Analysis of solder mask roughness and stencil shape influence on void formation in solder joints

Voids inside solder joints are empty spaces negatively affecting the mechanical, thermal, and electrical properties of the solder joint. This article deals with this problematics using two novel approaches to reduce voids’ occurrence. The first approach is the change of the roughness of the solder mask to influence the flux spreading around the solder pad. The second approach is the deposition of the same volume of the solder paste to the solder pads using a modified, thicker stencil with smaller apertures. The experiments were performed for solder pastes containing SAC305 alloy and two types of fluxes (ROL0/ROL1). Solder joints were inspected by X-ray imaging, and the shots were subsequently processed by image analysis. For determination of the spreading area of the flux around the solder pad during the reflow process, the tested boards were scanned by a confocal digital microscope. The results showed that more aggressive flux caused less voiding in terms of the average area of voids covering the soldering pad. Analysis of samples prepared with modified stencils showed a much lower proportion of voids than with the standard stencil. Furthermore, the results revealed that the selection of the solder mask type significantly influenced the voids’ formation within solder joints in the case of the unmodified stencil for solder paste deposition. On the other hand, when the modified stencils were used for sample preparation, the solder mask had no further impact on voiding. Therefore, modifying the stencil can be pointed out as a preferable and effective way.


Introduction
Solder joints serve as the electrical, thermal, and mechanical interconnections between the printed circuit board (PCB) and components in electronics assembly. One of the serious defects of the solder joints is the formation of voids (empty spaces formed inside the joints). They have a negative influence on all types of interconnections mentioned above and thus pose a certain degree of reliability risk. The study by Steiner et al. [1] shows that voids in solder joints significantly influence the joint's mechanical strength. The larger the void area in the joint, the more the mechanical strength of the joint decreases. The same applies to the thermal properties of the joint, as confirmed by Zhu [2].
Voids can be divided into several groups. These are microvoids, macrovoids (voids), pinhole voids, shrinkage voids, Kirkendall voids, and micro-vias voids. Their typical layout and shape are shown in Fig. 1.
There is no explicit agreement about the size of macrovoids. Voids with a diameter below 50 μm are sometimes referred to as microvoids [3].
Macrovoids are known as process-dependent voids because they are created during the manufacturing process when various conditions influence their formation. Therefore, to understand the formation of the voids, we must first understand the manufacturing-soldering process. A typical process of surface mount technology starts with stencil printing, when a solder paste is deposited to solder pads on the PCB. The printing process has many variable process parameters such as ambient conditions, aperture geometry, aperture finish, and solder paste rheological properties [4]. Hence, it is still not well understood in depth due to its complexity. Especially the insufficient quality of solder pastes and their rough rolling and filling of stencil apertures can cause absorption of the air bubbles. This can lead to inadequate deposition of solder paste on the substrate and the further formation of voids during reflow soldering [5].
The printing process is followed by mounting the components by an automatic pick-and-place machine on the solder pads with deposited solder paste. The components are pressed into the solder paste with force pre-defined by a mounting head. Improperly set pressure force can lead to a change in the shape of previously deposited solder paste and thus increase the air bubbles inside the solder paste [6].
After that, PCB with mounted components goes to the soldering oven, where the solder paste is reflowed. Many studies observed and confirmed that the formation of voids strongly depends on the reflow soldering parameters, including preheating zone, peak temperature, and soaking time [7]. The direction of heat flow and cooling can also have an influence on the properties and structure of the resulting solder joints and voids' formation, which was confirmed by [8][9][10].
Voids' formation is caused by the outgassing of flux and other gasses which cannot escape from the solder joint during reflow. The evaporation of rheological additives and solvents from solder paste during the heating process is the next source of outgassing substances [11].
Another source of void formation during the reflow can be the flux reaction with the soldering part metallization or with oxides present on soldering surfaces or in the solder paste. The effect of PCB surface finish on voids' formation was studied by Nurmi et al. [12]. They stated that the OSP (organic surface preservative) surface finishes promoted the formation of voids more than the Ni/Au surface finish. The impact of PCB surface finish type, solder paste type, particle size distribution, and flux type on voids' formation was confirmed in other studies [13,14].
Generally, the voids are formed from gases trapped in the liquid solder alloy, and they are frozen inside the solder joint during solidification [15].
Many ways are examined to find the methods to eliminate the void's formation and, thus, the reliability risks caused by them. Using a nitrogen atmosphere has a positive effect on reducing the formation of voids [16]. Moreover, a significant reduction of voids can be achieved by applying a vacuum during reflow soldering [17].
Dušek and Bušek pointed out in their article that flux activity and composition have a significant effect on the formation of voids in solder joints. The flux also causes the removal of oxides and impurities from the soldered surface, which results in better solder flow. The surface tension is reduced, and wettability of the solder alloy is better [18]. In our experiment, solder pastes with rosin-based fluxes were used. Their classification can be found in Table 1.
The IPC-610 standard specifies an acceptability criterion for voiding: less than 25% void area on top-view image obtained by transmission X-ray analysis [20]. The solder joints are acceptable when the size and number of voids are at a low level which removes the potential risk of poor conductive connections [17]. This article deals with two new approaches to reducing the formation of voids in the solder joints.
The first evaluated approach was based on the idea of reducing the amount of flux present in the solder joint during the liquid and solidification phase of the process. For that purpose, a higher level of flux spreading to the surroundings of the solder pad is necessary, which can be achieved by changing the surface tension equilibrium between liquid flux, ambient air, and the surface of the solder mask [21]. The liquid wetting and spreading behavior on the surface depend primarily on the liquid and   substrate properties [22]. However, the only left possibility to change in the equation is the roughness of the solder mask. Surface roughness affects the wetting of liquid [21]. In a previous study [23], the effect of solder mask roughness on flux spreading was also proved -with a rougher solder mask, the flux wet better the surrounding surface around the solder pad, and the spreading area is larger. This could lead to less flux in the liquid alloy volume and, thus, decrease the probability of the flux entrapment within the joint. The second approach consisted of modifying the design of the steel stencil for solder paste printing. Instead of one aperture per solder pad, whose dimensions are equal, the modified stencil contained two or four smaller apertures per solder pad. In terms of keeping the same volume of deposited paste as for the unmodified stencil, the modified stencil was thicker. The modification should cause a higher spreading activity and movement of the solder paste over the solder pad during the reflow process, so a higher amount of gas bubbles could be forced out of the joint volume. Furthermore, such a design creates air gaps and channels through which the gassed flux parts can leave the solder joint more efficiently.
Based on literature research, the presented approaches are novel, and the impact of using them against the voiding issue was not described and analyzed in detail anywhere else. Effects on void occurrence in solder joints that have been investigated in other studies include, for example, changing the type of surface finish of solder pads or temperature profiles of reflow changing, where the authors suggest that up to approximately 5% improvement can be achieved [1]. Furthermore, the particle size of the solder alloy was investigated in the previous study, yielding an improvement of around 2% [24]. Changing the flux chemistry can also contribute to reducing voids formation [3]. However, the advantages of chosen approaches in this work include no additional costs in the soldering process and easy implementation compared with those mentioned earlier.

Samples and methods
A special PCB with dimensions 46 mm × 33 mm was designed for our experiment. A total of 90 boards were produced and analyzed, five boards for each combination of stencil type, solder mask, and solder paste type.
The PCBs were designed so 20 resistors with a package size of 2010 were soldered and tested on each. ENIG (electroless nickel-immersion gold) was chosen as the surface finish of the solder pads since it represents one of the standard surface finishes. The PCBs were in three groups, one without a solder mask and two with solder masks differing in surface roughness. One was matt (black variant), and the other was glossy (white variant). The surface roughness measurement of the solder masks was also included in this study.
Three types of steel, laser-cut stencils for solder paste printing with different shapes of apertures were designed and manufactured for this experiment. The basic idea is to apply always the same volume of solder paste to each solder pad, and thus the thickness of the stencil and dimensions of the apertures were modified. The stencils with three different designs of apertures with the same volume were marked as stencil 1, stencil 2, and stencil 3 (see Fig. 2).
The stencil with the first design (stencil 1) is 0.1 mm thick, and the other two (stencils 2 and 3) are 0.2 mm thick. This ensured the deposition of the same volume of the solder paste. The solder paste printing was performed using the hand-operated stencil printer SAB 06 (ELPRO s.r.o., Slovakia) with an alignment error of less than 0.1 mm.
Two solder pastes with the same solder alloy composition (Sn96.5Ag0.5Cu3) were used, differing in flux activity (ROL1 and ROL0, according to the IPC standard J-STD-004). The ROL1 flux has a higher chemical activity regarding the reduction of the oxides (due to the higher content of halides) than ROL0. The complete parameters of the solder pastes used in the experiment are shown in Table 2, where solder paste with ROL1 flux is marked as paste A and solder paste with ROL0 flux as paste B.
The PCBs were reflowed in a continuous air convection oven Mistral 260 (Spide-SMT, Netherlands), which allows the setting of three temperature zones. Two zones are used to preheat the PCB and activate the flux, and the last zone is to melt the solder alloy. The speed of the oven conveyor was set at 15 cm/min. The temperature profile was set according to the manufacturer's recommendations and was the same for both solder pastes and, therefore, for all tested samples. The temperature profile measured by a thermocouple attached to the soldering pad on the tested PCB is shown in Fig. 3. No temperature profile parameter differed from the manufacturer's recommendations, as it could also affect voids formation and distort the results of our study. Changes in reflow profile parameters were not the subject of this study, and their influence on void formation in solder joints has been confirmed elsewhere [1]. The temperature profile measured by a thermocouple attached to the soldering pad on the tested PCB is shown in Fig. 3.
Since the voids are located inside the soldered joint, it was necessary to X-ray (Nanomex, Ge-Phoenix, USA) samples. The output of this process was the images in which bright spots represent voids in a soldered joint. The captured images were further processed by the software NIS Elements (Laboratory Imaging s.r.o., Czech Republic) using brightness analysis (distinguishing between light and dark areas in the image). Subsequently, the bright spots are marked (Fig. 4), and their number and size of the area are calculated. Due to a limited resolution of the images, only voids larger than 5 µm in diameter were detected. The void areas were detected with a measurement error of less than 0.5%. Because we also evaluated the occurrence of macrovoids, in our work, the minimum size limit for macrovoid was set to 100 μm.
To detect the spreading area of the solder flux on the PCBs during the reflow process, PCBs were photographed using a confocal digital microscope (VK-X1000, Keyence, Japan). The images were then processed again using the NIS Elements software utilizing the same method as for the voids' evaluation. In Fig. 5, the spreading area of solder flux outside the solder joint, and its marking for analysis in NIS Elements can be seen. The results are shown in Fig. 6. Further, the roughness of the used solder masks was also measured by the confocal microscope. Three roughness parameters were determined from the measured data: R a (average height of the surface profile), R L0 (developed length of a roughness profile expressed in % of expansion

Results
The obtained results from the performed experiments are presented in the following figures. At first, an evaluation of the spreading area of the flux around solder pads was conducted (Fig. 6). Regarding the first presented approach to reduce voiding -change of the PCB surface roughness -a larger spreading area should indicate a lower level of voids inside the joint. Table 3 presents an additional measurement of the surface roughness of the tested PCB variants. Subsequently, X-ray imaging and image analysis were performed. The analysis of a number of macrovoids in solder joints was done to see their ratio in the solder joint and the relationship between the total voids area and macrovoids area and how they influence the total voids area. The overall results of the average area of macrovoids in solder joints in the tested samples as a percentage of solder joint area are shown in Fig. 7. Microvoids made up only a minimal part of voids, and their occurrence was almost identical (in absolute numbers) for all samples examined. These voids are the result of interfacial reactions at the substrate-solder interface. Therefore, they were excluded from the final evaluation and graphs, since the study was mainly focused on process-related voids (macrovoids). All data were statistically tested with a Student T-test with a standard significance level α = 0.05.

Occurrence of microvoids in solder joints
When the complete data were analyzed, it was apparent that macrovoids (with a diameter more than 100 µm) had a significant contribution to the formed voids. This phenomenon applied to all observed samples. Only 1.89% (on average) of voids area belonged to microvoids and

Influence of flux type on voids' formation
The results show that samples, where the solder paste with ROL1 flux (paste A) was used, had a smaller voids area for all used combinations compared to samples with solder paste with flux ROL0 (paste B). The content of voids in solder joints made from paste A was 20% lower than from paste B. One possible reason for the smaller voids area in the case of ROL1 is its higher activity compared to ROL0. Higher flux activity allows better reduction of oxides from the surface of the soldering pad. Furthermore, the substances in this flux may be more volatile, which could result in their faster evaporation and escape from the solder joint volume. The results showing the positive effect of a high flux activity on voids' formation are in full agreement with [3]. At first glance, it is evident that the use of ROL1 flux (paste A) generally caused fewer voids, but it also had a higher spreading around the solder pad compared to the ROL0 flux (paste B) (see Fig. 6 and Fig. 7). It means not just the flux activity but also the spreadability of the flux to the surrounding of the soldering pad has an influence on the formation of the voids, and it should thus be taken into consideration during the evaluation of the voids' formation.
According to a Student T-test, the difference between paste A and paste B data was statistically significant in all cases.

Influence of different shapes of stencil apertures
The results show that the samples, where modified designs of stencil apertures (stencil 2 and stencil 3) were used, exhibit lower voids formation than in the case of the standard design of stencil apertures (stencil 1). In the case of stencils 2 and 3, the solder paste (melted alloy, respectively) tended to spread to the originally uncovered parts of the solder pad, which probably enabled better outgassing during the soldering process. On samples prepared with stencil 1, the spread of the melted solder was nearly negligible since the solder paste already covered the whole area of the solder pad after the solder paste deposition process. Figure 8 presents the differences between the samples prepared with different stencils. It is clear that the volume of the solder paste was maintained, but the covered area decreased for Stencil 2 and Stencil 3. The effect of using the stencil with special apertures on voids reduction was also confirmed by the study of Diehm et al. [20]. However, they did not change the thickness of the stencil, so the solder paste volume on the pads decreased compared to the original full-covered area. This may lead to a mechanical weakening of the solder joint. Furthermore, the investigation was performed only for joints reflowed in a vacuumequipped soldering oven, whereas our experiment simulated a more common soldering process in practice. Another question that may arise is the relation between the flux spreading to the surroundings and changing the size of the stencil's apertures. It is apparent that the flux spreading to the solder pad's surroundings is greatest for stencil 1, although the deposited volume of solder pastes and thus the volume of the flux on the soldering pads were the same for all cases.
To get a better view of this phenomenon, we made the analysis of several solder joints' reliefs, on which the crosssection dimensional profiles of solder joints on soldering pads were measured. The measurement was conducted on the confocal microscope. In Fig. 9, the differences in the dimensional profile between the solder joints prepared by stencil 1 and stencil 3 are obvious. For stencil 1, the flux residues appeared rather in the surroundings of the solder pad, whereas the flux residues on samples prepared by stencil 3 remained rather on the top of the solder joint. Such a difference in flux behavior is probably a consequence of different wetting processes, comparing the pads fully covered by solder paste (stencil 1) and pads covered only partially (stencil 3). Therefore, it can be stated that the main principle of void reduction differs from the approach based on different surrounding surfaces (an effort to increase the flux spreading area) and is based primarily on creating air gaps and channels through which the gas substances of the flux can escape the solder joint volume. Additionally, the liquid substances are forced onto the top part of the solder joint during the soldering process.
A Student T-test showed that only the difference between stencil 1 and stencil 2 data was not statistically significant (p = 0.099).

Influence of different types of solder mask
In the case of solder paste deposition via unmodified stencil, the roughness of the PCB surface significantly influenced the formation of voids within the solder joint (Fig. 7). The best results were achieved when PCBs were covered with no solder mask. In this case, the voids ratio to the total solder joint area averaged 10.2%, which is about a 20% reduction compared to the PCB with a glossy solder mask. The substrate without the solder mask exhibited the highest surface roughness from evaluated variants of PCB. As the previous studies state [21,25], the higher surface roughness generally increases the wettability of the liquid. The flux spreading area was thus larger (Fig. 6), and the flux was drained away from the solder joint volume to a greater extent. These thoughts are in agreement with the study of Hirman and Steiner [26], which stated that the probability of void occurrence increases if the flux does not leave the solder joint. The differences in flux spreading can also be found between matt and glossy solder masks, although the impact on the voiding is not as significant -according to the results of the statistical T-test, the p-value was equal to 0.08.
In the case of stencil 2 and stencil 3, the influence of the solder mask is not as apparent as in the case of stencil 1. The reason is the much lower spreading of the flux to the surroundings of the soldering pad, as discussed in Section 4.3.

Conclusion
The influence of different flux types (ROL0 and ROL1), solder pad surroundings (without solder mask or with solder mask of different roughness), and stencil designs (ensuring the same volume of deposited paste on the solder pads but varying the deposited paste area) on voids' formation were evaluated. Results show that macrovoids occupy a much larger area of solder joints than other types of voids, regardless of the tested factors. The voids' formation was largely affected by modification of the stencil for solder paste printing. When the modified stencil with smaller apertures was used, the process of wetting uncovered areas enabled better outgassing and led to a decrease in total void area. From the evaluated approaches, the stencil modification, consisting of decreasing the apertures' size and increasing the thickness, is preferable.
Based on the detailed results, the preferred combination of the flux type, stencil design, and solder mask type is more aggressive flux (ROL1) together with the stencil containing apertures divided into four smaller openings and no solder mask. In the case of the unmodified stencil, the preferable option is to use ROL1 and PCB with a rough surface (no solder mask or matt one). These combinations significantly reduce the formation of voids in the solder joints.
The approaches presented in this paper have a comparable impact on void formation in solder joints as the temperature profile, or particle size in solder paste [1,3,24]. Nevertheless, the presented approaches are easy to implement and do not cost extra compared to other mentioned solutions, where it is often necessary to change the soldering process parameters or the composition of solder paste.
Funding Open access publishing supported by the National Technical Library in Prague. This work was supported by the Grant Agency of the Czech Technical University in Prague, grant no. SGS21/159/ OHK3/3 T/13.

Conflict of interest The authors declare no competing interests.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.