Fast projections of two-dimensional light patterns using acousto-optical deflectors

This paper presents a fast, feedback-free method for projecting two-dimensional light patterns using acousto-optical deflectors with an incommensurately staggered frequency lattice to intrinsically suppress intermodulation artifacts, enabling high-speed, high-fidelity generation of both separable and non-separable structured light fields.

The Gist

Imagine you are trying to paint a beautiful, complex picture on a wall using a single, tiny, super-fast paintbrush. You want to create a detailed image of a cat, a landscape, or even a specific pattern of light to hold tiny atoms in place.

This is the challenge scientists face when using Acousto-Optical Deflectors (AODs). These are high-tech devices that can steer a laser beam incredibly fast—faster than the blink of an eye. However, there's a catch: if you try to paint a complex picture by just moving the brush around quickly, you run into a problem called "speckle."

The Problem: The "Static" Noise

Think of the laser beam like a choir of singers. If you ask them to sing different notes at the same time to create a chord, they usually blend beautifully. But in this laser world, if the notes (frequencies) aren't perfectly tuned, the sound waves crash into each other, creating a chaotic, noisy mess called interference.

In the past, to avoid this noise, scientists had to paint the picture line by line, like an old-school dot-matrix printer. They would draw one line, wait for the noise to settle, draw the next, and so on. This was accurate, but it was slow. It's like trying to paint a masterpiece by only moving your hand up and down, never side-to-side, until you've finished the whole canvas.

The Solution: The "Staggered" Dance

The authors of this paper found a clever way to paint the whole picture at once, instantly, without the noise. They call it an "incommensurately staggered frequency lattice." That's a mouthful, so let's use a better analogy: The Mismatched Clocks.

Imagine you have two giant clocks, one controlling the horizontal movement (Left-Right) and one controlling the vertical movement (Up-Down).

• The Old Way: Both clocks tick at the exact same rhythm. If they tick together, the laser spots line up perfectly, but they also crash into each other, creating that annoying "static" noise (the coherent artifacts).
• The New Way: The scientists set the two clocks to tick at slightly different, mismatched rhythms.

By making the timing of the horizontal and vertical movements "out of sync" in a very specific, mathematical way, the laser spots still land exactly where you want them to form the image. However, because the rhythms are mismatched, the "noise" spots that used to crash into each other now miss each other entirely. They are like two dancers who are so out of step that they never bump into each other, even though they are dancing on the same floor.

Why This Matters

This simple trick changes everything:

1. Speed: Instead of painting line-by-line (which takes time), they can now flash the entire image at once. It's the difference between typing a letter one letter at a time versus printing the whole page instantly.
2. No Feedback Needed: Usually, to get a perfect laser image, you need a camera to look at the result, tell the computer "that spot is too bright," and the computer tries again. This is a slow loop. The new method is so precise that it gets it right the first time, every time. It's like a master chef who can season a dish perfectly without ever tasting it first.
3. Versatility: They showed this works for simple shapes (like a square) and even complex, non-repeating images (like a pixelated version of a Mondrian painting).

The Real-World Impact

Why do we care about painting with light?

• Holding Atoms: Scientists use these laser "paintings" to trap and hold tiny atoms in mid-air. This is crucial for building quantum computers, where these atoms act as the processors.
• Microscopy: It allows scientists to look at cells and biological samples with incredible speed and precision, without damaging them.
• Materials Science: It can be used to "write" patterns into materials at a microscopic level, creating new types of sensors or circuits.

The Bottom Line

The authors discovered a mathematical "dance step" for laser beams. By intentionally making the timing of the laser's movements slightly mismatched, they eliminated the noise that used to ruin the picture. This allows them to project complex, high-speed, high-quality images instantly, opening the door to faster quantum computers and better medical imaging.

In short: They figured out how to make the laser "sing" a perfect chord without any of the screeching feedback, allowing them to paint with light faster than ever before.

Technical Summary

1. Problem Statement

The precise control of structured light fields is critical for applications ranging from optical trapping and quantum simulation to materials processing. While devices like Liquid-Crystal Spatial Light Modulators (LC-SLMs) and Digital Micromirror Devices (DMDs) offer high resolution, they suffer from slow response times (LC-SLMs) or low light efficiency and speckle noise (DMDs).

Acousto-Optical Deflectors (AODs) offer a superior alternative for high-speed applications due to their high damage thresholds and microsecond-scale response times. However, generating complex 2D patterns with AODs presents specific challenges:

• Sequential Scanning Limitations: Projecting complex patterns by sequentially scanning single spots is often too slow, inducing unwanted particle diffusion or heating in optical traps.
• Multi-tone Driving Artifacts: A faster alternative is "multi-tone driving," where multiple frequencies drive the AODs simultaneously to create a superposition of spots. In 1D, time-averaging naturally suppresses interference (speckle). However, in 2D, using two orthogonal AODs introduces coherent artifacts. When driving frequencies in the $x$ and $y$ axes are commensurate, specific pairs of spots interfere constructively in the time-averaged image, creating a static speckle pattern that degrades image fidelity.
• Current Workarounds: Existing solutions often rely on slow line-scanning protocols to avoid these artifacts, sacrificing the speed advantage of AODs.

2. Methodology

The authors propose a novel, feedback-free projection scheme based on incommensurately staggered frequency lattices to intrinsically suppress coherent artifacts without resorting to slow scanning.

Theoretical Framework

• Light Field Generation: The system uses two perpendicular AODs driven by RF signals $s_x(t)$ and $s_y(t)$ composed of multiple tones. The resulting time-averaged intensity $I_{avg}(x,y)$ consists of an incoherent part (desired spots) and a coherent part (artifacts).
• Artifact Mechanism: Coherent artifacts arise when the frequency difference between two spots in the $x$-axis equals the frequency difference between two spots in the $y$-axis ($f_x^k - f_x^l = f_y^q - f_y^p$). This condition creates a diagonal interference pattern ($\Delta x = -\Delta y$) that does not average to zero.
• Incommensurate Staggering: To suppress this, the authors restrict the driving frequencies to a specific subset of a refined lattice. Instead of using a regular grid where $\Delta n_x = \Delta n_y = 1$, they introduce staggering parameters $\Delta n_x$ and $\Delta n_y$ such that $\Delta n_y = \Delta n_x - 1$.
  • This creates an incommensurate relationship between the $x$ and $y$ frequency steps.
  • The condition ensures that the smallest common multiple of the frequency steps is maximized, effectively pushing the nearest interfering spots far apart in the image plane ($W_{min}$).
  • Since the interference amplitude decays with distance (due to the Point Spread Function tails), increasing $W_{min}$ drastically reduces the artifact magnitude.

Optimization and Control

• Amplitude Optimization: For separable patterns ($I(x,y) \approx I_x(x)I_y(y)$), the driving amplitudes are optimized via a simple least-squares minimization to match the target intensity.
• Phase Optimization: To maximize power efficiency and ensure the driving signals remain within the linear regime of the AODs, the phases of the driving tones are optimized using a modified Gerchberg-Saxton algorithm. This minimizes the peak-to-average power ratio of the RF signal.
• Non-Separable Patterns: For general non-separable images, the authors employ Non-negative Singular Value Decomposition (NNSVD). The target image is approximated as a sum of a small number of separable sub-images, which are then projected sequentially.

3. Key Contributions

1. Artifact Suppression Scheme: Introduction of a specific frequency lattice (incommensurate staggering) that intrinsically eliminates coherent artifacts in 2D AOD projections without feedback loops.
2. Speed vs. Accuracy Trade-off: Demonstration of a quantitative trade-off where increasing the staggering parameter ($\Delta n_x$) increases the driving period ($\tau$) but significantly improves image fidelity by increasing the distance between interfering spots.
3. Feedback-Free Operation: Unlike LC-SLM or DMD systems which often require iterative feedback to correct speckle, this method achieves high-fidelity projection through deterministic signal design.
4. Extension to Non-Separable Patterns: A hybrid strategy combining NNSVD decomposition with the staggered lattice approach, enabling the projection of arbitrary images (e.g., text, complex shapes) at speeds superior to traditional line-scanning.

4. Results

• Experimental Validation: The authors experimentally measured the coherent artifact amplitude as a function of spot distance. They confirmed that the artifacts decay according to a power law (due to the "fat tails" of the AOD PSF) rather than a Gaussian decay, making artifact suppression via distance critical.
• Separable Patterns:
  • Projections of a uniform square and a parabolic intensity pattern were tested.
  • Increasing the staggering parameter from $\Delta n_x = 1$ to $7$ reduced the Root Mean Square Error (RMSE) significantly, approaching the shot-noise limit.
  • The method achieved high accuracy with driving periods ($\tau$) in the range of tens of microseconds.
• Comparison with Line-Scanning:
  • The proposed method was compared against multi-line scanning protocols.
  • Result: The incommensurate staggering scheme was consistently faster than line-scanning for the same level of accuracy. Line-scanning requires dead times and repetitions to mitigate crosstalk, which drastically increases the total projection time.
• Non-Separable Patterns:
  • A pixelated image of a Mondrian painting was projected using NNSVD with a rank of 4.
  • The total projection time was approximately $216 \mu s$, significantly faster than the estimated $>236 \mu s$ (plus overhead) required for a full line-scan of the same image.

5. Significance and Outlook

• High-Speed Quantum and Atomic Physics: The ability to generate arbitrary 2D potentials in microseconds is crucial for manipulating ultra-cold atoms (e.g., in Rydberg arrays) and quantum simulators, where slow scanning induces heating and decoherence.
• Robustness: The feedback-free nature of the system makes it robust and easier to implement in complex experimental setups compared to iterative optimization methods.
• Dynamic Potentials: The short driving periods enabled by this method open the door to engineering time-dependent potentials (Floquet dynamics) with minimal micromotion and heating, a key limitation in current AOD applications.
• General Applicability: While demonstrated for optical trapping, the technique is applicable to any field requiring fast, high-fidelity 2D light shaping, such as materials processing and advanced microscopy.

In conclusion, this paper presents a fundamental advancement in AOD technology, transforming it from a device limited to simple scanning or separable patterns into a versatile, high-speed engine for arbitrary 2D light field generation.