Final Report: Development of Life Cycle Inventory Modules for Semiconductor Processing

EPA Grant Number: R828208
Title: Development of Life Cycle Inventory Modules for Semiconductor Processing
Investigators: Murphy, Cynthia F. , Allen, David T.
Institution: The University of Texas at Austin
EPA Project Officer: Karn, Barbara
Project Period: April 1, 2000 through March 31, 2003 (Extended to June 1, 2004)
Project Amount: $325,000
RFA: Technology for a Sustainable Environment (1999) RFA Text |  Recipients Lists
Research Category: Pollution Prevention/Sustainable Development , Sustainability


Semiconductor devices, also referred to as microchips, have become ubiquitous within the last decade. In addition to obvious applications in computers and other information technology equipment, they commonly are incorporated into phones, toys, household appliances, and automobiles. Therefore, any attempt to perform a life cycle analysis (LCA) on these and many other products requires information regarding the life cycle of microchips. After defining the scope of an LCA there are three sequential steps: (1) inventory development; (2) impact assessment; and (3) recommendations for improvement. The value of the results derived from the impact assessment and subsequent suggested improvements are highly dependent upon the quality of the inventory. Inventory development, in turn, requires detailed knowledge about mass and energy flows; in the case of microchips, these flows are dominated by the manufacturing phase.

Microchips are manufactured on thin, flat discs of silicon (wafers) in a process called wafer fabrication. Wafer fabrication is very resource intensive, using large amounts of energy, water, and chemicals. The life expectancy of an individual product, however, typically is quite short and a common concern is that in the time it takes to develop a life cycle inventory (LCI) for a particular product, the inventory no longer will be valid. A proposed solution to this dilemma is to develop process-based, generic, parametric modules that are derived from fundamental choices in equipment, material selection, and product design, which then can be applied to the process flow of a particular product. The primary objective of this research project was to develop LCI modules for activities performed during the manufacture of integrated circuits.

Although specific product designs and process flows are highly variable and change rapidly in semiconductor manufacturing, there are some aspects that are quite uniform between facilities and over extended periods of time. First, there are very few manufacturers of wafer fabrication process equipment; this severely limits the potential for variability between different fabrication lines (also referred to as “fabs”). Second, the unit operations associated with semiconductor manufacturing have remained relatively constant over the past 30 years. Finally, the largest inventories in all cases are associated with elemental gases, water, and energy. By focusing on these areas, the project goal is to provide a platform for a generic set of modules that can be applied broadly. As such, the research was intended to facilitate the establishment of standards, encourage the development of predictive rather than historical life cycle analyses, and potentially simplify communication along the materials/product supply chain.

Summary/Accomplishments (Outputs/Outcomes):


This research project was conducted by the University of Texas at Austin in cooperation with International Semiconductor Manufacturing Technology (ISMT), a consortium that includes most major manufacturers of semiconductor devices. The majority of the data used in this research was provided by the onsite operations at ISMT in Austin, Texas. In addition, two industry members of the consortium provided site-specific data from a wafer fabrication facility (fab) located in the United States and from a fab in Europe. Most data are gathered from equipment designed to process 200 mm (8 inch) diameter wafers, but 150 mm (6 inch) and 300 mm (12 inch) equipment data were considered as well.

Benchmarking of Life Cycle Inventory Activities

The project was conducted in two stages. First, the status of LCI activities within the industry was examined and benchmarked (Schuppe, 2002). This included the identification and evaluation of commercially available software and associated databases. In addition, the member companies of ISMT were interviewed to determine their needs and concerns in this area. It was found that although there is growing interest within the industry in performing LCA, there also is concern about the availability of adequate data. Progress is hampered by the lack of appropriate database structures, which need to allow for management at a unit operations level, and by lack of standards for establishing system boundaries and data reporting. In addition, many of the chemicals used in semiconductor manufacturing either are not included in existing LCI databases or are represented inadequately. At the material level, most databases report information assuming standard grade materials rather than the ultrapure chemicals that are used almost exclusively in wafer fabrication. Current inventory databases reporting at the product level also do not account for the significant amount of variability in usage rates that exist within the industry as a result of the wide range of product types (designs) and manufacturing practices (process designs).

A ranking of materials based on the need for improved inventory and use data as inputs to life cycle analyses is presented in Table 1. Note that most are specialty chemicals, either because of their composition and/or because of purity requirements. The two primary concerns expressed by ISMT member companies were toxicity and material and energy consumption related to the upstream manufacture of these substances. From the perspective of developing an inventory for wafer fabrication, however, it is less clear that these should be the materials of immediate interest. First, the consumption of these materials relatively is well known because they are expensive (a liter of resist may cost up to $500) and typically tied to specific operations, thus simplifying accounting procedures. Second, if the four broad areas of consumption for wafer fabrication are considered (energy, water, elemental gases, and specialty chemicals), specialty chemicals represent, by far, the smallest fraction of mass utilized. This is not to imply that developing inventories for these materials is not needed, but rather that the other three categories have tended to receive less attention, despite their significant contribution to the environmental load, and therefore represent a larger information gap. In addition, while the per-unit costs are low, the total costs of energy, water, and elemental gases can be significant. For example, on a wafer with 20 mask layers using resist at $125 per liter (more likely than the extreme of $500 per liter cited above) and 4 ml per layer, the resist contributes $10 to the total cost. In contrast, energy at 300 kWh per wafer and $0.06 per kWh contributes $18 to the total cost.

Table 1. Top 25 Materials of Interest for LCI Improvement Per ISMT Member Companies






Sulfuric Acid




Isopropyl Alcohol (IPA)




Phosphoric Acid




Hydrochloric Acid




Tetramethyl ammonium hydroxide




Nitric Acid




Hydrofluoric Acid




Hydrogen Peroxide




N-Methyl Pyrrolidone (NMP)




Nitrogen Trifluoride




Ammonium Fluoride




Ammonium Hydroxide








Ethyl Lactate




Hexamethyl disilazane (HMDS)




Silica, Amorphous




Sodium Hydroxide




Propylene Glycol Monomethyl Ether Acetate (PGMEA)




Acetic Acid








Photo Resists






Ethanol Amine (ACT-690c)











Parametric Module Development

The second stage of the project was the development of parametric modules that could be applied to a variety of product and process configurations using water, energy, and elemental gases. The methodology is demonstrated for consumption of energy and oxygen (O2); however, the approach can be applied to all material categories and emissions. The modules are based on the most commonly defined unit operations used in semiconductor manufacturing. These are furnace processes (including thermal oxidation, diffusion, and anneal), ion implant, chemical vapor deposition (which may overlap with furnace processes), lithography, etch, metallization, chemical mechanical polishing, and wafer clean. Organizing the inventory around unit operations has two advantages. First, this organization can be easily understood by the industry because manufacturing departments are structured along the same lines. Second, unit operations are process-centric, thus facilitating the development of a process-based LCI. Modules are developed for process flows made up of unit operations, for individual unit operations, and for levels forming a hierarchical relationship within the unit operation. These lower levels, illustrated in Figure 1, are equipment type, functions, and recipes.

A Tree Structure Illustrates the Relationship Between the Levels at Which Parametric Modules Can Be Developed. Parametric modules are developed at the function level using recipe parameters and at the unit operation level using either equipment type or function parameters.

Figure 1. A Tree Structure Illustrates the Relationship Between the Levels at Which Parametric Modules Can Be Developed. Parametric modules are developed at the function level using recipe parameters and at the unit operation level using either equipment type or function parameters.

Development of modules at the recipe level is relatively straightforward. A “recipe,” typically unique to a piece of equipment, defines a step-wise process and sets the process parameters (e.g., time, temperature, and material flows) for each step. The consumption of a specific material is first determined at each step by multiplying the flow rate by process time. Summing over all steps and dividing by the number of wafers processed determines the total mass per wafer.

Modules at the function level describe material and energy use for different product specifications, such as film thickness. As an example, consider the energy and oxygen requirements associated with growing an oxide layer in the function “copper insulation” within the “furnace” unit operation. Oxygen consumption and energy use data were collected from different recipes used to form a range of oxide layer thicknesses (Figure 2). The resulting empirically derived relationships are:

kWh/wfr = (6.0x10-7 • d2) – (3.1x10-3 • d) + 44, R2 = 0.998

O2(sL)/wfr = (1.0x10-7 • d2)) – (9.0x10-5 • d) + 2.6, R2 = 1


kWh/wfr = kilowatt-hours per wafer,
O2(sL)/wfr = standard liters of oxygen per wafer,
d = film thickness in angstroms (Å) or 10-10 meters, and
R2 = correlation coefficient.

These empirical parametric relationships can be used to estimate energy and oxygen consumption for a continuum of oxide layer thicknesses, ranging between 2,500 and 6,300 Å. To a limited degree, it also may be possible to extrapolate beyond the defined range.

Figure 2. Energy Consumption (top) and O2 Consumption (bottom) Per 8 Inch Wafer as a Function of Layer Thickness is Shown for Thermal Oxide Grown in a Horizontal Furnace in the Presence of Hydrogen (Wet Oxidation)

Discrete, rather than continuous relationships, often are the basis for module development at the equipment level. These inventory modules can be used to perform trade-off analyses between equipment designs, assuming that the same function is performed on each system. Wafer cleaning occurs before almost every process step and accounts for a significant amount of chemical usage and the majority of the water use in wafer fabrication. This unit operation also exhibits relatively little variation from company to company and is therefore easy to characterize. A comparison of the material consumption for two different wafer clean equipment types using one of the most common cleaning chemistries, referred to as the SPM/HF/RCA (from sulfuric acid/peroxide mixture, hydrofluoric acid, and Radio Corporation of America) clean, is presented in Table 2 (Murphy, Laurent, et al., 2003). This type of module results in discrete parametric relationships, best represented in tabular rather than graphical form. Modules also are developed at the unit operation level and at the process level by grouping unit operations modules together (Murphy, Kenig, et al., 2003).

Table 2. Material Consumption by Technology (Per 200 mm Wafer)


Wet Bench
milliliters per wafer

Spray Cleaning
milliliters per wafer

49% HF (hydrofluoric acid)



H2SO4 (sulfuric acid)



HCl (hydrochloric acid)



H202 (hydrogen peroxide)



NH4OH (ammonium hydroxide)



DI (de-ionized water)



IPA (isopropyl alcohol)



N2 (nitrogen)

No data



The data used to develop the examples described above are included in a database focused on wafer clean and furnace unit operations. These operations are the most uniform across the industry in terms of equipment-type, functions, and recipes. They are equally important for low-end products as well as high-end products, and they tend to have the least amount of proprietary data related to the use of specialty chemicals. For all these reasons they were considered to have the greatest potential for general application. They also are important operations in the overall accounting of energy and material flows for wafer fabrication. Six of the top eight specialty chemicals in Table 1 are used in the wafer clean operation and thus included in the database. Energy measurements were taken directly on equipment at ISMT. Water, elemental gases, and specialty chemical use was determined from analysis of available recipes.


The total amounts of energy, materials, and water consumed in semiconductor manufacturing worldwide are significant, highlighting the need to develop material and energy inventories for semiconductor device manufacture, both for use in improving environmental performance within the wafer fabrication facility and for use in life cycle assessment activities. There have been several barriers to the development of such inventories, including concern over protection of intellectual property and use of proprietary materials. The real challenge, however, is the tremendous variability between products and facilities as well as very short product lifetimes (2 years or less). The proposed solution to facilitating development of inventories for wafer fabrication is to develop parametric modules based on unit operations, which change at a much slower rate than the product or process designs. Many of the data are already available and their use for development of inventory modules is dependent primarily on organization and analysis. In many cases, the relationships are based on fundamental physical and chemical properties and processes, and data can be collected that relatively are independent of the specific product being manufactured.

The parametric modules developed within the context of this project are intended to support environmental analysis for wafer fabrication in the semiconductor industry. As the volume of micro- and nanotechnology-based manufacturing increases on a global scale, however, it is important that there are methods available for the environmental analysis of complex and variable systems, particularly when they involve the production of emerging, and therefore frequently changing, technologies. The methodology proposed herein could facilitate effective application within many economic sectors.

Journal Articles on this Report : 1 Displayed | Download in RIS Format

Other project views: All 10 publications 1 publications in selected types All 1 journal articles
Type Citation Project Document Sources
Journal Article Murphy CF, Kenig GA, Allen DT, Laurent J-P, Dyer DE. Development of parametric material, energy, and emission inventories for wafer fabrication in the semiconductor industry. Environmental Science & Technology 2003;37(23):5373-5382. R828208 (Final)
  • Full-text: ACS Full Text
  • Other: ACS PDF
  • Supplemental Keywords:

    wafer fabrication, parametric inventories, unit operations, material and energy flows, emerging technologies, life cycle analysis,, RFA, Scientific Discipline, Sustainable Industry/Business, Sustainable Environment, Environmental Chemistry, cleaner production/pollution prevention, Technology for Sustainable Environment, Economics and Business, Environmental Engineering, life cycle analysis, integrated circuit manufacturing, SIC 3674, sustainable development, cleaner production, environmentally conscious manufacturing, waste minimization, waste reduction, life cycle inventory, green process systems, electronics, life cycle assessment, semiconductor manufacturing, pollution prevention

    Progress and Final Reports:

    Original Abstract