Mid-infrared plasmonic biosensing with graphene

See allHide authors and affiliations

Science  10 Jul 2015:
Vol. 349, Issue 6244, pp. 165-168
DOI: 10.1126/science.aab2051

Graphene-based biosensors

The mid-infrared (mid-IR) range is particularly well suited for biosensing because it encompasses the molecular vibrations that identify the biochemical building blocks of life, such as proteins, lipids, and DNA. However, the resulting optical signal is extremely weak and often requires complex techniques to enhance the biological detection. Rodrigo et al. present a graphene-based biosensor that they dynamically tuned over a broad spectral range through electrical gating. The authors selectively probed protein molecules at different mid-IR frequencies using a single device.

Science, this issue p. 165


Infrared spectroscopy is the technique of choice for chemical identification of biomolecules through their vibrational fingerprints. However, infrared light interacts poorly with nanometric-size molecules. We exploit the unique electro-optical properties of graphene to demonstrate a high-sensitivity tunable plasmonic biosensor for chemically specific label-free detection of protein monolayers. The plasmon resonance of nanostructured graphene is dynamically tuned to selectively probe the protein at different frequencies and extract its complex refractive index. Additionally, the extreme spatial light confinement in graphene—up to two orders of magnitude higher than in metals—produces an unprecedentedly high overlap with nanometric biomolecules, enabling superior sensitivity in the detection of their refractive index and vibrational fingerprints. The combination of tunable spectral selectivity and enhanced sensitivity of graphene opens exciting prospects for biosensing.

Graphene has the potential to reshape the landscape of photonics and optoelectronics owing to its exceptional optical and electrical properties (13). In particular, its infrared (IR) response is characterized by long-lived collective electron oscillations (plasmons) that can be dynamically tuned by electrostatic gating, in contrast to conventional plasmonic materials such as noble metals (410). Furthermore, the electromagnetic fields of graphene IR plasmons display unprecedented spatial confinement, making them extremely attractive for enhanced light-matter interactions and integrated mid-IR photonics (1114). Specifically, biosensing is an area in which graphene tunability and IR light localization offer great opportunities.

The mid-IR range is particularly well suited for biosensing, as it encompasses the molecular vibrations that uniquely identify the biochemical building blocks of life, such as proteins, lipids, and DNA (15). IR absorption spectroscopy is a powerful technique that provides exquisite biochemical information in a nondestructive label-free fashion by accessing these vibrational fingerprints. Nevertheless, vibrational absorption signals are prohibitively weak because of the large mismatch between mid-IR wavelengths (2 to 6 μm) and biomolecular dimensions (<10 nm). To overcome this limitation, high sensitivity can be achieved by exploiting the strong optical near fields in the vicinity of resonant metallic nanostructures (1618); however, this comes at the expense of a reduced spectral bandwidth and is ultimately limited by the relatively poor field confinement of metals in the mid-IR (19).

Here, we report a graphene-based tunable mid-IR biosensor and demonstrate its potential for quantitative protein detection and chemical-specific molecular identification. Our device (Fig. 1A) consists of a graphene layer synthesized by chemical vapor deposition and transferred to a 280-nm-thick native silica oxide of a silicon substrate. Graphene nanoribbon arrays (width W = 20 to 60 nm and period P ≈ 2W) are then patterned using electron beam lithography and oxygen plasma etching (20). A scanning electron microscope image and an atomic force microscope profile for typical samples are shown in Fig. 1, B and C. We apply an electrostatic field across the SiO2 layer through a bias voltage (Vg) that is varied between 0 and 120 V to dynamically control the Fermi level (EF) of graphene. Extinction spectra of the device are acquired using Fourier transform infrared (FTIR) spectroscopy for the incident electric field polarized perpendicular to the nanoribbons. Figure 2A shows the extinction for a nanoribbon array with W = 30 nm, P = 80 nm, and different values of Vg (dashed curves). A prominent resonance is observed, which is associated with localized surface plasmons (LSPs) polarized across the nanoribbons. By changing Vg, the resonance frequency is tuned continuously from 1450 cm−1 to above 1800 cm−1. The ribbon width W = 30 nm is chosen so that the frequency tuning range sweeps across the target vibrational fingerprints (fig. S1).

Fig. 1 Tunable graphene mid-IR biosensor.

(A) Conceptual view of the graphene biosensor. An infrared beam excites a plasmon resonance across the graphene nanoribbons. The electromagnetic field is concentrated at the ribbon edge, enhancing light interaction with the protein molecules adsorbed on graphene. Protein sensing is achieved by detecting a plasmon resonance spectral shift (Δω) accompanied by narrow dips corresponding to the molecular vibration bands of the protein. The plasmonic resonance is electrostatically tuned to sweep continuously over the protein vibrational bands. (B) Scanning electron microscope image of a graphene nanoribbon array (width W = 30 nm, period P = 80 nm). Vertical nanoribbons are electrically interconnected by horizontal strips to maintain the graphene surface at uniform potential. (C) Atomic force microscope cross section of a graphene nanoribbon array.

Fig. 2 Mid-IR spectrum of the graphene biosensor.

(A) Extinction spectra of the graphene nanoribbon array (W = 30 nm, P = 80 nm) for bias voltages Vg from –20 V to –130 V before (dashed curves) and after (solid curves) protein bilayer formation. Extinction is calculated as the relative difference in transmission between regions with (T) and without (T0) graphene nanoribbons. Gray vertical strips indicate amide I and II vibrational bands of the protein. (B) Analytic calculation of the extinction spectra after fitting graphene and protein parameters to reproduce experimental data. (C) Graphene carrier density (ns) and Fermi energy (EF) extracted from experimental IR extinction spectra of the bare graphene nanoribbon array at different applied bias voltages Vg. (D) Permittivity of the protein bilayer extracted from the analytic fit to the experimental IR spectra (solid red curves) of the graphene biosensor compared to the permittivity extracted from IRRAS and ellipsometry measurements (dashed black curves). Upper and lower curves show the real and imaginary components, respectively.

We sought to detect protein molecules, the primary material of life enabling most of the critical biological functions. The main vibrational fingerprints of proteins are amide I and II bands (1660 and 1550 cm−1), which are primarily associated with the C=O stretch and N-H bend modes in the amide functional group. For demonstration of protein detection, we used recombinant protein A/G and goat anti-mouse immunoglobulin G (IgG). Incubation of A/G on the sensor surface allows the formation of a protein monolayer by physisorption, which is then used to bind IgG antibodies and form a well-defined protein bilayer (20). The extinction spectra of the sensor are presented in Fig. 2A before and after protein bilayer formation, showing radical changes upon protein immobilization. The first observed prominent effect is a red shift of the plasmonic resonance as a consequence of the change in the refractive index at the sensor surface. Despite the nanometric thickness of the protein bilayer, we detected frequency shifts exceeding 200 cm−1. The second prominent effect is the emergence of two spectral dips at 1660 cm−1 and 1550 cm−1 that are almost undetectable when they are far from the plasmonic resonance (e.g., for Vg = –20 V) and become progressively more intense with increasing spectral overlap (e.g., for Vg = –130 V). Their spectral positions coincide with the amide I and II bands, respectively, unambiguously revealing the presence of the protein compounds in a chemically specific manner. The decrease in extinction induced by the vibrational modes is the result of resonant coupling between plasmons and molecular vibrations (21).

To extract quantitative information on the protein optical parameters, we use an analytical model of the IR response of the graphene nanoribbon array (22). We model graphene in the electrostatic limit (W, P << λ) under the assumption that the ribbon response is dominated by the lowest-order transversal mode. The model involves a detailed account of the protein layer; however, a reasonable agreement is obtained in the limit of a thick protein layer. The transmission coefficient of the structure then reduces toEmbedded Image(1)whereEmbedded Image (2)is an effective graphene-ribbon polarizability that takes into account the complex refractive indices of the silica substrate n2 (23) and the material immediately above the ribbons n1, while the coefficient A is a function of P/W (in particular, A = 28.0 for P/W = 2.67). Here, t0 and r0 are the transmission and reflection coefficients of the interface between media 1 and 2 in the absence of graphene. The response of the latter enters through its frequency-dependent surface conductivity σ(ω), which we model in the local random-phase approximation (11). Finally, we compute the ratio of transmission in regions with and without graphene as Embedded Image, which is the magnitude measured in the experiments.

We first used the analytic model to extract the graphene parameters from experimental IR spectra for bare nanoribbons (i.e., with n1 = 1). The calculated spectra are reported in Fig. 2B (dashed curves) for the extracted relaxation time (τ = 15 fs) and Fermi energies (EF = −0.17 to −0.43 eV). We observe that the carrier density (Embedded Image) changes linearly with Vg (Fig. 2C) and graphene has an intrinsic doping EF0 = −0.17 eV produced by charge transfer from the silica. Next, the analytic model is used to retrieve the protein permittivity from experimental results by adjusting a Lorentzian permittivityEmbedded Image (3)Good agreement is observed between experimental and calculated spectra (Fig. 2B) for the protein Lorentzian parameters upon least-squares fitting. The extracted permittivity has a nondispersive term Embedded Image = 2.08 and shows two absorption peaks at 1668 and 1532 cm−1, matching the amide I and II bands, respectively (Fig. 2D). The fitted permittivity is also in good agreement with independent protein permittivity measurements from ellipsometry for Embedded Image and IR reflection absorption spectroscopy (IRRAS) for Sk, ωk, and γk (20). There is, however, a small discrepancy, which we attribute to a slight overestimate of plasmon-protein coupling in the theoretical model. These results indicate that the proposed graphene biosensor combines refractive index sensing, so far a prerogative of visible plasmonic sensors, with the unique chemical specificity of mid-IR spectroscopy, together with the extra degree of freedom enabled by the graphene electro-optical tunability.

The characteristics of our graphene biosensor become more evident by comparing its spectral response to that of a state-of-the-art metallic localized surface plasmon resonance (LSPR) sensor composed of a gold dipole-antenna array (Fig. 3). Both devices are first operated in a spectral range free of protein vibrational modes by setting graphene at Vg = –20 V and designing a gold dipole length L = 2.6 μm (Fig. 3A). Upon protein immobilization, we detect a resonance shift of 160 cm−1 for graphene, which is approximately 6 times the 27 cm−1 shift obtained with gold. Next, the operation spectrum is moved toward the protein amide I and II bands by setting graphene at Vg = –120 V and using a different gold sensor with L = 2.1 μm (Fig. 3B). Clearly, dynamic tunability of graphene is one of its main advantages over gold for surface-enhanced IR absorption (SEIRA), enabling sensing over a broad spectrum with a single device. In addition, for the SEIRA signal corresponding to the amide I band, the graphene sensor features a signal modulation of 27%, which is almost 3 times that observed with the gold sensor (11%).

Fig. 3 Graphene versus gold.

(A) Extinction spectra of graphene and gold nanoantenna arrays before (dashed curves) and after (solid curves) protein bilayer formation for plasmonic resonance peak away from the molecular vibration bands. The gold antennas have dimensions 2.6 μm × 0.2 μm × 0.1 μm while the graphene is biased to Vg = –20 V. The spectral shift of the plasmonic resonance (indicated by horizontal arrows) shows the refractive index sensitivity of the biosensors. (B) Extinction spectra of graphene and gold biosensors after protein formation (thick curves) and fitting (thin curves) for plasmon peak overlapping with the molecular vibration bands. The gold antennas have dimensions 2.1 μm × 0.2 μm × 0.1 μm and the graphene gate voltage is Vg = –120 V. The intensity of the spectral features at amide I and II bands (1660 to 1550 cm−1) indicates the SEIRA sensitivity of the biosensors. (C) Near-field enhancement distribution |E/E0| in the plasmonic sensors operating at 1600 cm−1 resonance frequency. (D) Percentage of space-integrated near-field intensity confined within a volume extending a distance d outside the nanoantenna. Inset shows a zoom-in for d between 0 and 40 nm.

The large spectral shifts and absorption signals confirm the unprecedented sensitivity of our graphene biosensor to the complex refractive index of the target molecule. For similar IR-frequency plasmons, the graphene atomic thickness leads to a higher confinement, resulting in a much larger spatial overlap between the mid-IR plasmonic field and the analyte. Figure 3C shows the near-field distribution of LSPR modes in graphene nanoribbons and gold dipole arrays calculated with a finite-element method. The field hotspots are located at the endpoints of the gold dipole and along the edges of the graphene nanoribbon. By computing the percentage of near-field intensity confined within a given distance d from the structure (Fig. 3D), we observe that 90% of the mode energy is confined within 15 nm from the graphene surface, whereas the same percentage is spread over a distance 500 nm away from the gold surface, thus confirming the tighter field confinement of graphene in the mid-IR. As the biosensing signal comes only from the field inside the target volume, we also calculate the field overlap with an 8-nm-thick protein bilayer, which is 29% for graphene versus only 4% for gold. The near-field intensity overlap can be experimentally extracted as the ratio of the relative resonance shift (Δω/ω) and the permittivity variation (εp – εair) (24). This estimate yields 26% and 5% field overlap for graphene and gold, in good agreement with simulations (see above). These results demonstrate the ability of graphene to provide stronger light-protein interactions beyond state-of-the-art metallic plasmonic sensors; further improvement in the graphene quality should lead to even better sensitivity and spectral resolution.

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S3

Reference (25, 26)

References and Notes

  1. See supplementary materials on Science Online.
  2. Acknowledgments: Supported by European Commission grants FP7-IEF-2013-625673-GRYPHON, Graphene Flagship CNECT-ICT-604391, and FP7-ICT-2013-613024-GRASP; the Spanish Ministry of Economy and Competitiveness (MINECO) “Fondo Europeo de Desarrollo Regional” (FEDER) through grant TEC2013-46168-R; NATO’s Public Diplomacy Division in the framework of “Science for Peace”; European Union’s Horizon 2020 research and innovation program under grant agreement No 644956; the Swiss National Science Foundation through project 133583; and Fundació Privada Cellex, the Severo Ochoa Program, and the Ramon y Cajal fellowship program. We also acknowledge École Polytechnique Fédérale de Lausanne and Center of MicroNano Technology for financial support and nanofabrication. This paper is dedicated to the memory of our friend and colleague, Julien Perruisseau-Carrier.
View Abstract

Navigate This Article