Demonstration of a Low-Cost, Single-Molecule Capable, Multimode Optical Microscope

Introduction

Optical measurements are known to provide important information about the physical and chemical properties of materials. Prior to technological breakthroughs in the late 1980s leading to single-molecule (SM) detection, the sensitivity of detectors, such as photomultiplier tubes and photodiodes, dictated optical measurements where averaging over many individual single-molecule optical events was unavoidable. (We use the term “single-molecule detection” to encompass the acquisition of optical signals from any object free of ensemble averaging). After advances in detector technology leading to avalanche photodiodes (APDs) and CCDs with high quantum efficiencies, it was recognized that such ensemble averaging obscures very interesting and unusual physics [1–4]. For example, SM surface-enhanced Raman scattering (SERS) [5], photon bunching [6], antibunching [7], spectral diffusion [8], and fluorescence intermittency [9, 10], are just a few of the many discoveries to arise from these advances over the last decade.

During the same period, there has been concomitant interest in the area of nanoscience and nanotechnology. Motivating this has been the push towards smaller electronic components (Moore’s law) [11] as well as fundamental questions about the behavior of matter with sizes in between that of atoms and molecules (the traditional realm of physics/chemistry) and bulk materials. Underlying this has been the realization that mesoscopic materials possess size-dependent properties stemming from quantum-confinement effects.

In order to address both the applied and basic research interests of nanostructures, there has been tremendous effort in the development of nanometer-sized metal and semiconductor systems having novel size- and/or shape-dependent optical and electrical properties [12, 13]. Examples of such systems include semiconductor quantum dots (QDs) or nanocrystals (NCs) made of CdS, CdSe, and PbSe, which behave as “artificial atoms” and possess size-dependent absorption and emission spectra [14]. Other systems of interest include semiconductor nanowires (NWs) with additional size- and shape-dependent properties [15]. In all cases, because of current synthetic limitations preventing the creation of ensembles of identical nanostructures, new approaches for studying such systems individually are needed. As a consequence, many modern optical experiments with nanostructures are often carried out using far-field SM microscopy alone or in conjunction with other techniques, such as atomic force microscopy [16–18].

The purpose of this paper is to describe the assembly, from commercial off-the-shelf components, of a low-cost (~$3300) optical microscope capable of single-molecule detection. Expansion of the microscope’s capabilities includes adding a fiber-based spectrometer to allow the acquisition of fluorescence spectra from specific areas of samples having high quantum efficiencies. It is hoped that such an inexpensive, yet efficient, microscope will provide undergraduates and high school students the means to access optical physics at the single-molecule/nanostructure level [19, 20]. Furthermore, by constructing such a microscope from the ground up, students will benefit from seeing commercial microscopes as more than just black boxes.

Using this low-cost microscope, we have detected the emission of single, 4.2-nm diameter, CdSe/ZnS core/shell QDs; the emission of individual CdSe NWs (7–10-nm diameter, 1–10-mm length) as well as the Brownian motion of 40-nm and 500-nm-diameter dye-doped polystyrene spheres. Examples of images taken in different microscope modes are provided in the text along with a general description of each and its relevance to single-molecule microscopy.

Basic Principles

In this section we discuss the basic principles underlying the ability to detect single fluorophores. Throughout this section we will consider two specific cases, namely, a single CdSe QD (a well-studied QD system) and a single rhodamine 6G (R6G) dye molecule (a common laser dye). The first subsection below provides an estimate of the sample coverage suitable for single-molecule detection. This is followed by expressions for the absorption cross-section, absorption/emission rates, and saturation intensity of a fluorophore. The last two subsections provide estimates for the microscope collection efficiency as well as the predicted signal-to-noise ratio in an experiment. It is hoped that through these “back-of-the-envelope” calculations, students will become cognizant of the physics dictating the feasibility of single-molecule detection.

(A) Sample-Coverage Estimate. In performing single-molecule experiments, it is imperative to reduce the number of molecules or particles in the field of view. This is accomplished by diluting the sample to suitable levels such that, on average, only one or, at most, several fluorophores are present in the microscope excitation region at any given time. Typical dilutions will range from 10^–9 to 10^–12 M. This range of concentrations can be independently established using the calculation described below.

First, we assume a R6G or CdSe QD solution with a concentration of 10^–9 M. In either case, the number of fluorophores present in solution is cN_A where c is the concentration in moles per liter and N_A is Avogadro’s number. Next, if 100 mL of this solution is dropcast onto a 1-in diameter microscope coverslip, the total number of fluorophores deposited is cN_AV; V is the sample volume used. Assuming that the solution spreads uniformly, the idealized surface density of fluorophores is

(1)

where A (5.07 ´ 10⁸ mm²) is the area of the substrate. A nM solution, therefore, results in a nominal surface density of r ~ 119 fluorophores/mm². Although relatively sparse, this coverage represents an upper limit to the single-molecule imaging regime given that typical detection areas range from 10 mm ´ 10 mm to 60 mm ´ 60 mm. If the same calculation is repeated for a pM solution, a surface density of r ~ 0.12 fluorophores/mm² is obtained. This coverage is clearly within the realm of single-molecule detection and helps establish our lower concentration limit.

The actual coverage, however, depends upon many variables; therefore, this calculation only provides a crude estimate to the range of sample concentrations suitable for single-molecule imaging. In practice, determining the correct sample concentration is achieved through trial-and-error, involving successive dilutions of the initial stock. Once appropriate coverages have been achieved, single-fluorophore detection is often established by observing single-step photobleaching and/or fluorescence intermittency (“blinking”). Together with the high sample dilutions, these observations make for compelling arguments supporting the detection of single molecules [4, 16].

(B) Converting Extinction Coefficients to Cross-Sections. In what follows, we establish the absorption cross-section of our fluorophore of interest. The absorption cross-section is a measure of how readily the molecule/particle absorbs light at a given wavelength and is analogous to its molar extinction coefficient. Units of cross section are generally cm² or Å² whereas the extinction coefficient is typically expressed in units of M^–1cm^–1. Because the latter is more commonly reported, we outline below a calculation that allows conversion from one to the other.

Extinction coefficients are often determined through Beer’s Law

(2)

where A is the absorbance or optical density of a sample, e is the stated molar extinction coefficient (M^–1cm^–1), c is the sample concentration (moles per liter) and l is the sample pathlength (cm). This expression can also be expressed as A = log(1/T) = log(I₀/I) where T is the transmission of the sample expressed as a ratio of incident (I₀) to transmitted (I) intensities (T = I/I₀). In parallel, an alternative expression for Beer’s law can be obtained from the exponential attenuation of an incident beam through a sample. The expression is

(3)

where s is the desired absorption cross section (cm²/molecule), n is the sample’s number density (molecules/cm³) and l is the optical pathlength (cm). A simple manipulation of this expression yields

When substituted into the previous expression for Beer’s law (eq 2), one obtains

giving a direct relation between the extinction coefficient and the absorption cross-section. If the concentrations n and c are both expressed using the same units (molecules/cm³), a relationship between the absorption cross-section and the molar extinction coefficient can be established as follows:

(4)

The factor of 1000 originates from a conversion between dm³ and cm³ and 1/log e = 2.303.

To provide concrete examples of this conversion, we consider both our CdSe and R6G solutions. In the case of the QD ensemble, the peak of the first transition in the absorption spectrum is related to the NC size [21]. When the first excited state is centered at 532 nm, and the diameter of the particles is 3.5 nm [22]. This allows us to obtain the QD molar extinction coefficient through size-dependent values tabulated in reference 23. The extinction coefficient of 3.5-nm diameter CdSe QDs at 532 nm is 1.6 ´ 10⁵ M^–1 cm^–1. By comparison, the reported molar extinction coefficient of R6G at 532 nm is 1.14 ´ 10⁵ M^–1 cm^–1[24]. Using eq 4, both extinction coefficients can now be converted to their corresponding absorption cross-sections. The values obtained are s(CdSe) = 6.12 ´ 10^–16 M^–1 cm^–1 and s(R6G) = 4.36 ´ 10^–16 cm². One can see that the cross-sections of both species are similar with the 3.5-nm QD being slightly more absorptive.

(C) Calculating the Experimental Excitation Intensity. We now calculate the excitation intensity of a given experiment. This will subsequently allow us, in conjunction with the above cross-sections, to determine the rate at which incident photons are absorbed by either fluorophore. In our example, we assume the specific case of a confocal diffraction-limited spot having a Gaussian intensity profile. Under these conditions, the full width at half maximum (FWHM) of the excitation spot is l = 2NA where is the wavelength of the light and is the numerical aperture of the microscope objective. NA is a measure of the objective’s light-gathering ability and typically varies from 0.9 to 1.45. In the microscope we have constructed, the objective’s NA is 1.25 resulting in a spot size FWHM of 213 nm. Next, assuming a Gaussian intensity profile,

where r₀ is the center of the excitation and s is its standard deviation, we obtain 1.18s = FWHM/2. The spread of the excitation in the diffraction-limited spot is, therefore,

At this point, to complete the calculation of the excitation intensity, we need the laser power incident on the sample. In practice, this is measured using a power meter positioned at the objective’s output. For the sake of example, however, we assume a measured excitation power of P = 500 nW (532 nm). Together with the spot’s Gaussian profile, we obtain an excitation intensity, I, through the relation

(5)

expressed in plane polar coordinates. Alternatively, this expression can be written in terms of Cartesian coordinates, yielding the same value of I. The solution to either integral yields

(6)

whereupon we solve for I to obtain a peak excitation intensity of 978 W cm^–2. This value (I ~ 1 kWcm^–2) is common to many single-molecule experiments. Its use as a measure of the actual excitation intensity seen by the sample is justified by asserting that the physical size of the fluorophore (~1 to 10 nm) is generally much smaller than the size of the excitation spot (~200 to 500 nm).

(D) Calculating the Rate of Absorption. Provided with the above excitation intensities and absorption cross-sections, we now calculate the rate at which the molecule/particle absorbs photons. As before, we consider the case of a single CdSe QD and a single R6G molecule. The rate of absorbed photons is determined by the product of their absorption cross-sections and their respective excitation intensities divided by the energy per photon. This expression is generally written as

(7)

where ħw is obtained as follows: The energy associated with 532-nm light is 2.54 eV (1240/532 nm = 2.54 eV). Furthermore, the number of Joules per eV is 1.602 ´ 10^–19 J eV^–1 leading to an energy per photon of ħw₅₃₂ = 3.73 ´ 10^–19 J per photon. Substituting this value into eq 7 yields the absorption rates in CdSe and R6G: A₅₃₂(CdSe)= 1.6 ´ 10⁶ photons per second and A₅₃₂ (R6G) =1.14 ´ 10⁶ photons per second.

(E) Calculating the Rate of Emission. The absorption rate must now be connected to the fluorophore emission rate. The relevant expression is

(8)

where the emission rate is simply a product of the absorption rate and the intrinsic quantum yield (QY) of the material, which is a measure of how effectively it converts absorbed light into emitted photons. In the case of ZnS overcoated CdSe QDs, ensemble quantum yields typically have values of QY~0.30 [25]. By contrast, R6G has a very high intrinsic quantum yield that approaches unity, QY = 0.95 [26]. Using this and eq 8, the rate of emitted photons from a single CdSe QD and a single R6G molecule is determined to be 4.8 ´ 10⁵ photons per second and 1.08 ´ 10⁶ photons per second, respectively.

(F) A Saturation Analysis. The signal-to-noise ratio and, correspondingly, the ability to detect single fluorophores improves if the number of photons emitted by the sample increases. The easiest way to achieve this is by increasing the excitation intensity in the experiment; however, such a strategy does not work indefinitely as the onset of saturation leads to an upper limit for the fluorophore emission rate. To illustrate this phenomenon, the saturation of a simple two-level system can be analyzed.

Consider a two-level system with a lower to upper transition rate of g₁₂ = Abs and a corresponding emission rate of g₂₁ = Emm = 1/t_rad; t_rad is the radiative lifetime of the excited state. For the case of our CdSe QD, the lifetime is experimentally found to be on the order of 20 ns at room temperature, t_rad » 20 ns [27], leading to g₂₁ (CdSe) = 5 ´ 10⁷ s^–1. In the case of R6G, t_rad » 4 ns resulting in g₂₁ (R6G) = 2.5 ´ 10⁸ s^–1 [26]. Next, we solve for the kinetics of the two-level system by considering the following rate equations

(9a)

(9b)

where P₁ and P₂ are the probabilities of being in the lower and upper states, respectively and . If we invoke a steady-state approximation, dP₁/dt = dP₂/dt = 0, which leads to g₁₂P₁ = g₂₁P₂. Using the above conservation of probability, this equality can be turned into an expression for P₂

(10)

This equation can be further manipulated by using the relation between the pump rate (g₁₂) and the excitation intensity (I), g₁₂ = A = Is/ħw to obtain an equivalent expression,

A saturation intensity, I_sat = ħwg₁₂/s can now be defined resulting in

With this at hand, the maximum emission rate of a fluorophore is then found to be the product of P₂ with the system’s intrinsic radiative rate, g₁₂. The desired expression for the intensity dependence of the emission rate is, therefore,

(11)

which shows saturation of the single-molecule/particle emission rate with increasing excitation intensity. Based on the numbers provided for both CdSe and R6G, their respective saturation intensities are I_sat(CdSe) » 100 kW cm^–2 and I_sat(R6G) » 200 kW cm^–2. It is, therefore, apparent that the excitation intensities considered here (I₅₃₂ » 1 kW cm^–2 are far from saturation in either system.

(G) Objective and System Collection Efficiency. We now consider the overall collection efficiency of the microscope. This will ultimately determine whether single-molecule detection is possible and the corresponding signal-to-noise ratio seen in the experiment. At the heart of this estimate is a calculation of the microscope objective’s collection efficiency. To carry this out, recognize that associated with each objective is a numerical aperture (NA), which provides a measure of its light gathering ability. Numerical apertures typically seen in single molecule experiments range from NA = 0.9 to 1.45, with higher NA objectives being able to gather more light. Occasionally objectives with NAs greater than 1.45 are used but only in conjunction with special high index immersion oils. In the microscope described here, the objective’s numerical aperture is NA = 1.25 and is related to the range of incident light angles it collects through the expression NA = nsinq. In this relation, n is the refractive index of the immersion oil/glass coverslip (n = 1.515) and qis the maximum incident light half angle. A NA of 1.25 leads to q = 55.6°.

Next, assuming an isotropic point source, the following calculation can be performed to determine the collection efficiency of the objective. For simplicity, we ignore the effects of an oriented transition dipole [28] as well as changes due to the actual emission pattern of a fluorophore when located at the interface between two regions of differing refractive index (air vs. glass) [29]. The collection efficiency of the objective, h_obj, is thus determined by the following expression

(12)

This formula can be readily integrated to yield h_obj = 0.22, which implies that the objective collects 22% of any isotropically emitted light. In practice, however, both the orientation of the transition dipole as well as the presence of a higher-refractive-index substrate will modify the objective’s actual collection efficiency.

The overall microscope detection efficiency is now a product of h_obj along with the transmission efficiencies of other optical elements in the system. In the microscope described here, the efficiencies of both a dichroic beamsplitter and a dielectric barrier filter must be considered. These values are h_dichroic » 0.95 and h_barrier » 0.95 (between 580 nm and 750 nm) based on transmission curves provided by the manufacturer. The last element to consider is the CCD whose quantum efficiency is h_ccd,. In our system, the actual value of h_ccd is not known, as it has not been provided by either the vendor or the manufacturer; however, if we estimate h_ccd= 0.10, which appears reasonable based on a comparison to other similar CCD chips, the total collection efficiency of the microscope is found to be approximately 2%

(13)

This value is in line with typical collection efficiencies of single-molecule imaging microscopes, which range from 1% to 10% [4, 16].

(H) Signal-to-Noise Estimate. Once the emitted light from the fluorophore has been collected, the signal-to-noise ratio is the last factor dictating whether single-molecule detection is feasible. Although there are (realistically) many potential sources of unwanted noise in an experiment, the major source of noise in a carefully constructed measurement is shot noise due to the signal itself. While background autofluorescence from other parts of the sample or from other undesired fluorophores will all act concertedly to reduce the signal-to-noise ratio, their contributions are generally suppressed by a careful choice of solvents, substrates, and sample preparation conditions. The detector also contributes noise through dark counts and read noise; however, the former can be eliminated by cooling the detector. The major detector contribution to noise is therefore CCD read noise, which is electronic in nature and originates from amplifiers in the CCD circuitry [16].

To determine the signal-to-noise ratio of our hypothetical experiment, we first calculate the number of detected photons, N. The relevant expression is

(14)

where t_int is the integration time and h_system, A, and QY all originate from previous calculations.A general expression for the signal to noise ratio, SNR, in our experiment, can now be written as

Figure 1. Microscope schematic.

Figure 2. Photograph of the actual microscope with the fiber-based spectrometer. The labels A through D are explained in the supporting materials.

(15)

where N is the number of detected photons, n_back is the background noise (n_back µ C_backIt_int; C_back is a proportionality constant, I is the excitation intensity, and t_int is the integration time), n_darkis the dark counts (n_dark = C_darkt_int; C_darkis another proportionality constant and t_int is the integration time), and n_read is the intrinsic CCD read noise [16]. This expression can be evaluated to yield a comprehensive SNR value for our experiment.

Because the upper limit to the SNR ratio is dictated by shot noise, , it alone can be used to determine the maximum possible SNR ratio. The desired expression is

(16)

For the two cases considered here, CdSe and R6G, a simple evaluation of the maximum SNR ratio using previously calculated emission rates and microscope collection efficiencies yields SNR_max(CdSe) » 22 and SNR_max(R6G) » 33, thus showing that it is possible to view single molecules and particles with reasonable contrast under the current conditions.

Microscope Overview

In designing our system we have followed a classic scheme (Figure 1) for an inverted microscope [30] and have extended it to encompass confocal, widefield epi-illumination, and objective-based total internal reflection (TIR) modes. The optical microscope shown in Figures 1 and 2 has been constructed from parts purchased from Thorlabs (mechanics and lenses), Edmund Scientific (objective and eyepiece), Chroma (dichroic beamsplitter, emission barrier filter, and Al mirrors), Ocean Optics (fiber-based spectrometer), Supercircuits (low-cost CCD), and ATI Technologies (video-capture card). A detailed description of the assembly and alignment of the microscope has been included in the supporting materials where we have provided a detailed list of parts, vendors, catalog numbers, and prices. We have grouped the items into two categories: parts required to build the microscope and additional components needed to upgrade it with a fiber-based spectrometer. Because the microscope has been assembled from commercial components, some modifications have been necessary to allow all pieces to work together. Appendix B describes these modifications in more detail. Appendix C describes the assembly/alignment of the system and Appendix D discusses the implementation of an optional fiber-based spectrometer.

The excitation part of the microscope consists of several major components. In particular, a commercial green laser pointer (1, Figure 1) emits 532 nm, partially polarized light to excite fluorophores of interest. The choice of excitation source is dictated by two consideratons: first, many fluorophores absorb green light and, second, this laser pointer is inexpensive and widely available. The laser pointer is mounted on a one-dimensional translation stage (2, Figure 1) allowing the user to switch between TIR and confocal modes (described below). Two Al mirrors (3a and 3b, Figure 1) are utilized to vertically as well as horizontally align the laser beam along the axis of the objective. A dichroic beamsplitter (11, Figure 1) then reflects this incoming excitation towards the objective (6, Figure 1), which focuses the beam onto the sample (8, Figure 1). Alternatively, a partially reflecting Al- or Ag-coated mirror can be used; however, the dichroic yields the best combination of excitation reflectivity (~100%) and transmittance (~95%) of the collected fluorescence.

The manual X,Y,Z stage (7, Figure 1) on which the sample sits is formed from two separate X,Y and Z stages glued together (Thorlabs, SM1Z, and HPT1). This combination allows precise vertical and horizontal positioning of samples over the objective. In both widefield epi-fluorescence and objective-based TIR modes, an additional plan-convex lens (5, Figure 1) with a +75-mm focal length is placed approximately 75 mm behind the objective’s back aperture. The lens is mounted inside an adjustable X,Y slip plate positioner (Thorlabs, SPT1) to allow both widefield epi-fluorescence and TIR modes. In either case, the lens creates a large excitation spot, which allows many fluorophores to be illuminated and viewed simultaneously. To attenuate the excitation, a set of absorptive neutral density filters (4, Figure 1) can be placed prior to the focusing lens.

At the heart of the microscope is a low-cost Edmund PLAN 100´/1.25NA, 160/0.17 oil-immersion objective. Because its numerical aperture (NA = 1.25) is larger than unity, immersion oil is needed between the objective and the coverslip to achieve efficient excitation of the sample and collection of its fluorescence. Furthermore, the substrate thickness should not exceed 0.17 mm and the refractive index of the oil must match the refractive index of the glass coverslip (n = 1.515). When the objective is used as specified, a 100´ magnified image of the sample is created ~160 mm away from the objective’s back aperture.

Still, inserting additional elements needed to complete the microscope (11–13, 15, 16, Figure 1) within this 160-mm length is problematic. Consequently, we use the objective in an “infinity-corrected” mode by bringing the coverslip closer to it, effectively reducing its working distance (WD). In this infinity-corrected mode, the resulting image of the sample moves towards infinity yielding a collimated output, simultaneously providing more space for additional optical components. Although it is possible that the resulting image quality has been degraded, no noticeable effects have been seen in any of our images. This approach yields a low-cost alternative to more expensive (commercial) infinity-corrected objectives. Alternatively, a negative ~160-mm-focal-length plan-concave lens can be inserted just in front of the objective’s back aperture to achieve the same effect [31].

The collection optics of the microscope are constructed as follows. In an inverted microscope, the objective used to excite the sample also collects its fluorescence and any reflected/scattered excitation light. Because the excitation and fluorescence typically have different optical frequencies (called a Stokes shift), a dichroic beamsplitter (11, Figure 1) and a barrier filter (12, Figure 1) can be used to separate the fluorescence from the excitation. In the current configuration, the dichroic beamsplitter (550DCXR) from Chroma has a 550-nm cut-off wavelength and passes ~95% of the fluorescence with wavelengths above this value. Some attenuation of the excitation light is provided by the dichroic beamsplitter, but not enough to yield high quality fluorescence images. An additional dielectric barrier filter is, therefore, required to reject any remaining 532-nm excitation light. The HQ555lp barrier filter from Chroma, located just after the beamsplitter, serves this purpose and blocks any remaining green light (OD > 6 at 532 nm), passing almost 100% of the fluorescence between 560 and 750 nm. Both the dichroic and barrier filter are relatively inexpensive and convenient to use because their broad reflectivity/passband allows the use of different excitation wavelengths without changing either element. This approach should be contrasted to the use of expensive, angle-sensitive holographic notch filters that are adjusted for specific laser wavelengths.

Below the barrier filter, a 100% reflective Al mirror (13, Figure 1) redirects the transmitted fluorescence towards a +150-mm-focal-length plan-convex lens (14, Figure 1). After this, a 50/50 Al mirror (15, Figure 1) splits the original fluorescence into two paths. One leads to a low-cost CCD (19, Figure 1) and the other to an eyepiece or optional fiber coupler (17, Figure 1). In the latter case, the small size of the fiber’s core (typically 200 mm) requires an X,Y stage (16, Figure 1) for precise coupling of the fluorescence into the fiber. A commercial video-capture card is used to record and save important observations directly to the computer hard drive as mpg movies. Commercial video-editing software is also used to crop single frames from the movies for additional analysis. Based on the data published in reference 31, the collection efficiency of the microscope, excluding the CCD quantum efficiency, is approximately 15%.

Materials

A variety of materials have been used to demonstrate the different operating modes of the microscope as well as its sensitivity. In particular, high-quality, colloidal CdSe/ZnS (core/shell) QDs (~4.2 nm diameter) have been synthesized according to the procedure described in reference 32 and have subsequently been overcoated with ZnS to improve their quantum yield. Alternatively, rather than synthesize these particles, such QDs can be purchased commercially [33, 34. Solution-based CdSe NWs have been prepared according to the procedure described in reference 35. They have diameters between 7 and 10 nm and lengths up to 10 microns. Dye-doped polystyrene beads with diameters of 40 nm and 500 nm have been purchased from Molecular Probes [36]. In all cases, sample solutions are created by diluting the initial QD/NW (bead) stock in toluene with deionized water. Samples are then prepared by spin coating or drop casting these solutions onto 1-in-diameter, 0.17-mm-thick glass or fused silica coverslips. Successive dilutions of the resulting sample solution are then used to empirically attain the optimal concentration for imaging individual fluorophores.

Microscope Modes

Because of its flexible design, the assembled microscope can operate in confocal, widefield epi-fluorescence, widefield objective-based TIR, and brightfield transmitted light modes. Each of these techniques has unique features useful for SM detection and their properties are described in succession below. Examples demonstrating each mode are also provided.

(A) Brightfield Transmitted Light Illumination. In the brightfield transmitted light mode [37], a large area of a transparent or partially transparent sample is illuminated from the direction above the specimen. Light passed through the sample is collected by the objective and contains contributions from absorption, scattering, and diffraction. In the absence of a barrier filter, the gathered light is subsequently focused onto a CCD or an eyepiece for observation (Figure 3a). Complimentary to the brightfield transmitted light mode, darkfield illumination, which utilizes light at higher incident angles, has been found to be useful for biological studies in conjunction with single Au or Ag nanoparticles (NPs). This is because the strong environmental sensitivity of the NP scattering spectrum enables ultra-low-concentration analyte sensing without the need for additional amplification protocols [38. 39].

In our microscope, the brightfield transmitted light mode utilizes an illumination source consisting of a fiber bundle (9, Figure 1) connected to a quartz tungsten halogen (QTH) lamp (10, Figure 1). Transmitted light collected by the objective passes through both the dichroic and the barrier filter whereupon a +150-mm lens produces an image of the sample on either the CCD or eyepiece. This is possible because the

Figure 3. (a) Cartoon schematic of brightfield illumination. (b) Image of a microscope calibration grid with square openings (image is approximately 46 ´ 62 mm²).

barrier filter does not attenuate any light with wavelengths above 550 nm.

The primarily application of the brightfield illumination here is to bring the coverslip into focus in preparation for other imaging modes. This can be accomplished by looking for dust particles on the coverslip. Other applications of the brightfield mode include calibrating recorded images taken in this and other modes. To illustrate, Figure 3b shows a test pattern consisting of ~9.1-by-9.1-mm² openings in a ~50-nm Au film. By comparing the length of the square with the full size of the CCD image, the dimensions of the active area can be determined (~46 ´ 62 mm²). This calibration also allows us to estimate both the excitation intensity/spot size utilized in other modes of the microscope (described below) as well as to provide better images of the overall sample morphology.

(B) Confocal Mode. In the confocal mode [40] light from an excitation source is focused into a diffraction-limited spot [41] and a pinhole (~50-mm diameter) is placed in front of the photodetector. Both limit the depth of field to ~0.6 mm and hence lower the signal from out of focus fluorophores and/or any system autofluorescence. This feature serves a number of purposes. For example, thick biological samples (1–2 mm) prepared for fluorescence microscopy often require reduction of the microscope detection volume in order to obtain high quality fluorescence images from a given focal plane [42–44].Using a high-NA oil-immersion objective in conjunction with a pinhole allows one to obtain images at different sample depths. Consequently, three-dimensional images of a specimen can be generated from a series of stacked two-dimensional cross sections [40]. In the case of nanostructures sparsely deposited on a substrate, the surface determines the only plane of observation and the pinhole is not necessary if no significant autofluorescence is expected from the coverslip. Nonetheless, the pinhole is useful for studying emission from nanostructures because it rejects any non-specific contributions to the detected signal.

Common methods for implementing confocal microscopy include raster scanning the excitation across the sample using galvomirrors or moving the sample relative to a stationary excitation spot with piezoelectric actuators [40]. In both cases, the fluorescence is collected with the same objective used to excite the sample and is eventually focused through a pinhole onto a photodetector (usually a photomultiplier tube (PMT) or avalanche photodiode (APD)). An image of the sample is subsequently generated point-by-point.

Without a beam or sample-scanning procedure, as is the case here, true images of a sample cannot be acquired; however, one can observe the morphology of samples with large scattering cross sections (for example, bundles of NWs) by simultaneously using brightfield transmitted light illumination during the confocal data acquisition. On the other hand, for fluorophores with small scattering cross sections, invisible to brightfield illumination (QDs, for example), widefield epi-fluorescence excitation (described below) is used to bring the desired part of sample into the focus for confocal measurements. Experiments possible with the current microscope in confocal mode include acquiring fluorescent trajectories (fluorescence intensity versus time) or fluorescence spectra from different parts of a sample.

(C) Widefield Epi-Illumination. Epi-illumination (also known as reflected light) mode [16] is designed for, but not limited to, observations of opaque samples where the objective illuminates and collects any reflected or emitted light. In contrast to the confocal mode, widefield epi-illumination yields an excitation spot that is spread out over a large area (~22-mm diameter in our case).

The widefield epi-fluorescence mode (Figure 4a) is especially useful for studying nanostructures that exhibit strong fluorescence but have weak scattering cross sections. For example, single CdSe NWs and CdSe/ZnS core/shell QDs are clearly apparent in Figure 4b and c under widefield epi-fluorescence conditions. By contrast, they are nearly invisible under brightfield illumination. Figure 4b shows CdSe/ZnS core/shell QDs deposited on a fused-silica coverslip excited at ~600 W cm^–2. In the excitation area (~300 mm²), many single

Figure 8. (Top): Brownian motion trajectory of a single 500-nm dye-doped polystyrene bead. (Bottom): Plot of RMS displacement versus the square root of time showing linear behavior.

resulting in an intermediate-sized excitation spot. Moving or removing the 75-mm lens, therefore, allows the diameter of the excitation to be varied from ~1.4 mm to ~21 mm.

At the same time, there are corresponding changes in the excitation intensity. To illustrate, at a constant laser power the excitation intensity at the substrate scales quadratically with the diameter, d, of the spot; therefore, the intensity changes by a factor of ~2 to 3 between the extremes encountered in the epi-fluorescence mode. More dramatically, a transition between confocal and widefield modes causes a 225 [(21/1.4)² = 225] times change in the excitation intensity.

To estimate the size of the epi-illumination excitation area we first calculate the angle, a, at which a 1-mm-diameter laser beam converges when focused with a 75-mm lens onto the objective’s BFP. In the current case, a equals ~1/75 radians. The product of this value and the objective’s back focal length, f, determines the size of the excitation area. From the objective’s specifications (160-mm back focal length, 100´ magnification), we determine that f is 160/100 = 1.6 mm; therefore, the excitation has a 1.6(mm) ´ 1/75(rad) @ 0.021-mm (21-mm) diameter, which is in good agreement with the size measured in Figure 4 determined using previous brightfield calibrations of the CCD image area (~22-mm diameter). For estimating the excitation intensity of a typical experiment, we assume that the incoming light has a Gaussian profile leading to nominal peak excitation intensities of ~1 kW cm^–2 and ~8 W cm^–2 for the QDs and NWs shown in Figure 4.

(D) Total Internal Reflection.Total internal reflection [47, 48] is another technique by which to improve the image contrast of fluorescent samples. It accomplishes this in a different way than does the confocal or epi-fluorescence modes. Namely, the TIR mode localizes the excitation volume at or immediately above the substrate. By contrast, both the confocal and epi-fluorescence modes have a finite, yet substantial, depth of field, which translates into a significant volume of the sample being excited. This localization is achieved when the incident laser beam encounters a substrate-air (or substrate-liquid) interface from the substrate side with an incident angle exceeding the TIR critical angle a [a = arcsin(n₁/n₂) where n₁ and n₂ are the refractive indices of the air (or liquid) and the substrate respectively; n₂> n_1;a_glass/air = 41.3° and a_glass/water = 61.4°]. Under these conditions, a majority of the light is reflected back into the substrate; however, an evanescent field is created above the interface, which can be used to excite highly localized fluorophores.

Several key properties of an evanescent excitation make it valuable in single-molecule optical experiments. First, superior suppression of background noise is possible in this mode, because the intensity of the evanescent wave decays exponentially within 50 to 200 nm above the substrate. Fluorophores within this narrowly defined layer are clearly visible, whereas fluorophores above it, for example those in solution, are not effectively excited and do not contribute to the detected fluorescence. Second, the evanescent wave provides additional opportunities for fluorescence polarization measurements. In the TIR mode, one can create an excitation polarization where the electric field of the light has a major component orthogonal to the surface, allowing fluorophores with transition dipoles normal to the substrate to be effectively excited. This makes the TIR mode mutually complimentary to the confocal or epi-illumination modes, in which a major component of the light’s electric field lies parallel to the substrate.

The evanescent excitation can be created using two different approaches: (a) with an external prism [49], or (b) using the same objective for both TIR excitation and fluorescence collection [47]. Both modes yield high quality data, but we prefer to use the objective-based TIR approach due to its more compact design (Figure 9a). In this mode, total internal reflection is achieved using a laser beam with a waist significantly smaller than the objective’s back input aperture. This beam enters the objective collinear with its axis but significantly off-center. Under such conditions and if the NA of the objective is higher than the refractive index of the opposing medium, n₁ [50],the beam encounters the substrate–air (or substrate–liquid) interface above the TIR critical angle. Because the numerical aperture of the microscope objective equals 1.25, conditions for evanescent excitation are easily achieved on a glass–air interface (n₁= 1). Unfortunately, evanescent measurements on a glass–water interface are beyond the objective’s capabilities because the refractive index of water is n₁= 1.33. Thus, to conduct objective-based TIR observations in this case, an objective with a higher NA must be used.

Similar to widefield epi-illumination, the size of the excitation focus in the objective-based TIR can be adjusted

Figure 9. (a) Cartoon schematic of widefield TIR illumination and corresponding images of (b) CdSe/ZnS core/shell QDs and (c) CdSe NWs.

using a +75-mm lens placed before the objective’s back input aperture. Without this lens, the laser beam creates an ~1.4 ´ 2.2-mm²excitation spot [41], which is larger than the confocal diffraction-limited value. With a +75-mm lens inserted, however, the excitation area expands to ~20 ´ 30 mm² and many fluorophores on the substrate can be observed simultaneously. It should be noted that this lens must be displaced laterally along with the incident laser beam in order to achieve proper TIR excitation conditions. Otherwise, the incident light might enter the objective at an improper angle.

In Figure 9b, fluorescence images of CdSe/ZnS core/shell QDs spin-coated on a fused silica coverslip have been obtained with an ~3 kW cm^–2 excitation intensity. Furthermore, Figure 9c shows the fluorescence of CdSe NWs obtained with a n~40 W cm^–2 excitation intensity. These images are similar to those presented in Figure 4 and illustrate QD blinking ( ing materials, CdSe_QDs_widefield_TIR.mpg) as well as differences in the morphology/optical behavior of NWs relative to QDs (supporting materials, CdSe_NWs_widefield_TIR.mpg). A similar analysis of the QD fluorescence intermittency, as described in the epi-illumination section, can be conducted using movies acquired in this mode.

Widefield TIR excitation intensities have been calculated using the numbers obtained from the Widefield epi-illumination section (same incident power, both cases). Two small corrections, however, have been introduced. The first accounts for the elliptical shape of the TIR excitation spot and the second involves a unitless prefactor, A, which accounts for the refractive index of the interface and the excitation’s incident angle as well as its polarization (typical range: ~3 to ~7 for a glass-air interface) [51]. These factors increase the TIR excitation intensity by ~3.3 because of ~1.5´ larger TIR spot size and A @ 5 in accord with reference 51.

Comparison of Modes.As discussed previously, each mode of the microscope has useful features for SM microscopy. Application of a particular mode is dictated by both the sample and by the desired experimental data. In this respect, whereas brightfield illumination proved to be ineffective at imaging samples with low scattering cross sections or with (overall) sizes smaller than ~100 nm, it could be used to observe transparent or partially transparent samples such as the calibration grid (or bundles of solution-based CdSe NWs). This could, in turn, be used to aid other imaging modes of the microscope. Similarly, the confocal mode was found to be useful for spectroscopic measurements due to its high spatial selectivity but was shown to be unsuitable for creating fluorescence images given that the current microscope lacks a sample (or beam) scanning procedure. Widefield epi-illumination allows such fluorescence imaging but, because of its finite (but tangible) depth-of-field, requires low autofluorescence substrates and dilute conditions to suppress undesired out-of-focus fluorescence contributions. Widefield TIR overcomes this problem by limiting the excitation volume to a thin layer above the substrate but, as a direct consequence, cannot image beyond this, preventing the observation of fluorophores in solution or their Brownian motion. In summary, all modes implemented in the current microscope (with the exception of brightfield illumination) can image individual CdSe QDs, NWs, and fluorescent beads (Figures. 4, 6, 7, and 9). The actual choice of imaging mode, however, is left to the reader and to the specifics of a particular experiment.

Conclusions

The present paper describes the construction of a low-cost, multimode optical microscope assembled from off-the-shelf commercial components. Various techniques employed in modern single-molecule microscopy, such as confocal, epi-fluorescence, objective-based total internal reflection, and brightfield transmitted light modes have been implemented to detect the blinking of single CdSe/ZnS QDs, the fluorescence of single CdSe NWs, and the Brownian motion of individual dye-doped polystyrene beads. The goal of this paper is to demonstrate applied optics and microscopy to both undergraduates and high school students alike through the assembly and alignment of this instrument. Furthermore, once assembled, this microscope provides a platform through which they can pursue optical measurements of chemical and physical processes at the single-molecule level.

Acknowledgment. The authors would like to thank Greg Hartland and Dani Meisel for a critical reading of the manuscript. We also thank Li Zeng for assistance in assembling the system. This work has been supported in part by the ACS PRF, the University of Notre Dame and the Notre Dame Radiation Laboratory

Supporting Materials. A list of the components needed to construct the microscope, modifications, assembly and alignment, optional spectrometer, CCD camera comments, and freeware comments as well as movies illustrating the experiments described are included in a Zip file as supporting materials. (http://dx.doi.org/10.1333/s00897050912a)

References and Notes

1. Moerner, W. E.; Kador, L. Phys. Rev. Lett. 1989, 62, 2535–2538.

2. Orrit, M.; Bernard, J. Phys. Rev. Lett. 1990, 65, 2716–2719.

3. Barkai, E.; Jung, Y.; Silbey, R. Annu. Rev. Phys. Chem. 2004, 55, 457–507.

4. Single-Molecule Optical Detection, Imaging, and Spectroscopy; Basche T.; Moerner, W.E.; Orrit, M.; Wild, U. P., Eds.; VCH: Germany, 1997.

5. Kneipp, K.; Wang, Y.; Kneipp, H.; Perelman, L. T.; Itzkan, I.; Dasari, R. R.; Feld, M. S. Phys. Rev. Lett. 1997, 78, 1667–1669.

6. Bernard, J.; Fleury, L.; Talor, H.; Orrit, M. J. Chem. Phys. 1993, 98, 850–859.

7. Basche, Th.; Moerner, W. E.; Orrit, M.; Talon, H. Phys. Rev. Lett. 1992, 69, 1516–1519.

8. Xie, X. S.; Trautman, J. K. Annu. Rev. Phys. Chem. 1998, 49, 441–479.

9. Kuno, M.; Fromm, D. P.; Hamann, H. F.; Gallagher, A.; Nesbitt, D. J. J. Chem. Phys. 2001, 115, 1028–1040.

10. Brokmann, X.; Hermier, J.-P.; Messin, G.; Desbiolles, P.; Bouchaud, J.-P.; Dahan, M. Phys. Rev. Lett. 2003, 90, 120601-1–120601-4.

11. Diebold, A. C. AIP Conf. Proc. 2003, 683, 81–96.

12. Kondo, Y.; Takayanagi, K. Science 2000, 289, 606–608.

13. Holmes, J. D.; Johnston, K. P.; Doty, R. C.; Korgel, B. A. Science 2000, 287, 1471–1473.

14. Gaponenko, S.V. AIP Conf. Proc. 2001, 560, 157–176.

15. Wu, Y.; Xiang, J.; Yang, C.; Lu, W.; Lieber, C. Nature 2004, 430, 61–64.

16. Moerner, W. E.; Fromm, D. P. Rev. Sci. Instr. 2003, 74, 3597–3619.

17. Protasenko, V.; Gallagher, A.; Nesbitt, D. Opt. Comm. 2004, 233, 45–56.

18. Gerton, J.; Wade, L.; Lessard, G.; Ma, Z.; Quake, S. Phys. Rev. Lett. 2004, 93, 180801-1–180801-4.

19. Zimmermann, J.; van Dorp, A.; Renn. A. J. Chem. Educ. 2004, 81, 553–557.

20. Harbron, E. J.; Barbara, P. E. J. Chem. Educ. 2002, 79, 211–213.

21. Murray, C. B.; Norris, D. J.; Bawendi, M. G. J. Am. Chem. Soc. 1993, 115, 8706.

22. Mikulec, F. V.; Kuno, M.; Bennati, M.; Hall, D. A.; Griffin, R. G.; Bawendi, M. G. J. Am. Chem. Soc. 2000, 122, 2532.

23. Yu, W. W.; Qu, L.; Guo, W.; Peng, X. G. Chem. Mater. 2003, 15, 2854.

24. Du, H.; Fuh, R. A.; Li, J.; Corkan, A.; Lindsey, J. S. Photochem. Photobio. 1998, 68, 141.

25. Dabbousi, B. O.; Rodriguez Viejo, J.; Mikulec, F. V.; Heine, J. R.; Mattoussi, H.; Ober, R.; Jensen, K. F.; Bawendi, M. G. J. Phys. Chem. B 1997, 101, 9463.

26. Magde, D.; Wong, R.; Seybold, P. G. Photochem. Photobio. 2002, 75, 327.

27. Fisher, B. R.; Eisler, H-J.; Stott, N. E.; Bawendi, M. G. J. Phys. Chem. B 2004, 108, 143.

28. Plakhotnik, T.; Moerner, W. E.; Palm, V.; Wild, U. P. Opt. Comm. 1995, 114, 83.

29. Hellen, E. H.; Axelrod, D. J. Opt. Soc. Am. B 1987, 4, 337.

30. Olympus IX70 Inverted Microscope Cutaway Diagram.http://www.olympusmicro.com/primer/techniques/fluorescence/ix70fluorescence.html(accessed May 2005).

31. Protasenko, V. V.; Gallagher, A.; Labardi, M.; Nesbitt, D. J. Proc. SPIE Int. Soc. Opt. Eng. 2003, 5188, 254–264.

32. Peng, Z.A.; Peng, X.G J. Am. Chem. Soc. 2001, 123, 183–184.

33. Evident Technologies. http://www.evidenttech.com/ (accessed May 2005).

34. QuantumDot. http://www.qdots.com (accessed May 2005).

35. Grebinski, J. W.; Richter, K. L.; Zhang, J.; Kosel, T. H.; Kuno, M. J. Phys. Chem. B 2004, 108, 9745–9751;Grebinski, J. W.; Hull, K. L.; Zhang, J.; Kosel, T. H.; Kuno, M. Chem. Mat. 2004, 16, 5260.

36. Molecular Probes. http://www.molecularprobes.com/ (accessed May 2005).

37. Abramowitz, M.; Contrast Methods in Microscopy. Transmitted Light; Basics and Beyond Series, Olympus America Inc.: Melville, NY, 1985.

38. Haes, A. J.; Van Duyne, R. P. Expert Rev. Mol. Diagn. 2004, 4, 527–537; Haes, A. J.;Van Duyne, R. P. Anal. Bioanal. Chem. 2004, 379, 920–930; McFarland, A. D.; Van Duyne, R. P. Nano Lett. 2003, 3, 1057–1062.

39. Raschke, G.; Brogl, S.; Susha, A. S.; Rogach, A. L.; Klar, T.A.; Feldmann, J.; Fieres, B.; Petkov, N.; Bein, T.; Nichtl, A.; Kurzinger, K. Nano Lett. 2004, 4, 1853–1857.

40. Confocal Scanning Optical Microscopy and Related Imaging Systems; Corle, T. R.; Kino, G. S.; Academic Press: New York, 1996.

41. The absolute diffraction limited spot is ~0.2 mm, estimated from l_excitation/2NA where NA = 1.25 and l_excitation = 532 nm. This assumes that the objective’s back aperture is fully filled; however, when the beam diameter is smaller than the objective’s input aperture (as is the case here), the effective NA of the objective is NA_eff ~ d/2f where f is the specified objective focal length and d is the diameter of the laser beam. For d = 1 mm and f = 2.74 mm, NA_eff @ 0.18, which is ~7 times smaller than the objective’s specified numerical aperture (NA=1.25). As a consequence, in the confocal mode the actual diameter of the excitation spot is ~1.4 mm, not ~0.2 mm. In all cases, though, the NA of the objective when collecting emitted light is still approximately 1.25.

42. Theory of Confocal Microscopy. Introduction to Confocal Microscopy http://www.olympusfluoview.com/theory/confocal intro.html (accessed May 2005).

43. White, J. G.; Amos, W. B.; Fordham, M. J. Cell Biology 1987, 105, 41–48.

44. Single Molecule Detection in Solution. Methods and Applications, 1st ed.; Zandler, Ch.; Enderlein, J.; Keller, R. A., Eds.; Wiley-VCH: New York, 2002.

45. Image J. http://rsb.info.nih.gov/ij/ (accessed May 2005).

46. Kirksey, H. G.; Jones, R. F. J. Chem. Educ. 1988, 65, 1091; Nakroshis, P.; Amoroso, M.; Legere, J.; Smith, C. Am. J. Phys. 2003, 71, 568 and references therein.

47. Stout, A. L.; Axelrod, D. Appl. Opt. 1989, 28, 5237–5242.

48. Total Internal Reflection Fluorescence Microscopy. http://www.olympusmicro.com/primer/techniques/fluorescence/tirf/tirfhome.html (accessed May 2005).

49. Kobitski, A.Yu.; Heyes, C. D.; Nienhaus, G. U. Appl. Surf. Sci. 2004, 234, 86–92.

50. According to Snell’s law n₁sinq₁ = n₂sinq₂, where n₁ and n₂ are the refractive indices of air (or liquid) and substrate, respectively, q₂ is the angle of an incident beam, and q₁ is the angle of a refracted light. When n₂> n₁, there is a range of q₂, where n₂sinq₂ > n₁, and the angle at which sinq₂ = n₁/n₂defines the TIR critical angle. Because NA = n₂sinq₂, we conclude that the objective NA must be larger than n₁ to achieve TIR.

51. Axelrod, D.; Methods Cell Biology 1989, 30, 245–270.