Uniroyal combines the latest quaternary and ternary compounds, with multiple quantum well (MQW) semiconductor technology, to provide the brightest multicandela class LEDs available. For the best in Visible and UV LED results, choose Uniroyal for “The Light Inside”™ your LED-based product. Uniroyal is your one-stop shop for Epitaxial Wafers, Package Ready Die (PRD)™, and custom LED semiconductors.
White Papers

Cost of Ownership Issues for Hydrogen Gas Purification

Eric Bretschneider
UNIROYAL Optoelectronics


In compound semiconductor growth, the choice of gas purification methods has traditionally been based solely on technology. Little attention has been paid to the operating cost associated with different gas purifiers. In order to compete in high volume markets, the operating cost and specifically the cost of ownership should be a crucial factor in selecting a gas purification method. Accordingly, a detailed study has been completed to compare the cost of ownership for a variety of gas purification methods at varying gas flow rates. Depending on the volume of gas being purified, the cost of ownership for different purification methods can vary widely. The impact of bulk compressed or liquid hydrogen and point of use versus bulk purification on the cost of ownership is shown to be particularly significant.


From its very beginning, decisions relating to the compound semiconductor industry (CSI) have been made based on technology. The early support of the defense industry that was vital to the development of today's compound semiconductor industry placed a premium on performance. There was little reward or incentive for manufacturers to reduce costs. As the industry has matured, remnants of this attitude remain. It is still not uncommon to hear sales pitches that focus on technology and performance with little attention paid to the reality of economics.

Continued incremental advances in silicon based technologies have pushed the performance of silicon based devices into realms that as recently as five years ago were thought to be the exclusive domain of compound semiconductors. Although compound semiconductors have managed to flourish in many niche markets such as optoelectronics where the indirect band gap of silicon places it at a significant disadvantage, manufacturers need to pay close attention to economic considerations in order to successfully compete against existing and emerging elemental semiconductor technologies.

Additionally, the technologies used for MOCVD growth of compound semiconductors have advanced to the stage where most of the commercial equipment is capable of growing materials and structures of comparable quality and performance. Under these circumstances, it behooves the manufacturer to pay close attention to the relative economic capabilities of different types of hardware. Indeed, it is not uncommon to find that many hardware manufacturers who supply the compound semiconductor industry to quote Cost of Ownership (CoO) metrics for their equipment. Most of these metrics are based on the SEMI E-35 standard for CoO. While the growing prevalence of CoO metrics helps to illustrate the relative maturity of the CSI, it is important to realize that these metrics are typically based on a generic version of a process and therefore may not correlate well to the actual economics of operating hardware in a manufacturing environment.

Cost of Ownership

One of the first steps in the design and specification of a new manufacturing facility or the expansion of an existing facility, should be an economic analysis to determine the magnitude of the potential revenue stream. At the concept stage, the uncertainty in this estimate may exceed 50%! Despite this large potential error, it is still critical to begin a cost model at the earliest stages of design. Careful attention to economics at all stages of design will help to maximize the revenue stream from a facility and may also help to avoid unnecessary design efforts and expenses. As the details of the design are refined, the estimated costs and revenues from the model will begin to converge. It is not unreasonable to achieve economic estimates within 5% of the actual costs and revenue streams before construction on a facility has even begun.

As mentioned above, many of the details required to calculate an accurate CoO metric are not available to hardware manufacturers. These include, but are not limited to: facilities design (including support equipment), layout, installation costs and specific details of the manufacturing process. The details of a manufacturing process potentially have the greatest impact on the economics of a facility. Key factors that may be affected include: total cycle time (wafer throughput), labor requirements, the maintenance schedule (uptime, spare parts inventory) and also impact the choice of raw materials (process yield, cost, safety stock, maintenance schedule).

In light of uncertainty of the CoO metrics obtained from equipment manufacturers, it would be useful to establish a new metric that encompasses all of the economic factors in a process. Such a Global Economic Metric (GEM) that incorporates all the factors that impact the economic viability of a process would provide a single figure of merit that would be perhaps the most useful and easily understood metric for comparing hardware, raw materials, facility and process configurations.

By necessity, any GEM model will require a tremendous amount of information that may or may not be available to equipment suppliers. Uncertainties in the input data are best handled by performing sensitivity analyses for input parameters and using statistical methods to generate an uncertainty factor. A full discussion of the preferred methods is beyond the scope of this article, but can be found in standard references on derivatives and statistics.

Hydrogen Delivery Systems

Before discussing the technologies available for hydrogen purification, it is important to discuss the choices for the storage/delivery of bulk hydrogen: compressed (BHY) and liquid hydrogen (LHY). As will be seen later, this decision may have a significant impact on the overall economics of hydrogen purification.

For small facilities, BHY is often the preferred form of delivery. Typical storage and delivery pressures are 3000 psig and 200 psig respectively. Depending on the supplier, delivery specifications may range from 99.95% to 99.995%. Facility configurations may require transfilling high volume cylinders that are permanently installed on site, or banks of high volume cylinders on flat bed trailers may be exchanged. Both of these options impose different requirements on the facilities design team. While permanently mounted cylinders minimize the total square footage required they result in the highest unit cost of material due to the inefficient delivery mode.

With a minimum purity specification of 99.9995% and a unit cost that is 30% to 50% that of BHY, LHY would seem to be the obvious choice for a new or existing facility. Unfortunately, the higher purity of a LHY system comes at a significant price with respect to the design of facility. It is vitally important that strict attention is paid to all OSHA , NFPA and local codes related to LHY systems at every stage of design. Adherence to these codes may require explosion proof electronics and no permanent openings or flammable materials within a significant radius of any component of the system. LHY systems also have a low delivery pressure of ~100 psig. Although this pressure may be increased by use of a hydrogen compressor, this option may negatively impact the delivered purity and may significantly increase maintenance requirements for the system.

One additional issue with operation of a LHY system is indirectly related to its extremely low temperature. Unless there is a constant draw of hydrogen, the system will vent an appreciable amount of hydrogen. This typically limits the use of LHY systems to facilities that operate on a 24/7 schedule. It should be noted that although LHY has a significantly lower unit cost than BHY, at low consumption rates, BHY provides the lowest operating cost. This is due to a combination of several factors including simpler facilities design, loss of hydrogen due to venting during low draw periods of operation on a LHY system, and a lower monthly service charge for hardware leased from a gas supplier.

Figure 1. Comparison of relative unit costs (arbitrary units) for compressed (+) and liquid (x) hydrogen.

As will be seen later, the lower delivery pressure of LHY systems may have a significant impact on the choice of technology for hydrogen purification. It is important to note that although the effective unit cost for hydrogen is to a first approximation a linear function of the monthly volume, the total cost for hydrogen is proportional to the square of the monthly volume. Thus at high monthly volumes, the lower effective unit cost of LHY will almost always result in the lowest operating cost for a facility provided that it operates on a 24/7 schedule.

Hydrogen Purification

In both the elemental and compound semiconductor industries, control of contaminants is a critical factor that may contribute significantly to both device performance and process yield. Although there have been recent reports on the use of nitrogen as a carrier gas , hydrogen is the dominant carrier gas used in the both the compound and elemental semiconductor industry. In the CSI, hydrogen purifiers are typically installed at the point of use, inside the tool cabinet. Due to this fact, the cost of the hardware, which is perhaps the simplest, most obvious economic metric, is generally not even visible, as it is incorporated into the cost of capital equipment. CoO associated with different methods/technologies has received almost no attention.

Pd cell

Of all the technologies available for hydrogen purification, Palladium (Pd) cell technology has the longest history and is the current method of choice in the CSI. Pd cells work by passing an impure hydrogen stream over a heated Pd alloy that acts as a semi-permeable membrane. At high temperatures, hydrogen will dissociatively dissolve into Pd and its alloys. If a pressure gradient is present across the membrane, atomic hydrogen will be transported through the membrane by bulk diffusion through the membrane where it will recombine to form molecular hydrogen on the other side. Throughput is dependent on the inlet pressure, membrane area, membrane thickness and to a slight extent, the temperature of the membrane.

Pd cell based purifiers have the benefit of a theoretically unlimited lifetime. In practice, of course, the lifetimes are much shorter. The typical failure mode is formation of pinholes or micro cracks in the Pd alloy membrane. These failures may be precipitated by a number of events including thermal cycling, exposure to high levels of impurities (>50 ppm oxygen) and cool down in the presence of hydrogen. Due to the catastrophic nature these failure modes, they are typically not detected until after at least one run of non-conforming wafers has been produced. As the revenue potential from some processes is measured in tens of thousands of dollars, this can add significantly impact the overall economics of a plant.

Heated Getter Another issue with PD cell failure is the possible contamination of expensive metal-organic sources. For large, production systems, the total value of metal-organics installed on the system may exceed $100,000. In order to address this concern, most of the system suppliers have made heated getter purifiers a standard component for their MOCVD systems. These units are typically installed directly upstream of the metal-organic gas distribution manifold and are designed to protect all of the metal-organic bubblers on a system from contamination in the event the hydrogen purifier fails.

Theses systems rely on an irreversible reaction to remove impurities. The getter material is typically maintained at high temperature to increase the rate of reaction and to allow impurities to diffuse into the bulk of the getter media. The end result is a highly effective purification system with a high capacity for impurity material. Other advantages of this technology include a low-pressure drop across the purifier and improved resistance to thermal cycling. The lifetime of these units is dependent on the impurity load and flow rate of hydrogen. This allows simple determination of maintenance intervals to change the getter cartridge. The low-pressure drop allows use of a LHY delivery system, which with its much higher purity can increase the lifetime of a cartridge by more than an order of magnitude.

Unfortunately, these systems still require some type of UPS feed for maximum reliability and the purification media is not regenerable. The maximum flow rate for these units is on the order of 50 slm, which may limit their use on some of the larger, commercial MOCVD systems.

Ambient Temperature Absorption A third option for hydrogen purification is use of an ambient temperature purifying media. These materials will reversibly react with impurities at low temperature. At elevated temperatures, the equilibrium point of the reaction shifts and the media may be regenerated. This minimizes maintenance requirements and allows for a theoretical uptime of 100%. An informal survey of multiple users of this technology in the silicon industry completed by Uniroyal Optoelectronics found that uptimes in excess of 99.9% are readily achievable. It is important to note this performance level is in fact extremely conservative, as several companies reported no downtime for units in operation for over eight years.

As the purification reactions take place at ambient temperature, loss of power will have no impact on the purification of process gases. The only power is for the electronics that control regeneration operation. The regeneration interval is based on typical flow rates and is inversely proportional to the impurity challenge.

These purifiers also have a low-pressure drop and are thus readily adaptable to use with a LHY system. This fact coupled with real world uptimes approaching 100% makes ambient temperature purification the obvious choice for manufacturing facilities requiring high volumes of ultra pure hydrogen.

Comparative Cost of Ownership In order to compare the relative cost of ownership for each of the technologies detailed above a theoretical epi process was created. Key factors in determining the CoO for each technology included: batch size, total cycle time, raw materials cost, revenue potential, average hydrogen flow rate, MTBF estimates for each technology and maintenance requirements. When factoring in maintenance requirements, it is important to factor in all of the economic parameters such as the accounting cost of safety stock items, parts, labor and lost revenue. As can be seen in figure 3 below, the choice of technology based on relative capital cost may change significantly with the total number of installed systems. This serves to highlight the importance of a detailed cost model for minimizing the total capital outlay over the life of a facility.

Figure 2. Comparison of relative monthly operating costs for compressed (+) and liquid (x) hydrogen.

Although ambient temperature purifiers that are not on-line regenerable are readily available, the capital costs and CoO of this type of purifier were found to essentially identical to that of heated getter systems. For this reason, results on this type of purifier are not presented. It should be noted that even the smallest regenerable, ambient temperature purifiers are capable of flow rates of hundreds of slm; equivalent to the flow rate of multiple MOCVD systems. For this reason, it was assumed that this type of equipment was installed as a facility purifier. This is an essential distinction as the capital and installation costs of these purifiers is very high compared to point of use (POU) purifiers. All other technologies were assumed to be installed at POU.

With respect to operational costs for the different purification technologies, the high inlet pressure required for Pd cells places them at a distinct disadvantage. While it is possible to operate Pd cells at inlet pressures of ~ 100 psig, this will require use of a larger (and more expensive) Pd cell or the use of two Pd cells in parallel. Both of these options would contribute significantly to installation costs. The data shown in figure 3 assumes a single Pd cell on each system. Use of a hydrogen compressor in conjunction with a LHY system was not considered due to the issues with potential contamination. The ability to readily use the relatively high purity and inexpensive feed from a LHY system result in a significant economic benefit for both heated getter and ambient temperature purifiers. As can be seen in Figure 4, the operating costs for both of these technologies is significantly lower than that of Pd cell technology.

Figure 3. Relative Capital Costs including installation and safety stock requirements for Pd cell (+), Heated Getter (x), and Ambient Temperature (•) purifiers.

Perhaps the greatest disadvantage to Pd cell technology is the catastrophic nature of its failure modes. There is as of yet no method to predict, in advance, when a cell will fail. This inevitably results in a loss of raw material, revenue and a significant amount of processing time to verify a failure and install a replacement unit.

Device Results

While the above discussions present a strong argument for use of non-Pd cell technology for hydrogen purification, there remains a strong bias against these alternative technologies. It is essentially impossible to argue against the theoretically perfect performance of a Pd cell. Only by looking at the economics associated with the use of this method of hydrogen purification can a reasonable case be made to use heated getter or ambient temperature technology. Today's high brightness (HB) LEDs provide an excellent test case for proving the efficacy of alternative technologies.

In the case of long wavelength (570 - 650 nm) LEDs, AlInGaP is the material system of choice for producing HB LEDs. In particular, the performance of devices that use a distributed bragg reflector (DBR) to improve light extraction provides an excellent benchmark. The DBRs in these structures typically consist of alternating layers of AlAs and AlGaAs. High aluminum containing alloys are known to be extremely sensitive to residual oxygen contamination. At growth temperature, both gallium and indium oxides will readily form, but due to their relatively high vapor pressures, they will not tend to incorporate into a growing epi layer. Aluminum, on the other hand, has both a stronger affinity for oxygen and a marked lower vapor pressure.

In the spring of 2001, Uniroyal Technology Corporation issued two press releases regarding product performance. The first was related to HB InGaN LEDs that exhibited brightness levels of 3.5 to 4.0 mW in the die form at dominant wavelengths between 450 and 470 nm. Due to the high refractive index of the epoxy used to encapsulate LEDs, packaging will typically result in power level between 2.2 - 3.0 times that of a bare die. This means that conventionally packaged, these device could be reasonably expected to produce lamps with power outputs between 8 and 10 mW when operated at 20 mA forward current. This is among the highest performance ever reported for InGaN materials.

The second press release was related to Uniroyal Optoelectronics' line of AlInGaP HB LEDs at dominant wavelengths between 618 nm and 626 nm. When operated at a forward current of 20 mA, these devices exhibit typical brightness levels of 140 to 200 mcd in die form. This performance level is among the highest ever reported for devices grown on absorbing substrates. It should be noted that the structure includes a DBR with layers containing >70% aluminum.

Both the InGaN and AlInGaP devices have been successfully grown using ambient temperature purification technology at these performance levels and with excellent wavelength uniformity. For InGaN devices grown on 2" sapphire substrates wavelength non-uniformities of less than 0.25% and a total wavelength variation of less than 4 nm have been achieved. Similarly, for AlInGaP devices on 3" GaAs substrates wavelength non-uniformities of less than 0.5% and total wavelength variations for an entire run of wafers of less than 4 nm are routinely achieved.


In summary, it has been shown that at least two economically viable alternatives to Pd cell technology for hydrogen purification exist. With a range of parameter space that included operation of a single, small capacity system operating 40 hours/week to multiple high volume systems operating 24 hours/day, no conditions were found where Pd cells provided the lowest GEM. The economic advantages of these alternative technologies are primarily a result of their compatibility with LHY delivery systems and the catastrophic nature of Pd cell failures.


The author would like to thank the following companies for providing information on hydrogen system and unit costs: Air Products, BOC - Edwards, Matheson Tri-Gas, and Praxair.


  1. M. S. Peters and K. D. Timmerhaus, Plant Design and Economics for Chemical Engineers. McGraw-Hill Book Company New York (1980) p157.
  2. G. E. P. Box, W. G. Hunter and J. S. Hunter, Statistics for Experimenters. John Wiley & Sons, New York (1978) pp107-150, 556-582.
  3. OSHA CFR 1910.103
  4. S. Guadagnuolo and G. Vergani "On the Suitability of Getter Purified Hydrogen for the LP-MOVPE of AlGaAs: A Comparison to Pd-Diffused Hydrogen", CS-MAX 2001.
  5. Uniroyal Technology Corporation press release dated April 10, 2001.
  6. Uniroyal Technology Corporation press release dated May 9, 2001.
  7. Non-uniformity defined as (standard deviation/mean) x 100.
Figure 4. Relative annual operating costs for Pd cell (+), Heated Getter (x), and Ambient Temperature (•) purifiers.

Contact Uniroyal Optoelectronics at (800) 634-8491, or customer.service@uniroyalopto.com
Copyright © 2001 Uniroyal Optoelectronics
This site is optimized for current versions of Internet Explorer or NetScape.