Scattering-type-scanning near-field optical microscopy (s-SNOM) is one of the most widely used near-field optical microscopy techniques. It enables nanoscale spatial resolution of the surface of a sample or at a specific depth below the surface without causing damage to the sample. Consequently, its application spans across diverse scientific disciplines, such as physics, chemistry, and biology. Through the photoexcitation at the tip-sample boundary, s-SNOM simultaneously generates weak near-field-scattering signals containing crucial near-field information and intense background scattering signals that can obscure this information. The challenge lies in suppressing the background noise, and various near-field signal extraction techniques have been proposed to address this issue. These techniques play a pivotal role in extracting genuine near-field information, enabling a more precise clarification of the sub-wavelength characteristics of samples in multiple research areas.

Such potency in revealing the intricate nanoscale underscores the importance of mastering these techniques. In the pursuit of this objective, it is imperative to extensively investigate the mechanism for generating near-field-scattering signals and techniques for extracting near-field optical signals and their underlying principles. Such detailed investigations help refine near-field extraction techniques and are significant in observing the physical changes that occur during s-SNOM experiments. This knowledge facilitates the thorough analysis of near-field optical research findings and their practical applications, promoting advancements in various scientific disciplines.

To advance the techniques for extracting near-field optical signals, researchers have examined the generation mechanism of near-field-scattering signals inherent in the near-field detection process. They established several physical models, including the point-dipole and finite-dipole models, to analyze tip-sample interactions. These models enable a more accurate analysis of near-field experimental results, even without resorting to computationally tasking simulation methods. In addition, research on these models provides theoretical support for the development of near-field extraction techniques: the near-field-scattering signal exhibits a more pronounced variation in higher-order harmonics with tip vibration, which distinguishes it from the background scattering signal. This dissimilarity in response to periodic modulation of the tip allows for effective demodulation at higher harmonics of the vibration frequency of the tip using a lock-in amplifier, achieving signal suppression against background noise. Building upon demodulation at higher harmonics, researchers have proposed techniques for extracting near-field optical signals, including intensity and electric-field detection techniques.

The intensity detection techniques, which were earlier and more extensively adopted, include prevalent techniques, such as the self-homodyne, (quadrature) homodyne, and pseudoheterodyne detection techniques. In addition, less-used techniques, such as phase-shifting interferometry, synthetic optical holography, and self-heterodyne detection, were developed. These intensity detection techniques often involve the addition of components, such as plane mirrors and acousto-optic modulators, to the interferometer system and generate reference signals with a specific phase or frequency. Such modifications effectively suppress the background noise, enabling the acquisition of a high signal-to-noise ratio and pristine near-field information. Some of these methods also facilitate the separation of near-field information into amplitude and phase components, which is crucial for the advancement of s-SNOM applications.

Electric-field detection techniques, including electro-optic sampling and photoconductive antennas, are applied in radio frequency and terahertz detection bands because of their simple experimental setups and electric-field detection characteristics. Although electric-field detection techniques may not achieve the same extent of background noise suppression as the intensity detection techniques, these techniques contribute to diversifying technological options for experiments employing s-SNOM, further enriching the potential applications of this innovative microscopy.

In this review, we first clarify the mechanism of near-field-scattering signals using tip-sample physical models and present an overview of the research development in tip-sample interactions. Subsequently, various techniques for extracting near-field optical signals developed to date are systematically introduced, with emphasis on crucial techniques, such as self-homodyne, (quadrature) homodyne, heterodyne, and pseudoheterodyne detection. Through comparative experiments, the distinctions between several widely used techniques, including self-homodyne, quadrature homodyne, and pseudoheterodyne detection, are intuitively demonstrated, consistent with the theoretical expectations. Finally, the state-of-the-art advancements in near-field optical signal extraction are presented, along with a prospective outlook on their future development.

Currently, several mainstream techniques for extracting near-field optical signals have distinct advantages in terms of the signal-to-noise ratio, experimental efficiency, and complexity, helping researchers to make selections based on the specific requirements of their experiments. Consequently, research efforts in near-field extraction techniques predominantly concentrate on the advancement of existing techniques. This can be achieved by incorporating additional computation or experimental components. A typical example is the pseudoheterodyne interferometry for multicolor near-field imaging, which achieves simultaneous imaging of multiple spectra in a single experiment by calculating coefficient variations corresponding to the separation of the amplitude and phase at different detection wavelengths. Enhancements such as this maintain the original near-field optical signal extraction capabilities and address a broader range of experimental requirements, such as the improvement of experimental efficiency and the increase in imaging resolution. These advancements contribute to the improvement of experimental efficiency and the richness of information acquisition in s-SNOM experiments, encapsulating the future trajectory of near-field optical signal extraction techniques.