In conventional far-field imaging, where all points in space are constantly illuminated during imaging, it is conventional to define the best theoretical lateral resolution by the Rayleigh Criterion, or that distance between two point scatterers where the central Airy disc of one overlaps with the first dark ring of the second, leading to a significant intensity dip (at least 26%) between the two: r = (0.61 x wavelength) / (numerical aperture) – say two thirds of a micron. The resolution in z (focus) is far worse, being on the order of tens of microns.
Confocal microscopy is a far-field technique designed to improve resolution by illuminating an isolated diffraction-limited volume and collecting the scattered light from the same volume: resolution here is conventionally defined as the full-width-half-height dimension of the measured scattering response: now r = (0.41 x wavelength) / (numerical aperture) perhaps a bit better than a third of a micron for lateral resolution and only about three times worse in z (focus). Near-field techniques aim to reduce the illuminated volume still further.
Near-field scanning optical microscopy (NSOM or SNOM, depending on whether you follow IBM or Bell Lab’s lead) employs a small aperture, at the very tip, in the metallization of a drawn optical fiber that allows light to leak out a very short distance (an evanescent filed), thus illuminating a volume the diameter of the aperture and only a couple of nanometers deep. Scattered light from this volume is then collected in the far-field.
Of course this process may be reversed with collection through the aperture, from the same volume that is illuminated in the far-field. It is even possible to both illuminate and collect through the fiber, although each pass attenuates the beam by some million times, so this can only be useful for very strong scattering.
Height feedback to track surface topography and maintain a fixed tip / surface separation, has traditionally been by shear-force feedback, where the tip is oscillated laterally while fixed to one tine of a tuning fork at resonance and fixing the phase shift to maintain constant interaction with the surface; this is slow somewhat insensitive and leads to aperture wear (enlargement) by collision with any significant topography of the surface. Lately, other schemes have been devised using a similar aperture at the apex of a hollow AFM tip to perform this type of NSOM while allowing somewhat more robust feedback.
Because smaller apertures that illuminate a smaller volume transmit fewer photons, this technique is limited, in practice, to resolutions of 100nm (or with more difficulty) perhaps half that. Also there are significant artefacts associated with, for instance: variable throughput of the aperture when held close to a surface or by variation in feedback height due to the width (aperture diameter + twice the metallization thickness) interacting with surface roughness; and polarization variation across the aperture. It should be noted that scanning these apertures over an edge (even in the absence of topography) generates an artefact with a size of half the aperture width, which is sometimes confused with resolution.
Scattering-SNOM (sSNOM), of which TERS is an example, uses a physical tip to define the spectroscopy volume passively, by its physical presence, or actively, by its photonic effects. For instance, a physical tip scanned within nanometers of the surface of a TIRF crystal, where a totallyinternally- reflected excitation beam provides a shallow evanescent field will generate light propagating to the far-field from the volume immediately surrounding the physical tip which, therefore, defines the resolution.
TERS and associated TEOS techniques employ a metal coating whose photonic activity causes it to act as an antenna which mediates energy transfer from the far-field excitation beam to the near-field. So-called “hot spots” at the tip are non-propogating fields of extremely high electric dipole, which effectively illuminate a volume with only a couple of nanometers in diameter. Whilst it is still necessary to deconvolve the diffraction-limited far-field response from the nanometer-scale near-field response, this is possible and in the case of strong scatterers not strictly necessary since the tip-enhancement can be on the order of 107, or more.
There is currently much discussion among practitioners on definitions of resolution for TEOS techniques and there remains theoretical work to be done to fully understand which of the contrast mechanisms seen are useful.
In this example presented Fig. 7, the TERS image (100 × 100 nm scanning area, with a pixel step size of 1.3 nm, total acquisition time < 9 min, 100 ms integration time per pixel), showed nanoscale chemical imaging of a single carbon nanotube with a spatial resolution of 8 nm, confirmed from the section analysis of the TERS intensity.
As a first approximation, this resolution is dependent on the radius of curvature of the TERS tip itself, those "8 nanometers" result from convolution of the tip radius. Thus, the achievable TERS resolution can be approximated and considered equal to half the radius of curvature of the tip.