 In the setup, a single photodetector was used to detect reflectivity change due to desorption of molecular adsorbates. The fast time response was achieved because of the use of an ac coupled photodetector. Such an ac coupled photodetector can also be used in the present setup to speed up the response time to ns. The method that uses a LDA or CCD has a typical response time of ms which is much slower than either the single cell or bicell photodetector methods. The response time of the AOTF method is limited by the response time of the photodetector as well as by the modulation frequency of the AOTF and the bandwidth of the lock-in amplifier.12 D. Examples
Using high angular resolution SPR, we have studied electron transfer induced conformational changes in cyt. c, a redox protein whose main function is to shuttle electrons between membrane bound enzymes in the mitochondrial respiratory chain.18 A basic question that is directly related to the electron transport function of this important protein is whether there is a large conformational difference between its oxidized and reduced states.19–23 Experiments such as and small-angle x-ray scattering hydrodynamic21 measurements23 detected large differences between the two states which were attributed to a large conformational difference. However, with x-ray crystallography24 and nuclear magnetic resonance 共NMR兲 spectroscopy25 only small structural differences between the two states have been observed. Directly adsorbing proteins on bare metal surfaces often results in denaturation. So a monolayer of mercaptopropionic acid 共MPA兲 self-assembled on the gold film was used as a cushion to prevent cyt. c from denaturation.26,27 The SPR angle versus potential and the simultaneously recorded cyclic voltammogram 共CV兲 共inset, Fig. 5兲 of the MPA-coated gold film in 200 L 50 mM phosphate buffer 共pH⫽7.0兲 are shown in Fig. 5共a兲. The shift in the SPR angle can be attrib-
FIG. 5. 共a兲 SPR dip shift of a MPA-coated gold electrode in 50 mM phosphate solution as the electrode potential is scanned between ⫺0.2 and 0.3 V at a rate of 0.1 V/s. The inset shows the simultaneously recorded cyclic voltammogram. 共b兲 SPR dip shift as a function of time before, during, and after cyt. c was introduced into the solution cell. The two arrows mark the moments when the cyt. c was introduced and then replaced with buffer, respectively. 共c兲 SPR dip shift of cyt. c immobilized on a MPA-coated gold electrode in 50 mM phosphate buffer as the electrode potential was scanned between ⫺0.2 and 0.3 V at a rate of 0.1 V/s. The inset shows the simultaneously recorded cyclic voltammogram in which a pair of peaks corresponds to an electron transfer reaction of the protein. The dashed lines in 共c兲 indicate changes due to double layer charging.
uted to a change in the electron density of the gold film.16 The electron density induced SPR shift observed here is about 0.001°/100 mV, which is much smaller than that for a bare gold electrode.16 This is most likely because the presence of MPA decreases the surface capacitance and therefore the electron density change for a given potential change. We note that the noise in the SPR angle versus potential plot came largely from the noise in the electrochemical potential. After characterizing the MPA-coated gold electrode in buffer, 20 L cyt. c solution 共27 M cyt. c in 50 mM phosphate buffer兲 was introduced into the solution cell and the subsequent change in the SPR angle was monitored as a function of time 关Fig. 5共b兲兴. The angle increased and reached a stable value in about 15 min. Replacing the cyt. c solution with buffer solution did not change the angle much, showing that the change was not due to a change in the solution index of refraction and that the adsorbed cyt. c layer was stable.26,27 The CV of the adsorbed cyt. c exhibits a pair of peaks corresponding to the well known electron transfer reaction
4660
Tao et al.
Rev. Sci. Instrum., Vol. 70, No. 12, December 1999
involving Fe3⫹↔Fe⫹2 关inset, Fig. 5共c兲兴. The coverage of the cyt. c estimated from the peak area was (6⫾2) ⫻10⫺12 cm2, which is close to complete monolayer coverage. The simultaneously measured SPR angle shows a sigmoidal decrease as the protein was switched from the oxidized to the reduced states. The change was reversible when switching the protein back to the oxidized state. Before and after the sigmoidal change near the redox potential, a small linear shift in the resonant angle was visible 关dashed lines in Fig. 5共c兲兴. This shift is believed to be a residual effect that is observed for the MPA-coated gold electrode in buffer solution. Subtracting this effect, the electron transfer induced change in the resonant angle is about 0.006°. The SPR shift can be attributed to a conformational change in the protein induced by the electron transfer reaction. This change can change both the average thickness and the index of refraction of the protein layer which cannot be determined based on a SPR angular shift at a given wavelength alone. For a rough estimate, we can use the Lorentz– Lorentz relation28 which relates the change in the index of refraction (⌬n) of the protein layer to the thickness ⌬d. Using the relation, we have estimated that a 0.006° shift in the dip angle due to electron transfer corresponds to an ⬃0.3 Å change in the thickness. The change is rather small which does not support the conclusion based on hydrodynamic and small angle x-ray measurements22 but is in good agreement with the results of x-ray crystallography24 and NMR studies.25 IV. DISCUSSION
We have demonstrated simple high resolution SPR spectroscopy using a differential detection technique. The angular resolution and response time achieved are 10⫺5 deg (10⫺8 RIU兲 and 1 s, respectively. The setup has good linearity, and is compact and immune to noise due to ambient light. Using the setup, we have studied subtle conformational changes due to the electron transfer reaction in cyt. c immobilized on a solid electrode.
ACKNOWLEDGMENTS
Financial support through grants from the NSF 共CHE9818073兲, AFSOR 共F49620-96-1-0346兲, and the donors of the Petroleum Research Fund 共33516-AC5兲 administered by the ACS is acknowledged. 1
For a recent review see, for example, J. Homola, S. S. Yee, and G. Gauglitz, Sens. Actuators B 54, 3 共1999兲. For a recent review, see, for example, A. G. Frutos and R. M. Corn, Anal. Chem. 70, 449A 共1998兲. 3 E. Kretschmann, Z. Phys. 241, 313 共1971兲. 4 A. Otto, Z. Phys. 216, 398 共1968兲. 5 J. G. Gordon II and J. D. Swalen, Opt. Commun. 22, 374 共1977兲. 6 C. Nylander, B. Liedberg, and T. Lind, Sens. Actuators 3, 79 共1982兲. 7 M. M. B. Vidal, R. Lopez, S. Aleggret, J. Alonso-Chamarro, I. Garces, and J. Mateo, Sens. Actuators B 11, 455 共1993兲. 8 S. Herminghaus and P. Leiderer, Appl. Phys. Lett. 58, 352 共1991兲. 9 E. Kretschmann, Opt. Commun. 26, 1 共1978兲. 10 W. Hickel and W. Knoll, Thin Solid Films 199, 367 共1991兲. 11 M. J. Jory, G. W. Bradberry, P. S. Cann, and J. R. Sambles, Meas. Sci. Technol. 6, 1193 共1995兲. 12 F. Caruso, M. J. Jory, G. W. Bradberry, J. R. Sambles, and D. N. Furlong, J. Appl. Phys. 83, 1023 共1998兲. 13 G. Meyer and N. M. Amer, Appl. Phys. Lett. 53, 1045 共1988兲. 14 S. Alexander, L. Hellemans, O. Marti, J. Schneir, V. Elings, P. K. Hansma, M. Longmire, and J. Gurley, J. Appl. Phys. 65, 164 共1989兲. 15 W. N. Hansen, J. Opt. Soc. Am. 58, 380 共1969兲. 16 R. Kotz, D. M. Kolb, and J. K. Sass, Surf. Sci. 69, 359 共1977兲. 17 B. Liedberg, C. Nylander, and I. Lundstrom, Biosens. Bioelectron. 10, R1 共1995兲. 18 R. A. Scott and A. G. Mauk, Cytochrome c: A Multidisciplinary Approach 共University Science Books, Sausalito, CA, 1995兲. 19 E. Margoliash and A. Schejter, Adv. Protein Chem. 21, 113 共1966兲. 20 P. Chin and S. S. Brody, Biochemistry 14, 1190 共1975兲. 21 W. R. Fisher, H. Taniuchi, and C. B. Anfinsen, J. Biol. Chem. 248, 3188 共1973兲. 22 D. Eden, J. B. Matthew, J. J. Rosa, and F. M. Richards, Proc. Natl. Acad. Sci. USA 79, 815 共1982兲. 23 J. Trewhella, V. A. P. Carlson, E. H. Curtis, and D. B. Heidon, Biochemistry 27, 1121 共1988兲. 24 A. M. Berghuis and G. D. Brayer, J. Mol. Biol. 223, 959 共1992兲. 25 Y. Feng, H. Roder, and S. W. Englander, Biochemistry 29, 3494 共1990兲. 26 T. M. Nahir, R. A. Clark, and E. F. Bowden, Anal. Chem. 66, 2595 共1994兲. 27 Z. Feng, S. Imabayashi, T. Kakiuchi, and K. Niki, J. Chem. Soc., Faraday Trans. 93, 1367 共1997兲. 28 C. Kittle, Solid State Physics 共Wiley, New York, 1996兲, p. 390. 2
