| HS Code | 916192 |
| Dielectric Constant Dk 10ghz | 3.3 |
| Dissipation Factor Df 10ghz | 0.0015 |
| Glass Transition Temperature Tg | >200°C |
| Thermal Decomposition Temperature Td 5 Weight Loss | >350°C |
| Peel Strength To Copper Foil | 0.8 N/mm |
| Moisture Absorption | 0.1% |
| Coefficient Of Thermal Expansion Cte | 10 ppm/°C |
| Thermal Conductivity | 0.5 W/m·K |
| Flexural Strength | 180 MPa |
| Flammability Rating | UL94 V-0 |
| Surface Resistivity | 1×10^12 Ω |
| Volume Resistivity | 1×10^14 Ω·cm |
As an accredited M9 Ultra-Low Loss Hydrocarbon Resin for AI Server High-Speed Boards factory, we enforce strict quality protocols—every batch undergoes rigorous testing to ensure consistent efficacy and safety standards.
| Packing | Packaged in 25 kg nitrogen-filled steel drums with moisture-proof inner lining, labeled with batch number and safety handling instructions. |
| Container Loading (20′ FCL) | M9 ultra-low loss resin loaded in 20′ FCL, drummed and palletized, securely fastened, protected from moisture for safe transport. |
| Shipping | M9 Ultra-Low Loss Hydrocarbon Resin ships in sealed, moisture-proof containers to preserve dielectric integrity. Handle with care, avoid extreme heat or impact. Standard ground or air freight acceptable; no special hazardous classification. Ensure proper labeling and documentation for traceability. Store upright in dry, ventilated areas until use. |
| Storage | Store in a tightly sealed, original container in a cool, dry, well-ventilated area, away from direct sunlight, heat sources, and oxidizers. Keep temperature stable between 10–30°C. Avoid moisture contamination and prolonged exposure to air. Use explosion-proof equipment if dust forms. Under recommended conditions, shelf life is typically 12 months from manufacture date. |
| Shelf Life | Shelf life is 12 months from manufacture when stored sealed, cool, and dry, away from sunlight and contamination. |
In insertion-loss budgeting for a 112 Gb/s PAM4 channel on a 1.6 mm, 32-layer AI accelerator mainboard, the glass style is fixed to 1035 spread-glass with a fabric Dk of 3.5, and the resin matrix contribution is isolated by comparing measured channel loss against simulated conductor loss. The M9 Ultra-Low Loss Hydrocarbon Resin is incorporated into a polyphenylene ether/TAIC varnish at 18–28 wt% of total organic resin solids. Filler pre-drying at 120 °C for 4 h is mandatory when ambient RH exceeds 60%. Below 18 wt% the cured laminate crosses above a loss tangent of 0.0030 at 10 GHz, while above 28 wt% the B-stage becomes brittle enough to crack during sheeting and resin flow falls below 2.0%. Published production data for the exact upper-bound threshold at 28 wt% remains limited; qualification batches are typically bracketed from 18 wt% to 25 wt% before committing to the upper edge. Compliance verification is performed to IPC-4101E/126 with dielectric testing by IPC-TM-650 2.5.5.5 at 10 GHz and IPC-TM-650 2.5.5.13 at 10 GHz, dimensional stability by IPC-TM-650 2.4.24.2, and flammability by UL 94 V-0. In production, the varnish is mixed in a jacketed high-shear disperser at 35–45 °C for 90–120 min using MEK/toluene to a final viscosity of 120–220 cPs at 25 °C; the fabric is impregnated on a horizontal treater with 16 oven zones, first stage 90–110 °C, final stage 150–170 °C, with B-stage gel time controlled to 180–240 s at 171 °C and resin flow of 2.0–4.5%. The varnish is incompatible with amine-based accelerators; contact with primary or secondary amines causes exothermic viscosity build and premature gelation before the treater. Layup is pressed in a vacuum hot press at 190–210 °C under 2.0–3.5 MPa for 120–150 min, followed by post-bake at 200 °C for 60 min to complete crosslinking. The cured CCL is converted into 24- to 56-layer AI accelerator mainboards, OAM 2.0 compute modules, UBB baseboards, and 800 Gb/s switch line cards.
| M9 addition (wt% of organic resin) | Dk at 10 GHz | Df at 10 GHz | B-stage resin flow (%) |
|---|---|---|---|
| 15 | 3.4 | 0.0031 | 5.0 |
| 20 | 3.3 | 0.0027 | 3.5 |
| 25 | 3.2 | 0.0024 | 2.8 |
| 30 | 3.1 | 0.0026 | 1.7 |
For a 48-layer switch backplane, the interface between a bonding sheet and an etched innerlayer core is a discrete insertion-loss discontinuity, and the M9 loading in the bonding sheet is reduced to 15–22 wt% of total organic resin to hold cured resin flow between 2.0% and 3.5%. The reduced loading maintains enough melt viscosity to fill 3.5 mil trace geometries over 0.5 oz copper without excessive squeeze-out into clearance regions, but it still retains a post-cure Df below 0.0025 at 10 GHz when measured by IPC-TM-650 2.5.5.5. Process control on production treaters specifies a final prepreg resin content of 68–74 wt% on 1067 or 1080 low-Dk glass, B-staging at 150–165 °C, and sealed polyethylene storage at 5–10 °C for no longer than 30 days; outlife at 23 °C and 50% RH is limited to 72 h because moisture uptake above 0.10% has been associated with increased CAF incidence after 85 °C/85% RH bias testing. Vacuum press operators observe that residual solvent above 0.5% in the B-stage creates resin splash and non-wetting at the copper interface during the 200 °C ramp, producing microvoids that are visible at 500x cross-section inspection and are not eliminated by a subsequent post-bake. Compliance for the bonding sheet is documented under IPC-4101E/126 and IPC-TM-650 2.4.24.2 for thermal expansion, with flammability classification UL 94 V-0. The bonding sheets are converted into 32- to 48-layer backplanes, midplanes, and switch line cards for 800 Gb/s data-center switching fabrics.
The substitution of low-profile RTF copper (Rz < 2.5 μm) for high-profile HTE copper on a 56-layer AI spine switch backplane shifts the adhesion burden toward the resin matrix; the low-profile nodule geometry does not create the same mechanical interlock as HTE copper, so the M9 addition is held at 20–27 wt% of total organic resin to balance crosslink density against peel strength. Below 20 wt%, copper peel after thermal stress measured to IPC-TM-650 2.4.8.3 drops below 0.70 N/mm on RTF foil, while above 27 wt% the cured resin modulus exceeds the strain window of the low-profile foil during innerlayer reflow. Vacuum lamination is performed at 30–50 mbar with a ramp of 2–3 °C/min to 190–200 °C, held at 2.0–3.0 MPa for 120–150 min, with press cycles staged to allow the resin to fill 3.5 mil lines over 0.5 oz copper before gelation. Compliance is maintained under IPC-4101E/126 and IPC-TM-650 2.4.24.2 for z-axis expansion, with copper adhesion verified after 288 °C solder float for 20 s. The cured laminate is used for 48- to 56-layer AI spine switch backplanes and 800 Gb/s fabric line cards.
Halogen-free resin formulations for AI storage backplanes replace brominated flame retardants with 20–35 phr of precipitated silica and 5–10 phr of aluminum hydroxide in a varnish system containing M9 at 18–25 wt% of organic resin solids. The M9 addition is kept below 25 wt% because higher loadings reduce the flow needed to encapsulate flame-retardant filler particles, and the resulting microporous region at the glass-fiber interface has been observed in 85 °C/85% RH CAF testing to initiate conductive anodic filament growth. Compliance is verified under IEC 61249-2-26 for halogen content, IPC-4101E/126 for laminate classification, UL 94 V-0 for flammability, and RoHS 2011/65/EU with REACH 1907/2006/EC SVHC screening. In production, the flame-retardant filler is dispersed in a high-shear mixer at 40 °C for 60 min before the M9 resin is added, avoiding a viscosity spike that narrows the coating window on the horizontal treater. Impregnation uses 106 or 1080 glass, and B-staging is held at 150–160 °C to keep residual solvent below 0.4%. Curing is performed at 190–205 °C under 2.5–3.0 MPa for 120 min. Moisture absorption after 24 h water soak at 23 °C is limited to 0.15% by the filler loading, and the cured laminate is used in 20- to 36-layer AI storage backplanes, edge-server mainboards, and top-of-rack switch line cards.
A 5-stage sequential lamination process for a 20-layer HDI mezzanine board in an AI server chassis subjects the M9-modified dielectric to 4 additional cure excursions at 180–200 °C, and the formulation is adjusted to 22–30 wt% M9 to preserve uncured resin flow for each subsequent lamination. The B-stage surface after first lamination is exposed to UV laser drilling at 355 nm, 9–12 W; alkaline permanganate desmear at 70–80 °C removes drill smear and creates micro-roughening for electroless copper at 0.3–0.8 μm. The higher M9 loading within the 22–30 wt% range reduces the loss tangent of the dielectric build-up film to below 0.0028 at 10 GHz, as measured by IPC-TM-650 2.5.5.13, while remaining below the threshold for brittle B-stage cracking. Qualification is conducted to IPC-6012 Class 3 with acceptance inspection to IPC-A-600 Class 3, and the finished boards are assembled into HDI mezzanine modules, accelerator interposers, and NVMe storage backplanes inside dense AI server enclosures.
Thermal reliability verification of M9-containing CCL for AI server boards uses 288 °C solder float for 20 s followed by cross-section inspection at 200x and 500x. The acceptance boundary is the absence of measling, delamination, or resin recession at glass bundle tips, and the M9 addition is fixed at 20–25 wt% for the final board construction because lower addition fails the TMA expansion limit after 3 thermal cycles, while higher addition increases the elastic modulus of the resin beyond the copper elongation window. Verification is performed to IPC-TM-650 2.4.8.3 for copper peel after thermal stress and IPC-TM-650 2.4.24.2 for thermal expansion, with flammability confirmed by UL 94 V-0. The compliance matrix is applied to each production lot before release, and the cured CCL is routed into final AI server board constructions including compute modules, switch line cards, and power-delivery backplanes requiring Class 3 workmanship acceptance.
| Test | Method | Condition | Acceptance boundary |
|---|---|---|---|
| Dielectric constant/loss | IPC-TM-650 2.5.5.5 | 10 GHz | Df ≤ 0.0025 |
| Copper peel after thermal stress | IPC-TM-650 2.4.8.3 | 288 °C/20 s | ≥ 0.70 N/mm |
| Thermal expansion | IPC-TM-650 2.4.24.2 | TMA | CTE α1 ≤ 45 ppm/°C |
| Flammability | UL 94 | Vertical burn | V-0 |
| Halogen content | IEC 61249-2-26 | Combustion IC | Cl 900 ppm, Br 900 ppm, total 1500 ppm |
Competitive M9 Ultra-Low Loss Hydrocarbon Resin for AI Server High-Speed Boards 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 high-layer-count switch ASIC carrier boards and AI accelerator daughtercards operating at 112 Gb/s and 224 Gb/s PAM-4 signaling, dielectric loss tangent Df and signal attenuation per unit length directly determine channel reach and equalizer complexity. In a 2.0 mm total-thickness backplane, a Df difference of 0.001 at 10 GHz can shift insertion loss by more than 0.5 dB/in at the Nyquist frequency when conductor roughness is held constant. M9 Ultra-Low Loss Hydrocarbon Resin is supplied as the matrix resin for prepreg and core constructions intended for AI-server high-speed boards. Manufacturer-stated target data for M9 include dielectric constant Dk of 2.35 at 10 GHz and Df of 0.0014 at 10 GHz when tested per IPC-TM-650 2.5.5.5. These values are obtained on a 0.8 mm thick laminate produced with low-Dk E-glass. Df after three reflow cycles at 260°C typically increases by less than 0.0002 when conditioned at 23°C and 50% RH for 24 h, but this target should be confirmed by the fabricator because copper-foil roughness and glass style influence the measured value. Independent verification for a specific AI-server stack-up remains limited until controlled impedance coupons are measured per IPC-TM-650 2.5.5.13 and conductor roughness corrections are applied. Within the same hydrocarbon resin family, earlier designations M7 and M8 are associated with higher Df targets; published data for direct substitution into AI-server boards are limited.
Base chemistry of M9 differs from first-generation dicyclopentadiene-derived hydrocarbon resins in three measurable ways: lower melt viscosity at comparable molecular weight, higher vinyl reactivity, and broader tolerance for fused-silica filler loading. The resin is supplied as a hydrocarbon oligomer with number-average molecular weight Mn of 1,800 g/mol and polydispersity index 2.1 by gel permeation chromatography against polystyrene standards. Melt viscosity is 750 Pa·s at 150°C measured on a controlled-stress rheometer with 25 mm parallel plates at 1 rad/s. Batch-to-batch variation is specified at ±15%; values outside this band have been associated with inconsistent prepreg impregnation on a 40:1 L/D twin-screw extruder. The target filler loading is 55–62 wt% fused silica with an epoxy-functional trimethoxysilane surface treatment at 0.8 wt% on filler. Excess silane above 1.2 wt% produces volatile condensate during lamination and has been observed to increase void content. At the target filler loading, M9 retains a spiral-flow length of 18 mm at 170°C and 7 MPa clamping pressure. Standard hydrocarbon systems with similar Df typically accept less than 45 wt% filler before flow length falls below 10 mm. The higher filler loading contributes to a lower in-plane coefficient of thermal expansion without requiring a more rigid backbone that would increase Dk.
The prepreg is typically manufactured as a 65% resin-content prepreg on 1067, 1078, and 2116 E-glass styles, with controlled gel time of 120 ±10 s at 170°C by stroke-cure method. In roll-to-roll immersion impregnation, the resin bath viscosity is held at 150 ±20 Pa·s at 40°C to maintain resin-content uniformity. The melt viscosity at lamination drops to 350 Pa·s at 160°C before gelation; if viscosity falls below 300 Pa·s, resin starvation into 50 μm copper clearances becomes difficult to control.
| Property | Test method | Target value | Test condition / notes |
|---|---|---|---|
| Dielectric constant Dk at 10 GHz | IPC-TM-650 2.5.5.5 | 2.35 | 23°C/50% RH, reinforced laminate |
| Dissipation factor Df at 10 GHz | IPC-TM-650 2.5.5.5 | 0.0014 | 23°C/50% RH, reinforced laminate |
| Glass transition temperature Tg | ASTM D3418 | 210°C | DSC, midpoint |
| CTE x/y below Tg | IPC-TM-650 2.4.24 | 13 ppm/°C | 25–150°C |
| CTE z below Tg | IPC-TM-650 2.4.24 | 45 ppm/°C | 25–150°C |
| Water absorption | IPC-TM-650 2.6.2.1 | 0.12% | 24 h immersion |
| Copper peel strength | IPC-TM-650 2.4.8 | 0.78 N/mm | 1 oz reverse-treated foil |
| Decomposition temperature 5% mass loss | ASTM E1131 | 415°C | nitrogen, 10°C/min |
The target values in Table 1 are manufacturer-stated design data and are not independent lot-certified values. Procurement specifications should fix the exact tolerance for Df because batch-to-batch filler dispersion can shift Df by 0.0003 if silane treatment is not optimized.
On a production-scale vacuum-assisted hydraulic press, the M9 prepreg system has a defined processing window. The lamination temperature plateau is 200°C ±5°C; excursions above 205°C accelerate free-radical crosslinking before the resin has fully filled 50 μm trace gaps, leading to edge voiding and a drop in characteristic impedance consistency. The gel point at 200°C is 280 s by moving-die rheometry. Full pad pressure of 2.4 MPa must be reached before 220 s to maintain flow into dense routing areas. In fabricator trials on a 40:1 L/D twin-screw extruder, barrel temperatures above 180°C produced melt-pressure excursions above 2,400 kPa at the die and resulted in edge voids in 0.1 mm core laminates. At lamination temperatures below 195°C, incomplete crosslinking produces low Tg cores and increases moisture sensitivity after two-sided reflow. Conversely, sustained temperature above 205°C causes dielectric spacing variation in buried stripline structures because the resin begins to crosslink before final thickness is reached.
Desmear following laser drilling is acceptable with permanganate chemistry; etchback is held below 2 μm because the silane-treated fused silica is inert. Plasma processing with CF4/O2 at 80:20 and 500 W for 20 min yields hole-wall roughness Ra of 0.25 μm when measured on a stylus profilometer. Mechanical drilling of 0.2 mm diameter vias at 40 kRPM spindle speed and 25 μm/rev infeed produces acceptable nail-heading when the material is pre-dried at 105°C for 4 h if RH exceeds 60%. Electroless copper deposition after desmear at 45°C for 20 min gives a backlight rating of 8.5 for 0.25 mm through-holes at 10:1 aspect ratio. After solder shock at 288°C for 6 cycles per IPC-TM-650 2.6.8, peel retention remains above 70% of the initial value on oxide-free copper. The use of oxide passivation is not required for M9 adhesion but does not degrade Df if oxide thickness is below 1 μm. Additives containing reactive amines should not be introduced into the prepreg bath because they interfere with the peroxide-initiated cure and delay the gel point by more than 25 s.
In stack-up design, the choice among M9, PTFE/ceramic, and epoxy/PPO is governed by insertion loss, dimensional stability, and fabrication-line capability. PTFE/ceramic systems often provide Df below 0.001 and Dk near 2.2, but require lamination temperatures above 350°C and specialized sodium-etch or plasma surface treatments. Epoxy/PPO systems process at lower temperatures but exhibit Df near 0.005 and higher moisture absorption, reducing margin at 224 Gb/s. M9 is positioned between these extremes: it does not achieve the lowest possible Df of PTFE, but it avoids the high-temperature lamination and bond-surface preparation constraints of PTFE. Table 2 compares manufacturer-published typical values at 10 GHz.
| Parameter | M9 target | PTFE/ceramic typical | Epoxy/PPO modified typical |
|---|---|---|---|
| Dk at 10 GHz | 2.35 | 2.20 | 3.40 |
| Df at 10 GHz | 0.0014 | 0.0009 | 0.0050 |
| Tg by DSC | 210°C | 250°C | 180°C |
| CTE x/y | 13 ppm/°C | 24 ppm/°C | 45 ppm/°C |
| Water absorption | 0.12% | 0.05% | 0.40% |
| Lamination plateau | 200°C | 370°C | 190°C |
| Surface preparation | Reverse-treated foil; no oxide passivation required | Sodium etch or plasma overbond treatment | Standard oxide or bondfilm |
The comparison shows that M9 does not match PTFE/ceramic on Df and water absorption but enters the fabrication line through conventional hydrocarbon processing and does not require a sodium-etch step. M9 is compatible with standard FR-4 inner-layer bonding. Relative to first-generation hydrocarbon resin systems with Df near 0.0025 at 10 GHz, M9 improves the loss budget by approximately 0.4 dB/in at 28 GHz when laminate construction is held constant. Published independent loss data for exact AI-server stack-ups is limited; fabricators should run insertion-loss qualification coupons with the actual copper foil roughness profile and report the root-mean-square roughness after oxide treatment.
Incoming resin qualification for M9-based laminates is commonly aligned to IPC-4101E/99 and includes the following test designations: IPC-TM-650 2.5.5.5 for Dk/Df, 2.5.5.13 for insertion loss, 2.4.24 for CTE, 2.6.2.1 for water absorption, 2.4.8 for peel strength, and 2.6.25 for CAF resistance. Halogen content is reported under IEC 61249-2-21 with limits of 900 ppm chlorine, 900 ppm bromine, and 1,500 ppm total halogens as typical acceptance values. Because the resin is a non-brominated hydrocarbon system, RoHS recast 2011/65/EU applies to the finished laminate rather than to the resin alone. REACH SVHC content should be below 0.1% w/w per candidate list. UL 94 V-0 data at 0.4 mm thickness are manufacturer-reported and should be verified on the final laminate construction. If laminates are stored above 60% RH, pre-drying at 105°C for 4 h is required before lamination to prevent blistering at 260°C reflow.
Accelerated aging of M9-based laminates at 150°C for 500 h in air indicates that Df at 10 GHz increases by 0.0006 to 0.0008 from the initial target, while Dk increases by less than 0.03. The increase is attributed to oxidative polar species formation at the filler-resin interface rather than bulk chain scission, based on the unchanged decomposition temperature. Aging at 105°C for 1,000 h produces less than 0.0003 Df drift. For AI-server boards operating at continuous junction temperatures near 110°C, the 105°C aging condition is more relevant; qualification should include post-aging insertion-loss coupons rather than relying only on as-pressed Df values. Flexural modulus per ASTM D790-17 remains above 85% of initial after 500 h at 150°C, and Tg by ASTM D3418 decreases by less than 5°C. Published data for this specific configuration under high-humidity biased aging is limited; fabricators should conduct 85°C/85% RH biased CAF testing at 65 V for 1,000 h per IPC-TM-650 2.6.25 before design freeze.
In a 2.4 mm thick AI server backplane stack-up using 0.8 mm cores and 1067 glass prepregs, the M9 resin system is applied in the signal-prepreg layers to control insertion loss; the ground-plane cores can use a lower-cost hydrocarbon resin only if impedance simulation accounts for the Dk mismatch. For a 22-layer board with 0.5 oz signal copper and 1 oz ground copper, insertion-loss qualification at 14 GHz should compare measured values to the fabricator’s de-embedded coupon target. For boards with dense 0.8 mm pitch ball-grid arrays, the higher filler loading of M9 reduces pad cratering relative to standard epoxy/PPO because the z-axis expansion below Tg is lower. However, the same filler loading reduces copper adhesion after multiple reflow cycles if the copper foil treatment is not matched to the matrix; therefore, foil-substrate compatibility must be qualified under IPC-TM-650 2.4.8 after 6 solder floats at 288°C. M9 is not recommended for buried capacitance layers where a ferroelectric-filled high-Dk resin is needed to achieve 10 nF/in², nor for high-power RF power amplifier boards where peak temperatures exceed 250°C because Dk stability is not verified beyond 230°C.
