Access Official Technical PDF

Share Summary:LinkedInEmail

As semiconductor devices approach dimensions where individual atoms and defects can influence performance, knowing the average dopant level is no longer enough. Engineers need to understand how dopant concentration varies across a device and, ideally, with depth. Semiconductor dopant profiling with scanning probes offers a way to measure local electrical properties at small scales, but turning an instrument signal into a reliable concentration requires careful modeling and calibration.

Scanning capacitance microscopy (SCM) and near-field scanning microwave microscopy (NSMM) are two radio-frequency techniques used for this purpose. Both can probe local electrical behavior, while NSMM also offers sensitivity to subsurface features. Their measurements are complementary to methods that characterize surface potential, work function, or resistance.

Why local dopant profiling matters

Device scaling makes performance more sensitive to the location of dopants and defects, not simply their overall concentration. A spatially resolved profile can help characterize semiconductor structures and assess whether dopant distributions align with device dimensions. The need extends beyond silicon to emerging low-dimensional materials, although profiling approaches developed for silicon may not transfer directly to other material systems.

Many scanning-probe dopant measurements are effectively two-dimensional: they characterize the surface or near-surface region. To investigate depth-dependent profiles with these methods, a sample may need to be prepared as a cross section, for example through focused ion beam milling. NSMM’s subsurface sensitivity makes it a candidate for three-dimensional measurements, but quantitative depth profiling still depends on suitable models and calibration.

How the probe-sample junction affects measurements

In many practical measurements, a metal probe contacts a semiconductor covered by a thin dielectric, such as native oxide. This forms a metal-oxide-semiconductor (MOS) junction. The measured capacitance depends on the oxide and semiconductor, as well as probe geometry and parasitic electrical contributions. A simple parallel-plate capacitor can provide an initial approximation, but it does not capture the complete tip shape or the semiconductor’s response.

In a semiconductor, charge redistribution near the interface can create a depletion region. Its width and capacitance vary with applied bias and dopant density. Measuring capacitance as a function of voltage (C-V), or its differential response (dC/dV), therefore provides information that can be used to estimate local doping when relevant material parameters are known.

With an oxide present, the junction response includes both oxide and depletion capacitance. For an n-type semiconductor, positive bias tends to produce accumulation of electrons at the interface; reducing the bias can move the device into depletion; and sufficiently negative bias can produce inversion, with holes becoming the dominant carriers near the interface. The voltage polarity is reversed for p-type material. These regimes shape the measured C-V curve and provide context for interpreting it.

Semiconductor dopant profiling with scanning capacitance microscopy

SCM detects changes in local capacitance, commonly using a radio-frequency sensor and a small alternating voltage applied alongside a swept DC bias. A lock-in amplifier measures the capacitance change associated with the modulation. The resulting signal is related to local electrical properties, but extracting a dopant profile requires a model of the tip-sample interaction.

Model inputs can include tip shape, oxide thickness and permittivity, and operating voltage. Simplified approaches can reduce computation time, while three-dimensional finite-element calculations can provide a more detailed treatment of the electric potential and charge distribution. Model choice matters: different assumptions can produce different conversions from dC/dV to estimated dopant concentration. Lateral charge redistribution, air gaps, and condensed water can also affect the response in some measurement conditions.

Reference samples make measurements quantitative

Without calibration, SCM signals are generally more useful for comparing relative changes than for assigning absolute dopant concentrations. A useful reference is topographically smooth and contains regions with known, contrasting doping levels across the range of interest. Multilayer samples with uniformly doped stripes can provide these regions in cross section. Separate p-type and n-type standards may be used.

Comparing SCM measurements on a reference sample with an established method such as secondary ion mass spectrometry (SIMS) can help establish a conversion to absolute concentration. Reported best-case lateral resolution is on the order of 10 nm, though sensitivity and resolution depend strongly on tip geometry.

Using NSMM for local capacitance-voltage measurements

NSMM can measure the voltage-dependent reflection coefficient of a probe-sample junction. To use that response for dopant profiling, the workflow has three essential stages:

  1. Measure the reflection response: Record the complex reflection coefficient, S11, as the tip bias changes.
  2. Calibrate the response: Use reference features with known impedance and a model of the tip-sample interaction to convert signal changes into capacitance.
  3. Estimate dopant concentration: Interpret the calibrated C-V or dC/dV data with an appropriate semiconductor model.

One calibration approach uses microcapacitors formed by conductive pads on oxide layers of known thickness. Measurements on a pad and on the adjacent oxide provide a relative signal that can be related to total capacitance. The method depends on assumptions about the oxide, reference geometry, and parasitic capacitance. The reference and test sample should have sufficiently similar electromagnetic properties for the model and calibration to apply.

Calibration also needs to remain valid over time. Microwave measurements can respond to environmental changes, and mechanical wear can alter tip geometry—and therefore capacitance. Changes to the setup or probe may require calibration to be checked or repeated.

Complementary scanning-probe methods

Other techniques provide different views of semiconductor properties. Scanning spreading-resistance microscopy measures DC resistance between a conductive probe and a backside contact; under suitable contact conditions, local spreading resistance is inversely related to carrier concentration. Scanning Kelvin probe microscopy measures surface potential, from which work function and, with an appropriate model, carrier concentration can be assessed. Scanning tunneling spectroscopy measures tunneling conductance as a function of voltage, enabling local electronic density-of-states measurements and characterization of individual atoms under suitable conditions.

These methods are generally surface-sensitive and do not directly provide depth profiles without cross-sectional sample preparation. Like SCM and NSMM, quantitative interpretation depends on calibration, reference samples, and physical models suited to the material and measurement geometry.

Practical implications for metrology

For engineers and semiconductor metrology teams, the key lesson is that a high-resolution image is not automatically a quantitative dopant map. The instrument response must be connected to concentration through a defensible model and calibrated against suitable reference structures. Probe geometry, oxide properties, parasitic electrical paths, and interface behavior all influence that connection.

Established MOS-based approaches have been widely applied to silicon, whose silicon-silicon oxide interface is comparatively well behaved. Compound semiconductors such as GaN and AlGaN can have interface states, trapped charges, and Fermi-level pinning that affect surface band bending and capacitance. Extending profiling methods to these materials therefore requires models that account for their interface physics rather than relying uncritically on idealized silicon assumptions.

SCM and NSMM can add valuable local electrical information to semiconductor characterization, especially when paired with well-designed references and carefully validated analysis. Selecting the method, sample preparation, and model together is essential to distinguish relative contrast from an absolute dopant profile and to understand what the measurement can—and cannot—establish.

Access Official Technical PDF

About The Author