Subscriber access provided by MITEnvironmental Science & Technology is published by the American Chemical Society.1155 Sixteenth Street N.W., Washington, DC 20036Correspondence/RebuttalResponse to Comment on “The 1.7 Kilogram Microchip: Energyand Material Use in the Production of Semiconductor Devices”Eric D. Williams, Robert U. Ayres, and Miriam HellerEnviron. Sci. Technol., 2004, 38 (6), 1916-1917• DOI: 10.1021/es049890z • Publication Date (Web): 14 February 2004Downloaded from http://pubs.acs.org on April 1, 2009More About This ArticleAdditional resources and features associated with this article are available within the HTML version:• Supporting Information• Access to high resolution figures• Links to articles and content related to this article• Copyright permission to reproduce figures and/or text from this articleResponse to Comment on “The 1.7 KilogramMicrochip: Energy and Material Use in theProduction of Semiconductor Devices”Shadman and McManus (1) reject our conclusion that “thewidespread assumption that the microchip represents aprime example of dematerialization is misleading at bestand probably false” (2). The reason put forward is that “thestudy’s conclusion is itself misleadingbecause using weightas the basis for comparison is arbitrary, nonscientific, andinaccurate”.Ifwehadbeencomparingmicrochipswithotherproductsonthebasisofweight,thisobjectionmighthavesomevalidity.Our point was, and is, simply that the amount of materialsusedtomanufactureacomputerchipthesedaysishundreds,if not thousands, of times greater than the quantity actuallyembodied in the chip. This makes the weight of the chip amisleading indicator of the amount of materials used, andit means that people like Alan Greenspan and FrancesCairncross, who have cited microelectronics as an exampleof radical “dematerialization”, have misunderstood thesituation.Assessing the microchip industry vis-a`-vis other indus-tries or technology alternatives in terms of social/environ-mental benefits and costs was not the purpose of ourpaper. Such a synthetic analysis would be a completelydifferent paper, and we believe one not yet possible towrite given that understanding of the environmental im-plications of information technology is still in its infancy.We do believe there are many applications of informa-tion technology that are actually or potentially beneficial tothe environment (such as telecommuting). Indeed, someof us are engaging in research to understand and maxi-mize these benefits. However, the environmental manage-ment of production, use, and disposal of hardwaresandassociated production wastessis research that can andshould be addressed on its own, as distinct from itsapplications.Given our intention to characterize the resource in-tensity of semiconductor manufacturing, we believe ouranalytic method and choice of indicators are appropri-ate. Widely accepted definitions of “dematerialization”refer to weight of the materials or energy used to manufac-ture industrial end products. We have expressed this re-source (materials/energy) intensity of the network of manu-facturing processes in an easily understood quantitativeform. That this intensity is high is surely no surprise to theindustry, as environmental reports of all major firms in-clude data and substantial discussion on their efforts toreduce energy and materials use and emissions. If theenvironmental intensity per unit mass of chips producedwas comparable to conventional goods, energy and ma-terials use would be a nonissue, which clearly is not thecase.While notthe topicof ourpaper, we agreewith Shadmanand McManus (1) that the semiconductor industry hasbeen improving its environmental performance. We haveyet to see, however, that this progress has been sufficientto counteract growth in demand so as to realize net de-creases in material and energy used by the industry. Ac-cordingtotheU.S.Censusstatisticsandconsultingfirmdata,wafer, chemical, and energy use of the semiconductorindustry increased 6-10% annually in the late 1990s. Whilethis rate of increase is less than the economic growth of theindustry (about 15% on average), it is still substantial. Theevolution of materials intensity at the product level is aninteresting question as well. While the smaller feature sizeof newer generation chips could imply less materials useper transistor, the increased complexity of processes, therequirement for ever-declining defect densities, and theneed for purer starting materials have the opposite effect:the ratio of indirect to direct materials consumption mayactually be increasing. This would be an interesting study,which we think should be done. In fact, we invite Intel, theworld’slargestmanufacturerofchips,toprovidetherelevantdata.Regarding the Shadman and McManus (1) critique ofthe microchip-automobile comparison, we suggest thatthey check our paper and its main references more care-fully. Our figure for energy intensity of automobile manu-facturing comes from a life cycle study that includesproductionofrawmaterials,parts,andassembly,notsimplythe final assembly stage as they suggest. If anything, thedifference in intensity is larger than we reported, as ouranalysis gives a lower bound on chip manufacturing thatdoes not include the energy and materials associated withtheproductionof high-gradechemical inputs,the lifecyclesof manufacturing tools and other equipment, and otherancillary processes such as transport of materials throughthe supply chain. Regarding data quality, one reason thestudy generated such widespread interest was preciselybecause we went far beyond most prior studies in the field,collectingseveral datasources formostprocesses, includingone modern chip fabrication facility (that shall remainanonymous). We must admit that we did not make directmeasurements ourselves.Given thatwe reliedon secondarysources, we cannot rule out the possibility that there wascollusionbetweenreporterstotheU.S.BureauoftheCensus,theindustrycommitteethatcontributedtotheMCClifecyclestudy (which included Motorola and SEMATECH), and theSemiconductorIndustryAssociation(SIA)tooverstateenergyuse.Finally, Shadman and McManus (1) misquote a sen-tence from our paper, which they then claim proves thatwe have admitted a lack of pertinent input/output data.The correct statement in the paper was “Quantitativeinformation on input/outputs to the assembly processis scarce” (italics added). Assembly is one process stepamong many