Atomic force microscopy (AFM) is a high-resolution scanning probe technique widely used to quantitatively measure surface roughness and morphology in semiconductor materials. Unlike electron microscopy methods, AFM does not rely on electron beams but instead uses a sharp probe mounted on a cantilever to physically scan the surface with nanoscale precision. The interaction forces between the probe tip and the sample surface are measured, allowing for three-dimensional topographic mapping. AFM operates in multiple modes, including contact mode, tapping mode, and non-contact mode, each suited for different material properties and measurement requirements.

Surface roughness parameters such as Ra, Rq, and Rz are derived from AFM height data to quantify topography. Ra, the arithmetic average roughness, is calculated by averaging the absolute deviations of height values from the mean plane over the scanned area. Rq, the root mean square roughness, provides a more sensitive measure by squaring deviations before averaging, making it more responsive to extreme peaks and valleys. Rz, the average maximum height, measures the average difference between the highest peaks and lowest valleys over small sampling lengths. These parameters are critical for evaluating semiconductor surfaces, where roughness can impact device performance, interfacial adhesion, and thin-film uniformity.

AFM offers several advantages over optical profilometry and stylus-based methods. Optical profilometry relies on light interference or reflection, which can suffer from diffraction limits and difficulty in resolving steep sidewalls or highly reflective surfaces. Stylus profilometers physically drag a tip across the surface, risking damage to delicate samples and lacking the lateral resolution of AFM. In contrast, AFM achieves sub-nanometer vertical resolution and lateral resolution down to a few nanometers, making it ideal for analyzing nanoscale features. Additionally, AFM does not require conductive coatings or vacuum conditions, enabling measurements in ambient air or liquid environments.

Thin films in semiconductor applications often require precise roughness control to ensure uniformity and minimize defects. AFM can characterize films such as chemical vapor-deposited silicon nitride or atomic layer-deposited oxides, where roughness impacts dielectric properties and interfacial quality. For example, studies have shown that ALD-grown aluminum oxide films with Ra values below 0.5 nm exhibit superior breakdown voltages compared to rougher films. AFM can also detect pinholes or grain boundaries in polycrystalline films, which are critical for applications like gate dielectrics or barrier layers.

Nanostructures, including quantum dots, nanowires, and nanopillars, present unique challenges for morphology analysis due to their small dimensions and high aspect ratios. AFM can resolve individual nanostructures and measure their height, diameter, and spacing with high accuracy. For instance, in GaN nanowire arrays, AFM has been used to confirm uniform heights within ±5% deviation, a key parameter for optoelectronic device performance. The technique is also valuable for assessing sidewall roughness in etched nanostructures, which affects scattering losses in photonic devices.

Patterned semiconductor wafers, such as those used in integrated circuit manufacturing, require stringent roughness control to ensure lithographic fidelity and device yield. AFM can profile trenches, vias, and contact holes with sub-nanometer precision, identifying edge roughness or residual material that may impair etching or deposition processes. In extreme ultraviolet (EUV) lithography, line-edge roughness below 1 nm is often targeted, and AFM provides direct measurements without the averaging effects seen in optical techniques. Furthermore, AFM can map step heights in multilayer structures, such as fin field-effect transistor (FinFET) fins, where conformal growth and etching processes must be tightly controlled.

The quantitative nature of AFM data enables statistical analysis of surface features, such as power spectral density (PSD) calculations, which decompose roughness into spatial frequency components. This is particularly useful for identifying periodic patterns or distinguishing between intrinsic material roughness and process-induced artifacts. For example, PSD analysis of chemical-mechanical polished wafers can reveal polishing scratches or slurry particles that contribute to high-frequency roughness.

Despite its advantages, AFM has limitations, including slower scan speeds compared to optical methods and potential tip artifacts when probing very sharp features. Careful tip selection and calibration are necessary to ensure accurate measurements, especially for high-aspect-ratio structures. However, ongoing advancements in high-speed AFM and automated tip characterization have mitigated some of these challenges.

In summary, AFM provides unparalleled quantitative data on semiconductor surface roughness and morphology through parameters like Ra, Rq, and Rz. Its high resolution, versatility, and ability to operate under ambient conditions make it indispensable for analyzing thin films, nanostructures, and patterned wafers. By offering direct topographic measurements without the limitations of optical or stylus methods, AFM plays a critical role in semiconductor research and manufacturing quality control.