Underfill Adhesive Coefficient of Thermal Expansion (CTE) Evaluation Testing: Methods, Standards, an

The coefficient of thermal expansion (CTE) of underfill adhesives is a decisive parameter governing the thermo-mechanical reliability of flip-chip, BGA, and advanced semiconductor packages. This article presents a comprehensive evaluation experiment for underfill CTE measurement using thermomechanical analysis (TMA) in accordance with ASTM E831 and ISO 11359-2, covering specimen preparation, test conditions, data interpretation, and acceptance criteria. The resulting data enable packaging engineers to minimize CTE mismatch, suppress solder joint fatigue, and extend product service life across thermal cycling environments.

Technical Overview

Underfill adhesives are epoxy-based encapsulants dispensed beneath flip-chip dies, chip-scale packages (CSP), and ball grid array (BGA) components to redistribute thermo-mechanical stress away from fragile solder interconnects. During temperature cycling, the silicon die (CTE approximately 2.6 ppm/°C), the organic substrate (CTE typically 14–17 ppm/°C), and the copper/solder metallization expand and contract at different rates. The underfill must bridge these mismatches without itself becoming a source of stress, which makes its coefficient of thermal expansion one of the most critical material properties to quantify.

The CTE evaluation experiment described here measures dimensional change of a cured underfill specimen as a function of temperature using thermomechanical analysis. Two distinct expansion regimes are characterized: the glassy-state expansion below the glass transition temperature (α1, typically 20–45 ppm/°C for commercial underfills) and the rubbery-state expansion above Tg (α2, typically 80–150 ppm/°C). The inflection point between these regimes also provides an independent determination of Tg, which is essential because excessive expansion above Tg during reflow or high-temperature operation can induce delamination, fillet cracking, and die backside stress.

A well-designed CTE evaluation program supports material selection, incoming quality control, supplier lot-to-lot comparison, cure schedule optimization, and qualification against industry reliability requirements such as temperature cycling from -40°C to +125°C per JESD22-A104. All testing described below can be performed by accredited third-party laboratories or in-house reliability labs equipped with calibrated TMA instrumentation.

Testing Samples

Standard Cured Plaque Specimens

The primary specimen format is a bulk plaque cast from liquid underfill and cured according to the supplier-recommended schedule (for example, 150°C for 60 minutes or 165°C for 30 minutes in a convection oven). After curing, the plaque is machined or die-cut into rectangular specimens with flat, parallel, and smooth end faces. Typical dimensions are 3 mm to 6 mm in thickness and 5 mm to 10 mm in cross-section, with perpendicularity within 1 degree to avoid probe seating artifacts. Quasi-isotropic cure shrinkage should be verified, and any specimen exhibiting voids, bubbles, or cure gradients visible under magnification is rejected.

Post-Cure and Environmental Conditioning Subsets

To evaluate thermal stability, a second set of specimens is subjected to defined preconditioning prior to measurement, which may include 168 hours at 85°C/85% relative humidity per J-STD-020 moisture sensitivity leveling, multiple reflow simulations at 260°C peak temperature, or extended post-mold bake at 175°C. Comparing the CTE of conditioned specimens against the as-cured baseline reveals whether moisture uptake or additional crosslinking shifts the expansion behavior and Tg of the underfill system.

Specimen Conditioning Prior to Test

Unless otherwise specified, all specimens are conditioned at 23 ± 2°C and 50 ± 5% relative humidity for at least 24 hours before testing. Specimens are then desiccated and marked with a scribe line to define the measurement axis, ensuring that successive runs track the same geometric direction and that any filler-induced anisotropy is consistently captured.

Testing Items

The evaluation experiment encompasses the following measurands, each reported with mean value, standard deviation, and specimen count:

• Coefficient of thermal expansion below Tg (α1), expressed in ppm/°C over the linear range from 25°C to a temperature at least 30°C below the anticipated Tg.
• Coefficient of thermal expansion above Tg (α2), expressed in ppm/°C over the linear range starting at least 20°C above the transition midpoint.
• Glass transition temperature (Tg) determined from the intersection of the extrapolated α1 and α2 slope segments, providing a dilatometric Tg that can be cross-checked against DSC results.
• Total dimensional change (ΔL/L0) across the full programmed temperature range, an indicator of cumulative expansion relevant to gap filling and fillet geometry.
• Measurement repeatability and specimen-to-specimen reproducibility, quantified to establish the measurement uncertainty budget.

Testing Methods

Thermomechanical Analysis (TMA) — Primary Method

Thermomechanical analysis per ASTM E831 and ISO 11359-2 is the industry-standard technique for underfill CTE evaluation. The specimen is placed on the quartz platform of the TMA furnace, and a low-mass quartz expansion probe applies a static force of 0.02 N to 0.10 N to maintain contact without deforming the cured epoxy. The furnace is purged with nitrogen or dry helium at 30 mL/min to 50 mL/min to prevent oxidative degradation, and the specimen is equilibrated at the starting temperature for 5 minutes to 10 minutes to eliminate thermal lag and probe settling drift.

Test Conditions and Parameters

A representative test program uses the following parameters:

• Temperature range: -50°C to 250°C, or a narrower window such as 25°C to 200°C when the maximum operating temperature is known.
• Heating rate: 3°C/min to 5°C/min, selected to keep specimen temperature gradients below 1°C and to resolve the Tg inflection cleanly.
• Probe force: 0.05 N nominal, with a preload hold until dimensional drift falls below 5 μm/min.
• Specimen count: minimum three specimens per material lot, with duplicate runs on each specimen to confirm stability.
• Data acquisition: dimension change sampled every 1°C, converted to instantaneous CTE via secant calculation over 10°C intervals.

Data Reduction and Calculation

The CTE at temperature T is calculated as the slope of the normalized dimensional change versus temperature curve: α = (1/L0) × (dL/dT), where L0 is the initial specimen length measured at the starting temperature. A least-squares linear regression is fitted to the pre-Tg segment and separately to the post-Tg segment; the intersection of the two extrapolated lines defines the dilatometric Tg. The first thermal ramp may be used as a conditioning run, with reported values taken from the second ramp to erase residual cure and stress history, and this protocol must be stated explicitly in the test report.

Verification and Cross-Check Methods

For high-stakes qualification programs, TMA results can be corroborated by push-rod dilatometry per ASTM E228 on larger specimens, which offers excellent accuracy at moderate expansion levels, or by laser interferometric dilatometry for low-expansion verification. Internal consistency is further checked by comparing the dilatometric Tg with the DSC Tg per ASTM E1356 and the DMA storage-modulus Tg per ASTM D4065, since a coherent dataset across three independent techniques provides high confidence in material characterization.

Measurement Uncertainty and Traceability

The laboratory verifies instrument performance using a certified reference material such as fused silica or aluminum with known expansion behavior, establishing measurement uncertainty typically below ±1 ppm/°C for α1 and below ±3 ppm/°C for α2. Thermocouple calibration, probe force verification, and furnace temperature mapping are documented, and all equipment is traceable to national metrology institute standards, satisfying the traceability expectations of ISO/IEC 17025-accredited laboratory practice.

Applications

Underfill CTE evaluation data directly support a wide range of electronics packaging and reliability engineering activities:

• Flip-chip and chip-scale package material selection, where matching underfill α1 to the effective solder joint and substrate CTE minimizes shear strain on interconnects during thermal cycling.
• Automotive and aerospace electronics qualification, in which packages must survive thousands of cycles between -40°C and +125°C per JESD22-A104 and related profile requirements.
• Corner bonding and edge-bond adhesive specification for large BGA devices in handheld and portable products, where controlled CTE preserves drop-test and bend performance.
• Advanced 2.5D/3D packaging and heterogeneous integration, where underfill CTE compatibility with through-silicon vias, interposers, and microbump arrays governs warpage control.
• Cure schedule and process optimization, since incomplete cure manifests as depressed Tg and elevated α1, enabling process engineers to validate oven profiles before mass production.
• Incoming inspection and supplier lot qualification, using statistically defined CTE acceptance windows to guard against formulation drift between resin batches.
• Failure analysis, where a measured CTE shift after field exposure or temperature cycling indicates degradation, moisture uptake, or post-cure changes contributing to observed delamination or cracking.

FAQ

What is a typical CTE target for capillary underfill adhesives? Most commercial capillary underfills specify α1 between 25 ppm/°C and 40 ppm/°C with Tg of 110°C to 150°C. No-flow and wafer-applied underfills may range slightly wider. The optimal value depends on the die size, bump pitch, substrate CTE, and the temperature range of the intended application, and it should always be selected based on simulation plus experimental verification rather than a single generic number.

Why must both α1 and α2 be measured instead of a single average CTE? The underfill operates below Tg during normal service, so α1 governs in-use stress. However, during reflow, burn-in, or high-temperature operation near or above Tg, expansion jumps to the much larger α2, which drives fillet cracking and interface delamination. Reporting both values, along with Tg, gives a complete picture of behavior across the entire assembly and test temperature envelope.

How long does a single CTE evaluation take? A typical program covering three specimens with two ramps each from -50°C to 250°C at 5°C/min, including equilibration and cooling between runs, requires approximately 8 to 12 hours of instrument time, with a full report including data reduction, uncertainty statement, and cross-technique Tg comparison delivered within three to five working days by most accredited third-party laboratories.

Can CTE be measured directly on a populated assembly instead of a bulk plaque? Bulk plaque measurement per ASTM E831 remains the reference approach because geometry and cure state are controlled. For assemblies, complementary techniques such as digital image correlation, shadow moiré warpage measurement, or finite-element model calibration using the plaque data are recommended, because extracting a clean material CTE from a constrained multi-material stack-up introduces significant systematic uncertainty.

What acceptance criteria should be written into a material specification? A robust specification defines α1 with a tolerance band (for example, 30 ± 5 ppm/°C), a minimum Tg at least 20°C above the maximum continuous operating temperature, a maximum α2 ceiling, and a maximum permissible shift in α1 and Tg after moisture preconditioning and reflow simulation. Statistical lot acceptance should be based on three or more specimens with results compared against the documented measurement uncertainty.