Novel Light-Field Imaging Device with Enhanced Light Collection for Cold Atom Clouds

JINST 17 P08021

This work was supported by the Department of Energy, Laboratory Directed Research and Development program at SLAC National Accelerator Laboratory, under contract DE-AC02-76SF00515

Sanha Cheong on behalf of SLAC MAGIS Group

CPAD Workshop 2022

Session WG4: Quantum and Superconducting Detectors

Nov. 30^th, 2022

Atom Interferometers

Atom Interferometer

• Optical interferometer “flipped”
• De-localized matter-wave over two paths
• Lasers for excitation, momentum transfer, etc.
• Phase sensitive to spacetime area, properties of two states, etc.

Gradiometer or “Differential” Interferometer

• Two atom interferometers separated by some distance
• Driven by a common laser ⇒ Laser noises cancel
• Differential phase sensitive to baseline length and energy splitting

Image from�Hogan Lab Website

Image from�arXiv:2104.02835

2 / Light-field Imaging for Atom Clouds / CPAD 2022

MAGIS Concept

MAGIS Experiment

• Sr clock gradiometer separated by large baseline
• Sensitive to photon- or electron-coupling ultra-light dark matter
• Sensitive to mid-band GW
• 100m experiment currently under construction @ Fermilab
• Science paper: arXiv:2104.02835
• Also highlighted in Snowmass CF2 whitepaper recently

3 / Light-field Imaging for Atom Clouds / CPAD 2022

Imaging Requirements at Atom Interferometers

Expected for MAGIS and typical for many atomic experiments:

• Fluorescing cold atom clouds
• O(mm) in overall size
• O(100 µm) feature sizes
• Capturing 3D light field
• Intensity + direction approach to imaging
• Increase the light gathering ability of imaging
• Challenges
• Limited number of viewports in experiments
• More light → larger lens → shallow depth of field
• Common limitation in many atomic experiments

arXiv:2203.14915

4 / Light-field Imaging for Atom Clouds / CPAD 2022

Mirrors to the Rescue

Geometry of a single “view”

Redirect light that is not traveling to the lens with mirrors

Each view direction from the object is mapped to a unique point on the sensor

5 / Light-field Imaging for Atom Clouds / CPAD 2022

Spatially Multiplexed Light Field Imaging

Array of mirrors around the atom cloud

• Place virtual objects on the focal plane
• Design to capture light from one hemisphere

Images of different views through different mirrors

Capture 3D information & collect more light

6 / Light-field Imaging for Atom Clouds / CPAD 2022

Demonstrator

• 3D printed mechanical support for 5mm mirrors
• Target held with 100 µm fibers
• Optical alignment with In-situ grid ~ 1^o

7 / Light-field Imaging for Atom Clouds / CPAD 2022

Test Object: DOE Cube

• 3D printed test object�(projection micro-stereolithography)
• Absorbs at 405 nm
• Fluoresces at 430 - 550 nm

(This is 150 µm thread, final images taken with 100 µm threads)

Microscope images of each face

CAD model

1 mm

1 mm

1 mm

143 µm

8 / Light-field Imaging for Atom Clouds / CPAD 2022

3D Reconstruction & View Synthesis

• Very popular in computer vision/graphics community
• Take multiple views from different angles
• Learn underlying scene/volume
• Generate new views @ different angles
• NeRF: Neural Radiance Fields
• NN to encode radiance field over volume
• NeuS: Neural Implicit Surfaces
• Uses “signed distance function”
• Focuses on surface reconstruction

9 / Light-field Imaging for Atom Clouds / CPAD 2022

3D Reconstruction Demo Examples

10 / Light-field Imaging for Atom Clouds / CPAD 2022

Results: Raw Images

• Patch extraction of each view
• Select good alignment and illumination
• 77/90 selected

11 / Light-field Imaging for Atom Clouds / CPAD 2022

Results: Learning 3D Structure via NeuS

12 / Light-field Imaging for Atom Clouds / CPAD 2022

Results: 3D Reconstruction

• 90 mirrors, 77 views selected for reconstruction
• Reconstruction using NeuS

Reconstructed mesh surface

Comparison of CAD, microscope,�and learned depth map

Interpolated views (SDF + Color)

13 / Light-field Imaging for Atom Clouds / CPAD 2022

Results: Reconstruction Performance

• Fourier Shell Correlation (FSC) as a measure of 3D reconstruction performance
• Correlation between two density distributions
• Comparison between two independent reconstructions (using half of the dome)
• System achieves design specifications!

14 / Light-field Imaging for Atom Clouds / CPAD 2022

Summary

• Successful demonstration of spatially multiplexed�light-field imaging for 1mm³ object with 100 µm features
• Paper published at JINST 17 P08021
• Broad applicability to atom interferometry and beyond
• Ongoing / Future works
• Installing at Stanford lab and imaging real atom clouds
• Full 3D reco. of MOT and dipole traps

​

​

​

​

• Laser wavefront aberration studies using 3D atom clouds�(collaboration with Kovachy lab @ Northwestern)

15 / Light-field Imaging for Atom Clouds / CPAD 2022

Thank You

Ariel Schwartzman

Murtaza Safdari

Sanha Cheong

Michael Kagan

Sean Gasiorowski

Maxime Vandegar

Joseph Frisch

MAGIS-100 - BSM Physics Potential

https://arxiv.org/abs/2203.14915

17 / Light-field Imaging for Atom Clouds / CPAD 2022

Determining Mirror Parameters

On Axis View

Off Axis View

18 / Light-field Imaging for Atom Clouds / CPAD 2022

Mirror Parameters - Using The Magic of Ellipses!

Reflections, Directrix, Rays

Working Example

19 / Light-field Imaging for Atom Clouds / CPAD 2022

Results: Alignment Tolerance

• Fourier Shell Correlation (FSC) as a measure of 3D reconstruction performance
• Simulation study with tomographic atom cloud density
• 100 um resolution possible with upto 0.6 mm or 2^o alignment errors

20 / Light-field Imaging for Atom Clouds / CPAD 2022