- Email: [email protected]
transmission spectroscopy
Analysis of particle emittance, composition, size and temperature by FT4.r. emission/transmission spectroscopy* Peter R. Solomon, James R. Markham
Robert M. Carangelo, Philip and David G. Hamblen
E. Best,
Advanced Fuel Research, Inc., 87 Church Street, East Hartford, (Received 20 October 1986; revised 28 January 7987)
CT 06108,
USA
A new method was developed for online, in-situ monitoring of particle laden streams to determine chemical composition, concentration, particle size and temperature of the individual components. The technique uses an FT-i.r. spectrometer to perform both emission and transmission spectroscopy along a line of sight through the reacting medium. The technique has been applied to measure particle properties under a variety of circumstances and several examples are presented for measurements of temperature, emittance, composition and size. (Keywords: instrumental methods of analysis; FT-i.r.; particle properties)
Many chemical and energy conversion processes involve multi-phase feed or product streams. Streams can consist of mixtures of solids, liquids and gases and in some cases the important species may be in transition between phases (e.g. liquid fuel combustion which involves liquid fuel droplets, vaporized fuel and soot). Process monitoring requires the measurement of parameters for the separate phases. There are requirements for monitoring feedstocks and samples of the process stream as well as for in-situ monitoring. Fourier Transform infrared (FT-i.r.) absorption spectroscopy has been used previously as an in-situ diagnostic for both gas species concentration and gas temperature’ -9. Recently, a new method was developed for online, in-situ monitoring of particle laden streams to determine chemical composition, concentration, particle size and temperature of the individual components. The technique uses an FT-i.r. spectrometer to perform both emission and transmission (E/T) spectroscopy along a line of sight through the reacting medium. The theoretical foundation for the method was considered by Best et al.“. Applications of E/T spectroscopy were recently reported for: measuring particle temperatures in nonmeasuring particle emitcombusting media10-‘3; tance1’*14; measuring temperature and concentration of gases and soot in both combusting and non-combusting infrared analysis of gasmedia’ 5 ; and quantitative suspended solids”. An important application of the method was to address a controversy concerning coal optical properties recently raised by Brewster and Kunitomo”. Their recent measurements suggest that previous determinations for coal of the imaginary part of the index of refraction, k, may be too high by an order of magnitude. If so, the calculated coal emissivity based on these values will also be too high. The E/T technique was employed to * This paper was presented at the American Chemical Society Symposium ‘New Applications of Analytical Techniques to Fossil Fuels’. held at New York City, USA, April 1986
0016-2361/87/070897-12$3.00 0 1987 Butterworth & Co. (Publishers)
Ltd.
determine directly the spectral emittance, E,, of coal and char particles as a function of coal rank, extent of pyrolysis and temperature. For one coal, the measured spectral emittance was compared with a calculated value obtained from the material’s complex index of refraction N =n+ ik and particle size using Mie theory. The complex index of refraction was determined using an extension of Brewster and Kunitomo’s method for measuring k. The technique employs KBr and CsI pellet spectra and uses Mie theory and the Kramers-Kronig transforms to separate scattering and absorption effects. Good agreement has been obtained between the calculated and measured values of E”, confirming the contention that previously measured values of k are too high. EXPERIMENTAL Apparatus The apparatus employed in the experiments consists of an FT-i.r. spectrometer coupled to a reactor, such that the FT-i.r. focus passes through the sample stream as shown in Figure I. Emission measurements are made with the movable mirror in place. Transmission measurements are made with the movable mirror removed. The Fourier transform technique, in contrast to wavelength dispersive methods, processes all wavelengths of a spectrum simultaneously. For this reason it can be used to measure spectral properties of particulate flows, which are notoriously difficult to maintain at a constant level. The emission and transmission can be measured. for the same sample volume. The technique is extremely rapid; a low noise emission or transmission spectrum at low resolution (4 cm- ‘) can be recorded in under a second. Typically 128 scans are co-added to improve signal to noise. Also, radiation passing through the interferometer is amplitude modulated, and only such radiation is detected. Because of its unmodulated nature, the particulate emission passing directly to the detector
FUEL, 1987, Vol 66, July
897
Analysis
of particle properties 1
r-----------
I
by FT-i.r. emission/transmission
I I I I I
Canwter
I
I I
d <
et al.
n lr
Loser
(2)
R, = &IW
Source
> _TroosmIsslon Detector
P. R. Solomon
way :
llnterferaneter
I
I
spectroscopy:
where: WV= the instrument response function measured using a black-body cavity. The radiance measurement detects both radiation emitted by the particle itself, as well as wall radiation scattered or refracted by the particle as illustrated in Figure 2b. The normalized radiance, R”,, is calculated in the following way:
1
R’: = R,/( 1 - r,)
(3)
The complete analysis for the normalized emission has been presented elsewhere lo. For this discussion, the limit where particles are sufficiently large that they effectively block all the radiation incident on them and their diffraction pattern falls completely within the angular acceptance aperture of the spectrometer is considered. Then the normalized radiance has the simple form:
Sample Stream
external cell or reactor. Figure 1 FT-i.r. spectrometer with (Reproduced from Solomon, P. R. et al., Appl. Spectrosc. 1986, 40(6), 746, with the permission of Applied Spectroscopy)
does not interfere with the measurements of scattering or transmission. In an ideal emission/transmission experiment both measurements are made on the identical sample. Here, the emission and transmission measurements were made sequentially in time along the same optical path, for a sample flowing through the cell in a nominal steady condition. Measurements were made with a Nicolet model 7199 FT-i.r. using Mercury-Cadmium-Telluride (MCT) detectors.
R’:=(l
-E,)R,~(T,)+E,R,~(T,)
(4)
E, =the spectral emittance; and R:(T,) and R,b(T,) = the theoretical black-body curves corresponding
where:
to the temperature of the wall (T,) and particle (T,), respectively. This equation will be used in the discussion of results to follow. For the 200 x 325 mesh size fraction used for most of the experiments, this approximation gives measured E, values which are z 25 % too small at 1600 cm-‘.
Cl)
Measurements
To determine the temperature, emittance, size or composition, measurements are made of the transmittance and the radiance, from which a quantity called the normalized radiance is calculated. The normalized radiance is the radiance divided by (1 -transmittance) which in the absence of diffraction effects is a measure of intersected particle surface area. This normalization allows the radiance from particles which till only a fraction of the FT-i.r. aperture to be compared with a black-body standard obtained for the full aperture. The transmittance, rV,at wavenumber v, is measured in the usual way: r, = IV/L,
FUEL, 1987, Vol 66, July
Incident Beam
b)
(1)
where: I,, = the intensity transmitted through the cell in the absence of sample; and I, = that transmitted with the sample stream in place. The geometry for the transmittance measurement is illustrated in Figure 2a. With a particle in the focal volume, energy is taken out of the incident beam by absorption and scattering. The figure illustrates scattering by refraction of energy at the particles’ surfaces. Scattering will also be caused by reflection and diffraction. For particles > 50pm in diameter, whose index of refraction is sufficiently different from the surrounding medium, almost all the energy incident on the particle is absorbed or scattered. To measure the sample radiance, the power from the sample with background subtracted, S,, is measured, and converted to the sample radiance, R,, in the following
898
To
Figure 2 Measurement geometry. (a) Geometry of transmission measurement showing the incident beam refracted by the particle out of the beam; and (b) geometry of radiance measurement showing wall radiance refracted through a particle to the detector. (Reproduced from Solomon, P. R. et al., Appf. Spectrosc. 1986,40(6), 746, with permission of Applied Spectroscopy)
Analysis of particle properties
by FT-i.r. emission/transmission
RESULTS
spectroscopy: 1
5 65
2
Temperature and emittance
et al.
16'20
MICRONS
To illustrate these measurements several simple cases are considered, the first being hot particles such as coal or char surrounded by cold walls (T,$ T,). Then according to Equation (4): R’:=&,R,b(T,)
P. R. Solomon
(5)
If T, is known, then E, can be determined and vice versa. The first example is for char under conditions where the particle temperature is known. To achieve this condition, the particles are brought into thermal equilibrium with a heated tube reactor (HTR) and then exit the reactor and immediately pass the FT-i.r. aperture. The reactor has been described previously l1 . The gas-particle mixture cools about 75°C between the reactor and the FT-i.r. focus as determined with a thermocouple’ ‘. Figure 3 shows the radiance, (R,), 1 -transmittance, (1 -T,,), and the normalized radiance, (R:), from char emerging from the HTR at a temperature of 983 K. (1 -T,) is a measure of the intersected surface area of particles, i.e., 3.7% of the beam is blocked. The normalized radiance was compared with a number of grey body curves at different values of constant E. The E= 0.7 curve at 1000 K gives the best fit with the experimental data. The temperature is in good agreement with the thermocouple measurement. The second example is for lignite particles at a temperature of 782 K for a short time (IV0.1 s), before any pyrolysis has occurred. The normalized radiance spectrum in Figure 4a is compared with a grey-body (c=O.90) at the average thermocouple temperature of 782 K. While previously formed char shows a grey-body shape close to the average temperature, the lignite does not have a grey-body shape. The grey-body and experimental curves are close only in the range 160& lOOOcm_’ where the emittance, calculated from Equation (5) (Figure 4b), is close to 0.9. As will be demonstrated below, the emissivity of the particle is size and temperature dependent. For the size of coal particles used here, only specific bands between 1600 and 1000 cm - ’ (corresponding to the strongest absorption bands in coal) have a spectral emittance near 0.9. The result can be understood by considering that a particle’s emittance is related to its absorbance. Particles will emit only where they have sufficiently strong absorption bands. As discussed by Hottel and Sarofim”, for large particles (where diffraction can be neglected) the absorption can be calculated from geometrical optics considering all possible rays through the particles. The absorption within the particle can be calculated using absorbance values measured by the KBr pellet method 4,1g020.Figure 4c shows the absorbance measured by the KBr pellet method. The correspondence between the high absorbance bands and the regions of high emissivity are apparent. The most significant difference between the spectra of Figures 4a and c is the presence of a steeply sloping background going toward large wavenumbers in the pellet spectrum (Figure 4c) and its absence in the emittance spectrum (Figure 4b). The pellet spectrum is from a transmission measurement which does not distinguish between the absorption and scattered components of the total extinction cross section. This problem is treated in the section on Calculated Emissivity. By comparing the normalized radiance with a grey-
6SbO
4SbO
zsbo
sbo
WAVENUMBERS (cm-l)
k&O
zsbo
sbo
WAVENUM~ERS (cm-l 1
i3. 6SbO
4sbo &JO sbo WAVENUMBERS (cm-l 1
(a) Radiance; (b) lOO(1-transmittance); and (c) normalized from char formed at 1300°C in a previous test and subsequently measured at an asymptotic tube temperature of 1073 K (measured by thermocouples in the centre of the flow and at the wall), 115cm reaction distance (200 x 325 mesh) R/(1 -5)
FUEL, 1987, Vol 66, July
899
Analysis
of particle properties
by FT-i.r. emission/transmission
a
0.9
BB(782K)
b
spectroscopy:
Figure 4 Spectra for - 500 mesh size fraction of Zap, North Dakota lignite: (a) normalized radiance; (b) spectral emittance; and (c) KBr pellet spectrum. (Reproduced from Best, P. E. et al., Combustion and Flame 1986, 66, 47, with permission of the Combustion Institute)
900
FUEL, 1987, Vol 66, July
et al.
body curve for particles at the same temperature, it can be seen in Figure 5 that the emittance (the ratio of the normalized radiance to a black-body curve at the particle temperatures) varies with particle size. R: is smallest for small particle sizes and approaches the black-body curve for large size particles. The same comparison in Figure 6 shows that the spectral emittance varies with rank. An important feature is the emittance above 1700 cm-’ which increases at >90% carbon (Figure 6f). The increase in emittance at > 90% carbon is consistent with a corresponding increase in the broad sloping absorbance observed in KBr pellet spectra for high carbon coals. This absorbance is believed to be due to electronic absorption of multi-ring aromatic hydrocarbons”. The emittance of anthracite is close to grey-body. The spectral emittance variation with rank is consistent with the functional group variation with rank. Having demonstrated that the FT-i.r. method does give an appropriate temperature assuming E2 0.9 in the 1600lOOOcm_’ region, the technique was used to determine particle temperatures for non-isothermal conditions in the HTR reactor”. Figure 7 shows the E/T spectra for increasing time in the reactor. The FT-i.r. temperatures, which were used in conjunction with species evolution and reaction times to determine pyrolysis rates, were in good agreement with calculated temperatures and temperatures measured with a thermocouple’ ‘. Changes in emittance of particles as a function of the extent of pyrolysis can also be seen from Figurr 7. At 883 K (Figure 7a) all the absorption bands present in raw coal can be seen, except that the hydroxyl peak is noticeably depleted because of a reduction in hydrogen bonding. At 963 K (Figure 7b) a broad continuum, characteristic of char, is beginning to grow. By 1050K (Figure 74 almost a grey-body continuum can be seen with CO2 and H,O peaks superimposed. From the spectra discussed so far it can be said that the spectral emittance of lignite of this size range increases continuously with pyrolysis, reaching a maximum of 2 0.9 when primary pyrolysis is complete and then reducing to 0.7 at high temperatures, as the char becomes more graphitic. The emissivity of graphite in the i.r. region between 2 and 10 pm is reported to be 0.6-0.8 (ref. 22). The relationship of the development of the continuum char spectrum with the extent of pyrolysis will be the subject of a future investigation. The other examples are from a reactor where the particles are surrounded by hot walls. For the case where the particles are at wall temperature, T, = T,, Equation (4) reduces to R’: = R:( T,)
WAVENUMBERS
P. R. Solomon
(6)
Figure 8a displays Ri in one such case. In this case particle emittance does not have to be included in the analysis, as it can be shown, and Equation (6) indicates, that any lack of emittance must be made up by scattered and reflected radiation. Indeed, the normalized emission is close in shape and amplitude to a black-body at the measured window height wall temperature (1225 K). The determination of TP when it is of the same order of magnitude as T, requires a knowledge of E, at some wavenumber and use of the complete expression for R’: (Equation (4)). Examples are discussed in the literaturelO. Two additional examples are shown for the case TP<
Analysis of particle properties by FT-i.r. emission/transmission T,. Then Equation (4) reduces to: R: = (1 - I:, )R$ <, 1
(7)
spectroscopy:
Size information can be obtained from the (1 -z,) or RfJ spectra. Figure 5 shows the increase in the emissivity with -- - _~____..._ *This technique was developed by John McClelland of Ames Lahoratory
and the spectra
were supplied
by him
et al.
a NORTH DAKOTA LIGNITE 100 x 170 mesh
Figure 8h is for KCI, a non-absorbing particle, and Figure h’c is for coal which absorbs in specific bands. From Equation (71, Ra = R,b(T,) for KCI, since c = 0. R: matches rhe black-body curve at the wall temperature of IO60 K. The normalized radiance spectrum of the coal sample matches the wall black-body in regionswhere the coal is transparent (Le.. E: 0) but is lower where the coal has ,tbsorption bands. The emissivity c, = 1 - R:/Rf(T, ) can he determined from these spectra. The emissivity for the lignite of Fi~ur~~~ is shown in F~~~r~~~. when compared with the emlttance for the same coal at 782 K in Figure 4h, I he most significant difference is the larger hydroxyl band .:t low temperatures due to hydrogen bonding.
As illustrated in Figrtre _7b,the normalized radiance, R:, 1~the wall radiance Rf(T,,) attenuated over the chord ct at the particle’s absorbance bands. RC (e.g. Figure 8~) can, therefore, be used to obtain the absorbance of the particle it’ the average chord length, tl, is known. This has been done using a ray optics model and a computed ~)robab~~itydensity function Pfrl) for all possible chords cf. A more detailed discussion of the method has been presented elsewhere’h. Figuw 9 shows a comparison between the absorbance spectra obtained for the same cl)al by three methods: 1, a quantitative spectrum (i.e., for a known sample density) of a finely dispersed coal in a pressed K5r pellet; 2, a non-quantitative photoacoustic spectrum for fine partides suspended on a thin membrane*; and, 3, a quantitative ‘absorbance’spectrum determined by E/T spectroscopy considered to arise from the black-body furnace spectrum being attenuated by the particle over some effective sample thickness related to the shape and size of the particle. The details of the analysis to obtain the E/T absorbance spectra from the n
P. R. Solomon
I
b NORTH DAKOTA LIGNITE 200 x 325 mesh
0.9 BB(770K)
NORTH DAKOTA LSGNITE
b’igure 5 Normalized Dakota lignite
radiance
for different
size fractions
FUEL, 1987, Vot
of North
66, July
901
Analysis
of particle properties
by FT-i.r. emission/transmission
spectroscopy:
a
P. R. Solomon
et al.
d NORTH DAKOTA LIGNITE 66.52 C (DAF)
ROSEBUD SUBBlTUMlNOUS 72.12 C (DAF)
0.9 BB(67OK)
On9 BB(64OK) b
b
3
0
l-
x
b ILLINOIS16 73.9% C (DAF)
KENTUCKY #9 81.72 C (DAF)
a v; ? S
C
LOUER KITTANNING 91.2% C (DMF)
WAVENUMBERS Figure 6
902
Comparison of normalized emission (radiance) wuth theoretical grey-body curves (E= 0.9) for coal of different rank (200 x 325 mesh)
FUEL,
1987,
Vol 66, July
Analysis
of particle properties
6500 5500
4500
3500
2500
by FT-i.r. emission/transmission
1500
WAVENUNBERS Figure 7
6500
5500
45-W
spectroscopy:
3500
25bo
P. R. Solomon
1560
et al.
!
WAVENUMBERS
Comparison of normalized radiance with theoretical grey-body curves (E= 0.9) for chars at increasing extents of pyrolysis (200 x 325 mesh)
particle size, especially the region above 1700 wavenumbers. The indicated black-body temperatures correspond to thermocouple measurements taken at the FT-i.r. focus. The average particle size can be determined from RCif the complex index of refraction of the material is known. Alternatively, (1 - z,) contains size information for particles with D < 80 pm. For example (1 -T,) in Figure 3b increases at low wavenumbers (long wavelengths) due to diffractions effects. The shape may be calculated using Rayleigh theory for large particlesz4 or Mie theory for small particles 25. While the effect is small for particles near 80pm diameter, the diffraction effect increases as the particles decrease in size and are very sensitive to size in the l-30pm range. Figures IOU and b illustrate results for tine (< 20 pm diam.) mineral particles. The spectra show the characteristic absorption bands of the minerals (peaks in the 5~1500 wavenumber region). The spectra also illustrate the reduced attenuation at large wavelengths (i.e., low wavenumbers) characteristic of line particles. The silica particles (Figure IOU) which have a very tight size distribution show an increase in attenuation at 3500 wavenumbers (- 3 pm). This is a resonance effect where
the wavelength of the radiation matches the diameter of the particle. In this case, because the silica is very porous, its effective diameter is smaller than the observed diameter of 11 pm. The resonance is not so apparent for the ash (Figure ZOb) because it has a very broad size distribution. Figure IOc shows the results for a sample of coal with large size particles (i.e., 40pm diam.). The characteristics of such large particles are that they scatter or absorb essentially all of the light incident on them. Any radiation hitting the particle which is not absorbed will be scattered. Consequently, the attenuation is proportional to the area of the particle and no absorption effects can be seen. There is an increase in attenuation at long wavelengths due to diffraction effects. Figure 11 illustrates the use of theoretical reference spectra to interpret the observed spectra. The reference spectra were generated using Mie theory. The theory and computer code were published in Bohren and Huffman25, and so no further discussion of the theory will be given here. Mie theory is a general solution of Maxwell’s equations for an isotropic, homogeneous sphere (diameter D) with a complex index of refraction m = n + ik in a medium of different index of refraction. The complex
FUEL,
1987,
Vol 66, July
903
Analysis of particle properties by FT-i.r. emission/transmission spectroscopy: P. R. Solomon 9
2
et al.
-C
0
ci 0
d
0
ti
0
rd
0 -i
0
cj c C
6500 5500
1500
4500
500
100 5500
4500
3500
2500
1500
b
/ “1.
00 5500
4500
3500
2600
1500
WAVENUMBERS
6500
5500
4500
3500
2500
1500
500
WAVENUMBERS
Figure 8 Comparison of normalized radiance with a theoretical black-body curve for several cases: (a) coal, T, = TP;(b) KCI, TW= 1060% Tp; (c)coal T, = 1056% TP; and (d) emittance derived from (c) using Equation (7)
index of refraction contains all of the electromagnetic properties of non-magnetic materials. The complex index of refraction has been determined using a transmission method discussed in the next section for the samples of silica, ash and the Ohio No. 6 coal (for which spectra are shown in Figure IO)12,14. The absorbance used in Figure II is defined as (-log, 0 2,). When Mie theory is used to calculate the extinction properties of monodisperse spheres, it is found that surface resonances and other interference effects dominate the spectra. Such features are not observed in the data given here, and the scattering spectra should be compared with those calculated for scattering by a distribution of particle sizes, as was done by Bohren and Huffman2’. That is the approach used here. Use of a size distribution also makes the application of a theory for spheres to represent irregularly shaped particles a reasonable approximation. The particle size distribution which was chosen is a two-parameter (b and q) distribution, where the number of particles, P, having diameters between D and D + dD is given by:
904
FUEL, 1987, Vol 66, July
P=TDqv4exp(-bDy)dD which is the Rosin-Rammler function26. Calculated extinction spectra for several characteristic size distributions are shown in Figures 1 la, c and e. The size distribution parameters (b,q, the diameter of maximum intensity, D(peak), and width AD at half the intensity) are presented in Table 1. The spectra are normalized to give the same amplitude at 65OOcm-’ and the value of Fi500 is given in Table 1. The observed spectra normalized to the same amplitude at 6500 cm- ’ can then be compared with the reference spectra. Figure lla shows the results for silica. The reference spectra are for a tight size distribution as measured for this sample. The best fit between the spectrum and the reference spectrum is for an average diameter of 5 pm (spectrum c). This is substantially smaller than the observed diameter of 11 pm. The discrepancy is caused by the material’s porosity (x 65 % void fraction). This changes the effective index of refraction. Calculations were made using an index of refraction based on an
Analysis
of particle properties
by FT-i.r. emission/transmission
spectroscopy:
P. R. Solomon
et al.
measured spectra for the two size fractions of Ohio No. 6. These are expected to have an average difference of 18 pm in diameter. The reference spectra which best lit are for 60 and 80 pm diameters. These do show the right trends but were for a larger average radius than expected. CALCULATED
EMISSIVITY
The determination of spectral emittance is important for two reasons. The first is that knowledge of the spectral emittance is necessary for the measurement of particle temperatures. The second is that the spectral emittance or the emissivity (the average emittance over all
81 a
Bands Resonance
IO
3sbo esbo sbo WRVENLWERS
to
Figure 9 Comparison of absorbance spectra obtained by KBr pellet, photoacoustic spectroscopy (PAS), and Eflmethods. The KBr and E/T spectra are normalized to a 1 mg sample in a 1.33 cm’ area. The PAS spectrum is nonquantitative. The E/r spectrum was smoothed using a 25 point Savistky-Golay algorithm available on theNicolet 7199 FT-i.r. (Reproduced from Solomon, P. R. et al., Appl. Spectrosc. 1986, 40(6), 746. with permission of Applied Spectroscopy)
average between the silica and air (the voids). This average employs the Garnett relation for inhomogeneous materials given in Ref. 25. The resulting spectrum f calculated for 12 pm diameter spheres (58 y0 voids) is now in excellent agreement with the observed spectrum. Figure ZZc presents the results for ash. Here a wide distribution was used in agreement with the SEM results. The same problem of void fraction exists here as the powder has a density of 0.75 gcmw3 which suggests a void fraction of 57 %. While the best tit for solid material is for a diameter of 3 ,um, the fit for 57 % void fraction is 7 pm, in good agreement with the observed distribution. Results for coal are presented in Figure lle. For particles larger than 20,um diameter, the spectra are reasonably smooth with a slight upward slope going toward low wavenumbers and with a sharper rise approaching 500 cm - ‘. The variation with particle size is not drastic, so the sensitivity of the fitting procedure is much lower than for the small particles. Nevertheless, the observed spectrum for two size fractions of coal (200 x 325 and 325 x 400) do show the expected difference, and can be used to infer relative size. Figure Ile compares the
Diffraction
6500
4500
2500
500
WAVENUMBERS Figure 10 Infrared extinction lOO(1-T,) for particles exiting the heated tube reactor at an asymptotic tube temperature of 25°C. (a) Silica (1.4 g min -I feed rate); (b) fly ash (3.4 g mitt-’ feed rate); and (c) Ohio No. 6 coal (2.Og min-’ feed rate)
FUEL,
1987,
Vol 66, July
905
Analysis
of particle
properties
by FT-i-r.
emissionftransmissjon
spectroscopy:
P. R. Solomon
et al
9.25
WAVENUMBERS
18.5
DIAMETER (microns)
d
*
.
6500 7.0
,
0.0
1
4500
2500
7
500
6.85
WAVE~BERS
13.7
DIAMETER (microns) I
,
s-
6500
4500
2500
500
WAVE~~S
0
81.5
163
DIAMETER (microns)
Figure 11 Infrared extinction coefficients for particles. The spectra are arbitrarily scaled to give the same amplitude at 6500 wavenumbers. (a) Comparison of measured and computed extinction efficiency for silica. Theory curve c is for D= 5 pm, solid particles; f is for D= 12 pm with a pore volume of 65 %. (b) Particle size distribution for simulation $ (c)Comparison of measured and computed extinction efficiency for ash. Theory curve (i) is for D = 3 I”, solid particles; k is for D = 7 pm with a pore volume of 57 %. (d) Particle size distribution for simulation k. (e) Comparison 01 measured and computed extinction efficiency for coal particles of two size cuts (200 x 325 and 325 x 400). (f) Particle size dist~bution for simulation m
Table 1
Particle distribution parameters Porosity D(peak) (void fraction) w)
F6s00
2.4 1.3
12 5
0.58 0
1.96 1.53
3 7
0 0.57
2.64 1.95
-
1.12 1.10
AD b
4
9.36 4.45 x lo-” lo-l6
14
lum)
Silica F Ash
i
8.23 x 1O-4 1.35x1o-5
5 4.2
3.1 13.5
Coal m n
2.04x lo-l4 1.53x lo-‘3
7 7
40 30
906
FUEL, 1987, Vol 66, July
k
z
wavelengths) must be known to calculate the rate of particle heat-up or the radiative energy released by the particle during gasification or combustion. For small particles, the spectral emittance is usually not measured. Rather, it is calculated using the complex index of refraction, N= n +ik, from the standard equations of electromagnetic theory. For spherical particles these calculations are performed using Mie theory. Unfortunately, some controversy surrounds the value of k for coal. The problem was recently considered by Brewster and Kunitomo ” . Large variations in the value of k have been reported in the literature ranging from < 0.1 to > 0.5. Brewster and Kunitomo have suggested that a possible reason for these discrepancies is that
Analysis-of
particle properties
by FT-i.r. emission/transmission
previous measurements of k for coal using reflection techniques were highly inaccurate due to an inherent limitation in the ability to get sufficiently smooth, and homogeneous, coal surfaces. Using a transmission technique for small particles in KBr, they obtained values of k which are more than an order of magnitude lower in most regions of the spectrum. If the previously measured values of k are too high, then values of emissivity based on them are also too high. Based on the values of emissivity presented above, this is indeed the case. In this section, the measured values of E, are compared with the predictions of Mie theory. For these calculations, the value of n and k were determined using KBr and CsI pellet spectra for coal, in an extension of the method of Brewster and Kunitomo “. To determine N= n, + ‘,$ from pellet spectra, Mie theory and the Kramers-Kromg transform are used. In general, the complex index of refraction contains all of the electromagnetic properties of the material, but the real and imaginary parts, n, and k,, must be Kramers-Kronig transforms of each other2’. Physically, this relationship arises by requiring causality, i.e., the material cannot respond to radiation until after the incident radiation hits the material. In practice, we determine k,, perform a Fourier Transform into the time domain, force causality in the time domain, and then transform back to the frequency domain to obtain n (Ref. 27). The complete description of the method will be the subject of a future publication, but briefly the procedure for determining n and k is as follows: 1, With trial values of II, = n, (a constant) and k, = 0 and a particle diameter D, for coal in KBr, Mie theory is used to calculate a smooth scattering curve. 2. This scattering curve, plus a constant to account for reflection from the pellet surface, is subtracted from the experimental spectrum and the result is assumed to be a pure absorption, A,, where: 4nk
volume of coal
2.30261
area of pellet
Ay=px
3. Calculate k, from Equation (9) and use the KramersKronig transform to determine the correction to n,. 4. Using these values of nvr k, and D, Mie theory is used to predict a CsI pellet spectrum. Since the scattering >4OOOcm- ’ depends largely on the product (ncoalnmedium)D,this fourth step allows D and n, to be determined. This procedure is repeated until both the KBr and CsI spectra can be predicted with the same n, k and D. The results of this procedure are shown in Figure 12. Figures 12a and b are the KBr and CsI pellet spectra for a Montana Rosebud coal. Values of no = 1.66 and D = 2.6pm were found to give reasonable fits to the data. Figure 12c shows the calculated n and k; and Figure 12d shows the absorbance due to k (which now depends only on the coal, not the pellet material). Figures 22a and b also show the calculated scattering contribution to the absorbance spectrum (which depends on the difference between n, for the coal and the medium). The structure in the scattering curves is the Christiansen effect, which results from a decrease in n on the high frequency side of an absorption band and an increase in n on the low frequency side. As can be seen from Figures 12a and b, the effect on the scattering is in the opposite direction for the KBr (lower n than the coal) and CsI (higher n than the coal) pellets. Figure Z2d shows the calculated total spectrum for the CsI pellet which is the sum of the absorption and scattering contributions. It is to be compared with the measured spectra (Figure 12b). The
P. R. Solomon
et al.
agreement is quite good, most obviously for the Christiansen effect on the minimum near 15OOcm-‘. With a particle radius of 25 pm, Mie theory is used with the n, and k, from Figure 12~ to predict the emittance spectrum of cold coal, Figure Z2e. This is compared with the measured emittance in Figure 22f: The agreement is good. As mentioned above, the measured value is expected to be =c25 % too low at 16OOcm-’ owing to the use of the large particle approximations. CONCLUSION The FT-i.r. emission/transmission method has been used successfully for determining composition, spectral emittance, size and temperature of gas-suspended particles. The method appears to be accurate, reliable and easy to use. It has applications to online process monitoring and in-situ diagnostics. ACKNOWLEDGEMENTS The general E/T method development and measurements of temperature and emittance were performed under DOE contract No. DE-AC21-81FE05122, the measurement of composition and size under NSF SBIR Grant No. CPE-8460379, the measurement of index of refraction and calculating of emissivities under DOE SBIR contract No. DE-ACOl-85ER80313. The authors wish to thank Woodrow Fievland and Richard Wassel of Babcock and Wilcox for helpful discussions on the coal emissivity problems and John McClelland of Ames Laboratory for supplying the PAS spectrum. REFERENCES 1
Liebman, S. A., Ahlstrom, D. H. and Grifliths, P.R. Appl. Spectrosc. 1976, 30(3), 355
2
(9)
spectroscopy:
3
4 5 6 7 8
9
10 11 12
13 14
15
Erickson, M. D., Frazier, S. E. and Sparacino, C. M. Fuel 1981, 60, 263 Solomon, P. R., Hamblen, D. G., Carangelo, R. M. and Krause, J. L. Coal Thermal Decomposition in an Entrained Flow Reactor; Experiment and Theory’, 19th Symposium (International) on Combustion, The Combustion Institute, Pittsburgh, PA, 1982, p. 1139 Solomon, P. R., Hamblen, D. G. and Carangelo, R. M. Am. Chem. Sot. Symp. Ser. 1982, 205, 4, 77 Ottensen, D. K. and Stephenson, D. A. Combustion and Flame 1982, 46, 95 Ottensen, D. K. and Thorne, L. R. International Conference on Coal Science, Pittsburgh, PA, 1983, p. 621 Solomon, P. R. ‘Pyrolysis’, in ‘Chemistry of Coal Conversion’ (Ed. R. H. Schlosberg), Plenum, New York, 1985, p. 121 Solomon, P. R. and Hamblen, D. G. ‘Measurements and Theory of Coal Pyrolysis’, A Topical Report for US Department of Energy, Oflice of Fossil Energy, Morgantown Energy Technology Center, Contract No. DE-AC21-81FE05122, 1984 Solomon, P. R., Hamblen, D. Cl. and Carangelo, R. M. ‘Analytical Pyrolysis’ (Ed. K. J. Voorhees), Butterworths, London, 1984, Chapter 5, p. 121 Best, P. E., Carangelo, R. M. and Solomon, P. R. Combustion and Flame 1986,66,47 Solomon, P. R., Serio, M. A., Carangelo, R. M. and Markham, J. R. Fuel 1986,65, 182 Solomon, P. R., Carangelo, R. M., Best, P. E., Markham, J. R. and Hamblen, D. G. Am. Chem. Sot. Div. Fuel Chem., Prepr. 1986, 31(l), 141 Best, P. E., Carangelo, R. M. and Solomon, P. R. Am. Chem. Sot. Div. Fuel Chem., Prepr. 1984, 29(6), 249 Solomon, P. R., Carangelo, R. M., Best, P. E., Markham, J. R. and Hamblen, D. G. ‘The Spectral Emittance of Pulverized Coal and Char’, presented at the 21st Symposium (Int) on Combustion Meeting, Munich, Germany, 1986 Solomon, P. R., Best, P. E., Carangelo, R. M., Markham, J. R., Chien, P.-L., Santoro, R. J. and Semerjian, H. G. ‘FT-IR Emission/Transmission Spectroscopy for In-Situ Combustion
FUEL,
1987,
Vol 66, July
907
Analysis
of particle properties
by FT-i.r. emission/transmission
spectroscopy:
f? R. Solomon
et al.
I
ti
6
C 6500
5500
YSOO
I 3500
2500
,500
JO
WAVENUMBERS Figure 12 (a) Measured spectrum for Montana Rosebud in KBr with part of spectrum due to scattering. (b) Measured spectrum for Montana Rosebud in CsI with part of spectrum due to scattering. (cc)Real and imaginaq parts of the index of refraction, N = n + i/c. (d) Upper curve is calculated spectrum for Montana Rosebud in Csf, lower curve is calculated absorbance of Montana Rosebud without scattering. (e) Measured emittance and (f) calculated emittance for Montana Rosebud.
16
17 I8 19
Diagnostics’, presented at the 21st Symposium (Int) on Combustion Meeting, Munich, Germany, 1986 Solomon, P. R., Carangelo, R. M., Hamblen, D. G. and Best, P. E. Appl. Spectrosc. 198640, 746 Brewster, M. Q. and Kunitomo, T. Trans. ASME 1984,106,678 Hottel, H. C. and Sarotim, A. F. ‘Radiative Transfer’, McGraw Hill, New York, 1967 Solomon, P. R., Hamblen, D. G., Carangelo, R. M., Markham, .I. R. and Chaffee, M. R. Am. Chem. SK Div. Fuel Chem., Prepr. 1985, 30(l), 1 Solomon, P. R. and Carangelo, R. M. Fuel 1982,61, 663 van Krevelen, D. W. ‘Coal’, Eisevier, Amsterdam, 1961
908 FUEL, 1987, Vol 66, July
23
Foster, P. J. and Howarth, C. R. Curbon 1968, 6, 719; and references therein Laufer. G., Huneke, J. T., Royce. B. S. H. and Teng, Y. C. Appf.
24
Gumprecht, R. 0. and Sliepceivch, C. M, J. Phys. Chem. 1953,
22
Phys. Lett.,
1980, 37, 517
57,90
25 26 27
Bhoren, C. F. and Huffman, D. R. ‘Absorption and Scattering of Light by Small Particles’, John Wiley, NY, 1983 Chigier, N. ‘Energy Combustion and Environment’, McGraw Hill, NY, 1981, pi 26t Peterson, C. W. and Knight, B. W. .I. Optimal Sot. America 1973, 63(10), 1238
Robert M. Carangelo, Philip and David G. Hamblen
E. Best,
Advanced Fuel Research, Inc., 87 Church Street, East Hartford, (Received 20 October 1986; revised 28 January 7987)
CT 06108,
USA
A new method was developed for online, in-situ monitoring of particle laden streams to determine chemical composition, concentration, particle size and temperature of the individual components. The technique uses an FT-i.r. spectrometer to perform both emission and transmission spectroscopy along a line of sight through the reacting medium. The technique has been applied to measure particle properties under a variety of circumstances and several examples are presented for measurements of temperature, emittance, composition and size. (Keywords: instrumental methods of analysis; FT-i.r.; particle properties)
Many chemical and energy conversion processes involve multi-phase feed or product streams. Streams can consist of mixtures of solids, liquids and gases and in some cases the important species may be in transition between phases (e.g. liquid fuel combustion which involves liquid fuel droplets, vaporized fuel and soot). Process monitoring requires the measurement of parameters for the separate phases. There are requirements for monitoring feedstocks and samples of the process stream as well as for in-situ monitoring. Fourier Transform infrared (FT-i.r.) absorption spectroscopy has been used previously as an in-situ diagnostic for both gas species concentration and gas temperature’ -9. Recently, a new method was developed for online, in-situ monitoring of particle laden streams to determine chemical composition, concentration, particle size and temperature of the individual components. The technique uses an FT-i.r. spectrometer to perform both emission and transmission (E/T) spectroscopy along a line of sight through the reacting medium. The theoretical foundation for the method was considered by Best et al.“. Applications of E/T spectroscopy were recently reported for: measuring particle temperatures in nonmeasuring particle emitcombusting media10-‘3; tance1’*14; measuring temperature and concentration of gases and soot in both combusting and non-combusting infrared analysis of gasmedia’ 5 ; and quantitative suspended solids”. An important application of the method was to address a controversy concerning coal optical properties recently raised by Brewster and Kunitomo”. Their recent measurements suggest that previous determinations for coal of the imaginary part of the index of refraction, k, may be too high by an order of magnitude. If so, the calculated coal emissivity based on these values will also be too high. The E/T technique was employed to * This paper was presented at the American Chemical Society Symposium ‘New Applications of Analytical Techniques to Fossil Fuels’. held at New York City, USA, April 1986
0016-2361/87/070897-12$3.00 0 1987 Butterworth & Co. (Publishers)
Ltd.
determine directly the spectral emittance, E,, of coal and char particles as a function of coal rank, extent of pyrolysis and temperature. For one coal, the measured spectral emittance was compared with a calculated value obtained from the material’s complex index of refraction N =n+ ik and particle size using Mie theory. The complex index of refraction was determined using an extension of Brewster and Kunitomo’s method for measuring k. The technique employs KBr and CsI pellet spectra and uses Mie theory and the Kramers-Kronig transforms to separate scattering and absorption effects. Good agreement has been obtained between the calculated and measured values of E”, confirming the contention that previously measured values of k are too high. EXPERIMENTAL Apparatus The apparatus employed in the experiments consists of an FT-i.r. spectrometer coupled to a reactor, such that the FT-i.r. focus passes through the sample stream as shown in Figure I. Emission measurements are made with the movable mirror in place. Transmission measurements are made with the movable mirror removed. The Fourier transform technique, in contrast to wavelength dispersive methods, processes all wavelengths of a spectrum simultaneously. For this reason it can be used to measure spectral properties of particulate flows, which are notoriously difficult to maintain at a constant level. The emission and transmission can be measured. for the same sample volume. The technique is extremely rapid; a low noise emission or transmission spectrum at low resolution (4 cm- ‘) can be recorded in under a second. Typically 128 scans are co-added to improve signal to noise. Also, radiation passing through the interferometer is amplitude modulated, and only such radiation is detected. Because of its unmodulated nature, the particulate emission passing directly to the detector
FUEL, 1987, Vol 66, July
897
Analysis
of particle properties 1
r-----------
I
by FT-i.r. emission/transmission
I I I I I
Canwter
I
I I
d <
et al.
n lr
Loser
(2)
R, = &IW
Source
> _TroosmIsslon Detector
P. R. Solomon
way :
llnterferaneter
I
I
spectroscopy:
where: WV= the instrument response function measured using a black-body cavity. The radiance measurement detects both radiation emitted by the particle itself, as well as wall radiation scattered or refracted by the particle as illustrated in Figure 2b. The normalized radiance, R”,, is calculated in the following way:
1
R’: = R,/( 1 - r,)
(3)
The complete analysis for the normalized emission has been presented elsewhere lo. For this discussion, the limit where particles are sufficiently large that they effectively block all the radiation incident on them and their diffraction pattern falls completely within the angular acceptance aperture of the spectrometer is considered. Then the normalized radiance has the simple form:
Sample Stream
external cell or reactor. Figure 1 FT-i.r. spectrometer with (Reproduced from Solomon, P. R. et al., Appl. Spectrosc. 1986, 40(6), 746, with the permission of Applied Spectroscopy)
does not interfere with the measurements of scattering or transmission. In an ideal emission/transmission experiment both measurements are made on the identical sample. Here, the emission and transmission measurements were made sequentially in time along the same optical path, for a sample flowing through the cell in a nominal steady condition. Measurements were made with a Nicolet model 7199 FT-i.r. using Mercury-Cadmium-Telluride (MCT) detectors.
R’:=(l
-E,)R,~(T,)+E,R,~(T,)
(4)
E, =the spectral emittance; and R:(T,) and R,b(T,) = the theoretical black-body curves corresponding
where:
to the temperature of the wall (T,) and particle (T,), respectively. This equation will be used in the discussion of results to follow. For the 200 x 325 mesh size fraction used for most of the experiments, this approximation gives measured E, values which are z 25 % too small at 1600 cm-‘.
Cl)
Measurements
To determine the temperature, emittance, size or composition, measurements are made of the transmittance and the radiance, from which a quantity called the normalized radiance is calculated. The normalized radiance is the radiance divided by (1 -transmittance) which in the absence of diffraction effects is a measure of intersected particle surface area. This normalization allows the radiance from particles which till only a fraction of the FT-i.r. aperture to be compared with a black-body standard obtained for the full aperture. The transmittance, rV,at wavenumber v, is measured in the usual way: r, = IV/L,
FUEL, 1987, Vol 66, July
Incident Beam
b)
(1)
where: I,, = the intensity transmitted through the cell in the absence of sample; and I, = that transmitted with the sample stream in place. The geometry for the transmittance measurement is illustrated in Figure 2a. With a particle in the focal volume, energy is taken out of the incident beam by absorption and scattering. The figure illustrates scattering by refraction of energy at the particles’ surfaces. Scattering will also be caused by reflection and diffraction. For particles > 50pm in diameter, whose index of refraction is sufficiently different from the surrounding medium, almost all the energy incident on the particle is absorbed or scattered. To measure the sample radiance, the power from the sample with background subtracted, S,, is measured, and converted to the sample radiance, R,, in the following
898
To
Figure 2 Measurement geometry. (a) Geometry of transmission measurement showing the incident beam refracted by the particle out of the beam; and (b) geometry of radiance measurement showing wall radiance refracted through a particle to the detector. (Reproduced from Solomon, P. R. et al., Appf. Spectrosc. 1986,40(6), 746, with permission of Applied Spectroscopy)
Analysis of particle properties
by FT-i.r. emission/transmission
RESULTS
spectroscopy: 1
5 65
2
Temperature and emittance
et al.
16'20
MICRONS
To illustrate these measurements several simple cases are considered, the first being hot particles such as coal or char surrounded by cold walls (T,$ T,). Then according to Equation (4): R’:=&,R,b(T,)
P. R. Solomon
(5)
If T, is known, then E, can be determined and vice versa. The first example is for char under conditions where the particle temperature is known. To achieve this condition, the particles are brought into thermal equilibrium with a heated tube reactor (HTR) and then exit the reactor and immediately pass the FT-i.r. aperture. The reactor has been described previously l1 . The gas-particle mixture cools about 75°C between the reactor and the FT-i.r. focus as determined with a thermocouple’ ‘. Figure 3 shows the radiance, (R,), 1 -transmittance, (1 -T,,), and the normalized radiance, (R:), from char emerging from the HTR at a temperature of 983 K. (1 -T,) is a measure of the intersected surface area of particles, i.e., 3.7% of the beam is blocked. The normalized radiance was compared with a number of grey body curves at different values of constant E. The E= 0.7 curve at 1000 K gives the best fit with the experimental data. The temperature is in good agreement with the thermocouple measurement. The second example is for lignite particles at a temperature of 782 K for a short time (IV0.1 s), before any pyrolysis has occurred. The normalized radiance spectrum in Figure 4a is compared with a grey-body (c=O.90) at the average thermocouple temperature of 782 K. While previously formed char shows a grey-body shape close to the average temperature, the lignite does not have a grey-body shape. The grey-body and experimental curves are close only in the range 160& lOOOcm_’ where the emittance, calculated from Equation (5) (Figure 4b), is close to 0.9. As will be demonstrated below, the emissivity of the particle is size and temperature dependent. For the size of coal particles used here, only specific bands between 1600 and 1000 cm - ’ (corresponding to the strongest absorption bands in coal) have a spectral emittance near 0.9. The result can be understood by considering that a particle’s emittance is related to its absorbance. Particles will emit only where they have sufficiently strong absorption bands. As discussed by Hottel and Sarofim”, for large particles (where diffraction can be neglected) the absorption can be calculated from geometrical optics considering all possible rays through the particles. The absorption within the particle can be calculated using absorbance values measured by the KBr pellet method 4,1g020.Figure 4c shows the absorbance measured by the KBr pellet method. The correspondence between the high absorbance bands and the regions of high emissivity are apparent. The most significant difference between the spectra of Figures 4a and c is the presence of a steeply sloping background going toward large wavenumbers in the pellet spectrum (Figure 4c) and its absence in the emittance spectrum (Figure 4b). The pellet spectrum is from a transmission measurement which does not distinguish between the absorption and scattered components of the total extinction cross section. This problem is treated in the section on Calculated Emissivity. By comparing the normalized radiance with a grey-
6SbO
4SbO
zsbo
sbo
WAVENUMBERS (cm-l)
k&O
zsbo
sbo
WAVENUM~ERS (cm-l 1
i3. 6SbO
4sbo &JO sbo WAVENUMBERS (cm-l 1
(a) Radiance; (b) lOO(1-transmittance); and (c) normalized from char formed at 1300°C in a previous test and subsequently measured at an asymptotic tube temperature of 1073 K (measured by thermocouples in the centre of the flow and at the wall), 115cm reaction distance (200 x 325 mesh) R/(1 -5)
FUEL, 1987, Vol 66, July
899
Analysis
of particle properties
by FT-i.r. emission/transmission
a
0.9
BB(782K)
b
spectroscopy:
Figure 4 Spectra for - 500 mesh size fraction of Zap, North Dakota lignite: (a) normalized radiance; (b) spectral emittance; and (c) KBr pellet spectrum. (Reproduced from Best, P. E. et al., Combustion and Flame 1986, 66, 47, with permission of the Combustion Institute)
900
FUEL, 1987, Vol 66, July
et al.
body curve for particles at the same temperature, it can be seen in Figure 5 that the emittance (the ratio of the normalized radiance to a black-body curve at the particle temperatures) varies with particle size. R: is smallest for small particle sizes and approaches the black-body curve for large size particles. The same comparison in Figure 6 shows that the spectral emittance varies with rank. An important feature is the emittance above 1700 cm-’ which increases at >90% carbon (Figure 6f). The increase in emittance at > 90% carbon is consistent with a corresponding increase in the broad sloping absorbance observed in KBr pellet spectra for high carbon coals. This absorbance is believed to be due to electronic absorption of multi-ring aromatic hydrocarbons”. The emittance of anthracite is close to grey-body. The spectral emittance variation with rank is consistent with the functional group variation with rank. Having demonstrated that the FT-i.r. method does give an appropriate temperature assuming E2 0.9 in the 1600lOOOcm_’ region, the technique was used to determine particle temperatures for non-isothermal conditions in the HTR reactor”. Figure 7 shows the E/T spectra for increasing time in the reactor. The FT-i.r. temperatures, which were used in conjunction with species evolution and reaction times to determine pyrolysis rates, were in good agreement with calculated temperatures and temperatures measured with a thermocouple’ ‘. Changes in emittance of particles as a function of the extent of pyrolysis can also be seen from Figurr 7. At 883 K (Figure 7a) all the absorption bands present in raw coal can be seen, except that the hydroxyl peak is noticeably depleted because of a reduction in hydrogen bonding. At 963 K (Figure 7b) a broad continuum, characteristic of char, is beginning to grow. By 1050K (Figure 74 almost a grey-body continuum can be seen with CO2 and H,O peaks superimposed. From the spectra discussed so far it can be said that the spectral emittance of lignite of this size range increases continuously with pyrolysis, reaching a maximum of 2 0.9 when primary pyrolysis is complete and then reducing to 0.7 at high temperatures, as the char becomes more graphitic. The emissivity of graphite in the i.r. region between 2 and 10 pm is reported to be 0.6-0.8 (ref. 22). The relationship of the development of the continuum char spectrum with the extent of pyrolysis will be the subject of a future investigation. The other examples are from a reactor where the particles are surrounded by hot walls. For the case where the particles are at wall temperature, T, = T,, Equation (4) reduces to R’: = R:( T,)
WAVENUMBERS
P. R. Solomon
(6)
Figure 8a displays Ri in one such case. In this case particle emittance does not have to be included in the analysis, as it can be shown, and Equation (6) indicates, that any lack of emittance must be made up by scattered and reflected radiation. Indeed, the normalized emission is close in shape and amplitude to a black-body at the measured window height wall temperature (1225 K). The determination of TP when it is of the same order of magnitude as T, requires a knowledge of E, at some wavenumber and use of the complete expression for R’: (Equation (4)). Examples are discussed in the literaturelO. Two additional examples are shown for the case TP<
Analysis of particle properties by FT-i.r. emission/transmission T,. Then Equation (4) reduces to: R: = (1 - I:, )R$ <, 1
(7)
spectroscopy:
Size information can be obtained from the (1 -z,) or RfJ spectra. Figure 5 shows the increase in the emissivity with -- - _~____..._ *This technique was developed by John McClelland of Ames Lahoratory
and the spectra
were supplied
by him
et al.
a NORTH DAKOTA LIGNITE 100 x 170 mesh
Figure 8h is for KCI, a non-absorbing particle, and Figure h’c is for coal which absorbs in specific bands. From Equation (71, Ra = R,b(T,) for KCI, since c = 0. R: matches rhe black-body curve at the wall temperature of IO60 K. The normalized radiance spectrum of the coal sample matches the wall black-body in regionswhere the coal is transparent (Le.. E: 0) but is lower where the coal has ,tbsorption bands. The emissivity c, = 1 - R:/Rf(T, ) can he determined from these spectra. The emissivity for the lignite of Fi~ur~~~ is shown in F~~~r~~~. when compared with the emlttance for the same coal at 782 K in Figure 4h, I he most significant difference is the larger hydroxyl band .:t low temperatures due to hydrogen bonding.
As illustrated in Figrtre _7b,the normalized radiance, R:, 1~the wall radiance Rf(T,,) attenuated over the chord ct at the particle’s absorbance bands. RC (e.g. Figure 8~) can, therefore, be used to obtain the absorbance of the particle it’ the average chord length, tl, is known. This has been done using a ray optics model and a computed ~)robab~~itydensity function Pfrl) for all possible chords cf. A more detailed discussion of the method has been presented elsewhere’h. Figuw 9 shows a comparison between the absorbance spectra obtained for the same cl)al by three methods: 1, a quantitative spectrum (i.e., for a known sample density) of a finely dispersed coal in a pressed K5r pellet; 2, a non-quantitative photoacoustic spectrum for fine partides suspended on a thin membrane*; and, 3, a quantitative ‘absorbance’spectrum determined by E/T spectroscopy considered to arise from the black-body furnace spectrum being attenuated by the particle over some effective sample thickness related to the shape and size of the particle. The details of the analysis to obtain the E/T absorbance spectra from the n
P. R. Solomon
I
b NORTH DAKOTA LIGNITE 200 x 325 mesh
0.9 BB(770K)
NORTH DAKOTA LSGNITE
b’igure 5 Normalized Dakota lignite
radiance
for different
size fractions
FUEL, 1987, Vot
of North
66, July
901
Analysis
of particle properties
by FT-i.r. emission/transmission
spectroscopy:
a
P. R. Solomon
et al.
d NORTH DAKOTA LIGNITE 66.52 C (DAF)
ROSEBUD SUBBlTUMlNOUS 72.12 C (DAF)
0.9 BB(67OK)
On9 BB(64OK) b
b
3
0
l-
x
b ILLINOIS16 73.9% C (DAF)
KENTUCKY #9 81.72 C (DAF)
a v; ? S
C
LOUER KITTANNING 91.2% C (DMF)
WAVENUMBERS Figure 6
902
Comparison of normalized emission (radiance) wuth theoretical grey-body curves (E= 0.9) for coal of different rank (200 x 325 mesh)
FUEL,
1987,
Vol 66, July
Analysis
of particle properties
6500 5500
4500
3500
2500
by FT-i.r. emission/transmission
1500
WAVENUNBERS Figure 7
6500
5500
45-W
spectroscopy:
3500
25bo
P. R. Solomon
1560
et al.
!
WAVENUMBERS
Comparison of normalized radiance with theoretical grey-body curves (E= 0.9) for chars at increasing extents of pyrolysis (200 x 325 mesh)
particle size, especially the region above 1700 wavenumbers. The indicated black-body temperatures correspond to thermocouple measurements taken at the FT-i.r. focus. The average particle size can be determined from RCif the complex index of refraction of the material is known. Alternatively, (1 - z,) contains size information for particles with D < 80 pm. For example (1 -T,) in Figure 3b increases at low wavenumbers (long wavelengths) due to diffractions effects. The shape may be calculated using Rayleigh theory for large particlesz4 or Mie theory for small particles 25. While the effect is small for particles near 80pm diameter, the diffraction effect increases as the particles decrease in size and are very sensitive to size in the l-30pm range. Figures IOU and b illustrate results for tine (< 20 pm diam.) mineral particles. The spectra show the characteristic absorption bands of the minerals (peaks in the 5~1500 wavenumber region). The spectra also illustrate the reduced attenuation at large wavelengths (i.e., low wavenumbers) characteristic of line particles. The silica particles (Figure IOU) which have a very tight size distribution show an increase in attenuation at 3500 wavenumbers (- 3 pm). This is a resonance effect where
the wavelength of the radiation matches the diameter of the particle. In this case, because the silica is very porous, its effective diameter is smaller than the observed diameter of 11 pm. The resonance is not so apparent for the ash (Figure ZOb) because it has a very broad size distribution. Figure IOc shows the results for a sample of coal with large size particles (i.e., 40pm diam.). The characteristics of such large particles are that they scatter or absorb essentially all of the light incident on them. Any radiation hitting the particle which is not absorbed will be scattered. Consequently, the attenuation is proportional to the area of the particle and no absorption effects can be seen. There is an increase in attenuation at long wavelengths due to diffraction effects. Figure 11 illustrates the use of theoretical reference spectra to interpret the observed spectra. The reference spectra were generated using Mie theory. The theory and computer code were published in Bohren and Huffman25, and so no further discussion of the theory will be given here. Mie theory is a general solution of Maxwell’s equations for an isotropic, homogeneous sphere (diameter D) with a complex index of refraction m = n + ik in a medium of different index of refraction. The complex
FUEL,
1987,
Vol 66, July
903
Analysis of particle properties by FT-i.r. emission/transmission spectroscopy: P. R. Solomon 9
2
et al.
-C
0
ci 0
d
0
ti
0
rd
0 -i
0
cj c C
6500 5500
1500
4500
500
100 5500
4500
3500
2500
1500
b
/ “1.
00 5500
4500
3500
2600
1500
WAVENUMBERS
6500
5500
4500
3500
2500
1500
500
WAVENUMBERS
Figure 8 Comparison of normalized radiance with a theoretical black-body curve for several cases: (a) coal, T, = TP;(b) KCI, TW= 1060% Tp; (c)coal T, = 1056% TP; and (d) emittance derived from (c) using Equation (7)
index of refraction contains all of the electromagnetic properties of non-magnetic materials. The complex index of refraction has been determined using a transmission method discussed in the next section for the samples of silica, ash and the Ohio No. 6 coal (for which spectra are shown in Figure IO)12,14. The absorbance used in Figure II is defined as (-log, 0 2,). When Mie theory is used to calculate the extinction properties of monodisperse spheres, it is found that surface resonances and other interference effects dominate the spectra. Such features are not observed in the data given here, and the scattering spectra should be compared with those calculated for scattering by a distribution of particle sizes, as was done by Bohren and Huffman2’. That is the approach used here. Use of a size distribution also makes the application of a theory for spheres to represent irregularly shaped particles a reasonable approximation. The particle size distribution which was chosen is a two-parameter (b and q) distribution, where the number of particles, P, having diameters between D and D + dD is given by:
904
FUEL, 1987, Vol 66, July
P=TDqv4exp(-bDy)dD which is the Rosin-Rammler function26. Calculated extinction spectra for several characteristic size distributions are shown in Figures 1 la, c and e. The size distribution parameters (b,q, the diameter of maximum intensity, D(peak), and width AD at half the intensity) are presented in Table 1. The spectra are normalized to give the same amplitude at 65OOcm-’ and the value of Fi500 is given in Table 1. The observed spectra normalized to the same amplitude at 6500 cm- ’ can then be compared with the reference spectra. Figure lla shows the results for silica. The reference spectra are for a tight size distribution as measured for this sample. The best fit between the spectrum and the reference spectrum is for an average diameter of 5 pm (spectrum c). This is substantially smaller than the observed diameter of 11 pm. The discrepancy is caused by the material’s porosity (x 65 % void fraction). This changes the effective index of refraction. Calculations were made using an index of refraction based on an
Analysis
of particle properties
by FT-i.r. emission/transmission
spectroscopy:
P. R. Solomon
et al.
measured spectra for the two size fractions of Ohio No. 6. These are expected to have an average difference of 18 pm in diameter. The reference spectra which best lit are for 60 and 80 pm diameters. These do show the right trends but were for a larger average radius than expected. CALCULATED
EMISSIVITY
The determination of spectral emittance is important for two reasons. The first is that knowledge of the spectral emittance is necessary for the measurement of particle temperatures. The second is that the spectral emittance or the emissivity (the average emittance over all
81 a
Bands Resonance
IO
3sbo esbo sbo WRVENLWERS
to
Figure 9 Comparison of absorbance spectra obtained by KBr pellet, photoacoustic spectroscopy (PAS), and Eflmethods. The KBr and E/T spectra are normalized to a 1 mg sample in a 1.33 cm’ area. The PAS spectrum is nonquantitative. The E/r spectrum was smoothed using a 25 point Savistky-Golay algorithm available on theNicolet 7199 FT-i.r. (Reproduced from Solomon, P. R. et al., Appl. Spectrosc. 1986, 40(6), 746. with permission of Applied Spectroscopy)
average between the silica and air (the voids). This average employs the Garnett relation for inhomogeneous materials given in Ref. 25. The resulting spectrum f calculated for 12 pm diameter spheres (58 y0 voids) is now in excellent agreement with the observed spectrum. Figure ZZc presents the results for ash. Here a wide distribution was used in agreement with the SEM results. The same problem of void fraction exists here as the powder has a density of 0.75 gcmw3 which suggests a void fraction of 57 %. While the best tit for solid material is for a diameter of 3 ,um, the fit for 57 % void fraction is 7 pm, in good agreement with the observed distribution. Results for coal are presented in Figure lle. For particles larger than 20,um diameter, the spectra are reasonably smooth with a slight upward slope going toward low wavenumbers and with a sharper rise approaching 500 cm - ‘. The variation with particle size is not drastic, so the sensitivity of the fitting procedure is much lower than for the small particles. Nevertheless, the observed spectrum for two size fractions of coal (200 x 325 and 325 x 400) do show the expected difference, and can be used to infer relative size. Figure Ile compares the
Diffraction
6500
4500
2500
500
WAVENUMBERS Figure 10 Infrared extinction lOO(1-T,) for particles exiting the heated tube reactor at an asymptotic tube temperature of 25°C. (a) Silica (1.4 g min -I feed rate); (b) fly ash (3.4 g mitt-’ feed rate); and (c) Ohio No. 6 coal (2.Og min-’ feed rate)
FUEL,
1987,
Vol 66, July
905
Analysis
of particle
properties
by FT-i-r.
emissionftransmissjon
spectroscopy:
P. R. Solomon
et al
9.25
WAVENUMBERS
18.5
DIAMETER (microns)
d
*
.
6500 7.0
,
0.0
1
4500
2500
7
500
6.85
WAVE~BERS
13.7
DIAMETER (microns) I
,
s-
6500
4500
2500
500
WAVE~~S
0
81.5
163
DIAMETER (microns)
Figure 11 Infrared extinction coefficients for particles. The spectra are arbitrarily scaled to give the same amplitude at 6500 wavenumbers. (a) Comparison of measured and computed extinction efficiency for silica. Theory curve c is for D= 5 pm, solid particles; f is for D= 12 pm with a pore volume of 65 %. (b) Particle size distribution for simulation $ (c)Comparison of measured and computed extinction efficiency for ash. Theory curve (i) is for D = 3 I”, solid particles; k is for D = 7 pm with a pore volume of 57 %. (d) Particle size distribution for simulation k. (e) Comparison 01 measured and computed extinction efficiency for coal particles of two size cuts (200 x 325 and 325 x 400). (f) Particle size dist~bution for simulation m
Table 1
Particle distribution parameters Porosity D(peak) (void fraction) w)
F6s00
2.4 1.3
12 5
0.58 0
1.96 1.53
3 7
0 0.57
2.64 1.95
-
1.12 1.10
AD b
4
9.36 4.45 x lo-” lo-l6
14
lum)
Silica F Ash
i
8.23 x 1O-4 1.35x1o-5
5 4.2
3.1 13.5
Coal m n
2.04x lo-l4 1.53x lo-‘3
7 7
40 30
906
FUEL, 1987, Vol 66, July
k
z
wavelengths) must be known to calculate the rate of particle heat-up or the radiative energy released by the particle during gasification or combustion. For small particles, the spectral emittance is usually not measured. Rather, it is calculated using the complex index of refraction, N= n +ik, from the standard equations of electromagnetic theory. For spherical particles these calculations are performed using Mie theory. Unfortunately, some controversy surrounds the value of k for coal. The problem was recently considered by Brewster and Kunitomo ” . Large variations in the value of k have been reported in the literature ranging from < 0.1 to > 0.5. Brewster and Kunitomo have suggested that a possible reason for these discrepancies is that
Analysis-of
particle properties
by FT-i.r. emission/transmission
previous measurements of k for coal using reflection techniques were highly inaccurate due to an inherent limitation in the ability to get sufficiently smooth, and homogeneous, coal surfaces. Using a transmission technique for small particles in KBr, they obtained values of k which are more than an order of magnitude lower in most regions of the spectrum. If the previously measured values of k are too high, then values of emissivity based on them are also too high. Based on the values of emissivity presented above, this is indeed the case. In this section, the measured values of E, are compared with the predictions of Mie theory. For these calculations, the value of n and k were determined using KBr and CsI pellet spectra for coal, in an extension of the method of Brewster and Kunitomo “. To determine N= n, + ‘,$ from pellet spectra, Mie theory and the Kramers-Kromg transform are used. In general, the complex index of refraction contains all of the electromagnetic properties of the material, but the real and imaginary parts, n, and k,, must be Kramers-Kronig transforms of each other2’. Physically, this relationship arises by requiring causality, i.e., the material cannot respond to radiation until after the incident radiation hits the material. In practice, we determine k,, perform a Fourier Transform into the time domain, force causality in the time domain, and then transform back to the frequency domain to obtain n (Ref. 27). The complete description of the method will be the subject of a future publication, but briefly the procedure for determining n and k is as follows: 1, With trial values of II, = n, (a constant) and k, = 0 and a particle diameter D, for coal in KBr, Mie theory is used to calculate a smooth scattering curve. 2. This scattering curve, plus a constant to account for reflection from the pellet surface, is subtracted from the experimental spectrum and the result is assumed to be a pure absorption, A,, where: 4nk
volume of coal
2.30261
area of pellet
Ay=px
3. Calculate k, from Equation (9) and use the KramersKronig transform to determine the correction to n,. 4. Using these values of nvr k, and D, Mie theory is used to predict a CsI pellet spectrum. Since the scattering >4OOOcm- ’ depends largely on the product (ncoalnmedium)D,this fourth step allows D and n, to be determined. This procedure is repeated until both the KBr and CsI spectra can be predicted with the same n, k and D. The results of this procedure are shown in Figure 12. Figures 12a and b are the KBr and CsI pellet spectra for a Montana Rosebud coal. Values of no = 1.66 and D = 2.6pm were found to give reasonable fits to the data. Figure 12c shows the calculated n and k; and Figure 12d shows the absorbance due to k (which now depends only on the coal, not the pellet material). Figures 22a and b also show the calculated scattering contribution to the absorbance spectrum (which depends on the difference between n, for the coal and the medium). The structure in the scattering curves is the Christiansen effect, which results from a decrease in n on the high frequency side of an absorption band and an increase in n on the low frequency side. As can be seen from Figures 12a and b, the effect on the scattering is in the opposite direction for the KBr (lower n than the coal) and CsI (higher n than the coal) pellets. Figure Z2d shows the calculated total spectrum for the CsI pellet which is the sum of the absorption and scattering contributions. It is to be compared with the measured spectra (Figure 12b). The
P. R. Solomon
et al.
agreement is quite good, most obviously for the Christiansen effect on the minimum near 15OOcm-‘. With a particle radius of 25 pm, Mie theory is used with the n, and k, from Figure 12~ to predict the emittance spectrum of cold coal, Figure Z2e. This is compared with the measured emittance in Figure 22f: The agreement is good. As mentioned above, the measured value is expected to be =c25 % too low at 16OOcm-’ owing to the use of the large particle approximations. CONCLUSION The FT-i.r. emission/transmission method has been used successfully for determining composition, spectral emittance, size and temperature of gas-suspended particles. The method appears to be accurate, reliable and easy to use. It has applications to online process monitoring and in-situ diagnostics. ACKNOWLEDGEMENTS The general E/T method development and measurements of temperature and emittance were performed under DOE contract No. DE-AC21-81FE05122, the measurement of composition and size under NSF SBIR Grant No. CPE-8460379, the measurement of index of refraction and calculating of emissivities under DOE SBIR contract No. DE-ACOl-85ER80313. The authors wish to thank Woodrow Fievland and Richard Wassel of Babcock and Wilcox for helpful discussions on the coal emissivity problems and John McClelland of Ames Laboratory for supplying the PAS spectrum. REFERENCES 1
Liebman, S. A., Ahlstrom, D. H. and Grifliths, P.R. Appl. Spectrosc. 1976, 30(3), 355
2
(9)
spectroscopy:
3
4 5 6 7 8
9
10 11 12
13 14
15
Erickson, M. D., Frazier, S. E. and Sparacino, C. M. Fuel 1981, 60, 263 Solomon, P. R., Hamblen, D. G., Carangelo, R. M. and Krause, J. L. Coal Thermal Decomposition in an Entrained Flow Reactor; Experiment and Theory’, 19th Symposium (International) on Combustion, The Combustion Institute, Pittsburgh, PA, 1982, p. 1139 Solomon, P. R., Hamblen, D. G. and Carangelo, R. M. Am. Chem. Sot. Symp. Ser. 1982, 205, 4, 77 Ottensen, D. K. and Stephenson, D. A. Combustion and Flame 1982, 46, 95 Ottensen, D. K. and Thorne, L. R. International Conference on Coal Science, Pittsburgh, PA, 1983, p. 621 Solomon, P. R. ‘Pyrolysis’, in ‘Chemistry of Coal Conversion’ (Ed. R. H. Schlosberg), Plenum, New York, 1985, p. 121 Solomon, P. R. and Hamblen, D. G. ‘Measurements and Theory of Coal Pyrolysis’, A Topical Report for US Department of Energy, Oflice of Fossil Energy, Morgantown Energy Technology Center, Contract No. DE-AC21-81FE05122, 1984 Solomon, P. R., Hamblen, D. Cl. and Carangelo, R. M. ‘Analytical Pyrolysis’ (Ed. K. J. Voorhees), Butterworths, London, 1984, Chapter 5, p. 121 Best, P. E., Carangelo, R. M. and Solomon, P. R. Combustion and Flame 1986,66,47 Solomon, P. R., Serio, M. A., Carangelo, R. M. and Markham, J. R. Fuel 1986,65, 182 Solomon, P. R., Carangelo, R. M., Best, P. E., Markham, J. R. and Hamblen, D. G. Am. Chem. Sot. Div. Fuel Chem., Prepr. 1986, 31(l), 141 Best, P. E., Carangelo, R. M. and Solomon, P. R. Am. Chem. Sot. Div. Fuel Chem., Prepr. 1984, 29(6), 249 Solomon, P. R., Carangelo, R. M., Best, P. E., Markham, J. R. and Hamblen, D. G. ‘The Spectral Emittance of Pulverized Coal and Char’, presented at the 21st Symposium (Int) on Combustion Meeting, Munich, Germany, 1986 Solomon, P. R., Best, P. E., Carangelo, R. M., Markham, J. R., Chien, P.-L., Santoro, R. J. and Semerjian, H. G. ‘FT-IR Emission/Transmission Spectroscopy for In-Situ Combustion
FUEL,
1987,
Vol 66, July
907
Analysis
of particle properties
by FT-i.r. emission/transmission
spectroscopy:
f? R. Solomon
et al.
I
ti
6
C 6500
5500
YSOO
I 3500
2500
,500
JO
WAVENUMBERS Figure 12 (a) Measured spectrum for Montana Rosebud in KBr with part of spectrum due to scattering. (b) Measured spectrum for Montana Rosebud in CsI with part of spectrum due to scattering. (cc)Real and imaginaq parts of the index of refraction, N = n + i/c. (d) Upper curve is calculated spectrum for Montana Rosebud in Csf, lower curve is calculated absorbance of Montana Rosebud without scattering. (e) Measured emittance and (f) calculated emittance for Montana Rosebud.
16
17 I8 19
Diagnostics’, presented at the 21st Symposium (Int) on Combustion Meeting, Munich, Germany, 1986 Solomon, P. R., Carangelo, R. M., Hamblen, D. G. and Best, P. E. Appl. Spectrosc. 198640, 746 Brewster, M. Q. and Kunitomo, T. Trans. ASME 1984,106,678 Hottel, H. C. and Sarotim, A. F. ‘Radiative Transfer’, McGraw Hill, New York, 1967 Solomon, P. R., Hamblen, D. G., Carangelo, R. M., Markham, .I. R. and Chaffee, M. R. Am. Chem. SK Div. Fuel Chem., Prepr. 1985, 30(l), 1 Solomon, P. R. and Carangelo, R. M. Fuel 1982,61, 663 van Krevelen, D. W. ‘Coal’, Eisevier, Amsterdam, 1961
908 FUEL, 1987, Vol 66, July
23
Foster, P. J. and Howarth, C. R. Curbon 1968, 6, 719; and references therein Laufer. G., Huneke, J. T., Royce. B. S. H. and Teng, Y. C. Appf.
24
Gumprecht, R. 0. and Sliepceivch, C. M, J. Phys. Chem. 1953,
22
Phys. Lett.,
1980, 37, 517
57,90
25 26 27
Bhoren, C. F. and Huffman, D. R. ‘Absorption and Scattering of Light by Small Particles’, John Wiley, NY, 1983 Chigier, N. ‘Energy Combustion and Environment’, McGraw Hill, NY, 1981, pi 26t Peterson, C. W. and Knight, B. W. .I. Optimal Sot. America 1973, 63(10), 1238