1. Introduction

Surface plasmon resonance (SPR) has been extensively used since the seventies to probe minute index variations at the interface between a gold film and a dielectric medium [1–3]. Many geometries including metal films, stripes, nanoparticles, nanorods, holes, slits . . . have been used to sustain surface plasmon polaritons (SPP) which offer a wide range of original properties, such as field enhancement and localization, high surface sensitivity and sub-wavelength localization. This explains why SPP have found applications in a variety of fields such as spectroscopy, nanophotonics, biosensing, plasmonic circuitry, nanolasers, sub-wavelength imaging. SPR is also very popular for the detection of molecular adsorption of small molecules, thin polymer films, DNA or proteins, self-assembled layers . . . [4, 5]. The principle of SPR detection takes advantage of the high sensitivity of SPP to refractive index gradients generated by the adsorption of molecules on a metal substrate. The plasmon wave undergoes a modification of both its amplitude and its phase when it meets an obstacle of different refractive index.

SPP are transverse magnetic (TM or p-polarized) surface waves that propagate along a metal-dielectric interface, typically at visible or infrared wavelengths [3]. SPP are guided waves since both the electric and magnetic amplitudes decay exponentially in the Z direction normal to the interface into both the metal and the dielectric material. Because of dissipative losses in the metal, SPP are also damped in their propagation along the X-direction (parallel to the gold surface and normal to the magnetic field).

More elaborated SPR microscopy techniques with enhanced sensitivity have been proposed, based on Mach-Zehnder interferometry [6,7] or dark-field microscopy [8]. The limitation of all methods using the Kretschmann configuration comes from a few micron lateral resolution due to the guided mode propagation of the plasmon inside the thin metal layer [9]. Coupling light with plasmon modes thanks to oil-immersion [10–12] or solid-immersion [13] objective lenses allows to circumvent this difficulty by shaping and confining the plasmon wave laterally to the gold surface. The excitation of SPR with high numerical aperture lenses combined with an interferometric device has pushed down SPR microscopy resolution to sub-wavelength dimensions [14–19]. Two methods have been proposed so far: a wide-field SPR microscope [12, 16] and a scanning surface plasmon microscope (SSPM) [14, 15]. We have more specifically focused on the second method during the past ten years [15, 17–26].

Conventional SPR machines typically operate in the visible and near-visible wavelength range, from 500 to 800 nm. Surface-plasmon waves in the visible range, characterized by a shallow penetration of less than a few hundreds of nm, are useful for studying monomolecular layers in contact with a sensor surface. This is why SPR is popular for quantitative studies of dynamic interactions in thin biolayers, including molecular recognition of binding events. However, visible-range SPR is not optimal for studying living cells because the cell size may considerably exceed the penetration depth of visible-range surface-plasmon waves. As an alternative, infrared (IR) surface plasmons penetrate much deeper and are more appropriate for studying cells. In particular, the penetration depth of near to mid-IR surface-plasmon waves is a few micrometers, that is just about typical adhering cell height. Traditional glass lenses and optical components can still be used in the very near IR (up to two microns) [27] whereas for larger wavelengths other optical materials such as silicon [28, 29], chalcogenide glass [30–32] and ZnS [33, 34] are requested.

In this study, for the first time to our knowledge in the literature, we develop a near IR SSPM (1550 nm), combined with a visible SSPM (632.8 nm). To keep the possibility of high resolution imaging with SPP excitation, at reasonable cost, using the same objective lens, we select 1550 nm as the IR wavelength. We describe the V (Z) response of this microscope in terms of a Fourier transform of the reflected electric field filtered by the objective lens pupil window and we show that the width of this window is a key parameter for maintaining a good microscope contrast since we combine two wavelengths. Because the precise characteristics of the objective lens pupil window are quite impossible to know a priori, it is also important to develop a method to quantify the impact of this filtering on the microscope response. Therefore we propose a wavelet-based space-frequency analysis to decipher the characteristic frequencies of the V (Z) curves and explain how and how much the optical elements can impact the performance of the SPR detection by this microscope

The paper is organized as follows. Section 2 is devoted to the description of the two wavelength SSPM set-up and of the method of blood cell sample preparation. In Section 3 are introduced the principles of SPR and SSPM with their typical R_p(θ) and V (Z) functions. SPR properties in both visible and near IR ranges are described in order to enlighten the capabilities of IR SSPM. The general shape of the V (Z) response is interpreted in terms of a superimposition of different frequencies produced by either the surface plasmon resonance or the confinement of the incident light angles. Since in experimental situations the decoding of these different frequencies is not intuitive, especially in near IR, we use in Section 4 the Morlet wavelet transform to build a space-frequency representation of the V (Z) response. We show that this space-frequency decomposition is well suited to reveal the characteristic frequencies of the experimental V (Z) curves, and this even though the objective lens pupil profile is not known. Finally we apply this wavelet decomposition to theoretical and experimental V (Z) curves obtained from respectively numerical simulations and direct recording from our two-wavelength SSPM. We illustrate the performance of this microscope by comparing a series of images of a blood cell obtained with both visible and near IR lights.

2. Materials and methods

2.1. Optical set-up

Fig. 1 shows the optical set-up combining two SSPMs working at 632.8 nm and 1550 nm respectively, with an inverted microscope stand. On each line we have introduced a heterodyne interferometer equipped with reference mirrors (M1 and M2), beam splitters (BS1 and BS2) and two acousto-optic modulators (AOs). The two polarized laser sources (He-Ne (λ = 632.8 nm) and near IR laser with external cavity (λ = 1550 nm)) are collimated, expanded so that the beam diameters are twice the objective lens (OL) aperture before focusing on the sample. The SSPM image is formed by scanning the sample with a XY piezoelectric stage at a fixed defocus Z. We use a 1.45 NA 60x Olympus objective for both wavelengths. The principle of SPP excitation is analog to Kretschmann configuration [3]: the coupling medium of refractive index n₀ is made by an OL, an immersion oil and a coverslip. The gold film sample was prepared by thermal evaporating under vacuum a 35 nm gold layer on a 3 nm chromium layer - (ACM - Application couches minces - Villiers St Frédéric, France). The final dielectric medium with refractive index n₃ is air or water (here we use only air). The incoming rays that fall on the glass-gold interface with an angle θ_SP are coupled to SPP; they propagate at the gold/dielectric interface and reradiate in symmetrical rays back to the objective lens and to the interferometer. The rays which are not coupled to SPP are reflected back into OL without SPP phase retardation. The refractive index n₂ and the size of the objects localized at the metal/final media interface lead to local variations of the coupling angle θ_SP and of the SPP wave vector k_SP.

 

Fig. 1 (a) Global view of the optical set-up. (b) Zoom on the inverted microscope stand and the 3 axis piezo-table (green arrows) with the gold coated coverslip. S1 and S2 are respectively the red (He-Ne - 632.8 nm) and the near-IR (1550 nm) sources. BS1 and BS2 are the two beam splitters. M1 and M2 are the two mirrors of the two interferometers. AOs are acousto-optic modulators; each interferometer has two AOs, designed for each wavelength. On each interferometer, before the injection in the microscope stand, we have interposed beam expanders (BE). Finally the two red (1) and near-IR (2) beams are injected inside a microscope stand (MS) toward the microscope objective lens (OL) which focuses onto the sample coverslip (CV). D1 (resp. D2) is an amplified silicon (resp. germanium) detector.

Download Full Size | PDF

2.2. Cellular sample preparation

Glass coverslips were cleaned in a caustic detergent for one hour at 50°C, and rinsed twice with distilled water (at least 2 hours each) at room temperature; ultimately they were quickly rinsed with ultrapure water (18MΩ.cm⁻¹), absolute ethanol and dried under a nitrogen flow. They were stored in clean compartments and sent to a French company (ACM Application couches -ACM - Villiers St Frédéric) for thermal evaporation of a 3 nm adhesion layer of chromium and 35 nm of gold. If not used immediately, gold coverslips were stored into a vacuum chamber to prevent pollution due to atmospheric impurities. The gold samples where treated by O₃ and UV light before being brought to the hematology laboratory for the preparation of the blood extract by blood smearing. EDTA-anticoagulated fresh peripheral blood or bone marrow samples were processed using the whole-blood lysis technique and a drop of the cell suspension was placed at one end of the clean gold-coated coverslip, and the edge of a second slide was drawn across the drop at an angle so that capillary action spreads the drop along the edge. The second slide was then pushed in one smooth motion to the opposite end of the first slide, spreading the drop across the slide to make the smear. This method, although apparently simple, requires some dexterity and several hours of training to produce samples with a single layer of cells and to avoid crashing the cells onto the coverslip.

3. Modeling the scanning surface plasmon microscope

3.1. Reflectivity curves: from visible to IR

Propagation of the plasmon excitation. The wave vector (k_SP) for surface plasmon in a Kretschmann configuration can be written as:

The fact that k_SP is parallel to the gold interface imposes that the incident electric field is p-polarized (transverse mode (TM) for the magnetic field). For each wavelength, the matching condition for SPR is fulfilled for a single angle of incidence θ_SP, the coupling angle, which increases when decreasing the wavelength λ. The lateral propagation length L_X of the plasmon wave is directly related to the inverse of the imaginary part of k_SP. If one neglects, in a first approximation, the correction to k_SP due to the prism coupling (Δk), the lateral propagation length behaves as:

 

Fig. 2 Plot of the evolution of L_X (a) and L_ZSP (b) with the wavelength λ. The dielectric permittivities of gold ε_m(λ) were taken from reference [35].

Download Full Size | PDF

 

Fig. 3 Reflectivity curves (R_p) computed for a 35 nm gold film with a 3 nm chromium sublayer, with air as dielectric medium. (a) Modulus of R_p versus θ. (b) 3D plots of the real and imaginary parts of R_p and the angle of incidence θ: [ℜ(R_p), ℑ(R_p), θ]. In red (resp. black) are plotted the curves for λ = 632.8 nm (resp. λ = 1550 nm). The dielectric permittivities ε_m(λ) of gold were taken from reference [35], and the ones of glass are those of BK7.

Download Full Size | PDF

SPR and evanescent waves. Evanescent waves occurring for angles of incidence greater than the critical (total reflection) angle θ_ATR are described by a penetration depth which quantifies the confinement of the field in the vicinity of the interface. This penetration depth depends on the indices of the two media bounding the interface, for instance the index of glass n_c and the index of the dielectric medium n_D[36]. When the angle of the incident beam θ increases fromθ_TR to π/2, l_Z decreases from ∞ to $l_Z=\frac{\lambda }{2\pi \sqrt{n_c^2-n_D^2}}$. When SPR is excited, namely when θ = θ_SP, the penetration depth in the dielectric medium reads:

3.1.1. SPR reflectivity curves

Two reflectivity curves of |R_p| versus the incident angle θ are shown in Fig. 3(a) for λ = 632.8 nm (in red) and λ = 1550 nm (in black) respectively. They were computed for a 35 nm layer gold film in contact with air, with a 3 nm adhesion chromium layer, to closely match the experimental conditions that will be described later (see reference [36] for the formula of the transmission and reflection coefficients of a multi-layer system). We selected a 35 nm thickness of gold rather than 45 nm, to increase the signal response of SSPM in near IR range. We observe in Fig. 3 that both reflectivity curves show a drop in between θ = 40 ° and θ = 50 ° (44.26° for λ = 632.8 nm and 42.05° for λ = 1550 nm), corresponding to the plasmon resonance and that the shape of this resonance is definitely much sharper with a near IR excitation, as previously mentioned. We notice also that, below the total internal reflection angle (θ < θ_ATR), the reflectivity in p polarization is higher with near IR than with red wavelength. The plasmon resonance corresponds to a fast variation of the phase of the complex reflectivity R_p, which also depends on the sharpness of the resonance; the sharper the resonance, the closer this phase drop to 2π. We illustrate this property of the surface plasmon resonance on a 3D plot of the real and imaginary components of R_p and the angle θ in Fig. 3(b). At resonance (θ ≳ 40°), the red (visible) and black (near IR) R_p curves make a loop of about one turn, corresponding to this 2π phase drop. We remark also that the red loop is not complete, meaning that the phase drop does not reach 2π, whereas the black loop is quite complete, meaning that with near IR light the phase drop reaches a value closer to 2π.

The angle of plasmon resonance θ_SP changes with the wavelength λ as:

 

Fig. 4 Plot of the evolution of θ_SP (a) and Δθ_SP (b) with the wavelength λ, computed for a 35 nm gold film with a 3 nm chromium sublayer, with air as dielectric medium. The red curve in (a) corresponds to the total internal reflection angle θ_ATR. The dielectric permittivities ε_m(λ) of gold were taken from reference [35], and the ones of glass are those of BK7.

Download Full Size | PDF

We have pointed out just above that the decrease of the radiative losses of the plasmon wave when increasing λ should materialize in a narrowing of the plasmon resonance curve |R_p(θ)|. Actually the angular width of the surface plasmon resonance is determined by the imaginary part of the surface plasmon wave vector [27]:

3.1.2. Sensitivity of the plasmon reflectivity curves with an adsorbed layer

Indeed, the plasmon resonance is very sensitive to adsorbed layers on gold, which explains that it has been extensively used during the past decades for sensing monolayers. It is therefore interesting to compare this sensitivity, in particular the drift of θ_SP in the presence of an adsorbed layer. To this purpose, we added on the gold film a dielectric layer with optical index 1.4 and thickness e varying from 0 to 250 nm and we computed the modulus of R_p. We reconstructed in Fig. 5, a color coded image of |R_p| (0 corresponding to dark blue, and 1 to red), in the two dimensional space [θ, e].

 

Fig. 5 Color maps of the reflectivity modulus (|R_p|) in the two dimensional space [θ, e], for e varying from 0 to 250 nm. |R_p| is color-coded from dark blue (0) to carmine red (1). (a) λ = 632.8 nm; (b) λ = 1550 nm. As explained in the text, we took a thin 35 nm gold film on a 3 nm sublayer of chromium, and a dielectric layer with thickness e in contact with air.

Download Full Size | PDF

The drift of the plasmon resonance to larger θ values when the thickness of the added dielectric layer is increased is more important for visible (632.8 nm, Fig. 5(a)) than for near IR (1550 nm, Fig. 5(b)) lights. For both wavelengths, we note that the drop of |R_p| at resonance widens when increasing the dielectric layer thickness. Given the fact that the range of incident angles is limited by the objective numerical aperture, we conclude from the numerical results reported in Fig. 5 that the smallest thickness variations will be better detected from the significant drift in θ_SP observed with the visible light. However, beyond 100 nm, the resonance drop fades and it crosses the limit angle of the objective lens (arcsin(NA/n_c) ∼ 73° with NA=1.45 and n_c = 1.5151), making impossible the use of SPR excitation mode for microscopic imaging. Meawhile, the slower drift of the drop of |R_p| with the dielectric thickness e in near IR excitation allows us to characterize much thicker samples, typically up to 700 nm thickness. In the case of near IR excitation light, for dielectric thickness larger than 700 nm, the plasmon resonance angle θ_SP of a 35 nm gold film in air saturates at about 70°, inside the range of angle afforded by a 1.45 NA objective lens.

It is also very instructive to investigate how the phase drop at plasmon resonance varies in the presence of an adsorbed dielectric layer. We use in Fig. 6 the same 3D representation of the complex reflectivity R_p, with the incident angle θ along the vertical axis as in Fig. 3(b). This representation shows that in visible light, the phase drop at plasmon resonance starts from a value close to 2π for e = 0 nm and progressively decreases to values close to π for e = 250 nm. For near IR light, the phase drop changes much slower than before when increasing e, showing that the plasmon resonance sensor keeps its high sensitivity for a wider range of thickness values. Thus, depending on the thickness of the sample to be characterized, the wavelength of operation will have to be optimized, the thicker the sample, the larger the wavelength.

 

Fig. 6 3D plots of the real and imaginary parts of R_p and the angle of incidence θ: [ℜ(R_p), ℑ(R_p), θ]. The differents curves were computed for a 35 nm gold film with a 3 nm sublayer of chromium and a dielectric layer (index 1.4) of increasing thickness from 0 nm (red) to 250 nm (blue). (a) λ = 632.8 nm; (b) λ = 1550 nm.

Download Full Size | PDF

3.2. V (Z) curves

SSPM is a combination of an heterodyne interferometer, a high numerical aperture objective lens and a (X, Y, Z) piezoelectric scanning device (Fig. 1). It is commonly used to record the so-called V (Z) response [14,15]; this is illustrated in Fig. 7 where are shown two typical experimental |V (Z)| curves obtained with our two-wavelength optical device. We notice immediately that the interpretation of these curves for positive Z values, is not at all intuitive since they look like a mixture of non stationary oscillatory behavior that cannot be separated by simple Fourier analysis. We will therefore need an adapted space-frequency analysis to identify the underlying frequencies and to interpret them in terms of plasmon mode frequencies. V (Z) is an integral over the radial θ (θ_min, θ_max) and azimuthal φ (0, 2π) angles of the backward reflected electric field after plasmon excitation at the gold dielectric interface. The reflected field is described by R_p (resp R_s) in p (resp. s) polarization configuration. This method is directly inspired from scanning acoustic microscopy [14, 37, 38].

Ideally, this pupil function P should be as flat as possible (constant). In our experiment it will be modeled by a rectangular window with smooth gaussian borders of the form (Fig. 8):

 

Fig. 7 Plot of the experimental |V (Z)| curves obtained with the two-wavelength SSPM set-up. (a) Lin-lin plots of the normalized modulus responses |V (Z)|/V₀. (b) Log-lin plot of |V (Z)|. In red: visible light (λ = 632.8 nm); in black: near IR light (λ = 1550 nm).

Download Full Size | PDF

 

Fig. 8 Modeling the objective lens pupil from a rectangular shape (P = 1) to a smooth profile. The σ parameter describes the smoothing of P (Eq. (10)). The dashed black (resp. solid red) vertical lines correspond to the resonance plasmon angles θ_SP for near IR (resp. red) light.

Download Full Size | PDF

With the following change of variables [39]:

Table 1. Table of spatial frequencies of V (Z) computed from a constant pupil function and a numerical aperture lens NA=1.45

View Table | View all tables in this article

Note that the modulus of V (Z)

As shown in Fig. 9, when the pupil borders are smoothed (Fig. 8), the periodic modulation with frequency ν₁ of |V (Z)| vanishes to leave a simple 1/Z decay. One would then think that the smoothest P would be best for surface plasmon microscopy because it would avoid the mixing of the pupil window frequencies with those of the plasmon resonance itself. In practice, there is a compromise to reach between this smoothing and the fact that there should be enough light power in the range of angles corresponding to θ_SP. A similar issue has been recently raised by Zhang et al[40] who proposed to introduce a spatial light modulator confocal pinhole in a wide field SPR microscope to control the microscope pupil function. Although this system could be a more stable and compact design than the heterodyne interferometer reported here, its loss of signal to noise ratio could be detrimental to its performance in near IR.

 

Fig. 9 Theoretical |V (Z)| curves (in log-lin representation) computed for a constant reflectivity R_p = 1 (panels (a) and (b)) and for a 35 nm gold film with a 3 nm sublayer of chromium in air (panels (c) and (d)). The P shape parameter σ (from 0 to 1) accounts for the smoothing of the objective lens pupil (Fig. 8). (a, c) λ = 1550 nm. (b, d) λ = 632.8 nm.

Download Full Size | PDF

3.3. V (Z) response to a phase drop

The phase of the reflectivity undergoes a very sharp drop at plasmon resonance. Its impact on V (Z) can be computed assuming that the reflectivity modulus |R_p| = 1 everywhere, except when θ = θ_SP, where |R_p(θ_SP)| tends to zero at resonance (Fig. 3(a)). This is fairly realistic in case of no radiative metal loss. The phase drop can be described by a Heaviside function, which means that below resonance ϕ = 0 and above resonance Δϕ_SP = ±2π +ε_ϕ. This small quantity ε_ϕ is important since it describes the sharpness of the plasmon resonance. Accordingly:

Then the integral in the definition (Eq. (12)) of V (Z) splits in two integrals:

Taking cosθ_min = 1, we obtain:

Table 2. Table of spatial frequencies (Eq. (23)) corresponding to the combination of plasmon resonance and objective pupil lens aperture, computed for a 35 nm gold film with a 3 nm sublayer of chromium.

View Table | View all tables in this article

Let us point out the presence of a sinc function in the first term in Eq. (22), so that the corresponding period of its modulus is the half period of the sine:

4. Deciphering the frequencies of V (Z) curves using a Morlet wavelet transform

We have shown in Section 3 that the V (Z) response curve involves a superimposition of different frequencies produced by either the surface plasmon resonance or the confinement of the incident angles. In experimental situations it is very difficult when the light source wavelength is varied to know precisely the evolution of these frequencies and to separate surface plasmon modes from other effects coming from the optical device itself. In this section, we apply a wavelet-based space-frequency method to achieve this decomposition. The continuous wavelet transform(CWT) is a mathematical technique introduced in signal analysis in the early 1980s [41,42]. Since then, it has been the subject of considerable theoretical developments and practical applications in a wide variety of fields [43–58]. Some optical wavelet device has been designed that performs the CWT thanks to Fourier windowing principles [59, 60].

The CWT amounts to project an arbitrary signal s(t) onto a family of functions {g_ba(t)}, where b ∈ ℜ is the time parameter and a ∈ ℜ⁺ the scale parameter [44, 52, 53]:

4.1. Morlet space-frequency maps of V (Z) curves for visible light

The Morlet decomposition of the V (Z) function is performed in Fourier space, by a direct product of the Fourier transform of the analyzing Morlet wavelet with the Fourier transform of ln|V (Z)|. Taking the reverse Fourier transform gives directly the wavelet transform at given scale a. This operation is repeated for different values of the frequency ν = 1/a (in μm⁻¹), from which we construct a space-frequency map using a color coding from blue (minima value of the wavelet transform modulus) to red (maxima value of the wavelet transform modulus). Such a space-frequency map is shown in Fig. 10(a) for the V (Z) function computed for a gold film of 35 nm, with a sublayer of 3 nm of chromium, and a constant pupil window P = 1 for 0 < θ < θ_max = arcsin(1.45/1.5151) ∼ 73° in visible light (λ = 632.8 nm). We recognize in this representation the frequency 2ν₁ ∼ 3.4 μm⁻¹ corresponding to the sharp pupil borders at θ_max; it manifests as a wide red band for negative Z values and for Z larger than 15 μm. In the remaining range of Z values emerge two other spatial frequencies, one in between 1 and 2 μm⁻¹ and the other one around 2 μm⁻¹. The smaller one corresponds to 1/ΔZ_SP = 2ν_1SP ∼ 1.36 μm⁻¹ whereas the larger one corresponds to 2ν_3SP ∼ 2.04 μm⁻¹ (Tables 1 and 2) and results from some mixing of the plasmon frequency with the pupil border frequency (Eq. (23)). When the objective pupil is smoothed (Figs 10(b) and 10(c)), the influence of the pupil borders progressively vanishes and we recover the spatial frequency given by the plasmon resonance only. This space-frequency representation shows also that it is necessary to have a strong smoothing of the pupil function of the objective lens to recover the characteristic spatial frequency of the plasmon resonance for Z values smaller than 5 μm.

 

Fig. 10 Morlet wavelet transform of the V (Z) functions computed for a 35 nm gold film with a 3 nm chromium sublayer, for visible light (λ = 632.8 nm). (Bottom) ln|V (Z)| vs Z; (Top) |T_ln|V|(Z,ν = 1/a)| as coded using 256 colors from blue (minimal value) to red (maximal value). (a) Rectangular pupil function P (σ = 0). (b) Smoothed pupil function P (σ = 0.17). (c) Smoothed pupil function P (σ = 0.81). θ_max = arcsin(NA/n_c) with NA= 1.45 and n_c = 1.5151.

Download Full Size | PDF

4.2. Morlet space-frequency maps of V (Z) curves for near IR light

In near IR light, the same scenario is observed in the space-frequency maps when the pupil window becomes smoother. In Fig. 11(a), we see mainly the frequency 2ν₁ ∼ 1.44 μm⁻¹ (red horizontal band) corresponding the sharp boundary of the pupil (Table 1). This band is very flat for negative Z values, and shows extra modulations for positive Z values that originate from the presence of the plasmon resonance. The band corresponding to the plasmon frequency ν_SP = 2ν_1SP ∼ 0.498 μm⁻¹ is quite imperceptible if the pupil window is not smoothed enough (Fig. 11(a) and 11(b)). Smoothing further the pupil window cancels completely the 1.44 μm⁻¹ modulation and strengthens the band corresponding to ν_SP that definitely emerges from the background (Fig. 11(c)).

 

Fig. 11 Morlet wavelet transform of the V (Z) functions computed for a 35 nm gold film with a 3 nm chromium sublayer, for near IR light (λ = 1550 nm). Same representation as in Fig. 10. (a) Rectangular pupil function P (σ = 0). (b) Smoothed pupil function P (σ = 0.17). (c) Smoothed pupil function P (σ = 0.81). θ_max = arcsin(NA/n_c) with NA= 1.45 and n_c = 1.5.

Download Full Size | PDF

4.3. Experimental data

In Fig. 12 are shown the experimental V (Z) curves obtained with our two-wavelength SSPM device for a 35 nm gold film prepared by thermal evaporation on a 3 nm chromium layer. The two ln|V (Z)| signals are plotted at the bottom of Figs 12 (a) and (b). We notice that the |V (Z)| obtained for visible wavelength λ = 632.8 nm reveals the characteristic spatial frequency of plasmon resonance for Z values lower than 5 μm. The strong similarity observed between the Morlet space-frequency maps reconstructed from the experimental (Fig. 12(a)) and the numerical (Fig. 10(c)) V (Z) functions suggests a strong experimental smoothing of our objective lens pupil.

 

Fig. 12 Morlet wavelet transform of the experimental V (Z) functions obtained with the SSPM optical set-up at (a) visible (λ = 632.8 nm) and (b) near IR (λ = 1550 nm) wavelengths. These V (Z) curves were recorded for a 35 nm thick gold film with a 3 nm sublayer of chromium in air. Same representation as in Fig. 10

Download Full Size | PDF

The experimental V (Z) curve recorded with near IR light (λ = 1550 nm), for the same gold film sample, the same day, with the same objective lens, the same matching index liquid and without disassemblying the optical set-up (since both visible and near IR light can be run simultaneously) is shown in Fig. 12(b). It turns out to be very different from the one obtained with visible light, and this not only for the values of its spatial frequencies, but also for their extensions in Z. In particular, two spatial frequencies are evidenced by the Morlet space-frequency map, a lower one around 0.4 μm⁻¹, likely corresponding to the plasmon resonance ν_SP, and a larger one, close to 1 μm⁻¹ that probably corresponds to the objective pupil itself (2ν₁). When comparing the experimental (Fig. 12(b)) and the numerical (Fig. 11) Morlet space-frequency maps, we can conclude that in near IR light, the objective lens has a numerical aperture that is smaller than the value given for visible light, NA_IR = 1.5 * sin(arccos(1 − 1.550/(2 * 1.5))) ∼ 1.31 < 1.45, and its pupil function is flatter than in visible light. This apparent decrease of the objective lens numerical aperture in the near IR probably results from the fact that this Olympus M Plan Apochromat lens was aberration corrected for blue, green and red wavelengths and not for near IR wavelengths.

4.4. Application to blood cell imaging

The two sets of four images shown in Fig. 13 were captured with the two-wavelength SSPM set-up from blood cell smearing on the gold film (Section 2). The imaged cell was selected because, on the one hand it showed adhesion filopodia that strongly adhered on the gold surface (Figs 13(a)–13(c)), and on the other hand it had also a small filamentous body that came out of the gold plane surface (Figs 13(a’)–13(c’)). Since the SSPM offers the possibility to scan the sample at different penetration depths thanks to the defocus parameter Z, four values of Z are represented in Fig. 13 that correspond to focusing inside the glass (Z < 0, Figs 13(a), 13(a’)), on the gold film (Z ∼ 0, Figs 13(b), 13(b’)), on the cell media (Z > 0, Figs 13(c), 13(c’)) and on air above the cell surface (Z > 0, Figs 13(d), 13(d’)).

 

Fig. 13 Comparison of SSPM images collected from a fixed lymphocyte in air with (a–d) visible light (λ = 632.8 nm) and (a’–d’) IR light (λ = 1550 nm). (a) Z = −1.7μm, (b) Z = −0.2μm, (c) Z = 1.25μm, (d) Z = 3.95μm. (a’) Z = −2μm, (b’) Z = −0.5μm, (c’) Z = 1.3μm, (d’) Z = 4.35μm. The scale bar is 10 μm.

Download Full Size | PDF

The comparison of the visible and near IR images confirms that the integration depths for the two wavelengths are very different. For instance, the picture obtained for λ = 632.8 nm and Z = 1.25 μm (Fig. 13(c)) must be compared to the picture obtained with near IR λ = 1500 nm and Z = −0.5 μm (Fig. 13(b’)). The depth of integration of the cell body through the evanescent field afforded with the near IR light is actually much larger; in particular it allows the observation of a filamentous body on the top of the cell (Fig. 13(d’)), that cannot be captured with the visible light (Fig. 13(d)). However, for capturing cell bodies that are firmly anchored on the gold surface, the visible light SSPM (Fig. 13(b)) is more appropriate than the near IR SSPM (Fig. 13(b’)). For larger Z defocus values, the near IR SSPM can still provide images that are not limited by diffraction effects. The near IR image in Fig. 13(d’) shows some granulations on the cell border that could not be observed with visible light in Fig. 13(d) since they are protruding outside the range of the visible light evanescent field. The combination of both these wavelengths gives therefore instructive and complementary views of the cell, without the need of any external dying agent that could disturb and spoil its structure.

5. Conclusion and future prospects

In this paper, we have shown that high resolution scanning surface plasmon microscopy can be carried out in near IR range, with the same objective lens as in visible range, providing a deeper scanning in the Z direction. This microscope is more suitable to investigate thicker samples, such as cells and tissues. We have shown that the two variables ν and Z defined in Eq. (12) are conjugated entities in this microscope. As a consequence, when the surface plasmon resonance is sharper in ν = ((2n_c/λ)cosθ), as observed in near IR light, the V (Z) function is less localized in Z, which allows the characterization of thicker samples.

We have illustrated the performance of this dual wavelength microscope to image a lymphocyte cell. The implementation of a space-frequency analysis of the response function V (Z) of this microscope has proved effective in deciphering the underlying spatial frequencies that can be related to characteristic angles such as the surface plasmon resonance angular position and the angular boundaries defined by the numerical aperture of the objective lens. The advantage of this wavelet decomposition is its ability to perform a local analysis of the frequency response, which cannot be done with classical Fourier method. This space-frequency analysis has enlightened the crucial role played by the shaping the light profile (in particular the pupil of the objective lens) in surface plasmon microscopy. In particular, as regards to previous studies where annular diaphragms were used to shape the incoming light [10, 63, 64], it could be interesting to select the width and the shape of the diaphragm with the help of this wavelet based space-frequency decomposition to optimize the contrast of the surface plasmon response depending on the index and thickness of the sample.

Finally, the operation of this dual wavelength microscope in liquid media is an experimental challenge, since in a liquid medium the plasmon resonance angles increase and get closer to the angular limits of the objective lens. The superimposition of the corresponding spatial frequencies makes then harder to separate them. We are currently testing two methods to achieve this goal, namely the introduction of an objective lens with a larger numerical aperture (which is more expensive) and the modification of the index of the glass/gold/dielectric interface to recover resonance curves within an affordable angular range compatible with lower numerical aperture objective lenses.