Issue



Low cost AlN substrate technology for HBLED and power semiconductors


04/01/2012







JONATHAN HARRIS, CMC Laboratories, Tempe, AZ


A technology has been developed that allows AlN to be sintered at lower temperatures. This allows the material to be sintered and flat fired in a continuous furnace very similar to furnaces used for alumina.


Packaging requirements for high brightness LED (HBLED) technology is pushing the current material envelop for both low cost and high thermal performance. The desire to shrink package size is driving LED substrate requirements toward higher and higher heat dissipation. And the commercial imperative to decrease the $/Watt figure of merit for light output is putting cost pressure on the LED packaging technology to utilize lower cost substrate alternatives.


HBLED devices are bonded to a ceramic "tile" that consists of a ceramic substrate that has been metallized with thick-plated copper. Connection between the top surface with the active device and the backside, which is surface-mounted to a high thermal conductivity metal core printed circuit board, is accomplished with Cu-filled vias. Thus heat conduction from the active device occurs through both the Cu vias and the ceramic. The ceramic material provides electrical isolation between the different polarity inputs that drive the LED.


Traditionally, 96% Al2O3 has been used as the ceramic substrate in HBLED applications because of its low cost and good mechanical stability. However, with a thermal conductivity of only 20 W/m-K, alumina does not contribute significantly to heat transport in the tiles. This brings in the opportunity for using other ceramic materials with higher thermal performance such as AlN or Si3N4.


The downfall for both of these alternatives has been much higher cost than alumina.


Aluminum Nitride


Aluminum Nitride (AlN) is a polycrystalline, high melting temperature (refractory), ceramic material with an advantageous set of properties for die level packaging of high brightness LEDs and power semiconductors. These critical properties include:


??? Good electrical insulation


??? High thermal conductivity


??? High flexural strength


??? Stable up to very high temperature


??? Able to be laser drilled, metallized, plated and brazed


A more detailed list of properties is shown in Table 1 below.



As power densities of semiconductor devices increase, the need for packaging to remove generated heat increases, particularly for devices such as LEDs, which are sensitive to increasing temperature. AlN, which has a thermal conductivity that is 8-9 times higher than competitive materials such as Al2O3, becomes an excellent technical solution to increasing thermal demands on first level packaging materials.


Applications with high and increasing thermal demand include: RF power components for cellular infrastructure, HBLED, power semiconductors for motor control, packaging for highly concentrated photo-voltaic installations, and packaging for semiconductor lasers used in telecommunications.


AlN ceramic substrates are typically 15 to 60 mils thick, and up to 4.5" square (larger for some specialized applications). These substrates are fabricated using conventional ceramic processing technology. A typical fabrication sequence is given in Table 2 below.



As evident from the brief discussion above, AlN has a range of very beneficial properties for high thermal demand applications. However, there is one very key drawback of AlN which has limited its utilization. The key issue is the cost of AlN substrates relative to lower performance materials such as alumina. Typically, AlN costs 5-7 times more than alumina on a cost/square inch basis.


Below is a list of the key contributors to this higher cost structure:



  1. Currently available AlN powder is approximately 20 times more expensive than alumina powder of comparable quality (purity, particle size).

  2. AlN tape must be fired in a non-oxidizing atmosphere. This means that binder removal, which is typically done through oxidation, must be done in a separate furnacing step (at a temperature well below the sintering temperature). A thick film continuous furnace can be used. For alumina, binder removal can be accomplished in the sintering furnace in one furnace step.

  3. AlN is sintered in a batch furnace with much lower throughput than continuous furnaces used for alumina. In addition, these batch furnaces are constructed using Mo and W metal heat shields and heating elements because of the extremely high sintering temperatures (>1800??C), so the overall furnace cost is very high.

  4. AlN can also be sintered in graphite batch furnaces. Though lower capital cost than W furnaces, the sintering fixtures for this type of furnace are very high cost and the throughput is still low due to batch processing. Also, the interaction of AlN with the carbon containing atmosphere is a graphite furnace must be limited to produce high quality product.

  5. The considerations of furnace cost and low throughput for sintering are also a factor for flat fire, so there is essentially a "double hit" for using batch processing.

  6. Alumina can be processed in an aqueous environment. This makes the tape fabrication less expensive than the AlN process which must utilize non-aqueous solvents. This is a significant factor for tape casting.


The focus of this article will be to discuss a new technology that has been developed at CMC Laboratories, Inc. which addresses items 3, 4 and 5 in the list above. This new technology allows AlN to be sintered at lower temperatures which allows the material to be sintered and flat fired in a continuous furnace very similar to furnaces used for alumina.


HBLED grade AlN


Table 3 below compares key properties for the low temperature sintered, lower cost "HBLED Grade" AlN compared to the standard, high temperature sintered, higher cost material that is currently commercially available.



It is clear from this graph that all of the properties are very similar, except that the thermal conductivity of the HBLED grade material is about 24% lower, but is still 6+ times higher than alumina. This makes the HBLED grade material suitable for all but the highest thermal demand applications for AlN.


HBLED grade AlN is made with the same basic processing steps outlined in Table 2 that are used for the high temperature material. The key difference is the sintering additives which allow the material to densify at 1675-1690??C as compared to the conventional 1820-1835??C. Tape binder formulations, tape casting conditions and the binder burn out process are also the same as, or very similar, to conventional material.





Figure 1. Low temperature sintered AlN substrate.



Fig. 1 shows a picture of a 4.5" x 4.5" x 20 mils substrate made from HBLED grade material that was fired at 1690??C in a nitrogen gas atmosphere with a hold time at sintering temperature of 3 hours.


Sintering aids for AlN ceramics perform two key functions: (1) they form a liquid phase at the sintering temperature which increases the rate of densification ("Liquid Phase Sintering" process); and (2) they getter oxygen from the AlN grains during sintering. Since the oxygen content of the AlN grains controls AlN's thermal conductivity, effective oxygen gettering is key to achieving the highest possible thermal performance.


The sintering temperature must be high for two reasons for the. First, the temperature must be high enough to melt the additive phase to form a liquid which enhances the rate of sintering by orders or magnitude. Second, the temperature must be high enough so that oxygen can diffuse out of the AlN grains during sintering to enhance the thermal conductivity of the AlN ceramic.


There is a third critical requirement for the additive phase during AlN sintering. While a liquid, the Y-Al-O phase will completely surround each AlN grain. If we define a wetting angle between the AlN and Y-Al-O measured at the 3 grain junctions, the microstructure has a very low wetting angle that is less than 60???. This type of microstructure is shown in the SEM micrograph in Fig. 2a. The dark grains in this figure, which are about 10 microns large, are the AlN. The bright phase is the Y-Al-O.







  

Figure 2a. Wetted microstructure- high temperature AlN


 


Figure 2b. De-wetted microstructure- high temperature AlN

There are two critical performance issues with a wetted microstructure. First, because AlN fracture is inter-granular, the presence of a Y-Al-O phase between the grains lowers the tensile strength of the ceramic by a large factor. The second problem is that a wetted microstructure results in Y-Al-O covering large portions of the surface of the substrate. This reduces the consistency of AlN metallization processes.


So a key requirement for the oxide second phase during AlN sintering is that the oxide phase de-wet the ceramic grains during the later stages of the sintering process so that the final microstructure will have a de-wetted Y-Al-O phase as shown in the micrograph in Fig. 2b.


??? These same basic considerations for sintering of high temperature, conventional AlN are relevant to designing a low temperature sintering process:


??? The sintering additive must melt at the sintering temperature to facilitate liquid phase sintering kinetics.


??? The temperature must be high enough for oxygen to diffuse out of the AlN grains during sintering. This consideration puts somewhat of a lower limit on how low AlN can be sintered to produce high thermal conductivity.


??? The liquid phase must de-wet from the AlN grains after densification to form a de-wetted microstructure and thus high flexural strength.


??? This de-wetting is also required to produce ceramic with high electrical resistivity





Figure 3. Microstructure of AlN, sintered at 1675??C.



Fig. 3 shows the microstructure of a low temperature formulation that was fired at 1675C. This has a modified sintering additive package which will melt at much lower temperature than the conventional Y-Al-O additives, but still has a strong chemical driving force to getter oxygen from the AlN grains.


As in the previous micrographs, the dark grey areas are the AlN ceramic grains, about 3-5 microns in size, and the bright areas are the oxide sintering additive phase.


Furnacing considerations


The motivation for developing an AlN formulation that sinters below 1700??C is the new furnace options that this lower temperature opens up. At 1700??C or below, a continuous tunnel kiln can be utilized. This furnace runs in a N2 atmosphere with a small amount of H2 present to protect the heating elements from oxidation. The heat shields are constructed of alumina and the heaters are made from Mo. The substrates are stacked on alumina plates which are continuously pushed through the furnace. The rate of travel depends on the length of the hot zone and, the required time at sintering temperature (about 3-5 hours). Thus the longer the hot zone, the faster the speed through the furnace and the higher the sinter through-put. Since a continuous furnace runs in steady state, there is no time required for the furnace to heat up and cool down. The heat up and cool down cycles are the key rate limitations in a batch furnace.


Conclusion


The five major cost factors for AlN substrates (compared to Al2O3) were discussed: (1) higher cost powder; (2) separate BBO cycle; (3) batch sintering cycle; (4) batch flat fire cycle and (5) non-aqueous processing. By adopting a low temperature sintering configuration, cost factors 4 and 5 are addressed bringing the sintering and flat-firing operations in line with the process for alumina.


Of course, this process will only be appropriate for applications where a thermal conductivity of 130 W/m-K is acceptable. This thermal conductivity should be acceptable for most HBLED, RF and power semiconductor applications. For laser diode telecommunications applications, 130 W/m-K will most likely be too low and conventional higher cost AlN will continue to be utilized.


The availability of a low temperature, continuous sintering process also provides strong motivation for the next phase of cost reduction for AlN, utilization of lower cost, lower performance AlN powder. Again, with a focus on HBLED and power semiconductor applications, sensitivity to impurities such as Fe and Si, which drive up AlN powder costs, may not be anywhere as stringent as applications such as RF and microwave (where dielectric properties at high frequencies are important). The combination of lower cost powder and a continuous sintering process would move AlN substrate pricing much more in line with alumina.


Jonathan Harris, PhD, is president of CMC Laboratories, Inc., Tempe, AZ E mail: jharris@cmclaboratories.com; web: www.cmclaboratories.com


Solid State Technology, Volume 55, Issue 3, April 2012


More Solid State Technology Current Issue Articles

More Solid State Technology Archives Issue Articles