Skip to content
Data-driven sample preparation for semiconductor failure analysis

We put the last manual preparation step
of semiconductor failure analysis
under closed-loop numerical control:angle-accurate lapping, polishing and CMP
of microchip samples.

Better preparation results,

Patestra Una is a new class of benchtop system:
numerical control and native data integration,
brought to a process step that has never had either.

The last non-digitised step in semiconductor analysis.

Full-area chemical-mechanical preparation is in daily use in every analysis lab, and few methods are as versatile. Its principle is simple: abrasion and surface chemistry, tuned to each other. Yet it still runs by hand, without digitisation or process control, whereas every neighbouring step moved to automated, data-integrated systems years ago.

Among today’s problems, three weigh heaviest.

Time to result +
Thin, measure, repeat. Every pass costs bench time and a queue at the microscope, while a production line, a liability case or a patent analysis waits on the answer.
Sample risk +
The specimen is usually unique: a field return, or the one failing die out of a production lot. Material removal is one-way, and a pass too far leaves no way back.
Unwritten knowledge +
The process lives in practised hands. Such experience is hard to hire and takes years to build, and it stays with the bench where it was learned.

Your measurement data
deserves better than

Our proposal

One closed control loop for your sample preparation process.

Your measurement data goes in. The machine starts where your metrology left off.

Orientation and contact force are held numerically through the whole run, under full process load rather than merely under static conditions.

Every parameter is logged. Results thus repeat across operators and sites.

The missing piece between metrology and analysis, not another machine beside them.

Drawing: US 6,672,947 B2

Your process works.
Yet it wasn’t built for
what’s next.

Your trusty polishing machine: semi-automatic, hand-set,
data-blind, patented decades ago.
Still the state of the art. You know it shouldn’t be.

Patestra Una’s architecture is

Every device generation is denser and more three-dimensional, and asks more of the mechanical step than a practised hand can give. Patestra Una takes over exactly there: one benchtop machine for what a lab runs on several today, your measurement data driving the result, and the process written down instead of carried in someone’s head.

Field of application

From package level down to the die, from frontside through backside to the edge. Six examples of what Una is designed for; the range of lab processes it supports is even wider.

Package level
Die level
Frontside
Backside
Edgeside
Chip-on-demand
Single dies recovered intact from finished packages, through the stacked dies, interposers and hybrid bonds of 3D integration. Ready for re-use or further analysis.
Data in: X-ray · profilometry
Pre-etch removal +
Heatsinks, leadframes and interposers cleared first, so the wet etch starts from a favourable plane instead of crawling under barely soluble material for hours.
Data in: X-ray · profilometry
Localisation access +
Thermography and quantum sensing resolve only as well as they can get close: every micron of package material between sensor and active region costs resolution. Una approaches under a set residual-material budget with a hard stop — the device stays alive.
Data in: X-ray · profilometry
Delayering +
Metal and dielectric layers taken off in deliberate alternation with the dry etch: delayer, planarise, repeat. Uniform from edge to edge of the die — the wedge error is controlled, not set once — so contact geometry stays within tolerance through the whole sequence.
Data in: film thickness · profilometry
Backside thinning +
Silicon reduced to a set residual thickness across the batch, in far fewer thin-polish-measure iterations. Highly selective slurries end the removal at the target layer by themselves — but only under homogeneous pressure. Una holds that working point across the full die and the full run, so advances on the consumables side become fully usable.
Data in: film thickness · profilometry · stop-layer design data
Building on published CMP results: Herfurth et al., ISTFA 2024 [10.31399/asm.cp.istfa2024p0416] · IHP process patent EP 4 546 397 A1
Cross-sectioning +
A cut placed on target: the plane comes straight from the die face image and is held under load, so the usual grind-and-check iterations are no longer needed. This includes removing the substrate first and fine-polishing the remaining layer stack, comparable to a trusty Sagitta Centar.
Data in: die face image · X-ray · layout
Improving the workflows of labs across
IDM Foundry Fabless OSAT Analysis service providers Automotive & aerospace Defence Process & materials development Research institutes Authorities & forensics

What changes on your bench.

Today With Una

A process you can cope with today.

A workflow that stays future-proof.

01

Data

Measurement data can’t come in. Target geometry is therefore set by hand.

Profilometry, optical imaging, film thickness, X-ray and design layout go straight in. The target plane is derived automatically.

02

Accuracy

> 100 arcsec

manual · uncorrected

real-time + feedforward (design target)

03

Knowledge

Parameters aren’t logged, so nothing is reproducible across operators or sites.

Recipes transfer across people and sites. Onboarding no longer takes months of training.

04

Batch

Every specimen set up from scratch, one at a time.

At package level, the first specimen sets the recipe. The rest of the batch then runs it like a macro.

05

Calibration

Dial gauges and patience, repeated for every setup.

A software-guided workflow, set up and calibrated in under a minute.

06

Footprint

3 full lab workstations

several machines

1 full lab workstation

one Patestra Una

A closed loop,

Setting the plane accurately has never been the difficult part. What counts is how well that setting holds once the process is running, under load and over the length of a run. Una measures and corrects in real time, throughout.

Four subsystems,
one loop:

data coupled, orientation held, contact defined, patent pending.

Give me some tech detail!
Data

Native data interface

Measurement data passes straight in. The target plane is then set within a few clicks. When the run completes, set values, traces and the result are written back to your database. The preparation step finally speaks your lab’s language.

Orientation

Closed-loop orientation control

A high-stiffness parallel-kinematic stage carries the specimen. Process forces vary with the material stack and the abrasive, and they act on the specimen throughout the run. Real-time feedback and feedforward hold the pose against them: < 5 arcsec under all process conditions (design target), where manual machines typically drift past 100. Edge roll-off, commonly attributed to abrasive mechanics alone, is in large part a symptom of that drift; hold the pose, and the artefact is reduced with it.

Contact

Force-controlled vertical axis

A dedicated precision axis regulates contact force to < 0.1 N, target- or force-guided. Contact becomes a numerically defined event that repeats from run to run. The approach-check-correct loop is then shortened to a fraction of its passes, and abrasive consumption becomes measurable.

Consumables

Standard 8″ / 200 mm consumables

Una runs on the plates, pads, cloths and suspensions your lab already stocks. No proprietary consumables, no change of supplier.

Condition

Self-diagnosis

Every axis, force and metrology channel is sensed and logged. Maintenance becomes condition-based, and drift localises itself to the axis at fault. Angular deviation is known at any moment during the run; today it can only be reconstructed afterwards, from the finished surface.

Automating the obvious.

Metrology and imaging went digital years ago; mechanical preparation remained a craft. Patestra exists to close that gap, so that failure-analysis specialists and lab managers in semiconductor companies, research institutes and analysis providers can rely on preparation as they rely on the analysis that follows it.

Ulrich Wendorff, founder of Patestra, Munich
Photo © Cosima Teschemacher
Ulrich Wendorff
linkedin.com/in/ulrich-wendorff ↗

Why we build this.

Mechanical sample preparation is daily practice in every lab: a bench, a plate and a controlled contact take any material stack from millimetres to atomic layers, all at room temperature, with no vacuum, no plasma, no beam and no hazardous chemicals.

Its ceiling was never the physics; it is the machine that holds the plane. Having spent years at that bench, we strive to break the ceiling. Una wires this step into the lab and unlocks the value of your existing metrology. That is what Patestra is being founded for.

Supported by
Hochschule München University of Applied Sciences SCE — Strascheg Center for Entrepreneurship Zeidler Forschungsstiftung

One email starts the loop.

You know where preparation costs you time and specimens. Una addresses exactly that, and it is still at the stage where your requirements can shape the system. Write us what would make the difference in your lab; a question, an idea or a plain ‘hi’ works just as well.

discover@patestra.com