Global environmental impact assessment of the Pb-free shift

The Authors

Anders S.G. Andrae, National Institute of Advanced Industrial Science and Technology, Tsukuba, Japan

Norihiro Itsubo, National Institute of Advanced Industrial Science and Technology, Tsukuba, Japan

Atsushi Inaba, National Institute of Advanced Industrial Science and Technology, Tsukuba, Japan

Acknowledgements

The staff of the LCA methodology group at AIST is acknowledged for their support leading to the completion of this paper.

Abstract

Purpose – Using two different conceptual approaches to environmental life-cycle assessment, attributional and consequential, the purpose was to test the hypothesis that a typical lead free solder paste Sn95.5Ag3.8Cu0.7 is worse than Sn63Pb37 as far as global environmental impacts are concerned.

Design/methodology/approach – Single index weighting indices within the impact methodology Life cycle Impact Assessment Method based on Endpoint Modelling (LIME) impact methodology, were applied to the flows of three life cycle inventory models and their globally related flows. The LIME results based on three environmental impact categories, i.e. resource consumption, global warming and ozonelayer depletion are presented and discussed.

Findings – The attributional LCA (ALCA) results point towards a larger impact for Sn95.5Ag3.8Cu0.7 than Sn63Pb37 mostly due to the higher Sn and Ag content. This study confirms earlier similar ALCAs. The system expansion for the Consequential LCA (CLCA) did not change this conclusion.

Research limitations/implications – The present study has not included the affected microelectronics packaging parts of electronic products, nor has it included toxic effects as they are local. ALCA was considered to be equal to CLCA for Sn95.5Ag3.8Cu0.7, where no Ag nor Sn recycling was included.

Originality/value – For the first time a global environmental impact assessment of the shift to Pb-free solder paste using the LIME weighting method applied to a CLCA is reported. Environmental life-cycle investigations intended to support decisions of an ecological nature in the microelectronics packaging industry should benefit from the consequential approach.

Article Type:

Research paper

Keyword(s):

Solders; Modelling; Life cycle costs; Environmental management.

Journal:

Soldering & Surface Mount Technology

Volume:

Number:

Year:

2007

pp:

18-28

Copyright ©

Emerald Group Publishing Limited

ISSN:

0954-0911

Introduction

Since, the beginning of electronics production, solders have been used as interconnection materials and Pb has been widely used in solder alloys for its proven reliability in eutectic combination with tin (Plumbridge, 1996). However, the heavy metal Pb is under certain conditions harmful to humans, and it has therefore been banned from use in petrol, water pipes and paint. In Europe, an EU directive on restrictions on the use of hazardous substances required the elimination of Pb in electronics, in all but a number of special applications, by July 1, 2006 (European Union, 2003). Solder is the main Pb use within electronic products, therefore Pb-free solders, anticipated to reduce toxic environmental impact, have and will come into extensive use (Itsubo et al., 2004a). Although the Pb-free solders are not entirely new materials, they have not yet been extensively used. An analysis of the environmental impacts of Pb-free solders as compared to Sn-Pb solders would help in avoiding worse alternatives. The most likely high-volume Pb-free alloys have similar physical properties to Sn95.5Ag3.8Cu0.7 (LF), however, several Pb-free material combinations using Sn in combination with Bi, Cu, Co, In, Sb, Zn, have previously been compared to each other and to Sn63Pb37 (TL), e.g. using environmental life-cycle assessment (LCA). Schoenung et al. (2005) presented an excellent review of the current scientific status of more or less the whole complex problem related to the industrial ecology of Pb-free electronics. For LCA, they conclude that previously performed case studies lack transparency and therefore could not be easily validated (Schoenung et al., 2005). Nevertheless, several authors indicate that the Pb-free shift will be disadvantageous from a global warming point of view, but advantageous from a toxic emissions point of view (Verhoef et al., 2004; Itsubo et al., 2004a; Ekvall and Andræ 2006; Geibig and Socolof, 2005; Andrae and Liu, 2006). However, the resource issues, for example the possible scarcity of metals replacing Pb, could also be important for certain metals, as in 2003 around 120 Gg of electronic solders were produced of which approximately 12 Gg were used to produce solder paste.

The global environmental impacts, and a tradeoff between them, can be quantified using methodologies such as the Life Cycle Impact Assessment Method based on Endpoint Modelling (LIME) (Itsubo et al., 2004b). Further, to the author's knowledge, no entire electronic product has so far been evaluated, using LCA, when comparing Pb and Pb-free solders, but that can be explained because most of the Pb that will be replaced in electronics originates from solders. The hypothesis for the present research is that there will be negative environmental impact assessment results globally, as a consequence of introducing LF paste to replace TL paste. The problems addressed in order to attempt to falsify this hypothesis are:

H1. What are the attributional LIME scores, expressed in Japanese Yen, for an LCA comparison between LF and TL?

H2. What is the consequential change in LIME score when a global change between TL and LF is made?

Environmental life-cycle assessment of solders in electronics

The total global environmental load and impact on the biosphere, troposphere, and stratosphere is mainly the result of the industrial metabolism connected to product systems in which resources continuously are converted into useful products and services demanded by human societies. Environmental LCA is most commonly used by universities and companies as a method to evaluate the mass balance of inputs and outputs of specific product systems and to organise and convert those inputs and outputs into environmental themes or categories relative to resource use, human health, and ecotoxicity explained by Rebitzer et al. (2004) and Pennington et al. (2004). In Figure 1 the phases of LCA and their inter-relationships are schematically shown.

In this context, an example of a LCA goal could be to compare the life cycle impact of a mobile phone using Pb and one without Pb. The scope, which means what is intended to be included with the system boundary, could be just the solder life-cycle or also include the life cycles of all those parts of the phone using Pb, e.g. printed wiring board finishes, termination finishes, ball grid array interconnections, and internal chip-to-substrate interconnects for controlled collapse chip technology (Garner et al., 2000). The functional unit, the basis of the calculation, must be chosen and must reflect the function of the life-cycle. In this case, it could be “the average use of one mobile phone during three years”. The scope decides which unit processes to quantify in the inventory analysis within the system boundaries, of which examples are shown in Figure 2. In LCA a unit process is defined as the smallest system for which data is collected (Andræ et al., 2005). One variant of LCA is the so-called attributional LCA (ALCA), sometimes also referred to as retrospective or accounting LCA. The ALCA aims to specify how much of the global environmental load within the system boundaries belongs to a certain human activity. ALCA commonly uses average data in contrast to specific data, and is used to compare two or more alternatives and also to find the most environmentally relevant unit processes. Another variant of LCA is the so-called Consequential LCA (CLCA) where the consequences of decisions, such as phasing out Pb from electronics, are evaluated. These consequences can also relate to activities outside of the system boundaries, should these activities be affected (Ekvall and Andræ 2006). In CLCA, the change in the global environmental load as a result of adding or removing a specific human activity is studied. The CLCA make use of marginal data as it is the marginal producers and consumers that are affected by a small change. One of the practical problems with the CLCA is how to identify who these marginal actors will be. An attempt was made to model the Pb and Pb scrap markets (Ekvall and Andræ 2006). The inventory flows (obtained in the data collection step as emissions, resource consumptions, and waste amounts) from either an ALCA or CLCA, are classified according to which possible environmental impacts, e.g. global warming, they could cause. Anthropogenic and potential environmental impacts, which may be global, regional, local or a combination of these, include global warming, ozonelayer depletion, photo-chemical oxidant creation, acidification, local air pollution, human toxicity, ecotoxicity, eutrophication, and resource consumption. For example, the CFC's are classified as being able to contribute both to global warming and to ozone layer depletion. After the classification the flows are characterised according to their relative importance for each environmental impact indicator. The present research will focus on the integration of the environmental impact indicators for resource depletion, global warming, and ozonelayer depletion as these environmental effects are global. In Figure 2 the scope of the present attributional solder paste life-cycle model, as currently used by most LCA practitioners, is shown. The use of the electronic product is outside the system boundary.

Figure 3 shows the scope of the present consequential model from the perspective of the global shift to LF from TL.

Methodology for impact assessment – LIME

Based on a finished inventory analysis where all data sources are given (Ekvall and Andræ 2006), the method “Life-cycle Impact assessment Method based on Endpoint modelling, LIME” was applied (Itsubo et al., 2004b). The origin of this methodology is a study conducted by the LCA National Project of Japan aiming at the development of a Japanese version of a damage-oriented impact assessment method. In LIME, the potential damage is measured for four safeguard objects: human health, the utilisation of non-renewable resources (social assets), the increase of extinction risk (biodiversity), and the loss of primary production caused by mining of resources (primary productivity) are individually measured. Modelling socio-economic impact was based on the concept of user-cost, which accounts for the equity of future generations. Interviews were performed in Japan where a statistically representative population answered questions on how they valued different types of damage to the environment. This approach made it possible to make a monetary weighting between, e.g. human health and biodiversity. The Yen scores in this context are what the Japanese society is willing to pay to avoid a unit of damage, caused by the environmental loadings, to the safeguard objects shown in Figure 4. The present research made use of weighted LIME factors expressed in Yen/kg enabling a comparison and integration of the damage derived from different impact categories such as global warming and ozonelayer depletion. In Figure 4 the LIME concept is shown with a focus on global impacts. Pb emissions are considered a local impact, as opposed to compounds belonging to global impact categories, and are therefore not included within the scope.

For LIME, the resource consumption indices are originally based on the resource characterisation as developed by the French company Conception Development Durable Environment CODDE (2007) for a Raw Material Depletion (RMD) Indicator. Equation (1) explains how the indicator is calculated: Equation 1 RMD is the total characterisation factor; R _i the available reserve base of a resource (i) such as a high-grade metal ore, coal, etc. which realistically can be extracted; Y _i the number of years left of resource (i) considering the rate at which it is currently depleted; M _i the production of resource (i); and I _i the inventory flow of resource (i). The inventory flow is the amount of, e.g. silver needed to produce the functional unit. Table I shows the figures used to calculate the RMD values. At this stage the recycling of metals is not included, but would presumably delay the decrease of Y _i.

Owing to lack of comparable cradle-to-gate resource data for Ag, Pb, Cu and Sn production, it was assumed that the metal resource consumption was only the produced metal. For example, for 1 kg of Ag produced, 1 kg of Ag metal resources was assumed as input, whereas the extraction and processing energy inputs were taken from the literature. For Ag a mixed model assuming Ag to be a by-product mainly of Canadian Zn/Pb and Australian Au production was used (Teck Cominco Metal Ltd, 2004; Stewart and Petrie, 2006; New Boliden, 2005), for Pb a model assuming Pb to be a product of primary Pb production (Althaus and Classen, 2005), for Cu a model assuming Cu to be a by-product of Ni production (Althaus and Classen, 2005), and for Sn a model assuming Sn to be a product of primary Sn production (Althaus and Classen, 2005). However, the most relevant LCI data would have been a global average for ALCA and the marginal for CLCA. Further, for the ALCAs, the outflow of printed board assemblies from the preparation process has a positive economic value. Therefore, it was not regarded as waste but a raw material inflow to the life-cycles where the recycled metals are used. Thus, for ALCA, the metal recycling processes belong to subsequent life-cycles. Owing to lack of economic and marginal data, e.g. for the Sn market, the inventory result for the consequential LF study was assumed to be identical to the one for the attributional LF study. All LIME results and others presented below are expressed per 0.53 cm³ solder paste (corresponds to 2.5 g TL having a density of 4.7 g/cm³, 90 wt% metal alloy and 10 wt% flux) applied to a printed board assembly before the reflow soldering process. The quantity 2.5 g was chosen based on the USEPA study of a typical printed circuit board assembly (Geibig and Socolof, 2005).

Inventory analysis results

Three LCI's covering 416 substance flows were investigated. Totally 97 resources and seven gases were relevant for global impacts. In Table II the results for nine selected flows are displayed and they were selected based on an initial LIME screening of the product systems. Pb is included though it is presently not quantified for global impact categories. The column farthest to the right represents the consequential inventory result of the shift.

Impact assessment results

The results were obtained by multiplying the obtained inventory data by the corresponding weighting factors (Itsubo et al., 2004b). The characterisation indices and LIME factors used in the present study are shown in Table III.

Overall, 98 mass% of the resource input flows, 100 mass% of the greenhouse gases, and 100 mass% of the ozone depleting gases, had a corresponding LIME index. The total LIME scores for the CLCA comparison between TL and LF were around 1.8 and 3.4¥, respectively. The total LIME results for the ALCA were also near 1.8 and 3.4¥, respectively. In Table IV it is shown that Sn and Ag resources from Sn and Ag production, respectively, were the hot spots in this study.

Figure 5 shows the LIME result where resource consumption, global warming and ozonelayer depletion are weighted into a single index.

The consequence is that the total LIME score will increase by 1.6¥ or by 90 per cent.

Sensitivity check

A life-cycle impact assessment method which can be compared to LIME is the Eco-indicator ‘99 (Eco-i. 99) method as it is based on the disability adjusted life year (DALY), concept (Abou-Zahr, 1999). The DALY concept combines in one score the time lived with disability and the time lost due to premature mortality. As shown by Table III Eco-i. 99 has not reported a weighted index for damage to resources caused by extraction of Ag and some other resources. The results shown in Figure 6 were obtained by multiplying the obtained inventory data by the corresponding weighted damage factors. The major similarity is that consumption of Sn resources is one of the dominating flows.

Completeness check

The robustness of the results in Figure 5 are hampered by at least five factors:

not all emissions to air and the resources, which theoretically could have globally related LIME indices, were reported by Itsubo et al. (2004b), e.g. carbon monoxide (Holloway et al., 2000);
too low resolution of the inventory result, which means more flows of for example CFC's could possibly be “hidden” inside the system boundaries;
poor precision in the numerical values of emissions and resources, which means different kinds of temporal and measurement uncertainties;
poor precision of the LIME indices, which means different kinds of temporal, spatial, and geographical uncertainties; and
poor representation of unit processes, which means that a greater number of factories in reality represent the average presently used.

However, considering what turned out to be the dominating issues of the study, Ag and Sn resources, the global perspective of the study, as well as the relatively small product systems, it is unlikely the results would change drastically should the effects of these five factors be greatly improved.

Discussion and interpretation

The CLCA and the ALCA results both indicate that the overall LIME scores in Yen will increase considerably per functional unit. The most important differences between TL and LF are two-fold: Sn and Ag resource consumption. The Ag resources from Ag production are significant due to the relatively high LIME factor for Ag, about 6,900¥/kg. CO₂ emissions from electricity production are also noticeable. However, considering global warming results alone, it was earlier reported for a similar inventory to the present study that the solder application process and the Sn production were the processes mainly affected (Ekvall and Andræ 2006). The marginal Pb usage, as a result of the ban on the use of Pb in solder, will be where the competition is the largest and where the Pb consumers are most sensitive to a Pb price change. On the margin, in remote areas, diesel combustion to generate electricity is expected to be replaced by Pb back-up batteries (the marginal Pb usage) when Pb is banned in solders. All impacts attributable to Pb production will not disappear as, on the margin, Pb will be used in Pb-acid batteries (instead of solder pastes) which in combination with photovoltaic cells will replace diesel combustion for electricity generation. In fact, CLCA helped identify this offset in the impact related to the Pb production. Figure 6 shows the relative importance for different processes as evaluated by LIME compared to Eco-i. 99 and the three environmental impact categories on which they are based. Considering RMD results alone, they show a big resemblance to the overall LIME results. The present CLCA study predicts that the LIME score will rise as the avoidance of the environmental impacts connected to diesel combustion does not outweigh the increased resource consumptions and emissions from Ag, Sn, and energy production.

Further, it is likely that the Eco-i. 99 H, A weighted factor for Ag resources would be relatively high, and obviously would indicate more similarities between overall and individual Eco-i. 99 and LIME scores.

Moreover, electricity production could be of higher importance, although not likely to change any conclusions. The data quality is poor for the model of the scrap Pb market and waste management of competing sources of scrap Pb. However, this part of the model does not have a big influence on the results.

Other analyses of the shift to Pb-free solders

USEPA performed an attributional cradle-to-grave LCA comparison between Pb solders and Pb-free solders (Geibig and Socolof, 2005). The present study cannot be easily compared to the one by USEPA, among other things due to the lack of inventory transparency. Nevertheless, two similar solder pastes to the present ones were evaluated by the USEPA. These were Sn63Pb37 (SnPb) and Sn95.5Ag3.9Cu0.6 (SAC). The results from the environmental impact categories “Nonrenewable resource use (NRR)” “global warming (GW)” and ”Ozone depletion (OD)” used by USEPA were used for a comparison with the present LIME analysis. The USEPA functional unit was 1,000 cm³ of solder metal alloy applied before the reflow oven, compared to the present 0.53 cm³ of solder paste. Figure 7 shows the comparative results for the two studies expressed per USEPA functional unit, where 9,300 g of TL and 8,170 g of LF correspond to that measure. The units are kg resources for NRR, kg CO₂-equivalents for GW, and kg × 10⁶ CFC-11-equivalents for ODP.

The NRR difference is mainly due to the “inert rock” resource consumption used in the USEPA model of electricity generation. The relatively low ODP value for TL is due to effects identified by the use of CLCA, where the ODP from alternate Pb usage (battery production) offsets ODP connected to electricity and Sn production. When the top contributing inventory flows for NRR, GW and OD for SnPb and SAC were multiplied by the respective LIME single index, the Zn-Pb-Cu resource consumption and CO₂ emissions related to electricity generation for solder application were the dominating flows for SAC and SnPb, respectively. This comparison shows that USEPA did not characterise the resource consumption, but just reported the magnitude and top contributors. For example, “inert rock” having a moderate LIME index, dominated the NRR which merely reports the amount of resources used. Should another type of characterisation by for example RMD have been performed, other flows would have appeared as important. Furthermore, the analysis of the USEPA study confirmed that stratospheric ozone depletion is of minor importance for the LIME score.

Verhoef et al. (2004) and Reuter and Verhoef (2004) showed that dynamic modelling, as a life-cycle inventory of the total upstream system, could be useful for assessing the environmental aspects of the manufacturing of solders. The system boundaries included the ore processing, metal production and solder production activities for solder alloys including Sn60Pb40 and several Pb-free combinations. No inventory data were reported, but it was evaluated using the Eco-i. 99 method. However, the question was raised, whether a ban on use of Pb will lead to an environmental benefit, as the manufacturing of solders could be considered a global open loop recycling and production system. Verhoef argues that the governmental decisions to introduce Pb-free will push the system out of the existing steady state globally, and while production of Pb, Cu and Sn would not be affected significantly, local changes could be important. Nevertheless, the resource depletion dominated Eco-i. 99 scores were higher for SnAgCu than for Sn-Pb, but became smaller as the authors changed the weighting of resource depletion to 5 per cent as compared to the initial settings of Eco-i. 99. The findings indicated that preventing human toxicity will instead lead to resource damage. An important concluding remark is that there exists a limited production infrastructure for co-products in Pb ore processing. In case of an extended ban on lead, both the availability and recovery of a range of metals will be affected (Verhoef et al., 2004; Reuter and Verhoef, 2004).

Effect of recycling

It is uncertain how much of the Sn, Ag, Pb, and Cu solder metals that will be recycled will specifically be used for production of new solders. Quantification of this is especially important for Ag and Sn. Based on this research increased recycling of Ag and Sn could be significant in decreasing the global impact. In the present research it was not possible to apply the LIME factors for other than global effects, as the local LIME is adjusted for Japanese conditions. This screening of the global situation was however useful as it strengthens earlier results showing that the social and economic impacts, due to the consumption of resources for Sn and Ag production, could rise as a result of the shift to Pb-free solder paste (Itsubo et al., 2004c, p. 441 Figure 2(b)-(c)).

Consequential LCA including Ag and Sn

An important discussion is whether a consequential model for LF would lead to different conclusions. How much will global Sn and Ag usage rise as far as electronics solder is concerned? Using Deubzer's replacement scenario from the year 2003 to 2006 (Deubzer, 2007) the rise could be from about 68 Gg before the shift (29 mass% of the global Sn consumption) to about 113 Gg (35 mass%) for Sn after the shift, and from 0.075 Gg (0.2 mass%) to 3.6 Gg (12 mass%) for Ag (it is uncertain exactly how much electronic solder is globally used, but it was about 120,000 tonnes in 2003 and probably more than 10 per cent more in 2006). It mainly depends on the primary production, recycling, pricing and electronics market consumption of these metals, where economies such as that of China Govern more and more the world market trends for Sn. The shift to Pb-free solder could lead to a decreased use of Ag and Sn in a mix of other products. Which marginal Sn consumers, having the possibility to substitute Sn, are most sensitive to a change in Sn price? Which marginal Ag consumers, having the possibility to substitute Ag, are most sensitive to a change in Ag price? A change in the Sn and Ag prices will affect the uses of these metals. For some products the Sn or Ag cost is a small part of the total production cost, while for others Sn or Ag have important functional advantages, making the demand for Sn and Ag less sensitive to changes in the Sn or Ag price. If the marginal products using Sn cannot replace Sn in the long run, the total Sn production will go up. This is under the condition that all other major users of Sn, for example producers of Sn-coated cans, also use the same amount of Sn after the shift to Pb-free solder. However, if the Sn price increases too much, materials such as Al, glass, paper, plastic and Sn-free steel can substitute for Sn in, for example, pet food cans. On the other hand, Sn can also find new markets for example as alloys in automotive balance weights.

As for annual Ag consumption, industrial and decorative uses, photography and jewelry and silverware represent more than 95 percent, and the electronics and photography industries are the main consumers (Lanzano et al., 2006). The unique properties of Ag restrict its substitution in most applications. Further, long-term analyses of the Sn and Ag markets are required to forecast what can happen.

Comparing the ALCA results for TL and LF is not the same as predicting the consequences of shifting from TL to LF. Occasionally, as seems to be the case regarding the solders in the present study, the two techniques, ALCA and CLCA, more or less provide the same conclusion. It is obvious that ALCA is too rudimentary for estimating future environmental impacts, but on the other hand CLCA needs to be more developed in co-operation with econometricians to be more accurate and comprehensive.

Conclusions

The following conclusions can be drawn based on the research carried out in the present study:

as far as globally related environmental impacts are concerned, the shift from TL to LF solder paste is likely to increase them;
a significant increase was detected due to increased Ag and Sn production;
no significant increase in the LIME score could be related to the increased generation of electricity for the LF solder application processes as compared to TL;
the LIME score is highly dependent on the Ag and Sn weighting factors;
ozone depleting substances had an insignificant influence on the LIME scores; and
this study confirms earlier work reporting that the resource consumption will be higher for SnAgCu solder pastes than conventional SnPb.

For LIME it has earlier been shown, in a scenario for Japan, that the toxicity of Pb was most important leading to the overall environmental superiority of Pb-free solders. The next step would be to make a trade-off between the global impact of the locally related impacts, such as those originating from the use of human-toxic chemicals, and the known global impacts. For example, the existence of industrial Pb aerosols could make way for updated LCA indices (Rankin et al., 2005).

Globally focused CLCA's of solders also need to be improved in a number of areas:

the design of quantitative partial economic equilibrium models for the electricity, Sn and Ag markets and the corresponding scrap markets for the metals;
identification of marginal consumers for Sn, Ag and other solder metals;
marginal data for primary metal production need to be collected; and
system expansion including Au and Ag containing printed wiring board surface finishes could be worthwhile.

Equation 1

Anders S.G. Andrae

Norihiro Itsubo

Atsushi Inaba

Figure 1The phases in an LCA

Figure 2The scope of the attributional model for solder pastes

Figure 3The scope of the consequential model of the global shift

Figure 4Conceptual figure of LIME

Figure 5The consequential LIME results expressed per functional unit obtained when subtracting the CLCA TL from the ALCA (= CLCA) LF

Figure 6The consequential LIME results compared to Eco-i. 99 and environmental impact categories GWP20

Figure 7Comparative results for the USEPA solder LCA study and the present regarding three different environmental impacts

Table IThe basis for RMD indices

Table IISelection of LCI results expressed per functional unit of TL and LF

Table IIIIndices used

Table IVTop contributors to LIME single index given per solder paste type

References

Abou-Zahr, C. (1999), "Disability adjusted life years (DALYS) and reproductive health: a critical analysis", Reproductive Health Matters, Vol. 7 No.14, pp.118-29.