🔬 XPS IRF Simulator & Resolution Explorer

数学的基礎と装置関数の厳密な記述

▶ アプリを開く (Streamlit)

1. 概要:なぜこのアプローチが新しいか

🎯 従来手法との違い

従来のアプローチ本シミュレータ
装置関数 = ガウシアン(ブラックボックス)幾何学的要因を分解してパラメータ化
結果(FWHM)だけ測定原因(κ, θ, α, σ等)を個別に制御
メーカー任せの補正Forward/Inverse モデルを自分で構築
1D畳み込みのみ2D空間積分 + 座標変換

XPS装置メーカーは幾何学的補正を実装していますが、その詳細は公開されていません。本シミュレータは、これらの効果を物理的に分解し、パラメータ化することで、装置関数の「中身」を理解可能にします。

2. シミュレーションの流れ

真の物性
\(f_{\text{FD}}(E, T)\)
X線光源による2D発光
\(I_{\text{source}}(E, y)\)
検出器歪み+座標変換
\(\mathcal{D}_{\kappa,\theta}\)
Y軸積分+分解能畳み込み
観測スペクトル
\(S_{\text{obs}}(E)\)

3. フェルミ・ディラック分布

\[ f_{\text{FD}}(E, T) = \frac{1}{1 + \exp\left(\frac{E - E_F}{k_B T}\right)} \]
パラメータ説明
\(k_B\)ボルツマン定数\(8.617333262 \times 10^{-5}\) eV/K
\(T\)試料温度UI: Temperature (K)
\(E_F\)フェルミエネルギー基準点 (0 eV)

4. X線光源のモデル化

4.1 非対称ガウス分布(Skew-Normal)

\[ \phi_\gamma(x; \sigma, \gamma) = 2 \cdot \frac{1}{\sigma\sqrt{2\pi}} \exp\left(-\frac{x^2}{2\sigma^2}\right) \cdot \frac{1}{2}\left[1 + \text{erf}\left(\frac{\gamma x}{\sigma\sqrt{2}}\right)\right] \]

4.2 2D楕円ガウス分布(回転対応)

回転変換:

\[ \begin{pmatrix} E' \\ y' \end{pmatrix} = \begin{pmatrix} \cos\theta_{\text{rot}} & -\sin\theta_{\text{rot}} \\ \sin\theta_{\text{rot}} & \cos\theta_{\text{rot}} \end{pmatrix} \begin{pmatrix} E \\ y \end{pmatrix} \]

4.3 2D発光像の生成

\[ I_{\text{source}}(E, y) = f_{\text{FD}}(E - \alpha y, T) \cdot \phi_{\gamma_y}(y; \sigma_y, \gamma_y) \]
パラメータ説明UI名
\(\sigma_x\)エネルギー方向スポットサイズSpot Size X (meV)
\(\sigma_y\)空間方向スポットサイズSpot Size Y (mm)
\(\gamma_x, \gamma_y\)非対称性パラメータSpot Skew X/Y
\(\theta_{\text{rot}}\)スポット回転角度Spot Rotation (deg)
\(\alpha\)エネルギー勾配 (dE/dy)Energy Gradient

5. 検出器のモデル化

5.1 幾何学的歪み

\[ E_{\text{src}} = E \cos\theta_{\text{tilt}} + y \sin\theta_{\text{tilt}} \] \[ E_{\text{curved}} = E_{\text{src}} - \kappa \left(\frac{y}{y_{\max}}\right)^2 \]
スマイル歪み (κ): 半球型アナライザーの球面収差により、等エネルギー線が湾曲する効果。Y軸積分後に高BE側へ伸びる非対称な裾野を生成します。

5.2 1D投影と分解能畳み込み

\[ S_{\text{obs}}(E) = \left[\int_{-\infty}^{\infty} I_{\text{distorted}}(E, y) \, dy\right] \ast G_{\text{res}}(E; \sigma_{\text{res}}) \]
パラメータ説明UI名
\(\kappa\)スマイル曲率係数Smile Curvature
\(\theta_{\text{tilt}}\)検出器傾き角度Detector Tilt (deg)
\(\sigma_{\text{res}}\)装置固有分解能Intrinsic Res (meV)

5.3 分解能の合成

本シミュレータでは、ソース分解能と検出器分解能を独立に畳み込み、合成分解能を以下の式で計算します:

\[ \sigma_{\text{combined}} = \sqrt{\sigma_{\text{source}}^2 + \sigma_{\text{detector}}^2} \]
成分説明UI名
\(\sigma_{\text{source}}\)X線源のエネルギー方向分解能Source Resolution (meV)
\(\sigma_{\text{detector}}\)検出器の固有分解能Detector Resolution (meV)
\(\sigma_{\text{combined}}\)合成分解能(二乗和平方根)Combined Resolution
物理的意味: 2つの独立したガウシアン畳み込みの合成はガウシアンとなり、その標準偏差は各成分の二乗和の平方根になります。これにより、ソース分解能と検出器分解能の寄与を定量的に分離できます。

6. ノイズモデル

\[ S_{\text{final}}(E) = \frac{\text{Poisson}\left(S_{\text{clean}} \cdot \lambda\right)}{\lambda} + \mathcal{N}(0, \sigma_{\text{readout}}) \]
ノイズ種別特性UI名
ポアソンノイズ信号強度依存(ショットノイズ)Poisson Noise Level
ガウシアンノイズ信号強度非依存(読み出しノイズ)Gaussian Readout Noise

7. 統一的な数式表現

完全な観測スペクトル

\[ \boxed{ S_{\text{obs}}(E) = \left[ \int_{-\infty}^{\infty} f_{\text{FD}}\left(\mathcal{T}(E, y; \kappa, \theta) - \alpha y, T\right) \cdot \phi_{\gamma_y}(y; \sigma_y, \gamma_y) \, dy \right] \ast G_{\text{res}}(E; \sigma_{\text{res}}) + \eta(E) } \]

ここで座標変換は:

\[ \mathcal{T}(E, y; \kappa, \theta) = \left(E \cos\theta + y \sin\theta\right) - \kappa \left(\frac{y}{y_{\max}}\right)^2 \]

演算子表記

\[ S_{\text{obs}} = \mathcal{P}_y \left[ \mathcal{D}_{\kappa,\theta} \left[ \mathcal{S}_\alpha \left[ f_{\text{FD}} \right] \cdot W_y \right] \right] \ast G_{\text{res}} + \eta \]

8. 実装の検証

✓ 検証済み項目

ボルツマン定数: CODATA 2018準拠
フェルミ・ディラック式: 標準形式
歪正規分布: 係数2を含む正しい定義
回転行列: 反時計回り正(標準)
スマイル歪み: y²依存(球面収差の一次近似)
2Dスキューガウス: 1D積の近似(実用上十分)
文献との整合性: 装置関数をVoigt関数(ガウシアン×ローレンツィアン)で扱う文献は存在しますが、幾何学的要因を個別にパラメータ化した公開文献は見当たりません。装置メーカーは補正を実装していますが、詳細は非公開です。

9. 参考文献

  • Jain, V. et al., Applied Surface Science 447, 548-553 (2018) - Voigt関数とXPS
  • Hemispherical Analyzer - Wikipedia

🔬 XPS IRF Simulator & Resolution Explorer

Mathematical Foundation and Rigorous Description of Instrumental Response

▶ Open App (Streamlit)

1. Overview: Why This Approach is Novel

🎯 Comparison with Conventional Methods

Conventional ApproachThis Simulator
IRF = Gaussian (black box)Decomposed geometric factors with parametrization
Measure only results (FWHM)Control individual causes (κ, θ, α, σ, etc.)
Vendor-dependent correctionsBuild your own Forward/Inverse model
1D convolution only2D spatial integration + coordinate transform

XPS instrument manufacturers implement geometric corrections, but the details are proprietary. This simulator decomposes and parametrizes these effects, making the "internals" of the instrumental function understandable and controllable.

2. Simulation Pipeline

True Physics
\(f_{\text{FD}}(E, T)\)
X-ray Source 2D Emission
\(I_{\text{source}}(E, y)\)
Detector Distortion + Transform
\(\mathcal{D}_{\kappa,\theta}\)
Y-axis Integration + Resolution Convolution
Observed Spectrum
\(S_{\text{obs}}(E)\)

3. Fermi-Dirac Distribution

\[ f_{\text{FD}}(E, T) = \frac{1}{1 + \exp\left(\frac{E - E_F}{k_B T}\right)} \]
ParameterDescriptionValue
\(k_B\)Boltzmann constant\(8.617333262 \times 10^{-5}\) eV/K
\(T\)Sample temperatureUI: Temperature (K)
\(E_F\)Fermi energyReference (0 eV)

4. X-ray Source Modeling

4.1 Skew-Normal Distribution

\[ \phi_\gamma(x; \sigma, \gamma) = 2 \cdot \frac{1}{\sigma\sqrt{2\pi}} \exp\left(-\frac{x^2}{2\sigma^2}\right) \cdot \frac{1}{2}\left[1 + \text{erf}\left(\frac{\gamma x}{\sigma\sqrt{2}}\right)\right] \]

4.2 2D Elliptical Gaussian with Rotation

Rotation Transform:

\[ \begin{pmatrix} E' \\ y' \end{pmatrix} = \begin{pmatrix} \cos\theta_{\text{rot}} & -\sin\theta_{\text{rot}} \\ \sin\theta_{\text{rot}} & \cos\theta_{\text{rot}} \end{pmatrix} \begin{pmatrix} E \\ y \end{pmatrix} \]

4.3 2D Emission Image Generation

\[ I_{\text{source}}(E, y) = f_{\text{FD}}(E - \alpha y, T) \cdot \phi_{\gamma_y}(y; \sigma_y, \gamma_y) \]
ParameterDescriptionUI Label
\(\sigma_x\)Energy-direction spot sizeSpot Size X (meV)
\(\sigma_y\)Spatial-direction spot sizeSpot Size Y (mm)
\(\gamma_x, \gamma_y\)Asymmetry parametersSpot Skew X/Y
\(\theta_{\text{rot}}\)Spot rotation angleSpot Rotation (deg)
\(\alpha\)Energy gradient (dE/dy)Energy Gradient

5. Detector Modeling

5.1 Geometric Distortion

\[ E_{\text{src}} = E \cos\theta_{\text{tilt}} + y \sin\theta_{\text{tilt}} \] \[ E_{\text{curved}} = E_{\text{src}} - \kappa \left(\frac{y}{y_{\max}}\right)^2 \]
Smile Distortion (κ): Due to spherical aberration in hemispherical analyzers, iso-energy lines become curved. After Y-axis integration, this produces an asymmetric tail extending toward higher binding energy.

5.2 1D Projection and Resolution Convolution

\[ S_{\text{obs}}(E) = \left[\int_{-\infty}^{\infty} I_{\text{distorted}}(E, y) \, dy\right] \ast G_{\text{res}}(E; \sigma_{\text{res}}) \]
ParameterDescriptionUI Label
\(\kappa\)Smile curvature coefficientSmile Curvature
\(\theta_{\text{tilt}}\)Detector tilt angleDetector Tilt (deg)
\(\sigma_{\text{res}}\)Intrinsic resolutionIntrinsic Res (meV)

5.3 Resolution Combination

This simulator applies source resolution and detector resolution as independent convolutions, and calculates the combined resolution using:

\[ \sigma_{\text{combined}} = \sqrt{\sigma_{\text{source}}^2 + \sigma_{\text{detector}}^2} \]
ComponentDescriptionUI Label
\(\sigma_{\text{source}}\)X-ray source energy resolutionSource Resolution (meV)
\(\sigma_{\text{detector}}\)Detector intrinsic resolutionDetector Resolution (meV)
\(\sigma_{\text{combined}}\)Combined resolution (root sum of squares)Combined Resolution
Physical Meaning: The convolution of two independent Gaussians results in a Gaussian whose standard deviation is the square root of the sum of squares of each component. This allows quantitative separation of source and detector resolution contributions.

6. Noise Model

\[ S_{\text{final}}(E) = \frac{\text{Poisson}\left(S_{\text{clean}} \cdot \lambda\right)}{\lambda} + \mathcal{N}(0, \sigma_{\text{readout}}) \]
Noise TypeCharacteristicsUI Label
Poisson noiseSignal-dependent (shot noise)Poisson Noise Level
Gaussian noiseSignal-independent (readout noise)Gaussian Readout Noise

7. Unified Mathematical Expression

Complete Observed Spectrum

\[ \boxed{ S_{\text{obs}}(E) = \left[ \int_{-\infty}^{\infty} f_{\text{FD}}\left(\mathcal{T}(E, y; \kappa, \theta) - \alpha y, T\right) \cdot \phi_{\gamma_y}(y; \sigma_y, \gamma_y) \, dy \right] \ast G_{\text{res}}(E; \sigma_{\text{res}}) + \eta(E) } \]

Where the coordinate transform is:

\[ \mathcal{T}(E, y; \kappa, \theta) = \left(E \cos\theta + y \sin\theta\right) - \kappa \left(\frac{y}{y_{\max}}\right)^2 \]

Operator Notation

\[ S_{\text{obs}} = \mathcal{P}_y \left[ \mathcal{D}_{\kappa,\theta} \left[ \mathcal{S}_\alpha \left[ f_{\text{FD}} \right] \cdot W_y \right] \right] \ast G_{\text{res}} + \eta \]

8. Implementation Verification

✓ Verified Items

Boltzmann constant: CODATA 2018 compliant
Fermi-Dirac formula: Standard form
Skew-normal distribution: Correct definition with factor of 2
Rotation matrix: Counterclockwise positive (standard)
Smile distortion: y² dependence (first-order spherical aberration)
2D skew-Gaussian: Product of 1D approximation (sufficient for practical use)
Literature Consistency: While literature treating the IRF as a Voigt function (Gaussian × Lorentzian) exists, no published work parametrizing individual geometric factors was found. Instrument manufacturers implement corrections, but details remain proprietary.

9. References

  • Jain, V. et al., Applied Surface Science 447, 548-553 (2018) - Voigt functions in XPS
  • Hemispherical Analyzer - Wikipedia