Source signature and HM’s Enrichment in PM2.5
Emission from Clinker Production
Paper # 05-A-579-AWMA
C. Gutierrez-Cañas, J.A. Legarreta, M. Larrion, J. R. Vega,
Escuela Superior de Ingenieros, U. País Vasco, A. de Urquijo s/n , 48013 Bilbao (Spain)
E. García and S. Astarloa
AIR g, M. Diaz de Haro 68, Bajo, Pab. 3-4-5, 48920 Portugalete (Spain)
ABSTRACT
Heavy metals from high-T stationary sources and PM2.5 are correlated. Data
representative of current designs and operating practices regarding emission of PM2.5
from clinker and cement production are very sparse, several studies cover these topics
for other stationary sources.
Although the emission of primary PM2.5 is believed to be very low, this range
concentrates several trace pollutants –regulated- and becomes more relevant according
to the improvement of gas-solid separation devices. The tendency of stricter emission
limits has not been encompassed by the experimental knowledge about fine fractions.
National emission limit values differ not only on designated metals but also in
measurement conditions, methods, time interval and frequency.
Open questions in currently available source profiles for stationary sources include: to
identify chemical species that differentiate among sources, to measure following nonstandard or non-equivalent methods, to know the source operation and dynamics; and to
report variability, inaccuracy and uncertainty of chemical abundance measurements.
This work present an experimental approach and initial field results for characterization
of HM and PM2.5 emissions from clinker production covering the entire dynamic (circa
24 h/cycle) of the process.
PROBLEM STATEMENT: HEAVY METALS AND PM2.5
The concentration of individual heavy metals has a strong dependence on size fraction.
Fractional efficiency criteria will substitute the conventional ones (based on
extrapolation, similarity, and on simple physical descriptions, such as the overall
efficiency-based models) for equipment and process design and rating. Nowadays, the
equipment for gas-solid separation is designed using some theoretical guidelines, but
really a number of relevant decisions are based to a great extent on empiricism.
However, the current needs to micron range efficient removal in industrial processes
require a deep insight on size distributions, and on the size-distributed chemical
composition.
According to this problem statement, the following particle physico-chemical properties
of particulate emissions were measured:




Particle mass size distribution
Particle number size distribution
Elemental composition
Morphology
The mass concentration and mass size distribution were measured, with cascade
impactors (Dekati PM10, Andersen Mark III and Kalman heads). The number
concentration and number size distribution was measured with ELPI, which also
separate sample to further analysis. The size distribution -aerodynamic diameter- of the
particulate matter collected in the electrostatic precipitator was measured by means of a
time-of-flight analyzer (TSI3603). Samples of the main streams were also collected for
elemental analysis. The relative abundance of the different metals depends on the fuel
and secondary raw materials as well as combustion conditions.
Current and short-term trends in emission limit values focus on trace pollutants, as is
the case of heavy metals. On the other hand, new air quality standards for PM2.5 will
drive new processes to identify emission sources, and to develop new techniques for
fine particle control. However, these new standards and cost-effective abatement
technology depend on an adequate understanding about the relationship between the
supposed adverse health effects and the physicochemical properties of the fine PM
governing or causing them.
Additionally, improved characterization of source profiles is required for reliable
PM2.5 source apportionment1,2 . The source signatures must take into account the
intrinsic variability of chemical abundances due to process dynamics or operational
changes; not only from discontinuous processes –such as electric arc furnace
steelmaking, for example- but also for continuous processes. This is the case of
clinker production.
Cement companies have improved traditionally the cost-effectiveness of the clinker
production by using secondary raw materials and low-quality combustibles. Recently,
the interest and role of wastes as fuels or secondary raw materials on cement industry
is rapidly growing. It has been estimated (Cem. British Assoc., 1998) a European
mean value of 11% for the fuel substitution rate, although large discrepancies among
regions are also reported. These discrepancies (ranging from a share of 2% in an EU
state to a 20% in other one, as reported by the same manufacturing group) are closely
related to the structure of the regional market, and local policies on permits. The same
fact is observed considering regional differences concerning the use and addition
proportion as secondary materials of a wide range on wastes from other industrial
processes, ranging from an incipient reuse to locations where it is almost
indispensable from a social, economical and technical point of view.
In Europe the IPPC (Integrated Pollution Prevention Control) 96/61/EC Directive lays
down procedures for granting operation permits for cement plants. It aims at achieving a
high level of protection for the environment as a whole by means of measures “designed
to prevent or, where that is not practicable, to reduce emissions” to air, water and land.
The use of hazardous waste as an alternative fuel in cement kilns is regulated at EU
level by Directive 94/67/EC. While the Directive sets out rules for the burning of
hazardous wastes in dedicated plants for incineration of waste, it also recognises and
provides for the procedure of co-incineration, that is the burning of wastes in industrial
furnaces (such as cement kilns) not exclusively designed for such purposes.
IPPC Directive will entry into force in 2007. The emission limit values will be set using
an integrated approach by the locally competent authority with due consideration of: the
technical characteristics of the plant, the geographical location; the local environmental
conditions and the best available technique (to be taken from the respective BAT
Reference Document). The environmental impact of the cement manufacturing process
is mainly restricted to atmospheric emissions thus not representing an environmental
“multi-media” impact process. Representative emission data (long term average values)
from European cement kilns in operation are summarised in the Table13.
Table 1 Long-term averaged emission values from clinker kilns*,**
*
Emission limits refer to reference conditions in the dry or wet gas (i.e. 0° C and 1013
mbar), sometimes adjusted to a specified oxygen level. Limit values also refer to
averaging periods which may be different from country to country (half-hourly, hourly,
daily, annual average). Specific conditions for short term non-compliance (exceedances)
also apply.
**
In Europe, cement clinker (about 78%) is predominantly burnt in dry-process plants in
rotary kilns; shaft kilns are seldom used.
The Directive 94/67/EC states emission limits for cement kilns as co-incineration plants
with special provisions. However, the tendency of emission limits, also as the tendency
of achievable emissions -corresponding to the current available technology and/or to the
best industrial practices, as reported almost fifteen years ago by the VDI (1985)- have
not been encompassed by the experimental knowledge about heavy metals mass
balance, more precisely about their unequal distribution among process outlets.
There is still a lack of European reference methods or standardised practices, which
could provide a common basis for the interpretation of several sampling, analysis, and
result reporting approaches. Moreover, there is also a lack of equivalence between
results obtained by different sampling trains, by different sample treatments and/or by
different analytical techniques. This uncertainty has been reflected also in regulatory
standards and directives; for example concerning the “measurement techniques” the
Directive on Incineration state that they “have to be carried out representatively”,
following CEN standards and, if not available, ISO standards or other national or
international procedures.
The central questions in terms of a industrial perspective are, therefore, whether the
clinker production process can be efficiently monitored, in order to ensure the
simultaneous compliance of regulatory requirements and product quality constraints. Insight on this point can be acquired neither by the use of conventional emission
factors, nor by extrapolation of data from other conversion processes, nor by
modelling-, whether the incentives for waste treatment are realistic, or coherent with
emission limits and product quality, whether cement industry contributes to a life
cycle of other products processes in an operation compatible with environmental
requirements, and whether the need for innovative design, operation and management
tools, incorporating sustainability criteria, could be implemented.
EXPERIMENTAL METHODOLOGY
Sampling and Analysis
The following paragraphs refer to a plant for 2200 t clinker/day, dry-process with long
rotary kiln, precalcinator and a four-stage cyclone preheater. The main raw materials are
limestone, marl, and as Fe-carrier normally wastes from steel rolling; and the routine
fuels are petroleum coke and used tyres.
The emission from the clinker production line could be briefly described as a high
temperature aerosol of complex chemical composition, long residence time and
complex gas-to-particle physicochemical interactions along the process line. In this
plant, the mean residence time of the gas in the kiln, precalciner, preheater and
electrofilter is around 40s with temperatures falling from 1100 to circa 130ºC at the
stack. An estimate for the material retention time is 35 min. The dust collected in the
electrostatic precipitator is recycled to the cyclone stages near the kiln. This closed-loop
lead to the build-up of characteristic metals –such as thallium- and it is maintained
stable as the temperature profile remain unchanged.
However, the operation is not strictly stable, because it could be broken down into three
successive periods from the point of view of the particle-laden gas flow sources: 1)
Clinker kiln, 2) Clinker kiln and crude mill, 3) Clinker kiln, crude mill and coke mill.
These subregimes must be individually described leading to a separate discussion about
emission characterization, compliance and gas cleaning requirements. In the case of
metals confined in the above-mentioned closed loops, it may give rise to unexpected
short-time emissions.
The instruments, techniques and procedures were as follows. The mass size distribution
was measured by means of cascade impactors (4 stage Dekati PM10 or Kalman heads)
as heads of a conventional sampling train –EPA5 or VDI 20664- under constant flow
rate. The operational problems, such as the possibility of particle bounce and
reentrainment, interstage wall losses, and the non-ideal collection characteristics of the
substrate, has been widely reported and specific procedures were applied to QA/QC..
Sampling periods must allow the collection of a sufficient sample in each stage –for
further chemical analysis- while avoiding any overloading. A continuous recording of
stack velocity was used as emission regime stability criteria.
The number size distribution was measured by an ELPI. In this case, due to the high
concentration the impactor must be used with a two-stage dilution train (Dekati
ejectors). The first one was maintained to about 180 ºC and the second one operate
without heating. No condensation was observed and the losses along the line –before
the first diluter- are considered negligible. The averaging period was set to 10 s, in this
way, the signal could be considered as nearly “real-time” size distribution measurement.
The ELPI was operated in stable conditions and also during transients or operational
changes. In the first case, each stage was analyzed. The sampling periods lasted
approximately from 1 to 2 hours in order to avoid impactor loading collecting sufficient
mass to perform chemical analysis. Gas flow through the ELPI was set around 10 lpm.
For the size ranges of interest, the use of a low-pressure cascade impactor allows to
work with fine particles because the lower cut-off aerodynamic diameter limit of the
conventional non compressible flow impactor is approximately 0.3 µm. The last stage
was maintained to 100 mbar absolute pressure.
In general, particle bounce can be a severe problem in high-velocity impactors and
volatile particles may also evaporate under the low- pressure conditions of the lowpressure impactor. Is this case, the presence of semivolatiles could be a big concern for
the analysis of Hg, taking into account that the solid-vapor behaviour depends strongly
on the species properties and can not be estimated on the basis of pure element
behavior.
The basic operation principle of the ELPI combines inertial classification of the low
pressure impactor with real-time electrical detection. It has been reported that for
gravimetric measurements fine particle losses to the upper stages are not critical, as the
mass introduced by diffusion deposition is insignificant when compared to the mass of
particles actually impacting to the stage. When using an electrical detection instead of a
gravimetric method, the current carried by the fine particles can be significant when
compared to the signal caused by particles impacting to the stage. Losses are on the
order of 0.1 to 2 % when the stage cut-off diameter is much larger than the
aerodynamic diameter of the particle5
Sample treatment and analysis of heavy metals are conducted under the VDI3868
(1999), VDZ recommended practices for mass flows characterisation in cement
production. For trace analysis, a comparison among AAS, ICP/AES, ICP/MS and FRX
was also carried out. The results show that available FRX was not adequate to this kind
of task, despite the advantages to the direct use of solid samples. Due to the few amount
of mass sample, the results here presented have been obtained by ICP/AES and
ICP/MS.
Mass size distribution
Figure 1 shows the ratios of the mass collected on each stage of the cascade impactor
using as reference basis the total mass collected in each run, for a series of eight
successive samples in a month. Stages are designated by its nominal cut-off diameter,
despite the fact that corrections must be made due to stack conditions (pressure and
temperature).
The observed variability of relative PM(i)/PM(i) has no or little sense without taking
into account the operational subregimes. The SEM micrographs and elemental analysis
of the collected samples shows that the coarse mode (larger than 5m) consists mainly
of fragments of the inorganic raw materials originated during the conminution of crude
feed. The fine mode is composed of agglomerates resulting from gas-to-solid
conversion along the process line (kiln and the four-stage cyclone preheater) and,
consequently, is very sensitive to process conditions and dynamics.
Its morphology6 shows fractal-like structures correspond to a rather slow growth
process enriched in the intermediate range condensing metals- and spherical structures
corresponding to rapid cooling and condensation of volatile metals –mainly thallium
was detected- after the last cyclone stage. In any case, it is to be pointed out that PM2.5
could range roughly from 40% to 80% mass.
Figure 1 Influence of operating regime: Specific ratios PM(i)/PM(i).
100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%
Sampling run
(PM2,5-PM10)/TSP
(PM1-PM2.5)/TSP
PM1/TSP
The operational changes may lead to a transient behaviour of the emission. The record
of emission evolution following a shutdown of the coke mill is represented in the Figure
2. The typical delay for a maximum value of about 150 mg/Nm3 is showed. Due to the
temperature increase, this transient is expected to be related with a short-time emission
of volatile metals, but this assumption is not yet confirmed because of the limited
sampling time and, hence, mass sample.
Figure 2 ELPI record after a coke mill shutdown
An example of the number size distribution is given in the Figures 3 and 4. In the Figure
3 are represented data from two situations: the emission is composed of gases coming
from the kiln and the crude mill, and the second one, coming from the kiln and the coke
mill. In the Figure 4 are represented two records of number size distribution obtained
along the same sampling period, illustrating the typical variations observed after and
electrostatic precipitator.
Figure 3 Number size distributions. Left: Subregime 1: Crude mill on. Coal mill off.
Right: Subregime 2: Crude mill off. Coal mill on
In full-scale particulate separation equipment, there is usually a penetration maximum
roughly in the same fine size fraction (0,1-1 µm) due to a decreased particle impaction
and weak electrical collection in that size range. It is also possible to find fine particle
emissions caused by the control devices themselves. Electrostatic precipitators are
reported to emit a ultrafine particles in the presence of other components of the gas,
such as SOx, due to particle formation inside corona region7.
Figure 4 Subregime 3: Crude mill off. Coal mill off
Size-distributed heavy metals
This presence of a submicron mode was described at first for combustion systems8,9 as
well as the trace element concentration in the fine fraction10 . There are available
valuable fields studies for combustion systems11,12. Data for other main stationary
sources are sparse13 and, in the case of clinker production they are not found6.
The literature to date on partitioning and transformations of heavy metals in solid fuel
conversion processes is quite extensive. Although there are valuable experimental
studies for several fuels14,15,16, in the absence of reliable monitoring methods for
industrial monitoring there are only sparse data about on-line systems.
In the Figure 5 are summarized the long-term averaged enrichment factors obtained17
for several feed patterns (secondary raw materials and alternative fuels). Enrichment
factors are calculated for each metal as the ratio of concentration in the stream to the
corresponding in the crude feed (rather low variability in medium-range time scale).
It must be emphasized that this results do not confirm the hypothesis that instead of the
direct measurement of chemical composition of the emitted dust, the chemical
composition of collected matter in the cleaning device could be used to follow the
plume impact. On the opposite, in this case the CKD and the emission shows a rather
dissimilar partitioning of metals or, in other terms, chemical fingerprint. The use of
CKD as a surrogate for the estimation of the chemical composition of emitted PM can
only be carefully used on the basis of a in-situ well-established correlation.
Figure 5 Mean enrichment factors
While the application of analytical techniques to collected aerosol material could be
considered as relatively advanced, on-line methods for the measurement of aerosol
chemical properties represent a serious gap in existing aerosol instrumentation18.
Consequently, the control of composition must be periodic and by extractive sampling
of a representative sample. A problem arise when the chemical composition of emitted
particles is inferred from historical data; that is, the extrapolation or the absence of
documentation on process streams and operating conditions.
The Figure 6 summarize the short- to medium-term variability ( a few months) of heavy
metals in respect to the historical data (blue bars). This is mainly due to substitution by
secondary raw materials and alternative fuels. The experimental design is discussed
elsewhere17.
Figure 6 Variability of enrichment factor for the emission of heavy metals in the shortand long-term
It was also observed a marked bias of all the designated heavy metals to the PM1, with
the exception of Sb. In the Figure 7 are represented data of heavy metal concentration –
in terms of the above defined enrichment factor-. Despite of the variability of the total
enrichment factors- see Figure 7- the size-distributed variability show normally the
same pattern.
Figure 7 Size-distributed heavy metals enrichment factors
The aerosol collected –and size-distributed- was partially solubilized with water and
further analyzed by ICP-AES (Na, Mg, Al, P, S, K, Ca and Fe) and ICP-MS (Li, Be, B,
Sc, Ti, V, Cr, Mn, Co, Ni, Cu, Zn, Ga, Ge, As, Se, Rb, Sr, Y, Zr, Nb, Mo, Cd, Sn, Sb,
Cs, Ba, lanthanides, Hf, Ta, W, Hg, Tl, Pb, Bi, Th and U)17. Volatile heavy metals (Se,
Bi, Pb, Tl and Cd, because Hg has not be considered in this analysis due to its high
partition coefficient) show a significant higher concentration in the fine fraction. The
higher concentrations correspond to the finer fractions in either the soluble and
insoluble fraction. Generally, the metals are present mostly in the insoluble fraction,
with the exception of Tl, probably combined as halogenide.
CONCLUSIONS
Decreasing total particle emissions does not necessarily decrease fine particle
emissions. There are no direct emission limit values for fine particles, but it is possible
in the future as a consequence of PM2.5 air quality standards. However, in fact, they
could be considered as implicitly regulated through emission limit values for pollutants
that are present mainly in this size range.
The size stratification of selected trace metals in the clinker dust aerosol, as in other
aerosols arising from high-T processes, implies that an efficient control of heavy metals
emissions must be based on fractional efficiency criteria. Cost-effective control
techniques for the fine particle emission require understanding the formation
mechanisms and aerosol dynamics, combustion efficiency and the influence of gas
composition.
The dependence of emissions upon process dynamics is not efficiently measured by
conventional on-line PM monitors. Evidence should be acquired by intensive sampling.
The use of cascade impactors as sample preseparators has been proved useful for
analytical purposes, despite the limitations to achieve a mass size function distribution
for complex particulate matter due to the differences of physical diameters on the same
aerodynamic stage.
Source signature is determined by feed and fuel nature and flows, by operating
practices, by particle control devices performance and finally by the effect of secondary
particles (from NOx, SOx, and alkalis). More data are needed and open questions about
currently available source profiles for stationary sources include: to identify the
chemical species that differentiate among sources, to identify the chemical species that
can reflect an adequate feed pattern –to be used as tracers and as basis for waste
materials acceptance criteria-, to overcome the lack of standardisation or equivalences
among methods and measuring devices; to reflect the source operation and dynamics
and to report the variability, inaccuracy and uncertainty of chemical abundance
measurements.
As open questions are to explain the correspondence of emission limits with health
effects, and to decide the metrics: mass versus other properties. Because of the complex
transformations occurring under diverse local conditions (meteorological, influence of
other sources, among others) is difficult to establish the actual contribution of fine
particle emitters. The performance assessment of the determination of the size-resolved
composition before aging and the dilution sampling trains require more experimental
data.
ACKNOWLEDGEMENTS
This work was done in a co-operation Project with Cementos Lemona S.A. Authors are
indebted to Mr. C. Urcelay for assistance. The research was partially supported by the
Ministerio de Ciencia y Tecnología, Madrid (Spain) and by Industri Saila, Basque
Government (Vitoria-Gasteiz)
REFERENCES
1 England G.E., Zielinska B., Loos K., Crane I. Ritter K., Fuel Proc. Technol., 2000,
65-66, 177-188
2 BAT Reference Document 2000-2003, Cembureau, Brussels (2004)
3 Chow J.C., Koutrakis P. (Eds.), Proceedings, PM2.5: A Fine Particle Standard,
Pittsburgh, PA: Air and Waste Management Assoc., 1998.
4 VDI 2066 Part 5. "Particle size selective measurement by impaction method -cascade
impactor". Verein Deutscher Ingenieure (1994)
5 Marjamäki, M., Virtanen A. and Keskinen J., J. Aerosol Sci. 1998 29 S441-442. 5th
International Aerosol Conference 1998, Edinburgh
6 Gutierrez-Cañas C., Larrion M., Garcia E., Parra Y., Rodriguez J. Guede E, J.
Aerosol Sci, S 637, 6th Internacional Aerosol Conference, Taipei, 2002
7 Mohr M., S. Ylatalo, N. Klippel, E.I. Kauppinen, O. Riccius, H. Burtscher, Aerosol
Sci. Technol. 1996, 24, 191–204.
8 M.W. McElroy, R.C. Carr, D.S. Ensor, G.R. Markowski, Science, 1981, 215, 13–19.
9 Markowski, G.R., Ensor, D.S. and Hooper, R.G., Environ. Sci. Technol. 1980, 14,11,
1400-1402.
10 Markowski, G.R. and Filby, R., Environ. Sci. Technol. 1859, 19, 796 - 804.
11 Kauppinen, E.I. and Pakkanen, T.P.Environ. Sci. Technol.1990, 24, 1811-1818.
12 Moisio, M., Laitinen A., Hautanen J., and Keskinen J.. J. Aerosol Sci., 5th
International Aerosol Conference 1998, Edinburgh
13 Ylatalo, J. Hautanen, Aerosol Sci. Technol., 1998, 29, 17–30.
14 Reed G.P., Dugwell D.,R. Kandiyoti, Energy Fuels, 2001, 15, 794-800
15 Miller B.B., Dugwell D.R., Kandiyoti R., Fuel, 2002, 81, 159-171
16 Lind T., Kauppinen E., Jokiniemi J., Maenhaut W., Proc. Combust. Inst, 1994, 25,
201-209
17 García E., Larrión M., Astarloa S., Chomón K., Gutierrez.Cañas C., J. Aerosol Sci.,
2004, S1183-S1184, Abstracts of the European Aerosol Conference 2004, Budapest.
18 Monkhouse P, Progress Energy Comb. Sci., 2002, 28, 331-381