Restoring missing low scattering angle data in two-dimensional diffraction patterns of isolated molecules

This paper presents an iterative algorithm that utilizes Fourier and Abel transforms along with real-space constraints to successfully restore missing low scattering angle data in anisotropic two-dimensional diffraction patterns of isolated molecules, requiring only approximate knowledge of the molecule's internuclear distances.

The Gist

Imagine you are trying to take a photograph of a spinning, dancing molecule to see exactly how its atoms are moving. To do this, scientists shoot a beam of tiny particles (like electrons) at the molecule. These particles bounce off the atoms and hit a detector, creating a pattern of dots and rings. This pattern is like a "fingerprint" of the molecule's shape.

However, there's a big problem with taking these photos: The camera has a blind spot.

The Problem: The Missing Center

In these experiments, the beam of particles is so strong that if it hits the detector directly, it would break the camera. So, scientists put a tiny shield (a "beam stop") right in the middle to block the direct beam.

The Analogy: Imagine trying to take a picture of a fireworks display, but you have to hold a thick pole right in front of your camera lens to stop the blinding flash from burning your eyes. The result? You get a beautiful picture of the outer sparks, but the center of the explosion is completely black and missing.

In the world of molecules, this "black hole" in the middle of the data means scientists are missing the most important information: how the atoms are arranged relative to each other. Without the center, they can't reconstruct the full 3D shape of the molecule in real space. It's like trying to solve a jigsaw puzzle but throwing away all the pieces that show the faces of the people in the picture.

The Old Way vs. The New Way

Previously, scientists tried to fix this by guessing. They would look at the edges of the puzzle and try to "smoothly interpolate" (draw a guess) for the missing middle. Or, they would use complex computer simulations to guess what the middle should look like.

But this paper introduces a clever iterative algorithm (a step-by-step guessing game) that doesn't just guess; it learns the answer.

The Solution: The "Shrink-Wrap" Game

The authors, Yanwei Xiong and Martin Centurion, developed a method that works like a game of "Hot and Cold" or a digital version of sculpting clay. Here is how it works, using simple metaphors:

1. The Two Worlds: Think of the data as existing in two different worlds:

   • World A (The Diffraction Pattern): This is the raw data with the missing center (the black hole).
   • World B (The Real Shape): This is the actual 3D map of where the atoms are.

2. The Magic Bridge: The algorithm uses a mathematical "bridge" (Fourier and Abel transforms) to jump back and forth between World A and World B.

3. The "Shrink-Wrap" Constraint:

   • The scientists know one simple fact about the molecule: It has a specific size. They know the shortest distance between any two atoms and the longest distance (e.g., "The molecule is no bigger than a grape and no smaller than a pea").
   • When the algorithm jumps to World B (the real shape), it applies a "Shrink-Wrap" rule. It says, "Any part of this shape that is outside the known size of the molecule is fake noise. Cut it off!"
   • This "noise" is actually the error caused by the missing data in the center. By cutting off the parts that don't fit the known size, the algorithm forces the data to become more accurate.

4. The Loop:

   • The algorithm takes the "cleaned up" shape from World B, jumps back to World A, and fills in the missing center hole based on what it learned.
   • It jumps back to World B, checks the size again, cuts off more noise, and jumps back again.
   • It does this over and over (50 times in their test). With every loop, the "guess" for the missing center gets closer and closer to the truth, and the noise gets smaller.

The Results: A Crystal Clear Picture

They tested this on a molecule called Trifluoroiodomethane (CF3I).

• The Simulation: First, they used a computer to create a perfect fake molecule with a known missing center. Their algorithm successfully recovered the missing data, making the "fake" puzzle look exactly like the "real" one.
• The Real Experiment: Then, they used a real laser to align real CF3I molecules and took actual photos. Even with the messy, noisy real-world data, the algorithm successfully filled in the missing center.

Why This Matters

This is a huge deal because:

• It's Simple: You don't need to know the exact chemical structure beforehand. You just need to know the rough size (min and max distance between atoms).
• It's General: It works for both "isotropic" (round, boring) patterns and "anisotropic" (spiky, directional) patterns. Most modern experiments use lasers that make molecules line up, creating these complex, directional patterns. This method unlocks the data from those advanced experiments.
• It Reveals the Invisible: It allows scientists to see the full 3D dance of atoms during chemical reactions, which was previously impossible because the "center" of the data was lost.

In summary: The authors built a smart, self-correcting digital tool that fills in the missing center of a molecular photo by repeatedly checking if the resulting shape fits the known size of the molecule. It's like having a magic eraser that removes the guesswork and reveals the true structure of the molecular world.

Technical Summary

1. Problem Statement

In gas-phase ultrafast electron diffraction (GUED) and ultrafast X-ray diffraction (UXRD) experiments, researchers aim to capture the structural dynamics of isolated molecules with femtosecond and sub-angstrom resolution. However, a critical limitation exists in these experiments:

• Data Loss at Low Angles: To protect detectors from the intense direct beam, a beam stop is used, which blocks scattering signals at low momentum transfer ($s < s_{min}$).
• Consequence: The missing low-angle data is essential for reconstructing the real-space pair distribution function (PDF) and obtaining accurate interatomic distances. Without this data, the real-space representation suffers from significant artifacts (ringing effects) and lacks the necessary long-range structural information.
• Anisotropy Challenge: While methods exist to restore missing data for isotropic (1-D) signals, many modern experiments use linearly polarized lasers to align molecules, creating anisotropic 2-D diffraction patterns. These patterns contain richer information (e.g., bond angles, angular distributions) but require complex 2-D Fourier and Abel inversions. Existing isotropic restoration algorithms cannot be directly applied to these anisotropic 2-D patterns.

2. Methodology

The authors propose an iterative retrieval algorithm designed specifically to restore missing low-scattering-angle data in 2-D anisotropic diffraction patterns. The method combines phase retrieval concepts with the specific physics of electron scattering from aligned molecules.

Key Algorithmic Steps:

1. Initial Guess: The missing region ($0 \le s < s_{min}$) is initially filled with a guess (e.g., linear interpolation) to create a complete 2-D signal $\mathcal{M}_e(\mathbf{s})$.
2. Forward Transform (Momentum to Real Space):
   • The signal is transformed from momentum space to real space using a 2-D Fourier Transform followed by an Abel Inversion.
   • A damping function ($e^{-ds^2}$) is applied to the momentum signal to minimize artifacts from the $s_{max}$ cutoff.
   • This yields a real-space map $\mathcal{P}_e(r, \alpha)$, which contains the true signal plus an artifact $\mathcal{E}(r, \alpha)$ caused by the missing low-angle data.
3. Real-Space Constraint (Support Constraint):
   • The algorithm applies a "support constraint" based on a-priori knowledge of the molecule's geometry. Specifically, it assumes the molecule has a known minimum ($r_1$) and maximum ($r_2$) interatomic distance.
   • A 2-D band-pass filter (Gaussian-like function) is applied to truncate any signal outside the range $[r_1, r_2]$. This effectively suppresses the artifact $\mathcal{E}(r, \alpha)$, which typically extends beyond the physical bounds of the molecule, while preserving the true signal $\mathcal{P}(r, \alpha)$.
4. Inverse Transform (Real to Momentum Space):
   • The constrained real-space signal is transformed back to momentum space using the Abel Transform followed by the 2-D Fourier Transform.
5. Legendre Decomposition and Stitching:
   • The 2-D pattern is decomposed into Legendre polynomials ($\ell_i(s)$) to handle the cylindrical symmetry of the laser-aligned sample.
   • The low-$s$ components of the newly calculated signal are stitched with the high-$s$ components of the measured experimental data.
   • A rescaling factor is applied to ensure continuity at the stitching point ($s_{min}$).
6. Iteration: The process repeats (steps 2–5) until the error metric (sum of squared differences between iterations) converges.

Required Inputs:

• The measured diffraction pattern (from $s_{min}$ to $s_{max}$).
• Approximate a-priori knowledge of the molecule's shortest and longest interatomic distances ($r_{min}, r_{max}$).

3. Key Contributions

• Extension to 2-D Anisotropic Signals: The primary contribution is the adaptation of iterative phase retrieval techniques from 1-D isotropic signals to 2-D anisotropic patterns, enabling the recovery of missing low-angle data in complex molecular alignment experiments.
• Minimal A-Priori Requirements: The algorithm is robust and requires only the approximate range of interatomic distances, avoiding the need for complex structural models or genetic algorithms.
• General Applicability: The method is applicable to both GUED and UXRD and works for both anisotropic (aligned) and isotropic (random) samples.
• Separation of Scattering Types: The framework theoretically allows for the separation of elastic and inelastic scattering components in GUED signals.

4. Results

The authors validated the algorithm using both simulated data and experimental data from trifluoroiodomethane (CF$_3$I) molecules.

• Simulated Data (CF$_3$I):

  • Using a simulated pattern with a missing region from $0$ to $1.6$Å$^{-1}$, the algorithm successfully restored the signal after 50 iterations.
  • The restored Legendre projections ($\ell_0$ and $\ell_2$) matched the true theoretical signals with high fidelity.
  • The real-space representation (Modified Pair Distribution Function, MPDF) clearly resolved atom-pair rings corresponding to C-F, C-I, F-F, and F-I distances, which were obscured by artifacts in the initial guess.
  • Error metrics ($\mathcal{S}_n$ and $\mathcal{R}_n$) converged rapidly, plateauing after ~15 iterations.

• Experimental Data (CF$_3$I Alignment):

  • The algorithm was applied to experimental GUED data of CF$_3$I aligned by an impulsive IR laser pulse.
  • The retrieved 2-D patterns and real-space maps after 50 iterations showed excellent agreement with theoretical calculations derived from the time-dependent Schrödinger equation (TDSE).
  • The method successfully recovered the missing low-$s$ data, allowing for a clear visualization of the molecular orientation and bond distances in the real-space domain.

• Robustness Test (Appendix A):

  • The algorithm was tested with a more restricted momentum transfer range ($1.6$ to $5.0$Å$^{-1}$), typical of UXRD experiments. It successfully restored the low-angle data, demonstrating versatility across different experimental setups.

5. Significance

This work addresses a fundamental bottleneck in ultrafast diffraction science: the loss of low-angle data. By enabling the restoration of this missing information in 2-D anisotropic patterns, the algorithm:

• Enhances Structural Resolution: It allows for the accurate retrieval of real-space pair distribution functions and atom-pair angular distributions, which are critical for understanding complex molecular dynamics and reaction pathways.
• Maximizes Information Content: It unlocks the full potential of anisotropic diffraction patterns, which contain more structural information than isotropic averages.
• Simplifies Analysis: It provides a computationally efficient, model-free (or minimally model-dependent) alternative to complex genetic algorithms or simulation-heavy fitting procedures.
• Broad Impact: The technique is immediately applicable to current and future GUED and UXRD experiments studying chemical reactions, photodissociation, and molecular alignment, facilitating more accurate "movies" of molecular motion.