Result of high accelerated stress test of organic substrates made by integrated dry process

Semi Additive Process(SAP) is a superior process to manufacture the PCB packages. It has the feature to manufacture the PCB with reasonable cost and SAP realizes the FLS and narrow via pitch that is required by semiconductor industry rules. The recent precision requirements have become smaller while the miniaturization of the semiconductor and the shrinkage of the package size are demanded year by year.

In future FLS wiring, the current SAP may have some issues with thinned metal wiring in seed layer etching such as the connectivity between dielectric resin and Cu metal, the defects of the resist pattern depend on wrong coherency and the undercut of etching and so on.

Particularly, it is difficult challenge that the lower surface roughness on the dielectric resin to achieve the reduction of high frequency electrical transmission loss is compatible with the improvement of the coherency.

However the process of 2 microns FLS remains unusable if this challenge is not achieved.

Dry process is proposed as alternative to wet process.

Dry process is the environment friendly as it does not depend on chemical reaction and chemical solution penetration.

Wet desmear process comes to the limit in the device miniaturization recently and Dry process has been studied instead of the wet process. [1] We have developed a new fundamentally desmear process applied with Vacuum ultra violet light (named Photodesmear process) for several years. [3–7].

We have shown that Photodesmear is effective from coupon-sized samples to 500mm width samples that is the practical level, from previous studies.

Photodesmear is a smear removal process consisting of the process irradiated by the vacuum ultra violet (VUV) light and the process cleaned by the water without any damage to dielectric surface.

We have built a large VUV irradiation tool. We have successfully demonstrated Photodesmear of a 500mm square full-sized panel. We have evaluated the cleanliness of small-diameter via holes and the process uniformity in a full-sized panel (substrate).

We have confirmed the high peel strength between the dielectric and Cu plate and the roughness control of a dielectric surface. In addition, we also have discussed the mechanism of the smear removal by the VUV irradiation processing in Photodesmear [9].

In this paper, we present the investigation result of the influence that Photodesmear process gave on printed circuit board on the standpoint of the reliability. We produced a reliable test vehicle by ourselves design. The reliability test result of the test vehicle was given as a result of same level in comparison with a conventional process.

We confirmed the effectiveness of Photodesmear process by this result. With the development of the tool and the process, we have evaluated reliability of Photodesmear process. This process can respond to the market demand of future high density packages.

Fig.1 shows the mechanism of VUV (172nm) light emission in a Xe excimer lamp. This lamp consists of 2 (inner and outer) quartz tubes as dielectrics. The metal electrode is mounted both inside of inner and outside of outer quartz tubes. An area between 2 quartz tubes is filled with a discharged gas (Xe gas). Once a high AC voltage is applied to the electrodes, and a number of fine-wire discharge plasmas (it is called the dielectric barrier discharge) are generated between the two dielectrics. This discharge plasma excites the gas atoms to instantaneously produce the excimer state (Xe2*). When the excited state of atoms returns to the original (ground) state, the wavelengths specific to excimer radiation are emitted (excimer emission). The emitted spectrum can be set by the filled discharged gas. For example, if the lamp is filled with Xe gas, the lamp emits a monochromatic wavelength light having the central wavelength of 172nm. The excimer lamp has many features such as, high photon energy, single wave length, low temperature operation, instant turn on/off and is also mercury free.

Fig.2 shows the schematic mechanism of Photodesmear. The 172nm light has very high photon energy at 7.2eV which is capable of breaking bonds in organic molecular chains with molecular bonding energies lower than 172nm. 172nm photons also creates ozone from atmospheric oxygen. The ozone has strong oxidization power. Furthermore the 172nm photon energy can further break down ozone to create excited state atomic oxygen which has oxidation power exceeding that of ozone. Ozone and excited state atomic oxygen then react with C and H atom that were decomposed by 172nm photon energy from the molecular fragments. It finally creates H₂O and CO₂. [3–9].

Fig.3 shows the cross section schematic diagram of a high-power VUV irradiation chamber. The high-power excimer lamps which were newly developed for this tool are lined multiple so as to cover the irradiation region.

A substrate is placed on the work stage which has the heat control system.

The process chamber is sealed and kept the controlled temperature and filled with the process gas.

The substrate catches VUV irradiation from high-power lamps via the window glass in the process chamber.

The test vehicle has two layer structures. The substrate material we use for this experiment is 0.8 mm thick glass/epoxy core with CZ treated Cu film. At first we formed the pattern of the lower level by lithography. ABF (GX-T31, 25μm, produced by Ajinomoto Fine-Techno Co., Inc.) is laminated and CO₂ laser (LUC-2K21/1C, manufactured by Via Mechanics, Ltd.) is used to prepare blind via holes about 100μm diameter with trough the polyethylene terephthalate film.

The substrate is set on the work stage and exposed by a high-power vacuum ultra violet light. After the exposure processing, the substrate is taken out and ultrasonic cleaning with deionized water to remove the remaining filler or the like inorganic materials [5]. We formed titanium/copper seed layer on thickness of 100/300nm by sputter tool (TOPAZ, manufacture by Tango Systems Inc.). And we conducted electrical copper plating on thickness of 25 microns.

After lithography process and resist strip process, we conducted to soft etch for titanium seed layer in large sized experimental bath (Melstrip TI-3991, produced by Meltex Inc.). The etching condition was selected with Research Center for Tree-Dimensional Semiconductors.

In another process, the sample boards were produced at a cause public-like printed circuit board production line of the cooperation of OK PRINT Corporation Inc.

We designed the daisy chain pattern for a reliability test. It was connected with laser via between layer 1 and layer 2 wiring across the dielectric resin. As wiring pitch is 1000 micron (100 micron line width), the pad diameter is 400 micron. The total wiring length is 400mm (Fig 4).

Blind via hole is located in 20mm square area. The test pattern comprises a terminal for resistance measurements every 40 via to evaluate uniformity of the electrical connectivity.

We designed the comb formed pattern for an electrical migration test. The electrode of tooth of the comb is patterned 60 micron pitch in turn. It is formed in layer 1 on dielectric epoxy resin.

The substrate after Photodesmear, is applied a dry seed layer of Ti and Cu using a sputtering (TOPAZ, manufactured by Tango Systems Inc.). After the dry seed sputtering, 25μm Cu was electric plated. We pulled out one piece of sample board from trial lot. The peel strength is measured using a peel tester (EZ-LX, manufactured by Shimadzu Corporation).

We made cross section sample for optical microscope and SEM. We observed the interface between copper pad and copper filled via.

We evaluate reliability of test vehicle under two conditions. The first is high temperature and humidity chamber test with electric bias. The other is high accelerated temperature and humidity stress test.

We put a test vehicle in constant temperature and humidity chamber (Type PR-2K, ESPEC CORP.) of 85 degree C and related humidity 85 % at atmospheric pressure for 168 hours to evaluation reliability of the via. We loaded DC 12V/100mA at the both end of daisy chain pattern.

And we used high accelerated temperature and humidity stress test tool (HASTEST PC-242HSR2, HIRAYAMA MANUFUCTURING CORPORATION).

We measured the resistance of the daisy chain in digital-ohm tester (VOAC7521A manufactured by IWATSU ELECTRIC CO., LTD.) in before and after high temperature and high humidity test.

We put a test vehicle in high accelerated temperature and humidity stress test tool (HASTEST PC-242HSR2, HIRAYAMA MANUFUCTURING CORPORATION) to evaluate electrical migration. The condition is 85% humidity, 130 degree C, atmospheric pressure 230kPa, 168 hours. During the test, we look bias of DC20V and always monitored it by migration tester (Model MIG-87B, manufactured by IMV CORPORATION).

We evaluated the connectivity between Cu pat and Cu electroplating in Via bottom by using SEM/EDX (Model JSM-IT300HR, JOEL) and XPS Analysis (Model PHZ Quantera SXM, ULVAC-PHI, Inc.).

We got 6 pieces of test vehicle of 140mm square from 510mm by 340mm of printed circuit board (Fig. 6). As for these reliability tests, we produced a reference sample by conventional processing at same time to become the competitive test.

Figure 8 is SEM image of the cross section of the daisy chain. We could observe that layer 1 and layer 2 connected with copper filled via (Fig. 9).

Fig.9 shows the cross section SEM image of Copper Filled Via.

It can be observed that the status of connectivity between the pat and the electroplating is well.

Fig. 10 shows the cross section SEM image of electrical connection interface of via bottom and each EDX mapping data of Carbon, Silicon and Copper. Existence of each silicon element and carbon element (smear components) are not detected in the interconnecting region between the copper electroplating and the copper pad. EDX mapping data of the titanium is shown from the seed layer between the copper electroplating and copper pad.

We conducted the peel strength test to see the adhesive strength between the copper electroplating layer and the dielectric layer after copper electroplating.

We conducted this test in 9 point on the panel.

Fig. 11 shows the test result.

The mean peal strength was good result (average 5.72 N/cm).

Fig.12 shows an example of the evaluation of the via connectivity result.

The vertical axis means resistance value and the horizontal axis means the position of the resistance measurement between the pads of daisy chain.

Fig.12 shows that relation between the resistance and the pad position is linear on red line.

It means that the electrical connection of the copper pad is located with uniformity in the daisy chain circuit.

The red line shows the line on Fig.12 before high temperature and high humidity chamber test.

The blue line shows the line on Fig.12 after the test.

Both red and blue line are almost same line on graph.

The change of the resistance was only 0.15% in 168 hours on this test.

85 degree C / 85 % related humidity at atmospheric pressure and DC12V bias

The table 1 shows the comparison table between Photodesmear test vehicle sample and the conventional desmear sample on the test result after the reliable test.

Table 1 compare the electrical resistance value before and after the reliable test in this evaluation.

The difference between the electrical resistance value before test and that of after test is 0.15% on the sample made by Photodesmear process

On the other hand, that of conventional wet desmear process is 0.19%.

It means that the reliability of Photodesmear process is not inferior in comparison with conventional wet desmear process on this reliable evaluation.

130 degree C / 85 % related humidity at 230kPa pressure for 168 hours

The table 2 shows the comparison of the high accelerated temperature and humidity stress test between test vehicle and reference sample data.

The electrical resistance of the test vehicle treated by Photodesmear process a little increase by the titanium layer or by becoming small diameter of via in comparison with that of the conventional wet process sample.

We believe that this difference does not serious influence for the reliability.

However it is important for us to grasp these results.

Figure 13 shows a result of the reliability test (electro migration test). The test vehicle was examined in high accelerated temperature and humidity stress test tool under the condition of 130 degree C, related humidity 85%, pressure 230kPa, 168 hours.

The test vehicle has the comb-tooth shaped circuit pattern. This circuit was applied with DC20V during the test.

Figure 13 indicates the variation of the electrical resistance during the test.

The comb-tooth shaped circuit is basically the open circuit.

The electrical resistance should be infinity or more than 1E06 ohm.

It means that the circuit pattern was broken by the electro migration if the electrical resistance became short circuit during the test.

The test vehicle treated by Photodesmear process could keep high resistance during the test.

We understand Both Conventional Wet dsmear and New Photodesmear method oxidize the copper surface when processing.

Oxidized surface on Cu pad influence the electro connectivity between Cu pad and Cu electroplating.

Therefore, we evaluated oxidization status on copper for each process using by XPS analysis.

Figure 14 shows the peak signal of the oxygen atom on the copper surface after both desmear process by XPS analysis.

Left side graph shows the copper surface treated by Photodesmear process. Right side graph shows that of the conventional wet desmear process.

Both graph has 7 signals. Top signal shows to analyze the existence of the oxygen atom in the depth of 18.9nm from the copper surface. In this case, No oxygen atom exists in the depth of 18.9nm because the signal is flat.

Second signal is that of 12.6nm. Third signal is that of 6.3nm. Fourth signal is that of 3.2nm. Fifth signal is that of 1.6nm. Sixth signal is that of 0.8nm. Seventh signal is that of 0nm (Surface).

Top signal of both Photodesmear process and the wet desmear process is flat. It means that no oxygen atom exist in the depth of 18.9nm.

No oxygen atom exists in the depth of 12.6nm on both process similarly.

Third signal of Photodesmear process shows no existence of oxygen atom in the depth of 6.3nm.

However third signal of the wet desmear process shows the existence of oxygen atom in the depth of 6.3nm.

Fourth signal of Photodesmear process also shows the oxygen atom exists in the depth of 3.2nm.

It means that Photodesmear process oxidizes the copper in depth of 3.2nm from the surface.

On the other hand, it means that the wet desmear process oxidizes the copper in depth of 6.3nm from the surface deeply in comparison with that of Photodesmear process.

It means that the etching process to remove the oxidization layer of the copper can be shorted because the oxidization layer of the copper that is treated by Photodesmear process is thinner than that of the wet desmear process.

The summary of this report is as described below,

• We understands that it is possible to replace conventional desmear process and seed layer formation process with Photodesmear process.

• We understand that the test vehicle treated by Photodesmear process can maintain the quality in the production.

• We confirmed that Photodesmear process can maintain a cleanness of via bottom and high peel strength between epoxy resin and copper electroplating.

• We confirmed that the test vehicle treated by Photodeesmear process can maintain an uniformity and connectivity of via by resistance measurement of the daisy chain consisting of two layers.

• There were a few resistance changes before and after the high accelerated temperature and humidity stress test.

• The result of electrical migration of the test vehicle treated by Photodesmear process was good.

• Photodesmear process oxidizes the copper surface.

• The oxidization layer on the copper surface treated by Photodesmear process is thinner than that of the oxidization layer on the copper surface treated by the conventional desmear process.

As mentioned above, we have made a test vehicle which was treated with the Photodesmear process. The conclusion after accelerated temperature and humidity stress testing is that the Photodesmear process has a high level of reliability.

We have verified high peel strength, via electrical connectivity, evaluating electrical migration combining Photodesmear process and dry seed process such as next generation technology.

Semiconductor packaging market goes to more cost reduction in manufacturing process. In future, the trend of miniaturization of semiconductor packaging process faces the threshold on the current process. On the other hand, the market keep to need to a fine patterned panel level packaging. The appearance of the new material, new method and new tool are expected.

Photodesmear process does not use harmful chemicals. It is expected as a next generation technology because of environmentally friendly process.

We are going to optimize the process and evaluate a printed circuit board with a silicon tip. We will continue to investigate the dry process technology for economy and reliability.

Finally, we would like to say that a cooperation of Ajinomoto Fine-Techno Co., Inc., Tango systems Inc., OK Print Corporation, Research Center for Tree-Dimensional Semiconductors are highly appreciated on this paper.