Barium Titanate on 300 mm silicon

Proven performance for next-generation photonics.

Production oxide MBE from DCA Instruments makes BTO-on-silicon photonics manufacturable – FOUP-to-FOUP automation, consistent results wafer after wafer, and high-quality epitaxial BTO on a proven, reliable platform.

  • 1.6 Tb/snet (4 × 448 Gb/s PAM4) on BTO grown on DCA MBE – OFC 2026 [3]
  • >100BTO wafers with 3σ lattice-parameter variation ≈ 0.1 %
  • 24/7FOUP-to-FOUP batch processing with SECS/GEM factory connection
  • 25,000+transfer cycles validated on the automated UHV robot

Why BTO? A strong Pockels material made for silicon

  • Fully CMOS compatible

    Epitaxial BTO (BaTiO3) grows on silicon via an MBE-grown SrTiO3 (STO) seed. The integration steps are commonly available in CMOS back-end-of-line processing, and BTO has been monolithically integrated in a 300 mm SiPh foundry process with 95 % functional yield. [1, 2]

  • Smaller, denser modulators

    BTO’s Pockels coefficient (~923 pm/V) is one of the largest on silicon: 1.75 mm phase shifters reach 105 GHz, four modulators fit on a 5.1 × 3.5 mm² die, and 350 µm phase shifters give a full π shift in a 116-actuator mesh. [3, 5]

  • Low power, no heat

    BTO is field-driven, with no carrier injection. DC bias tuning stays below 5 mW, against 13 mW per π for a thermal phase shifter. Static power is 560 nW per π, with no thermal crosstalk. [3, 4, 5]

Uniform BTO film grown on a 300 mm silicon wafer
Uniform BTO grown on a 300 mm silicon wafer.

Read why BTO is emerging for integrated photonic platforms

BTO beyond datacom

  • Non-volatile photonic memoryFerroelectric domains hold an optical state with no holding power; multilevel non-volatile phase shifters (16 levels shown). [6, 5]
  • Programmable photonics and optical computing116-actuator BTO gate array with ~80 ns switching, for filters, 4×4 linear transforms and routing. [5]
  • Quantum photonicsLow-loss BTO switches for photon routing (0.33 dB·V) [2]; Pockels effect retained at 4 K (~200 pm/V). [7]
  • Robust operationModulation maintained above the bulk Curie point (tested to 130 °C); no degradation after 250 °C exposure; 24 h at 90 °C. [8]

Proven results

At OFC 2026, McGill, Ciena and Lumiphase showed a BTO/silicon-photonics transmitter carrying a net 1.6 Tb/s (4 × 448 Gb/s PAM4 over 2 km) with 105 GHz bandwidth and consistent performance from 20 to 85 °C – built on epitaxial BTO grown on DCA metal-oxide MBE systems. A companion DP-IQ modulator reached a net 1 Tb/s per wavelength. [3, 4]

Four-channel BTO silicon-photonics transmitter chip
Four-channel BTO–SiPh transmitter.
Measured bandwidth of the BTO modulator, 105 GHz
105 GHz modulator bandwidth.
Fully processed 200 mm BTO-on-silicon-photonics wafer
Fully processed 200 mm BTO-on-silicon-photonics wafer.

Productive: delivering output, wafer after wafer

3σ statistics over more than 100 BTO wafers on silicon, grown by solid-source MBE.

Lattice parameter (indicator of Ba/Ti stoichiometry)
≈ 0.1 %3σ variation
Thickness
< 1 %3σ variation, films > 150 nm
Refractive index
< 0.2 %typical 3σ variation
Graph: 3σ variation of BTO lattice parameter on 300 mm wafers
Lattice parameter – 3σ variation.
Graph: 3σ variation of BTO film thickness on 300 mm wafers
Thickness – 3σ variation.
Graph: 3σ variation of BTO refractive index on 300 mm wafers
Refractive index – 3σ variation.
Graph: BTO lattice parameter from centre to edge of a 300 mm wafer
Lattice parameter – across the wafer.
Graph: BTO refractive index from centre to edge of a 300 mm wafer
Refractive index – across the wafer.

Uniform from centre to edgeLattice parameter within ±0.2 % and refractive index within ±0.4 % across the 300 mm wafer. 300 mm wafers double device throughput compared with 200 mm wafers.

Automated: a fab tool, not a lab tool

In 2019 DCA built the world’s first production oxide MBE for 300 mm wafers. Wafers go in and come out in FOUPs under a fully automated 24/7 batch process.

Wafer size8″ (200 mm) or 12″ (300 mm)
LoadingEFEM with dual FOUP ports; wafer ID recognition and flipping
CapacityLoad lock and storage for 10 or 25 wafers
UHV transferAutomated transfer robot (CDC), validated for over 25,000 transfer cycles
ScalabilityOne robot serves multiple growth chambers (e.g. dual M1000 MBE); additional CDCs extend the cluster tool to up to four M1000 growth chambers
Temperature uniformity±2.0 °C at 600 °C for precise, repeatable BTO growth
Process controlSurface preparation and MBE growth with closed-loop feedback; real-time flux and growth control with integrated RHEED and XBS
SoftwareDCA MOSAIC runs every BTO route: wafer ID and tracking, automated flux calibration and closed-loop source control for optimal stoichiometry, parallel layer-based growth recipes, safety interlocks, 24/7 batch processing and SECS/GEM
CleanlinessUltra-low particle, SEMI-FAB compliant. Particle count over a 300 mm Si wafer: 25 after transfer to chambers, 91 after heating to 800 °C with rotation (M-Series data)
DCA M1000 production MBE with automated transfer system and EFEM for FOUP loading
M1000 production MBE with automated transfer and EFEM, compatible with a Tier 1 SEMI-fab environment.

Every BTO route on one flexible, fully automated DCA platform

The enabling step: an MBE-grown SrTiO3 seed

Without an MBE-grown SrTiO3 (STO) seed layer, high-quality epitaxial BTO films on silicon cannot be achieved. The few-atomic-layer seed locks into registry with silicon and templates single-crystal BTO; sputtered and hybrid-MBE BTO also build on it.

Three-step schematic: BTO grown directly on bare silicon forms randomly oriented grains; an MBE-grown SrTiO3 seed a few atomic layers thick locks onto the silicon lattice; single-crystal BaTiO3 then grows on the seed.
Why the MBE-grown seed matters: (1) directly on bare silicon, BTO grows as randomly oriented grains; (2) an SrTiO3 seed only a few atomic layers thick locks onto the silicon lattice; (3) single-crystal BTO grows on top. Schematic, not to scale. Click to enlarge.
RouteDCA sourceStrength
Solid-source MBETi from DCA HTEC high-temperature effusion cell or e-beam evaporatorHighest controllability, excellent material quality
Hybrid MBETi from DCA MOPI metal-organic injectorFaster growth; proven on 200 mm wafers, with Vπ and propagation loss comparable to solid-source MBE
UHV sputteringDCA APEX-L linear magnetron for 8″–12″Simpler processing, higher throughput, on the MBE seed

BTO grown on 200 mm wafers by Hybrid MBE - proven device performance

Box plot per wafer: normalised V-pi of BTO devices, three wafers grown by solid-source MBE and six by hybrid MBE with the MOPI source
Vπ per wafer, solid-source MBE vs MOPI (normalised).
Box plot per wafer: normalised propagation loss of BTO devices grown by solid-source MBE and by hybrid MBE with the MOPI source
Propagation loss per wafer, solid-source MBE vs MOPI (normalised).

Faster growth, comparable devicesThe MOPI source supplies titanium as a metal-organic precursor, for faster growth and higher throughput. Device Vπ and propagation loss are comparable to solid-source MBE.

In production flows today

Three business models, one enabling step: precision epitaxy on production-grade DCA tools.

  • Lumiphase

    Device maker, Switzerland

    Manufactures BTO-enhanced silicon-photonics PICs based on proprietary technology.

    MBE-grown BTO in 200 and 300 mm SiPh flows.

  • PsiQuantum

    Photonic quantum computing, USA

    Manufactures photonic components, including single-photon sources, single-photon detectors and ultra-high-performance optical switches.

    300 mm foundry flow.

  • La Luce Cristallina

    Merchant BTO/STO wafers, USA

    Supplies epitaxial BTO and STO wafer platforms on silicon and SOI, in multiple wafer sizes, orientations and thicknesses.

    DCA MBE for the STO seed, DCA UHV sputtering for BTO.

Dr Sabina Hatch, Sales Director, DCA Instruments

Talk to us about BTO production on 200 / 300 mm

Dr Sabina Hatch, Sales Director – sabina.hatch@dca.fi

References

  1. S. Abel et al., “Large Pockels effect in micro- and nanostructured barium titanate integrated on silicon”, Nature Materials 18, 42–47 (2019). doi:10.1038/s41563-018-0208-0
  2. Alexander et al. (PsiQuantum), “A manufacturable platform for photonic quantum computing”, Nature 641, 876–883 (2025). doi:10.1038/s41586-025-08820-7
  3. B. Qiu et al., “Barium Titanate Enabling Net 1.6T (4×448 Gbps PAM4) On a Silicon Photonics Platform”, OFC 2026, postdeadline paper Th4B.3.
  4. B. Qiu et al., “Barium Titanate DP-IQM Enabling Net 1 Tbps/λ ZR and Coherent-Lite Data Center Networks”, OFC 2026, paper Th3J.4.
  5. C. Catalá-Lahoz et al., “High-speed non-volatile barium titanate field-programmable photonic gate array”, Nature Photonics 20, 932–940 (2026). doi:10.1038/s41566-026-01934-y
  6. J. Geler-Kremer et al., “A ferroelectric multilevel non-volatile photonic phase shifter”, Nature Photonics 16, 491–497 (2022). doi:10.1038/s41566-022-01003-0
  7. F. Eltes et al., “An integrated optical modulator operating at cryogenic temperatures”, Nature Materials 19, 1164–1168 (2020). doi:10.1038/s41563-020-0725-5
  8. A. Messner et al., “Plasmonic ferroelectric modulators”, Journal of Lightwave Technology 37(2), 281–290 (2019). doi:10.1109/JLT.2018.2881332

Further reading: BTO poling dynamics in high-speed communications, Optica.

About DCA

News

Service and Support

Products

Applications