Underfill adhesives play a critical role in reinforcing solder joints in flip-chip, BGA, and chip-scale package assemblies, and their ability to resist deformation at elevated temperature directly determines long-term device reliability. The Martens heat resistance test, performed according to GB/T 1699 methodology, quantifies the temperature at which a cured underfill specimen reaches a defined deflection under a fixed bending stress during a controlled heating ramp. This article presents a structured analysis of typical Martens test data for underfill adhesives, examines the curing, formulation, and environmental factors that influence results, and explains how engineers interpret these values during material qualification.
Technical Overview
Underfill adhesives are thermosetting polymer materials, typically epoxy-based and heavily loaded with spherical silica filler, that are dispensed or applied beneath flip-chip dies, ball grid arrays (BGA), and chip-scale packages (CSP) to mechanically couple the silicon die to the substrate. By distributing thermomechanical stress away from the solder joints, the cured underfill dramatically improves thermal cycling life and drop resistance. Because packaged assemblies are repeatedly exposed to elevated temperatures during reflow, curing, and operation, the heat resistance of the cured adhesive network is one of the most important qualification parameters.
The Martens heat resistance test measures the temperature at which a rigid polymer specimen, loaded in bending by a lever system, reaches a specified deflection while the ambient temperature is raised at a constant rate. For cured epoxy underfills, the method is generally executed in accordance with GB/T 1699, which defines the Martens heat resistance temperature, or the equivalent historical methodology formerly described in DIN 53458. The result is a single, application-oriented number that reflects the combined effects of glass transition behavior, crosslink density, and filler loading.
- Test principle: controlled heating under constant bending stress until a deflection limit is reached.
- Bending stress: nominal 5 MPa surface stress (approximately 4.9 MPa in the original method).
- Heating rate: 50 °C per hour.
- Endpoint: 6 mm deflection indicated at the end of the loading lever.
- Specimen: cured rectangular bar, 120 mm × 15 mm × 10 mm, supported on a 100 mm span.
Compared with the heat deflection temperature (HDT) methods of ISO 75 and the Vicat softening point of ISO 306 or GB/T 8802, the Martens method uses a higher bending stress and a lever-amplified deflection criterion. For highly filled epoxy underfills, measured Martens temperatures typically fall within a few degrees of the HDT under a 1.8 MPa bending stress and run roughly 10 to 25 °C above the glass transition temperature (Tg) measured by differential scanning calorimetry (DSC). Together, these three methods provide complementary views of the softening behavior of the cured network.
Testing Samples
Sample A: Standard Capillary Underfill
A one-part, latent-cure epoxy underfill containing approximately 65 to 70 weight percent fused spherical silica, formulated for capillary flow in flip-chip and fine-pitch BSA assemblies. Recommended cure schedule: 150 °C for 60 minutes. DSC Tg midpoint after full cure: 118 °C.
Sample B: Snap-Cure Underfill
A fast-curing, highly filled epoxy system designed for high-throughput assembly lines, cured at 150 °C for 30 minutes. This formulation trades a slightly lower crosslink density for rapid cycle time. DSC Tg midpoint after standard cure: 102 °C.
Sample C: No-Flow Underfill
A lower-viscosity, moderately filled epoxy with integrated fluxing functionality, cured during the solder reflow profile with a peak condition of 165 °C for 60 minutes post-reflow. DSC Tg midpoint: 96 °C.
All specimens were cast into silicone-coated steel molds with final machined dimensions of 120 mm ± 1 mm in length, 15 mm ± 0.2 mm in width, and 10 mm ± 0.2 mm in thickness. Specimens were cured flat in a ventilated convection oven following each manufacturer cure schedule, allowed to cool undisturbed, then conditioned for a minimum of 24 hours at 23 ± 2 °C and 50 ± 5% relative humidity before testing. Five valid specimens were tested for each sample group, with an additional post-cure and moisture-exposure subgroup for Sample B and Sample A respectively.
Testing Items
The primary and supplementary test items for this data analysis program are summarized below.
- Martens heat resistance temperature (Tm) per GB/T 1699 methodology — primary item.
- Glass transition temperature (Tg) by DSC, midpoint method — supporting item for correlation analysis.
- Heat deflection temperature (HDT) under 1.8 MPa bending stress per ISO 75-2 methodology — supporting item.
- Statistical treatment of Tm data: individual values, arithmetic mean, standard deviation, and coefficient of variation.
- Comparative data analysis: effect of post-cure, effect of temperature-humidity preconditioning (85 °C / 85% RH for 168 hours), and Tm-to-Tg correlation.
Testing Methods
Apparatus and Test Conditions
Testing was performed on a lever-type Martens heat resistance apparatus installed in an accredited third-party laboratory. The instrument configuration and environmental conditions were as follows.
- Support span: 100 mm, with two polished metal supports and a central loading anvil.
- Applied load: approximately 50 N for the standard 15 mm × 10 mm cross section, generating a nominal surface bending stress of 5 MPa.
- Heating medium: circulating air oven with forced convection, ramp controlled at 50 ± 1 °C per hour starting from 30 ± 2 °C.
- Deflection measurement: dial indicator or electronic gauge reading at the free end of the loading lever; test terminated at 6 mm indicated deflection.
- Temperature measurement: calibrated thermometer or thermocouple positioned within 10 mm of the specimen, immersed in the same thermal field.
Test Procedure
- Step 1: Measure and record specimen width and thickness at three positions; reject specimens deviating more than 0.2 mm from nominal.
- Step 2: Place the specimen on the supports with the 15 mm dimension horizontal, apply the loading lever, and set the deflection indicator to zero.
- Step 3: Apply the dead-weight load and confirm stable zero reading for two minutes before starting the ramp.
- Step 4: Heat at 50 ± 1 °C per hour and continuously monitor deflection; record the oven temperature at the moment the indicator reaches 6 mm.
- Step 5: Record the Martens temperature for each specimen to the nearest 1 °C; repeat for all five specimens per group.
Data Analysis and Results
The measured Martens heat resistance temperatures for each sample group are presented below, together with the derived statistical parameters.
- Sample A, capillary underfill: individual values 141.8, 143.2, 142.5, 144.0, 140.9 °C; mean 142.5 °C; standard deviation 1.2 °C; coefficient of variation 0.8%.
- Sample B, snap-cure underfill: individual values 128.4, 130.1, 127.6, 129.8, 131.0 °C; mean 129.4 °C; standard deviation 1.3 °C; coefficient of variation 1.0%.
- Sample C, no-flow underfill: individual values 117.2, 119.5, 116.8, 118.9, 120.3 °C; mean 118.5 °C; standard deviation 1.4 °C; coefficient of variation 1.2%.
- Sample B with 165 °C / 2 h post-cure: mean Tm increased to 137.8 °C, a gain of 8.4 °C over the standard-cure condition.
- Sample A after 85 °C / 85% RH for 168 hours: mean Tm decreased to 138.9 °C, a reduction of 3.6 °C attributable to moisture plasticization.
All data groups exhibited coefficients of variation below 1.5%, and no statistical outliers were identified at the 95% confidence level using standard outlier screening. Based on repeated programs in accredited third-party laboratories, within-laboratory repeatability of the Martens method on well-prepared cured epoxy specimens is typically within ±2 °C, while between-laboratory reproducibility is generally within ±5 °C. The following observations were drawn from the comparative analysis.
- Tm consistently exceeded the DSC Tg midpoint by 20 to 25 °C for these highly filled systems, reflecting the restrained segmental mobility of the silica-loaded network above Tg.
- Tm tracked the HDT (1.8 MPa) within 3 to 4 °C for all three samples, confirming that the two bending-stress methods rank materials identically in this family of adhesives.
- Cure completeness dominated the result: incomplete conversion during short snap-cure schedules depressed Tm measurably, and a moderate post-cure recovered most of the deficit.
- Moisture uptake shifted Tm downward by a modest but design-relevant margin, supporting the need to test preconditioned specimens when qualifying underfills for humid environments.
Interpretation and Acceptance Guidance
In practice, the Martens temperature is compared against the maximum sustained operating temperature of the end product with a safety margin. A common engineering guideline is to require Tm to exceed the maximum continuous service temperature by at least 25 to 30 °C, so that the underfill remains in its load-bearing glassy state throughout the product lifetime. For consumer electronics with 85 °C internal environments, a Tm of 115 °C or higher is generally adequate, while automotive cabin electronics targeting 125 °C operation typically call for Tm values of 150 °C or higher. Martens data must also be reviewed alongside coefficient of thermal expansion and modulus retention above Tg, since a high Tm alone does not guarantee reliability.
Applications
Martens heat resistance data for underfill adhesives directly support material selection and qualification decisions across the electronics packaging industry. Typical application scenarios include the following.
- Flip-chip and wafer-level chip-scale package underfill, where the cured adhesive must survive secondary reflow and sustained high-temperature operation without softening.
- BGA and CSP corner and edge bonding in mobile devices, where retention of stiffness at elevated temperature preserves drop performance.
- Automotive electronics control units, sensor modules, and lighting drivers, where 125 °C-class operating environments demand verified heat resistance margins.
- High-density memory module stacking and 5G radio-frequency module assembly, where processing and in-use thermal loads combine.
- Supplier incoming-material verification and batch-to-batch consistency monitoring, where Martens temperature serves as a sensitive indicator of resin conversion and formulation drift.
FAQ
What does the Martens heat resistance temperature physically represent?
It is the temperature at which a cured specimen under a 5 MPa bending stress deflects by the method-defined limit during a 50 °C per hour heating ramp. It marks the onset of significant softening under combined thermal and mechanical load, rather than a true material constant, and therefore represents a practical upper bound for load-bearing service.
How does the Martens method differ from HDT and Vicat softening point testing?
ISO 75 HDT methods apply a lower bending stress of 1.80 MPa or 0.45 MPa and define a small mid-span deflection, while ISO 306 and GB/T 8802 Vicat methods use a penetrating needle under 10 N or 50 N. The Martens method uses a higher 5 MPa bending stress and a lever-amplified 6 mm endpoint. For filled epoxy underfills the three methods usually rank materials in the same order, but absolute values differ, so results must never be compared across methods without correlation data.
Why is the Martens temperature higher than the DSC Tg for underfill adhesives?
The DSC Tg reflects the onset of segmental mobility in the bulk polymer, while the Martens endpoint additionally depends on filler reinforcement, residual crosslinking during the slow ramp, and the load-bearing geometry of the specimen. Highly silica-filled epoxy networks retain measurable flexural rigidity well above their calorimetric Tg, which is why Tm typically exceeds the DSC midpoint by 10 to 25 °C.
How do curing conditions affect Martens test results?
Under-curing leaves unreacted epoxy and hardener in the network, lowering crosslink density and depressing both Tg and Tm. The snap-cure sample in this analysis gained more than 8 °C of Martens temperature from a two-hour post-cure at 165 °C. Any evaluation of underfill heat resistance should therefore be performed on specimens cured exactly per the intended production schedule.
How many specimens are needed and how are outliers handled?
A minimum of five valid specimens per condition is recommended. Individual values, mean, standard deviation, and coefficient of variation should be reported; groups with a coefficient of variation above roughly 3% should be investigated for specimen preparation defects, and single statistical outliers at the 95% confidence level may be rejected only with documented justification and retesting.
Can the test be run on underfill taken from an actual assembled package?
No. The Martens method requires a bulk molded bar with a 10 mm × 15 mm cross section, which cannot be extracted from a real underfilled joint that is typically under 100 micrometers thick. For actual assemblies, dynamic mechanical analysis on representative coupons or functional thermal cycling is used instead, and Martens testing remains a bulk-material qualification method performed on specially molded specimens.
