Environ. Sci. Technol. 2010, 44, 8750–8757
Life-Cycle Nitrogen Trifluoride Emissions from Photovoltaics VASILIS FTHENAKIS* Brookhaven National Laboratory and Columbia University, New York, New York, United States DANIEL O. CLARK, MEHRAN MOALEM, AND PHIL CHANDLER Applied Materials, Santa Clara, California, United States ROBERT G. RIDGEWAY, FORREST E. HULBERT, DAVID B. COOPER, AND PETER J. MAROULIS Air Products and Chemicals Inc., Allentown, Pennsylvania, United States
Received February 9, 2010. Revised manuscript received September 26, 2010. Accepted September 30, 2010.
Amorphous- and nanocrystalline-silicon thin-film photovoltaic modules are made in high-throughput manufacturing lines that necessitate quickly cleaning the reactor. Using NF3, a potent greenhouse gas, as the cleaning agent triggered concerns as recent reports reveal that the atmospheric concentrations of this gas have increased significantly. We quantified the lifecycle emissions of NF3 in photovoltaic (PV) manufacturing, on the basis of actual measurements at the facilities of a major producer of NF3 and of a manufacturer of PV end-use equipment. From these, we defined the best practices and technologies that are the most likely to keep worldwide atmospheric concentrations of NF3 at very low radiative forcing levels. For the average U.S. insolation and electricity-grid conditions, the greenhouse gas (GHG) emissions from manufacturing and using NF3 in current PV a-Si and tandem a-Si/ nc-Si facilities add 2 and 7 g CO2eq/kWh, which can be displaced within the first 1-4 months of the PV system life.
1. Introduction Chemical vapor deposition (CVD) operations or plasma deposition operations require cleaning the reactor between operations to maintain the purity of the deposited layers. Until the mid- to late-1990s, cleaning was accomplished either by off-line manual scrubbing or by dry-etching using a perfluorocarbon gas (PFC), mainly CF4 or C2F6. Excitation of the gas into plasma creates fluorine radicals that bond with the Si-based residue, converting it to SiF4 that can be vented. However, increased demands on the size and throughput of semiconductor fabrication facilities (fabs), coupled with the low dissociation of the PFCs drove the move to using nitrogen trifluoride, NF3. This gas has a higher dissociation rate than PFCs, supporting faster throughputs and lower emissions in manufacturing integrated circuits and thin film transistorbased displays (1, 2). Emissions were reduced further by employing remote plasma sources (RPS) that dissociate NF3 into fluorine radicals before they enter the chamber (3-8). This combination of lower emissions, better process per* Corresponding author e-mail:
[email protected]. 8750
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 44, NO. 22, 2010
formance, faster clean times, and reduced equipment maintenance led the semiconductor industry to adopt NF3 as their principal cleaning chemical in the last 15 years, as lately did the emerging a-Si/crystalline-Si thin-film photovoltaics industry. The indirect carbon emissions in the life-cycle of photovoltaics (PVs), due to using fossil fuels, are extremely small compared to those from fossil-fuel cycles (9). However, concerns arose about the PV industry’s direct use of potent greenhouse gases (10). A few researchers attempted to quantify the carbon footprint from employing fluorinated gases to clean the reactor during PV manufacturing (11-14). Thus, Wild-Scholten et al. (11) determined that the majority of NF3 emissions originate during manufacturing. In contrast, Schottler et al. (12, 13) concluded that the use phase, not production, may dominate the impact and highlighted the potential value of abatement downtime in significantly lowering the overall environmental impact of a fab (14). Ours is the first study of NF3 emissions in the PV life-cycle in which actual measurements of NF3 concentrations were collected from the effluent streams in major manufacturing and end-use facilities. Production and end use are the main two stages where emissions occur. We deemed that transportation was unlikely to contribute; thus, there are no reports of transportation accidents involving NF3 emissions, and a risk analysis of transportation using International Organization for Standardization (ISO) module packages for the PV end use showed a very small probability of leakage (Supporting Information, Appendix S1).
2. NF3 Concentrations in the Atmosphere Nitrogen trifluoride is a potent greenhouse gas, with a projected atmospheric lifetime of 550 years and an estimated global warming potential (GWP100) of 16 800 (15). Weiss et al. (16) demonstrated an increase in the atmospheric concentrations of NF3 from 0.02 ppt in 1978 to 0.454 ppt in 2008. Their latest measurements reveal a rate of increase of 0.053 ppt year-1 or about 11% per year, corresponding to about 620 t of global NF3 emissions annually. Estimating a production of 4000 tons per year, these authors forecast an emission rate of about 16%. However, they also noted a recent slow-down of the rise in concentration. To assess the relative impact of cumulative NF3 emissions and gauge the potential impact of future ones, we use radiative forcing (RF), attributable to the current atmospheric NF3 burden. Radiative forcing is defined as an externally imposed perturbation in the Earth’s energy balance, induced by changes in concentrations of greenhouse gases and aerosols, the Earth’s albedo, and solar energy (17). For long-lived greenhouse gases, such as NF3, we estimate the globally averaged change in RF from the following equation: ∆Fnew ) R(X - X0) Here, R is a gas-specific constant for radiative efficiency (W/ m2-ppbv), X is the atmospheric concentration in parts per billion by volume (ppbv), and X0 is the unperturbed concentration. For NF3, R is 0.21 W/m2-ppbv (17), yielding a net radiative forcing of 0.9 × 10-4 W/m2 based on the current global atmospheric concentration of 0.454 ppt and zero background concentrations (16). The total radiative forcing for all anthropogenic greenhouse-gas emissions from 1750 to 2005 reportedly is 2.63 ( 0.26 W/m2 (18), of which nitrogen trifluoride represents 0.003%. Compared to CO2 alone, the current NF3 radiative forcing from the atmospheric burden 10.1021/es100401y
2010 American Chemical Society
Published on Web 10/25/2010
FIGURE 1. Annual worldwide NF3 production, 1995-2008. of 0.454 ppt is 0.005% of the total anthropogenic CO2 RF of 1.66 ( 0.17 W/m2.
3. NF3 Emissions in Production and Distribution A few manufacturers worldwide generate NF3 via the fluorination of ammonia or the electrolysis of ammonia and HF. The estimated annual production capacity of the largest producer, Air Products, is over 3000 tons. Other major producers are Sodiff, Kanto Denka, Formosa Plastics, and Mitsui Chemicals. We estimated that the global production of NF3 in 2008 was about 7200 tons, that grew by an average of 41% per year since 1995 (Figure 1). Both the rate of NF3 production and its atmospheric concentrations rose over the past decade; however, the former increased annually by an average of 41% and the latter by 11%. Therefore, expressing atmospheric NF3 concentrations as a function of the global annual NF3 production shows a reduction in the NF3 emission factor (Figure 2). On the basis of the estimated global production and the atmospheric concentrations measured by Weiss et al., the emission factor in 2008 was 9%, i.e., below their estimate of 16 ( 4%. In the following, we highlight past and current practices in the various stages of the NF3 lifecycle to identify the underlying reasons for this fall in emission factors. 3.1. Past and Current Practices in NF3 Production. NF3 emissions occur during manufacturing from known or fugitive sources. The former are those from returned product containers and various vent streams operating during manufacturing (e.g., vents for the product-fill manifold, reactor start-up or shut-down, analyzers, and product purification). Fugitive emissions come from leaks occurring during manufacturing of the gas from pressurized manifolds, compressors, and during product fills. Depending upon the size of the customer’s fabrication facility, NF3 is distributed in cylinders or ISO containers. Air Products manufactures
NF3 at four facilities worldwide where it is filled into appropriate ISO containers for transportation to end users and to transfills (distribution centers). At their global transfill facility, the gas is transferred into smaller cylinders for shipment to customers. However, with the increase in customer consumption driven by larger fabrication facilities, the substantial majority of NF3 now is shipped to customers directly in ISO containers; cylinders are required only for a very small percentage of the product. Thus, the majority of product fill occurs at the manufacturing sites. Ambient NF3 monitoring is conducted at the plant using continuous gas monitors, such as those based on nondispersive infrared (NDIR) sensors, able to detect concentrations as low as a few ppmv. Residual product returned by customers is a potential source for fugitive emissions. Estimates suggest that most fabs leave about 5-10% of the total gas volume in the container to avoid contamination. Over the last 8 years, Air Products phased out their venting of product residuals. When containers are returned, their residuals are analyzed, and if purity specifications are met, the containers are topfilled with new product; otherwise, the residuals are recycled into the manufacturing process. 3.2. NF3 Emissions Monitoring Program. Air Products Inc. reported a continuing decline in NF3 emission factor at their production facilities in Hometown, PA, over the last 12 years. The estimated total emission factors were about 7% in 1997, 5% in 2004, and 2% in 2006-2009; they are targeted to decline to 0.5% (19). The Japanese manufacturer, Kanto Denka, also reported a fall from 3.8% in 2006 (when they started a major emissions-reduction program) to 3.1% in 2007, 2.5% in 2008, and 2% in 2009. This company aims for further reductions, likely toward Air Product’s 0.5% target (20). To assess the fraction of product emitted to the atmosphere during operations, Air Products conducted two VOL. 44, NO. 22, 2010 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
8751
FIGURE 2. Decline in NF3 emission factors over the 8 years, 2000-2008. These data were calculated from the values presented in Figure 1 and the measurements reported by Weiss et al. on NF3 atmospheric concentrations over the same period. independent calculations based on a combination of engineering data and analytical measurements. The first calculation was based on material balances and included process parameters such as process-flow rates, process efficiencies, process yield, and product produced and transferred into containers. Critical measurements are the mass flow of F2 to reactors, the efficiency of process conversion, and the quantity of NF3 transferred to containers. Fluorine flow is determined by Coriolis-based mass-flow meters, with the manufacturer’s stated precision of (0.05%. The scales used to weigh NF3 transferred to containers also are very precise ((0.5%). Process-conversion efficiencies are determined using process analytical metrology, such as gas chromatography and Fourier transform infrared (FTIR); their estimated precision is
