Here we give the full abstract and summary of the ASME report on chlorine input and dioxin output from different kinds of - mainly full scale - incinerators. This is the latest worldwide scientific evidence about the (absence of a) link between chlorine and dioxin formation in incinerators.

Parts of this page


CRTD-Vol. 36

Prepared by

H. Gregor Rigo
Rigo & Rigo Associates, Inc.

A. John Chandler
A. J. Chandler & Associates, Ltd.

W. Steven Lanier
Energy & Environmental Research Corp.

Under the direction of Subcommittee on the

Relationship Between Chlorine in Waste Streams and Dioxin Emissions From Combustors of the ASME Research Committee on Industrial and Municipal Wastes

Reviewed and approved by ASME Research Committee on Industrial and Municipal Wastes

The American Society of Mechanical Engineers United Engineering Center
345 East 4th Street / New York, NY 10017



This study investigated the impact that waste feed chlorine content and waste stream components, including PVC plastics and salts, have on PCDD/F concentrations in the flue gases at waste combustion facilities. The effort was not intended to develop emission factors, evaluate control system performance or assess PCDD/F relationships in liquid effluents or solid residues. The data were mostly obtained from full-scale tests. Seminal laboratory studies were used to suggest the form relationships might take and help identify significant confounding factors which must be taken into account.

Over 1,900 Municipal Waste Combustor [MWC], Hazardous Waste Incinerator [HWI], Medical Waste Incinerator [MWI], Hazardous Waste Fired Boiler [HWB], Cement Kiln [CK], Biomass Combustor [BMC], and Laboratory- Bench- and Pilot-Scale Combustor [LBP] test results from 169 facilities, many with multiple units, were analyzed to determine whether there is a relationship between stack gas PCDD/F concentrations and chlorine input or uncontrolled gas phase chlorine concentration as a surrogate. Chlorine feed concentration ranged from less than 0.1 percent for many BMC up to 80 percent for some HWI. The chlorine feed concentration was between these two extremes for most of the other sources and generally spanned an 8:1 range in each category.

The data were carefully checked to identify errors and statistical outliers, standardized to common reference conditions and analyzed to determine if there were any changes in either the signature (composition) or quantity of PCDD/F concentrations. Changes that were greater than measurement method imprecision were studied to determine whether they were related to differences in waste stream or flue gas chlorine content.

Detailed analysis of the signatures was done using Cluster Analysis. Signatures for data collected at the same sampling point within a combustor family (i.e., MWC, HWI, etc.) were all the same unless the run produced a significant number of below detection limit values. Some facilities displayed signature differences exceeding PCDD/F measurement method imprecision between sampling locations. No observable changes in PCDD/F profile were associated with differences in chlorine content.

A general review, using a combination of Canonical Correlation Analysis and simple Linear Regression, found no statistically significant relationship between chlorine input and PCDD/F stack gas concentrations for the majority (80 percent) of the 90 facilities which had sufficient simultaneous data to detect a statistically significant trend. Eleven percent displayed an increase; 9 percent of the facilities displayed decreasing PCDD/F concentrations with increasing chlorine.

Analysis of Variance [ANOVA] and combustion engineering models were combined using Multivariant Linear and Autoregressive Regression to analyze three major controlled experiments at Municipal Waste Combustors during which PVC, mixed plastics or salt was spiked into the feed. No statistically significant relationship was found between PCDD/F concentrations and chlorine in two of the experiments. In the third, a relationship was found when the PVC was sufficient to raise the chlorine content to about 10 times that normally found in MWC's. This data set displays severe autocorrelation (run-to-run memory effects), so the observed effect may be due to something other than the spiking.

Similar techniques were employed to analyze the results of the major parametric MWC studies conducted under the auspices of Environment Canada, Environmental Protection Agency, New York State Energy Research and Development Authority and the Great Lakes Regional Council of Governors. No statistically significant relationship between PCDD/F concentrations in the products of combustion and chlorine input was found in the data from these studies.

The same mixed picture is found for MWI's with 15 percent showing an increase and 7 percent showing a decrease in PCDD/F concentrations in the stack gases with increasing chlorine feed.

HWI's taken as a whole show no observable relationship between PCDD/F and chlorine concentrations in the waste being burned. Examination of individual facilities reveals that some have been tested over a 10:1 range of chlorine contents, up to almost 80 percent chlorine, and display no discernible relationship. Eighteen percent of the HWI facilities, however, showed an increase, while 18 percent showed a decrease with increasing chlorine feed.

Cement Kilns displayed no apparent stack gas PCDD/F concentration increase with increasing chlorine feed rate. This is not surprising, since chlorine is frequently added to remove natural alkalis and produce Portland cement meeting federal and state construction standards.

The waste fired boilers showed decreasing PCDD/F concentrations with increasing chlorine, but boiler design and cofiring fuel differences confound this finding.

Biomass combustors produce PCDD/F, but there is too little simultaneous PCDD/F and chlorine data to determine if there is a general trend. One test replacing salt laden wood with higher chlorine sludge displayed reduced PCDD/F concentrations in the stack gases.

While some laboratory experiments show that there is a functional relationship between chlorine input and PCDD/F concentrations in the products of combustion under certain conditions, the effect is much smaller than the effect of confounders like combustor design, operating practices and the normal variability found in emission measurements made at commercial scale systems.

Whatever effect chlorine has on stack gas PCDD/F concentrations from waste combustors is masked by these other variables. Discernible changes and consistent improvements in PCDD/F waste combustor stack concentrations are unlikely to be realized by reducing waste chlorine content.




The hypothesis that fuel chlorine content and combustor flue gas PCDD/F concentrations are related was not confirmed by the data analyzed in this study.

Of the more than 1900 PCDD/F test results collected at 169 facilities in the database, PCDD/F and chlorine are simultaneously characterized at 107 units in 90 facilities. Seventy-two facilities (80%) showed no statistically significant relationship between chlorine input and PCDD/F measured in the gas streams. For the data sets with an apparent relationship, ten displayed increasing PCDD/F concentrations with increasing chlorine, while eight demonstrated a decrease.

The failure to find simultaneous increases in most cases and finding a few inverse relationships, indicates that whatever effect waste feed chlorine has on PCDD/F concentrations in combustor flue gases, it is smaller than the influence of other causative factors. Any effect chlorine has on PCDD/F concentrations in commercial scale systems is masked by the effect of air pollution control system [APCS] temperature, ash chemistry, combustion conditions, measurement imprecision, and localized flow stratification.

The hypothesis that the amount or type of chlorine in the waste fed to combustion units is directly related to gaseous PCDD/F concentrations measured at the combustor outlet, part way through the air pollution control system or at the stack is not supported by the preponderance of the data examined by this study.



Combustion systems burning generally low chlorine fossil fuels usually exhibit low PCDD/F concentrations. In contrast, waste combustion systems burning hazardous or municipal wastes, which can have much higher chlorine contents, frequently exhibit higher PCDD/F concentrations. These observations have led some to hypothesize that PCDD/F concentrations are determined by fuel chlorine content. Some regulatory agencies, public interest groups and individuals are even calling for limits on the input of chlorinated compounds into combustors.

Prior to promulgating such regulations, it is prudent to examine the available data to determine if the hypothesized relationship is supported by field experience.



The purpose of this study was to use existing test data and information to investigate the effect that waste feed chlorine content - especially components like PVC, mixed plastics and salt - has on PCDD/F concentrations in the gaseous products of combustion at waste combustion facilities.

This effort was not intended to:

The study purposely focused on information from full-scale installations. Available data from municipal, medical, hazardous and biomass (agricultural and wood waste) fired combustors were assembled and analyzed. New test data was not developed as part of this effort.

Laboratory-, bench- and pilot-scale test data were reviewed, however, to assess their relevance to understanding the behavior of full-scale facilities. No attempt was made to include all the published laboratory-, bench- and pilot-scale test results and theoretical studies. Rather, seminal work, defined as frequently quoted original texts, was utilized. Only those phenomena that could be evaluated using available data were relevant to this effort. Consequently, limitations of existing studies (no ash catalyst chemistries (copper and iron), no mass balances, incomplete data, etc.) are inherent to this effort.

A complete listing of the standardized data is provided in Appendix C. Others are encouraged to use these data to perform additional analyses.



Data were assembled from in-house sources, original test reports, and summaries appearing in the open literature. PCDD/F, chlorine and relevant process information were entered into a database for Municipal Waste Combustors [MWC], Medical Waste Incinerators [MWI], Hazardous Waste Incinerators [HWI), Hazardous Waste Fired Boilers [HWB], Cement Kilns [CK], Biomass Combustors [BMC] and Laboratory-, Bench- and Pilot-scale Combustors [LBP].

All database entries were checked for transcription and data reduction errors and statistical outliers were flagged. The data were standardized to USEPA reference conditions (i.e., 20 °C, 1 atm, 7% 02, dry) and expressed in common units (ppmdv, mg/dsm3, ng/dsm3, etc.). PCDD/F concentrations were expressed as moles rather than nanograms (mass) for statistical analyses as required by chemical equilibrium and kinetics. For similar reasons, the logs of emission measurements and chlorine concentrations were used in evaluations to minimize the adverse statistical effects of data non-normality.

Two objective hypotheses were studied to address the central study question (i.e., what is the relationship between chlorine and gaseous PCDD/F concentrations from waste combustors?). They are:

Does the distribution of PCDD/F congeners and homologues (signature) change within a source category as the amount of chlorine present varies


Does the amount of PCDD/F in flue gases change as the chlorine content in feed varies?

If the answer to both of these testable questions is no, then there is no discernible relationship between stack gas PCDD/F concentrations and feed stream chlorine content.

The statistical analyses needed to objectively answer the above questions were first performed at the individual plant level and then for groups of like facilities. Confounding variables (i.e., sampling location, flue gas temperature, intentional experimental changes, salt versus organochloride spiking, design differences, etc.) were controlled to the extent possible in all analyses.

Cluster Analysis was used to analyze the PCDD/F signatures. Signatures were deemed different if the cluster coefficient was greater than that which the errors and uncertainty inherent in the PCDD/F sampling and analysis method explain at the 95% statistical confidence level.

Quantity change assessment began by graphing the data to identify visible trends. This was followed by a variety of methods. Canonical correlation, ordinary least squares and lag-one autocorrelation linear regression and Principal Components Analysis [PCA] were used to identify statistically significant relationships. Parameters like chlorine and temperature that might be contributing to observed concentration changes were evaluated using Analysis Of Variance (ANOVA]. Blocking variables that indicate the presence or absence of an effect (e.g., PVC plastics added or not) were used to account for incremental changes in operating parameters and configurations in the ANOVA. Continuous variables were used to account for continuous parameters like temperature and flue gas oxygen content. The statistical significance of each coefficient was used to determine whether the associated variable caused a measurable effect.



The data assembled for this study come from the major types of waste combustors. Input chlorine concentrations range from 0.1 % to over 76% (Figure 1) and having gaseous HCl concentrations measured before the APCS from 20 to nearly 8,000 ppmdv @ 7% 02 (Figure 2). This range of chlorine concentrations and the differing nature of the PCDD/F congener distributions make it inappropriate to analyze these data as a complete set. Rather, the data were analyzed by combustor type.

Municipal Waste Combustors [MWC]

Data from the 63 MWC facilities in the database were used in this portion of the study. Several of the facilities were subjected to parametric tests that focused on PCDD/F formation. A number of spiking tests added mixed plastics, PVC and salt to the normal feed to explicitly determine the relationship between PCDD/F concentrations and chlorine. Finally, many tests included some measurement of chlorine feed rate.

For spiking tests without measured uncontrolled HCl concentrations (i.e., HCl data before the acid gas control system), the variables representing step changes in feed composition were used to identify these intentional interventions in the analysis.

Comparison of the individual signatures obtained from each facility found that some test runs apparently differ from the bulk of the test runs by more than nominal method imprecision.

These differences are usually caused by different numbers of below detection limit [BDL] results. The remaining signature differences are associated with tests performed at different locations (i.e., boiler outlet, stack, etc.). After accounting for these differences, there is no indication that the normal range of operating conditions or wide variations in feed material, including spiking runs with elevated or depressed chlorine content, cause any measurable change in PCDD/F signatures. Also, little difference exists in the signatures between different facilities.

It can, therefore, be concluded that MWC facilities have similar signatures and that changes in chlorine content do not measurably affect the composition of PCDD/F in the flue gases at these facilities.

With respect to the quantity of PCDD/F's generated in MWC facilities, the parametric and spiking studies are particularly helpful in examining effects of varying chlorine feed conditions. Detailed statistical analyses were individually conducted on data from 18 facilities. It was concluded that there is no observable relationship between PCDD/F concentrations measured at a particular location and variations in chlorine normally observed at MWC's or from plastics, PVC or salt spiking. In the Hørsholm study where PVC, salt and lime were spiked, the data are strongly autocorrelated (i.e., the preceding run explains a large portion of the current run's response) and contain unexplained statistical outliers that are inconsistent with replicate run results. As such, no conclusions about the relationship between PVC or salt spiking and PCDD/F concentrations in the stack gases can really be drawn from this statistically designed experiment. Another full-scale study, Würzburg, found no difference between the PCDD/F concentrations normally observed in the stack cases at this state-of-the-art facility when burning MSW and those measured when the system was burning MSW augmented with 7.5 and 15% mixed plastics. A counter-example to a global relationship between PVC and stack gas PCDD/F concentrations is provided by the Pittsfield experiment. The combustion of PVC-free material produced the same type and amount of PCDD/F in the products of combustion at specific locations as did combustion of regular waste and PVC-spiked waste after controlling for other intentional operational differences. An overall analysis of the aggregated MWC database, controlled for confounding factors, also failed to find a statistically significant relationship between chlorine input and flue gas PCDD/F.

In summary, neither normally observed changes in chlorine content, nor intentionally induced increases and decreases in the chlorine content of waste feed via adjusting the mixed plastics content of MSW or spiking with PVC and salt, discernibly affect PCDD/F concentrations in MWC flue gas.

Since few chlorine induced changes in either stack gas PCDD/F composition or quantity were found, it is concluded that there is no consistent observable relationship between chlorine in the waste feed and MWC flue gas PCDD/F concentrations.


Medical Waste Incinerators [MWI]

The signatures used to compare composition were generally the same at individual sampling locations within each of the 31 MWI facilities in the database. The signatures before and after the air pollution control equipment were sometimes different; however, the changes were not related to chlorine content.

No observable relationship was found between chlorine content in the waste feed and PCDD/F concentrations in the flue gases at MWI's. When the data from all MWI's in the database are combined, there is a statistically significant negative relationship between chlorine content as represented by HCl measurements and the logarithm of molar PCDD/F concentrations; that is, if the relationship is true, increasing the chlorine fed to MWI's decreases PCDD/F concentrations. This negative relationship is probably a data artifact and not a real effect, since current PCDD/F formation theories do not provide a scientifically defensible explanation for this finding.

Based upon the above analyses, there is no statistically significant relationship between the composition or amount of PCDD/F concentrations in the gases emitted from MWI's and the amount of chlorine in the waste feed.


Hazardous Waste Combustors [HWC]

The complete dioxin signatures are essentially the same for all hazardous waste incinerators [HWI] regardless of their design, the air pollution control system employed, or the chlorine concentration in the feed. A few signatures however, are different. The signature differences observed are mostly attributable to different numbers of congeners with BDL values. There was no relationship between chlorine and signature arrangement in the clusters.

Scatter plots were used to display the relationship between total molar PCDD/F concentrations in the stack gas and percent chlorine in the feed. A variable relationship was found; 18 of 28 units with simultaneous PCDD/F and chlorine characterization information display no statistically significant relationship. Five facilities show an increase in PCDD/F concentrations with increased chlorine in the feed and five facilities show a decrease.

Hazardous waste fired boiler [HWB] signatures for the five units in the database were separated by plant. Since these boilers range from modified package boilers firing only hazardous waste to a pulverized coal-fired boiler with 10% liquid hazardous waste co-firing, the separation is probably the result of combustor design differences. A scatter plot of the logarithm of molar PCDD/F concentrations versus percent chlorine in the waste feed displays a statistically significant decrease in stack gas PCDD/F concentrations with increasing chlorine. Given the differences in boiler design, this is probably a data artifact since there is presently no PCDD/F formation theory to explain this behavior.

Cement Kilns [CK] displayed no relationship between the amount of chlorine fed per ton of clinker produced and PCDD/F concentrations. The signatures for different types of cement making processes are separated by a greater distance coefficient than method imprecision explains. Thus, kiln type (e.g., long wet, long dry, preheater, preheater/precalciner) makes a difference. The gaseous PCDD/F concentrations generally did not display a meaningful relationship with chlorine in waste feed. The one plant that showed any statistically significant effect displayed a reduction in PCDD/F concentrations with increasing chlorine feed. This is most likely a data artifact and not a real phenomenon.

HWI's generally show no relationship between stack gas PCDD/F concentrations and chlorine, but some individual installations display either an increase or decrease with increased chlorine in the feed. HWB's show a decrease in stack gas PCDD/F concentrations with increased chlorine which may be due to boiler design differences rather than being a real phenomenon. CK's display either no relationship or decreasing stack gas PCDD/F concentrations with increasing chlorine in waste feed.


Biomass Combustors [BMC]

The biomass combustor data in the database show that neither PCDD/F composition nor concentration are related to chlorine content in waste feed within the data for each of the seven systems located in five plants with simultaneous PCDD/F and chlorine characterization information. Even deliberately spiking with Pentachlorophenol [PCP] at Northwood, B.C. produced only two detectable PCDD/F homologues at the highest spiking levels. At Elk Falls, salt laden wood waste was augmented with even higher chlorine content pulp mill sludge; however, PCDD/F concentrations decreased. Data from the other 3 BMC's in the database show comparable gaseous PCDD/F concentrations; however, the chlorine content of the biomass fuel was not measured and simultaneous HCl measurements are not available, so conclusions regarding a potential relationship between PCDD/F concentrations and chlorine cannot be drawn from the balance of the tests.

BMC spiking tests indicate that adding chlorine to the waste feed does not affect the stack gas PCDD/F concentrations at individual plants; however, one facility showed a decrease when pulp mill sludge was substituted for lower chlorine content salt laden wood chip.

Laboratory, Bench and Pilot Combustors [LBP]

Laboratory-, bench-, and pilot-scale facilities produced some discernible trends between PCDD/F gas side concentrations and chlorine feed rate. Laboratory tube furnace tests have found that PCDD/F formation appears unrelated to HCl concentration, but Cl2 concentration causes increases once a threshold that is about five times the normally measured Cl2 in MWC's is exceeded. Bench-scale fluidized bed combustor results also indicate that there may be a threshold chlorine concentration needed to produce an effect. Strong evidence that PCDD/F formation is influenced by temperature has been uncovered. A pilot-scale MWC test found that chlorine effects may not be discernible until chlorine concentrations ran well outside the levels normally found in municipal solid waste.

LBP tests are useful for mechanistic studies, but do not necessarily represent the performance of full-scale facilities. LBP facilities cannot simultaneously match full-scale facilities' fluid dynamics, time-temperature histories and physical feed characteristics (i.e., fuel must be sized and frequently pelletized for use in these furnaces). LBP results must be cautiously used as guidance for interpreting full-scale installation test results. Full-scale facility confirmation tests are necessary to confirm that any observed phenomena are not artifacts of the LBP equipment.

While this effort has evaluated the relationship between PCDD/F concentrations in flue gas streams with feed chlorine concentration, it did not address the potential for a relationship between chlorine and PCDD/F concentrations in liquid effluents (i.e., residue quench waters, wet scrubber blowdown, etc.) or solid residues leaving facilities because adequate data was not available.



There is a wealth of information in Appendix C and in the main body of this report that can be used to address other PCDD/F formation and APCS performance questions. Others are encouraged to use this work as a starting point to address such questions.

If future research testing is conducted, it should be designed with sufficient stabilization time between different test conditions to minimize the between-run memory effects observed by several researchers. At a minimum, the same time period should elapse between changes in conditions and the start of experimental runs to facilitate at least rudimentary statistical treatment of this problem. A complete description of actual operating conditions should be included in all test reports so that the test remains useful as our understanding grows and new theories can be tested. Finally, simultaneously characterizing all solid, liquid and gaseous effluent streams will enable mass balances to help quantify experimental error and distinguish between PCDD/F destruction (inhibited formation) and media transfer.

American Society of Mechanical Engineers December, 1995


You are at level three of the Chlorophiles pages

Created: November 2, 1996.
Last update: February 28, 1999.

Welcome page

To the Home Page of the Chlorophiles

Up: Chlorine input and dioxin emissions

Right: Chemistry of dioxin formation

For any comment on the ASME report or on any of the Chlorophiles pages:

Mail to: