This article appeared in a journal published by Elsevier. The attachedcopy is furnished to the author for internal non-commercial researchand education use, including for instruction at the authors institution

and sharing with colleagues.

Other uses, including reproduction and distribution, or selling orlicensing copies, or posting to personal, institutional or third party

websites are prohibited.

In most cases authors are permitted to post their version of thearticle (e.g. in Word or Tex form) to their personal website orinstitutional repository. Authors requiring further information

regarding Elsevier’s archiving and manuscript policies areencouraged to visit:

http://www.elsevier.com/authorsrights

Author's personal copy

Identifying critical materials for photovoltaics in the US: A multi-metricapproach

Michele Goe, Gabrielle Gaustad ⇑Golisano Institute for Sustainability, Rochester Institute of Technology, 111 Lomb Memorial Drive, Rochester, NY 14623, United States

h i g h l i g h t s

� Ever increasing non-fuel material consumption has heightened energy security concerns.� Sustainability related metrics enable policymakers a more comprehensive approach.� Single score and metric aggregation oversimplify or confound indicator trends.

a r t i c l e i n f o

Article history:Received 28 June 2013Received in revised form 28 November 2013Accepted 8 January 2014Available online 11 February 2014

Keywords:Energy securitySupply riskThin-film photovoltaicsIndiumGalliumTellurium

a b s t r a c t

There are increasing concerns that physical material constraints threaten energy security and the growthof emerging technologies. Traditional approaches to quantify material criticality utilize single-score met-rics which are narrowly focused on physical scarcity and often lead to command-and-control policies.However, a broader definition of criticality that goes beyond physical scarcity to include sustainabilitymetrics e.g. embodied energy, political instability, economic value can provide policymakers with a morecomprehensive perspective of the complex and highly interconnected relationships between indicators.We use the case of solar photovoltaic materials to demonstrate the challenges and opportunities in crit-ical materials policy and indicator choices. For silicon-based and thin-film photovoltaics in particular, Ge,Pt, As, In, Sn and Ag were found to be the most critical relative to the 17 materials studied. Multi-metricanalysis for these materials reveals tradeoffs that suggest friction between sustainable economics, polit-ical stability of supply, and environmental quality objectives.

� 2014 Elsevier Ltd. All rights reserved.

1. Introduction

The United States is a dominant consumer of primary energyand materials in the world. However, the growth of emergingeconomies such as China and India and their increasing consump-tion of energy and materials have begun to draw attention towardsmaterials availability and criticality concerns. Further deepeningthese concerns is the recognition of the United States’ import reli-ance on primary energy fuels and some primary materials; of par-ticular relevance are rare earth metals with applications inemerging electrical and energy technologies that are mined inadversarial or socio-politically unstable nations [1]. One emergingtechnology that may be essential to US energy security and climatechange mitigation is solar photovoltaics (PV). With respect to lifecycle carbon emissions and land use, PV technologies have lessenvironmental impact than traditional energy technologies i.e. coalpower plants [2,3]. This implies that broad PV deployment would

significantly reduce global greenhouse gas emissions and its asso-ciated climate impacts. However, it has been suggested that mate-rial availability is a potential constraint for broad deployment of PV[4–7]. For example, current silicon-based and thin-film solar PV’score technology depends on several primary materials i.e. In andTe which were recently determined to be of high importance forthe development of a clean energy economy and at near-criticalor critical supply risk by the U.S. Department of Energy (DOE)[8]. Recent PV research also assesses the broader impacts of criticaland non-critical material choice [9–14].

Concerns over material availability, especially for emergingtechnologies, are not new and over the last 70 years have sparkeddebates as well as national policies aimed at securing critical mate-rials [15]. These policies continue to be implemented despite thelack of a broader definition of criticality. For example, the most re-cent Department of Defense (DoD) Strategic and Critical Materialsreport per the Strategic and Critical Materials Stockpiling Act [16]uses material consumption, production, and projected future de-mand to determine the severity of material criticality. Similarly,in previous literature [4,17–20], the material availability is

http://dx.doi.org/10.1016/j.apenergy.2014.01.0250306-2619/� 2014 Elsevier Ltd. All rights reserved.

⇑ Corresponding author. Tel.: +1 585 475 6089.E-mail address: gabrielle.gaustad@rit.edu (G. Gaustad).

Applied Energy 123 (2014) 387–396

Contents lists available at ScienceDirect

Applied Energy

journal homepage: www.elsevier .com/locate /apenergy

Author's personal copy

determined primarily by physical scarcity, however, systems levelconsiderations such as the production share of politically instablenations, toxicity, embodied energy, or the value to the economyare not considered. The use of a broader definition of criticalitywould likely increase the scope to include energy intensive mate-rials such as aluminum and silicon that are not physically scarcebut have broad economic and environmental implications.

1.1. Aims of this study

Earlier literature claimed material criticality concerns at thepolicy level were waning by pointing to increased foreign mineralreliance and decreased domestic mining [21–23]. Similar circum-stances have motivated recent interest in identifying critical mate-rials. Several nations including those in the European Union (EU)have recently identified materials that are common to photovolta-ics (e.g. In, Ga, and Ge) as critical in terms of supply risk and eco-nomic importance [24–28]. However these studies lacksensitivity of results to data uncertainty and organization; theyalso rely on relative rather than normative determinations of crit-icality which lack context for (future) supply risks. For example,the Centre for Policy Related Statistics’ aggregation of productgroups masks supply chain dependencies. The Morley et al. studycontains no clear environmental metric and aggregates similarmetrics (e.g. depletion time, reserve base) to determine a singlecriticality ‘‘score’’ which ignores the interdependence of data.Other criticality studies have proposed methodology to ascertainthe supply risk from a corporate, national, and/or global perspec-tive [29,30]. Furthermore, none of the studies mentioned above ad-dress uncertainty as to the impact of a limited supply of basemetals such as Cu, Al, or Zn on the criticality of their by-productmetals (e.g. Te for Cu, Ga for Al, and In for Zn). Lastly, these studiesare limited in the breath of criticality metrics especially with re-gards to economic and environmental risks which would providepolicymakers with a more systemic perspective.

Several questions arise from the afore mentioned literaturegaps: What metrics are useful for policy-makers in assessing andregulating criticality issues? What policies would address metalcriticality while at the same time continue to encourage solar PVadoption?

Addressing criticality in policy is challenging due to the com-plex, highly interconnected geopolitical relationships of supplychains, infrastructure lock-in, and increasing material demand thatmust be balanced with low carbon supply. This work aims to quan-tify and compare a uniquely broad set of criticality metrics for sil-icon-based and thin-film i.e. cadmium telluride (CdTe), copperindium diselenide (CIGS), amorphous silicon (a-Si) PV technologiesthat focus on a more comprehensive or life cycle systems approachwhich is unique in its inclusion of environmental metrics. Thisanalysis highlights comparisons between metrics and combina-tions of metrics. In addition, we suggest how to depart from tradi-tional command-and-control policies utilizing the aforementionedmetrics to mitigate criticality in the short and long term.

1.2. Criticality definition and materials considered

Material criticality, as defined here, is a relative concept in thatit compares materials against each other to determine which mate-rials have the greatest risks of disruption to supply. In this analysis,the focus is on PV materials and also includes impacts on the econ-omy and the environment. In order to evaluate the criticality of so-lar PV materials from the perspective of the US. we characterizethree areas of criticality: supply risk (Section 3.1), economic risk,(Section 3.2) and environmental risk (Section 3.3). This is a semi-dynamic study in that we include select data for materials over a20-year period (1992–2012) commenting on their trends in the

context of the decision making for policy. The solar PV materialsconsidered in this study and their previously identified criticalityissues are summarized in Table 1.

2. Methodology

In order to evaluate example risks to supply, the environment,and the economy, several criticality components were selectedfor these broad criticality risk groups. Many indicators or metricsexist for any of these components; the selection of the indicatorslisted in Table 2 was motivated by broad applicability to the PVmaterials of interest and data availability. A key challenge inassessing criticality is to synthesize and appropriately weigh indi-cators of various scales and units. Previous studies have aggregatedand weighed multiple indicators based on national priority or arbi-trarily [8,33]. For a clear comparison, this work uses percentages ornormalization to characterize the various criticality indicators.Many of the calculated metrics are universal in nature such asembodied energy, material reserve bases, or political stability,however, this particular analysis often takes a United States basedapproach, for example using national primary prices, using the USas the denominator for calculating import reliance, and using tox-icity scores developed by the U.S. Environmental Protection Agency

Table 1Potential critical solar PV metals considered for this study.

Material Previously identified criticality issues Source

Aluminum (Al) Economic importance [24,31]Defense/Military importance

Arsenic (As) Toxicity [30]High import reliance

Cadmium (Cd) ToxicityCopper (Cu) Defense/Military importance [31]Iron (Fe) Global demand [27]Gallium (Ga) Low substitutability [24,28,32]

Recycling constraints [8,27]Producer trade restrictionsImport relianceImportance to ‘‘clean energy’’Carbon footprint of mining and production

Germanium (Ge) Economic importance and supply risk [24,32]Substitutability [27]Carbon footprint of mining and production

Gold (Au) Carbon footprint of mining and productionIndium (In) High demand from emerging technologies [24]

Technical difficulty of recycling andsubstitution

[28,32]

Import reliance [8]Secondary production constrainedImportance to ‘‘clean energy’’Geological scarcity

Molybdenum(Mo)

Economic importance [24]Limited number of mining corporationsSubstitutability

Platinum (Pt) Regional concentration of mining [32]Recycling restrictionRapid demand growth

Selenium (Se) Net import reliance [24]Silicon (Si) Recycling constrained [27]

Global demandSilver (Ag) Toxicity [27]

Political instability of producersClimate change vulnerability of producers

Tellurium (Te) Economic importance [24]Recycling constraints [8,32]Importance to ‘‘clean energy’’ [27]Geological scarcity

Tin (Sn) Substitutability, political instability ofproducers

[27]

Zinc (Zn) Economic importance [24,31]Defense/Military importance [27]Political instability of producers

388 M. Goe, G. Gaustad / Applied Energy 123 (2014) 387–396

Author's personal copy

(EPA). This methodology could be applied to any individual nationor even used at the global level, in order to better reflect the policygoals of other stakeholders. For example, in the European Union,Restriction of Hazardous Substances (RoHS) could be used as a sub-stitute for the EPA toxicity scores and each individual nation couldsubstitute their own import reliance.

2.1. Calculation of supply risk indicators

Here supply risk metrics refer to two components of scarcityidentified by Alonso et al. [42]: physical resource constraints andinstitutional inefficiency. Physical scarcity refers to the actualamount of economically extractable material available as quanti-fied by its reserves and reserve base. Physical scarcity also takesinto account resource quality and effort required to mine and re-fine it. Institutional inefficiency refers to scarcity issues that arisenot due to actual physical constraints but to other supply disrup-tions such as war, strikes, natural disasters, and other supply bot-tlenecks. The indicators selected to represent this second type ofscarcity are net import reliance and the Hirfindahl–Hirshmann In-dex, a measure of socio-political instability.

Net import reliance is defined as the ratio between net importsto apparent consumption, see Table A.1 in Appendixfor values. Netimport is defined by the U.S. Geological Survey (USGS) as the dif-ference between imports, exports, and stock changes. Apparentconsumption is defined as the summation of production, imports,and stock changes minus exports. For some non-PV materialsapparent consumption has been shown to significantly underesti-mate total consumption due to the imports and exports of productsthat contain large amounts of a material [43].

We follow previous studies [24,44,45] that make use of the Hir-findahl–Hirshmann Index (HHI) to characterize the relative supplyrisk related to socio-political stability of material and mine produc-ers. The assumption here is that the greater the socio-political sta-bility the smaller the risk of supply disruption. This index indicatesthe concentration of mineral production (SC) obtained from eachcountry (C) with low World Governance Indicators (WGI) PoliticalStability Absence of Violence (PSAV) scores. The WGI utilizes sur-vey and expert data from 30 sources to develop an index scorefor 213 countries. The score rates political stability on a scale frompoor (�2.5) to good (+2.5). For example, in 2010, the US, ranked56th percentile with a score of 0.31 ± 0.23 while China ranked inthe 24th percentile with a score of �0.76 ± 0.23. The WGI scorewas scaled and inverted so that a score of zero indicates good sta-bility and poor stability is a score of 10. A single score metric forsuch a complex characterization is challenging and some studieshave cited issues with data completeness and uncertainty[46,47]. However, this metric provides a widely accepted first passindicator for national stability.

HHI ¼X

C

ðSCÞ2WGIC ð1Þ

The supply risk indicators chosen to represent physical scarcity is-sues are the ratio of production to reserves and recycling rate. TheUSGS defines a reserve base as resources that have a ‘‘reasonablepotential for becoming economically available within planninghorizons beyond those that assume proven technology and currenteconomics’’. Similarly, reserves are defined as the part of the reservebase that could be ‘‘economically extracted or produced at the timeof determination’’. Reserve base and reserve data is important toquantify in relation to future demand (apparent consumption) inorder to extrapolate potential depletion time scales. A high recy-cling rate indicates the availability of a robust secondary supplychain which can mitigate supply-chain disruptions arising fromphysical scarcity of primary resources. Average 2011 US recyclingrates were used from the USGS.

2.2. Calculation of economic risk indicators

Three indicators to represent economic risk were quantified forthis work: primary price, domestic consumption, and contributionto the economy. The market price of a primary material is one indi-cator of the cost of mining, extracting, refining, and transportingmaterials. The price signals both the product value and the costto acquire an alternative; here USGS average prices were used[35]. To calculate a proxy metric for contribution to the economy,the largest market sectors by product volume for each metal wereobtained from USGS and Graedel et al. [48]. In order to quantify thevalue to the economy of each material we estimated the contribu-tion of each industrial sector to the U.S. Gross Domestic Product(GDP) using Gilmore et al. [49] shown in Table 3. For most materi-als, such as indium, product applications spanned multiple indus-trial sectors and therefore were allocated to multiple GDPcategories. A total percentage was then calculated by adding theindividual product or industry amounts for a particular PV materialand then dividing by the total GDP.

2.3. Calculation of environmental risk indicators

Environmental risk spans a wide array of impacts to humanhealth and the natural environment resulting from energy use,consumption, and toxicity of materials throughout their life-cycle.Three specific indicators were chosen for this work to representenvironmental risk: embodied energy, an Environmental Protec-tion Agency toxicity score, and energy savings obtained throughrecycling. High primary material embodied energy (EPE) can bean indicator for environmental risk (as well as economic); disrup-tions in the energy or fuel supply chain will cause a strong rippleeffect throughout the life-cycle of these high EPE materials. Onestrategy to mitigate this supply chain risk is recycling which pro-vides a domestic supply located near demand. The difference be-tween the energy required to produce a material from primaryversus secondary (i.e. recycled) sources is the energy savings. High

Table 2Criticality related risk, broad risk components, selected indicators, and data sources.

Criticality related risk Components Indicators Data sources

Supply Institutional inefficiency Net import reliance [34,35]Hirfindahl–Hirshmann index of primary material and ore producers

Physical scarcity Recycling rate [30,36,37]Ratio of production to reserves

Environmental Human toxicity CERCLA points [38,39]Energy intensity Primary embodied energy

Energy savings

Economic Material specific Primary material price [35,40,41]Economy-wide Domestic consumption

Economic value by sector

M. Goe, G. Gaustad / Applied Energy 123 (2014) 387–396 389

Author's personal copy

energy savings indicate a strong motivation for recycling and lowerrisk given that a recycling infrastructure is in place for that specificmaterials. Primary embodied energy (EPE) was obtained from Sim-aPro 7.3.3 using the Ecoinvent database v2.0 according to [50]; sav-ings can generally be assumed to be linearly related to the primaryembodied energy (although this may not always be the case).Materials where both primary and secondary energy data wereavailable in the Ecoinvent database were regressed to determineregression coefficient a and B in Eq. (2). These were found to be�0.9762 and 16.361, respectively, with high correlation to existingdata (R2 = 1). Equation 2 was then used to estimate energy savingsfor materials where secondary embodied energy data was notavailable (i.e. Se, Cd, Zn, Te, In, Ga, Mo, As) [51].

Energy savings ¼ aðEPEÞ þ B ð2Þ

The risk to human health is quantified by the Environmental Protec-tion Agency’s (EPA) Comprehensive Environmental Response, Com-pensation, and Liability Act (CERCLA) 2011 data according to [38].CERCLA evaluated and ranked the toxicity of 859 compounds basedon three equally weighted criteria: frequency of occurrence at na-tional priorities list (NPL) or Superfund sites, toxicity, and the po-tential for human exposure. Toxicity criteria score ranges from 0to 600 and was evaluated by the compound’s ignitability or reacta-bility, aquatic toxicity, chronic toxicity, carcinogenicity, and radio-nucleotides. The total score ranges from 0 to 1800. Gradel et al.[30] have made use of midpoint and endpoint metrics to calculatea Life-Cycle Impact Assessment called ReciPE in order to evaluateenvironmental implications of material decisions and many othercriticality studies have left out this aspect completely [8,26,52,53].Similar to our approach, Gradel et al. represents potential humanand aquatic toxicity from mineral extraction and refinement stages.However, Gradel et al. is limited because it does not include histor-

ical frequency of use or concentration in the environment. Othercommon environmental health and safety metrics include: permis-sible exposure limit (PEL), recommended exposure limit (REL), EPAcarcinogen classification, reference dose (RfD), threshold limit value(TLDV). Of these, all but TLDV, which is specific to material exposurefor a worker, is considered in the CERCLA toxicity score.

3. Criticality policy and indicators

3.1. Things governments can do to address criticality

There are a myriad of things that governments can do to ad-dress criticality issues. We argue here that a comprehensive ap-proach to criticality requires a critical look at the implications ofusing a single indicator or a particular set of narrowly focused indi-cators. For the PV materials studied, we find that the use of multi-ple economic, environmental, and supply indicators revealspotential opportunities and tradeoffs of policy actions. We haveproposed seven policy categories, discussed throughout Section 3.2,each that suggest a potential policy mechanism using a single sup-ply, economic, or environmental indicator in Table 5. These policiesare aligned vertically along a sliding scale from direct or command-and-control to moderate strategies. Each category then has two po-tential policy mechanism examples: a high expense and low ex-pense option.

The most direct policies of imposing critical materials importand export taxes have important supply, environmental, and eco-nomic consequences. On the one hand, import taxes could increasedomestic production and encourage the development of substi-tutes for critical materials. Alternatively, import taxes could alsoincrease the domestic price high enough to slow the US transitionto renewable energy, a consequence that has broad human healthand climate change impacts. Additionally, taxing imports fromadversarial nations could isolate and further marginalize thesesocieties whose economic cooperation may be one strategy to in-crease their political stability. Similarly, waste export tax, could in-crease recycling rate but also increase toxicity issues in currentlandfills if these materials are not properly recovered. See Appen-dix A.3 for more discussion of direct strategies to addresscriticality.

Whereas imposing a tax serves as a deterrent for certain activ-ities, offering subsides or grants uses economic incentives toencourage activities with the same goal of reducing material criti-cality. A subsidy of domestic mining of critical materials wouldincentivize domestic production or otherwise potentially lead toan increase in prices and primary energy when reserves are moreexpensive to develop domestically than abroad. However, if theUS has domestic expertise in mining for a particular material then,the primary material cost and embodied energy can be reduced.Another subsidy we propose is for R&D of electronics recoveryand product manufacturing from secondary materials. The goalof this policy is to increase the recycling rate and availability ofcritical materials. Potential advantages include reduced humanhealth impacts when materials formerly disposed in landfills arecollected for end-of-life recovery. In addition, R&D investmenthas the potential of increasing economic activity from the creationof new markets that make use of secondary materials.

Another policy aimed at reducing supply risks without directlyinfluencing markets is information. For example, disseminatinginformation on a streamlined patent process for technologies thatutilize substitutes. Like R&D investment this policy has the poten-tial to increase economic activity in renewable energy sectors andother new markets. Another form of information policy is develop-ing an online secondary materials exchange or published reportson critical material availability. There are many advantages to

Table 3GDP market sector allocation for products and material combinations.

GDP Market Sector Material – product orindustry

Manufacturing In – Pb free soldersFe – industrial machinesGe – catalystsZn – galvanization, alloysTe – chemicals, alloysSe – alloysMo – chemicals, steel,stainlessCd – pigmentsAg – industrial

Construction Fe – constructionAl – buildingsSe – glasses

Retail trade Ag – photography,jewelryCd – batteriesZn – alloysSe – glasses

Transportation and warehousing Cd – batteriesFe – transportationSe – glasses

Agriculture Se – agriculture

Information communication technologyproducing industries

As – photovoltaicsGe – fiber opticsIn – LCD (TV), monitors,Ga – IC, optoelectronicsGe – fiber optics, infraredopticsTe – electronicsSe – glasses

390 M. Goe, G. Gaustad / Applied Energy 123 (2014) 387–396

Author's personal copy

encouraging knowledge accumulation of the domestic supply anddemand flows; the downside is the risk of exposing domestic sup-ply vulnerabilities to potential terrorists.

Education on the other hand, focuses on domestic expertisebuilding by providing training to material recovery facilities(MRFs) in secondary recovery techniques, to local state govern-ments on ways to mitigate climate change without the use of crit-ical resources, and to MRFs on increasing resource recoveryefficiency of critical materials. Education policies have the advan-tage of potentially increasing recycling rate and reducing toxicityrisks. However, an unintended consequence to discouraging theuse of critical materials may be the divestment from current PVtechnology.

3.2. Comparison of single category metrics

Table 4 shows the individual ranking results for each of 10 indi-cators; it also ranks from 1 to 17 in order of greatest to least theaggregated economic, supply, and environmental risk indicators.In this normalization scheme, the higher the rank number, thesmaller the relative criticality risk for that indicator. These individ-ual metrics are described in detail previously in Sections3.2.1–3.2.3. The aggregated ranking assumes equal weighting foreach of the indicators, which skews the results toward supply riskconsiderations as they make up 40% of the total. It should be notedthat results can be skewed to favor any set of indicators, given thegoals and agenda of the decision-maker. This particular analysis ismeant to focus on supply chain risk while considering the influ-ence of environmental and economic indicators in a systems per-spective. This approach determined PV materials in order of mostto least critical for this particular combined set of indicators are:Ge, Pt, As, In, Sn, Ag, Se, Si, Te, Cd, Zn Au, Ga, Cu, Mo, Al, and Fe.

Our criticality designation is based on an ordinal ranking and islimited in its ability to measure how far apart two materials are.For this reason, we also plot risk indicators along an axis (Figs. 3–5) to gain a perspective about relative risk compared to normalizedmetrics e.g. %GDP. Unfortunately, there is no clear benchmark orline which we can draw that determines whether a material is crit-ical or not from the US perspective. Criticality determinations ulti-mately depend on stakeholder priority, available information, andfuture demands. For example, if the US prioritizes short term ac-cess and availability of materials for national security as indicatedin recent literature [31], then Pt, Se, Te and In may be determinedto be non-critical materials (due to either a low import reliance or

a high production to reserve ratio). Here we attempt to align PVmaterials relative to one another rather than to make absolutejudgments on criticality.

3.2.1. Supply risk indicator resultsAccording to the political instability indicator all of the PV

materials and base metals studied have high supply risk whencompared various distribution scenarios, see Appendix A.1. Histor-ical data indicates that the trends of unequal distribution and theconcentration of production have become more pronounced forall PV materials over the last decade (except Au, Se, Pt, and In).These trends indicate a shift in production towards single countrydominance. For example, in 2012, 11 of 17 materials studied hadone producer, China, which held 30–60% share of primary produc-tion as compare to the rest of the world (ROW) as shown in Fig. 1.In general, as the concentration of production increased, the pro-ducer political instability increased. The presence of one or severalextremely unstable non-dominant producers e.g. Somalia has littleimpact on the overall political instability of the supply chain wheresingle country dominance is most severe. This implies that for allmaterials, policies that enable equal distribution among producersis more effective at decreasing the supply risk (as measured byHHI) than any other ’single country’ approaches e.g. encouragingmore production from individual producers that are very stableor improving conditions of very unstable non-dominant producers.

Similar to HHI, net import reliance (NIR) is an indicator of thequality and effort required to obtain physical resources. All PVmaterials except Au, Cd, Mo, Fe, and Se, are dependent on importsto meet greater than 25% of apparent consumption. Similar to glo-bal production, the bulk of domestic imports are obtained from oneor two producers as shown in Fig. 2. From 2005 to 2012, NIR of amajority of PV materials have remained constant as shown inTable A.1 of the Appendix; Al and Si have decreased reliance, whileZn and Ge have increased reliance. Improving NIR requires an in-crease in domestic production and a decrease in imports. There-fore, this indicator encourages policies that increase domesticcontrol of key material resources e.g. protective tariffs, miningindustry subsidies. In general, one would expect that greaterdomestic control increases security. However, in the case ofunforeseen domestic supply disruptions e.g. weather, extremeadherence to this strategy would decrease security both domesti-cally and globally. Furthermore, because NIR narrow focus ondomestic supply chain quality, it provides a false sense of materialsecurity. That is, due to the complexity and interdependence of the

Table 4Criticality priority of PV materials and indicator values where arrows indicate historical trends.

Rank Mat. HHI Net importreliance (%)

Recyclingrate (%)

Ratioproduction toreserves

CERCLAscore

Relative primaryembodied energy(Fe = 1)

Energysavings(%)

Primary price($1000/ton)

Consumption($M)

Econ. sectorvalue (%GDP)

1 Ge 3.28 0.90 0.30 4 255 4 0.60 1380 55 0.162 Pt 3.05 0.89 0.65 85 – 10,515 0.95 50,798 8331 0.179 Te 1.56 0.80 <0.01 120 196 6 0.87 155 31 0.167 Se 2.25 1.00 0.38 49 778 103 0.97 567 51 0.163 As 2.28 1.00 <0.01 19 1665 25 0.95 2 11 0.046 Ag 1.90 0.73 0.22 23 608 9 0.90 21 875 0.16

11 Zn 3.00 <40% <0.01 168 919 41 0.96 3 1834 0.164 In 0.76 E <0.01 2 288 2 0.63 128 61 0.295 Sn 0.71 0.64 0.32 20 488 166 0.97 999 5892 0.178 Si 0.62 0.99 0.18 53 – 121 0.97 556 19 0.04

10 Cd 1.00 0.74 0.27 23 1319 1 0.40 2 1824 0.1713 Ga 1.41 <1% 0.33 43 112 2 0.67 28 497 0.1116 Al 0.98 E 0.14 158 685 2 0.60 2 1 0.2515 Mo 0.59 E 0.18 42 442 12,511 0.98 53,434 8015 0.1012 Au 1.25 0.20 0.36 20 – 6 0.78 2 9099 0.1514 Cu 1.56 0.07 0.41 42 805 1 0.96 1 65,448 0.2117 Fe 0.67 0.34 0.30 53 – 1 0.71 8 14 0.19

M. Goe, G. Gaustad / Applied Energy 123 (2014) 387–396 391

Author's personal copy

global materials markets, a supply disruption can be catastrophicto the entire material supply chain despite any particular domesticreliance.

HHI and NIR are indicators for institutional inefficiency; how-ever, more straightforward scarcity metrics i.e. ratio of reservesto production and recycling rate measure actual physical quanti-ties of resources available. Where primary reserve data for PVmaterials is unavailable due to abundance e.g. Si or scarcity e.g.Ga it has been estimated from a combination of base metal re-serves, ore grade, and refining efficiency assumptions, as discussedin Appendix A.1.3. When compared to global production, reservesare 2–120 times greater for most PV materials (with the exceptionbase metals Al, Zn and Fe). If we assume no recycling, constant pro-duction, and no changes to stock or reserve estimates, reservescould be depleted in a few generations for half of the PV materialsstudied. Under these assumptions, In, Ag and Ge have the greatestrisk of depletion in the next 20 years. Since demand is increasing,reserves are likely to be depleted even sooner. Although, increasingprice, technological efficiency, the discovery of new reserves, andincreasing recycling rates are all also delaying the steady March to-wards resource depletion. In order to further quantify supply risksdue to future consumption Angerer et al. [37] posited that by 2030consumption levels of Ga, In, and Pt would exceed current produc-tion by 2–6 times. Historically, future consumption of PV materialshas exceeded past production by 1.8–2.5 times for 20 and 30-yearoutlooks since 1980. An exception to this steady trend is Ga whoserecent consumption is nearly 11 times the production of 30 yearsprior. This rapid increase in demand was due the expansion of elec-tronics e.g. computers that required Ga for integrated circuits andoptoelectronics e.g. light emitting diodes (LEDs) and solar cells. Theindicators of physical scarcity for primary resources discussedabove can elucidate broad implications for disruptions in supply.For example, supply disruptions for Ga could have slowed develop-ment and innovation of information and communications technol-ogy (ICT) that now account for 650 billion dollars or 4% of U.S. GDP.However, the growth of new industries such as ICT is often unpre-dictable. Therein lays the problem of relying on physical scarcityindicators to evaluate supply risk: future uncertainty makes anyassertions about depletion unreliable. Despite this uncertainty,

Table 5Things that governments can do to mitigate criticality issues of PV materials.

China ROW

0% 25% 50% 75% 100%

ZincCopper

SeleniumCadmium

AluminumTin

Iron (raw IndiumArsenicSilicon

Germanium

Share of World Production

Fig. 1. Share of world primary production held by a single producer, China, ascompared to the rest of the world (ROW) of various PV materials.

China

China

China

China

Chile

Canada

Canada

Canada

Canada

Canada

Germany

Germany

Mexico

Russia

Australia

Belgium

Belgium

UK

others

others

others

others

others

others

others

others

others

others

0% 25% 50% 75% 100%

Indium

Selenium

Copper

Cadmium

Aluminium

Silicon

Iron ore

Tellurium

Gallium

Germanium

Share of U.S. Imports

Fig. 2. Share of US imports held by the top one or two producers.

392 M. Goe, G. Gaustad / Applied Energy 123 (2014) 387–396

Author's personal copy

the use of these indicators has driven aggressive policies such asstockpiling and investing in the development of new reserves thatseek to avoid short-term supply disruptions due to depletion.Alternatively, physical scarcity indicators can be utilized to driveless aggressive policies such as increasing new and old scrap recy-cling which seek to delay depletion until more abundant substi-tutes are developed.

Recycling rates of base metals i.e. Zn, Fe, Cu and precious metalsi.e. Au, Ag, Pt are between 27–41% and 18–65%, respectively; thehighest of all PV materials. In general, as observed by Graedelet al. [54], recycling rates are more closely related to material appli-cations (i.e. use volume and ease of recovery) than physical scar-city; materials embedded in small amounts in complexelectronics e.g. Si, As have lower recycling rates than those usedin large volume products with less material mixing e.g. Cd in batter-ies and Sn in cans. Exceptions to this trend occur with In, Ga, and Gefrom electronics which is likely due to its high primary (and there-fore secondary) price. Use phase barriers to recycling also includethe level of dissipative use, reuse in markets without a recoveryinfrastructure, and product lifetime. Recycling rates reflect not onlythe use phase but also end-of-life barriers to secondary production.For example, all materials studied have a nearly zero recycling ratewith respect to its use in PV applications, despite the rapid deploy-ment in PV applications and attention to metal criticality concernsover the last few decades. Lack of PV recycling has been attributedto low economic incentives [55], in adequate recovery technologyor infrastructure [56–58], and the lack of policy incentives [59].Shortages in base or by-product metals could impact energy secu-rity; PV recycling initiatives may be able to delay these impacts un-til more abundant substitutes are developed. Recognizing thisopportunity, the EU has included solar PV modules in its corporatetake back mandate for electronics [60] and the National RenewableEnergy Laboratory (NREL) has funded research in alternative PVcompositions that avoid physically scarce materials [61]. Recent lit-erature has also suggests that recycling photovoltaics could signif-icantly reduce lifecycle energy [62].

3.2.2. Economic risk indicator resultsPrecious metals, Au, Pt, and Ag, have the highest primary price

of the PV materials studied followed by In, Ga, and Ge. As previ-ously discussed, these metals also have greater physical and insti-tutional scarcity issues as compared to other PV materials. Weobserve, in general, high volume materials e.g. Fe, Al, and Zn aretypically lower priced; and low volume materials e.g. In, Ga, Ge,and Au are higher priced. Exceptions include materials with lowprice and low volume i.e. Si, Cd, and As. In addition, high volumematerials also have higher consumption value in dollars. It is there-fore expected that those materials with higher consumption willcontribute more to the U.S. economy as measured by share of sec-tor GDP. However, this is not necessarily the case. We observe thatdue to the methodology chosen the more economic sectors thatutilize a material, the greater a material’s economic importanceindicator. For example, although Ag, Cd, and Se have much lowervolumes than Al they are utilized in more economic sectors andtherefore are determined to have a greater share of sector GDP. De-spite the limitations of this method, high volume materials such asFe, and Zn were also determined to be among the most importantmaterials to the economy when using share of sector GDP as anindicator. From a policy perspective, the indicators that identifymaterials of greatest importance to the economy can drive strate-gies to shield markets from the economic ramifications of lack ofmaterial availability. These strategies may include governmentfinancing of vulnerable materials markets, enforcing price controls,or offering subsidies for sectors transitioning to abundant substi-tutes or recycling waste materials.

3.2.3. Environmental risk indicator resultsThe environmental indicators measure two aspects of risk: en-

ergy intensity and toxicity, discussed below. Arsenic has been his-torically used as a chemical weapon, insecticide, and a medicinebefore being identified as a carcinogen. Cadmium dust inhalationand zinc ingestion are also toxic, having been known to cause poi-soning and the absorption disruption of essential minerals. Notsurprisingly, materials such as As, Cd, and Zn have the greatest tox-icity issues, while Pt, Si, and Fe each have the least toxicity. Most(i.e. 10 of 14) of the PV materials included in this study have CER-CLA scores in the top 51 percentile of the 859 materials list. Ofthese, 7 materials are in the top 25 percentile. What separatesthe materials with more acute environmental impacts are theirexposure score, a portion of the calculated CERCLA metrics. Over6 years (from 2005 to 2011) there has only been minor movementof CERCLA scores between 1% and 5% for PV materials; Si, Ge, andGa are the notable exceptions. Although Si maintains a ranking of699 of 859 materials, its CERCLA score has doubled in this periodof time due to its increased concentration in the environment. BothGa and Ge have moved from up in ranking due to increases inexposure, toxicity and frequency in the environment. In and Te alsosaw minor CERCLA score increases due to concentration in theenvironment over this 6-year period. These trends show that PVactive materials are becoming more prevalent at Superfund sites.Increases in domestic PV production, adoption, and end-of-lifelandfilling will likely increase this trend. Policy can mitigate theseenvironmental impacts by encouraging secondary production sothat materials avoid landfilling, investing in more efficient pollu-tion control technology for production facilities, and increasingthe penalty for environmental dumping of critical materials.

Primary embodied energy and energy savings (from the use ofsecondary materials) measure the energy intensity of environmen-tal risk. When PV materials are compared to more common mate-rials such as Fe, they all exhibit greater energy intensity; this isespecially the case for Pt, Au, Ag, and Ga, which are several ordersof magnitude greater than Fe. Zn, Cu, Cd, and Se have the lowestprimary energy intensity of within 10–90% of Fe. In terms of energysavings from the use of secondary materials, Pt, Au, Ag, and Gayield the greatest benefits, while Zn, Ge, Cd, and Mo yield the leastenergy savings. This is expected since in general, the greater theprimary energy the greater the energy savings from the use of sec-ondary materials. These energy savings are due to avoidance ofhigh energy processing e.g. extraction from mines, electrolysis,refining that is not needed for most secondary processes which in-volve physical separation and re-melting.

However, unexpectedly there is also an empirical relationshipbetween CERCLA score and primary energy intensity. As the pri-mary energy intensity of a material increases the CERCLA score de-creases. When coupled with material price information, thisrelationship may explain why less energy intensive and less expen-sive materials are found with greater frequency and concentrationat Superfund sites despite their high toxicity e.g. As, Zn, Cu and Se:their recovery is not efficient from an energy or economic stand-point. These relationships can also motivate policy that uses eco-nomic mechanisms to drive secondary production by increasinglandfill tip fees, taxing waste exports, or mandating a portion ofgovernment technology purchases meet a recycled contentstandard.

3.2.4. Multi-metric resultsTraditional strategies aimed at addressing criticality concerns

have relied on single metric command and control policies e.g.stockpiling, direct governmental take-over of mineral producers,protective tariffs, etc. These command-and-control policies havebeen criticized for their narrow focus on physical scarcity and theireconomic inefficiency [63–65]. Our intent is to show the use of

M. Goe, G. Gaustad / Applied Energy 123 (2014) 387–396 393

Author's personal copy

multiple metrics with a lifecycle perspective can lead to a systemslevel approach to criticality policy that addresses not only physical

scarcity but also institutional inefficiency, environmental impacts,and economic risks.

Figs. 3–5 shows simultaneously the economic, supply, and envi-ronmental risk of PV materials using four different sets of indica-tors. The x-axis for Figs. 3 and 4 are the same; therefore, we canobserve vertical shifts and diameter size changes across figuresto understand the impact of different economic, supply and envi-ronmental indicators on the overall criticality order. For Figs. 3and 4, increasing economic and environmental risk is towardsthe top right corner of the figure. The circles represent the supplyrisk indicators; circle diameter size changes from small to largeindicate increasing supply risk. In Fig. 3, As, Cd, and Zn are the mosttoxic according to their CERCLA score as described in the method-ology; they appear higher on the y-axis, while precious metals Au,Pt are the least toxic and therefore lowest on the y-axis. Au and Ptare the furthest to the right on the primary price x-axis as they arethe most expensive. Combining these two indicators, Fe and Si arethe least expensive and least toxic i.e. closest to the origin. The cir-cles show a supply side risk indicator: socio-political instability viathe Hirfindahl–Hirshmann Index as described in Section 2.1; thelarger circles indicate relative increased supply risk. Ge, In, Si andAs have the highest producer socio-political instability. For thisset of three metrics there is a tradeoff between price and CERCLAscore for PV materials. As previously observed, this relationshipmay explain why less energy intensive and less expensive materi-als e.g. As, Zn, Cu, Se are found with greater frequency and concen-tration at Superfund sites despite their high toxicity: their recoveryis not efficient from an energy or economic standpoint. To a lesserextent there is also a relationship between price and producerpolitical instability, in general, the higher the price the greaterthe producer political instability. As stated above, since price is alsoinversely related to domestic consumption, materials with higherpolitical instability also have lower domestic consumption. Theserelationships suggest that policies aimed at reducing a single met-ric may impact multiple attributes of energy security. These resultswould suggest that a policy aimed at, for example, reducing pro-ducer political instability, would likely end up influencing environ-mental impacts to human health as well as economic impacts.Taking a multi-metric approach to criticality policy can highlightthese tradeoffs and provide sensitivity analysis. One such exampleis the promotion of sustainable mining in politically unstable na-tions where the actions to reduce negative environmental impacts(e.g. groundwater contamination of mining operations) are cou-pled with technology efficiency improvements, stable-livingwages, and community stakeholder involvement.

These criticality tradeoffs are more pronounced in Fig. 4 where,the y-axis is primary embodied energy as our environmental indi-cator, and the circles represent the ratio of reserves to production,a different supply risk indicator. The x-axis is the same as Fig. 3:primary price as an economic indicator. Given this particular setof indicators, Au, Pt, In, Ga, and Ag are the most critical while againFe is the least. For this set of metrics the risk increases with respectto price and primary energy but decreases with respect to the ratioof reserves to production. These relationships show that in general,energy intensive materials are at greater risk for depletion and aremore expensive. As previously stated, the greater the primary en-ergy, the greater the energy savings from secondary production.These relationships suggest that strategies aimed at increasingrecycling may work to simultaneously address physical scarcityand energy consumption for economically valuable materials mar-kets. Alternatively, strategies that seek to open new mine reservescould decrease price while increasing negative human health andenvironmental impacts of certain critical materials.

Fig. 5 focuses on several supply chain risk indicators, the y-axisuses net import reliance as a metric to measure institutional inef-ficiency, a scarcity indicator and the x-axis uses recycling rate as a

Mo

Ga

Ag

Au

Ge

SeSnAl

Cd

Fe

In

PtSi

As

Te

Cu

Zn

(200)

300

800

1,300

1,800

1.E+02 1.E+04 1.E+06 1.E+08

Env

iron

men

tal r

isk

: CE

RC

LA

sco

re

Economic risk: primary price ($/ton)

=1.0 producer instability score

Fig. 3. Relative criticality of PV materials using CERCLA score as an environmentalindicator (y-axis), primary price as an economic indicator (x-axis), and socio-political stability as indicated by the Hirfindahl–Hirshmann Index as a supply riskindicator (size of the circles).

Ag

Au

Cu

Ga

MoSe

Sn

AlCd

Fe

Ge

In

Pt

SiAs

Te

Zn1

10

100

1,000

10,000

100,000

1,000,000

100 10,000 1,000,000 100,000,000

Env

irio

nmen

tal r

isk:

pr

imar

y em

bodi

ed e

nerg

y (M

J/to

n)

Economic risk: primary price ($/ton)

= 50 ratio of reserves to production

Fig. 4. Relative criticality of PV materials using primary embodied energy as anenvironmental indicator (y-axis), primary price as an economic indicator (x-axis),and the ratio of reserves to production as a supply risk indicator (size of the circles).

Ag

Au

Cu

Ga

MoSe

Sn

Al

Cd

Fe

Ge

In

Pt

Si

As

TeZn

(0.1)

0.1

0.3

0.5

0.7

0.9

1.1

(0.1)0.1 0.3 0.5 0.7

Inst

itutio

nal i

neff

fici

ency

:ne

t im

port

rel

ianc

e (%

)

Physical scarcity: recycling rate (%)

=100CERCLA score

Fig. 5. Relative criticality of PV materials using net import reliance, recycling rate,and CERCLA score.

394 M. Goe, G. Gaustad / Applied Energy 123 (2014) 387–396

Author's personal copy

physical scarcity indicator. Low recycling rates indicate a lack ofsecondary supply; a robust recycling infrastructure can mitigateboth physical scarcity and supply disruptions. Increasing supplychain risk is towards the top right corner of the figure; the increas-ing size of the circles represents increasing environmental risk asquantified via the CERCLA score. Appendix A.3 contains additionaldiscussion about multi-metric supply risk tradeoffs. Given this setof metrics, As, Te, Ga, Ge, and In are the most critical while Fe is theleast. Institutional inefficiency (supply chain disruption) and envi-ronmental risk are weakly related for some PV materials giventhese metrics. In general, as import reliance increases, the toxicityrisk and frequency at Superfund sites decreases. This relationshipsuggests that domestic environmental risks related to miningdiminish as our reliance increases. The global impacts of this activ-ity include greater environmental impacts (e.g. climate change andhuman health problems) that are being shifted to other parts of the(developing) world where there may be less stringent environmen-tal standards and lower technology efficiency. Policymakers canutilize this set of metrics to develop comprehensive strategies thatpromote secondary production, domestic mining, or investment inthe efficiency and environmental safety of foreign miningoperations.

The relationships observed from Fig. 5 demonstrate how aggre-gating an indicator may conceal the underlying reasons for relativerisk rankings. In the case of supply risk determinations, somematerials (for this analysis Ge, As, Si, and In) have acute producerpolitical instability which may cause supply disruption via strikes,upheaval, or revolt whereas others have more import dependenceissues which may cause supply disruptions via tariff introductionor new import/export regulation (in this case In, Sn, Zn, Se, andAs). Other materials lack recycling infrastructure which presentsa potential physical scarcity issue such as Au, Se, and Te. Aggrega-tion of these metrics may confuse or cancel the individual effects.Si and Pt demonstrate another interesting case in terms of critical-ity determinations: that the use of particular indicators may dropsome materials off the list of concern. For example, Pt does not ap-pear at Superfund sites and therefore has no CERCLA score. Si is anabundant material; its reserve data is highly uncertain due to alack of concern; however, the high purity Si required for PV andother electronics applications requires abundant energy to refineit potentially making it vulnerable to energy supply disruptions.These types of tradeoffs make it difficult to accurately map Ptand Si relative to other materials using these metrics.

4. Conclusions

Our analysis determined PV materials in relative order of mostto least critical for these metrics are: Pt, Ge, Te, In, As, Si, Sn, Se, Mo,Ag, Cd, Zn, Ga, Au, Al, Fe and Cu. Of these, Se, Fe, Cd, Ag and Zn arethe most important materials to the economy in terms of the num-ber and size of industrial sector applications. In terms of consump-tion, less expensive materials e.g. Fe, Al have greater value dueeither to larger volumes. Compared to best and worst case scenar-ios, nearly all of the PV materials studied have high producer polit-ical instability due to the high concentration of production amongone or two producers. When PV materials are compared to morecommon materials such as Fe, they all exhibit greater energy inten-sity and most are present at Superfund sites. Multi-metric analysisreveals tradeoffs that suggest friction between sustainable eco-nomics, political stability of supply, and environmental qualityobjectives. We have proposed moderate, long-term policies e.g.education, subsidies, information and aggressive, short term po-lices e.g. expanding defense, financing, regulation, and taxes aimedat delaying or mitigating criticality issues of PV materials. Moder-ate policies may require coordination between federal and state

governments whereas aggressive policies are more confrontationaltowards foreign governments, possibly sparking further conflictwith adversarial nations in this realm.

There are questions remaining in this area particularly aroundfuture criticality, policy, and resource management. Current trendsindicate that many countries are moving towards aggressive ac-tions to secure resources necessary for economic growth and infra-structure development while material prices, and consumptionrates increase. These factors point to increasing conflicts over theappropriate and effective actions that may shield domestic mar-kets from supply disruption. Future work in this area is also re-quired to understand the global dynamics of criticality, forexample, what is the impact of increasing populations and afflu-ence of emerging economies on physical scarcity and the environ-ment? What politically feasible policies can decouple economicgrowth, resource depletion, and environmental impacts to mitigatefuture criticality risk? In addition, several policy interventions havebeen proposed here that could be explored further such as whetherencouraging recycling could mitigate criticality or if the U.S. shouldinvest in domestic mining to ensure future materials availability?Additionally, several previous studies have joined us in proposingthe use of substitutes, however, the supply, environmental, and so-cio-political tradeoffs of the proposed substitutes is unclear.

Appendix A. Supplementary material

Supplementary data associated with this article can be found, inthe online version, at http://dx.doi.org/10.1016/j.apenergy.2014.01.025.

References

[1] Wagner L. Materials in the economy-material flows, scarcity, and theenvironment. USGS Circular 2002;1221:1–34.

[2] Fthenakis VM, Kim HC. Land use and electricity generation: a life-cycleanalysis. Renew Sust Energy Rev 2009;13:1465–74.

[3] Sims RE, Rogner H-H, Gregory K. Carbon emission and mitigation costcomparisons between fossil fuel, nuclear and renewable energy resources forelectricity generation. Energy Policy 2003;31:1315–26.

[4] Houari Y, Speirs J, Candelise C, Gross RJK. A system dynamics model oftellurium availability for CdTe PV. Prog Photovolta: Res Appl2014;22(1):129–46.

[5] Candelise C, Winskel M, Gross RJK. Implications for CdTe and CIGS technologiesproduction costs of indium and tellurium scarcity. Prog Photovolta: Res Appl2012;20:816–31.

[6] Feltrin A, Freundlich A. Material considerations for terawatt level deploymentof photovoltaics. Renew Energy 2008;33:180–5.

[7] Andersson BA. Materials availability for large-scale thin-film photovoltaics.Prog Photovolta: Res Appl 2000;8:61–76.

[8] DOE US. Critical Materials Strategy. Washington, DC: US DOE USGS (UnitedStates Geological Survey) MCS; 2010.

[9] Desideri U, Campana Pietro Elia. Analysis and comparison between aconcenetrating solar and photovoltaic power plant. Appl Energy2014;113:422–33.

[10] Zhang W, Zhu R, Bin Liu, Ramakrishna S. High-performance hybrid solar cellsemploying metal-free organic dye modified TiO2 as photoelectrode. ApplEnergy 2012;90:305–8.

[11] Vats K, Tiwari GN. Energy and exergy analysis of a building integratedsemitransparent photovoltaic thermal (BISPVT) system. Appl Energy2012;96:409–16.

[12] Anctil A, Fthenakis VM. Critical metals in strategic photovoltaic technologies:abundance versus recyclability. Prog Photovolta: Res Appl 2012;21:1253–9.

[13] Zardetto V, Mincuzzi G, De Rossi F, Di Giacomo F, Reale A, Di Carlo A, et al.Outdoor and diurnal performance of large conformal flexible metal/plastic dyesolar cells. Appl Energy 2014;113:1155–61.

[14] Mercaldo LV, Addonizio ML, Noce Della M, Veneri PD, Scognamiglio A, PrivatoC. Thin film silicon photovoltaics: architectural perspectives and technologicalissues. Appl Energy 2009;86:1836–44.

[15] McClure JA. Stockpiling of strategic and critical materials. Idaho Law Rev1983;19:417–53.

[16] 50 USC 98. n.d.[17] Candelise C, Speirs JF, Gross RJK. Materials availability for thin film (TF) PV

technologies development: a real concern? Renew Sust Energy Rev2011;15:4972–81.

[18] Kapilevich I, Skumanich A. Indium shortage implications for the PV and LCDmarket: technology and market considerations for maintaining growth.

M. Goe, G. Gaustad / Applied Energy 123 (2014) 387–396 395

Author's personal copy

Presented at the Photovoltaic Specialists Conference (PVSC), 2009 34th IEEE2009:2055–60.

[19] Zweibel K. The impact of tellurium supply on cadmium telluride photovoltaics.Science 2010;328:699–701.

[20] Fthenakis VM. Sustainability of photovoltaics: the case for thin-film solar cells.Renew Sust Energy Rev 2009;13:2746–50.

[21] Govett G, Govett MH. Mineral resource supplies and the limits of economicgrowth. Earth-Sci Rev 1972;8:275–90.

[22] Jones GK. United States dependence on imports of four strategic and criticalminerals: implications and policy alternatives. BC Envtl Aff L Rev 1987;15:217.

[23] Santini J. Growing crisis in the strategic and critical minerals of the UnitedStates. J Legis 1980;7:63.

[24] European Commission. Critical raw materials from the EU: report of the Ad-hoc working group on defining critical raw materials. European Commission.Brussels; 2010.

[25] Reller A. Criticality of metal resources for functional materials used inelectronics and microelectronics. Phys Status Solid 2011.

[26] Centre for Policy Related Statistics. Critical Materials in the Dutch Economy:Preliminary Results, Statistics Netherlands, The Hague; 2010.

[27] Morley N, Eatherly D. Material security: ensuring resource availability for theUK economy. C-Tech Innovation Ltd.; 2008.

[28] National Research Council. Minerals, critical minerals and the US economy.The National Academies Press; 2008. p. 1–263.

[29] Graedel TE, Barr R, Chandler C, Chase T, Choi J, Christoffersen L, et al.Methodology of Metal Criticality Determination. Environ Sci Technol2012;46:1063–70.

[30] Nassar NT, Barr R, Browning M, Diao Z, Friedlander E, Harper EM, et al.Criticality of the geological copper family. Environ Sci Technol2012;46:1071–8.

[31] Thomason JS, Atwell RJ, Bajraktari Y, Bell JP, Barnett DS, Karvonides NS, NilesMF, Schwartz EL. From National Defense Stockpile (NDS) to Strategic MaterialsSecurity Program (SMSP): Evidence and analytic support. Volume1. Alexandria, VA: Institute of Defense Analysis (IDA); 2010.

[32] Buchert M, Schüler D, Bleher D. Critical metals for future sustainabletechnologies and their recycling potential. Nairobi: United NationsEnvironment Programme; 2009.

[33] Pfleger P, Lichtblau K, Bardt H, Reller A. Rohstoffsituation Bayern: KeineZukunft ohne Rohstoffe. Germany: IW Consult Munich; 2009.

[34] Kaufman D, Kraay A, Mastruzzi M. The Worldwide Governance Indicators.World Bank Policy Research Working Paper No. 5430; 2010.

[35] US Geological Survey. Mineral Commodity Summaries 2013. US GeologicalSurvey; 2013.

[36] USGS National Minerals Information Center. Iron and steel; 2012. p. 1–2.[37] Angerer G, Erdmann L, Marscheider-Weidemann F, Scharp M, Lullmann A,

Handke V, et al. Raw materials for emerging technologies 2009:1–19.[38] ATSDR C. CERCLA Priority List of Hazardous Substances. Agency for Toxic

Substances and Disease Registry; 2007.[39] Frischknecht R, Jungbluth N, Althaus H, Doka G, Dones R, Hischier R, Hellweg R,

Nemecek T, Rebitzer G, Spielmann M. Overview and Methodology. ecoinventreport no. 1. Dubendorf: Swiss Centre for Life Cycle Inventories; 2007.

[40] Graedel TE. On the future availability of the energy metals. Annu Rev MaterRes 2011;41:323–35.

[41] US Bureau of Economic Analysis. Table 5. Value added by industry group as apercentage of GDP; 2013. p. 1–18.

[42] Alonso E, Field FR, Kirchain RE. A case study of the availability of platinumgroup metals for electronics manufacturers. Electronics and the environment.ISEE 2008. In: IEEE international symposium; 2008. p. 1–6.

[43] Wilburn DR, Buckingham DA. Apparent consumption Vs. total consumption-Alead-acid battery case study. Reston (Virginia): US Geological Survey (USGS);2006.

[44] Rosenau-Tornow D, Buchholz P, Riemann A, Wagner M. Assessing the long-term supply risks for mineral raw materials—a combined evaluation of pastand future trends. Resour Policy 2009;34:161–75.

[45] Erdmann L, Graedel TE. Criticality of non-fuel minerals: a review of majorapproaches and analyses. Environ Sci Technol 2011;45:7620–30.

[46] Langbein L, Knack S. The worldwide governance indicators: six, one, or none? JDev Stud 2010;46:350–70.

[47] Thomas MA. What do the worldwide governance indicators measure? Eur JDev Res 2009;22:31–54.

[48] Graedel TE, Erdmann L. Will metal scarcity impede routine industrial use?MRS Bull 2012.

[49] Gilmore TL, Morgan ET, Osborne SB. Annual industry accounts: advancestatistics on GDP by industry for 2010. Survey of Current Business2011;91:8–24.

[50] Frischknecht R, Jungbluth N, Althaus H, Doka G, Dones R, Hischier R, et al.Overview and methodology. Final report ecoinvent data v2.0. Dubendorf, CH:Swiss Centre for Life Cycle Inventories; 2007.

[51] Wang X, Gaustad G. Prioritizing material recovery for end-of-life printedcircuit boards. Waste Manage 2012:1–11.

[52] Commission E. Criticial raw materials for the EU 2010:1–84.[53] National Research Council (US). Committee on Critical Mineral Impacts on the

US Economy, National Research Council (US). Committee on Earth Resources,National Research Council (US). Board on Earth Sciences and Resources,National Research Council (US). Division on Earth and Life Studies. Minerals,critical minerals, and the US economy; 2008.

[54] Graedel TE, Allwood J, Birat J. What do we know about metal recycling rates? JInd Ecol 2011;15:355–66.

[55] Choi J-K, Fthenakis VM. Economic feasibility of recycling photovoltaicmodules. J Ind Ecol 2010;14:947–64.

[56] Fthenakis VM, Choi J-K. Design and optimization of photovoltaics recyclinginfrastructure. Environ Sci Technol 2010;44:8678.

[57] Berger W, Simon F-G, Weimann K, Alsema EA. A novel approach for therecycling of thin film photovoltaic modules. Resour Conserv Recy2010;54:711–8.

[58] Marwede M, Berger W, Schlummer M, Mäurer A, Reller A. Recycling paths forthin-film chalcogenide photovoltaic waste – current feasible processes. RenewEnergy 2013;55:220–9.

[59] Fthenakis VM. End-of-life management and recycling of PV modules. EnergyPolicy 2000;28:1051–8.

[60] European Parliment and of the Council. on waste electrical and electronicequipment (WEEE); 2003.

[61] CRSP Funds 10 New Advanced Solar Research Projects, Announces FirstSponsored Research Program Project with Konarka Technologies; 2010. p. 1–2.

[62] Goe M, Gaustad G. Strengthening the Case for Recycling Photovoltaics: AnEnergy Payback Analysis. Appl Energy 2014. http://dx.doi.org/10.1016/j.apenergy.2014.01.036.

[63] Tullock G. The welfare costs of tariffs, monopolies, and theft. Econ Inq1967;5:224–32.

[64] Nichols AL, Zeckhauser RJ. Stockpiling strategies and cartel prices. Bell J Econ1977:66–96.

[65] Alonso E, Gregory J, Field F, Kirchain R. Material availability and the supplychain: risks, effects, and responses. Environ Sci Technol 2007;41:6649–56.

396 M. Goe, G. Gaustad / Applied Energy 123 (2014) 387–396