Effects of Plasma Treatment on Ni-Plated Copper as a Heat Spreader for fcBGA-H Device

Plasma treatments change surface roughness and activate the functional groups on them. This study analyzes plasma effects on nickel-plated copper heat spreader with a reactive gas comprising argon and oxygen mixture and investigates the effects of plasma treatments and their resultant adhesion properties using atomic force microscopy (AFM), goniometer, X-ray photoelectron spectroscopy (XPS), and scanning acoustic tomography. After plasma treatment, the surface roughness and surface-area difference percentage (SADP) of the heat spreaders increase from 4.605 to 5.207 nm and from 1.876% to 2.668%, respectively. When compared with bare heat spreader surfaces, the surface energy of those that assist adhesion increases from 22.93 to 72.53 mN/m, and the amount of activated functional groups on them also increases by 1.79 × after plasma treatment. With an increase in SADP, surface energy, and the amount of activated functional groups, the thermal interface material (TIM) coverage increases and the delamination between TIM and the heat spreader improves after plasma treatment. Therefore, plasma treatment on heat spreaders is an effective method to improve the adhesion properties of TIMs, indicating improvements in the heat dissipation performance of flip-chip ball grid array-heat spreader (fcBGA-H) devices.

SECTION I. Introduction

The rising need for multifunctional high-performance electrical devices has increased the requirements of output power and levels of integration in semiconductor packaging [1]. The flip-chip technology is an appropriate option for satisfying all requirements as it is a commoditized technology that bonds a chip with a substrate. Due to the high output power requirement for multifunctionalities and high performance, the heat generated from a flip chip causes several issues. The operation of electrical devices at high temperatures not only degrades their performance but also shortens their lifetime, leading to unwanted results caused by heat. For stable and long-term use of electrical devices, thermal management is required to dissipate heat [1], [2].

To manage the heat generated by flip-chip devices, researchers suggest the use of heat spreaders (or base plates) that spread heat from hotspots at flip-chip locations by effectively transferring the heat from flip chips to heat sinks. Industrial applications use heat spreaders for spreading and dissipating heat through a structure called a flip-chip ball grid array-heat spreader (fcBGA-H). Representative materials having high thermal conductivity can be used as a heat spreader, including diamond (2200 W/m-K), silver (419 W/m-K), copper (385 W/m-K), graphite (240 W/m-K), aluminum (205 W/m-K), tungsten (175 W/m-K), and brass (147 W/m-K) [3], [4]. Despite the very high thermal conductivity of diamond and silver, their commercial use as a heat spreader is rare owing to the high cost of raw materials. However, some industries such as automotive, aerospace, or defense industries do not consider the raw-material cost, and several researchers have attempted to apply such materials as heat spreaders. Akhtar [1] investigated copper-diamond composites as heat spreaders and obtained desirable properties such as low coefficient of thermal expansion (CTE), low density, and improved structural response. Abyzov et al. [2] studied a composite of diamond particles with tungsten coating on a copper matrix. The composite was reported to have a high thermal conductivity of 500–900 W/m-K when synthetic or natural diamonds were added to the copper matrix. Nunes et al. [5] investigated the fabrication of composite materials comprising copper and nanodiamond with a focus on contamination and optimal nanoparticle dispersion by changes in the milling conditions. Kim et al. [6] fabricated a hybrid material comprising reduced graphene oxide (rGO) and silver nanowires (AgNW) on an Al₂O₃ substrate by ultrasonically spraying at cold temperatures. They anticipated high adhesive performance due to spraying at cold temperatures.

Unlike diamond and silver, copper is a common material which makes its cost reasonable for commercial applications. Therefore, several industries have adopted copper as a heat spreader, with numerous active research underway. Yu et al. [7] applied copper as a heat spreader by plating or paste-filling a large fan-out ball grid array (BGA) solution using the fin field-effect transistor (FinFET) process. Zha et al. [8] investigated copper heat spreaders on light-emitting diodes with bright and stable operation. Kuc et al. [9] studied the impact of copper heat spreaders on the thermal performance of III-N-based laser diodes. They reported a considerable increase in power conversion efficiency. Moon et al. [10] attempted to increase the performance of vertical-cavity surface-emitting lasers (VCSEL) by applying copper heat spreaders. They reported a decrease and increase in thermal resistance and peak power to 46.33% and 12.07% compared with those of a VCSEL without a heat spreader.

Copper is classified as a highly reactive and corrosive material that can be oxidized in air. As oxidation and corrosion are critical issues in semiconductor devices, surface coating or plating with anti-corrosive materials is suggested. For anti-corrosive coating, corrosion-protective mechanisms can be classified into barrier protection, surface passivation, and sacrificial protection (using the galvanic effect). The classification of anti-corrosive coating based on materials can be divided into organic, inorganic, and metallic coatings [11]. For anti-corrosive coating on heat spreaders, metallic coating with stable and unreactive materials is considered appropriate for barrier protection due to heat dissipation. Metals such as zinc (112.2 W/m-K), nickel (97.5 W/m-K), chrome (94 W/m-K), cobalt (69.2 W/m-K), or molybdenum (35.5 W/m-K) [3], [4] can be used as anti-corrosive coating materials. Among these metals, nickel offers high thermal conductivity and its atomic properties are highly similar to those of copper, including atomic period, atomic radius, ionization energy, and crystal structure. However, the chemical stability (corrosion or oxidation) of nickel is quite distinguishable from that of copper. Therefore, nickel can be considered a good anti-corrosive coating for copper.

To effectively dissipate heat, thermal interface materials (TIMs) require consideration [12]. Heat is transferred by conduction from a flip chip to a heat spreader through their contact area, although heat also transfers through air gaps by radiation and convection [13]. Both flip chips and nickel-plated copper heat spreaders are solid materials with imperfectly flat and smooth surfaces. Therefore, the area under contact is much smaller at approximately 1%–2% of apparent contact area [14]. To mitigate such a phenomenon, TIM materials having high thermal conductivity and viscoelastic property are applied between flip chips and heat spreaders. Gwinn and Webb [13] applied TIM between the central processor unit (CPU) and a heat sink to measure thermal interface resistance. Xu et al. [15] applied pressing force between flip chips and heat spreaders after dispensing TIMs between them to create hermetic contacts. The schematics of heat flow from a flip chip to a heat spreader with and without TIM are depicted in Fig. 1. The thermal conductivity of TIM along with the hermetic contact and adhesion between flip chip–TIM–heat spreader is important to effectively dissipate heat.

Fig. 1.

Heat flow estimation diagram (a) without TIM and (b) with TIM.

All electrical devices including a fcBGA-H face inevitable issues due to warpage while following technical trends of electric devices including lighter, thinner, shorter, and smaller devices. Warpages cause differences between maximum and minimum displacement values in specific areas. The difference can be measured by an optical microscope (OM) or a laser profiler but dynamic trends as per temperature cannot be measured using these instruments. A thermal shadow moire (TSM) is used to measure the warpage of devices as per temperature. Some studies have reported results on the warpage of fcBGA-H using TSM [16]. Generally, a larger variation due to warpage can cause frequent defects and delamination in electrical devices resulting in decreased effectiveness of heat transfer. Thus, filler-type TIMs including adhesive components are considered appropriate for thin fcBGA-H devices [14], [15]. Due to the high modulus and rigid properties of heat spreaders along with the adhesion component, variations caused by warpage are expected to be reduced. Moreover, the junction interface between a heat spreader and TIM is heterogeneous when filler-type TIM (Si-based) is used. Herein, the adhesion of a heterogeneous interface is naturally poorer than that of a homogeneous interface, and research on methods to improve adhesion between heat spreaders and TIMs is limited.

Some studies suggest that plasma treatments can mechanically change surface roughness and activate the functional groups on them, thereby assisting adhesion [17], [18]. Yang et al. [19] reported that a clean surface typically has a high roughness and a small contact angle (C/A). Plasma treatments remove some of the organic or inorganic contaminants on surfaces through physical etching or chemical reactions. Therefore, the surface can be roughened and enlarged when compared with untreated areas. When surfaces are enlarged, the bondable area between adhesive and adherend (heat spreaders, in this article) gets enlarged, and the bonding force increases. Djennas et al. [20] analyzed the effects of plasma on delamination and cracking of plastic packages. They reported that nonreactive plasma was effective in reducing contaminations. Furthermore, the surface that was plasma treated with oxygen-reactive gas showed a larger amount of functional-group activation than that treated with argon or nitrogen reactive gas [21]. Although numerous research results on plasma effects are available, research on the effect of plasma treatment on heat spreaders is limited.

Consequently, the properties of heat spreader surfaces before and after plasma treatments and their adhesion performance with TIM are analyzed in this study. Widely used reactive gases, argon and oxygen are coincidently applied as a mixture during plasma treatment. The mechanical, morphological, and chemical properties of a nickel-plated copper heat spreader surface and the resultant adhesion property with TIM are analyzed in this study.

SECTION II. Experimental Detail

In this study, a fcBGA-H package containing a plastic circuit board (PCB; also called the substrate) of dimension $30 \times \, 30$ mm² and a flip chip of dimension $15 \times \, 15$ mm² was employed. A nickel-plated copper heat spreader was used to dissipate the heat generated from the flip chip to evaluate the plasma treatment. Fig. 2 and Table I present the dimensions and material information of the device, while the fabricating procedure is depicted in Fig. 3. All materials used for the experiments were cleaned by the material manufacturing companies and provided to us in vacuum packaging. To strictly control oxidation, all materials were managed with a short shelf time and stored in an N₂ dry cabinet.

TABLE I Dimensions and Materials of Device

Fig. 2.

Structure of fcBGA-H (a) top and (b) section structural view of fcBGA-H.

Fig. 3.

Experimental procedure including sample fabrication and inspection.

A. Flip Chip Attach and Reflow

Flip chips were soldered on PCBs by a massive reflow (MR) soldering process, whose temperature profile is presented in Fig. 4. The peak temperature, soak time, and dwell time of MR were 250 °C, 120 s, and 100 s, respectively.

Fig. 4.

Reflow profile for flip chip soldering.

B. Under Fill and Oven Cure

After MR soldering of the flip chip, the flux-material remains were cleaned following a flux cleaning process (de-flux process), which employed deionized (DI) water with a small amount of detergent at 50 °C–70 °C, followed by drying process with air. To protect the solder bumps that connect the flip chip and PCB, under-fill (UF) fluids were dispensed adjacent to the flip chip and diffused by capillary effect. The UF fluids were cured in a convection oven with target temperature of 125 °C. The temperature rising time was 30 min, curing time was 2 h at peak temperature, and cooling time was 30 min. To mitigate oxidation of all surfaces exposed to ambient environment, N₂ gas was continuously flowed into the oven chamber at approximately 50 standard cubic feet per hour (SCFH).

C. Plasma Treatments on Heat Spreader

To study the effect of plasma treatment on the nickel-plated copper heat spreader, three types of samples were prepared, as listed in Table II. Expecting physical change by argon plasma and chemical activation by oxygen plasma, a mixture of argon and oxygen was used as the reactive gas for plasma treatment in the experiment.

TABLE II Plasma Conditions of Experimental Matrix

To effectively compare the mechanical changes on the surfaces induced by plasma treatment, reactive ion etching (RIE) mode direct plasma treatment was applied, whose power (PWR) electrode size was $320 \times \, 370$ mm², ground (GND) electrode size was approximately 1.7 × PWR electrode size, and distance from GND electrode to PWR electrode was 27 mm. The mimetic diagram of the plasma chamber and the changes before and after plasma treatment on the heat spreader surfaces are depicted in Fig. 5. The heat spreader surface was expected to be activated, while the functional groups that favor adhesion and surface roughness were anticipated to increase due to O${}_{2}^{+}$ and Ar⁺ plasma ions, respectively.

Fig. 5.

Mimetic diagram of (a) heat spreaders in RIE mode plasma chamber and heat spreader surface (b) before and (c) after the plasma treatment.

D. Heat Spreader Attach and Cure

Heat spreaders without and with $1 \times $ or $2 \times $ plasma treatments were attached to the flip chips on PCBs with exactly same conditions, as presented in Table III, after dispensing the same volume of TIM on each flip chip. The TIM spread easily at 145 °C due to its viscous properties and on-set point being slightly lower than 160 °C. Specific temperature conditions were applied while the heat spreader was attached and during TIM curing process. All fcBGA-H devices underwent TIM oven curing for the experiments, and the process conditions for TIM oven curing are presented in Table III.

TABLE III Process Conditions for Heat Spreader Attach and TIM Curing

E. Inspections

The nickel-plated copper heat spreaders were inspected using atomic force microscopy (AFM), C/A, and X-ray photoelectron spectroscopy (XPS) to analyze changes in their morphological, mechanical, and chemical properties before and after plasma treatment. Bare and plasma-treated heat spreaders were inspected for their surface morphology, roughness, and surface-area difference percentage (SADP) using AFM (Parks Systems, XE-100) in the noncontact inspection mode, wherein the cantilever had noncontact high frequency (NCHR) and the back-side reflex was coated. The dimension of the inspected area was $1 \times \, 1 \, \mu \mathrm {m}^{2}$ . The C/A of water droplets of volume approximately $3 \, \mu $ L on the heat spreaders was inspected using a goniometer (SEO; Phoenix-MT), with wait time of approximately 30 s after water was dropped on the surface. Before and after the plasma treatment, chemical properties such as the elements of surface and activated functional groups were inspected using XPS (Thermo Fisher, Nexsa). For all inspections, the sample size of heat spreaders was 5 ea per each condition. After the heat spreaders were attached onto the flip chips and TIMs were oven cured, scanning acoustic tomography (SAT; Hitachi, FineSAT) was performed to analyze the coverage of TIM between the flip chips and the heat spreader.

SECTION III. Results and Discussion

Before and after the plasma treatment, surface morphologies were observed using OM and secondary electron microscopy (SEM), as represented in Fig. 6. Both OM and SEM could not detect the difference between the bare and plasma-treated surfaces. The maximum resolutions applied for the experiments were 670 and 75 nm for OM and SEM, respectively. The degree of mechanical changes induced by plasma treatment could be smaller than the changes detectable by OM and SEM. To observe the mechanical changes by plasma treatment, other inspection method having finer resolution than SEM were required.

Fig. 6.

Surface of heat spreader inspected using OM and SEM as per the plasma conditions.

AFM offered finer resolution than SEM; the maximum lateral and vertical resolutions were approximately 0.2–0.3 nm and 10 pm, respectively. The surface morphology of heat spreaders inspected by AFM is presented in Fig. 7. The lateral resolution applied during the inspection was 3.9 nm, and the mechanical change by plasma treatment was evidently detected due to the fine resolution of AFM. The surface morphologies for bare, $1 \times $ (30 s), $2 \times $ (60 s), $4 \times $ (120 s), and $8 \times $ (240 s) plasma-treated heat spreaders were inspected. The bare heat-spreader surface was blunt shaped with a bigger mass, but the surface changed to a pointed with smaller mass (grit) surface as per the plasma treatment. The surface with $1 \times $ plasma condition appeared rougher than other heat spreader surfaces.

Fig. 7.

Surface of heat spreaders inspected by AFM as per the plasma conditions; (a) bare, (b) $1 \times $ , (c) $2 \times $ , (d) $4 \times $ , and (e) $8 \times $ plasma treated heat spreader surface.

The surface roughness of heat spreaders was represented by an arithmetic average of heights (Ra). The Ra values for heat spreaders are depicted in Fig. 8, where the values increased after $1 \times $ plasma treatment. However, the Ra values decreased with longer plasma conditions. The standard deviation values were 0.281, 0.565, 0.372, 0.266, and 0.512 nm for bare, $1 \times $ , $2 \times $ , $4 \times $ , and $8 \times $ plasma conditions, respectively. The SADP values of the heat spreaders are shown in Fig. 8. The values were calculated as the ratio of 3-D surface (Surface, in the following equation) to the 2-D area (Area, in the following equation) as \begin{equation*} \mathrm {SADP= }\frac {\mathrm {Surface-Area}}{\mathrm {Area}} \times { 100\%}. \tag {1}\end{equation*}

Fig. 8.

Surface roughness and SADP of heat spreaders inspected by AFM as per the plasma conditions.

Reportedly, SADP is considered more important for adhesion and fluid flow than surface roughness [19]. Both heat-spreader surfaces with $1 \times $ and $2 \times $ plasma conditions exhibited higher SADP than the bare heat-spreader surface, implying an increase in effective surfaces after plasma treatment. The standard deviation values were 0.348%, 0.393%, 0.448%, 0.101%, and 0.209% for the bare, $1 \times $ , $2 \times $ , $4 \times $ , and $8 \times $ plasma treatment conditions, respectively. When the standard deviation values for Ra and the SADP values were considered, both surface roughness (Ra) and SADP values tended saturation. Therefore, the heat spreader surface with $1 \times $ plasma condition exhibited the highest roughness and SADP value, even on considering their standard deviation values.

The C/A of water droplets on heat spreaders before and after plasma treatments were inspected using the goniometer, and the resultant surface energy was calculated using Young’s equation, which is one of the formulas that describes the correlation between C/A and surface energy; the equation is expressed as [22].\begin{equation*} \sigma _{s}=\sigma _{sl}+\sigma _{l} \times \mathrm { cos}{\theta } \tag {2}\end{equation*} where $\sigma _{s}$ denotes the surface energy of a solid, $\sigma _{sl}$ denotes the interfacial tension between the liquid and solid, $\sigma _{l}$ denotes the surface tension of the liquid, and $\theta $ denotes the C/A between the liquid and solid. For comparing C/A and surface energies, five heat spreaders were used for each condition and three points were inspected for every heat spreader. The C/A of water droplets and the surface energies of heat spreaders are presented in Fig. 9. The bare heat spreaders exhibited a much higher C/A of 71.64°, with surface energy of 22.93 mN/m being lower than that of other spreaders. After $1 \times $ plasma treatment, C/A greatly decreased to 3.89° and the surface energy increased to 72.53 mN/m. For longer plasma conditions of $2 \times $ , C/A slightly increased and surface energy decreased as compared with that of $1 \times $ plasma treatment condition. The resultant trend was identical with that of SADP obtained from AFM. On considering surface energy and SADP values, $1 \times $ plasma condition showed better adhesion performance than bare and $2 \times $ plasma conditions because the liquid employed in the goniometer was water droplets. The definable surface energy through the C/A was limited to that for water, 72 mN/m. Thus, the surface energies of the two spreader types with $1 \times $ and $2 \times $ plasma-treatment conditions had similar values and both were slightly higher than that for water. Thus, the surface energy of the $1 \times $ plasma-treated heat spreader was higher than 72.53 mN/m.

Fig. 9.

C/As and surface energies of heat spreader surface as per the plasma conditions.

The elements present on the surface of heat spreaders were inspected through XPS, and the results are presented in Table IV. The heat spreader comprised copper with its surface plated in Ni. Generally, the XPS information depth is approximately 10 nm. Therefore, no copper elements were detected on the surface, instead, carbon and oxygen were detected on all heat spreader surfaces. The ratio of carbon to other elements, C/(O+Ni), defined the degree of contamination on that surface. From the observed results, the bare heat spreader had more contaminants than other spreaders [21]. After plasma treatment, C/(O+Ni) largely decreased, but $2 \times $ plasma-treated heat spreaders exhibited a slightly higher ratio than the $1 \times $ plasma-treated heat spreaders, corresponding to the SADP value from the AFM, C/A, and surface energy. Therefore, the $1 \times $ plasma-treated heat spreader was expected to show better adhesion performance than the $2 \times $ plasma-treated and bare heat spreaders.

TABLE IV Atomic Contents on the Surfaces of Heat Spreaders as per the Plasma Conditions

The XPS survey results are depicted in Fig. 10. As per the plasma conditions, the major, C1s, O1s, and Ni2p₃ peaks exhibited slight changes. With $1 \times $ plasma condition, the intensity of the C1s peak decreased and the peak binding energy moved to a lower energy value. In case of the O1s peak, the intensity increased and the peak binding energy moved to a higher energy value. These binding energies represented chemical status and core binding, which were determined by electrostatic interactions [23]. The binding energy decreased due to the addition of a valence electron charge, such as the electrostatic shielding of nuclear charge from all the other electrons in an atom or due to the addition of an electronic charge as a result of change in bonding. The withdrawal of valence electron charge, such as oxidation, increased the binding energy [21], [23], [24]. When the binding energy decreased due to the addition of a valence electron charge in a bonding, the chemical composition could become unstable, thereby increasing the reactivity of the bonding. Accordingly, the increased binding energy in O1s spectra for $1 \times $ plasma condition could imply a more stable and less-reactive heat spreader surface than other conditions; the reason required verification as per the functional groups.

Fig. 10.

XPS survey results of heat spreader surfaces as per the plasma conditions.

The XPS narrow spectra for C1s, O1s, and Ni2p₃ are depicted in Fig. 11. The C1s spectra comprised three functional groups; C–C (carbon covalent bonding), C–O (carbon–oxygen covalent bonding), and C = O (carbonyl bonding). The O1s spectra comprised NiO (Ni²⁺-O^2- ionic bonding), O–O (oxygen covalent bonding), and O = C (carbonyl bonding). The Ni2p₃ spectra comprised metallic Ni, NiO (Ni²⁺-O^2- ionic bonding), Ni₂O₃ (Ni³⁺-O^2- ionic bonding), and satellite peaks. The amount of each functional group detected at the heat spreader surfaces is listed in Table V, and the binding energies of each functional group in the C1s and O1s spectra are presented in Table VI. The amount of detected functional groups was presented as CPS (Counts/s) values, which was similar to the representation of each functional group area in Fig. 11. Bare heat spreader exhibited higher peak heights at C1s and Ni2p₃ spectra, while that for O1s was the lowest among the measured heat spreaders. After plasma treatment, the peak heights for C1s and Ni2p₃ decreased, while that for O1s greatly increased. However, with longer ($2 \times $ ) plasma treatment time, the C1s peak height increased and the metallic Ni peak in the Ni2p₃ spectra disappeared, implying oxidation of Ni atoms at the heat spreader surface. NiO is harder than pure Ni metal and the disappearance of metallic Ni peaks supported the reason behind the saturation of surface roughness and SADP beyond $2 \times $ plasma treatment.

TABLE V Amount of Detected Functional Groups on Heat Spreader Surfaces

TABLE VI Peak Binding Energies of Detected Functional Groups in C1s and O1s Spectra on Heat Spreader Surfaces

Fig. 11.

Narrow spectra (C1s, O1s, and Ni2p₃) and functional group fit spectra of XPS results on heat spreader surfaces as per the plasma conditions (a) bare, (b) $1 \times $ plasma, and (c) $2 \times $ plasma.

The intensity of the O-C peak, whose binding energy was higher than that of other functional groups, was over four times larger than that of the bare and $2 \times $ plasma-treated heat spreaders, as shown in Fig. 11(b). The high-intensity O-C peak caused the O1s survey spectrum peak to move to a higher binding energy level for $1 \times $ plasma condition. When binding energy was considered to analyze surface activation through chemical shift, the $2 \times $ plasma condition exhibited the lowest binding energy for all functional groups excluding C–C, implying that the $2 \times $ plasma-treated heat spreader surface was the most activated surface in the experiment. However, the difference in binding energy was much lower than 1 eV and the risk of data misinterpretation was particularly high in this case [23]. Although the small binding energy difference was considered as the degree of reactivity, the amount of detected functional groups on the $1 \times $ plasma-treated heat spreader was overwhelmingly more than the other conditions.

The amount of functional groups activated on the heat spreader surfaces is shown in Fig. 12 as bar charts. As shown in Table V and Fig. 12, the CPS values for the C1s spectra decreased, and that for the O1s spectra increased after plasma treatment, implying that the plasma treatment successfully activated the heat spreader surfaces. Considering the Ni2p₃ spectra, with an increase in plasma treatment time, the amount of metallic Ni peaks detected on the heat spreader surfaces decreased. Several functional groups such as C–O and C = O could help improve the adhesion with TIM. The functional groups exhibit higher bonding energy, such as 358 and 799 kJ/mol for C–O and C = O, respectively [25], [26]. Bonding energy as 346 kJ/mol for C–C was smaller than that for others; thus, the heat spreaders having more C–O and C = O bonding rather than C–C bonding exhibited higher surface energy and adhesion with TIM because the TIM used in the experiment comprised Si-based filler type, which also included epoxy. When only the functional groups of C–O and C = O were considered, the $1 \times $ plasma-treated heat spreaders showed larger CPS values, bare heat spreaders exhibited 1.79 ×, and the $2 \times $ plasma-treated heat spreaders exhibited 1.5 ×, as shown in Fig. 12(d). Thus, the $1 \times $ plasma-treated heat spreaders exhibited better adhesion performance than other conditions of bare and $2 \times $ plasma treatment.

Fig. 12.

CPS values of each functional groups as per the plasma conditions and XPS spectra. (a) C1s spectra. (b) O1s spectra. (c) Ni2p₃ spectra. (d) Interested functional groups.

After TIM curing, three devices per each plasma condition were selected; the devices were fabricated to end-of-line (EOL) such that the BGA was soldered by MR under conditions similar to that shown in Fig. 4. Their TIM coverages were inspected through SAT, and the results are presented in Fig. 13(a)–(c). The first images for each plasma condition showed the flip chip area on the first device. In case of the devices whose heat spreaders were not treated by plasma, dark lines were observed, indicating squeezing out of TIM from the original area to flip chip area by a pressing force. The central area appeared darkish in color, indicating presence of TIM between the flip chip and heat spreader. However, the corner areas of the flip chip were much brighter than the central area, indicating delamination between the heat spreader–TIM or flip chip–TIM. In the experiment of this study, delamination between heat spreader–TIM was expected to occur much more frequently than between the flip chip–TIM due to the composing materials of TIM. The fcBGA-H devices with $1 \times $ plasma-treated heat spreaders exhibited no brighter area at the corner of the flip chip, but some of the fcBGA-H devices with $2 \times $ plasma-treated heat spreaders exhibited brighter areas at the corner of the flip chips. However, the area was smaller in size than that exhibited by the bare heat spreaders. One of the devices that exhibited brighter areas at the corner positions was cross sectioned and the resultant image is presented in Fig. 13(d). As expected, the area showed delamination between heat spreader–TIM. The SAT results showed that plasma treatment on heat spreader could be effective. As expected from XPS inspection results, activated functional groups through plasma treatment affected the adhesion property. The fcBGA-H devices with $2 \times $ plasma-treated heat spreaders exhibited unstable results on SAT, while those with $1 \times $ plasma-treated heat spreaders exhibited more stable SAT result. The resultant trend was exactly the same as the XPS results.

Fig. 13.

SAT results after heat spreaders are attached on flip chips with TIM as per the plasma conditions of (a) bare, (b) $1 \times $ plasma treatment, and (c) $2 \times $ plasma treatment. (d) Delamination trend at the corner of flip chip between TIM and heat spreader.

SECTION IV. Conclusion

In this study, the effects of RIE mode direct plasma treatment with Ar+O₂ gas were investigated by optical, mechanical, and chemical inspections. Representative nondestructive inspection methods such as OM and SEM could not detect any changes induced by plasma treatment. AFM, having a resolution much finer than SEM, detected the changes by plasma treatment, such that the surface shapes of heat spreaders changed from blunt type to a somewhat pointed shape. The results of AFM inspections and goniometer measurements of water droplets matched well, as the heat spreaders with $1 \times $ plasma condition exhibited the cleanest surfaces and largest SADP, along with the smallest C/A and highest surface energy among other conditions considered in the experiments. The chemical properties inspected by XPS showed that $1 \times $ plasma treated heat spreader had most clean surface among the test legs in the experiment. Due to the presence of O₂ in the plasma reactive gas, functional groups affecting adhesion with TIM were successfully activated on the heat spreader surfaces. With surface activation and high surface energy, the fcBGA-H devices with $1 \times $ plasma-treated heat spreaders exhibited the most stable adhesion between the heat spreader and TIM in the experiment.

The obtained results demonstrate that RIE mode Ar+O₂ direct plasma treatment on the heat spreader surfaces is effective for improving the adhesion performance of the heat spreader with TIM and increasing the heat dissipating performance of the heat spreader.