EMS Measurement of the Valence Spectral Function of Silicon - a test of Many-body Theory

High-resolution electron momentum spectroscopy measurements of silicon's valence spectral function demonstrate that while both GW and cumulant expansion calculations accurately describe band dispersions, only the cumulant expansion approach successfully reproduces the observed satellite structures, validating it as a superior method for modeling high-energy excitations.

The Gist

The Big Picture: Taking a "Molecular X-Ray" of Silicon

Imagine you have a perfect, invisible city made of atoms (Silicon). For decades, scientists have been trying to map this city. They knew where the buildings (energy levels) were located, but they didn't really know what the "streets" looked like or how the people (electrons) were actually moving around.

Most previous maps were like looking at a city from a satellite: you could see the general layout, but you couldn't see the individual cars or how fast they were driving. This paper is about taking a much closer, 3D snapshot of the electrons inside Silicon to see exactly how they behave, not just where they sit.

The Tool: The "Electron Billiard" (EMS)

To see these invisible electrons, the scientists used a technique called Electron Momentum Spectroscopy (EMS).

Think of it like a high-speed game of billiards, but played with subatomic particles:

1. The Cue Ball: They shoot a super-fast electron (the cue ball) at a thin slice of Silicon.
2. The Collision: The cue ball hits an electron inside the Silicon, knocking it out of the atom.
3. The Catch: Two detectors catch the original cue ball and the newly knocked-out electron.

By measuring the speed and direction of these two outgoing electrons, the scientists can work backward to figure out exactly how fast and in what direction the electron was moving before it got hit. It's like a detective reconstructing a car crash by measuring the speed of the two cars after they collide to figure out how they were driving before the crash.

The Discovery: It's Not Just a Smooth Slide

The scientists compared their new, high-definition map against two different computer models:

1. The "Ideal World" Model (Independent Particle Approximation)
Imagine a model where every electron is a lonely hiker walking on a perfectly smooth, empty path. They don't talk to each other; they just walk their own way.

• The Result: The scientists found that this model was actually pretty good at predicting where the electrons were (the energy levels). It was like a good map of the city's streets.

2. The "Crowded Party" Model (Many-Body Theory)
In reality, electrons are like people at a crowded, chaotic party. They bump into each other, push, and create ripples. This is called "electron correlation."

• The Result: The "Ideal World" model failed here. The real electrons weren't just walking on smooth paths. They were creating a huge mess of "echoes" and "shadows."

The "Ghost" Echoes (Satellites)

Here is the most fascinating part of the paper.

When an electron gets knocked out of Silicon, it doesn't just leave a single, clean spot. Because the electrons are so crowded and interactive, knocking one out creates a ripple effect.

• The Analogy: Imagine dropping a stone into a calm pond. You see the main splash (the main electron peak). But then, you see a series of smaller ripples spreading out (the "satellite" structures).
• The Finding: The scientists found that a huge chunk of the electron's "identity" (about 40% when the electron is slow) is actually hidden in these ripples. The electron isn't just one thing; it's a main splash plus a cloud of ripples.

The "Blurry" Photo (Lifetime Broadening)

The paper also talks about how "sharp" the picture is.

• The Analogy: Imagine taking a photo of a hummingbird. If the bird is sitting still, the photo is sharp. If the bird is flapping its wings wildly, the photo is blurry.
• The Finding: In Silicon, electrons at low speeds are like that frantic hummingbird. They interact so much with their neighbors that they change their state very quickly. This makes their "photo" very blurry (broad). As the electrons move faster (higher momentum), they become more like a calm bird, and the photo gets sharper.

Did the Computer Models Get It Right?

The scientists tested two advanced computer theories to see if they could predict these ripples and blurriness:

1. The GW Approximation: This is a popular, sophisticated model. It was great at predicting the main splash (the main electron), but it completely missed the ripples. It was like a weather forecast that predicted the rain but forgot the thunderstorms.
2. The Cumulant Expansion: This is a newer, more complex model. It was much better! It predicted the ripples and the blurriness much more accurately. However, even this model underestimated how big the ripples were. It saw the storm, but it thought the storm was smaller than it actually was.

The Bottom Line

This paper is a victory for "Many-Body Theory." It proves that to understand how Silicon works (which is crucial for making better computer chips), we can't just look at electrons as lonely individuals. We have to account for the chaotic, crowded party they are having.

While our best computer models are getting closer to the truth, this experiment shows that nature is still a bit more chaotic and complex than our current equations can fully capture. It's a reminder that even in a material as common as Silicon, there are still deep mysteries waiting to be solved.

Technical Summary

1. Problem Statement

The electronic properties of semiconductors, particularly silicon, have long been studied using Angle-Resolved Photoemission Spectroscopy (ARPES) and theoretical band structure calculations. However, a significant gap exists in the validation of many-body theories:

• Limitations of Existing Methods: ARPES is highly surface-sensitive and struggles to disentangle initial-state properties from final-state effects. Furthermore, extracting the full spectral function $A(q, \omega)$ (which includes quasiparticle lifetimes and satellite structures) from ARPES is difficult due to matrix element dependencies and background noise.
• Theoretical Deficiencies: While Density Functional Theory (DFT) and the $GW$ approximation successfully predict quasiparticle energies (band dispersions), they often fail to accurately describe electron correlation effects, specifically:
  • The finite lifetime of quasiparticles (leading to peak broadening).
  • The formation of satellite structures (e.g., plasmon satellites) arising from electron-electron interactions.
• Objective: The authors aim to measure the full Spectral Electron Momentum Density (SEMD), $A(q, \omega)$, of silicon's valence band using Electron Momentum Spectroscopy (EMS). This provides a direct, bulk-sensitive test of many-body theories by comparing experimental momentum densities and energy widths against predictions from Independent Particle Approximations (LDA) and advanced Many-Body Perturbation Theory (GW and Cumulant Expansion).

2. Methodology

Experimental Technique: Electron Momentum Spectroscopy (EMS)

• Principle: The experiment utilizes high-energy, high-momentum-transfer $(e, 2e)$ ionizing collisions. A 50 keV electron beam strikes a thin silicon target, ejecting a valence electron. The incident electron and the two outgoing electrons are detected in coincidence.
• Kinematics: By conserving energy and momentum, the binding energy ($\omega$) and recoil momentum ($q$) of the ejected electron are determined:
  • $\omega = E_0 - E_1 - E_2$
  • $q = k_1 + k_2 - k_0$
• Apparatus: A non-coplanar symmetric geometry was used at the Australian National University. The setup allows for the measurement of momentum densities along high-symmetry directions ($\langle 100 \rangle$, $\langle 110 \rangle$, $\langle 111 \rangle$) through the $\Gamma$ point.
• Sample Preparation: A self-supporting single-crystal silicon film (~20 nm thick) was prepared via ion implantation, wet chemical etching, and low-energy argon sputtering to ensure bulk sensitivity and minimize surface effects.
• Data Correction:
  • Inelastic Scattering: Multiple scattering events (e.g., plasmon excitation) were corrected using a parameter-free deconvolution based on measured energy loss spectra.
  • Diffraction: Coherent elastic scattering (diffraction) shifts momentum by reciprocal lattice vectors ($G$). The authors developed a method to identify and subtract these diffracted contributions by rotating the sample and analyzing regions where spectral density should theoretically be zero.

Theoretical Framework

• Independent Particle Model: Calculations were performed using the Full-Potential Linear Muffin-Tin Orbital (FP-LMTO) method within the DFT-LDA framework. This provides the baseline band structure and momentum density assuming no electron correlation.
• Many-Body Models:
  • $GW$ Approximation: Used to calculate the self-energy ($\Sigma$) to improve quasiparticle energies.
  • Cumulant Expansion: An advanced approach applied to the one-hole Green's function to account for multiple plasmon creation and vertex corrections, aiming to better describe satellite structures and quasiparticle lifetimes.

3. Key Contributions

1. First Full SEMD Measurement: The paper presents the first comprehensive experimental mapping of the valence spectral momentum density of silicon, covering both the quasiparticle peak and the satellite structures across multiple high-symmetry directions.
2. Diffraction Correction Protocol: The authors established a rigorous method to distinguish between "direct" spectral density and diffraction-induced artifacts in EMS data, allowing for the extraction of the true bulk spectral function.
3. Quantitative Benchmarking: The study provides a direct comparison between experimental data and state-of-the-art theoretical models (LDA, $GW$, and Cumulant Expansion), specifically testing the ability of these models to predict lifetimes (peak widths) and satellite intensities.

4. Key Results

• Band Structure (Dispersion):

  • The experimental band dispersion (peak positions) along $\langle 100 \rangle$, $\langle 110 \rangle$, and $\langle 111 \rangle$ directions matches the FP-LMTO (LDA) calculations extremely well.
  • The measured valence band width is 11.85 ± 0.2 eV, consistent with both LDA theory (11.93 eV) and previous ARPES data.
  • EMS data provided smoother and better-defined band structures compared to available ARPES data.

• Quasiparticle Lifetimes and Broadening:

  • Experimental spectra showed significant broadening (FWHM) of quasiparticle peaks, particularly at low momenta (near $\Gamma$), which was not captured by the independent particle model.
  • The width decreases as momentum increases, reaching a minimum (~1.35 eV) near the $X$ and $L$ points where dispersion extrema occur.
  • Theory Comparison: Both $GW$ and Cumulant models underestimated the experimental linewidths, though the Cumulant expansion provided a slightly better fit. The discrepancy suggests that current many-body approximations still underestimate the strength of electron correlation effects on lifetimes.

• Satellite Structures:

  • A significant portion of the spectral density (up to 40% at $q=0$) was found in satellite structures extending ~20 eV above the quasiparticle peak.
  • $GW$ Failure: The $GW$ approximation failed to reproduce the satellite structure, predicting a single satellite at an incorrect energy.
  • Cumulant Success: The Cumulant expansion successfully reproduced the shape and energy distribution of the satellite structure, confirming its superiority over $GW$ for describing correlation effects. However, it still underestimated the total intensity of the satellites, particularly at low momenta.

• Momentum Dependence:

  • The relative contribution of satellite structures to the total density decreases as momentum increases. By $q \approx 1$ a.u., the satellite structure essentially disappears, and the spectrum is dominated by the quasiparticle peak.

5. Significance

• Validation of Many-Body Theory: This work demonstrates that while DFT-LDA is excellent for ground-state band structures, it is insufficient for describing excited-state dynamics (lifetimes and satellites). It validates the Cumulant Expansion as a necessary step beyond the $GW$ approximation for modeling electron correlations in semiconductors.
• Methodological Advance: The study proves that EMS is a superior tool for testing many-body theories compared to ARPES because it measures the full spectral function in momentum space without the surface sensitivity and matrix element complications of photoemission.
• Future Directions: The remaining discrepancies between the Cumulant model and experimental data (specifically the underestimation of satellite intensity) highlight the need for even more sophisticated treatments of electron-electron interactions. The authors suggest that quantitative comparisons of EMS data with wave function representations could lead to new levels of understanding in condensed matter physics.

In summary, the paper successfully uses EMS to map the valence spectral function of silicon, confirming the accuracy of standard band theory for dispersion while revealing the critical importance of many-body effects for quasiparticle lifetimes and satellite formation, thereby providing a stringent testbed for advanced theoretical models.