CASE STUDY 03 / SOLAR DEVICE DESIGN

Solar absorber discovery and CdTe replacement challenge.

FluxMateria ran a full solar absorber-and-contact workflow locally in 28.003 seconds, then ran a stricter replacement challenge in 51.78 seconds. The result: a literature-grounded CdTe stack winner, plus novel CuS and CuSe replacement lanes that cleared the defined incumbent-replacement gate.

Bulk winner: InP Device winner: CdTe + ZnO + Mo Build winner: Si Replacement lanes: CuS / CuSe Scenario alignment: 4/4
Download full white paper Download public data packet
28.0s
Local full workflow
51.8s
Replacement challenge
+20.65%
Top replacement gain
4/4
Scenario alignment
2
Novel lanes cleared gate

Why this matters

Solar engineering teams often split absorber choice, contact engineering, and build planning across separate tools or teams. This case study kept them inside one workflow and showed that the answer changes when the full stack is allowed to matter.

Winner Shift

Bulk and device answers diverged

The absorber-only screen picked InP. The device-stack screen picked CdTe. That is the central result of the study.

Engineering Value

Contacts changed the decision

CdTe moved from bulk rank 4 to device rank 1 once the front and back contacts were modeled directly.

Build Reality

The best first build was different again

Si emerged as the strongest conservative build package, which means the workflow now separates upside from practical first fabrication.

Pipeline

FluxMateria did not stop at absorber ranking. The workflow moved from bulk screening to contacts, then into a solar-native device layer, benchmark-aware uncertainty, and a build handoff.

1

Curated absorber framing

The workflow started from a known photovoltaic pool spanning crystalline silicon, III-V, CdTe, chalcopyrite, and exploratory thin-film families.

8 absorbers
2

Bulk absorber screen

Absorber-only ranking tested band-gap fit and bulk photovoltaic suitability before any interface correction was allowed.

Winner: InP
3

Batch contact sweep

Front and back contacts were screened in batch rather than one pair at a time, which cut runtime sharply and made full-stack scoring practical.

ZnO / Mo surfaced
4

Solar-native device ranking

Absorber quality, contact balance, stability, cost, manufacturability, and handoff readiness were ranked together inside a dedicated solar module.

Winner: CdTe
5

Calibration and uncertainty

The module was benchmarked against known references, then used to attach confidence, active-learning value, and experiment planning to the shortlist.

4/4 aligned scenarios
6

Prototype handoff

The workflow separated the strongest device-stack answer from the strongest conservative build package instead of forcing one winner for every question.

Build winner: Si
Absorber-Only
InP
The best direct bulk fit inside the curated pool.
Best Device Stack
CdTe + ZnO + Mo
The strongest full-stack answer once contact engineering entered the problem.
Best First Build
Si
The most conservative prototype package once build risk and handoff were considered.

Benchmark Snapshot

The dedicated solar module was benchmarked before the final rerun. The public claim is not that every metric improved equally; it is that the module stayed coherent across references and aligned on all public-facing scenarios.

1.0
Holdout family accuracy
The calibrated module kept the photovoltaic-family distinctions correct on the holdout set.
4 / 4
Scenario alignment
Balanced-stack, low-hazard, efficiency-frontier, and active-learning scenarios all aligned with expectation.
1.887
PCE proxy MAE
The calibrated module remained in the right range on the holdout set while preserving family behavior and scenario alignment.
0.0667
Manufacturing MAE
Calibration tightened the practical engineering layers that matter most for shortlisting and build planning.

Final Shortlist

The public shortlist focuses on known absorber families and the contact stacks they led to. It is designed to show the final engineering answer clearly and keep the story centered on the real device decisions.

Top full-stack candidate
CdTe
Cadmium telluride
21.1%
PCE proxy
72.0
Device-readiness score
ZnO
Front contact
Mo
Back contact
Front mismatch: 0.024 eV Back mismatch: 0.094 eV Rank shift: +3

The clearest proof that full-stack scoring changed the answer. Bulk-only screening would not have picked this device stack first.

Top bulk-origin candidate
InP
III-V semiconductor
20.29%
PCE proxy
71.2
Device-readiness score
ZnO
Front contact
Al
Back contact
Bulk winner Strong III-V signal Stayed in top tier

The absorber-only winner remained highly credible. The workflow did not invalidate it; it showed why it no longer stayed first once contacts mattered.

Top immediate build candidate
Si
Crystalline silicon
19.81%
PCE proxy
79.0
Prototype handoff priority
ZnO
Front contact
NiO
Back contact
High manufacturing readiness Strong cost signal Conservative build path

This is the strongest practical build package in the workflow, which is different from the highest-upside full-stack answer.

High-efficiency comparator
GaAs
III-V semiconductor
19.27%
PCE proxy
64.4
Device-readiness score
ZnO
Front contact
NiO
Back contact

A serious comparator that reinforces the point: the shortlist stayed inside real photovoltaic families throughout the workflow.

Important Edge Case

CuInSe2 did not survive the strict public shortlist, but it remains scientifically important. It is a literature-backed chalcopyrite absorber family and a useful calibration target for future solar runs.

The value of this case study is not that every literature-known solar material ranked first. The value is that FluxMateria kept real families in view, exposed where contact engineering changed the answer, and left behind a sharper prototype shortlist plus a clear set of refinement targets.

Second-Pass Replacement Challenge

After the stack workflow identified CdTe + ZnO + Mo as the strongest device answer, FluxMateria asked a harder question: can a novel absorber lane beat that incumbent on a broader replacement objective?

Lead Replacement Lane
CuS
Best stack: ZnSe front / Al back
79.0 Replacement score
+20.65% Vs. CdTe baseline
78.4 Readiness score

CuS became the highest-upside replacement hypothesis because it combined a strong absorber window, defect-tolerant chemistry, low passivation burden, cost advantage, and credible architecture fit.

Second Replacement Lane
CuSe
Best stack: ZnO front / Mo back
75.4 Replacement score
+15.09% Vs. CdTe baseline
78.4 Readiness score

CuSe is the more literature-adjacent replacement lane. It kept the ZnO / Mo contact logic from the CdTe baseline while reducing hazard and supply-chain pressure in the model.

Why this is a validation point: FluxMateria was not pointed at a desired answer. In a few local passes, it converged on copper and tin chalcogenide directions that the photovoltaic literature has spent years exploring as earth-abundant thin-film alternatives. That does not prove a commercial device yet; it proves the computational search is landing on real, literature-adjacent physics rather than decorative novelty.

Literature Context

The shortlist and replacement lanes converged on photovoltaic families the field already treats as serious, while still pushing beyond the most obvious incumbents.

Established Families

Si, CdTe, and III-V families are real photovoltaic lineages

NREL best-cell-efficiency tracking keeps crystalline silicon, CdTe, and III-V families on the same chart of serious photovoltaic technologies. That supports the idea that FluxMateria stayed inside real solar design space.

CdTe Contacts

CdTe + ZnO is physically serious

CdTe front-contact band alignment has been studied directly, including ZnO-based contact layers. That makes the winning CdTe + ZnO front-contact pairing more than a generic software preference.

Silicon Contacts

Why silicon stayed strong on build readiness

Modern silicon-device performance depends heavily on passivating and selective contacts. That is why silicon can remain a top conservative build path even when a different material wins the stack-level score.

Chalcopyrite Back Contacts

Why CuInSe2 remains important

CuInSe2 and related chalcopyrite absorbers are strongly linked to Mo / MoSe2 back-contact behavior in the literature, which is why CuInSe2 remains a meaningful refinement target even though it did not make the final public shortlist.

Copper Chalcogenides

CuS and CuSe are not random chemistry

CuS has appeared in thin-film photovoltaic device work, while CuSe has been studied in photovoltaic devices and as a CdTe absorber-layer modifier. FluxMateria's CuS / CuSe replacement result lands near real research directions, but the exact replacement stacks still require fabrication.

Tin Chalcogenides

The near-miss lane is strongly literature-backed

SnS, SnSe, SnSx, and mixed Sn-S-Se absorber families are active low-cost solar research lines. In this campaign they scored strongly as frontier candidates, but did not clear the stricter CdTe replacement gate.

Careful interpretation: the literature does not say that CuS + ZnSe + Al or CuSe + ZnO + Mo are already proven commercial replacements for CdTe. It says the chemical neighborhoods FluxMateria selected are scientifically serious. The novelty is that FluxMateria connected absorber choice, contact fit, passivation burden, cost, and replacement scoring into one rapid decision layer.
NREL Best Research-Cell Efficiencies
https://www.nrel.gov/pv/assets/pdfs/best-research-cell-efficiencies.20210926.pdf
Band alignment of front contact layers for high-efficiency CdTe solar cells
https://www.osti.gov/servlets/purl/1534330
Silicon heterojunction solar cells with passivating contacts: Classification and advanced fabrication strategies
https://www.sciencedirect.com/science/article/abs/pii/S1369702124002088
CuInSe2 thin films: preparation, structure, properties, and solar cells
https://www.osti.gov/biblio/5168746
Mo/Cu(In,Ga)Se2 back interface chemical and optical properties for ultrathin CIGSe solar cells
https://www.sciencedirect.com/science/article/abs/pii/S0169433211017880
CuS thin-film photovoltaic device context
https://www.mdpi.com/1996-1073/13/3/688
CdSe/CuSe flexible photovoltaic device context
https://pubs.rsc.org/en/content/articlelanding/2013/cp/c3cp50435b
Earth-abundant chalcogenide thin-film solar cell review
https://www.frontiersin.org/journals/chemistry/articles/10.3389/fchem.2019.00297/full
SnSe/SnS bilayer absorber context
https://www.sciencedirect.com/science/article/pii/S0038092X23008083
What This Study Achieved

The remaining step is device fabrication and measurement.

That is exactly the point of the study: not to replace solar-device fabrication, but to make that fabrication sharper, cheaper, and more defensible by turning a broad absorber question into a much smaller set of higher-quality full-stack decisions.

Download white paper Download public data packet
This is an in-silico engineering-screening study. Real device fabrication, efficiency certification, durability testing, and manufacturing validation still require laboratory work and production-scale engineering.