DRAM optical metrology proof of concept

From die in hand to hidden geometry.

A 90 nm DRAM die used as a test object for non-destructive, below-Abbe optical metrology - revealing hierarchical architecture, periodic cell-scale signatures, and a path toward phase-derived geometry.

Discuss a DRAM sample Explore the proof point
Visible-light capture Millimeter-to-nanometer scale chain FFT and phase analysis
The test object

A readable hierarchy inside a 13.5 × 5 mm die.

The proof of concept follows the same DRAM die across physical scales: from the object in hand, through array and peripheral regions, into a global optical texture, and finally to a candidate cell/contact lattice.

The central question is not whether an optical image resembles a conventional die photograph. It is whether coherent, repeatable optical structure can be connected to known DRAM geometry and converted into useful metrology.

Powers of ten

The same device, read across scale.

Each view reveals a different physical hierarchy. The scale chain moves from the complete die toward periodic geometry associated with the active array.

13.5 mmPhysical DRAM die in hand
100-200 µmArray and peripheral architecture
~10 µmGlobal coherent optical texture
13.9 µmInferred field of view containing a cell-scale signature
~180 nmCandidate half-pitch / cell-feature interpretation
DRAM die fan bank and control strip regions used to navigate from macro architecture toward local geometry
Macro-to-local navigation through a DRAM die architecture
Reading the die like a map

Use known architecture to guide the optical path.

Long amplification regions, array fields, shared routing, peripheral blocks, and measured anchors provide a coordinate system for moving from the complete die toward the optical corridor of interest.

  • Peripheral control regions
  • Memory-array fields
  • Shared routing spine
  • Measured macro anchor
Where the story changes

A dense optical field resolves into repeated local geometry.

At an inferred field width of approximately 13.9 µm, the image produces staggered lobes and periodic information across the field. Local FFT analysis provides a second way to test whether the observed pattern is coherent rather than incidental texture.

Working interpretation: the image aligns most strongly with a staggered contact or landing-pad lattice. This is a plausible optical signature of 90 nm DRAM array geometry - a layout match, not yet direct proof.
Green optical field revealing a repeated staggered signature across a DRAM array region
Coherent green optical signature / inferred field width approximately 13.9 µm
POC observations

Geometry inferred from calibrated image and frequency analysis.

6.25 nm/pxWorking image scale derived from a 6 µm span across 960 pixels.
0.30-0.33 µmLocal FFT pitch measured in the repeated corridor texture.
154 × 94 nmApproximate dimensions of observed optical lobes in the working interpretation.
~29 nmObserved fringe spacing, treated as a likely moiré or harmonic feature rather than direct linewidth.

These values are proof-of-concept observations and interpretations from the supplied DRAM dataset. They are not production specifications or a final claim of resolved device geometry; direct correlative validation is the next step.

Phase-unwrapped and strain-related maps generated from the DRAM optical signal
Phase-unwrapped geometry and derived field interpretation
From HoloMoiré to HoloStrain

Convert a structured optical signal into metrology insight.

The longer-term opportunity is a measurement chain that turns the optical response into phase-derived displacement, pitch, strain, and residual-field maps.

  1. Capture the coherent optical signal across the target field.
  2. Estimate periodicity and spatial harmonics using FFT analysis.
  3. Unwrap phase and calculate local displacement or gradient behavior.
  4. Correlate derived geometry with SEM, layout, or other reference data.
The validation opportunity

Test whether optical field signatures can become a useful semiconductor metrology layer.

The next proof point is a focused correlation study: registered DRAM regions, known layout or SEM references, repeated optical captures, and measurable acceptance criteria for pitch, geometry, or process variation.

Plan a DRAM validation pilot