1. Introduction
Solid-state light sources become prevailing devices in the lighting technique [
1,
2,
3]. An essential component of them is the power LED (Light Emitting Diode) [
1,
2,
3,
4], in which the conversion of electrical energy into light takes place. During this conversion, some energy is changed into heat and increases the internal temperature of the diodes as a self-heating phenomenon [
4,
5,
6,
7,
8]. The increase in internal temperature decreases the emitted luminous flux and worsens the luminous efficiency (
ηΕ) of solid-state light sources [
8,
9]. At very high temperatures (e.g., 120 °C), the
ηΕ can decrease several times [
10]. Therefore, the efficient heat management of power LEDs is crucial [
4,
10,
11]. The efficiency of heat removal from the LED can be described using the thermal resistance of the power LED assembly:
where
Ta is the ambient temperature (K),
Tj is the internal temperature (K), and
Pth is the dissipated thermal power of the semiconductor device [
12,
13]. The
Rth is determined by the elements of the device through which the heat flows, e.g., the die of the chip, the package of the component, the substrate of the assembly, the heat-sink, and the case of the device [
14]. The thermal proprieties of the connecting elements between these different parts are also important, such as the thermal interfaces and solder joints [
15]. Other important factors that influence the efficiency of heat dissipation from electronic components are the soldering method, circuit layout, and solder alloy composition [
16,
17]. Previous studies have proved the influence of the assembly process, and the thermal pad dimension of Schottky diodes’ die on the thermal resistance of the diodes [
13,
18]. The influence of the soldering method, related to the void ratio in the solder joint, on the thermal resistance of the power LED was also investigated [
15]. Other thermal problems and solutions of power LEDs are reported in [
19,
20]. These problems are connected to the thermal modelling of power LEDs and multi-domain modelling of such semiconductor devices. However, the influence of the modification of solder joints on the thermal properties of power LEDs is still questionable.
Up to 2006, tin-lead alloys (e.g., Sn63Pb37 or Sn60Pb40) were mainly used in the electronics industry. Transition to lead-free technologies introduced the widespread use of tin-silver-copper (SnAgCu, SAC) solder alloys, e.g., Sn96.5Ag3Cu0.5 (SAC305) or Sn95.5Ag4Cu0.5 (SAC405). However, the main disadvantage of lead-free SAC alloys is their poor mechanical performance due to their high Ag content [
21]. Therefore, new research is being conducted to reduce the Ag content in SAC solders below the hypereutectic composition (up to an Ag content of 1.35 wt.%), maintaining their melting point between 217–221 °C. Moreover, because of the rapid miniaturization of electronic products, the available soldering technology may not guarantee the required performance, e.g., the I/O counts of IC chips [
22]. Solder alloys with better mechanical and thermal properties are required [
22]. One solution is to add ceramic powders and nanopowders to the solder alloys as a reinforcement. The variety of ceramics was already applied, like TiO
2, ZnO, Si
3Ni
4, SiC, La
2O
3, ZrO
2, etc. [
23,
24,
25]. Unfortunately, some of these reinforcements can be excluded from the solder joint during the soldering process [
26]. The retained ratio is usually higher when the composite solder was prepared using powder metallurgy than when using a solder-paste blending method [
26].
Recently, TiO
2 is the most frequently used reinforcement material. Grain refinement is a well-established process resulting in enhanced metallurgical properties such as toughness, ductility, and yield strength [
27]. The reinforcement addition could modify the surface instability [
28], the variation in the physical properties, and the grain boundary/interfacial characteristics [
28,
29]. The solder joints of the SAC alloy are composed of β–Sn matrix and Intermetallic Compounds (IMCs), usually Cu
6Sn
5, Cu
3Sn, and Ag
3Sn [
30]. TiO
2 particles can change several parameters of the solder alloy. First of all, the addition of TiO
2 influences the IMC growth. On the one hand, it decreases the thickness of the IMC layer [
31]; on the other, it influences the size of the Ag
3Sn grains, as well as the spacing between the Ag
3Sn grains [
27,
31]. The reduction of the spacing between the Ag
3Sn grains significantly increases the microhardness of the material [
27]. TiO
2 influence in the microstructural changes is related to the presence of enriched (with TiO
2) regions at the grain boundaries and increased microporosity [
31,
32]. The microhardness of the solder joint is influenced by TiO
2 by pinning the grain boundaries and thus impeding sliding of the grain boundaries, the increase in dislocation densities, and obstacles to restrict the motion of dislocation and the hardening mechanism of the matrix and TiO
2 nanopowders [
33,
34]. An addition of TiO
2 nanopowders has a significant effect on the ultimate tensile stress of lead-free SAC composite solders, increasing the value of this parameter of 25%. It also causes a 0.2% offset strain (0.2 YS) increase of 30%, as compared to noncomposite solder alloys. However, the total elongation of the lead-free SAC solder containing TiO
2 nanopowders is less than that of the lead-free SAC solder [
31]. Tsao et al. [
31] found that the solidus (TS) and liquidus (TL) temperatures of lead-free SAC solder reinforced with 1 wt.% of TiO
2 increased slightly from 220.1 to 220.9 °C and from 223.6 to 226.8 °C. Since TiO
2 has a high melting point, this increase in the melting point of TiO
2-reinforced composite solder alloys suggests the TiO
2 particles are dissolved locally in the molten solder. Ramli et al. [
28] observed decreases in the CTE value with increasing TiO
2 wt.% because the dimensional stability of the Sn–0.7Cu–0.05Ni composite solder is better than that of the non-reinforced solder. Better thermal-fatigue reliability is recommended, so a lower value of CTE is preferable [
28,
35].
The effect of TiO2 reinforcement on the mechanical parameters of the solder joint has been widely researched. However, the influence of TiO2 on the thermal properties of the solder joints and the whole assembly has never been described. Therefore, the aim of this study is to describe the influence of the composite solder alloy on the thermal properties of power LED assemblies by characterizing the relationship between the microstructure of the composite solder joints and their thermal properties.
3. Results and Discussion
Electric, thermal, and optical parameters of power LEDs were measured at the power supply of the diodes used as a direct current. The maximum admissible forward current of the diodes (
IFmax = 3 A) was increased to 5 A with the forced cooling system. It resulted in a higher junction temperature of the LEDs over the ambient temperature, and it allowed for more accurate measurement of
Rth and
Rthe. The measurement error of the thermal resistance is diminished due to the use of a high forward current, which resulted in a high temperature inside the LED [
36]. This allowed subtle changes in the thermal resistances to be observed, which resulted from modifying the composition of the soldering paste.
The values of electric thermal resistance
Rthe, thermal resistance
Rth, and the luminous efficiency
ηF in the case of the different samples are presented in
Figure 3.
The electric thermal resistance Rthe of the reference sample (SACX0307) and the SACX0307-TiO2 sample were almost equal. These values ranged between 4.05 K/W and 4.15 K/W. These deviations between the values of Rth correspond to the measurement error. However, the Rthe value of SAC0307-nanoTiO2 samples was 20% lower than those of the other samples. The decrease was 0.8 W/K. The whole electric thermal resistance between the junction of the LED and surroundings decreased by 16% as well. Rthe values were lower than the Rth values in all cases. The difference was as high as 40%. It suggests that as much as 40% of the electrical energy received from the power source was changed into light. The efficiency of the conversion of electrical energy into the luminous flux ηF varied depending on the type of soldering paste. The improvement of the thermal properties slightly increased the luminescence efficiency of SACX0307-TiO2. However, the minor increase in the thermal resistance of the SACX0307-nanoTiO2 solder joints resulted in a 12% luminescence efficiency decrease. This suggests that the size of the reinforcement particle is significant. The worst thermal and electrical parameters (the highest thermal resistance and the lowest luminous efficiency) were observed for the samples soldered with SACX0307-nanoTiO2 and the best for the samples soldered with SACX0307-TiO2.
One of the possible reasons for such changes could be the number of solder voids in the joints. The addition of TiO
2 influences the wettability of the solder alloy [
28]. As was proved by Ramli et al. [
28], the wettability of the composite solder alloys is better than non-composite solder alloys. This is related to the lower surface tension of composite solder alloys, which decreases the contact angle and thereby improves the wettability. TiO
2 can act as an agent to decrease the contact angle due to the segregation of TiO
2 particles near the wetting triple point [
28]. Lower surface tension significantly reduces the amount of solder voids [
37]. However, in our research, the area ratio of solder voids (lighter gray spots in
Figure 4) was similar for all of the samples.
The void content was 11–13%, which is acceptable in the case of the big component with large cooling pads. The 11–13% of void content does not influence the thermal parameters of the LED circuit.
The differences in thermal parameters were probably caused by the changes in the solder joint microstructure related to the presence of TiO
2 ceramics.
Figure 5 shows the microstructure of the cross-sectioned solder joints.
The use of both types of TiO
2 ceramic particles (nano and submicron) significantly suppressed the Sn grain growth. The reference samples (SACX0307) were investigated using polarized optical imaging and BSE-SEM as well (
Figure 5A,B) to localize the large Sn grains. The grain size in these solder joints was as large as hundreds of micro-meters, which is typical for simple solder alloys. The whole solder joint contained only a few grains with different orientations. On the contrary, in the case of composite solder alloys (both types SACX0307-TiO
2, as well as SACX0307-nanoTiO
2), the average grain size decreased to 3–5 μm, which means grain refinement of 2 orders of magnitude (
Figure 5C,D). Grain size differences were not observed between composite solder alloys; this means it was equal for the samples SACX-TiO
2 and SACX-nanoTiO
2.
Figure 6 presents the incorporation of the nanoTiO
2 particles into the solder matrix, and
Table 1 presents the elemental composition of the different phases in the solder joints (oxygen and carbon were excluded from the EDX results).
Figure 6A,B shows the differences between the IMC layer regions of the composite and the reference solder joints. In the composite solder joints, the TiO
2 particles are located on the surface of the IMC layer.
Figure 6C,D shows the same FIB cut in the solder bulk of a SACX0307-nanoTiO
2 sample by SE and BSE detectors. When an SE detector is used, the IMC particles and the grain boundaries are visible. When a BSE detector’s used, the nanoTiO
2 particles are visible. A comparison of the two micrographs (
Figure 6C,D) reveals that the TiO
2 particles are located at the grain boundaries and around the IMC particles. The TiO
2 particles cannot be dissolved in the molten solder, so during the solidification of the solder joint, they precipitate at the grain boundaries and around the IMCs. It might cause the suppressing effect of the TiO
2 particles on the IMC layer and Sn grain growth in the composite solder joints, discussed previously.
The thickness of the IMC layer, as well as the volume fraction of the IMC particles, are important parameters from the point of view of the electrical and thermal conductivity of the solder joints. The structure of the solder joints can be distinguished because it is rather like “a sandwich” structure, with upper and bottom parts. The upper part means the interface between the component (diode) and the solder joint, the bottom part corresponds to the interface between the solder joint and the substrate. Both interfaces contain a Cu layer, which takes part in the formation of IMC layers.
The average thicknesses of the upper and bottom IMC layers were calculated. No significant differences were found between the samples from the point of view of the upper IMC layer. This was between 2.2 ± 0.2 μm, and it was continuous round-type (
Figure 7A–C).
The upper IMC layers probably remained the same since the volume fraction of the TiO
2 decreased at the upper region of the solder joints, as was found by Nasir et al. [
38]. However, the ceramic particles suppressed the Cu
6Sn
5 IMC layer growth significantly at the bottom. The average bottom IMC layer thicknesses were 2.72 ± 0.32 μm (SACX0307), 1.94 ± 0.17 μm (SACX0307-TiO
2), and 2.1 ± 0.91 μm (SACX0307-nanoTiO
2) (
Figure 7D–F). The use of the ceramic particles caused a ~30% decrease in the bottom Cu
6Sn
5 IMC layer thickness, the thickness of the Cu
3Sn layer did not change. The type of the bottom IMC layer was an elongated scallop-type in the case of SACX0307 and a rough scallop-type in the case of SACX0307-TiO
2 and SACX0307-nanoTiO
2 samples. Tang et al. [
39] also found that an increase in TiO
2 nanoparticles from 0.02 to 0.1 wt.% changes the morphology of the Cu
6Sn
5 IMC layer from an elongated scallop-type to rough scallop-type. A further increase in the TiO
2 particle proportion (0.3 and 0.6 wt.%) can cause the growth of wicker-type Cu
6Sn
5 grains [
27,
39].
In the case of the composite solder joints, wicker-type IMC grains formed. These wicker-type Cu
6Sn
5 grains were several ten to several hundred micrometers in length. Tsao et al. [
31] wrote about the appearance of wicker-type Cu
6Sn
5 IMCs in the case of adding TiO
2 nanoparticles to Sn3.5Ag0.25Cu solder alloy. They found that wicker-type IMCs grow from the top of scallop-type IMCs, so nucleation is not required. It was explained that the growth is initiated by a local concentration increase in Cu. The whicker-type IMC elongates into the solder matrix due to the reaction of Sn with this Cu. The presence of TiO
2 particles in the molten solder could influence the sudden growth in Cu concentration at given places, although further research into this issue is necessary.
The changes in the Cu
6Sn
5 IMC layer in the composite solder joints can be explained by the following: while being soldered, the dispersed TiO
2 particles in the molten solder precipitate at the bottom of the solder joint. According to heterogeneous nucleation theory, the presence of TiO
2 particles decreases the thermodynamic energy of Cu
6Sn
5 nucleation since the Cu
6Sn
5 prefers to nucleate on the TiO
2 particles. The increase in the volume fraction of TiO
2 particles means more and more nucleation sites and the increase in nucleation probability of Cu
6Sn
5 grain growth. This increases the nucleation rate and results in the refinement of Cu
6Sn
5 grain by blocking the sequential grain ripening. According to Ostwald ripening theory, the size differences between neighboring IMC grains cause the solder flux to diffuse towards the bigger grains and away from the smaller ones. This results in the further shrinkage of smaller grains and enlargement of bigger grains, so the large grains grow by absorption of atoms which have diffused from the small ones [
39]. This phenomenon was also observed in our research in the composite solder joints, namely some larger Cu
6Sn
5 grains formed between the refined Cu
6Sn
5 grains (
Figure 7E,F).
Furthermore, the suppression of Cu
6Sn
5 IMC layer growth can also be explained by the adsorption theory of TiO
2 particles as a surface-active material. The surface energy of the Cu
6Sn
5 grains is minimized in the equilibrium state. An increase in the adsorption of an active element like TiO
2 particles decreases this surface energy. According to the Gibbs free energy equation, surface energy reduction decreases the growth velocity of the Cu
6Sn
5 grains and results in the suppression of the whole Cu
6Sn
5 IMC layer growth. It also has to be noted that the presence of TiO
2 particles in the solder matrix can act as diffusion barriers of Cu and Sn atoms, which are responsible for the further growth of the IMC layer in the solid-state of the solder joint. According to our observations, the thickness of the IMC layer was also decreased by the spalling of the IMCs into the solder bulk. The traces of the more intensive spalling are clearly visible in the composite solder joints (
Figure 7E,F). This spalling/braking phenomenon of the IMC grains has also been reported by other researchers [
39,
40]. Tang et al. [
39] showed that the IMC grains contain twice the number of weight percentages of Ti and O than the surface of the Cu
6Sn
5 IMC grain. This can cause the more intensive spalling/braking of the IMCs in the composite solder joints.
Some refinement of the Ag
3Sn IMCs was observed in the SACX0307-TiO
2 solder joints. The Ag
3Sn particles were evenly distributed, and they were isolated with larger spacing between them (
Figure 7E) than in the case of the other two sample types. In the case of the reference SACX0307 and SACX0307-nanoTiO
2 solder joints (
Figure 7D,F), Ag
3Sn particles frequently agglomerated to the groups. The reason could be that during the solidification process, the TiO
2 particles ensure a high nucleation density of the second phase in the eutectic colony [
41]. Chung et al. [
42] and Tsao et al. [
31] observed the same phenomenon in the case of different composite solders joints. Interestingly, this effect was not observed in the case of nano TiO
2 particles.
Although the TiO
2 nanoparticles decreased the average bottom IMC layer thickness to 2.1 ± 0.91 μm, the deviation was very high. The reason for the high deviation was examined further, and it was found that the IMC layer thickness could differ significantly, even in one given sample. The most extreme IMC layer thickness differences were found over the thermal pad, where most of the dissipated heat leaves the power LED.
Figure 8 shows a sample part of the SACX0307-nanoTiO
2 solder joint, where the IMC layer thickness changed suddenly and considerably.
On the right-hand side of the solder joint, the IMC layer growth was suppressed effectively, the average thickness was around 1 µm. On the left-hand side of the solder joint, the IMC layer was even thicker than in the reference SACX0307 sample, the average thickness almost reached 3.5 µm. The Kirkendall voids were also more frequent at these thicker IMC layer regions. Considerable local increases in the bottom IMC layer thickness and Kirkendall voids were frequent over the thermal pad. No direct evidence was found which could explain these phenomena. It might also be related to the local concentration increase in Cu, like in the case of wicker-type IMC growth, since, in both cases, considerable IMC growth occurred. However, the role of the TiO2 particle size is unclear, and why it occurred only in the case of nanoparticles; further research into this issue is necessary.
The Sn grain refinement in the composite solder joints results in considerable mechanical stability improvement, as was reported by Nasir et al. [
38]. However, the size and orientation of the Sn grain do not have a significant effect on the heat conduction ability of the solder joints. Nevertheless, the IMC layer acts as a continuous heat insulator layer in the solder joints between the contact pads and the solder bulk. The specific heat conductivity of the Sn and the Cu
6Sn
5 is 73.3 W/mK and 34.1 W/mK, respectively. It could decrease the
Rth and
Rthe values of power LED assemblies soldered with SACX0307-TiO
2. The spalling of IMCs does not change the volume fraction of the Cu
6Sn
5 in the solder joints. The IMCs are only more distributed in the solder joint and not concentrated on the Cu and Sn interface. More distributed Cu
6Sn
5 IMCs ensure heat paths with lower heat resistance.
The average IMC layer thickness also decreased in the SACX0307-nanoTiO
2 solder joints (compared to the reference SACX0307), but a considerable and local increase in the bottom IMC layer thickness and the Kirkendall voids were frequently found over the thermal pad. The Kirkendall voids also decrease the thermal and electrical conductivity of the solder joints [
43]. This could explain that the thermal properties of the SACX0307-nanoTiO
2 solder joints were not improved compared to the reference SACX0307 solder joints. The minor refinement of the Ag
3Sn IMCs in the SACX0307-TiO
2 solder joints probably has no considerable effect on the heat conductivity of the solder joints, but it could mean further mechanical improvement [
21].