Image fidelity for single-layer and multi-layer silver superlenses

In response to increasing interest in the area of subdiffraction-limited near-field imaging, the performance of several different realizable and theoretical superresolving silver-based lenses is simulated for a variety of different input object profiles. A computationally-efficient T-matrix technique is used to model the lenses, which consist of layers of silver with total width of 40 nm sandwiched between layers of polymethyl methacrylate and silicon dioxide. The lenses are exposed to nonperiodic bright- and dark-slit input patterns, with feature size varied between 1 nm and 2.5 microm. The performance of the lenses is characterized in terms of transfer function, contrast profile, error profile, and input-to-output correlation. It is shown that increasing the number of layers in a lens increases the lens' transmission coefficients at high spatial frequencies; however, this does not always lead to better imaging performance. The main reasons for this are lens-specific resonances that distort features at certain spatial frequencies, and the increased attenuation of the DC component of transmitted images, which reduces image fidelity, particularly for dark-line features. This suggests that, to achieve optimum results, the design of the superresolving lens system should take into account the characteristics of the images that it is expected to transmit.