Quantitative BSE Imaging: Methods & Insights

• Quantitative backscattered electron imaging is a family of methods that measure BSE signals through calibrated detector geometries and normalization processes to yield quantitative material descriptors.
• It integrates approaches like ROI normalization, event-counting detection, and energy-resolved measurements to reveal microstructural details in advanced materials.
• Applications include anti-phase-domain mapping, particle analysis, and multimodal fusion with EDS/EBSD, offering actionable insights despite challenges in absolute calibration.

Quantitative backscattered electron imaging is the use of backscattered-electron (BSE) signals in electron microscopy as measurable observables rather than solely as qualitative image contrast. In the literature represented here, the term encompasses several distinct but connected practices: direct or indirect estimation of the backscattering coefficient and reflected-electron spectra; ROI-based normalization of orientation contrast in scanning electron microscopy (SEM); event-counted, pixelated, and angularly selectable BSE detection with direct electron detectors; large-area BSE mosaicking converted into spatially resolved microstructural descriptors; and multimodal workflows in which BSE images are fused with EDS or EBSD for constituent-specific or boundary-specific quantification (Taioli et al., 2024). Rather than denoting a single standardized protocol, it denotes a family of methods in which detector geometry, beam energy, angle, signal normalization, and post-processing are controlled well enough that BSE intensity, spectrum, or distribution can be interpreted comparatively or, in some settings, physically.

1. Definitions, scope, and measurable quantities

Within this body of work, the most formal BSE quantity is the backscattering coefficient. One review defines the electron backscattering coefficient as the ratio of the primary electrons of the beam, which emerge from the surface of an irradiated solid with a kinetic energy of more than 50 eV, to the total number of primary electrons incident on the target (Taioli et al., 2024). The same review also treats the total electron yield as the sum of secondary electron yield and backscattering coefficient, and emphasizes that the 50 eV division is operational rather than fundamental, since there are secondary electrons with energies above a predefined threshold and backscattered electrons below 50 eV (Taioli et al., 2024). This is central for quantitative imaging because any detector that does not enforce a strict energy threshold measures a signal defined jointly by specimen physics and detector acceptance.

The quantitative observable need not be a calibrated backscattering coefficient. In direct orientation contrast imaging of anti-phase domains in zinc-blende III-V materials, the measured quantity is an ROI-based normalized contrast derived from in-lens SEM signals whose physical origin is tied to backscattering and channeling, even though the detected electron population may include BSE-induced secondary electrons (Corre et al., 17 Apr 2026). In direct-electron-detector SEM platforms, the observable can instead be the count of individual detected events with detector-plane position and time-of-arrival, from which BSE or ECCI images are reconstructed computationally (El-Khairaoui et al., 18 May 2026). In large-area morphology mapping, the measurable outputs are not absolute BSE yields but tile-wise descriptors such as lamellar structure proportion, lamellar width, and BCC phase proportion extracted from stitched BSE panoramas (Lang et al., 2024). In correlated EBSD–BSE workflows, BSE becomes a quantitative particle-measurement channel providing particle size, count, spacing, and boundary association below the applied EBSD step size (Ånes et al., 2022).

This diversity suggests that “quantitative” in this domain has several levels. At the most stringent level it refers to physically defined yields or spectra. At a comparative level it refers to normalized contrasts, count images, or morphology-derived area fractions that are reproducible within a specified detector and acquisition geometry. Several papers explicitly state that they do not provide absolute detector calibration to BSE yield, and therefore support comparative rather than absolute quantification (Corre et al., 17 Apr 2026).

2. Contrast formation and transport physics

The physical basis of quantitative BSE imaging is the coupling of elastic angular redistribution, inelastic energy loss, and escape from a finite interaction volume. A review of secondary and backscattered electron yields expresses the differential elastic scattering cross section as

$\frac{d\sigma_{\mathrm{el}}{d\Omega}=|f|^2+|g|^2$

and the elastic mean free path as

$$\lambda_{\mathrm{el}}=\frac{1}{N \sigma_{\mathrm{el}}}\,,$$

with the total mean free path

$$\lambda=\frac{1}{N\,(\sigma_{\mathrm{el}}+\sigma_{\mathrm{inel}})}\,.$$

The same review gives the differential inverse inelastic mean free path as

$\frac{d\lambda_{\mathrm{inel}}^{-1}}{dW}\,=\,\frac{1}{\pi a_0 E}\,\int_{q_-}^{q_+}\frac{dq}{q}\mbox{Im}\left[\frac{1}{\bar\varepsilon(q,W)}\right]\,,$

linking BSE spectral degradation directly to the dielectric response and energy loss function of the solid (Taioli et al., 2024). In this framework, BSE grayscale is a projection of an underlying energy–angle distribution whose shape depends on elastic cross sections, inelastic losses, and the detector’s acceptance.

Analytical models make the atomic-number dependence explicit. One expression for the backscattering coefficient is

$\eta\,=\,\frac{a-1+0.5^a}{a+1}\,,$

where $a=0.045Z$ (Taioli et al., 2024). For layered or evolving systems, Monte Carlo transport studies show that BSE intensity also depends strongly on thickness and density. In simulations of 5 keV electron transport during electron-beam-induced deposition of tungsten oxycarbide nanostructures on SiO$_2$, the interaction volume shifts from a substrate-controlled regime to a deposit-controlled regime as the deposit thickens, and a larger deposit density leads to enhanced electron backscattering (Salvat-Pujol et al., 2013). The same study reports attenuation depths of about 500 nm in SiO$_2$ and about 150 nm in W, together with an approximately factor $\sim 2$ increase in emitted/backscattered intensity for thick W-containing deposits relative to substrate alone (Salvat-Pujol et al., 2013). This is a direct quantitative reminder that BSE contrast in heterogeneous systems cannot be reduced to composition alone.

Orientation-sensitive BSE contrast adds a diffraction and channeling term. In anti-phase-domain imaging of non-centrosymmetric zinc-blende III-Vs, the orientation/domain contrast is interpreted as directly attributable to a BSE-related channeling/orientation mechanism: opposite crystal polarities bring polar planes such as $(111)$ into different channeling conditions, so one domain yields more backscattering and less channeling than the other near a Bragg condition (Corre et al., 17 Apr 2026). That study explicitly distinguishes this polarity-sensitive orientation contrast from compositional $$\lambda_{\mathrm{el}}=\frac{1}{N \sigma_{\mathrm{el}}}\,,$$0-contrast, since the two domains are the same material (Corre et al., 17 Apr 2026). A more angularly resolved version of the same idea appears in EBSD-based “spherical-angular dark field imaging,” where virtual images are formed by integrating intensity within one Bragg angle of a selected band profile,

$$\lambda_{\mathrm{el}}=\frac{1}{N \sigma_{\mathrm{el}}}\,,$$1

and the normalized phase contrast is written as

$$\lambda_{\mathrm{el}}=\frac{1}{N \sigma_{\mathrm{el}}}\,,$$2

That work treats EBSD patterns as angularly resolved backscattered-electron distributions and demonstrates diffraction-conditioned, virtual BSE-like imaging (McAuliffe et al., 2020).

3. Detector architectures and acquisition geometries

The cited work spans conventional semiconductor or in-lens SEM channels, direct electron detectors, and EBSD-pattern-based virtual detection. In III-V anti-phase-domain imaging, two SEM instruments were used: a Thermo Fisher Verios G4 HP with the Through Lens Detector in both “SE” and “BSE” modes, and a Thermo Fisher Apreo 2C using the T1 in-lens detector (Corre et al., 17 Apr 2026). A key practical finding is that TLD-SE and TLD-BSE modes show similar angular dependence, but the BSE mode exhibits substantially lower contrast than the SE mode, and the Apreo T1 detector is interpreted as being sensitive to secondary electrons induced by backscattered electrons rather than to backscattered electrons themselves (Corre et al., 17 Apr 2026). This sharply separates the origin of contrast from the actual detected electron population.

A more explicit event-counting architecture is provided by a Timepix3-based direct electron detector integrated into a Zeiss Supra 40 SEM. In this platform, the detector can be translated and rotated so that the same hardware is used for conventional EBSD, off-axis RKD, on-axis TKD, off-axis TKD, near-axis TKD, BSE, ECCI, and STEM-in-SEM without detector replacement or hardware modification (El-Khairaoui et al., 18 May 2026). In the BSE configuration, realized in off-axis RKD with $$\lambda_{\mathrm{el}}=\frac{1}{N \sigma_{\mathrm{el}}}\,,$$3, the detector is positioned near the pole piece for tilt-free analysis of flat samples (El-Khairaoui et al., 18 May 2026). The fundamental BSE-relevant detector output is an event stream in which the $$\lambda_{\mathrm{el}}=\frac{1}{N \sigma_{\mathrm{el}}}\,,$$4 detector coordinates and time-of-arrival of each detected electron are recorded, with a temporal resolution of 1.52 ns (El-Khairaoui et al., 18 May 2026). This makes the detector a software-defined collector: diffraction patterns are reconstructed from the $$\lambda_{\mathrm{el}}=\frac{1}{N \sigma_{\mathrm{el}}}\,,$$5 distributions, whereas electron images are reconstructed using time-of-arrival information (El-Khairaoui et al., 18 May 2026).

Monolithic direct electron detectors also enable energy-resolved BSE measurement. In energy-resolved EBSD, a DE-SEMCam monolithic active pixel sensor was calibrated by direct-beam measurements from 7 to 15 keV, yielding representative ADU/keV values of 7.84 at 8 keV, 8.11 at 10 keV, 8.12 at 12 keV, and 8.35 at 14 keV, with detector energy resolution of approximately 1 keV FWHM (Ventura et al., 27 Jul 2025). Under sparse-event conditions, a centroid-based counting algorithm identifies contiguous clusters, sums event ADU, and converts cluster signal to electron energy (Ventura et al., 27 Jul 2025). This is not yet routine general-purpose BSE imaging, but it provides direct experimental access to the energy of individual BSEs contributing to a diffraction pattern.

The practical implication is that detector choice determines what can be quantified. Conventional and in-lens detectors can deliver strong comparative contrast but may mix BSE and BSE-induced SE. Pixelated direct detectors preserve detector-plane distributions and enable virtual sectors or event counting. Energy-sensitive direct detectors add a spectral dimension. EBSD-pattern detectors, when treated as angular maps of backscattered intensity, enable crystallographically defined virtual apertures (Ventura et al., 27 Jul 2025).

4. Quantification workflows, normalization, and image construction

Several distinct quantitative workflows recur in the literature. In anti-phase-domain imaging, the core contrast metric is a difference-over-sum ROI normalization. After linear-transformation renormalization of image intensity to compensate for tilt-dependent detector sensitivity, detector position and finite numerical aperture, and progressive carbon contamination, four ROIs are tracked across tilt angle and the contrast is computed as

$$\lambda_{\mathrm{el}}=\frac{1}{N \sigma_{\mathrm{el}}}\,,$$6

The authors compute $$\lambda_{\mathrm{el}}=\frac{1}{N \sigma_{\mathrm{el}}}\,,$$7 and $$\lambda_{\mathrm{el}}=\frac{1}{N \sigma_{\mathrm{el}}}\,,$$8, and the overall orientation contrast is the average of these two values, with uncertainty reported as a 70% confidence interval (Corre et al., 17 Apr 2026). This is a paradigmatic comparative BSE-style metric: intensity is not absolutely calibrated, but detector and geometry variations are normalized sufficiently to compare beam energy and tilt conditions.

In event-based direct-detection SEM, quantification begins with sparse acquisition of individual electron events. The Timepix3 workflow stores $$\lambda_{\mathrm{el}}=\frac{1}{N \sigma_{\mathrm{el}}}\,,$$9 for each electron, reconstructs diffraction patterns from detector coordinates, and reconstructs reflection-mode images using time-of-arrival (El-Khairaoui et al., 18 May 2026). Because the same raw event stream can be spatially segmented after acquisition, software-defined virtual detectors can be constructed. The authors explicitly state that by applying appropriate spatial selection of backscattered electrons from specific channelling regions of the pattern, high-contrast ECCI micrographs of dislocations can be generated entirely through post-processing (El-Khairaoui et al., 18 May 2026). This is a detector-segmentation strategy rather than an analog-current measurement strategy.

Large-area BSE imaging introduces a different type of quantitative workflow. In automated high-resolution macroscopic BSE mapping of AlCoCrFeNi$$\lambda=\frac{1}{N\,(\sigma_{\mathrm{el}}+\sigma_{\mathrm{inel}})}\,.$$0 eutectic high-entropy alloy, low-definition BSE videos were acquired during stage translation, frames were extracted with 30% to 60% overlap, low-definition frames were restored by an ESRGAN-inspired generator–discriminator architecture, and 3902 images from 18 videos were stitched into a panorama of 123672 $$\lambda=\frac{1}{N\,(\sigma_{\mathrm{el}}+\sigma_{\mathrm{inel}})}\,.$$1 7848 pixels covering 2.75 cm $$\lambda=\frac{1}{N\,(\sigma_{\mathrm{el}}+\sigma_{\mathrm{inel}})}\,.$$2 0.175 cm (Lang et al., 2024). The degradation model is written as

$$\lambda=\frac{1}{N\,(\sigma_{\mathrm{el}}+\sigma_{\mathrm{inel}})}\,.$$3

and, for the two-stage mixed degradation,

$$\lambda=\frac{1}{N\,(\sigma_{\mathrm{el}}+\sigma_{\mathrm{inel}})}\,.$$4

After stitching, the panorama is divided into a $$\lambda=\frac{1}{N\,(\sigma_{\mathrm{el}}+\sigma_{\mathrm{inel}})}\,.$$5 grid of tiles, and each tile is converted into lamellar structure proportion, lamellar width, and BCC phase proportion (Lang et al., 2024). Here the quantified output is not detector yield but microstructure statistics derived from BSE morphology over a calibrated area.

Multimodal segmentation workflows rely on pixel-wise co-registration. At a concrete–epoxy interface, synchronous BSE and EDS maps are represented as matrices $$\lambda=\frac{1}{N\,(\sigma_{\mathrm{el}}+\sigma_{\mathrm{inel}})}\,.$$6 and $$\lambda=\frac{1}{N\,(\sigma_{\mathrm{el}}+\sigma_{\mathrm{inel}})}\,.$$7, and phase masks are built from thresholded logical rules involving Si, Ca, and BSE intensity: $$\lambda=\frac{1}{N\,(\sigma_{\mathrm{el}}+\sigma_{\mathrm{inel}})}\,.$$8 and

$$\lambda=\frac{1}{N\,(\sigma_{\mathrm{el}}+\sigma_{\mathrm{inel}})}\,.$$9

A $\frac{d\lambda_{\mathrm{inel}}^{-1}}{dW}\,=\,\frac{1}{\pi a_0 E}\,\int_{q_-}^{q_+}\frac{dq}{q}\mbox{Im}\left[\frac{1}{\bar\varepsilon(q,W)}\right]\,,$0 averaging kernel with $\frac{d\lambda_{\mathrm{inel}}^{-1}}{dW}\,=\,\frac{1}{\pi a_0 E}\,\int_{q_-}^{q_+}\frac{dq}{q}\mbox{Im}\left[\frac{1}{\bar\varepsilon(q,W)}\right]\,,$1 is then used for denoising by local majority filtering (Żak et al., 2022). This is quantitative BSE imaging in the sense that BSE is one channel in a rule-based classification model rather than a standalone grayscale image.

A further registration-based workflow is used in correlated EBSD–BSE particle analysis. After BSE segmentation in the native-resolution image, the BSE image is fused with a distortion-corrected EBSD map using a thin plate spline transformation with kernel

$\frac{d\lambda_{\mathrm{inel}}^{-1}}{dW}\,=\,\frac{1}{\pi a_0 E}\,\int_{q_-}^{q_+}\frac{dq}{q}\mbox{Im}\left[\frac{1}{\bar\varepsilon(q,W)}\right]\,,$2

Particle size is then reported as a stereologically corrected equivalent circular diameter

$\frac{d\lambda_{\mathrm{inel}}^{-1}}{dW}\,=\,\frac{1}{\pi a_0 E}\,\int_{q_-}^{q_+}\frac{dq}{q}\mbox{Im}\left[\frac{1}{\bar\varepsilon(q,W)}\right]\,,$3

The method also carries a binned map containing the number of particle pixels within each $\frac{d\lambda_{\mathrm{inel}}^{-1}}{dW}\,=\,\frac{1}{\pi a_0 E}\,\int_{q_-}^{q_+}\frac{dq}{q}\mbox{Im}\left[\frac{1}{\bar\varepsilon(q,W)}\right]\,,$4 block so that particle-size information is not destroyed by downsampling onto the EBSD grid (Ånes et al., 2022).

5. Established applications and quantitative outputs

A substantial part of the current literature uses quantitative BSE imaging to extract specific material descriptors rather than absolute electron-optical constants. In polarity-sensitive SEM imaging of III-Vs, the main application is anti-phase-domain mapping and optimization of contrast versus beam energy and tilt (Corre et al., 17 Apr 2026). On a 500 nm CMP-polished GaP/Si layer, geometric tilt correction, multistep binarization, and edge detection were used to derive surface-polarity and anti-phase-boundary statistics. The average surface polarity was written as

$\frac{d\lambda_{\mathrm{inel}}^{-1}}{dW}\,=\,\frac{1}{\pi a_0 E}\,\int_{q_-}^{q_+}\frac{dq}{q}\mbox{Im}\left[\frac{1}{\bar\varepsilon(q,W)}\right]\,,$5

the mean same-phase spacing as

$\frac{d\lambda_{\mathrm{inel}}^{-1}}{dW}\,=\,\frac{1}{\pi a_0 E}\,\int_{q_-}^{q_+}\frac{dq}{q}\mbox{Im}\left[\frac{1}{\bar\varepsilon(q,W)}\right]\,,$6

and the autocorrelation-based correlation length as

$\frac{d\lambda_{\mathrm{inel}}^{-1}}{dW}\,=\,\frac{1}{\pi a_0 E}\,\int_{q_-}^{q_+}\frac{dq}{q}\mbox{Im}\left[\frac{1}{\bar\varepsilon(q,W)}\right]\,,$7

Hough-transform analysis of the edge map then yielded anti-phase-boundary orientation densities, with strong occurrence along $\frac{d\lambda_{\mathrm{inel}}^{-1}}{dW}\,=\,\frac{1}{\pi a_0 E}\,\int_{q_-}^{q_+}\frac{dq}{q}\mbox{Im}\left[\frac{1}{\bar\varepsilon(q,W)}\right]\,,$8, $\frac{d\lambda_{\mathrm{inel}}^{-1}}{dW}\,=\,\frac{1}{\pi a_0 E}\,\int_{q_-}^{q_+}\frac{dq}{q}\mbox{Im}\left[\frac{1}{\bar\varepsilon(q,W)}\right]\,,$9, $\eta\,=\,\frac{a-1+0.5^a}{a+1}\,,$0, and $\eta\,=\,\frac{a-1+0.5^a}{a+1}\,,$1 (Corre et al., 17 Apr 2026).

In high-resolution macroscopic BSE mosaics of AlCoCrFeNi$\eta\,=\,\frac{a-1+0.5^a}{a+1}\,,$2, the principal outputs are spatially registered morphology descriptors and their correlations with hardness. The lamellar structure proportion is reported as approximately 15% to 33% with a main concentration near 26%, the lamellar width ranges from 550 to 750 nm mostly around 600 nm, and the BCC phase proportion ranges from 33% to 35% centered around 34.5% (Lang et al., 2024). Hardness varied from about 450 HV to 550 HV and the Pearson correlation coefficients between hardness and the image-derived descriptors were $\eta\,=\,\frac{a-1+0.5^a}{a+1}\,,$3 for lamellar width, $\eta\,=\,\frac{a-1+0.5^a}{a+1}\,,$4 for lamellar structure proportion, and $\eta\,=\,\frac{a-1+0.5^a}{a+1}\,,$5 for BCC phase proportion (Lang et al., 2024). This is an example of quantitative BSE imaging as spatially extensive property–microstructure mapping.

At a concrete–epoxy/RFA interface, joint BSE–EDS analysis is used for constituent-specific porosity. The method yields a mean total porosity of 5.08%, average resin porosity of 1.11%, resin porosity near the substrate boundary of 1.63%, average porosity of cement paste in the substrate of 8.55%, cement-paste porosity near the coating interface of 24.46%, and local porosity often exceeding 50% in narrow cement-paste layers around RFA fragments (Żak et al., 2022). The significance is not simply the magnitude of porosity but its assignment to the correct host phase, something the paper argues is unreliable under single-threshold BSE analysis when phases have strongly different average atomic number (Żak et al., 2022).

In EBSD–BSE correlative metallurgy, BSE contributes particle statistics below the EBSD step size. In a recovered Al–Mn alloy, the workflow detects particles down to 0.03 $\eta\,=\,\frac{a-1+0.5^a}{a+1}\,,$6m in diameter, with about 91% of particles smaller than 0.1 $\eta\,=\,\frac{a-1+0.5^a}{a+1}\,,$7m, i.e. below the EBSD step size (Ånes et al., 2022). Across all three BSE datasets, about 90,000 particles were detected, including about 11,000 inside EBSD ROIs, and in one example ROI 4972 particles were detected in BSE compared with 217 in the EBSD intensity map (Ånes et al., 2022). These BSE-derived particle data were inserted into a modified Smith–Zener drag analysis. Starting from

$\eta\,=\,\frac{a-1+0.5^a}{a+1}\,,$8

and using the subgrain-boundary form

$\eta\,=\,\frac{a-1+0.5^a}{a+1}\,,$9

with

$a=0.045Z$0

the analysis arrives at

$a=0.045Z$1

This couples BSE-derived dispersoid statistics directly to texture-dependent boundary drag (Ånes et al., 2022).

A more detector-physics-oriented application is energy-resolved EBSD. Under a 12 keV primary beam on Si(100), the measured BSE energy distribution contributing to the EBSD pattern extended down to about 3 keV, and the weighted average energy decreased from approximately 9.25 keV near the lower edge of the detector to approximately 7.71 keV near the upper edge (Ventura et al., 27 Jul 2025). Filtering within spectral windows 2–8 keV, 8–10 keV, and 10–13 keV showed that the 10–13 keV window produced the sharpest pattern and richest high-frequency Fourier content, while electrons in the 2–8 keV range still produced Kikuchi patterns despite having undergone substantial energy loss $a=0.045Z$2 keV (Ventura et al., 27 Jul 2025). This does not yet constitute conventional BSE imaging, but it provides direct experimental evidence that integrated BSE contrast is spectrally broad.

6. Limitations, misconceptions, and current directions

The most persistent misconception is that a channel labeled “BSE” necessarily measures pure backscattered-electron yield. The III-V anti-phase-domain study shows that physically BSE-related contrast can be strongest in channels dominated by secondary electrons induced by backscattered electrons, and that nominal TLD-BSE mode can exhibit substantially lower contrast than TLD-SE mode under otherwise similar conditions (Corre et al., 17 Apr 2026). Likewise, the concrete–epoxy interface study demonstrates that BSE darkness is not a unique marker of porosity in multiphase systems where low-$a=0.045Z$3 solids overlap pore-like gray levels; BSE intensity alone may require chemical context from EDS before thresholding becomes reliable (Żak et al., 2022).

A second limitation is that many “quantitative” BSE methods are comparative rather than absolutely calibrated. The direct orientation contrast work does not measure absolute BSE coefficients, does not calibrate detector angular acceptance, and does not separate direct BSE from BSE-induced SE in the in-lens channels (Corre et al., 17 Apr 2026). The Timepix3 platform demonstrates counting-based, pixelated, and angularly selectable BSE detection, but does not provide a formula for backscatter coefficient $a=0.045Z$4, a detector solid-angle normalization, an angular response function, or threshold-to-energy calibration for BSE (El-Khairaoui et al., 18 May 2026). Large-area stitched BSE mosaics report morphology descriptors and phase fractions but omit accelerating voltage, beam current, working distance, detector type, and intensity normalization protocol, so grayscale values are not instrument-independent physical quantities (Lang et al., 2024).

A third limitation is uncertainty propagation. The concrete–epoxy paper provides analytical threshold logic and some robustness discussion, including the statement that $a=0.045Z$5, $a=0.045Z$6, and $a=0.045Z$7 lie on relatively flat histogram regions and shifting them by even 10 gray levels does not visibly affect results, but it does not report confusion matrices or repeatability coefficients (Żak et al., 2022). The Al–Mn correlative workflow reports registration statistics and overlap fractions, but no formal false-positive or false-negative rates for BSE particle segmentation (Ånes et al., 2022). The energy-resolved EBSD paper reaches approximately 1 keV energy resolution and requires sparse conditions below roughly 0.05 electrons/pixel/frame to avoid coincidence loss, which limits throughput for mapping (Ventura et al., 27 Jul 2025).

A further issue is that strict BSE metrology still requires forward models coupling specimen transport and detector response. The transport review provides the specimen-side theory in unusual depth, combining ab initio dielectric input, elastic scattering theory, and Monte Carlo trajectory simulation, but it does not provide a full SEM image formation equation mapping local BSE yield to pixel gray level (Taioli et al., 2024). This suggests that the current state of the field is asymmetrical: specimen interaction physics is comparatively mature, whereas detector-specific inversion to absolute image intensity remains less standardized.

Current directions are therefore concentrated in three areas. One is detector-rich acquisition: event-based direct electron detectors, virtual sectoring, and energy-resolved direct detection extend BSE from an analog integrated signal into a countable, angularly distributed, and spectrally filterable dataset (El-Khairaoui et al., 18 May 2026). A second is multimodal fusion, in which BSE is combined with EDS or EBSD so that porosity, phase, particles, and boundary character can be assigned in a common coordinate system (Żak et al., 2022). A third is application-driven quantification, where BSE imaging is judged by the reliability of derived descriptors—domain polarity fractions, APB orientations, area fractions, lamellar widths, particle spacing, or boundary drag—rather than by absolute yield calibration alone (Ånes et al., 2022). This suggests that quantitative backscattered electron imaging is evolving not toward a single universal metric, but toward a layered methodology in which transport physics, detector characterization, normalization, and task-specific post-processing are all explicitly modeled.