| HS Code | 186948 |
| Df | <0.0003 |
| Dk | 2.5 |
| Water Absorption | <0.1% |
| Glass Transition Temperature | >350°C |
| Curing Temperature | 250°C |
| Thermal Stability | Stable up to 350°C |
| Adhesion | Excellent to silicon, metals, and dielectrics |
| Planarization | High degree of planarization for multilayer stacks |
| Copper Compatibility | Compatible with copper interconnects and RDL |
| Breakdown Voltage | >500 V/µm |
| Outgassing | Ultra-low outgassing |
| Viscosity | Low viscosity for spin-coating and lamination |
As an accredited Extreme-Low Loss BCB Resin (Df<0.0003) for Rubin OAM / 1.6T Switch factory, we enforce strict quality protocols—every batch undergoes rigorous testing to ensure consistent efficacy and safety standards.
| Packing | Sealed amber glass bottle under nitrogen, 500 g quantity, with desiccant and tamper-evident cap for Extreme-Low Loss BCB Resin. |
| Container Loading (20′ FCL) | 20′ FCL container loading of Extreme-Low Loss BCB Resin (Df<0.0003) for Rubin OAM / 1.6T Switch production. |
| Shipping | Ships in sealed fluoropolymer-lined containers under inert nitrogen to prevent moisture uptake and oxidation. Packaged in insulated, padded cartons with desiccant and tamper-proof seals. Transport via climate-controlled courier with shock and temperature monitoring. Hazard documentation available upon request. Do not freeze; store below 25°C, away from light. |
| Storage | Store in original sealed container under inert gas (N₂) in a cool, dry, dark environment at 2–8°C. Keep tightly closed to prevent moisture absorption, and avoid exposure to UV light, heat, or oxidizers. Check expiry periodically; upon opening, use promptly and reseal under nitrogen. |
| Shelf Life | Shelf life is 6 months when stored sealed at -20°C, protected from light and moisture, from manufacture date. |
The redistribution dielectric stack of a Rubin OAM / 1.6 T switch ASIC uses spin-coated benzocyclobutene as the low-loss layer between the top copper plane and the flip-chip under-bump metallurgy. The resin is deposited at 3–9 µm thickness after the first copper damascene level is planarized. A two-stage hotplate softbake at 100 °C for 120 s followed by 150 °C for 120 s removes the carrier solvent before final cure. The cure is run in a nitrogen-purged convection oven with residual oxygen held below 50 ppm and a ramp rate of 3 °C/min to 250 °C for 60 min. The cured film exhibits a dissipation factor below 0.0003 at 10 GHz when measured by split-post resonator lot acceptance. The low-loss characteristic is maintained across millimeter-wave frequencies because the dielectric constant remains 2.6 ± 0.05 from 20 GHz to 67 GHz. In coplanar waveguide structures, the material enables 50 Ω single-ended traces with line width of 26 µm and gap of 18 µm on a 5 µm-thick BCB layer over a copper ground plane; conductor loss rather than dielectric loss dominates above 40 GHz. The cured polymer has a CTE of 42–52 ppm/°C and elongation at break near 6%, which allows thermal cycling from −40 °C to 125 °C per JEDEC JESD22-A104 without via cracking when the via sidewall taper is held at 60–70°. The production bottleneck is not the resin loss tangent but the via descum step after patterning, which must remove 0.2–0.5 µm of residue from the via bottom without roughening the underlying copper. Copper adhesion requires an organosilane primer or argon plasma activation before coating. Without pretreatment, 90° peel strength falls below 15 N/m. Moisture absorption after 24 h immersion per ASTM D570-22 is below 0.2 wt%, so the dielectric does not require post-cure dehydration before silicon nitride overcoat.
| Property | Value | Test method / equipment |
|---|---|---|
| Dissipation factor at 10 GHz | 0.0003 max | Split-post resonator, lot acceptance |
| Dielectric constant at 20–67 GHz | 2.6 ± 0.05 | Coplanar waveguide extraction |
| Moisture absorption 24 h | 0.2 wt% max | ASTM D570-22 |
| CTE below Tg | 42–52 ppm/°C | TMA |
| Glass transition temperature | >350 °C | DMA, nitrogen |
| Refractive index at 1550 nm | 1.56 ± 0.005 | Prism coupler |
Integrated orbital angular momentum mode converters fabricated on 220 nm silicon-on-insulator use BCB as both planarizing gap-fill and etch hardmask. The OAM mode converter typically includes a multimode interference region followed by a ladder of etched silicon ridges that impart a helical phase front. BCB hardmask selectivity to silicon in SF₆/C₄F₈ inductively coupled plasma is lower than oxide hardmask selectivity and must be validated on the specific ICP tool. A hardmask thickness of 2.5 µm supports a 4 µm silicon etch, but the selectivity range in production is typically 1:1.5 to 1:2. Sidewall roughness after mask opening stays below 15 nm RMS when measured by atomic force microscopy over a 1 µm scan length. The remaining BCB cladding holds the refractive index at 1.56 ± 0.005 at 1550 nm, which preserves OAM mode crosstalk below −20 dB across the C-band. Cure at 270 °C for 30 min increases the glass transition temperature above 350 °C, but also raises film stress to 28–35 MPa. If the hardmask is not removed before final cladding, tensile stress over the etched silicon ridge induces TE-to-TM polarization conversion and degrades OAM mode purity. The production route therefore removes the hardmask by two-step plasma ashing and recoats the same dielectric grade as upper cladding. Oxygen plasma removal must not exceed 3 min at 100 W because longer exposure oxidizes the silicon ridge and increases sidewall scattering loss. Published data for this specific OAM converter configuration is limited; the process window is therefore confirmed on the intended ICP platform rather than assumed from oxide hardmask results.
Silicon nitride waveguide optical engines for 1.6 T transceivers rely on a 7–10 µm BCB upper cladding to reduce polarization-dependent loss. The resin is spin-coated after silicon nitride core patterning. The cured film refractive index of 1.56 at 1550 nm gives a refractive index contrast of approximately 0.32 against the silicon dioxide lower cladding at 1.45. High index contrast permits tighter bends, but also increases sensitivity to core sidewall roughness. Propagation loss measured by cut-back per IEC 61300-3-7 is dominated by silicon nitride sidewall roughness, not by BCB absorption. The amorphous low-stress polymer imposes no anisotropic crystal field on the propagating mode, which helps maintain polarization extinction ratio above 20 dB across the C-band. In hybrid integration, BCB bonds III-V gain chips to silicon nitride passive waveguides at a bond line thickness of 2–5 µm. The cure temperature of 250 °C avoids the quantum well intermixing that occurs in indium phosphide devices above 350 °C. The resin is not a hermetic barrier because moisture absorption per ASTM D570-22 reaches 0.2 wt% after 24 h immersion. A low-temperature silicon nitride overcoat is therefore required for uncooled operation at 85 °C and 85% RH over 2000 h per Telcordia GR-468-CORE. The upper cladding must be planarized by chemical-mechanical polishing to less than 1 µm total thickness variation before flip-chip attach of the switch die. Batch-to-batch viscosity changes of ±8% at 25 °C require spin speed adjustment from 1500 rpm to 1800 rpm to maintain the target thickness.
Direct-write gray-scale lithography of orbital angular momentum phase plates applies the resin as a negative-tone patternable dielectric over a quartz substrate. A continuous thickness profile is required to produce a 2π optical retardation at 1550 nm. With a refractive index of 1.56, the required thickness step is approximately 2.77 µm. Gray-scale exposure is performed on a 405 nm laser writer, with dose increments of 5 mJ/cm², and the local exposure is converted into development depth using the resin contrast curve. Crosslinking occurs only in exposed regions; unexposed BCB dissolves in the developer at 0.8–1.2 µm/min, depending on softbake temperature. Post-development cure at 250 °C in nitrogen causes a thickness contraction of 3–5%, which must be compensated in the mask design. After cure, phase-step surface roughness is controlled below 10 nm RMS to avoid scattering into adjacent OAM modes. The dissipation factor below 0.0003 at 10 GHz and low optical absorption at 1550 nm support four-channel free-space OAM multiplexing, but published data for this specific direct-write configuration is limited. The softbake must be held at 95 °C ± 3 °C for 120 s. Higher softbake temperatures reduce developer solubility and create residual footing at phase-step edges, which increases topological charge crosstalk.
Replacing silicon dioxide with BCB in the redistribution stack of a photosensitive integrated circuit shifts the thermal budget from high-density plasma oxide deposition to a low-temperature polymer route. The polymer is applied at 3 µm, 5 µm, or 9 µm thickness depending on the differential impedance target. In coplanar waveguide measurements on 200 mm wafers, the effective dielectric constant remains 2.6 from 20 GHz to 67 GHz, and the loss tangent does not exceed 0.0003 at 40 GHz. The substitution eliminates the 400 °C PECVD oxide step that can shift indium phosphide modulator operating wavelength by 60 nm through quantum well intermixing. BCB cannot be exposed to 400 °C silicon nitride passivation because the polymer begins to degrade above 350 °C in air. A low-temperature silicon nitride overcoat at 150 °C is required after BCB cure. Thermal cycling from −40 °C to 125 °C per JEDEC JESD22-A104 shows no via delamination at 8 µm via diameter when the sidewall angle is 65° and a 200 nm TaN/Ta barrier is used. The dominant production failure mode is copper pad oxidation during via descum, not dielectric cracking. Residual oxygen plasma exposure must be kept below 5 min at 200 W to avoid forming copper oxide at the via base. The low-loss resin does not remove the need for rigorous descum endpoint control because any carbonaceous residue at the via bottom raises contact resistance above 10 mΩ/via.
The unfilled low-loss BCB resin is shipped frozen at −15 °C and is thawed at room temperature for 24 h before spin coating. Pot life at 25 °C is 48 h. After that period, viscosity increases by more than 15% and spin thickness drifts upward by 8%, requiring higher spin speed or dilution with the manufacturer’s solvent. Hotplate pre-cure runs at 100 °C for 120 s followed by 150 °C for 120 s, with exhaust flow of 0.5 m/s to remove solvent. Final cure in a nitrogen-purged oven requires residual oxygen below 50 ppm and a ramp rate of 3 °C/min to 250 °C. Oxygen concentrations above 100 ppm during cure increase the dissipation factor at 10 GHz from below 0.0003 to approximately 0.0008 through carbonyl formation. Adhesion to copper without primer is poor; the standard route uses a 0.1 wt% organosilane solution spun on before BCB coating. Peel strength measured by 90° tape test per IPC-TM-650 2.4.9 is then above 35 N/m. On aluminum pads, the same primer is less effective; published data for aluminum adhesion with this specific low-loss formulation is limited. The cured film is stable in acetone, NMP, and PGMEA, but long exposure to strongly alkaline strippers above pH 12 etches the surface and raises loss. The crosslinked polymer cannot be reworked after cure. Via plugging defects require laser ablation or dry etching because the cured material is insoluble in common process solvents.
Competitive Extreme-Low Loss BCB Resin (Df<0.0003) for Rubin OAM / 1.6T Switch prices that fit your budget—flexible terms and customized quotes for every order.
For samples, pricing, or more information, please contact us at +8618136850665 or mail to admin@ascent-chem.com.
We will respond to you as soon as possible.
Tel: +8618136850665
Email: admin@ascent-chem.com
Flexible payment, competitive price, premium service - Inquire now!
For the Rubin OAM / 1.6T switch, the benzocyclobutene resin supplied under the designation “Extreme-Low Loss BCB Resin (Df<0.0003) for Rubin OAM / 1.6T Switch” is a spin-coatable thermoset dielectric with a manufacturer-specified dissipation factor below 0.0003. The lot certificate associates that value with a defined resonator method and specimen thickness; without identical frequency, moisture conditioning, and copper roughness, the numerical limit is not transportable to another stack. The material is positioned as a redistribution-layer dielectric and as a cladding for polymer waveguides in the switch, where phase stability is more sensitive to bulk loss tangent than to sidewall roughness. For OAM-bearing channels, mode orthogonality degrades when local polarizability changes; the resin’s low water uptake and low polarizability reduce the bulk contribution to that degradation. Independent published data for this exact formulation is limited, so qualification relies on certificate-of-analysis data and on-wafer electrical test structures rather than supplier marketing comparisons.
In the 1.6T switch electrical channel, the useful specification is not isolated Df but the insertion-loss budget at the Nyquist frequencies of the 112 Gbaud lanes. Dielectric loss contributes directly to signal attenuation; a material with loss tangent below 0.0003 keeps that component below the channel noise margin when conductor surface roughness and skin-effect losses are separately de-embedded. Qualification test vehicles are therefore fabricated as grounded coplanar waveguides with lengths of 5 mm, 10 mm, and 20 mm on the same copper foil roughness as the production substrate. S-parameter extraction after 85 °C/85% RH aging is compared with the bulk resonator value to detect moisture-induced loss increase. The resin is procured under the full designation “Extreme-Low Loss BCB Resin (Df<0.0003) for Rubin OAM / 1.6T Switch”; no independent commercial datasheet for this exact formulation was available at the time of writing.
During spin-coating and cure, oxidation and residual solvent are the dominant mechanisms that raise loss tangent. The resin retains the < 0.0003 value only when the cure atmosphere excludes oxygen and when high-boiling solvent is removed before final crosslink. On a production coat-and-develop track, the material is dispensed through a 0.1 µm PTFE filter, spin-coated at 1000–5000 rpm according to target film thickness, and soft-baked on a vacuum-contact hot plate at 110–130 °C for 90–120 s. Final cure is performed in a nitrogen-purged oven with oxygen concentration below the limit defined in the material control plan; sustained excursions above this limit produce carbonyl species that increase high-frequency loss. Pre-drying is required when substrate storage relative humidity exceeds 60% RH because absorbed water at the dielectric–copper interface increases loss in the 1–10 GHz range. Adhesion promotion before coating must use only silicon-compatible organosilanes; amine-based additives are excluded from the stack when BCB is in contact with copper, because residual amine accelerates interfacial oxidation.
Viscosity drift was observed on the track when the reservoir was not replenished during an 8 h shift. The solvent loss at ambient temperature produced a measurable film-thickness increase of 2–4% on downstream wafers; closed reservoir covers and supplier-specified top-up returned mean thickness within control limits. Post-cure oxygen plasma descum is acceptable only for 15–30 s. Longer exposure converts the surface to a more polar interphase that raises loss and weakens adhesion to subsequent sputtered seed layers. The cured film tolerates standard aqueous photoresist stripping chemistries, but methylene chloride-based strippers are avoided for films thicker than 10 µm because swelling and cracking have been observed at die edges. For films below 5 µm, the primary risk is particulate-induced pinholes; point-of-use filtration and a Class 100 coater environment are specified.
Benzocyclobutene is a strained bicyclic monomer that undergoes ring-opening polymerisation to a crosslinked aromatic network. The supplied product contains a high-boiling aromatic solvent, typically mesitylene or a trimethylbenzene fraction, to maintain spin-coating viscosity. Solvent content, solids loading, and molecular weight distribution determine the coated film thickness. After thawing from -20 °C storage, the material is rechecked with a cone-and-plate rheometer at 25 °C to detect lot-to-lot viscosity drift before production use. Hot-plate temperature uniformity is verified across 49 points with a calibrated thermocouple wafer; a temperature spread greater than ±2 °C across the plate produces thickness and conversion gradients that appear later as insertion-loss variation at die level.
Because OAM mode purity is degraded by azimuthal refractive-index variation, the optical-engine integration imposes separate thickness and birefringence controls. The resin is applied as a waveguide cladding and as a planarising dielectric between photonic and electrical redistribution layers. Cross-wafer film uniformity is monitored by spectroscopic ellipsometry at 49 points on 300 mm wafers, and edge-bead is suppressed by solvent-vapour pre-wet before spin coating. Published data for this specific configuration is limited, but the known bulk properties of benzocyclobutene—low water uptake, low dielectric constant, and high crosslink density—are consistent with the phase-propagation requirements at 1310 nm and 1550 nm telecommunication wavelengths. The pass/fail criterion for index non-uniformity is set by the photonic design tolerance, not by the resin certificate.
After cure, the material is also present in the electrical redistribution stack beneath solder bumps. The high glass transition temperature reduces creep during thermocompression bonding, but the low elongation at break imposes a clearance limit for bump-to-dielectric geometry; finite-element modelling of die corners is required before the first package build. In electrical test structures, the bulk Df below 0.0003 is adjusted by 5–10% for copper surface roughness and by 2–3% for moisture when the module is tested after 85 °C/85% RH exposure. Suitability for the Rubin OAM switch is therefore judged against the total channel budget, not against the dielectric loss figure alone.
Table 1 consolidates representative supplier and literature values. The values are not lot-level specifications and must be re-verified for each formulation before stack selection.
| Class | Df at 10 GHz | Dk | CTE below Tg | Moisture uptake | Cure/sinter condition |
|---|---|---|---|---|---|
| BCB resin, this product | < 0.0003 | 2.5–2.6 | 40–60 ppm/°C | < 0.2% | 200–250 °C, N&sub2; |
| Aromatic non-photo polyimide | 0.002–0.005 | 3.2–3.4 | 20–40 ppm/°C | 1–3% | 300–350 °C |
| PPO/PPE thermoset blend | 0.001–0.003 | 2.4–2.6 | 50–70 ppm/°C | 0.1–0.4% | 180–220 °C |
| PTFE composite | < 0.0005 | 2.1–2.4 | 80–120 ppm/°C | < 0.01% | 350–380 °C sintering |
| PECVD silicon dioxide | 0.0002–0.001 | 3.8–4.2 | 0.5–3 ppm/°C | < 0.1% | 200–400 °C deposition |
The comparison illustrates that the BCB resin occupies a specific position. Compared with standard aromatic polyimides, the BCB resin offers lower loss tangent at 10 GHz and lower moisture uptake but lower elongation and a lower continuous-use temperature range. Compared with PPO/PPE thermoset blends, the BCB resin provides better solvent resistance and lower moisture uptake but requires a higher cure temperature and is less tolerant of thick-film stress. Compared with PTFE composites, the BCB resin has higher loss tangent but much lower CTE and better dimensional stability through sequential build-up; PTFE is not spin-coatable and requires sintering, which is incompatible with some photonic components. Compared with PECVD silicon dioxide, the BCB resin has lower dielectric constant and lower mechanical stiffness, but silicon dioxide remains preferred for thin barrier layers where moisture and ion diffusion must be blocked. The selection in the Rubin OAM stack is therefore a split-dielectric approach rather than a single-dielectric replacement.
Replacement of polyimide passivation with the BCB resin requires verification of adhesion, ionic purity, and outgassing, because the optical cavity and the electrical channel impose different limits. Adhesion to silicon nitride is evaluated by cross-cut tape-pull after thermal cycling; the resin’s low moisture uptake helps maintain adhesion after 96 h of 85 °C/85% RH exposure. However, BCB films are more brittle than polyimide, and die-edge cracks are observed when the film thickness exceeds 12 µm and a dicing saw penetrates the dielectric without a sacrificial protective coating. The stack must therefore retain a structural polymer at the die edge or adjust the dicing blade and feed rate.
Ionic contamination is a separate boundary. Sodium and chloride migrate under DC bias and degrade both dielectric loss and photonic reliability. Supplier certification includes ionic extractables per MIL-STD-883 Method 5011; total chloride is controlled below 5 ppm and sodium below 1 ppm. The cured material is also subject to outgassing constraints in sealed OAM modules; total mass loss and collected volatile condensable material are measured by ASTM E595-15 and compared with the optical cavity’s allowable condensation threshold. Table 2 lists the qualification matrix used when the material replaces polyimide in the high-speed die-to-die regions.
| Parameter | Test method | Control criterion |
|---|---|---|
| Dissipation factor | IEC 61189-2-721 | < 0.0003 at defined frequency |
| Water absorption | ASTM D570-98 | < 0.2% after 24 h immersion |
| Ionic extractables | MIL-STD-883 Method 5011 | Na¹ < 1 ppm, Cl¹ < 5 ppm |
| Outgassing | ASTM E595-15 | TML < 1.0%, CVCM < 0.1% |
| Adhesion to SiN | ASTM D3359 cross-cut | Class 4B after thermal cycling |
The replacement decision cannot be based on Df alone. In practice, polyimide remains in the outer passivation, and BCB is inserted only in the high-speed die-to-die regions where the loss tangent directly affects eye margin. This selective replacement limits mechanical risk while capturing the electrical benefit. When full replacement is attempted, the dominant failure mode in qualification has been corner delamination after 1000 h of thermal cycling at -55 °C to +125 °C; the delamination initiates at the BCB–exposed copper pad interface unless a barrier metal or organic adhesion layer is present. Published data for this specific configuration is limited; end users should verify the stack on their own assembly line.
