The resonance condition of surface plasmon resonance (SPR) is very sensitive to the refractive index near the metal surface. According to this feature, SPR can be applied to provide a no-labeling and real-time biosensor in which the reactions of biomolecules are transduced as the change of refractive index to optical signal. As introduced in Chapter 2, some practical biosensors based on the propagating SPR with a flat gold film sputtered on a glass substrate have been developed and commercialized already. These instruments, however, require the optical setup of a total internal reflection to generate SPR. Furthermore, they need the rigorous control of flow system and inner temperature as they are also sensitive to the background noise factors in sample such as floating molecules and temperature variation. It results in commercialized SPR equipments that are large and very expensive or some cheap and small SPR equipments that are not satisfactory in terms of their sensitivity and usability. On the other hand, LSPR (localized SPR) is expected to eliminate the ambient noise factors due to its further confinement of sensing region. Though new sensing systems based on LSPR have been keenly studied, almost all of them cannot reach a practical use yet. One of the reasons to hog-tie LSPR commercialization is the reproducibility and cost of sensor device. While very high reproducible and low-cost devices are required especially for the disposable use, there still remains challenging barriers in the

device fabrication of metal nanostructures for LSPR generation. For instance, metal colloids immobilized on a substrate are commonly used as a sensor substrate. To realize a uniform quality in colloid diameter and shape, high process control in deoxidization of metal ions is necessary. In addition, the uniform immobilization for avoiding aggregations and density fluctuations of colloids is still a challenge in mass production. As a stable nanofabrication method, electron beam lithography can be another candidate. The patterns are, however, produced by scanning single electron beam on a wafer and it takes a long time to complete the whole pattern that results in a high cost of devices. Other methods such as nanosphere lithography also have the difficulties in the pattern reproducibility and process throughput. On the other side, a nanoimprint technology is focused as a novel fabrication method to produce nanostructures that are much smaller than 100 nm on a high process throughput as introduced in Chapter 1. Our group has been trying to apply this fabrication technique for preparing LSPR sensor device and to realize a small sized, low-cost and high sensitive biosensor. In the conventional LSPR, the metal free electrons are necessarily encaged in nanoscale region smaller than the diffraction limit of light since the metal configurations (colloids, nanorods, and so on) are confined in the metal island structure. For preparing the substrate by a very simple process with the nanoimprint method, the metal should be a continuous layer that is different from the conventional LSPR sensor. To realize this concept, the research was started by exploring the original nanoconfiguration that can hold LSPR characteristic with a continuous metal layer. And we have found that the periodic nanogap structure in a specific dimension can generate a standing plasmon resonance inside the nanogap. Also, the area of the intensified electric field can be easily tuned up by adjusting the dimensional parameters. The characteristics are found to be very suitable especially for the biosensor. In this chapter, the design and fabrication procedure of this device will be presented. 4.1 Design of Nanoimprint BiosensorThe eigenmodal solutions of LSPR are proved only in the case that the metal nanostructure is very simple (e.g., sphere and oval sphere). To design and analyze the plasmon resonance on the

arbitrary structure, photonic simulation tools play an important role. The basic design method and result of nanoimprint biosensor are introduced in this section. 4.1.1 Simulation Methods

To simulate the optical response characteristics of nanosized objects, not only the nature of the “ray” of light but also the nature of “wave” must be considered. Two simulation methods have been used for the optical design and analysis of the nanoimprint biosensor. The dynamic analyses were carried out by using FDTD (finite-difference time-domain) method. And the static analyses were conducted by RCWA (rigorous coupled-wave analysis) method. The features and overview of these simulation methods are described in this section. FDTD method In FDTD method, space is divided into the mesh, which is called Yee mesh. And the Maxwell’s equations are solved for each mesh in each time step. This method has been used primarily to design photonic devices such as photonic crystal. The advantages of this method are that an arbitrary structure can be analyzed in principle and that rigorous dynamic simulation can be conducted based on the Maxwell’s equations without any approximation. On the other hand, the accuracy depends on the fineness of the mesh size and time step. This means that larger computer memory and longer simulation period are necessary to demonstrate the simulation with higher accuracy. To overcome this problem, continuous improvements are accumulated in this method. When the simulation model is a periodic structure, the simulation space can be enormously downsized by setting the boundary condition as the periodic one. In recent simulation tools, the mesh sizes are nonuniform and are adjusted according to the structural variations. These techniques are very helpful especially in the case where the simulation space is three dimensional and the structural patterns are too small when compared with the whole object. In the analysis of SPR, it is necessary to simulate the behavior of the free electron in metal in particular. To realize it, some approaches are addressed. As a unique example, a large number of charged particles are prepared in the Yee mesh to simulate the electrons and ions. And the dynamic equations of these charged

particles and Maxwell’s equations are solved simultaneously (Usui et al., 2001). Although this method is a very promising technique since the behavior of each ions and electrons can be observed by this method, the large computer resources and simulation time is, however, inevitable. In another method, we have simulated the characteristic of the metal by adopting the Lorentz-Drude model. The characteristics of the metal are depicted in the complex dielectric coefficient using this method. When the pattern dimension is larger than the Debye length, the simulation result can be comparatively reasonable. RCWA method In RCWA method, the distribution of the dielectric constant of the periodic object is represented by a Fourier series expansion (Jarem and Banerjee, 1998). The static result can be calculated using this method. This has been generally used for analyzing the optical grating patterns as they are inherently periodic structure (Chambers and Nordin, 1999; Jarem, 2002). Since Maxwell’s equations are solved exactly, the analysis of SPR phenomena whose simulation model includes the complex dielectric constant is possible as long as the structure is periodic. It should be noticed that the analysis accuracy is dependent on the number of harmonic expansion term in Fourier series expansion. The advantage of this method is that relatively faster simulation analysis with smaller computer memory is possible as it deals only with a static and periodic model. However, to expect higher accuracy, larger number of harmonic expansion terms is necessary, which means longer simulation time and larger computer memory are demanded. The determination of the optimal simulation parameters is sometimes required before conducting simulation. In our work, these two methods are used complementarily. RCWA method is mainly used for analyzing the static result such as the calculation of reflection spectra. And FDTD method is used for examining the dynamic phenomena in detail. 4.1.2 Design Concept In the conventional LSPR sensor devices, gold colloid is mainly used to generate LSPR. Though they are relatively easy to be designed

and to be analyzed, the fabrication process stability and process cost are difficult issues for a practical use. As described above, we have proposed a different type of LSPR biosensor based on the nanoimprint fabrication process. The basic concept of this method is that the enhanced resonant electric field is localized around the periodic metal nanopatterns prepared by the nanoimprint and subsequent sputtering processes. Though this seems to be very simple, there are some bigger challenging points to achieve it. For instance, in the conventional LSPR, the electric field is confined in the metal island structure such as colloid, triangle pole, and so on. As a result, the localization of electric field is realized when the dimension of island structure is adequately smaller than the diffraction limit of light. If the metal film is covered continuously to the sensor surface, free electrons can basically move freely inside the film, and the confinement effect of free electrons seems to be challenging (Fig. 4.1).