The 1.7 kg Microchip Eric Williams, United Nations University Robert U. Ayres, INSEAD Miriam Heller, NSF
Motivations Growth of IT industry: macro-economic scale and continued high growth (average annual growth of global semiconductor industry is 16% per year in recent decades) . What are the environmental implications of this new industry? Are there general trends in relationship between high-tech economy and materials use/environment?
Life cycle inventory of microchip • Estimate life cycle inventory of energy and aggregate chemical use for production of common microchip. • Energy use is good indicator of impacts on climate change and fossil fuel use. Aggregate chemical use is poor indicator of impacts on local soil, air, water systems.
Guiding principles Only use publicly available sources, fully report all data and assumptions used. • Critically compare different data sources for different processes. • Compare final results with those from other groups and deconstruct differences.
Key Processes １．Wafer Fabrication 2. Quartz to Silicon wafers 3. Semiconductor-grade chemicals 4. Assembly
Chemical input :Compare data sources 1 Aggregate chemical input/emission
Energy use in fabrication 1 Various sources suggest 1.4-1.6 kWh of electricity consumed per cm2 of wafer processed, 80-90% of total energy use is electricity. Data reflects aggregate of national industries. Data sources: Census, JEIDA, Semiconductor Industry Association, Microelectronics and Computer Technology Corporation (MCC)
Water use in fabrication 1 Take “typical” figure as 20 liters/cm2
Stage Electenergy input/kg silicon Silicon Yield Data sources Quartz + carbon → silicon 13 kWh 90% Harben, 99; Dosaj, 97 Jackson, 96 Silicon → trichlorosilane 50 kWh 90% Takegoshi, 94; O’Mara et al, 90 Trichlorosilane → polysilicon 250 kWh 42% Tsuo et.al, 98; O’Mara, 90;Takegoshi, 94 Polysilicon → single crystal ingot 250 kWh 50% Takegoshi, 94 Single crystal ingot → silicon wafer 240 kWh 56% Takegoshi, 94; Lammers and Hara, 96 Process chain to produce wafer 2,130 kWh 9.5% From quartz to wafers 2 Production of silicon wafers requires around 160 times The energy required for “industrial” grade silicon
Chemical inputs to fabrication 3 • Semiconductor grade chemicals/gases typically 99.999-99.9999% purity, requires substantial purification, for which no data was available. • Data used reflects production of industrial grade chemicals (used Boustead database, other LCA databases same). • Distillation processes are, in general, energy intensive.
Assembly 4 Energy use: .34 kWh per cm2 of input silicon. Material inputs: packaging material (epoxy, ceramic), lead frame (copper, aluminum), processing chemicals. Data sources: MCC, JEIDA
LCA of 32MB DRAM chip Combine previous process data with information on wafer yields for 32MB DRAM chips (Semiconductor International, 1998): 1.6 cm2 of input wafer per chip.
Energy use for different stages in life cycle of 32 MB DRAM chip Breakdown of life cycle energy use in production and use of 1 2-gram memory chip
Fossil fuel, chemical, and water use For 1 memory chip, lower bounds are: • Fossil fuels consumed in production = 1,200 grams • Fossil fuels consumed in use = 440 grams • Chemicals “destructively” consumed = 72 grams • Water use is 36,000 grams per chip. Total fossil fuel and chemical use to produce 2 gram memory chip 1.7 kg
Secondary materialization Measure material and energy intensity: secondary materialization index (SMI): Secondary materials counted are only those obviously “destructively consumed”: fossil fuels and chemicals (water and elemental gases not included). SMI index for various products: Microchip: 640 Automobile： 1-2 Refrigerator: 2
Why so different? Despite trivial physical weight, “secondary” weight of chips is substantial. Why such a dramatic figure? Postulate: Because chips are exceedingly highly organized (low entropy) objects, the materials and energy required for processing is especially high.
Entropy analysis Estimate order of magnitude of entropy changes associated with final product and producing high-grade inputs: • Mesoscopic order of microchip: use S=-k ln W and checkerboard model. Cell length = 1 μm Board length = die size = 1 cm. result: Entropy (at room temp) = 9.5x10-20 J per memory chip • Ultra-high purity water (tap water – 100 ppm impurities, fab water - 1 ppb). Use entropy of mixing: ΔS= -R [(1-x) ln (1-x) + x ln x ] (x = impurity concentration) result Entropy change (at room temp) = 17 J/kg of pure water Magnitudes of entropy change much lower than energy use - does not explain practical experience of high energy needed for pure materials.
Third law of thermodynamics for purity as well as temperature? Third law of thermodynamics (Nernst, 1906): it is impossible to reach absolute zero in a finite number of reversible steps Analogous phenomenon for purity? Conjecture: energy efficiency of purification decreases as one approaches perfect purity. Conjecture: • 100% purity is impossible (no perfect vacuum) It follows that all purification processes have efficiency <1 and achieving higher purity with given process requires increasing # of steps (e.g. .9 x .9 x .9 ….)
Secondary materialization Many advanced materials/products are also low entropy. Does their proliferation imply increase in SMI of overall economy? The possibility of this is called secondary materialization Not known if significant, but suggests importance of life cycle materials studies to clarify. Need to carefully treat chemicals industry and purification/materials processing.
Hybrid LCI for desktop computer Analysis of energy use in production of desktop computer with 17-inch CRT monitor Hybrid method that splits estimation into process and economic IO pieces.
Commentary For desktop, production phase is 83% of total life cycle energy, very high share compared to other appliances such as refrigerator, which has 12% in production phase. Combination of high energy intensity in production and short lifespan imply that lifespan extension is key approach that should be pursued in policy for managing impacts of IT equipment.
Thank You Eric Williams Williams@hq.unu.edu