Advanced Search
RSS LinkedIn Twitter

Journal Archive

Johnson Matthey Technol. Rev., 2020, 64, (2), 165

doi:10.1595/205651320x15754757907469

Insights into Automotive Particulate Filters using Magnetic Resonance Imaging

Understanding filter drying in the manufacturing process and the effect of particulate matter on filter operation and fluid dynamics

    • J. D. Cooper, N. P. Ramskill, A. J. Sederman, L. F. Gladden
    • Department of Chemical Engineering and Biotechnology, University of Cambridge, West Cambridge Site, Philippa Fawcett Drive, Cambridge, CB3 0AS, UK
    • A. Tsolakis
    • School of Mechanical Engineering, University of Birmingham, Edgbaston, Birmingham, B15 2TT, UK
    • E. H. Stitt
    • Johnson Matthey, PO Box 1, Belasis Avenue, Billingham, Cleveland, TS23 1LB, UK
    • A. P. E. York*
    • Johnson Matthey, Blounts Court, Sonning Common, Reading, RG4 9NH, UK
    • Email: *ayork@matthey.com
SHARE THIS PAGE:

Article Synopsis

Understanding the manufacture and operation of automotive emissions control particulate filters is important in the optimised design of these emissions control systems. Here we show how magnetic resonance imaging (MRI) can be used to understand the drying process, which is part of the manufacture of catalysed particulate filters. Comparison between a wall-flow particulate filter substrate and a flow-through monolith (FTM) has been performed, with MRI giving spatial information on the drying process. We have also used MRI to study the fluid dynamics of a gasoline particulate filter (GPF). Inlet and outlet channel gas velocities have been measured for a clean GPF and two GPF samples loaded with particulate matter (PM) to understand the effect of PM on the filter flow profiles and porous wall permeability as soot is deposited.

Introduction

Particulate filters, comprising so-called wall-flow filter substrates, are of increasing importance in reducing pollutant emissions from vehicles to the levels required by legislation. Early legislation addressed the emissions of carbon monoxide, hydrocarbons and nitrogen oxides and the removal of these pollutants was achieved using FTM catalysts. Later, when PM emissions from diesel vehicles came under scrutiny, wall-flow filter devices were added to the emissions control system: often these were uncatalysed extruded cordierite or silicon carbide filter monoliths, though sometimes a catalyst was incorporated on the filter to widen their operating window. Wall-flow filters differ in their operation from FTMs, since adjacent channels are alternately blocked meaning exhaust gas must pass through the porous monolith wall to flow from inlet to outlet; in this way PM is deposited on the inlet channel walls. Previous PM legislation was based on particulate mass emissions, however, more recently the legislation, such as in Euro 6 (1), has turned to address particulate number emissions. Due to this change in emphasis both diesel and gasoline vehicles now require filter systems to achieve compliance. In gasoline applications, a GPF is typically a wall-flow filter, similar to that used for diesel, with a catalyst coating applied: by combining the catalyst and filter devices, multifunctional emissions control systems have also been developed resulting in reduced packaging volume.

In the manufacture of catalysed filter devices, first a catalyst coating is applied to the monolith substrate in the form of a slurry. The subsequent coated monolith is then dried and finally calcined to fix the catalyst coating. The drying process is both energy intensive and known to influence the final metal distribution within the catalyst throughout the catalysed filter and therefore also influences filter performance on a vehicle. Indeed, it has previously been shown (2, 3) that non-ideal drying in FTMs can result in the macroscopic redistribution of the catalyst. In the work of Vergunst et al. (3) and Wahlberg et al. (4) it was observed that the metal phase, when not bound to the catalyst support, will migrate to the surface where evaporation occurs. This results in varying degrees of inhomogeneous catalyst distribution depending on the method of drying used. Finally, enrichment of the catalyst in the washcoat layer of ceramic monoliths during drying has been observed using MRI (5). Traditionally, drying is studied using gravimetric methods, humidity and temperature measurements from which the drying kinetics can be determined (6). Although these techniques are well established in both industry and research, they are somewhat limited in that they are only able to provide macroscopic measurements and the process itself must be treated as a ‘black box’. Spatially resolved information has most often been obtained by sectioning the sample and then weighing the individual components. To gain a greater understanding of the intrinsic water migration characteristics during drying, in the first part of this work we use MRI to image the time-resolved water distribution during the drying. The data for the filter are compared with the same data acquired during drying of the related FTM.

The second part of this work uses MRI methods to measure the gas velocity in the channels of a clean and particulate loaded filter. During the operation of an automobile, PM-laden exhaust gas passes through the particulate filter and the particulate or soot is deposited inside. However, the micro- and macroscopic distribution of this soot deposition impacts the subsequent filtration behaviour, pressure drop, regeneration behaviour and ultimately the useful lifetime of the filter. Hence, a complete understanding of the filtration process is needed in order to optimise the function of particulate filters. The regeneration behaviour of particulate filters is coupled with the gas fluid dynamics. Both heat transfer and the mass transfer of oxidative species (oxygen for active regeneration and nitrogen dioxide for passive regeneration) are affected by the gas flow fields and will impact the efficacy and safety of the regeneration process. Large thermal gradients within the filter can be formed if the soot distribution is non-uniform that may damage the filter. While modelling and macroscopic measurements of these effects have been performed by many authors, there is a lack of experimental work focusing on the relationship between the filter structure, the gas transport and the perturbations of these by the loaded soot. As with drying, the range of techniques available to non-invasively measure flow in opaque filter systems is very limited, and most studies have used models (710). Magnetic resonance (MR) can provide spatial information on the effect of the PM on the filter operation.

MR techniques have gained prominence in chemical engineering research as they provide a non-invasive method of studying the chemistry and dynamics of a range of opaque systems (11). They are particularly useful for the study of porous media such as catalysts (1216), construction materials (1719) and pharmaceuticals (20, 21). Such techniques have also been used to provide information on the drying mechanism in a range of other applications such as detergent powders (22, 23), dehydration and preparation of foodstuffs (2426), evaporation from contaminated surfaces (2729) and fired-clay brick at elevated temperature (30). Applications of MR to drying and sorption have been covered in the review by Koptyug (31). While MR has previously been used to study drying and active component distribution in FTMs by Koptyug and coworkers (5, 15), no studies have looked at the drying process in particulate filters to date. This study investigates how the structure of the filter substrate influences the water migration behaviour during drying. MR techniques are used to characterise water migration behaviour in a wall-flow filter and FTM substrate under identical drying conditions. While MRI of liquids and their transport is common, imaging studies of nuclear magnetic resonance (NMR) active gas flows are relatively few. This is mainly due to challenges associated with the low signal-to-noise ratio (SNR) presented to the experimentalist. While hyperpolarised gases (for example, xenon-129) offer a significant boost to the SNR, they are costly and unsuitable for studying porous materials. The first application of imaging thermally polarised gases was demonstrated by Koptyug et al. (3234) who acquired two-dimensional (2D) velocity images of hydrocarbon gases at atmospheric pressure flowing through a cylindrical pipe and alumina monoliths of different channel geometries. In the monolith studies (32), from the spatially resolved profiles of the axial component of the velocity vector on the individual channel scale, information pertaining to shear rates and entry lengths were obtained which enabled useful insights into the mass transfer between the bulk gas flow and the porous channel walls in the monolith to be made. Codd and Altobelli (35) have also shown the application of using thermally polarised gases as a probe of the structure of porous materials. Ramskill et al. (36) performed an early study using sulfur hexafluoride gas to image flow profiles inside a clean emissions control filter. They implemented compressed sensing (CS) techniques, which enable spectra and images to be reconstructed with sufficient accuracy from relatively few data points and allow a reduction in data acquisition times (3741). Here we show a continuation of this study with more commercially relevant samples and with particulate loading.

This study highlights two ways in which MRI can be used to gain insight into automotive particulate filters. First a comparison of water migration during drying within a wall-flow particulate filter is reported and compared with the analogous process occurring within a FTM. The mechanism of water migration during drying will influence catalyst distribution in the manufactured filter and valuable insights are gained by being able to image this process in more than one dimension. Second, MRI is used to image gas fluid dynamics within a GPF both in the clean state and following two stages of soot loading. From these flow profiles, permeabilities as a function of space and time are predicted. Using MRI to gain a greater understanding of automotive particulate filters could eventually lead to improved catalyst properties, filtration efficiency and more efficient and improved manufacturing.

Principles of Magnetic Resonance Imaging

This section provides the reader with a brief introduction to the principles of MRI but the interested reader is directed to the texts by Callaghan (42) and Haacke (43) and review articles by Mantle and Sederman (44) and Caprihan and Fukushima (45) for further detail.

When NMR active nuclei (such as 1H or 19F) are placed in an external magnetic field (B0), the nuclei will precess at a characteristic frequency known as the Larmor frequency (ω0) as given by Equation (i):

(i)

where γ is the gyromagnetic ratio which is characteristic of the nuclei under observation. Considering the system in the rotating frame at reference frequency, ω0, spatial dependence of the precession frequency is achieved by applying a spatially-dependent magnetic field gradient (G) and can be expressed as follows (Equation (ii)):

(ii)

where ω(r) is the precession frequency at position vector r. The precessing nuclei in a volume element induce a voltage in the receiver coil and the complex NMR signal, S(t), is detected in the time domain. The time-dependent signal is then Fourier transformed into the frequency domain where it is represented by the spin-density function, ρ(r). The Fourier conjugate relationship between the time and frequency domain of the NMR signal is shown below in Equations (iii) and (iv):

(iii)
(iv)

From this, a spin-density map, i.e. an image, is obtained by taking the modulus of the complex function, ρ(r). In the case of the drying experiments the spin-density map provides a spatially-resolved measurement of the water content within the monolith sample.

Magnetic field gradients can be used to make the NMR signal sensitive to nuclei displacements in addition to position. This is achieved by first encoding the nuclei positions by applying a given magnetic field gradient (G) for time (δ), then decoding by applying the same magnetic field gradient in the opposing direction after an evolution time (Δ). Any static nuclei will be unchanged but any moving nuclei will create a phase shift in the NMR signal, ΔΦ proportional to their displacement Δr (Equation (v)):

(v)

Hence the gas velocity can be measured by consideration of the signal phase at each spatial location.

Materials and Methods

Comparison of Drying in a Particulate Filter and a Flow Through Monolith

A cordierite wall-flow particulate filter and a cordierite FTM were used. These are typical of substrates used commercially and the relevant properties of the two samples are listed in Table I. Porosity and mean pore size measurements were made using an AutoPore IV system mercury porosimeter (Micromeritics Instrument Corporation, USA). Drying of pure water from the respective substrates has been investigated under identical conditions i.e. with air at 20 l min–1 ± 2 l min–1 and temperature of 19.5°C ± 0.5°C.

Table I

Properties of the Diesel Particulate Filter and FTM Substratesa

Wall-flow FTM
Material cordierite
Length (L), mm 75
Core diameter, mm 26
Channel hydraulic diameter, mm 1
Substrate porosity (ɛ), % 48±4 24±2
Mean pore size, μm 13.8±7.8 2.8±1.1
Water content (mc), g 6.1±0.1 2.8±0.1

aThe mean pore size and porosity were determined by mercury porosimetry.

To saturate the samples, the substrates were immersed in deionised water for two minutes and then shaken to remove any water blocking the channels. A schematic of the experimental set up used is shown in Figure 1. The filter substrate was held within a PERSPEX® cell (Perspex International Ltd, UK) above an air distributor plate used to produce a uniform flow of air over the cross-sectional area of the filter. The relative humidity (RH) and temperature of the air flow at the inlet and outlet of the drying cell were recorded at 10 s intervals over the course of the process using a Humidiprobe (Pico Technology, UK). Temperature and RH measurements were recorded with an accuracy of ±0.5°C and ±2%, respectively. These measurements allow the total uptake of moisture by the air to be determined and thus the drying rate can be calculated through conservation of the total water mass (36, 46).

Fig. 1.

Schematic of the experimental setup for the drying experiments. A = compressed air line, B = pressure regulator, C = rotameter, D = air distributor plate, E = wall-flow filter or FTM substrate sample, F = imaging region, G = MRI spectrometer and H = magnetic field gradients

Schematic of the experimental setup for the drying experiments. A = compressed air line, B = pressure regulator, C = rotameter, D = air distributor plate, E = wall-flow filter or FTM substrate sample, F = imaging region, G = MRI spectrometer and H = magnetic field gradients

Effect of Soot Loading on Gas Fluid Dynamics in a Gasoline Particulate Filter

A cordierite GPF sample was prepared in the laboratory for this study. The cordierite substrate (55% porosity) was coated with a Pd/Rh alumina three-way catalyst typical of commercial catalysts used for GPF applications. The properties of the sample are shown in Table II, with the porosity and pore size of the catalyst coated filter listed for the front, middle and rear of the sample.

Table II

Properties of the GPF Samplea

Bare filter Catalyst coated filter
Material cordierite
Length (L), mm 145
Core diameter, mm 25 mm for soot loading, 6 mm for MRI
Channel hydraulic diameter, mm 1
Substrate porosity (ɛ), % 58 28.1, 25.9, 29.2
Mean pore size, μm 21 19, 16, 18

aPorosity and pore size are given for the front, middle and rear of the sample.

Samples were soot loaded using a 2 l, four-cylinder gasoline direct injection (GDI) turbocharged engine, and then subsequently removed and transferred to a different sample holder for the MRI flow experiments. Such engines are typical of current passenger automobiles. The engine was run at 2100 rpm, producing a torque of 60 Nm. Three 25 mm diameter cores were bored from the centre of the GPF, each acting as a sample for the soot loading. The filter sample to be loaded was held downstream inside the exhaust manifold. The manifold was surrounded by a furnace, allowing the filter to be held at different temperatures. A temperature of 300°C was chosen as representative of real-world gasoline exhausts; the temperature was measured using a thermocouple inserted into the manifold upstream of the filter. Two pressure transducers were placed either side of the filter sample, allowing measurement of the pressure drop during the loading process. Measurements were made at 180 ms intervals at both transducers. The pressure readings were subtracted and averaged over 2 min intervals to give the transient pressure drop. Three particulate loading protocols were used:

  • Protocol I: no soot loading

  • Protocol II: normal running of the engine for 50 min

  • Protocol III: Protocol II followed by 10 min of accelerated soot loading achieved by delaying the fuel injection by a crank-shaft angle of 50 degrees.

The backpressure profiles recorded for Protocols II and III are shown in Figure 2.

Fig. 2.

Pressure drop measurements for the GPF sample subject to particulate loading Protocols II and III

Pressure drop measurements for the GPF sample subject to particulate loading Protocols II and III

Magnetic Resonance Characterisation

For the drying experiments, the MR experiments were performed using a 2 Tesla (85 MHz for 1H) horizontal bore magnet controlled by an AV spectrometer (Bruker Corporation, USA). An 85 mm radio frequency (rf) coil tuned to a frequency of 85.1 MHz was used for excitation and signal detection and spatial resolution was achieved with magnetic field gradients with a maximum strength of 10.7 gauss cm–1. Three MR techniques were used as follows.

  • NMR spectroscopy was used to provide a quantitative measurement of the bulk water content through calibration with gravimetric measurements. Due to the short deadtime between rf excitation and detection, negligible relaxation weighting is associated with the spectra and hence they are directly proportional to the spin density of water. A recycle time of 3.5 s and eight scans for signal averaging were used, resulting in a total acquisition time of 0.5 min per spectrum

  • 2D images were acquired over the course of drying using the rapid acquisition with relaxation enhancement (RARE) pulse sequence (47). Images were acquired in the yz plane with a slice width of 10 mm, an in-plane field-of-view (FOV) of 80 mm × 30 mm and a data matrix symbol size of 32 × 32, giving an in-plane pixel resolution of 2.5 mm px–1 × 0.94 mm px–1 in the read (z) and phase (y) directions respectively. A RARE factor of four with eight scans were used, allowing acquisition of a full image in 3.5 min

  • One-dimensional (1D) profiles in the axial (z) direction were acquired using a spin-echo profiling sequence that integrates the spin density along the x and y directions (42). A FOV of 80 mm in the z direction and a matrix size of 128 points was used, giving a spatial resolution of 0.625 mm px–1. Eight scans were used for signal averaging, giving a total acquisition time of 0.5 min. The echo time between excitation and acquisition of the NMR signal was 10 ms, giving a maximum error of 3% for the relaxation times present in the system.

The MRI method used to measure gas velocity in the filter samples is described fully in Ramskill et al. (36). In the present study SF6 has been chosen as the NMR active gas to be used for velocity imaging due to its favourable MR properties in comparison with other potential candidates such as the hydrocarbon gases (48). Eleven images were acquired along the length of the GPF samples, each with a slice width of 6 mm. An SF6 gas pressure of 5.0 barg ± 0.1 barg and mass flow rate of 16 g min−1 was used for each sample. Axial velocity profiles were acquired for the GPF samples after all three soot loading protocols. The mean volume flow for each sample agreed with the value calculated from the mass flow rate to within 8.5%. The through-wall velocities were calculated for each based on the gas mass balance. Velocity profiles inside the inlet channels were extracted from the MR velocity images through the mid-point of the channels parallel to the filter wall.

Results and Discussion

Comparison of Drying in a Particulate Filter and a Flow Through Monolith

2D images in the yz plane of the wall-flow filter and FTM have been acquired over the course of drying as shown in Figure 3 and Figure 4 respectively. The initial water distribution along the two substrates has been determined by integrating the signal intensity in the first image of each sequence (Figure 3(a) and Figure 4(a)) in the y-direction. These data are shown in Figure 5. A uniform wetting of the channels of both substrates is seen with an area of relatively high signal intensity corresponding to the higher moisture content contained in the plugs of the wall-flow filter substrate. Figure 6 shows the moisture content at the three radial positions (marked as A, B and C on Figures 3 and 4) as a function of drying time. Apart from the longer drying time associated with the wall-flow filter substrate, the behaviour of the water saturation at the three positions across the monolith as a function of time show very similar behaviour. This result confirms that the 1D profiles in the axial (z) direction are sufficient to be able to study the drying mechanism in the wall-flow filter and FTM and thereby allow the process to be studied at a higher temporal resolution than would be permitted using 2D MRI.

Fig. 3.

2D images over the course of the drying of a wall-flow filter at: (a) 3.5 min; (b) 10.5 min; (c) 17.5 min; (d) 24.5 min; (e) 30.5 min and (f) 36.5 min. The signal intensity has been normalised relative to the maximum signal intensity in the filter substrate. Air flow is from the bottom with a volumetric flow rate of 20 l min–1. These images were acquired with a FOV of 80 mm × 30 mm in the zy plane corresponding to a spatial resolution of 2.5 mm × 0.94 mm

2D images over the course of the drying of a wall-flow filter at: (a) 3.5 min; (b) 10.5 min; (c) 17.5 min; (d) 24.5 min; (e) 30.5 min and (f) 36.5 min. The signal intensity has been normalised relative to the maximum signal intensity in the filter substrate. Air flow is from the bottom with a volumetric flow rate of 20 l min–1. These images were acquired with a FOV of 80 mm × 30 mm in the zy plane corresponding to a spatial resolution of 2.5 mm × 0.94 mm

Fig. 4.

2D qualitative images over the course of the drying of a FTM at: (a) 3.5 min; (b) 7 min; (c) 10.5 min; (d) 14 min: (e) 17 min and (f) 20.5 min. The signal intensity has been normalised relative to the maximum signal intensity in the filter material. Air flow is from the bottom with a volumetric flow rate of 20 l min–1. These images were acquired with a FOV of 80 mm × 30 mm in the zy plane corresponding to a spatial resolution of 2.5 mm × 0.94 mm

2D qualitative images over the course of the drying of a FTM at: (a) 3.5 min; (b) 7 min; (c) 10.5 min; (d) 14 min: (e) 17 min and (f) 20.5 min. The signal intensity has been normalised relative to the maximum signal intensity in the filter material. Air flow is from the bottom with a volumetric flow rate of 20 l min–1. These images were acquired with a FOV of 80 mm × 30 mm in the zy plane corresponding to a spatial resolution of 2.5 mm × 0.94 mm

Fig. 5.

1D profiles of signal intensity in: (a) the wall-flow filter obtained though numerical integration of the 2D images in the radial (y) direction; (b) FTM obtained though numerical integration of the 2D images in the radial (y) direction

1D profiles of signal intensity in: (a) the wall-flow filter obtained though numerical integration of the 2D images in the radial (y) direction; (b) FTM obtained though numerical integration of the 2D images in the radial (y) direction

Fig. 6.

Relative saturation taken at three radial (y) positions (A, B and C) from the 2D images (Figure 3(a) and Figure 4(a)) plotted over the course of drying for: (a) the wall-flow filter; and (b) FTM. Drying appears to be uniform in the radial direction

Relative saturation taken at three radial (y) positions (A, B and C) from the 2D images (Figure 3(a) and Figure 4(a)) plotted over the course of drying for: (a) the wall-flow filter; and (b) FTM. Drying appears to be uniform in the radial direction

Figures 7 and 8 show the quantitative drying curves and rate of drying for the filter and FTM, respectively. The water content data shown in Figure 7(a) and Figure 8(a) are quantitative and obtained directly from the traditional measurements (RH and temperature measurement) and integration of the signal from NMR spectroscopy; the error bars were calculated based on the instrument sensitivities and the standard deviation of repeat measurements respectively. The rate of drying data shown in Figure 7(b) and Figure 8(b) are obtained from the time derivative of the data in Figure 7(a) and Figure 8(a). It is seen that for both the wall-flow filter and FTM the trends in drying behaviour appear similar; in particular, a slow falling rate period followed by a faster falling rate as drying proceeds. The only significant difference is during the induction period of both samples; the NMR data shows an increasing rate whereas the traditional measurements show a decreasing rate. This is attributed to the temperature dependence of the NMR signal. As the sensitivity of the NMR signal depends on the population difference of nuclear spin energy levels, as described by the Boltzmann distribution, a reduction in temperature causes an increase in the observed NMR signal. During the induction period, a temperature drop of up to 6 K is observed due to the heat of evaporation, resulting in an increase of up to 2% in the observed NMR signal and hence a reduction in the calculated drying rate. While this is negligible for the modest temperature changes during most of the drying process, the temperature drop during the induction period is sharp and the increase in NMR signal decreases the measured drying rate by up to 20%. Thus, from a quantification standpoint we have successfully been able to validate the MRI using simple humidity measurements; however, the MRI is able to give spatial information, as will be shown.

Fig. 7.

(a) Drying; and (b) rate of drying curves for the wall-flow filter as determined from the RH/temperature and MR measurements. An air flow rate of 20 l min–1 was used

(a) Drying; and (b) rate of drying curves for the wall-flow filter as determined from the RH/temperature and MR measurements. An air flow rate of 20 l min–1 was used

Fig. 8.

(a) Drying; and (b) rate of drying curves for the FTM as determined from the RH/temperature and MR measurements. An air flow rate of 20 l min–1 was used

(a) Drying; and (b) rate of drying curves for the FTM as determined from the RH/temperature and MR measurements. An air flow rate of 20 l min–1 was used

For the wall-flow filter and FTM respectively, Figures 9 and 10 show the time evolution of the 1D profiles during the drying process, with 15 mm slices extracted from the data to show the average water saturation time evolution at 20 mm, 40 mm and 60 mm from the front of the substrates. In the case of the filter substrate (Figure 9), it can be seen that drying proceeds uniformly in the axial (z) direction up to a critical point at which a developing drying front is present until the filter is dry. In the initial stages, the rate of drying at the three axial positions is the same until ~20 min, after which the drying front develops and the front of the filter will dry more quickly than the middle and back sections. Between 20 min and 35 min, the middle and back sections of the filter continue to dry at the same rate until the drying front reaches the middle section and begins to dry more quickly than the back section. Finally, the remaining water is removed as the drying front moves through to the back of the filter and is completely dry after 50 min. In contrast, as is seen in Figure 10, for the FTM the drying front propagates through the substrate from the very beginning of the drying process.

Fig. 9.

(a) Time series of 1D axial (z) profiles over the course of the drying of a wall-flow filter; (b) average saturation over three 15 mm slices centred at 20 mm, 40 mm and 60 mm along the length of the sample. Air flow is from the bottom with a flow rate of 20 l min–1. FOV in the z direction was 80 mm and data matrix with 128 points thus corresponding to a spatial resolution of 0.625 mm px–1. The development of the drying front at ~20 min can be clearly seen

(a) Time series of 1D axial (z) profiles over the course of the drying of a wall-flow filter; (b) average saturation over three 15 mm slices centred at 20 mm, 40 mm and 60 mm along the length of the sample. Air flow is from the bottom with a flow rate of 20 l min–1. FOV in the z direction was 80 mm and data matrix with 128 points thus corresponding to a spatial resolution of 0.625 mm px–1. The development of the drying front at ~20 min can be clearly seen

Fig. 10.

(a) Time series of 1D axial (z ) profiles over the course of the drying of a FTM; (b) average saturation over three 15 mm slices at 20 mm, 40 mm and 60 mm along the length of the sample. Air flow is from the bottom with a flow rate of 20 l min–1. The FOV in the z direction was 80 mm and data matrix with 128 points thus corresponding to a spatial resolution of 0.625 mm px–1. It is evident that a drying front is present from the start of drying

(a) Time series of 1D axial (z ) profiles over the course of the drying of a FTM; (b) average saturation over three 15 mm slices at 20 mm, 40 mm and 60 mm along the length of the sample. Air flow is from the bottom with a flow rate of 20 l min–1. The FOV in the z direction was 80 mm and data matrix with 128 points thus corresponding to a spatial resolution of 0.625 mm px–1. It is evident that a drying front is present from the start of drying

The most striking result from these studies is that whilst the spatially-unresolved data (Figures 7 and 8) might suggest similar drying characteristics between the wall-flow filter and FTM substrates, the spatially-resolved measurements obtained from 1D and 2D MR measurements (Figures 6, 9 and 10) reveal significant differences. These data therefore provide important information on the microscopic contributions to the drying process which must be reflected in any computational model. Furthermore, the differences seen imply that care must be taken when drying particulate filters, since water migration and transport could result in a final catalyst location and properties that are different from those expected.

In order to reconcile the predictions of the bulk and spatially resolved measurements, the structure and operation of each sample must be considered. FTMs operate through the axial flow of gas through the channels of the structure, with no flow through the substrate itself. Thus, drying occurs through evaporation at the interface of the channels and the substrate, with water migrating from the saturated pores to the surface through capillary flow. Hence the rate of drying is determined by the humidity gradient in the channels. Air entering the front of the FTM has the lowest humidity, and so the rate of evaporation is greatest at the front and decreases along the length of the filter. As water is removed from the front, the humidity gradient shifts down the channels and the drying front propagates until it reaches the end of the filter as seen in Figure 10(a), causing the slowly decreasing drying rate seen between 5 min and 18 min in Figure 8(b). Once most of the moisture has been removed from each axial position, the residual is controlled by diffusive mass transport in the porous medium and is slower, resulting in the quickly decreasing drying rate after 18 min in Figure 8(b). This is consistent with the findings of Koptyug et al. (3) for a segmented monolith, suggesting that axial capillary flow is not significant due to the larger pore sizes present in this study.

In particulate filters, alternate channels are blocked meaning that gas is forced to flow through the porous substrate in order to exit the filter, adding extra complexity to the drying process. The air first removes water in the largest pores in the substrate, so the air is humidified in the walls and not the channels, so there is no humidity gradient along the length of the sample. These pores form the paths of least resistance between channels, allowing air to flow through the filter walls and become fully saturated. For this air flow rate, the transverse velocity through the walls is expected to be reasonably uniform (36, 49), and so the drying rate in this regime is highly uniform across the filter. From Figure 9, this takes around 20 min, at which point the air becomes less than fully saturated and a humidity gradient can form, starting the drying front that propagates through the wall-flow filter between 20 min and 50 min in Figure 9(a). Figure 7(b) shows no change at 20 min, indicating that despite the change in drying mechanism for the wall-flow filter, the rate-limiting process has not changed and is a similar rate to the FTM shown in Figure 8(b). The rest of the moisture is removed via a similar mechanism to that in FTMs; the transition from evaporation limited to diffusion limited drying can be seen in Figure 9(b) for each axial position.

This additional mechanism in the drying of wall-flow substrates may affect the distribution of catalyst in the monolith following the drying process. The propagation of a drying front in the filter material will create similar issues as have been observed in FTMs by Vergunst et al. (3). However, the effect of the uniform drying regime is uncertain due to the limited number of drying studies on particulate filters. Spatially uniform drying methods such as microwave drying (50) produce more homogeneous distributions of the active phase in the monolith, but as with static and mobile air drying, the mobility of the catalyst and the solvent leads to non-uniform distributions. Limiting or ceasing this mobility, either through fast drying protocols, deposition-precipitation (51) or freeze-drying (3), can improve the homogeneity of catalyst distribution at the expense of economic viability.

Effect of Soot Loading on Gas Fluid Dynamics in a Gasoline Particulate Filter

The velocity profiles measured for the catalyst coated wall-flow filter are shown in Figure 11 for (a) axial and (b) through-wall velocities. Under no soot loading (Protocol I) the axial velocity profile resembles those seen previously (36), with the characteristic U-shaped form of the through-wall velocity profile. After Protocol II of soot loading, the axial profile has only changed slightly, with a more linear change in channel velocities and the cross-over of the inlet and outlet velocities occurring further forward in the filter. A more uniform through-wall velocity is seen, although there is still a parabola-like section in the filter centre. After Protocol III, the axial velocity profile shows a linear decrease and increase in the inlet and outlet channel velocity respectively. The corresponding through-wall velocity profile is much more uniform. Velocity profiles were then extracted from the MRI measurements for the central inlet channel of the sample, and show the gas velocity radially across the cross-section of the single channel to show the evolution of the radial flow profile at different axial positions with increasing soot load (Figure 12): axial positions of z/L = 0.14, 0.33, 0.52, 0.70 and 0.87, referred to as P1, P2, P3, P4 and P5 respectively. The profiles at P1 to P4 are close together for the soot-free sample (Protocol I) but become more spaced out as the soot loading increases. Some profiles show step-like features towards the channel edge, for example P5 in Protocol I, P3 in Protocol II and P1 and P3 in Protocol III. At the highest soot loading (Protocol III), the shape of the flow profiles at P4 and P5 are narrower than the expected paraboloid.

Fig. 11.

MRI measurements (markers) of: (a) the inlet and outlet channel velocities; (b) through-wall velocities for the GPF sample with Protocol I, Protocol II and Protocol III

MRI measurements (markers) of: (a) the inlet and outlet channel velocities; (b) through-wall velocities for the GPF sample with Protocol I, Protocol II and Protocol III

Fig. 12.

Axial velocity flow profiles for the GPF sample after loading: (a) Protocol I; (b) Protocol II; and (c) Protocol III. Profiles are shown at axial positions of z/L = 0.14, 0.33, 0.52, 0.70 and 0.87. Lines are only shown as a guide

Axial velocity flow profiles for the GPF sample after loading: (a) Protocol I; (b) Protocol II; and (c) Protocol III. Profiles are shown at axial positions of z/L = 0.14, 0.33, 0.52, 0.70 and 0.87. Lines are only shown as a guide

Using a numerical 1D model (7) and the MRI data, it is possible to extract information regarding the axial properties of the filter wall in the form of a permeability from Darcy’s law. The permeability calculated is an average value for the combined effect of the filter wall and the particulate. However, because the through-wall velocity is known from the MRI measurements, it is possible to calculate a spatial permeability in the axial dimension (z) of the filter. The permeability profiles are shown in Figure 13. Some variability in the permeability profile for the clean filter is observed, but the permeability is much more uniform after loading Protocol II and especially after Protocol III. The results here show that as the GPF operates, regions of higher through-wall velocity will have a greater loading of soot due to the correspondingly high mass flow of PM; this leads to the ‘self-correction’ effect predicted by Bensaid et al. (52). The axial velocity profiles inside the inlet channels (Figure 12) show changes at the filter rear with increasing soot load. The profiles become ‘narrower’ with lower velocity towards the wall. This is similar to the profiles observed by York et al. (53) also using MRI for high soot loadings in a diesel particulate filter and may be consistent with the development of a soot cake layer in these regions. These regions also correspond to the largest reduction in wall permeability (Figure 13). Two limitations of the MRI method are that it cannot quantify the soot loading and it cannot differentiate between different diameters of soot particle. However, other techniques such as gravimetric analysis or microscopy may allow these to be related to the MRI results in the future.

Fig. 13.

Simulated permeability profiles for the GPF sample for loading: (a) Protocol I; (b) Protocol II; and (c) Protocol III. Lines are only shown as a guide

Simulated permeability profiles for the GPF sample for loading: (a) Protocol I; (b) Protocol II; and (c) Protocol III. Lines are only shown as a guide

Conclusions

A range of MR and traditional techniques, namely RH and temperature measurement, have been applied to study wall-flow filter substrates used for PM vehicle emissions control. Drying of the filter material has been compared with an FTM using 2D RARE imaging to investigate the effect of the structure on the drying kinetics. Since little deviation in the radial drying profile was seen the problem was reduced to 1D in the axial (z) direction. This allowed increased temporal resolution that revealed different drying mechanisms are associated with the wall-flow and FTM substrates, attributable to differences in the physical structures of the two autocatalyst substrates. MRI and velocimetry has also been used to investigate the effect of PM on filter channel fluid dynamics. In the gasoline system studied, any inhomogeneities in the filter wall permeability are ‘self-corrected’ by the particulate loading over the initial hour of filter operation. MRI is a method that can indirectly visualise soot location in the filter, while also providing important information on the flow characteristics and substrate properties. The application of MRI to study filters used in vehicle emissions control has provided new insight into their manufacturing and operation. The greater understanding of the drying process could ultimately result in an improved and more efficient drying process, while great understanding of their operation can lead to improved final product performance, for example higher filtration efficiency.

BACK TO TOP

References

  1. 1.
    Commission Regulation (EU) 459/2012, Official J. Eur. Union, 2012, 55, (L142), 16 LINK http://data.europa.eu/eli/reg/2012/459/oj
  2. 2.
    T. A. Nijhuis, A. E. W. Beers, T. Vergunst, I. Hoek, F. Kapteijn and J. A. Moulijn, Catal. Rev. Sci. Eng., 2001, 43, (4), 345 LINK https://doi.org/10.1081/CR-120001807
  3. 3.
    T. Vergunst, F. Kapteijn and J. A. Moulijn, Appl. Catal. A: Gen., 2001, 213, (2), 179 LINK https://doi.org/10.1016/S0926-860X(00)00896-6
  4. 4.
    A. Wahlberg, L. J. Pettersson, K. Bruce, M. Andersson and K. Jansson, Appl. Catal. B: Environ., 1999, 23, (4), 271 LINK https://doi.org/10.1016/S0926-3373(99)00083-1
  5. 5.
    Z. R. Ismagilov, S. A. Yashnik, A. A. Matveev, I. V. Koptyug and J. A. Moulijn, Catal. Today, 2005, 105, (3–4), 484 LINK https://doi.org/10.1016/j.cattod.2005.06.054
  6. 6.
    “Handbook of Industrial Drying”, 3rd Edn., ed. A. S. Mujumdar, Taylor and Francis Group LLC, 2007, 1280 pp
  7. 7.
    E. J. Bissett, Chem. Eng. Sci., 1984, 39, (7–8), 1233 LINK https://doi.org/10.1016/0009-2509(84)85084-8
  8. 8.
    A. G. Konstandopoulos and J. H. Johnson, SAE Technical Paper 890405, SAE International, Warrendale, USA, 1st February, 1989 LINK https://doi.org/10.4271/890405
  9. 9.
    E. J. Bissett, M. Kostoglou and A. G. Konstandopoulos, Chem. Eng. Sci., 2012, 84, 255 LINK https://doi.org/10.1016/j.ces.2012.08.012
  10. 10.
    A. P. E. York, T. C. Watling, M. Ahmadinejad, D. Bergeal, P. R. Phillips and D. Swallow, SAE Int. J. Fuels Lubr., 2009, 2, (1), 578 LINK https://doi.org/10.4271/2009-01-1266
  11. 11.
    L. F. Gladden, Am. Inst. Chem. Eng. J., 2004, 49, (1), 2 LINK https://doi.org/10.1002/aic.690490102
  12. 12.
    M. P. Hollewand and L. F. Gladden, Magn. Reson. Imaging, 1994, 12, (2), 291 LINK https://doi.org/10.1016/0730-725X(94)91538-5
  13. 13.
    L. Yu Khitrina, I. V. Koptyug, N. A. Pakhomov, R. Z. Sagdeev and V. N. Parmon, J. Phys. Chem. B, 2000, 104, (9), 1966 LINK https://doi.org/10.1021/jp994412
  14. 14.
    I. V. Koptyug, S. I. Kabanikhin, K. T. Iskakov, V. B. Fenelonov, L. Yu Khitrina, R. Z. Sagdeev and V. N. Parmon, Chem. Eng. Sci., 2000, 55, (9), 1559 LINK https://doi.org/10.1016/S0009-2509(99)00404-2
  15. 15.
    I. V. Koptyug, L. Yu Khitrina, V. N. Parmon and R. Z. Sagdeev, Magn. Reson. Imaging, 2001, 19, (3–4), 531 LINK https://doi.org/10.1016/S0730-725X(01)00286-7
  16. 16.
    A. A. Lysova, I. V. Koptyug, R. Z. Sagdeev, V. N. Parmon, J. A. Bergwerff and B. M. Weckhuysen, J. Am. Chem. Soc., 2005, 127, (34), 11916 LINK https://doi.org/10.1021/ja053456v
  17. 17.
    R. Enjilela, P. F. J. Cano-Barrita, A. Komar, A. J. Boyd and B. J. Balcom, Mater. Struct., 2017, 50, (2), 151 LINK https://doi.org/10.1617/s11527-017-1017-7
  18. 18.
    G. Guillot, A. Trokiner, L. Darrasse and H. Saint-Jalmes, J. Phys. D.: Appl. Phys., 1989, 22, (11), 1646 LINK https://doi.org/10.1088/0022-3727/22/11/013
  19. 19.
    K. M. Song, J. Mitchell, H. Jaffel and L. F. Gladden, J. Mater. Sci., 2010, 45, (19), 5282 LINK https://doi.org/10.1007/s10853-010-4572-7
  20. 20.
    M. D. Mantle, Int. J. Pharm., 2011, 417, (1–2), 173 LINK https://doi.org/10.1016/j.ijpharm.2010.11.035
  21. 21.
    Q. Zhang, L. Gladden, P. Avalle and M. Mantle, J. Control. Release, 2011, 156, (3), 345 LINK https://doi.org/10.1016/j.jconrel.2011.08.039
  22. 22.
    J. D. Griffith, A. E. Bayly and M. L. Johns, Chem. Eng. Sci., 2008, 63, (13), 3449 LINK https://doi.org/10.1016/j.ces.2008.03.043
  23. 23.
    J. D. Griffith, A. E. Bayly and M. L. Johns, J. Colloid Interface Sci., 2007, 315, (1), 223 LINK https://doi.org/10.1016/J.JCIS.2007.06.050
  24. 24.
    G. Adiletta, G. Iannone, P. Russo, G. Patimo, S. De Pasquale and M. Di Matteo, Int. J. Food Sci. Technol., 2014, 49, (12), 2602 LINK https://doi.org/10.1111/ijfs.12591
  25. 25.
    L. Manzocco, M. Anese, S. Marzona, N. Innocente, C. Lagazio and M. C. Nicoli, Food Chem., 2013, 141, (3), 2246 LINK https://doi.org/10.1016/j.foodchem.2013.04.068
  26. 26.
    T. Tsuruta, H. Tanigawa and H. Sashi, Dry. Technol., 2015, 33, (15–16), 1830 LINK https://doi.org/10.1080/07373937.2015.1036286
  27. 27.
    M. D. Mantle, N. C. Reis, R. F. Griffiths and L. F. Gladden, Magn. Reson. Imaging, 2003, 21, (3–4), 293 LINK https://doi.org/10.1016/S0730-725X(03)00157-7
  28. 28.
    N. C. Reis, R. F. Griffiths, M. D. Mantle and L. F. Gladden, Int. J. Heat Mass Trans., 2003, 46, (7), 1279 LINK https://doi.org/10.1016/S0017-9310(02)00395-2
  29. 29.
    N. C. Reis, R. F. Griffiths, M. D. Mantle, L. F. Gladden and J. M. Santos, Int. J. Heat Mass Trans., 2006, 49, (5–6), 951 LINK https://doi.org/10.1016/J.IJHEATMASSTRANSFER.2005.09.014
  30. 30.
    G. H. A. van der Heijden, H. P. Huinink, L. Pel and K. Kopinga, Chem. Eng. Sci., 2009, 64, (12), 3010 LINK https://doi.org/10.1016/J.CES.2009.03.012
  31. 31.
    I. V. Koptyug, Prog. Nucl. Magn. Reson. Spectrosc., 2012, 65, 1 LINK https://doi.org/10.1016/j.pnmrs.2011.12.001
  32. 32.
    I. V. Koptyug, S. Altobelli, E. Fukushima, A. V. Matveev and R. Z. Sagdeev, J. Magn. Reson., 2000, 147, (1), 36 LINK https://doi.org/10.1006/jmre.2000.2186
  33. 33.
    I. V. Koptyug, L. Yu Ilyina, A. V. Matveev, R. Z. Sagdeev, V. N. Parmon and S. A. Altobelli, Catal. Today, 2001, 69, (1–4), 385 LINK http://doi.org/10.1016/S0920-5861(01)00396-0
  34. 34.
    I. V. Koptyug, A. V. Matveev and S. A. Altobelli, Appl. Magn. Reson., 2002, 22, (2), 187 LINK https://doi.org/10.1007/BF03166102
  35. 35.
    S. L. Codd and S. A. Altobelli, J. Magn. Reson., 2003, 163, (1), 16 LINK https://doi.org/10.1016/S1090-7807(03)00111-3
  36. 36.
    N. P. Ramskill, A. P. E. York, A. J. Sederman and L. F. Gladden, Chem. Eng. Sci., 2017, 158, 490 LINK https://doi.org/10.1016/j.ces.2016.10.017
  37. 37.
    M. Lustig, D. Donoho and J. M. Pauly, Magn. Reson. Med., 2007, 58, (6), 1182 LINK https://doi.org/10.1002/mrm.21391
  38. 38.
    M. Lustig, ‘Sparse MRI’, PhD thesis, Department of Electrical Engineering, Stanford University, Stanford, USA, 2008, 131 pp
  39. 39.
    D. J. Holland, M. J. Bostock, L. F. Gladden and D. Nietlispach, Angew. Chem. Int. Ed., 2011, 50, (29), 6548 LINK https://doi.org/10.1002/anie.201100440
  40. 40.
    Y. Wu, C. D’Agostino, D. J. Holland and L. F. Gladden, Chem. Commun., 2014, 50, (91), 14137 LINK https://doi.org/10.1039/c4cc06051b
  41. 41.
    D. J. Holland and L. F. Gladden, Angew. Chem. Int. Ed., 2014, 53, (49), 13330 LINK https://doi.org/10.1002/anie.201400535
  42. 42.
    P. T. Callaghan, “Principles of Nuclear Magnetic Resonance Microscopy”, Oxford University Press, Oxford, UK, 1993, 492 pp
  43. 43.
    E. M. Haacke, R. W. Brown, M. R. Thompson and R. Venkatesan, “Magnetic Resonance Imaging – Physical Principles and Sequence Design”, John Wiley and Sons Inc, New York, USA, 1999, 914 pp
  44. 44.
    M. D. Mantle and A. J. Sederman, Prog. Nucl. Magn. Reson. Spectrosc., 2003, 43, (1–2), 3 LINK https://doi.org/10.1016/S0079-6565(03)00005-0
  45. 45.
    A. Caprihan and E. Fukushima, Phys. Rep., 1990, 198, (4), 195 LINK https://doi.org/10.1016/0370-1573(90)90046-5
  46. 46.
    “Perry’s Chemical Engineers’ Handbook”, 8th Edn., eds. R. H. Perry and D. W. Green, McGraw-Hill Education, New York, USA, 2008
  47. 47.
    J. Hennig, A. Nauerth and H. Friedburg, Magn. Reson. Med., 1986, 3, (6), 823 LINK https://doi.org/10.1002/mrm.1910030602
  48. 48.
    M. Schejbal, M. Marek, M. Kubíček and P. Kočí, Chem. Eng. J., 2009, 154, (1–3), 219 LINK https://doi.org/10.1016/j.cej.2009.04.056
  49. 49.
    M. Yu, D. Luss and V. Balakotaiah, Chem. Eng. J., 2013, 226, 68 LINK https://doi.org/10.1016/j.cej.2013.04.026
  50. 50.
    G. Bond, R. B. Moyes, S. D. Pollington and D. A. Whan, Stud. Surf. Sci. Catal., 1993, 75, 1805 LINK https://doi.org/10.1016/S0167-2991(08)64539-5
  51. 51.
    K. P. de Jong, Stud. Surf. Sci. Catal., 1991, 63, 19 LINK https://doi.org/10.1016/S0167-2991(08)64569-3
  52. 52.
    S. Bensaid, D. L. Marchisio and D. Fino, Chem. Eng. Sci., 2010, 65, (1), 357 LINK https://doi.org/10.1016/j.ces.2009.06.051
  53. 53.
    A. P. E. York, T. C. Watling, N. P. Ramskill, L. F. Gladden, A. J. Sederman, A. Tsolakis, J. M. Herreros and I. Lefort, SAE Technical Paper 2015-01-2009, SAE International, Warrendale, USA, 1st September, 2015 LINK https://doi.org/10.4271/2015-01-2009

Acknowledgements

The authors would like to thank the Engineering and Physical Sciences Research Council (EPSRC) for Cooperative Awards in Science and Technology (CASE) to N. P. Ramskill and J. D. Cooper. We would also like to thank Johnson Matthey, UK, for permission to publish.

The Authors


Jonathan Cooper has a Master’s degree in chemistry and recently completed his PhD at the Department of Chemical Engineering, University of Cambridge. His research focused on applying MRI methods to study the gas hydrodynamics in particulate filter systems. He now teaches chemistry at a school in London.


Nicholas Ramskill holds an MEng in Chemical Engineering from the University of Leeds, UK, and a PhD in Chemical Engineering from the University of Cambridge, UK. Nicholas completed his PhD at the Magnetic Resonance Research Centre in collaboration with industrial partner Johnson Matthey where his research focused on MRI studies of diesel particulate filters (DPFs). Subsequently, Nicholas was a Postdoctoral Research Associate in Cambridge where he worked in collaboration with Royal Dutch Shell, The Netherlands, on the development and application of MRI techniques to characterise enhanced oil recovery processes in the laboratory at representative reservoir conditions.


Andy Sederman is a Reader in Magnetic Resonance in Engineering at the University of Cambridge where he also gained his PhD in 1998. His research focus is on developing magnetic resonance techniques for application to engineering and materials. He has worked extensively in the area of velocity and transport measurement and methods to increase the imaging speed to be able to investigate transient systems, both by fast data acquisition and by utilising innovative reconstruction methods allied to data under-sampling. Areas of application for these methods have focused on single and multiphase flows, fluid flow in porous media and reaction and hydrodynamics in multiphase reactors.


Lynn Gladden is Shell Professor of Chemical Engineering at the University of Cambridge in the Department of Chemical Engineering and Biotechnology. She is recognised for her work on advancing magnetic resonance imaging techniques, originally developed for use in the medical environment, and using them in engineering research to gain a greater understanding of the physical and chemical phenomena that determine the performance of chemical processes and their resulting products.


Professor Athanasios Tsolakis has academic and industrial expertise in the field of low carbon energy carriers, environmental catalysts, combustion and pollutant control technologies. He works at the forefront of basic and translational research to improve fuel efficiency and reduce the environmental impact of the transportation and power generation sectors. Prior to his academic appointment at the University of Birmingham in 2005 he worked as a research scientist at Johnson Matthey in the design and characterisation of environmental catalysts for modern aftertreatment systems.


E. Hugh Stitt is a Scientific Consultant at Johnson Matthey, Chilton, UK. He is a Visiting Professor at the University of Birmingham; Fellow of the Institution of Chemical Engineers and a Fellow of the Royal Academy of Engineering. He has 30 years of industrial research experience across a variety of themes related to catalytic reaction engineering and catalyst manufacture with over 100 refereed publications.


Andrew York is a Research Manager responsible for the Emissions Control Reaction Engineering and Modelling group. He joined Johnson Matthey, Sonning Common, in 2000, and has had a variety of roles, including in gasoline and diesel catalyst research, and leading many collaborations with academia on various engineering and catalysis related projects.

Read more from this issue »

BACK TO TOP

SHARE THIS PAGE: