New Challenges for Higher Aspect Ratio: Filling Through-holes and Blind Microvias with Copper by Reverse Pulse Plating

Estimated reading time: 17 minutes

Editor’s Note: This article was originally published in the Proceedings of IPC APEX EXPO, San Diego, California, February, 28-March 1, 2012 and in the September 2012 issue of The PCB Magazine.

Abstract

This paper presents systematic investigations on complete through-hole filling for cores by a Cu electroplating process as an alternative to the common paste-plugging process. This electroplating process consists of two steps, a first process to merge both centers of the through-hole walls (X-plating) followed by filling up the resulting blind microvias. Processes and manufacturing technology are described as well as current limitations and requirements. Complete filling of through-holes is achieved by reverse pulse plating (RPP). This through-hole filling technology is targeting both at HDI production and also at the packaging level.

Through-hole filling by RPP offers a viable alternative to the standard paste plugging for core processing in substrate manufacturing. Current core manufacturing requires a paste plugging process for through-holes so that subsequent build-up layers can be produced by sequential lamination, and the flat core surface is essential for stacked via and also via-in-pad technology.

This paste plugging process requires additional process steps, each of which brings its own limitations and contributes to the overall cost. Filling the core through-vias by electroplating can eliminate the plugging process and significantly reduces the number of overall process steps which will also reduce costs. Moreover, it offers certain advantages such as potentially higher reliability in accelerated aging tests and an improved thermal management as the thermal conductivity of a completely copper-filled through-via is significantly higher than a paste-plugged through-via.

Today’s challenges in the so-called through-hole filling process are represented by through-holes with a high aspect ratio. Voids after X-plating occur easily for smaller through-hole diameter and higher board thickness. In addition, depending on designs, different pitches on one board increase the difficulty to achieve an acceptable plating uniformity. This paper presents systemic variations of some key parameters and concentrates on the performances of the second plating step, the blind microvias filling. The main focus is laid on recess distribution. Parameters as reverse pulse parameters, inorganic concentrations (Cu, Fe, and sulfuric acid), organic concentrations, electrolyte flow and temperature have been systematically varied and their influence on the filling performance are described.

Introduction

Through-hole filling by RPP offers a viable alternative to the standard paste plugging for core processing in substrate manufacturing. Current core manufacturing requires a paste plugging after Cu galvanization for through-holes so that subsequent build-up layers can be produced by sequential lamination, and the flat core surface is essential for stacked via and also via-in-pad technology.

Instead of plating only a conformal Cu barrel into the through-via, it is now possible to completely fill it with Cu and thus eliminate the need for the plugging process, reducing significantly the number of overall process steps and costs. Moreover, it offers certain advantages such as potentially higher reliability in accelerated aging tests and an improved thermal management as the thermal conductivity of a completely copper-filled through-via is significantly higher than a paste plugged through-via.

Today’s challenges in the so-called through-hole filling process are represented by through-holes with a high aspect ratio. Voids after X-plating occur easily for smaller through-hole diameter and higher board thickness. In addition, depending on designs, different pitches on one board increase the difficulty to achieve an acceptable plating uniformity. To understand the limitations of the process we are aiming at an identification of the key parameters. As there are two steps in the through-via filling process, we divided our experimental work accordingly. The first section presents an overview over the process itself (Figure 1), the second section concentrates on the formation of the X (merging through-via walls) and the third section deals with electrochemical filling of the resulting blind vias. Most of the tests have been conducted in a single board cell (volume 400 l), i.e., each plating cell is handling exactly one 18” x 24” sample board. Different test panels have been used and are described within the experiment itself. All panels have been electroplated in the so-called panel plating mode, i.e., there were no dry film patterns on the board during plating. This panel plate technology is being used in standard production processes by applying the resist and the conductor pattern in a subsequent step.

General Description of Through-Via Filling

Table 1 describes the process flow of the electrochemical through-via filling technology

Electrochemical  Cu Through-via filling

1. Drilling

1. Seed Layer Deposition ( E’less Cu 0.2 -0.5 µm)

1. Galvanic reinforcement (2-5 µm Cu)

1. Galvanic Cu Filling 

-X-Plating  ( 5-10 µm)

-Via filling  (10-15 µm)

1. Dry Film Application and subsequent Structuring

Table 1: Process steps characterizing the EC Cu through-via fill technology.

The EC plating process described in Figure 1 shows the two main plating processes of step 4 above.

Figure 1: Plating steps to a complete fill of a through-via with Cu, starting from a conformal plated layer. The second plating step is the formation of the X, resulting in two blind vias on both sides, which are then completely filled.

As alternative electrochemical route, one could consider a slow, conformal growth of copper on the side walls to a complete fill. That process has several disadvantages such as being slow (less productivity) and having a higher risk for voids (see Figure 2).

Figure 2: Conformal Cu growth to a complete fill may end up producing a long, ribbon-like void.

Moreover, this conformal technique would require too much Cu on the surface (approximately ½ hole diameter), whereas the above mentioned X-formation offers the possibility to significantly reduce the Cu overburden. That is achieved by applying a combination of RPP and etching in one electrolyte. The electrolyte contains not only Cu, but also Fe2+ and Fe3+. Cu is deposited as usual on the cathode (panel) whereas the Fe2+ ions are used to carry the current at the dimensionally stable anodes (DSA) and prevent oxidation of organic additives. The oxidation product Fe3+ is being used to dissolve the consumed copper chemically in a second compartment filled with pure Cu balls/clippings and reduced again to Fe2+. But the Fe3+ cannot be used only to replenish Cu, but also to lower the amount of the Cu deposition on the panel surface.

Cu2+ +2e- à  Cu         E = 0,340 mV vs. SHE             (I)

2 Fe3+  + 2e- à 2 Fe2+         E = 0,771 mV                   (II)

Reaction (II) is electrochemically favored against (I) and thus the amount of Fe3+ reduces and determines the amount of deposited Cu. A control unit between plating and dissolving tank regulates the exchange between both compartments and can control the amount of Fe3+ within ±0.1 g/l, which is a prerequisite for a good process control as can be seen later in this report. The exchange of electrolyte is much stronger on the surface than in a via, so that this effect is mainly observed on a surface and not in vias--especially not in a blind vias.

With this technology we were able to reduce the amount of Cu overburden for a 100 µm thick core from 50 µm down to 13 µm (Figure 3).

Figure 3: Development of through-via filling technology: Reduction of Cu overburden for a 100 µm thick core.

This reduction was a prerequisite for a production scenario since the Cu thickness determines in the end the achievable lines and spaces in the circuitization process of approximately 2 x Cu overburden:  i.e., 13 µm --> 26 µm l/s.

Nevertheless, the main core thickness for IC substrates is  >100 µm. Most of today’s products show cores of 400 µm thickness or more. These cores are extremely difficult to fill without void formation and require a long plating time. So it was the aim of this work to expand the technology to such cores by determining main parameters and improve each step of the process.

Experimental

We are using several RPP parameters in the following chapters and would like to explain notation.

Example: 5/40  80/2 -->

5 = Forward or cathodic current density (ASD)

40 = Reverse or anodic current density (ASD)-  

80 = Cathodic plating time (ms)

2 = Cathodic plating time (ms)

So 5/40 80/2 is an 80ms cathodic pulse with 5 ASD followed by a 2ms anodic pulse of 40 ASD.

X- Plating

As there are a huge number of parameters, we started a first series of experiment in a small lab tool to determine which of the following parameters might have a significant influence. We used a small 20 L test cell, 400 µm thick core panels with 100 µm hole through-holes and varied in several small DOEs of three parameters. We found that the parameters in italic showed a significant influence determined by statistical software tools and set up further experiments in a production tool on 400 µm panels with through-hole of 100 µm diameter.

• Current density and pulse parameters
• Fe3+ concentration
• Temperature
• Cu concentration
• Cl- concentration
• Electrolyte agitation
• Fe2+ concentration
• Sulfuric acid concentration
• Brightener (accelerator) and leveller (inhibitor) concentration (commercially available inpulse electrolytes) 

The nature and type of additive has a big influence on the plating performance, but in this specific case, chosen ratio of brightener and leveller does not have a significant influence. We also investigated the influence of the panel design on the X-formation, as it turns out that isolated and clustered through-holes (small pitch, many neighboring through-holes) may show different results as depicted in Figure 4.

Figure 4: Difference between isolated and clustered through-holes. Left: isolated, right-clustered TH with 200 µm pitch, middle through-holes at the edge of such a cluster.

Parameters for this experiment at 40°C were:

• Cu: 25 g/l
• H2SO4: 200 g/l
• Fe2+: 15.0 g/l
• Fe3+: 0.5 - 1.0 g/l
• Cl-: 100 mg/l
• RPP: 1.5/40 – 80/4

Commercially available additives were used according to data sheet. The differently located through-holes required different plating times to form the bridge in the middle. So the minimum time to form the X is always determined by the clustered through-holes.

Electrolyte Flow

The difference between these clustered and isolated THs could be minimized by increasing the electrolyte flow onto the panel. The flow direction was 90°C (rectangular) to the surface. Doubling the flow rate resulted in a homogenous X-formation with little difference. The effect of Fe3+ is even stronger as can be seen from Figure 5. Here, we varied the Fe3+ concentration only from 0.5 g/l up to 1.0 g/l. The other electrolytic parameters were, Cu: 50 g/l sulfuric acid. The result is that with a higher ferric concentration a much slower X- formation is observed. Both experiments have been conducted under identical parameters. The consequence of this result is that the control of ferric ions is extremely important for handling the process. Therefore, we installed the control unit between the plating and the dissolving tank so that the ferric ion concentration can be controlled within ± 0.1 g/l.

Figure 5: Effect of ferric ions on the X- formation. Top 1.0 g/l and bottom 0.5 g/l ferric, left isolated and right clustered through-vias.

The effect of Cu concentration on the X- formation is different to the usual blind via filling: In contrast to the latter process a lower Cu concentration turned out to be beneficial. The influence is significant both for the X- formation and for the distribution of the X between the isolated and clustered TH. An additional test run with 18 and 32 g/l Cu on very thick panels (3.2 mm with 0.3 mm hole diameter) confirmed the result. With these panels it is not possible yet to achieve through-via filling but we did not aim at a complete X- formation since differences between experiments can better be evaluated before the X- is finally formed. We stopped plating after a certain time period (55 min.) and measured the thickness in the middle of the hole compared to the surface. With the 18 g/l we increased the Cu thickness in the middle of the hole by a factor of 1.9.

The same effect had the increase of temperature from 20 to 40°C.

Summarizing our findings for the optimization of X-formation results in following proposals:

• Use strong reverse pulse. A 5/50 80/4 for 15 min leads to closure of the X- formation on 400 µm cores with 100 µm diameter
• Use lower Cu concentrations of 25 g/l for regular  distribution
• Use higher Cl- concentration 90 instead of 60 mg/l
• Use a strong electrolyte flow to minimize differences between clustered and isolated TH
• Use low ferric concentrations (0.5 g/l)
• Above mentioned parameters were used to plate a 400 µm/100 µm core as depicted in Figure 6. Additional parameters were 240 g/l sulfuric acid and 5 g/l Fe2+ added as ferrous sulfate.

Figure 6: 400 µm thick core with 100 µm TH plated within 15 minutes. Right detail showing 9 µm plated Cu on the surface while >50 µm Cu plated in the middle of the via.

Blind Via Filling

Most of the following experiments have been performed with boards having three different blind microvias with a depth of 40 µm and width of 75, 100, and 125 µm. These sample boards have been chosen because of their well-defined and reproducible via structures. Information gathered on these boards was then applied on through-vias to achieve the best possible result. All tests have been conducted with an inpulse electrolyte system (Cu/Fe) with the same commercially available additive system.

The following parameters have been investigated:

• Reverse and forward pulse time and current
• Temperature
• Electrolyte flow
• Fe3+ concentration.

The Cu concentration has not been investigated in this series due to the fact that its effect is already well known: A high Cu concentration is beneficial for via filling.

Reverse and forward current: The 5/40 80/2 parameter set has been chosen as a start point. The electrolyte was set up with 50 g/l Cu, 150 g/l sulfuric acid, 63 mg/l HCl with inpulse brightener (accelerator) and leveller (inhibitor) and plated at 45°C

Increasing the reverse time has a significant effect on the filling performance, as it changes from a deposit with a recess of a few µm into one with a convex dome on top of the via. As expected, the smaller vias are easier to fill than the larger ones. Shortening the cathodic pulse while keeping the anodic pulse constant, leads to a similar result. This effect is in line with the explanation that via filling is mainly achieved by a changing ratio of accelerating and inhibiting additives in a via and surface. As described elsewhere [3], the Cu accelerating agents are depleted during a reverse cycle mainly on the surface, but not in a via. There, a net accumulation, and thus, acceleration is observed. Net accumulation of accelerator in a via compared to the surface can be achieved either by intensifying the depletion rate from the surface (stronger anodic pulse) or by increasing the frequency of the anodic pulse.

Best results are achieved with lower forward current densities as it is expected. From via filling experiments in pure DC mode it is well known that lower current density usually leads to a better filling performance. This effect is also seen under reverse pulse conditions.

Electrolyte Agitation

The influence of electrolyte agitation on the filling performance has been investigated by using frequency controlled pumps that feed the spray bars. The electrolyte is pumped through nozzles directly onto the board (90° angle). We used the same electrolyte set up as before. The frequency is then changed during the plating procedure, as it turned out that this procedure gave the best performance. At first, a thin protecting Cu layer with moderate pulse parameters is deposited and then followed by the modification of the electrolyte agitation. The higher the pump frequency, the higher the agitation rate. At the end of the cycle again a moderate pulse layer is plated to achieve a smooth surface.

In general, one can observe with lower agitation rates a better via filling performance. With increasing pump frequency, significant dimples are observed so that the best results are achieved with low agitation rates.

Temperature Effects

Since the dimple performance is depending on the adsorption/desorption ratio between accelerator and inhibitor, one would expect a certain influence of the temperature. Therefore, we varied the temperature over a range from 25-50°C under the same electrolyte setup as before and used the best pump frequency program. As it turns out, there is an optimum temperature of about 35°C, but the influence of this parameter is not as strong as both aforementioned ones (pulse and agitation).

The panel surface is also affected. The surface becomes rough below 35°C and very smooth above 40°C, so that depending on requirements a compromise has to be found between good via filling and surface appearance.

The Fe3+ influence: It can be expected from equation (I) and (II) that the overall yield for the copper deposition is strongly affected by Fe3+. Figure 7 depicts this very strong dependency. This current yield is measured by weight difference against the theoretically possible Cu mass of 100%. A lower current yield than 100% usually needs to be counterbalanced by additional plating time. This is not necessarily the case for the via filling application since the amount of plated Cu on the surface is not equivalent to the amount in a via. The relevant parameter is the dimple, not the amount of plated Cu on the surface. Hence, the amount of plated Cu can vary with the Fe3+ concentration to achieve the best performance. The experiments shown in the graph 13 were conducted with aforementioned electrolyte, pulse and pump parameters at 45 °C.

Figure 7 (top): Yield (Cu) as a function of Fe3+ concentration. Figure 8 (bottom): Plated Cu thickness (columns) and dimple (red line) as a function of  Fe3+ concentration  for 75 µm x 40 µm blind microvias.

As it turns out, dimples are lowest for lower ferric concentrations. Contrary to this is again the surface appearance of the boards. The surface becomes smoother with increasing ferric concentrations. Again a compromise between dimple performance and surface appearance has to be found.

Summarizing the optimization experiments to fill blind vias leads to the following results:

• Apply CD of about 3-4 ASD cathodic and apply stronger reverse parameters to decrease the dimple (on cost of surface roughness): best compromise 3-4/40   80/4 via filling is achieved within 24 minutes
• Decrease electrolyte agitation for a smaller dimple
• Apply  low Fe3+ concentrations, depending on required surface roughness
• Work at 35-40°C to achieve best dimple 

Summary and outlook

Aim of the above presented optimization experiments was to identify the best possible parameter sets and combine it for through-vias thicker than 100 µm, especially for 400 µm thick cores. For some parameters a compromise between dimple and surface roughness had to be chosen.

This test series showed that for a successful fill of thicker core as 400 µm, a split into different electrolyte types is necessary with very different parameter sets.

Parameter

X-formation

Via fill

Pulse parameter

Strong reverse

Strong reverse

Overall current density

3-4 ASD

3-4 ASD

Temperature

40°C

35-40°C

Cu 2+

Low 20-30 g/l

High > 40 g/l

Fe3+

Very low 0.5g/l

> 1.5 g/l

Electrolyte agitation

Strong

Low

Time

Approx 15 min.

40-60 min.

Table 2: Process parameters for both TH fill steps.

At this moment we can successfully plate and fill through-vias for core with thickness of 300 µm or less within one hour and without occurrence of voids. Some examples are given in Figure 9.

Figure 9: Filled vias: 150 and 200 µm cores that have been plated with 22 resp. 23 µm Cu to achieve less than 10 resp. 5 µm dimple.

On the other hand, 400 µm cores show some voids still. These voids are usually small (10 µm or less) (Figure 10).

Figure 10: Completely filled TH showing small voids.

Investigations on the effect of these voids are under way and further planned. Especially, reliability measurements will be performed, which should show whether the voids will lead to cracks or stay inside the via as they seem to be perfectly encapsulated. Whether this technology can be expanded over 400 µm cores is still unclear. There is a great potential for completely filled Cu vias as thermal vias since the thermal conductivity of Cu is several magnitudes higher at every paste. But the greatest hurdle for this application is the use of very thick panels (>1mm) which require huge amounts of Cu, and thus long process times.

References: 1. M. Carano, “Via Hole Plugging Technology,” CircuiTree, May, 2007.2. Stephen Kenny and Bernd Roelfs, “Copper Electroplating Process for Next Generation Core Through-Via Filling” Proceedings of the Pan Pacific Microelectronics Symposium, Big Island Hawaii, February 10-12, 2009.3. Stephen Kenny and Bernd Roelfs, "High Aspect Ratio Package Core Production with Electrolytic Deposited Copper” IMAPS 2010 - Research Triangle, 43rd International Symposium on Microelectronics October 31-November 4, 2010.

  Mike Palazzola has more than 30 years experience in the PCB industry, and is currently a product manager for the electroplating and surface treatment product lines at Atotech USA. He has recently presented several papers at IPC conferences. Mike would like to acknowledge his co authors and colleagues: Nina Dambrowsky, Christof Erben, Stephen Kenny, and Bernd Roelfs from Atotech Deutschland GmbH.