Electrochemical capacitance based method applied to epoxy molded devices

Sensor devices combine a micro-mechanical structure (MEMS) with an electronic Application Specific Integrated Circuit (ASIC) chip for inertial application, pressure variation measurement, acoustic waves' detection, etc. to detect movements, pressure differences etc.

These Systems in Package (SiP) use mainly a full molded package solution for both consumer and automotive fields of application. Actually, epoxy molding compounds (EMCs) in plastic encapsulated microcircuits (PEMs) offer many advantages such as lower cost, lighter weight, and comparable performances over hermetic packages, despite of major moisture absorption after humidity exposure.

Moisture diffusion is one of the reliability concerns reported in literature [1,2,3,4] and many failure modes observed in these devices are ascribed to this phenomenon triggered by manufacturing, storage, or operation steps. Indeed, during the plastic package molding process, the silica fillers distribution in the polymer matrix is not uniform around the dies and the connections and it is influenced by their shapes.

For this purpose the complex geometry of the sensor device pads has been explored with the help of a high resolution X-Rays computed tomography tool.

This article describes how the real resin morphology influences the electrical signals provided by a MEMS to an ASIC in an inertial sensor SiP, focusing on the pads area of the MEMS, where a large surface of polysilicon is in direct contact with the resin.

The measurement method that quantifies and characterizes the local change of the resin properties is deployed starting from the standard available instruments following a simple R-C-R circuit to model the electrical behavior.

Some devices have been submitted to a thermal humidity stress (THS) cycle (85°C 85%RH 24hrs) after a baking session (125°C 24hrs) in order to evaluate the different moisture conditions.

The pure R-C-R model seems to give incomplete information because it is not considering any chemical aspect of the real phenomenon. On the other hand, specifications of the commercial epoxy compound list a percentage of ions that can contribute to an electrolytic solution in case of water presence. Therefore the electrical characterization must include a method that better fits with the reality; in particular the so-called Cyclic Stair Case Voltammetry (CSVC) technique allows the characterization of the moisture chemical behavior by considering the Randles cell model [5] as the appropriate equivalent electric circuit. The electrochemical measurements have been then extensively used to characterize the electrical behavior of the epoxy molded devices with some resin morphological irregularities, after submitting them to a humidity cycle.

The samples have been selected among a population of devices with an evident sensitivity to a damp environment by comparing the testing electrical values to the standard specification and to the initial dry condition.

To guarantee a small modification of the package a limited portion of the resin compound between the MEMS sensor and the ASIC has then been removed to allow a direct probing of the signals after cutting the devices connections (Fig. 1).

Fig. 1 shows the material and the devices involved in the electrical investigation. The ASIC die is an integrated circuit with a passivation layer that acts as an insulator and no significant parameters variations are expected from this side even if the resin is soaked. On the contrary, in the MEMS pads area a large polysilicon surface is in direct contact with the epoxy compound because of the specific architectural choice of these device.

Therefore the instruments setup has been prepared to run the measurements on the MEMS side (Fig. 2a) to extract the model parameters of a simple R-C-R circuit associated to each couple of nodes (Fig. 2b).

Of course the MEMS device adds to the network under investigation capacitances and resistances that must be isolated from the circuit of interest and afterward another cut has been conveniently done.

As the investigation concerns a small volume of material, the intrinsic order of magnitude of the electrical parameters is very low. In consequence the instruments setup has been built to match with the phenomenon under study in terms of current/voltage range and frequency. A typical commercial current amplifier with a current resolution of the order of few pA is suitable for evaluating the resistance value, according to the standard specifications of the epoxy compound. The DC measure across the nodes confirms that the resin resistance R_p value is in agreement with the official manufacturer document and it does not change in the presence of moisture absorption.

On the other hand the capacitances evaluation requires a more sophisticated setup. For this purpose, a trans-admittance network analyzer has been firstly preferred to qualitatively characterize electrical signals locally activated in the resin. The output signals (i_out) give the profile of a transfer function (i_out vs v_in) whose zero and pole are the base of the network study (Fig. 3a and 3b).

C_P is the intrinsic pad to pad capacitance that would be expected to be stable after submitting the sample to the humidity cycle.

Fig. 4 shows the experimental setup. A suitable high frequency current amplifier [9] has been connected to the signal analyzer [10] to evaluate the admittance function. The instrument applies a sinusoidal voltage stimulus of fixed amplitude by sweeping the frequency over a given range. In Fig. 5 the impedance behavior versus frequency has been reported before and after a THS cycle for one of the analyzed parts.

The first admittance function zero is dominated by R_P because R_P » R_S, but it cannot be used to calculate the parameter C_S with a sufficient precision grade because of the high measurements noise level.

The study of the pole frequency f_p=1/2π*C_S*R_S, placed in a range higher than 100kHz, is the right estimation way to calculate the parasitic network parameters.

An experimental board has been identified and added to the setup system to execute the high frequency measurements until 10MHz and the most relevant results have been reported in Fig. 6 and Fig. 7.

Fig. 6 displays the outcome of the measurements done on nodes whose electrical behavior is in line with the datasheet specifications, without significant changes after THS.

Fig. 7 shows the measurements results of electrodes causing the external parameters drift; these pads are affected by a visible humidity related behavior, although parameters estimation is not possible due to the poor equipment accuracy in the frequency range under evaluation.

A different approach is necessary to characterize the local moisture electrical behavior of the resin compound.

A measurements method based on the physical and chemical effect of the material modification after moisture absorption has been identified to completely model the phenomenon. By the way, the specifications of the commercial epoxy compound list a percentage of water extracted ions as Na⁺ and Cl⁻ that can activate electrochemical phenomena. Therefore the electrical characterization must include a method to better fit with the reality: the Cyclic Stair Case Voltammetry (CSCV) has been identified as the most appropriate technique to investigate the chemical phenomena from an electrical perspective [6, 7].

In fact, the physical-electrical system of the resin between each MEMS pads couple, in case of a differential voltage applied to the nodes, behaves similar to an electrolytic solution with water and free ions.

The above cited method adopted a regular potential steps scanning (Fig. 8).

The current between the nodes has been sampled just before the subsequent step. The voltage ramp slope, positive and then negative, resulting from the time duration and the amplitude of each step averaged along the measure period, has been selected in order to reduce the intrinsic capacitive contribution and the potential faradaic one.

The samples have been submitted to a first THS cycle and following bake. The typical diagrams types obtained from the current versus voltage measurements among the different nodes couples are reported in the following figures (9 to 14).

A second THS and bake cycle has been applied to the samples population too. The measurements graphs collected from this further investigation step may be represented by the typical diagrams reported in Fig. 15 and Fig. 16.

The epoxy material between the electrodes pairs of the Fig. 9–10 diagrams present an electrical characteristic coherent with the typical specifications of such compound that does not change its resistance like profile even if submitted to a moisture cycle.

To confirm the intrinsic behavior of the resin, an experiment has been run on a bulk formed from the same material of the samples, following the below setup design.

The samples of resin were placed between two copper plates as electrodes (Fig. 17); the voltage was applied measuring the flowing current.

Of the available bare materials, one was maintained in boiling water for 9 hours and another was maintained in a boiling water solution with sodium chloride (45g/l) for 9 hours.

Fig. 18 shows the electrical results.

According to the straight line, the epoxy behaves as a resistance in a pure water environment, while in an electrolytic solution type the obtained results are similar to those reported in the previous Fig. 11, from the devices investigation.

It looks like the resin material may locally presents electrical characteristics different from the typical bare material, anyhow similar to an electrochemical solution. In fact many scientific articles [5, 6, and 7] explain the electrochemical phenomena showing hysteresis loops diagrams by means of the double layer capacitance effect. An electrical double layer exists on the interface between an electrode and its surrounding electrolyte. This double layer is formed as ions from the solution adsorb onto the electrode surface. Separation of charges at the electrode forms a capacitor and can generate capacitive contribution to the circuit (Fig. 19).

The simplest model of the above described electrochemical interface [8] is the Randles circuit (Fig. 20 and Fig. 21).

The first order circuit model represented in Fig. 21 is the schematic network used to characterize the resin local properties following the results of the executed measurements. The C_dl parameter value can be evaluated by calculating the area subtended by the curves of Fig. 11, as highlighted in the Fig. 22.

The used expression is:

Where v=dv/dt is the potential scan rate and E1 and E2 are the cutoff potentials in cyclic voltammetry, i (E) is the instantaneous current, (E2 - E2) is the potential window width.

To check the resin morphology, some devices samples have been prepared to extract pads volume and submit them to High resolution X-Rays Computed Tomography (CT).

The results of the virtual cross sections across the X–Y plane, obtained by slicing along the Z axis, are summarized in Fig. 25.

Some Focused Ions Beam (FIB) cross sections have been run on the circled regions to identify the resin structure (Fig. 26).

The picture reported in Fig. 27 confirms the lack of homogeneity in the fillers distribution around the complex geometry of the sensor MEMS pads and the presence of micro-voids, ascribed to different injection conditions of molding material during assembling phase. Several examples of such defects have been found in the study and can be explained as a consequence of a shadow effect: fillers having dimensions similar to the distance between the pads block the smallest fillers and forbid complete molding. These defects increase the moisture absorption capacity of the resin and lead to higher humidity sensitivity. The electrical properties may be conditioned in the same way.

The Randles circuit and its capacitance part may be explained as the identified filler voids that act as an electrochemical solution (H₂O + ions) after the humidity exposure.

The electrochemical measurements method described in this work has been proven to be a powerful technique to accurately characterize the epoxy material properties, especially when this kind of resin is used in complex commercial devices SiP affected by a potential humidity exposure.

The Randles double layer capacitance effect study has been applied to a small portion of few suitable samples particularly sensitive to the moisture absorption by showing external electrical performances not aligned to the expected behavior. The identified equivalent circuit paves the way for a standard design tool and method that takes into account such sophisticated phenomenon.

The authors would like to thank Dr. Fabrizio Speroni, Dr. Fabio Biganzoli and Dr. Elio Calì for their continuous support and valuable hints to pursue the results and to correctly plan the several measurements sessions of the work.