Skip to content

Transfer of Freestanding Fluoropolymer Films for Advanced Semiconductor Devices

FieldValue
JournalACS Applied Materials & Interfaces
Published2026-08-27
AuthorsMohammad Monish, Koki Hino, Yosuke Sasama, Masato Urakami, Takehiro Ota
DOI10.1021/acsami.6c05045
Citations1

Freestanding CYTOP films formed a low-κ gate dielectric on hydrogen-terminated diamond. Their breakdown field was 8.0 ± 1.2 MV cm⁻¹. The resulting p-channel FET reached an effective hole mobility of approximately 400 cm² V⁻¹ s⁻¹, negligible hysteresis, and an interface trap density no greater than 3 × 10¹¹ cm⁻² eV⁻¹. These results show that preformed fluoropolymer can insulate a low-adhesion diamond surface without direct dielectric deposition.

Film formation and target lamination were separate steps. The authors prepared 3 wt % type-S CYTOP in CT-Solv.180 and stirred it for 2 h. They spin-coated 3 wt % PAA on silicon at 3000 rpm for 60 s and baked it at 155 °C for 2 min. CYTOP was coated at the same spin conditions. The anneal used 10 min at 50 °C, 30 min at 80 °C, and 10 min at 130 °C. This produced an approximately 90 nm film.

Kapton tape supported the film while PAA dissolved in deionized water at 50 °C. The film was mounted on a copper frame, baked at 100 °C for 5 min, and stored in a vacuum-sealed container. Optical alignment and three-axis positioning placed the pristine face against a target on a 110 °C stage. It adhered immediately on contact. Parallel alignment before contact was used to prevent misalignment or wrinkles. Minor wrinkles were removed by annealing at 130 °C. The authors attributed this recovery to residual-stress relaxation. Transfers were shown on silicon, CYTOP on silicon, sapphire, and hydrogen-terminated diamond.

AFM scans of the spin-coated CYTOP surface covered 20 µm × 20 µm at four locations on each of four samples. The resulting rms roughness was 0.45 ± 0.03 nm. Films spin-coated directly on silicon had nearly the same morphology, so PAA did not change the surface. Separate solutions with identical spin parameters produced thickness variation of ±5 nm. Measurements across one sample varied by only ±0.6 nm.

The MIM test films covered 25 to 221 nm by changing solution concentration and spin speed. This control provided dielectric layers thin enough for device measurements while retaining the smooth, featureless surface required for consistent electrical behavior.

Breakdown strength and dielectric constant

Section titled “Breakdown strength and dielectric constant”

Metal-insulator-metal structures used sapphire with 5 nm Ti/5 nm Pt bottom electrodes and 50 nm Al/100 nm Au top electrodes. Current density came from the electrode overlap area. Electric field was voltage divided by CYTOP thickness.

Before breakdown, leakage was typically below 10⁻⁷ A cm⁻². It stayed within the 10⁻⁶ A cm⁻² range in every device. Breakdown fields ranged from 5.9 to 10.4 MV cm⁻¹.

At a leakage current density of 10⁻⁵ A cm⁻², 13 MIM devices gave 8.0 ± 1.2 MV cm⁻¹. The 221 nm device was excluded from that average because it remained stable without breakdown up to the maximum applied field of about 9 MV cm⁻¹.

The mean breakdown field was comparable to directly spin-coated thick CYTOP. It was also among the highest values reported for transferred dielectrics. This result shows that lamination preserved the insulating behavior of the fluoropolymer.

Hydrogen-terminated diamond FET fabrication

Section titled “Hydrogen-terminated diamond FET fabrication”

The FET channel used polished standard grade IIa-type (100) undoped single-crystal CVD diamond. Nitrogen implantation isolated the active regions, and Ti/Pt contacts were 5 nm/5 nm. The diamond then received an H₂ anneal at 650 °C for 35 min. Hydrogen plasma followed at a stage temperature of 570 °C for 15 min. The plasma used 500 sccm H₂ at 27 Torr. The measured diamond surface reached about 700 to 750 °C. These steps formed TiC ohmic contacts and hydrogen-terminated the surface.

To avoid surface transfer doping, the diamond moved through a vacuum suitcase into an Ar glove box. The suitcase maintained pressure below 1 × 10⁻⁷ Torr. Oxygen and water were below 0.5 and 2 ppm, respectively. The CYTOP was baked inside the box at 80, 100, and 120 °C for 3 min each before lamination. Transfer to the glove box and lamination occurred within 5 h after leaving the MPCVD system.

A 90 nm CYTOP film served as the gate dielectric. A 50 nm Al/100 nm Au stack served as the gate electrode. This sequence kept the hydrogen-terminated surface away from ambient air and avoided applying the CYTOP solution directly to the diamond.

Transistor performance and interface quality

Section titled “Transistor performance and interface quality”

The representative Hall bar had a 117.5 µm gate length, an 8.4 µm width, and 29.8 µm between voltage probes. Room-temperature output and transfer data used a two-probe configuration.

MetricConditionResult
Normalized drain currentVDS = VGS = -30 V-23 mA mm⁻¹
On/off ratioVDS = -30 V5.7 × 10⁶
Normalized hysteresisGate range of 0.5 to -1 VBelow 5 mV (MV cm⁻¹)⁻¹
Subthreshold swingTransfer characteristic220 mV dec⁻¹
Interface trap densityTransfer characteristic≤3 × 10¹¹ cm⁻² eV⁻¹
Effective hole mobilityMaximum≈400 cm² V⁻¹ s⁻¹
Hole densityVGS = -30 V3.2 × 10¹² cm⁻²
Threshold voltageSheet-conductance fit from -5 to -15 V-2.3 V
Gate leakage3 MV cm⁻¹, source and drain shortedBelow 5 × 10⁻⁷ A cm⁻²

Four-probe sheet conductance peaked at 2.1 × 10⁻⁴ Ω⁻¹, corresponding to 4.8 kΩ. The hysteresis was comparable to the lowest values reported for 2D semiconductors with transferred gate insulators. The trap density was comparable to or below most H-terminated diamond FETs. Mobility fell within the upper range reported for diamond FETs using several gate insulators.

Capacitance-voltage measurements used a 50 mV AC signal from 10 kHz to 1 MHz on another three-terminal device. At 1 MHz, the ON-state capacitance was 0.126 µF cm⁻² and the OFF-state value was 0.106 µF cm⁻². Subtracting these values gave 0.020 µF cm⁻². For the 90 nm film, this corresponded to a relative permittivity of 2.0. The dielectric constant stayed consistent across frequencies, and forward and reverse C-V curves nearly overlapped. Long-term storage produced nearly identical output and transfer curves.

Limits and device engineering implications

Section titled “Limits and device engineering implications”

Several limits remain visible in the data. The main FET results center on one representative Hall-bar device, while a second device appears only in the Supporting Information. The main text does not provide a device population or the storage duration. The 221 nm MIM device also had no observed breakdown within the tested field, so it was excluded from the 13-device average.

The authors report slight surface conductivity after CYTOP lamination on hydrogen-terminated diamond. They state that the origin and energetic distribution are unclear. Local polymer-chain strain, residual solvent, or adsorbates are presented as possible sources of electron-trapping sites, not as established causes. They call for further experimental and theoretical work.

Type-M and type-A CYTOP were not tested. The discussion predicts stronger charge transfer for their higher electron affinities. It also proposes layered CYTOP as a future modulation-doping route, without reporting device measurements for that structure.

For diamond device engineering, the measured result supports (100)-oriented, hydrogen-terminated single-crystal diamond when a gate dielectric must avoid direct deposition. The authors argue that the measured relative permittivity of 2.0 and high breakdown field suit low-capacitance gate and interconnect insulation. Flexible, optoelectronic, and quantum-device uses remain proposals rather than demonstrated functions.

Thermal integration also matters. The reported flow includes a 110 °C lamination stage, a 120 °C CYTOP bake, and a 130 °C wrinkle anneal. The diamond had already received a much higher-temperature hydrogen-plasma treatment.

The demonstrated device needs a 2.2 mm × 2.2 mm × 0.5 mm (100)-oriented, hydrogen-terminated single-crystal diamond substrate with Ti/Pt contacts and a 90 nm CYTOP gate dielectric. 6CCVD’s diamond materials catalog lists MPCVD-grown single-crystal diamond, substrates up to 10 mm, and metallization with Ti or Pt. Confirm orientation, hydrogen termination, polish, and metal thickness against the paper through engineering support.

View Original Abstract

High-quality dielectric films are essential for fabricating advanced electronic devices, but their direct deposition often degrades the films and their underlying interfaces, which compromises device performance, especially on sensitive or low-adhesion surfaces. To overcome these limitations, film transfer methods enable the integration of high-quality dielectric films onto such surfaces without damaging the underlying interfaces. However, existing transfer methods have predominantly focused on high-dielectric-constant (high-κ) materials, leaving a critical gap for transferable, high-quality low-κ alternatives, which are required for enabling low-power and high-speed electronics. Herein, we address this need by demonstrating a method to integrate freestanding low-κ fluoropolymer dielectric films with smooth surface morphology onto diverse substrates, including low-adhesion surfaces like hydrogen-terminated diamond. The transferred films revealed high breakdown fields of 8.0 ± 1.2 MV cm-1, with leakage current density remaining typically below 10-7 A cm-2 before the breakdown. The incorporation of these fluoropolymer films as gate dielectrics in p-channel hydrogen-terminated diamond field-effect transistors resulted in transfer and output characteristics with negligible hysteresis, high channel mobility (≈400 cm2 V-1 s-1), and a low interface trap density (≤3 × 1011 cm-2 eV-1). These findings highlight the versatility of the transfer method and position freestanding fluoropolymers as a promising platform for forming high-quality dielectric/semiconductor interfaces for advanced electronics.