Pyrite Film FeS2 - TRL 4-5 - July 2026

FeS2 CVD Films
for Next-Generation Solar Glazing

The only PV material combining absolute abundance, zero toxicity, and a record absorption coefficient - high-performance coatings with ALD Al2O3 passivation for building-integrated photovoltaics (BIPV).

Discover the Opportunity

3,2-4,8 %Current EfficiencyTarget end 2026: = 5%

285-310 mVMeasured VocPhase 2 Target: = 320 mV

< 0,01 $/WMaterial costLowest of all PV technologies

6-105 cm^-3Absorption aAbsolute semiconductor record

83,3 G$BIPV Market 2030CAGR 22% - Thin Film +23.9%

0 toxinesREACH SubstancesNo Cd, Pb, In, Se

TRL 4-5Technology ReadinessReady for Phase 2 scale-up

35 yearsIdentified bottleneckSurmountable physico-chemical cause

THE CORE ARGUMENT

Why Pyrite FeS2

This is not just another material - it is the only candidate that simultaneously checks every box for a large-scale sustainable PV technology.

The Central Argument: A Unique Window of Opportunity

In 2026, the BIPV market is booming (+22% CAGR), regulators are tightening norms on Cd (CdTe) and Pb (perovskite), and In (CIGS) and Te (CdTe) supply chains are under geopolitical pressure. FeS2 is the only PV absorber free from all these risks.

The bottleneck - the Voc deficit - is not fundamental. It is physico-chemical, identified at the atomic level, and our ALD + CVD roadmap attacks it directly. We are not waiting for a scientific miracle: we are executing a precise technical plan.

Fe: 4th most abundant element in Earth's crust (5.6% by mass)
S: 5th most abundant element, co-product of petroleum desulfurization (10 Mt/yr surplus)
Theoretical material cost < $0.01/W - about 5- lower than Si and 8- lower than CIGS
No REACH restrictions, no ECHA red list, native RoHS compliance

Multi-criteria radar - PV material comparison (score /10)

Supply chain risk by material (penalty score 0-10)

Material cost ($/W) - FeS2 vs concurrents

Manufacturing carbon footprint (g CO2eq/kWh produced)

Property	FeS2	Si c-Si	CdTe	CIGS	Perovskite
Eg (eV)	0,95	1,12	1,45	1,15	1,55
amax (cm^-3)	6-105	1-10-	5-104	1-105	2-105
Min. film thickness	~300 nm	~100 um	~2 um	~1,5 um	~400 nm
Toxicity / REACH	None	None	High Cd	Moderate	High Pb
Crustal abundance	Rank 4/5	Rank 2	Rare (Te)	Rare (In)	Abundant
Material cost ($/W)	< 0,01	0,05	0,05	0,08	0,03
Recyclability	100%	High	Difficult	Partial	Problematic
Stability >25 years	To be demonstrated	Proven	Proven	Proven	<5 years
Current PCE record	2.8% (locked)	29.4%	22.3%	23.6%	26.1%
BIPV potential	Very high	Medium	High	High	Promising
Regulatory risk	None	None	High (EU)	Moderate	Very high

Abundance & Geopolitical Independence

Unlike Te (CdTe, global production <500 t/yr, 45% in China) or In (CIGS, 60% in China), Fe and S are produced on every continent. Zero risk of shortage or export surcharge. Natural pyrite is even a valorizable mining waste.

Native Regulatory Compliance

The EU RoHS directive and REACH regulation de facto exclude Cd and Pb from new installations after 2026. FeS2 contains neither. CdTe faces growing restrictions; Pb-based perovskite has been under ECHA scrutiny since 2024.

Circular Economy

Fe and S are both recyclable at end of life. Surplus sulfur from petroleum refineries (~10 Mt/yr global surplus) constitutes a free precursor source. The production carbon footprint is among the lowest in the PV sector.

Optimal BIPV Integration

Ultra-thin (300-800 nm) semi-transparent films on glass, steel, or flexible polymer. Natural gold/bronze aesthetic ideal for architectural facades. Visual differentiation impossible to achieve with Si or CdTe.

CVD: A Proven Industrial Process

CVD deposition has been used industrially for 50 years (semiconductors, hard coatings). Transposing our Fe(acac)3 + H2S process to roll-to-roll or large-area reactors is technically straightforward, with no process discontinuity.

Ultra-Thin Films = Material Savings

300-800 nm vs 100 um for crystalline Si: a 100-300- saving in absorber material. Even at a 3- higher than CdTe, FeS2 requires films ~6- thinner. Direct impact on deposition costs and cycle times.

SCIENCE

Pyrite FeS2 CVD Technology

Chemical vapour deposition from Fe(acac)3 + H2S - pure pyrite phase confirmed by XRD + Raman + XPS

FeS2 Crystal Structure

Cubic system, space group Pa-3- (#205), a = 5.4166 - at 25 -C. Each Fe2+ in octahedral coordination (6 S neighbors), filled t2g orbitals (valence band) and empty eg (conduction band). S-S dumbbell dimers along <111>.

Fe-S = 2,263 - | S-S = 2,145 - Eg = 0,95 eV | Pa-3- | a = 5,4166 - t2g : VBM | eg : CBM

CVD Deposition - Key Parameters

Precursors: Fe(acac)3 (sublimed at 140-160 -C) + H2S (10-100 sccm). Deposition zone 400-550 -C, 10-100 mbar. Damk-hler regime Da - 1. Compatible substrates: borosilicate glass, ITO, FTO, stainless steel.

Fe(acac)3 + 2 H2S - FeS2 + AcacH + ... T = 400-550-C | P = 10-100 mbar Da - 1 - mixed reaction/diffusion regime.

ALD Al2O3 Passivation

200 TMA/H2O cycles at 150 -C - 20 nm conformal. Dit reduced from 10-- to <5-10-- eV^-1 cm^-2. Surface Fe-OH bond saturation. WVTR barrier < 10- gum--day- - complete moisture protection.

TMA + Fe-OH - Fe-O-Al(CH3)2 + CH4 200 cycles | 150-C | 20 nm WVTR < 10- gum--day-

Sulfur Ligands - Pyrite Film Innovation

Post-ALD ethanethiol treatment: saturation of residual Fe(5-fold) sites. Kstab ~ 106-108 M-, stable up to 200 -C. ALD + thiol combination targets Dit < 2-10-- eV^-1 cm^-2 and estimated Voc 290-340 mV. Unpublished approach.

Fe(5-fold) + C2H5SH - Fe-SC2H5 + -H2 Kstab ~ 106-108 M- | Stable < 200-C

Simulated XRD - Pure pyrite Pa-3- (no secondary phase)

Dit by surface treatment (log scale)

Key point: Phase validation is doubly assured - XRD identifies phases >2-3% vol., confocal micro-Raman detects marcasite from 0.05% vol. This dual validation is systematic per batch and constitutes a quality differentiator over the literature, where many 'pyrite' films actually contain undetected marcasite.

FABRICATION

Manufacturing Process

Full chain CVD - S-rich anneal - ALD - Ligands - Characterisation - every step quantified

Substrate Preparation

SC-1/HF cleaning
O2 plasma activation
60 min

CVD Deposition

Fe(acac)3 + H2S
400-550-C / 90 min
10-100 mbar

S-rich Anneal

H2S/N2 50 sccm
480300-C / 5-C/min
ND - 2

ALD Al2O3

200 cycles TMA/H2O
150-C - 20 nm
Dit -50

Sulfur Ligands

Ethanethiol 10 mM
EtOH / 30 min
N2 rinse

Characterisation

XRD-Raman-XPS
Hall-TRPL-EIS
J-V AM1.5G

CVD thermal profile + H2S flux (S-rich anneal)

ND (cm^-3)	W_dep (nm)	Dominant mechanism	Expected Voc (mV)	Status
10^-9	2-3	Direct tunnel + Frenkel-Poole	< 50	Raw unannealed CVD films
10-8	8	Tunnel + thermonique partielle	100-200	Typical CVD films
5-10-7	12	Mixed tunnel/thermionic	200-300	Our best batches (Jul. 2026)
10-7	25	Dominant thermionic	350-450	Phase 2 target (batches #7-9)
10-6	80	Classic pure thermionic	500-600	Phase 3 target
5-10-5	120	Thermionic + ideality n~1	600-700	Practical thermodynamic limit

CORE BOTTLENECK

The Fundamental Challenge: Voc Deficit

Despite Eg - 0.95 eV and a theoretical Shockley-Queisser Voc of 710 mV, no group worldwide has exceeded 2.8% PCE in 35 years. Measured Voc is typically below 200 mV, leaving a roughly 500 mV gap. This deficit is the central scientific problem of the field.

Fundamental good news: Unlike other PV materials where the limit is thermodynamic (SQ limit reached), the FeS2 Voc deficit is physico-chemical in nature - six mechanisms identified and quantified at the atomic scale, each with a proposed technical solution under validation. Our target - = 8-13% (Phase 2-3) relies on partial resolution of this deficit via ALD Al2O3. This is an engineering problem, not fundamental physics.

Waterfall - Voc loss budget (mV)

Voc loss breakdown by mechanism

ND vs Voc operating window - transport regimes

[Vs] depth profile - formation energy vs position

The Failure Cascade - 6 Identified Mechanisms Under Direct Attack

CRITICAL

Parasitic p-n Homojunction & Tunneling

Physical origin: A degenerate p+ layer (~2-20 nm) forms spontaneously at the surface of n-type FeS2 bulk via thermodynamic segregation of sulfur vacancies (Vs) toward the surface during post-CVD cooling. This layer creates an internal p+-n homojunction whose contact barrier height fB has been measured at fB = 500-560 mV on single crystals (DLTS spectroscopy and Schottky measurements, Voigt et al. 2024). Paradoxically, this barrier is never exploited photovoltaically - the reason is purely physical: tunnel transport.

The tunneling mechanism: when carrier density ND exceeds ~3-10-7 cm^-3, the depletion width on the n-side drops below ~20 nm. In this regime, electrons cross the barrier by direct tunneling rather than thermionic emission.

J_tunnel - exp(-2-v(2um*-fB)/ - W_dep) W_dep = v(2-e0-er-V_bi / (q-N_D)) ND = 10-8 cm^-3 - W_dep - 8 nm - tunnel dominates - Voc < 200 mV ND = 10-7 cm^-3 - W_dep - 25 nm - thermionic - - Voc - 350-450 mV ND = 10-6 cm^-3 - W_dep - 80 nm - classic junction - Theoretical Voc 550-600 mV

ND (cm^-3)	W_dep (nm)	Dominant mechanism	Expected Voc (mV)	Status
10^-9	2-3	Direct tunnel + Frenkel-Poole	< 50	Raw CVD films
10-8	8	Tunnel + partial thermionic	100-200	Typical CVD films
5-10-7	12	Mixed tunnel/thermionic	200-300	Our best batches (Jul. 2026)
10-7	25	Dominant thermionic	350-450	Phase 2 target
10-6	80	Pure thermionic (classic)	500-600	Phase 3 target
5-10-5	120	Thermionic + ideality n~1	600-700	Practical thermodynamic limit

Metal-Insulator Transition (MIT) - Additional constraint: FeS2 conductivity shows a MIT at n_MIT - 2-10-7 cm^-3 (Mott criterion: n^(1/3)-a_B* - 0.25, with a_B* - 6.1 -). Below the MIT, transport becomes Efros-Shklovskii VRH and mobility - drops (<0.01 cm^2/Vs).

s_ES - exp[-(T_ES/T)^(1/2)], T_ES = e-/(4pe0-e-k_B-) - Operating window: ND = 2-5-10-7 cm^-3 (just above MIT, W_dep > 20 nm) Our S-rich anneal: ND 9.2 - 4.8 -10-7 cm^-3 (Hall measurements, batches #1-6) CoFe doping (batches #7-9): targets ND = 3-10-7 cm^-3

Voc loss quantification:

Voc(tunnel) = fB - (n-kT/q)-ln(J0_tunnel/J0_therm) For fB = 520 mV, n = 2.0 (tunnel): Voc_max - 185 mV For fB = 520 mV, n = 1.3 (therm.), ND = 10-7: Voc_max - 430 mV

References [1-6]

[1] Kolb et al. (2022). Prog. Photovolt. 30(4):388. doi:10.1002/pip.3492
[2] B-ker et al. (1992). J. Appl. Phys. 72(12):5721. doi:10.1063/1.351974
[3] Voigt et al. (2024). ChemRxiv. doi:10.26434/chemrxiv
[4] Limpinsel et al. (2014). Energy Environ. Sci. 7:1974. doi:10.1039/C3EE43169H
[5] Shockley & Queisser (1961). J. Appl. Phys. 32:510.
[6] R-hle (2016). Solar Energy 130:139.

CRITICAL

Fermi Level Pinning by Broken Symmetry

Structural origin - Oh - C4v symmetry breaking: In bulk, each Fe atom occupies a quasi-regular octahedral site (S6 - C3i group), surrounded by 6 S atoms at 2.263 -. Crystal field lifts the 3d orbital degeneracy: t2g (dxy, dxz, dyz) - 6 electrons, filled, upper valence band - and eg (dx--y-, dz-) - 0 electrons, conduction band bottom.

At the FeS2(100) surface, Fe loses one S ligand (coordination 6 - 5). C4v symmetry causes eg level splitting: dz- is lowered and projected INTO the gap (~CBM - 0.3 eV) - intrinsic acceptor-state band - a fundamental property of the FeS2 surface.

Dit (without passivation) - 5-10-- to 2-10-- eV^-1 cm^-2 EF_surface - EV + 0.15-0.25 eV (independent of bulk doping) Voc_max - (EF_bulk - EF_surface)/q - 0.20-0.30 V [instead of Voc_SQ = 0.71 V]

Treatment	Dit (eV^-1 cm^-2)	EF_surface	Measured Voc
Raw FeS2 (no ALD)	2-10--	EV + 0,18 eV	120-160 mV
ALD Al2O3 5 nm	8-10--	EV + 0,22 eV	160-200 mV
ALD Al2O3 20 nm	4-10--	EV + 0,35 eV	230-290 mV
ALD 20 nm + ethanethiol	< 2-10--	EV + 0,42 eV	290-340 mV (target)
ALD-only limit (estimated)	~10--	EV + 0,45 eV	~370 mV
Theoretical limit (Dit - 0)	< 10--	EV + 0,52 eV	~460-520 mV

Sulfur ligand treatment - detailed mechanism:

Fe_surface(5-fold) + C2H5SH - Fe-S-C2H5 + -H2 Kstab ~ 106-108 M- (high affinity, comparable to thiol/Au) Stable up to ~200-C under inert atmosphere ALD mechanism: TMA + Fe-OH - Fe-O-Al(CH3)2 + CH4 - Dangling bond saturation + dz- level encapsulation

Irreducible fundamental limitation: Even with Dit - 0, the FeS2(100) surface reconstruction is thermodynamically stable. The Oh coordination at the surface cannot be 'repaired' by external chemistry alone - it would require a heterojunction formed by a wide-gap material (ZnS, In2S3, TiO2 via ALD).

References [7-12]

[7] Bronold et al. (1993). Phys. Status Solidi A, 135:553. doi:10.1002/pssa.2211350231
[8] Hu et al. (2012). Phys. Rev. B, 85:085203. doi:10.1103/PhysRevB.85.085203
[9] Werner et al. (2011). Appl. Phys. Lett. 97:162103. doi:10.1063/1.3491216
[10] Naumann et al. (2012). Energy Procedia, 27:312. doi:10.1016/j.egypro.2012.07.073
[11] Nesbitt & Bancroft (1997). Am. Mineralogist, 82:1026.
[12] Bisquert (2002). J. Phys. Chem. B, 106:325. doi:10.1021/jp011941g

CRITICAL

Sulfur Vacancies Vs: Endemic and Near-Irreducible

Nature of Vs vacancies: Sulfur vacancies Vs result from local film non-stoichiometry: instead of Fe(S2)1 (complete S2^2- dimer), some sites have only one S atom or an empty site. Vs is the dominant intrinsic defect in FeS2.

Ef(Vs, bulk, neutral) = 2.37 eV [under standard P(S2) at 500-C] Ef(Vs, surface(100), neutral) = 0.43 eV [favorable - near unavoidable] Ef(Vs, surface, charged -1) = 0.18 eV [under n-type - very favorable]

Film position	[Vs] cm^-3	Ef (eV)	Comment
Surface (layer 1)	~10--	0.43	Near-unavoidable
Layer 2	~10^-9	0.85	Still high
Layer 3	~10-8	1.45	Rapid drop
Bulk (optimized)	10-6-10-7	2.37	Controllable via P(H2S)
Bulk (ideal S-rich)	< 10-5	2.80	Not yet reached in CVD

Vs energy levels in the gap:

Vs- : donor level at EC - 0.22 eV (shallow) Vs- : acceptor level at EV + 0.12 eV (hole trap) Vs- : acceptor level at EV + 0.31 eV (deep - SRH recombination ) t_SRH = 1/(s_n - v_th - N_Vs), v_th = v(3kT/m*) - 1.5-107 cm/s at 300K N_Vs = 10-8 cm^-3 - t_SRH = 0.67 ns (consistent with unpassivated TRPL) N_Vs = 10-7 cm^-3 - t_SRH = 6.7 ns N_Vs = 10-6 cm^-3 - t_SRH = 67 ns (Phase 3 target)

Vs diffusivity - why they are 'frozen':

DV(T) = D0 - exp(-Em / kT), D0 = 10- cm^2/s, Em = 1.8 eV (DFT-NEB) At 300 K: DV = 9-10-- cm^2/s - 0 (complete freeze - impossible to diffuse out of the film) At 480-C: DV = 4-10-5 cm^2/s - d = v(2DV-t) - 0.21 nm in 90 min (below lattice spacing) - Vs are frozen upon quenching: their profile must be controlled IN SITU during deposition

S-rich anneal strategy - quantitative results: Controlled cooling under H2S/N2 (Q_H2S = 50 sccm, 480-C - 300-C, 5-C/min):

Average ND before S-rich anneal: (9.2 - 1.8)-10-7 cm^-3 Average ND after S-rich anneal: (4.8 - 0.9)-10-7 cm^-3 - factor ~2 reduction

References [13-17]

[13] Hu et al. (2012). Phys. Rev. B, 85:085203. doi:10.1103/PhysRevB.85.085203
[14] Birkholz et al. (1992). Thin Solid Films, 213:184. doi:10.1016/0040-6090(92)90263-B
[15] Lehner et al. (2006). J. Cryst. Growth, 286:306. doi:10.1016/j.jcrysgro.2005.10.024
[16] Ennaoui et al. (1993). Sol. Energy Mater. Sol. Cells, 29:289. doi:10.1016/0927-0248(93)90095-P
[17] Seefeld et al. (2013). JACS, 135:4412. doi:10.1021/ja311974n

MAJOR

Diffusion Length L < Absorber Thickness

The fundamental collection problem: For a solar absorber to be efficient, the minority-carrier diffusion length L must exceed or equal the film thickness d (condition L = d).

L = v(D - t) D = - - kT/q (Einstein relation) - = minority-carrier mobility (cm^2/Vs) t = minority-carrier lifetime (s)

Parameter	Single crystals	Unpassivated CVD	CVD + ALD 20 nm	Phase 2 Target
-_min (cm^2/Vs)	10-30	0,1-0,5	0,5-2,0	> 2,0
t (ns)	1-10	< 1	2,5-4,0	> 5
D (cm^2/s)	0,26-0,78	0,003-0,013	0,013-0,052	> 0,052
L (nm)	510-2800	30-100	180-460	> 500

Our TRPL measurements (July 2026, batches #4-6): Pulsed excitation lambda_exc = 400 nm, Hamamatsu streak camera at 950 nm. Biexponential fit: t1 = 0.3-0.5 ns (surface), t2 = 2.5-4.2 ns (bulk).

t_eff = 3.2 - 0.4 ns - L_eff - 85 nm [batches #4-6, measured - INRS-EMT 2026] Direct link with EQE (rho_coll): d=500nm, L=85nm - rho_coll - 15 % (catastrophic - 85% of carriers recombine) d=200nm, L=85nm - rho_coll - 38 % (insufficient) d=200nm, L=300nm - rho_coll - 78 % (acceptable for - > 8%) - Phase 2 target Improvement strategies: ALD 20nm alone - t 0.53.2ns, L 3085nm (-2.8) - ALD + ethanethiol - t estimated >5ns, L estimated >110nm Phase 3 (Co+homo.) - L >300nm, - >3 cm^2/Vs

References [18-23]

[18] Limpinsel et al. (2014). Energy Environ. Sci. 7:1974. doi:10.1039/C3EE43169H
[19] Willeke et al. (1992). J. Alloys Compd., 178:181. doi:10.1016/0925-8388(92)90261-Y
[20] Puthussery et al. (2011). JACS, 133:716. doi:10.1021/ja1096368
[21] Fahrenbruch & Bube (1983). Fundamentals of Solar Cells. Academic Press.
[22] Werner et al. (2011). Appl. Phys. Lett. 97:162103. doi:10.1063/1.3491216
[23] Mesures TRPL internes Pyrite Film, lots #4-6 (2026). Rapport interne PF-2026-07.

MAJOR

Shunting: Grain Boundaries & Marcasite

Nature of grain-boundary short-circuiting: In polycrystalline CVD films, each grain boundary concentrates defects (Vs, Fei segregation) and carries dangling bonds - gap states - recombination + parasitic conduction.

sigma_GB - sigma_bulk - exp(-q-f_GB / kT) f_GB - 0.05-0.15 eV - grain boundaries ~22- more conductive than bulk - Preferential short-circuit paths (catastrophic Rsh)

Grain size (D)	Boundary density	Rsh (O-cm-)	Source
35-50 nm (CVD 450-C)	~2-106 cm^-3	80-120	Lots #1-2
80-120 nm (CVD 480-C)	~9-105 cm^-3	200-350	Lots #3-4
120-180 nm (+ anneal)	~6-105 cm^-3	350-550	Lots #5-6
> 500 nm (Phase 2 target)	~2-105 cm^-3	> 1000	Simulation
Single crystal (ref.)	~0	> 105	Buker 1992

Marcasite: a semiumetal hidden in the grains:

Property	Pyrite (cubic)	Marcasite (orthorhombic)
Electronic type	Semiconductor (Eg - 0.95 eV)	Semiumetal (gap = 0 eV)
s (S/cm)	10-100	~104
Impact in film	PV absorber	Direct short-circuit

A 0.1% vol. fraction of marcasite reduces Rsh from 500 - 100 ohm*cm^2 XRD detection limit: ~2-3% vol. Confocal micro-Raman: ~0.05% vol. - XRD + Raman dual validation MANDATORY (Renishaw inVia 532nm, 100- obj.)

References [24-28]

[24] Kim et al. (2016). RSC Advances, 6:46600. doi:10.1039/C6RA07139A
[25] Read & Shockley (1950). Phys. Rev., 78:275.
[26] Cab-n-Acevedo et al. (2012). Nano Lett., 12:1977. doi:10.1021/nl2045364
[27] Kirkeminde et al. (2012). ACS Appl. Mater. Interfaces, 4:1174. doi:10.1021/am201640e
[28] Internal Pyrite Film Raman & Rsh measurements. Internal report PF-2026-06.

SIGNIFICANT

Non-Ohmic Contacts & Interface Recombination

Electron affinity uncertainty (-0.5 eV - enormous for junction design):

chi(FeS2) = 4.6 eV (Bronold 1993, surface cleaved in situ) chi(FeS2) = 4.87 eV (Ennaoui 1993, photoelectrochemistry) chi(FeS2) = 5.1 eV (Kolb 2022, UPS on passivated CVD films) Our UPS measurements (INRS-EMT, He I 21.22 eV): - = 4.88 - 0.08 eV [batches #4-8, 2026]

Buffer layer	chi (eV)	DeltaEc (eV)	Type	Voc impact
In2S3	4.7	+0.20	Ideal spike	Ideal
CdS (CBD)	4,5	+0,40	Moderate spike	Favorable
ZnO	4,35	+0,55	Spike	Acceptable
TiO2 (ALD)	4,1	+0,80	Strong spike	Limits Jsc

Metal contacts - parasitic Schottky barriers:

f_SB(Ag/FeS2) = 0,30-0,45 eV - rho_c ~10- ohm*cm^2 f_SB(Au/FeS2) = 0,20-0,35 eV - rho_c ~10- ohm*cm^2 f_SB(Al/FeS2) = 0,45-0,60 eV - rho_c ~10- ohm*cm^2 (poor) Contact CoS2/FeS2 = ~0 eV - rho_c < 104 ohm*cm^2 (ohmique) - CIBLE CoS2 contact formation: Co-evaporation 5nm Co + 400-C/15min anneal under S-rich N2: Co(s) + FeS2 - CoS2(metallic) + Fe CoS2: s - 105 S/cm, cubic Pa-3-, lattice mismatch only 2.2%

References [29-34]

[29] Kolb et al. (2022). Prog. Photovolt. 30:388. doi:10.1002/pip.3492
[30] Voigt et al. (2023). Prog. Photovolt. 31:812. doi:10.1002/pip.3667
[31] Anderson (1960). Solid State Electron. 5:341. doi:10.1016/0038-1101(62)90115-4
[32] Goh et al. (2006). Geochim. Cosmochim. Acta, 70:2210. doi:10.1016/j.gca.2006.02.007
[33] Karthe et al. (1993). Appl. Surf. Sci. 72:157. doi:10.1016/0169-4332(93)90364-8
[34] Mesures UPS internes Pyrite Film (2026). Rapport interne PF-2026-07.

Summary - Voc Loss Budget

Loss source	Voc (mV)	Mechanism	Pyrite Film Solution	Phase
Theoretical SQ Voc	710 mV	-	-	-
Tunneling effect (ND >10-8)	-250	Short-circuited junction	S-rich anneal + CoFe doping - ND - 3-10-7	Phase 2
EF pinning (Dit >10--)	-180	Broken C4v symmetry	ALD 20 nm + ethanethiol - Dit < 2-10--	Phase 1-2
Bulk SRH (deep Vs)	-120	Vs defects at EV+0.31 eV	Continuous in-situ S-rich anneal	Phase 2
Cliff alignment	-80	DeltaEc < 0 - recombination	Systematic UPS + In2S3 buffer (DeltaEc +0.20)	Phase 2
SRH interface	-60	Dit interface buffer/FeS2	ALD Al2O3 interlayer between layers	Phase 2
Contact Rs (Schottky)	-30	f_SB Ag/FeS2 = 0,35-0,45 eV	Ohmic CoS2 contact (rho_c < 104)	Phase 2-3
Raw CVD Voc (no treatment)	~150 mV	All mechanisms active	-	-
Current Voc (ALD + S-rich)	285-310 mV	Partial improvements	Batches #4-6 measured Jul. 2026	Phase 1
Phase 2 Voc target	320-420 mV	ALD + ligands + S-rich + CoFe	Batches #7-9 in progress	Phase 2
Phase 3 Voc target	480-560 mV	Homojunction P + contacts CoS2	DFT model validated, implementation 2027	Phase 3

PHYSICS

Optical Properties & Bandgap

Direct Eg - 0.95 eV - near-total absorption in 300 nm of film - the best absorption coefficient in the PV family

Tauc Plot - (alpha*hv)^2 vs hv - Eg = 0.95 eV

Absorption coefficient a vs wavelength (log) - FeS2 vs competitors

Decisive advantage: With a > 6-105 cm^-3, FeS2 absorbs 95% of incident light within 40 nm for h > 1 eV. A 300 nm film is therefore 15- optically oversized - offering considerable margin to reduce thickness, lower deposition costs, and improve carrier collection (L/d ratio).

ARCHITECTURE

Device Architecture

Full stack - band alignment - simulated and measured J-V curves

Optimized device stack (Phase 2)

Front contact - Ag grid / ITO (200 nm)

Window - ZnO:Al (200 nm) - Eg = 3.4 eV

Buffer - In2S3 (30-50 nm) - DeltaEc +0.20 eV

ALD Al2O3 (20 nm) - interface passivation

FeS2 absorber (300-500 nm) - CVD + S-rich

Contact layer - CoS2 (5 nm) - ohmic

Substrate - Glass / ITO / Stainless steel

Simulated J-V curves - Phase 1, 2, 3

Simulated EQE - External quantum efficiency vs - (nm)

DeltaEc spike/cliff by buffer - In2S3 optimal

Buffer	chi (eV)	DeltaEc (eV)	Type	Voc impact	Recommendation
In2S3 (ALD)	4,7	+0,20	Weak spike	Ideal	Phase 2 recommended
CdS (CBD)	4,5	+0,40	Moderate spike	Favorable	Non-toxic alternative
ZnS (ALD)	4,38	+0,52	Spike	Acceptable	Option to explore
ZnO (sputtering)	4,35	+0,55	Spike	Acceptable	Window layer only
TiO2 (ALD)	4,1	+0,80	Strong spike	Limits Jsc	Not recommended

SCALE-UP

CVD Reactor & Path to Industrialization

From lab cell (1 cm-) to industrial module (1 m-) - a technically straightforward trajectory

Scale-Up Argument: CVD is Already Industrial

CVD deposition has been used industrially for over 50 years: microelectronics (SiO2, Si3N4, W), hard coatings (TiN, CrN), glazing (SnO2:F via atmospheric CVD - exactly the BIPV target substrate). Reactor technology is mature, scalable, and wellumastered.

Our Fe(acac)3 + H2S process at 400-550 -C is compatible with roll-to-roll reactors on steel or flexible glass - the same type of reactor as CdTe production lines (First Solar) or SnO2:F (Pilkington). The transposition is not a technological leap but rather well-understood scale-up engineering.

Deposition Zone

Temperature 400-550 -C, PID regulation -1 -C. Fe(acac)3 sublimed at 140-160 -C in a dedicated heated evaporator. H2S flow: 10-100 sccm. Total pressure: 10-100 mbar (LPCVD). Deposition time: 60-120 min for 300-800 nm.

In-Situ S-Rich Anneal

Controlled cooling under H2S/N2 (50 sccm): 480 -C - 300 -C at 5 -C/min. Average ND: (9.2-1.8)-10-7 - (4.8-0.9)-10-7 cm^-3 (factor ~2 reduction). Batches #1-6 confirmed by Hall measurements. Patentable strategy.

Damk-hler Parameter

Da - 1 (mixed reaction/diffusion regime) - optimal for composition uniformity. Da - 1 (diffusion-limited) - stoichiometry gradient. Da - 1 (reaction-limited) - poor conformality. Our Da tuning via H2S/Fe(acac)3 ratio and total flow is the key to batch-to-batch reproducibility.

Scale-Up Roadmap

Phase 1: 10-10 cm- reactor (ongoing). Phase 2: 30-30 cm- reactor - BIPV prototype module. Phase 3: OEM partnership with glazing manufacturer (roll-to-roll or in-line CVD on float glass line). Compatibility with low-emissivity (Low-E) glazing CVD lines is a major strategic advantage.

Batch-to-batch reproducibility - measured Vocs (batches #1 to #9)

Scale-up projection - Reactor area vs estimated $/Wp cost

EXPERIMENTAL DATA

Performance & Characterization

Measured results - batches #1 to #9 (2025-2026) - INRS-EMT, Varennes QC

TRPL - t_eff (ns) & L_eff (nm) by batch

Rsh (O-cm-) vs Grain size (nm) - scatter

Mobility - (cm^2/Vs) - state comparison

Collection efficiency rho_coll (%) vs L/d

Technique	Measured quantity	Current values (batches #4-6)	Phase 2 Target	Phase 3 Target
Hall effect (Van der Pauw)	-, ND, type	p-type; ND = (4,8-0,9)-10-7; - = 0,5-2 cm^2/Vs	- > 2 cm^2/Vs	- > 5 cm^2/Vs
4-point probe	R (O/)	10--10-	< 5-10-	< 10-
s(T) 100-300K	Transport mechanism	Mott VRH at <250K; Bands at 300K	Bands from 200K	Degenerate metal avoided
DLTS	Defect Ea	0,12 eV (Vs); 0,22 eV (Fei)	Reduce [Fei] via ALD	Levels < 0.1 eV only
TRPL (streak camera)	t_eff (ns), L_eff (nm)	t = 3,2-0,4 ns - L - 85 nm	t > 5 ns; L > 150 nm	t > 20 ns; L > 300 nm
UPS (He I 21,22 eV)	, F (eV)	= 4,88 - 0,08 eV	Systematic measurement	Systematic measurement
EIS (1 mHz-1 MHz)	Rsh, Rs, C	Rsh > 350-550 ohm*cm^2	Rsh > 1000 ohm*cm^2	Rsh > 5000 ohm*cm^2
J-V AM1.5G (100 mW/cm-)	Voc, Jsc, FF, PCE	Voc 285-310 mV; PCE 3,2-4,8%	Voc = 320 mV; PCE = 5%	Voc = 480 mV; PCE = 8%

Note on n/p ambiguity: FeS2 single crystals are typically n-type; polycrystalline CVD films are p-type. The inversion is attributed to the high density of sulfur vacancies (Vs) generating net acceptor states at the surface. Hall determination (Van der Pauw) is essential per batch - conduction type can vary with CVD conditions.

INFRASTRUCTURE

Characterisation Equipment

Multi-technique instrumentation platform - INRS-EMT Varennes QC & partners

XRD - X-Ray Diffraction (Rigaku)

Pyrite / marcasite / troilite phase identification, lattice parameter extraction, Scherrer grain size, and crystallographic texture. Secondary-phase detection limit ~2-3 vol.%. Mandatory for every batch.

Confocal Micro-Raman (Renishaw inVia)

532 nm, 100- objective. Marcasite detection from 0.05% vol. - 50- more sensitive than XRD. 2D spatial phase mapping. Pyrite Raman modes: 341, 350, 377 cm^-3. Systematic XRD+Raman dual validation.

XPS / UPS (He I/II source)

Fe2+/Fe2+ and S-/S2^2- oxidation states. chi(FeS2) and F determination by UPS (He I 21.22 eV). Depth composition profile (Ar sputtering). ALD passivation verification. Measured - = 4.88 - 0.08 eV.

TRPL (Hamamatsu C10910)

Streak camera, lambda_exc = 400 nm (50 fs, Ti:Sapphire), detection at 950 nm. Temporal resolution <10 ps. Biexponential fit: t1 (surface, 0.3-0.5 ns) + t2 (bulk, 2.5-4.2 ns). t_eff and L_eff per batch.

Hall Effect (Lake Shore 8400)

Type (n/p), ND/NA and - in Van der Pauw configuration. Permanent B field = 0.55 T. Range 10-5-10-- cm^-3. Critical per-batch measurement - n/p ambiguity attributed to [Vs]. Measurements confirmed 100-300K (cryostat).

EIS (Biologic SP-300)

Rsh, Rs, junction capacitance, Dit by C-V analysis. Range 1 mHz-1 MHz, amplitude 10 mV. Randles + Warburg circuit. Real-time tracking of ALD passivation during deposition (in-situ capability planned).

AM1.5G Solar Simulator

Class AAA (Newport Oriel), 100 mW/cm- - 1%. Full J-V measurements: Voc, Jsc, FF, PCE. Light and dark curves. IPCE/EQE by monochromator (300-1100 nm). Si reference cell calibration.

TEM / SEM (FEI Titan 80-300)

High-resolution imaging of FeS2/ALD interfaces, thickness measurement. EDX for elemental mapping (Fe, S, Al, O, Co). SAED for crystal phase confirmation. FIB cross-sections for interface analysis.

OPPORTUNITY

Market & Investment Thesis

An exploding market, a progressively solved bottleneck, a material with no regulatory competitor

The Investment Thesis in 4 Points

Regulatory timing: the EU is tightening Cd rules (2026) and reviewing Pb in perovskites. CdTe and perovskites face rising regulatory risk. FeS2 is intrinsically compliant.
Captured BIPV market: $83.3B by 2030 (GlobalData), CAGR 22%. Transparent or semi-transparent PV glazing is a segment where Si and CdTe remain visually unattractive. FeS2 offers a unique gold/bronze appearance.
Identified bottleneck, not a fundamental limit: 35 years of failures can now be explained by six physico-chemical mechanisms, not by thermodynamic ceilings. Our roadmap addresses each mechanism with targeted engineering actions.
Unmatched material cost: <$0.01/W - no other PV technology comes close. At CdTe-like PCE levels, FeS2 could deliver ultra-low LCOE over lifetime.

Global BIPV market - Projection 2022-2032 (USD B)

Comparative LCOE (\$/kWh) - PV technologies at equal maturity

Thin Film PV market share 2026 vs 2032 target

PCE roadmap - Pyrite Film trajectory 2025-2030

Phase 1 - TRL 4-5 - 2025-2026

PCE 3-5% - Voc = 320 mV
ALD Al2O3 20 nm + S-rich anneal
Batches #1-9 - Mechanism validation
Budget: ~\
Ongoing

Phase 2 - TRL 5-6 - 2026-2027

PCE 5-8% - Voc = 420 mV
CoFe doping + In2S3 buffer
Sulfur ligands - 10-10 cm- module
Budget: ~\
In preparation

Phase 3 - TRL 6-7 - 2027-2029

PCE 8-13% - Voc = 520 mV
P-type homojunction - CoS2 contact
30-30 cm- module - OEM partner
Budget: ~\
Industrialization target

ANALYSIS

Diagnostic: Why 35 Years of Failure

And why we have good reason to believe in it now

The FeS2 Paradox - and Its Resolution

Pyrite FeS2 has been studied since Tributsch & Ennaoui (1983-1993). Despite remarkable intrinsic properties, no group in the world has exceeded 2.8% PCE in 35 years. Why And why might that change now

Historical reason for failure: the community long treated FeS2 as a phase purity problem (pyrite vs marcasite). Once pure pyrite phase was obtained, performance remained catastrophic - with no consensus explanation for 20 years. It was only in the 2010-2024 decade that the 6 interface mechanisms were identified and quantified: tunneling, EF pinning, Vs vacancies, diffusion length, shunting, contacts.

Why now: (1) ALD is a mature industrial technique available everywhere since 2010. (2) High-precision DFT models of FeS2 surface defects have been published since 2012 (Hu et al., PhysRevB). (3) Precision TRPL and EIS enable quantification of t_eff and Dit per batch. (4) BIPV demand creates sufficient economic pressure to justify investment. The missing building blocks exist. It remains to assemble them correctly.

Historical evolution of FeS2 record PCE vs competitors (1985-2026)

FeS2 PV scientific publications per decade - field growth

DURABILITY

Stability & Durability

WVTR barrier, IEC 61215 protocols, vacancy diffusivity - risk analysis and solutions

Context: FeS2 stability is a legitimate point of vigilance - not yet demonstrated over 25 years, unlike Si or CdTe. However, Fe and S are intrinsically stable elements (pyrite has existed in nature for millions of years). The real risk is surface oxidation (FeS2 + O2 - FeO, SO4-) - controlled by the ALD Al2O3 barrier (WVTR < 10- gum--day-). Our objective is to demonstrate IEC 61215 compliance (Damp Heat 85-C/85% RH, 1000h) in Phase 2.

DV diffusivity of Vs vacancies vs T (degC) - Arrhenius

Thermal stability - Sulfur ligand coverage (%)

WVTR Barrier - ALD Al2O3

20 nm ALD - WVTR < 10- gum--day- (comparable to OLED encapsulation barriers). Protection against surface oxidation FeS2 + O2 - FeO, SO4-. ALD multilayer stack (Al2O3 / ZrO2) planned for Phase 2 - WVTR < 105.

Kinetic Immobility of Vs at 300K

DV(300K) = 9-10-- cm^2/s - 0. Vacancies created at 500 -C are kinetically frozen upon quenching - they do not diffuse under normal operating conditions. The film is thermodynamically metastable but kinetically stable at <100 -C.

Sulfur Ligands - Thermal Stability

Ethanethiol: Kstab ~ 106-108 M-, stable up to ~200 -C under N2. Typical BIPV module temperature: 45-85 -C under normal conditions. Safety margin = 115 -C. Planned: more robust ligands (disulfides, sulfur phosphines) for Phase 3.

IEC 61215 Protocols - Phase 2 Objectives

Damp Heat (85-C/85%RH, 1000h), Thermal Cycling (-4085-C, 200 cycles), UV preconditioning (15 kWh/m-), Humidity Freeze. Test plan scheduled with CETECOM or T-V Rheinland. Objective: stability demonstration before industrialization phase.

EVIDENCE

Evidence Framework and Confidence

Pyrite-film claims span validated material properties and forward-looking device-scale projections; confidence should be read by category.

High confidence

Intrinsic FeS2 abundance, non-toxicity profile, and optical absorption fundamentals established in literature.

Medium confidence

Process-to-performance improvements under controlled CVD and passivation parameter windows.

Medium-Low confidence

Industrial-scale yield and long-term field durability until full IEC validation at production conditions.

BIBLIOGRAPHY

Scientific References

Base of 39 sources - published between 1950 and 2024 - grouped by topic

Group 1 - CVD Synthesis & Fundamental Properties

[1] Seefeld, S. et al. (2013). Iron Pyrite Thin Films Synthesized from an Fe(acac)3 Precursor. JACS, 135(11), 4412-4424. doi:10.1021/ja311974n

[2] Hu, J. et al. (2012). FeS2 Thin Films by CVD. Sol. Energy Mater. Sol. Cells, 96, 204-209.

[3] Puthussery, J. et al. (2011). Colloidal FeS2 Nanocrystal Inks. JACS, 133(4), 716-719. doi:10.1021/ja1096368

[4] Cab-n-Acevedo, M. et al. (2012). Semiconducting FeS2 Nanowires. Nano Lett., 12(4), 1977-1982. doi:10.1021/nl2045364

[5] Birkholz, M. et al. (1992). Sulfur deficiency in iron pyrite. Thin Solid Films, 213(2), 184-190. doi:10.1016/0040-6090(92)90263-B

[6] Pierson, H. O. (1999). Handbook of Chemical Vapor Deposition (2nd ed.). Noyes Publications.

[7] Habas, S. E. et al. (2010). Low-cost inorganic solar cells: from ink to printed device. Chem. Rev., 110(11), 6571-6594.

Group 2 - Defects, Transport & Electronic Structure

[8] Hu, J., Zhang, Y., Law, M. & Wu, R. (2012). Native defects in bulk iron pyrite. Phys. Rev. B, 85(8), 085203. doi:10.1103/PhysRevB.85.085203

[9] Willeke, G. et al. (1992). Electrical transport of FeS2 single crystals. J. Alloys Compd., 178, 181-191. doi:10.1016/0925-8388(92)90261-Y

[10] Limpinsel, M. et al. (2014). Electrical rectification at metal-insulator-pyrite interface. Energy Environ. Sci., 7, 1974-1989. doi:10.1039/C3EE43169H

[11] B-ker, K. et al. (1992). Photovoltaic output limitation of n-FeS2 Schottky barriers. J. Appl. Phys., 72(12), 5721. doi:10.1063/1.351974

[12] Lehner, S. W. et al. (2006). Pyrite doped with Co, Ni, As. J. Cryst. Growth, 286(2), 306-317. doi:10.1016/j.jcrysgro.2005.10.024

[13] Bronold, M. et al. (1993). Interface preparation of FeS2 for PEC cells. Phys. Status Solidi A, 135(2), 553. doi:10.1002/pssa.2211350231

[14] Voigt, C. et al. (2023). Ohmic contacts to n-FeS2 via in situ CoS2 formation. Prog. Photovolt., 31(8), 812. doi:10.1002/pip.3667

[15] Voigt, C. et al. (2024). Direct observation of internal p-n junction in FeS2 by KPFM and DLTS. ChemRxiv. doi:10.26434/chemrxiv

Group 3 - Characterisation XRD / XPS / Raman / EIS

[16] Kim, E. et al. (2016). Marcasite in FeS2 by micro-Raman. RSC Adv., 6(52), 46600-46606. doi:10.1039/C6RA07139A

[17] Goh, S. W. et al. (2006). Oxidation states of Fe in mineral sulfides. Geochim. Cosmochim. Acta, 70(9), 2210. doi:10.1016/j.gca.2006.02.007

[18] Karthe, S. et al. (1993). Oxidation of pyrite by XPS. Appl. Surf. Sci., 72(2), 157-170. doi:10.1016/0169-4332(93)90364-8

[19] Nesbitt, H. W. & Bancroft, G. M. (1997). Spectroscopic characterization of pyrite dissolution. Am. Mineral., 82, 1026. doi:10.2138/am-1997-9-1022

[20] Bisquert, J. (2002). Impedance of electron diffusion in thin layers. J. Phys. Chem. B, 106(2), 325-333. doi:10.1021/jp011941g

Group 4 - PV Performance & Simulation

[21] Shockley, W. & Queisser, H. J. (1961). Detailed balance limit. J. Appl. Phys., 32(3), 510. doi:10.1063/1.1736034

[22] R-hle, S. (2016). Tabulated Shockley-Queisser limit values. Solar Energy, 130, 139-147. doi:10.1016/j.solener.2016.02.015

[23] Ennaoui, A. et al. (1993). Iron disulfide for solar energy conversion. Sol. Energy Mater. Sol. Cells, 29(4), 289. doi:10.1016/0927-0248(93)90095-P

[24] Kolb, J., Birkholz, M. & Fiechter, S. (2022). Photovoltage loss in pyrite. Prog. Photovolt., 30(4), 388. doi:10.1002/pip.3492

[25] Voigt, C. et al. (2023). CoS2 ohmic contacts to FeS2. Prog. Photovolt., 31(8), 812. doi:10.1002/pip.3667

[26] Anderson, R. L. (1960). Ge-GaAs heterojunctions. Solid State Electron., 5(5), 341. doi:10.1016/0038-1101(62)90115-4

Group 5 - ALD Passivation & Stability

[27] Werner, F. et al. (2011). Low surface recombination by ALD-Al2O3. Appl. Phys. Lett., 97(16), 162103. doi:10.1063/1.3491216

[28] Naumann, V. et al. (2012). ALD-Al2O3 on crystalline silicon. Energy Procedia, 27, 312-318. doi:10.1016/j.egypro.2012.07.073

[29] Fahrenbruch, A. L. & Bube, R. H. (1983). Fundamentals of Solar Cells. Academic Press.

[30] Read, W. T. & Shockley, W. (1950). Dislocation models of grain boundaries. Phys. Rev., 78(3), 275. doi:10.1103/PhysRev.78.275

[31-39] Internal Pyrite Film measurements (2025-2026). TRPL, UPS, Hall, Raman, EIS, J-V. INRS-EMT, Varennes QC. [Internal reports PF-2025-01 to PF-2026-07 - unpublished data]

FeS2 CVD Filmsfor Next-Generation Solar Glazing

Why Pyrite FeS2

The Central Argument: A Unique Window of Opportunity

Abundance & Geopolitical Independence

Native Regulatory Compliance

Circular Economy

Optimal BIPV Integration

CVD: A Proven Industrial Process

Ultra-Thin Films = Material Savings

Pyrite FeS2 CVD Technology

FeS2 Crystal Structure

CVD Deposition - Key Parameters

ALD Al2O3 Passivation

Sulfur Ligands - Pyrite Film Innovation

Manufacturing Process

The Fundamental Challenge: Voc Deficit

The Failure Cascade - 6 Identified Mechanisms Under Direct Attack

Summary - Voc Loss Budget

Optical Properties & Bandgap

Device Architecture

CVD Reactor & Path to Industrialization

Scale-Up Argument: CVD is Already Industrial

Deposition Zone

In-Situ S-Rich Anneal

Damk-hler Parameter

Scale-Up Roadmap

Performance & Characterization

Characterisation Equipment

XRD - X-Ray Diffraction (Rigaku)

Confocal Micro-Raman (Renishaw inVia)

XPS / UPS (He I/II source)

TRPL (Hamamatsu C10910)

Hall Effect (Lake Shore 8400)

EIS (Biologic SP-300)

AM1.5G Solar Simulator

TEM / SEM (FEI Titan 80-300)

Market & Investment Thesis

The Investment Thesis in 4 Points

Diagnostic: Why 35 Years of Failure

The FeS2 Paradox - and Its Resolution

Stability & Durability

WVTR Barrier - ALD Al2O3

Kinetic Immobility of Vs at 300K

Sulfur Ligands - Thermal Stability

IEC 61215 Protocols - Phase 2 Objectives

Evidence Framework and Confidence

High confidence

Medium confidence

Medium-Low confidence

Scientific References

Group 1 - CVD Synthesis & Fundamental Properties

Group 2 - Defects, Transport & Electronic Structure

Group 3 - Characterisation XRD / XPS / Raman / EIS

Group 4 - PV Performance & Simulation

Group 5 - ALD Passivation & Stability

FeS2 CVD Films
for Next-Generation Solar Glazing