科研 Skill

电子探针显微表征分析

microscopy

Covers electron and probe microscopy for materials characterization: SEM, TEM/STEM/EELS/EDS, and AFM, including drift correction, segmentation. Use for microscopy image analysis workflows.

发布者 Codar安全审核通过v0.1.0更新于 2026-11-18

SKILL.md

Microscopy

Overview

SEM, TEM, and AFM all produce spatially resolved images of a material's surface or internal structure, at scales from microns down to atomic resolution. They share a substantial image-analysis workflow — drift/distortion correction, segmentation, feature/particle sizing — even though the physical contrast mechanisms (secondary/backscattered electrons for SEM, transmitted/diffracted electrons for TEM, tip-sample force for AFM) are entirely different.

When to use this skill

Activate when the request involves:

  • SEM, scanning electron microscopy, TEM, transmission electron microscopy, STEM, EELS, EDS/EDX, AFM, atomic force microscopy
  • Terms: secondary electron, backscattered electron, bright-field/dark-field, particle sizing, segmentation, tip artifact
  • File formats: .dm3/.dm4 (Gatan), .emd, .tif (common microscope export), .ibw/.spm (AFM)
  • "Segment particles in this SEM image," "analyze this EELS spectrum image," "measure roughness from AFM"

Which reference to read

You have...Read
SEM images (secondary/backscattered electron, EDS elemental maps)references/sem.md
TEM images, diffraction, or STEM/EELS/EDS spectrum imagingreferences/tem.md
AFM topography, phase, or force datareferences/afm.md

For electron diffraction specifically (SAED, acquired in a TEM), see diffraction/references/electron-diffraction.md.

Core usage — shared image processing

Drift correction (cross-correlation registration across a stack)

import numpy as np
from skimage.registration import phase_cross_correlation
from scipy.ndimage import shift as ndi_shift

def correct_drift(image_stack):
    """image_stack: (n_frames, height, width). Registers each frame to the first."""
    reference = image_stack[0]
    corrected = [reference]
    for frame in image_stack[1:]:
        shift_estimate, _, _ = phase_cross_correlation(reference, frame, upsample_factor=10)
        corrected.append(ndi_shift(frame, shift_estimate))
    return np.array(corrected)

Particle/feature segmentation and sizing

from skimage.filters import threshold_otsu
from skimage.measure import label, regionprops
from skimage.morphology import remove_small_objects

def segment_and_measure(image, min_size_px=20):
    binary = image > threshold_otsu(image)
    binary = remove_small_objects(binary, min_size=min_size_px)
    labeled = label(binary)
    return [
        {"area_px": r.area, "equivalent_diameter_px": r.equivalent_diameter, "eccentricity": r.eccentricity}
        for r in regionprops(labeled)
    ]

Convert pixel measurements to physical units using the image's calibrated pixel size (from instrument metadata) before reporting — see Validation & Pitfalls.

Validation & Pitfalls

Canonical reference: Williams & Carter, Transmission Electron Microscopy (2nd ed., 2009), and Goldstein et al., Scanning Electron Microscopy and X-Ray Microanalysis (4th ed., 2017), for instrument-level physics across both techniques.

  • Pixel-to-physical-unit calibration is instrument- and magnification-setting-specific and must be read from image metadata, not assumed constant. Reporting particle sizes without confirming the actual calibration for that specific image (not a remembered "typical" value) is a common source of systematically wrong measurements.
  • Segmentation threshold choice (Otsu or any automatic method) can fail on images with uneven illumination, charging artifacts (SEM), or low contrast — visually verify segmentation results against the raw image on a representative sample before trusting automated measurements at scale.
  • "The image looks clear" is not the same as "the sample is representative" — imaging a small field of view at high magnification risks sampling bias; report how many fields/particles were measured and whether they were selected systematically (e.g. a grid pattern) rather than by eye, which biases toward visually striking (often larger, more distinct) features.

示例提示词

我有一批TEM图像需要做漂移校正和颗粒尺寸分布统计。

所属插件

材料表征、性能与材料信息学面向实验与数据驱动的材料研究:XRD/电镜/谱学/热分析/力学测试等表征,合成与加工,电化学与半导体表征,相场、CALPHAD 与连续介质力学,材料数据库、特征工程、贝叶斯优化与主动学习,选材与生命周期评价。

安装该插件会一并获得同领域的 42 个 Skill / MCP。

同插件的其他科研 Skill

常见问题

电子探针显微表征分析能做什么?

Covers electron and probe microscopy for materials characterization: SEM, TEM/STEM/EELS/EDS, and AFM, including drift correction, segmentation. Use for microscopy image analysis workflows。

如何安装电子探针显微表征分析?

在科搭 Codar 左侧点击「插件」,在「组件」页签搜索「电子探针显微表征分析」并安装;在 Claude Code 或 Codex 中可以发送本页的安装提示词,或下载 Skill 包后解压到 ~/.claude/skills 或 ~/.codex/skills。

电子探针显微表征分析需要哪些权限?

该科研 Skill没有声明额外权限;执行具体操作时仍由 Agent 按自身规则请求确认。

电子探针显微表征分析经过官方审核吗?

是。SpeedSkills 收录的每个版本都会记录内容摘要与权限声明,当前版本已通过安全审核。