Introduction

Implantable electronic devices (IEDs) including brain stimulations, pacemakers, and retinal implants have created the opportunity to improve diagnostic and therapeutic procedures in medical practice1,2,3,4,5. Despite the remarkable progress, long-term safe operations of IEDs are still, to a great extent, prohibited by potential corrosion threats due to the body fluid-covered service environment and the resulted water absorption by encapsulation material6,7, which lead to inner printed circuit boards (PCBs) corrosion and consequent catastrophic IED failures from both economy and user safety viewpoints. Besides, the high electrical field between two electrodes on a PCB induced by high integration and miniaturization of IEDs could further deteriorate its corrosion reliability8. The reliability of PCBs should be seriously valued to ensure a long-life IED since PCBs contain a series of components (i.e., capacitors, resistors, inductors) made up of different metals and alloys, which are corrosion susceptible under humid and voltage-biased environments. Therefore, extensive research9,10,11,12,13,14,15,16,17,18,19 has been conducted in an attempt to uncover the underlying mechanism of corrosion in electronic devices, which could inspire the design of IEDs in a protective manner for corrosion reliability control.

Dendrite formation induced by electrochemical migration (ECM) is a highly concerned corrosion problem published in the open literature10,19 that could considerably compromise the robustness of electronic devices and thus potentially lead to unexpected electronic failure. More specifically, dendrite formation and growth generally take place between two closely placed electrodes on a PCB surface in combination with electrolyte and bias potential9,10, where the dissolved metallic ions resulted from anodic reactions will migrate towards the cathode under the driving force of the electrical field and the concentration gradient and are subsequently redeposited as dendrites from the cathode to anode11,12. Once the dendrite bridges the two electrodes, the current will undergo a remarkable rise due to short circuit and then cause subsequent malfunction of the electronic devices. Most research are conducted under climate environment nowadays and although ECM has been proven to be a multi-parameter effect including the type and concentration of contaminations, humidity and temperature, bias form, type of alloy involved, etc., further studies on ECM behavior under other different environments still need to be conducted since current results may not be universal as ECM behavior is highly dependent on the migration environment13,14,15,16,17,18,19,20,21. Further, most researchers perform their studies employing alloy materials instead of an actual component on PCB, meanwhile relatively little attention has been paid to how ECM causing malfunction of electronic devices. In fact, dendrites and corrosion products formed during ECM process will significantly influence the component robustness on PCBs, which cause electrical functional issues of the devices. Therefore, it is important to conduct ECM investigations in the environment at which particulate devices are used, hence there is chance for PCB surface to get exposed to such conditions.

In this work, droplet tests are employed to simulate IEDs working under human implantation environment, where aqueous conditions will inevitably contact with the component on a PCB surface as a result of water diffusion through encapsulation materials of IEDs. Since passive components such as capacitors are problematic from the point of view of corrosion effects such as ECM, different size capacitors on a populated PCB are used as test components in the present work. Additionally, bias voltage is also varied in amplitude and form to understand the effect on ECM formation. Finite element method (FEM) analyses are preliminarily conducted to determine the exact voltage waveform used in further electrochemical tests, followed by further chronoamperometry (CA) tests and in-situ observations with the purpose to study the sensitivity of dendrite formation among different size capacitors under different potential amplitudes. Subsequently, the morphologies and compositions of the dendrites together with corrosion products are identified by means of SEM, TEM and XPS characterizations. Based on above results, the exact mechanism of dendrite formation is totally uncovered. Moreover, EIS technique is employed to assess the influence of ECM on capacitor reliability under different corrosion states, which also provides in-depth mechanism explanations for electronic corrosion.

Results and discussion

Feasibility verification by FEM

There are various voltage waveforms used in a real IED including the DC wave, square wave, the sawtooth wave and the sinusoidal wave22,23,24,25,26. To preliminarily determine the exact potential waveform used in further electrochemical tests, finite element analyses are implemented to simulate different voltage patterns originated from actual electronic devices, from which the electrode shape change and tin ions flux distribution are acquired as shown in Fig. 1. It is clearly seen from Fig. 1 that various voltage patterns could definitely induce shape change of the electrode in y direction as a consequence of dendrite formation, though the degree of such change varies slightly. More specifically, compared to the electrode under the square wave, sawtooth wave and sine wave which exhibits a much slighter shape change of 0.04 mm, 0.04 mm and 0.03 mm, respectively, the electrode under a 10 V—DC voltage manifests the most noteworthy electrode shape change of 0.07 mm (from 0.08 mm to 0.15 mm) in y direction within the simulation period, indicating an accelerated dendrite growth under the DC voltage. Besides, it is shown that tin ions migration driven by the concentration gradient and electric field is also the most pronounced under DC voltage, which could definitely account for the corresponding fastest dendrite growth. Furthermore, it is speculated here that such a difference in electrode shape change and tin ions migration is attributed to the integral values of different waveforms, which closely associates with the duty cycle and load level of a specific voltage pattern. Finally, the above results reveal that employing a DC voltage pattern is definitely feasible for further investigating the effect of voltage amplitude on ECM behavior as the DC voltage pattern is an accelerated test compared to other voltage waveforms. What’s more, it is also indicated that the triangle waveforms and square waveforms should be the preferred waveforms to potentially improve corrosion robustness for future IEDs as dendrites grow the slowest under these two waveforms.

Fig. 1: Electrode shape changes of the capacitor and tin ions flux distribution at t = 50 s under different voltage patterns.
figure 1

(a) DC voltage, (b) square wave, (c) sawtooth wave, (d) sine wave.

Full size image

ECM behavior of capacitors under different conditions

To exclude the influence of different material compositions on ECM behavior, the capacitors are first characterized using SEM with the surface and cross-section images of LC shown in Fig. 2 (The results of MC and SC are shown in Supplementary Figs 1 and 2). Figure 2a illustrates that the electrode material of the capacitor mainly consists of Sn, while the dielectric material between the two electrodes contains mostly Ba and Ti. Apart from that, the cross-sectional morphology of the electrode shown in Fig. 2b exhibits a three-layer structure, which contains Sn as the first layer, Ni as the middle layer and Cu as the third layer. Besides, both MC and SC present the same structure and material composition as LC in terms of surface and cross-section morphology (details see SI), thus excluding ECM sensitivity induced by different metal species. Notably, LC possesses the largest surface area among all the capacitors along with a relatively thick Sn layer (1.7 μm), both clearly indicating that LC endows the potential and capability to dissolve more tin ions in subsequent electrochemical tests.

Fig. 2: SEM images showing the surface and the cross-sectional morphologies of the large capacitor together with EDS mapping results.
figure 2

a SEM images and EDS mappings of the original large capacitor surface. b The cross-sectional morphologies and EDS mappings of the large capacitor electrode prepared by FIB.

Full size image

To verify the effect of bias voltage amplitude and pitch dimension on ECM behavior of capacitors, CA tests with different potential amplitude (3 V and 10 V) under DC voltage are conducted on different size capacitors and the results are shown in Fig. 3. Besides, to monitor the formation of corrosion products and dendrites during the test, in situ observation of ECM process using a Dino-Lite digital microscope is also carried out. For LC under 3 V voltage, it is clearly seen from Fig. 3a that some bubbles and white corrosion products occur firstly on the capacitor surface (point i), followed by the formation of dendrites bridging the two electrodes at 671 s (point ii). The current increases by 3 orders of magnitude induced by ions dissolution from the anode and electrolysis of water (both at anode and cathode due to potential bias) on the capacitor surface compared to the capacitor without droplet (about 10−6 mA, see Supplementary Fig. 3), while the current exhibits a much more remarkable rise resulted from the connection between the cathode and anode by the dendrite, almost by 6 orders of magnitude. Since the dendrite will burn off due to surge in current as a result of short circuit, the current falls back to 10−3 mA when the dendrite disconnects between the anode and cathode, thus giving a step-like current shape in Fig. 3a. Finally, as the droplet on the capacitor surface gradually evaporates, the current decreases accordingly to the value similar to the original capacitor before exposure to the droplet. By contrast, only corrosion products occur for MC and SC under 3 V, and no dendrite forms until the droplet has evaporated completely. In addition, for capacitors under 10 V, Fig. 3d–f illustrates that dendrites and corrosion products form on the surface of LC and SC with similar current variation as LC is under 3 V. However, it is worth mentioning that dendrite forms much faster under 10 V compared to capacitors under 3 V, implying a possibly more rapid ionic transport under a higher electric field. More specifically, the two electrodes are linked by the dendrite at 38 s and 10 s for LC and SC under 10 V, respectively, which is a quite shorter time in comparison with LC under 3 V (671 s). What’s more, the whole electronic device will definitely undergo some malfunction once the capacitor is shorted by dendrites, possibly leading to undesirable consequences.

Fig. 3: ECM behavior together with morphology changes of capacitors with different pitch dimensions under different bias potential.
figure 3

(a) LC 3 V, (b) MC 3 V, (c) SC 3 V, (d) LC 10 V, (e) MC 10 V, (f) SC 10 V.

Full size image

To gain more insights into the morphology and composition of dendrites and corrosion products formed under different conditions, SEM, TEM and XPS characterizations are further performed. Figure 4a–c display that dendrites form on LC under 3 V, LC under 10 V and SC under 10 V all exhibiting a branch feature with the composition to be metallic tin as confirmed by TEM diffraction pattern and EDS (Supplementary Table 1), meanwhile the dendrites seems to be surrounded by corrosion products, which is also supported by SEM images shown in Supplementary Fig. 4. As for corrosion products observed in the same condition, TEM diffraction patterns firstly prove an amorphous structure, while further XPS analyses reveal that the corrosion products are a mixture of SnO and SnO2 with different proportions. More specifically, Fig. 4g–i shows that corrosion products formed under 10 V possess less SnO2 compared to those under 3 V, which could be possibly attributed to the faster ion migration speed toward the cathode as a result of the higher bias potential. This phenomenon is also confirmed by CA tests given in Fig. 3d, f, where dendrites form almost instantaneously as the 10 V potential is applied on the capacitor, again indicating a more rapid ion transport speed. Apart from that, Sn4+ that reaches the cathode will subsequently change to [Sn(OH)6]2- due to the relatively faster stabilized alkaline area near the cathode for capacitors under 10 V (see Supplementary Fig. 5)10, therefore resulting in a depletion of Sn4+ and the resulted smaller amount of SnO2.

Fig. 4: Morphology and Composition characterizations of dendrites and corrosion products formed under different conditions using TEM and XPS.
figure 4

a, d TEM characterizations of dendrites and corrosion products formed on LC—3 V. g XPS characterizations of corrosion products formed on LC—3 V. b, e TEM characterizations of dendrites and corrosion products formed on LC—10 V. h XPS characterizations of corrosion products formed on LC—10 V. c, f TEM characterizations of dendrites and corrosion products formed on SC—10 V. (i) XPS characterizations of corrosion products formed on SC—10 V.

Full size image

Mechanism of ECM behavior on different capacitors

The above analyses have clearly illustrated that LC under 3 V, LC under 10 V and SC under 10 V present the strongest susceptibility to ECM behavior, thus leading to potential dendrite formation and subsequent electronic failure. To further uncover the underlying mechanism associated with ECM susceptibility and the resulted dendrite formation under different conditions, Fig. 5a is given below with a detailed description of the whole ECM process. In general, Sn2+ and Sn4+ will evolve near the anode due to anodic reactions which will subsequently migrate toward the cathode driven by the electric field, while some bubbles together with dendrites will appear around the cathode due to the reduction of H2O and metallic ions, respectively. To be specific, the possible anode and cathode reactions are listed below27,28,29,30,31,32:

Fig. 5: A comparison between the universal rule of dendrite formation and our results.
figure 5

a Schematic illustration of ECM behavior of capacitors. b Results of electric field and probability of dendrite formation for capacitors under different bias potentials.

Full size image

Main anodic reactions lead to tin dissolution:

$${Sn}\to {{Sn}}^{2+}+-2{e}^{-}$$
(1)
$${{Sn}}^{2+}\to {{Sn}}^{4+}+-2{e}^{-}$$
(2)

Main cathodic reactions lead to tin deposition:

$${{Sn}}^{2+}+2{e}^{-}\to {Sn}$$
(3)
$${{Sn}}^{4+}+4{e}^{-}\to {Sn}$$
(4)
$${{Sn}\left({OH}\right)}_{4}+2{{OH}}^{-}\to {[{{Sn}({OH})}_{6}]}^{2-}$$
(5)
$${[{{Sn}({OH})}_{6}]}^{2-}+4{e}^{-}\to {Sn}+6{{OH}}^{-}$$
(6)

(Note that the occurrence of Eqs. (5) and (6) depends on the pH change near the cathode due to the water hydrolysis process).

Apart from that, metallic ions will combine with OH- originated from water electrolysis and form tin hydroxides in the process of migration to the cathode, where the metallic ions could be partially or totally consumed, depending on the total amount of dissolved ions. The formation of tin hydroxides could be described as follows33:

$${{Sn}}^{2+}+2{{OH}}^{-}={{Sn}\left({OH}\right)}_{2}$$
(7)
$${{Sn}}^{4+}+4{{OH}}^{-}={{Sn}\left({OH}\right)}_{4}$$
(8)

In addition, most Sn(OH)2 and Sn(OH)4 will further dehydrate as a result of droplet evaporation and change to more stable tin oxides as demonstrated by XPS results shown in Fig. 4g–i, which can be expressed by the following reactions34:

$${{Sn}\,\left({OH}\right)}_{2}={SnO}+{H}_{2}O$$
(9)
$${{Sn}\,\left({OH}\right)}_{4}={{SnO}}_{2}+{2H}_{2}O$$
(10)

However, not all capacitors follow the universal rule as aforementioned since only certain occasions (LC under 3 V, LC under 10 V and SC under 10 V) exhibit the most sensitivity of dendrite formation, as shown in Fig. 5b. To further explain this, the electric filed which mainly drives ion diffusion is calculated for different capacitors under 3 V and 10 V bias potential (see Fig. 5b and SI), where all capacitors under 3 V posses a much smaller electric filed than those under 10 V and thus implying a much slower ion transport speed. For all capacitors under 3 V, corrosion products initially form on the capacitor surface due to relatively slow ion transport to the cathode, where ions will combine with OH− to first form corrosion products and in turn lead to a certain degree of ion depletion in the process of migration. Despite this, dendrite still forms subsequently on the surface of LC—3 V which could certainly be attributed to the sufficient amount of the dissolved metallic ions originated from the large electrode area as confirmed by Fig. 2. Meanwhile, the sluggish kinetic of subsequent dendrite growth on LC—3 V again proves a slow ion transport speed and thus demonstrating that the ECM process is dominated by the amount of ions in the droplet under 3 V. Apart from that, the pH gradient related to the capacitor size is also responsible for dendrite formation under 3 V. Compared to a more moderate pH gradient formed on LC surface, the tin ions on SC surface undergo a much sharper pH increase when migrating to cathode for the same distance due to its smaller size (see pH distribution shown in Fig. 5a. Consequently, tin ions on SC possess the priority to form hydroxides as less amount of tin ions are required to form tin hydroxides due to a higher pH value, while tin ions on LC will reach the cathode and then be reduced to dendrites (details see SI). Furthermore, compared with capacitors under 3 V, dendrites form on the surface of LC and SC under 10 V with a much more rapid speed (Fig. 3d–f), which implies a much faster ion migration to the cathode due to the higher electric field (see Fig. 5b). This phenomenon, together with the subsequent formation of corrosion products, indicates that the dendrite formation process is controlled by rapid ion migration speed under 10 V. Consequently, the amount of dissolved metallic ions originated from large electrode areas, along with the rapid ion transport induced by a high electric field, becomes the dominant factor of dendrite formation under low and high bias potential, respectively. Besides, it is speculated that neither the amount of dissolved ions nor the ion migration speed is large enough for MC thus no dendrite forms under all conditions. Moreover, it is deduced that dendrite formation and the resulting electronic malfunctions could also potentially be avoided thanks to the good coordination between the electrode area and the strength of the electrical field, which is quite beneficial for future design of high-reliable IEDs. However, detailed studies should be performed in the future to further prove this.

Evaluation of capacitor reliability

From the above CA analyses, it is demonstrated that the leakage current soars during the ECM process as a result of anodic corrosion, which would definitely exert an adverse effect on the capacitor reliability. Generally, four statuses could be recognized for all capacitors throughout the ECM process, which can be described as the original capacitor, capacitor with droplet, capacitor with oxides (corrosion products), and capacitor with dendrites (Fig. 6a). To further verify the effect of capacitor corrosion and the corresponding different capacitor statuses on capacitor reliability, EIS measurements of capacitors with different corrosion states are carried out and the obtained data are analyzed using the equivalent circuit models as given in Supplementary Fig. 6. (The corresponding Bode plots are presented in Supplementary Figs. 7–9). It is obvious from Fig. 6b that Nyquist plot of the original LC presents a straight-line characteristic, which is perpendicular to the x-axis and thus only has the imaginary part of the impedance. Under this circumstance, LC behaves as an open-circuit element, therefore giving a low current value as demonstrated by Supplementary Fig. 3. Compared with the original capacitor, a capacitive loop emerges in the Nyquist plot of both LC with droplet and corrosion products, indicating that the whole system begins to present the combined characteristic of resistance and capacitance. Subsequently, the capacitive loop changes to a semicircular shape due to the bridged electrodes by the dendrite, which implies a much more obvious feature of resistance. This phenomenon could be ascribed to the fact that the droplet and dendrites can be regarded as a resistor in parallel with the original capacitor, which has the priority to transport electrons compared to the original LC, thus accounting for the current variation in CA tests (Fig. 3). Apart from that, the fitting results given in Fig. 6c and Supplementary Table 2 illustrate that LC value remains constant during different corrosion states, while the resistance value sees a gradual decrease with the dendrite having the minimum value. This is also in accordance with the results shown in Fig. 3, where the current accordingly reaches the maximum value and experiences a sharp rise due to the formation of dendrites between the cathode and anode. Specifically, the current basically flows through the droplet before dendrites form thus the increase in ionic conductivity (details see SI) resulting from more ions dissolution together with water hydrolysis could contribute to the current rise for capacitor with droplet and capacitor with oxides, while the current mainly flows through the metallic dendrite once the dendrite forms therefore the leakage current soars due to the much smaller resistance of the dendrite. Apart from that, similar variations could also be observed in the Nyquist plot of MC and SC under different corrosion states, again suggesting the influence of capacitor corrosion on the capacitor reliability. Notably, it is apparent from the fitting results shown in Fig. 6g that SC value drops by 1 order of magnitude when dendrite bridges the cathode and anode, implying the possible severe capacitor damage due to anodic corrosion. As a result, EIS technique is proved to be an effective method to evaluate the capacitor reliability under a corrosive environment and provide some insights into mechanism analyses of electronic corrosion.

Fig. 6: Capacitor reliability under different corrosion states.
figure 6

a Different corrosion status on the capacitor surface during CA tests. b, c EIS diagrams on LC and corresponding fitting results. d, e EIS diagrams on MC and corresponding fitting results. f, g EIS diagrams on SC and corresponding fitting results (scatters: test data; lines: fitting data).

Full size image

From the above analyses, it is obvious that a resistor will be connected in parallel with the original capacitor (LC, MC, and SC) due to the emergence of the droplet, corrosion products, and dendrites, which will in turn lead to an obvious current rise of the capacitor. However, such an abnormal current rise caused by capacitor corrosion will definitely induce a series of unexpected consequences including malfunction or destruction of the entire electronic devices and even catching fire. Considering that capacitor corrosion could be revealed by different EIS diagrams, the EIS technique could be employed as a quick detection method of electronic corrosion and therefore avoid potential destructions accordingly.

In summary, the impacts of bias potential amplitude and pitch dimension on the ECM behavior of capacitors used in electronic devices are systematically investigated. From preliminary FEM analyses, an applied DC voltage pattern has been proved to be an accelerated test compared to other waveforms, thus is feasible to be employed for further electrochemical studies. Based on this, further CA tests along with in-situ morphology observation clearly show that LC under 3 V, LC under 10 V and SC under 10 V present the strongest sensitivity of dendrite formation among all the tested capacitors. By SEM, TEM, and XPS analyses, dendrites formed on the capacitors are proved to be metallic tin, while the corrosion products are demonstrated to be a mixture of SnO and SnO2. In addition, the in-depth mechanism of dendrite formation is also clearly explained as follows: the pitch dimension is the dominating factor of dendrite formation under low bias potential, while the ion migration speed also becomes the decisive reason affecting dendrite formation under high bias potential. Most importantly, EIS method is employed to assess the capacitor reliability of various corrosion states, which is proved to be an effective technique to detect electronic corrosion and offer some great insights into corrosion mechanisms. Our work not only opens a pathway to a more fundamental understanding of capacitor ECM behavior, but also offers some deeper insights into developing robustness of electronic corrosion by rational electronic design.

Methods

Materials

A populated PCB with three different sizes of capacitors is used in the present work. The dimensions of the different capacitors are 2.0 mm × 1.3 mm, 1.5 mm × 0.8 mm and 1.0 mm × 0.5 mm and thus named as LC, MC and SC accordingly, as presented in Fig. 7. Prior to all tests, the test PCBs are cleaned with isopropanol alcohol and then dried under compressing air.

Fig. 7: Capacitors with different pitch dimensions used in this work.
figure 7

(a) large capacitor, (b) medium capacitor, (c) small capacitor.

Full size image

The selected contamination of the capacitors in this work is 1.56 μg cm−2 NaCl and a droplet of about 2 μL electrolyte is added on the surface of the capacitor using a pipette. Three solutions containing different content of NaCl are prepared using analytical grade chemical and de-ionized water with a concentration of 0.02028 g L−1, 0.00936 g L−1 and 0.00390 g L−1, which are used for LC, MC, and SC respectively.

Simulations

To firstly prove the feasibility and rationality of applying a DC voltage pattern as a substitution for further study about the effect of voltage amplitude on ECM, the finite element method is conducted by COMSOL Multiphysics 6.0 to simulate the different voltage patterns used in real current electronic devices including the square wave, the triangular wave, and the sinusoidal wave. The second current distribution interface and the deformed geometry interface modules are employed to perform dendrite simulation on a PCB, while the transport in diluted species interface is also utilized to depict the mass transport of tin ions from the anode to cathode35. The initial geometry is presented in Supplementary Fig. 10. Accordingly, the shape changes of the electrode induced by dendrite formation together with the tin ions flux distribution are presented to clarify the influence of different voltage patterns on ECM behavior.

Electrochemical tests

A five-channel potentiostat (Biologic VSP, France) is employed to conduct electrochemical tests with the two electrodes on the capacitor as the anode and cathode, respectively. For chronoamperometry (CA) tests, two DC voltage patterns of 3 V and 10 V (Supplementary Fig. 11) are applied on the capacitor as a substitution of other voltage patterns in the current electronic devices, and the corresponding leakage current variation is recorded. Electrochemical impedance spectroscopy (EIS) measurements are carried out over a frequency range of 105–10−1 Hz with a 10-mV amplitude sinusoidal ac perturbation. The collected EIS data is then analyzed and fitted to the equivalent circuit models by the Zsimpwin software. To obtain reliable results, all electrochemical tests are repeated at least six times. After electrochemical tests, the corrosion products together with dendrites are collected by a pipette for further investigations.

Characterizations

When performing CA tests, a time-lapse video (Dino-Lite) is used to in-situ characterize morphology changes on the capacitor surface. The morphology and composition of capacitor electrode, together with corrosion products and dendrites formed on the capacitor surface under different conditions are imaged using a scanning electron microscopy/energy dispersive spectroscopy (SEM/EDS) element mapping (Quanta FEG 250, FEI). The cross-section of the capacitor electrode is prepared via focused ion beam (FIB) using Helios 5 Hydra UX DualBeam. The dendrite morphology, composition and diffraction pattern are further analyzed by transmission electron microscopy (TEM) performed on JEM-2100, Jeol, while the composition of corrosion products is characterized by X-ray photoelectron spectroscopy (XPS) using K-Alpha, Thermo Fisher Scientific.