MEASURING PARTICULATES IN HISTORIC BUILDINGS: A COMPARISON OF METHODOLOGIES

Barry Knight

English Heritage

In recent years, there has been an increasing amount of interest in the problem of dust deposition in museums. This has mainly been generated by large building projects, for instance at the V&A (Ford 1997), the British Museum (Kibrya 1999) and the Museum of Scotland (Eremin, Adams & Tate 2000). These surveys have mostly been carried out using the "loss of gloss" method developed by Adams (Adams 1997), a very simple technique which measures the decrease in reflectance of ordinary glass microscope slides after exposure to a dusty environment.

During February and March 2001 I received a 'Sharing Museums Skills' award from the Millennium Commission, which enabled me to spend six weeks working with David Howell at the Textile Conservation Studio in Hampton Court Palace. I carried out a project to compare the "loss of gloss" method with direct measurement of the numbers of particles deposited on glass slides and the surface area covered, using a microscope camera and image analysis software. I also compared this technique with the number of particles captured using black sticky pads, which are used for mounting small objects for examination in the scanning electron microscope.

Glass slides and sticky pads in the Pages' Chamber, Hampton Court Palace

Five glass slides and five sticky pads were put out in each of four locations in Hampton Court Palace, and one slide and one sticky pad were collected each week. For both the slides and the sticky pads, the number of particles and area covered were measured using the microscope camera, and the decrease in reflectance of the slides was subsequently measured by Stuart Adams. The slides were standard 26 x 76 mm glass microscope slides, while the sticky pads were double-sided self-adhesive carbon-coated pads, either 12 or 25 mm diameter. The pads were adhered to 12 or 25 mm aluminium stubs to facilitate handling, and were placed in wooden holders with holes drilled to accept the stems of the stubs.

For measurements on glass slides, a Nikon Labophot with a transmitted light base was used, with a x4 objective. The images were acquired with a Donpisha 3CCD colour vision camera module. For measurements on black sticky pads a Microvision MV120Z microscope camera was used, with a magnification of x240. This has an integral fibre optic illuminator which gives uniform vertical lighting, and is ideal for detecting light coloured particles on a dark background. With both cameras, the image was viewed on a TV monitor so that the area of interested could be identified and focused.

Image analysis was performed using IMAQ Vision Builder for Labview, version 4.0 This is a very versatile program which permits particles to be identified, counted, and their areas measured. It is also capable of discriminating particles on the basis of shape, so that, for example, fibres can be distinguished from other particles, although this was not done in these experiments. The analysis proceeds in several stages that can be individually selected and controlled before being incorporated into a procedure which runs automatically on the click of a button. These stages are:

1. Acquire new image.
2. Calibrate image - a 1 mm graticule was imaged and its length in pixels was recorded. This calibration factor enables measurements to be reported directly in micrometres.
3. Extract colour plane - only the red part of the colour image was used, to improve discrimination of particles from the background.
4. Threshold - only objects falling within a certain range of brightness on the grey scale from 0 (black) to 255 (white) are recorded. After this operation, the image consists of black particles on a white background. Different settings were required for bright field (glass slides) and dark field (sticky pads) measurements. It was quite difficult to choose values which enabled the program to identify all the particles that could be seen on the TV without generating spurious images. This was more difficult for the sticky pads, because it is more difficult to distinguish dark coloured particles on a black background than light coloured particles on a bright background.
5. Particle filter - particles covering fewer than 3 pixels (smaller than 6 µm²) are removed as these are mostly artefacts introduced by the thresholding step.
6. Look-up table: reverse - changes the image from black particles on a white background to white particles on a black background.
7. Image mask - the border of the image is removed as it contains white lines that confuse the counting process. The field of view was approximately 1 mm².
8. Particle analysis - counts the number of particles and measures their areas. The data are transferred to Excel where the total area can be calculated and the particle size distribution plotted.

Schematic layout of equipment

There was found to be considerable variation between readings, particularly when only a small fraction of the surface was covered (eg less than 1%). In order to obtain good statistics, 50 measurements were made for each slide or sticky pad. The glass slides were sampled at 5 mm intervals in a 10 x 5 grid, so as to get good coverage of the whole slide, while the sticky pads were examined more randomly, though trying to avoid overlapping fields as far as possible. It was found necessary to avoid the edges of the sticky pads because of finger marks which confused the image; it was also found that some of the sticky pads had been touched, making the readings unreliable.

The results for both numbers of particles per square millimetre and total area covered show a reasonably good fit to a Gaussian distribution calculated using the means and standard deviations obtained from the data. The fit is not perfect, but nevertheless it is good enough to be able to say that both numbers and areas do follow Gaussian statistics.

We can also look at the particle size distribution: it can be seen that the peak falls at about 50 µm², but the distribution is very skewed to larger sizes. It is worth pointing out that most particles are larger than 10 µm diameter, and in fact we do not measure particles smaller than about 1.5 µm diameter, so the particles which are of greatest significance for human health (< 1 µm diameter) are not significant for soiling, and conversely, the particles which are significant for soiling are too large to be inhaled.

However, if the number of particles is plotted against the logarithm of the particle area, we can see that the data fit a Gaussian curve much more closely: this is a lognormal distribution. Such distributions are encountered in many natural phenomena, and occur when an effect is produced not by the addition of many small random effects, but by their multiplication. In this case, we can imagine that large particles are broken down into smaller ones by many successive impacts, so that while there are many small particles there is still a residue of larger ones (Limpert, Stahel and Abbt 2001).

It can be seen that there is a steady increase in both the number of particles per square millimetre and the fraction of the surface area covered, and that there are clear differences between the different locations in Hampton Court. Only the results using the dust slides are shown, because, as explained earlier, the results from some of the sticky pads were unreliable because of finger marks on the surface. However, the trend of the results is the same.

Consistently more particles are seen on the dust slides than on the sticky pads, and their average size is smaller. As explained above, this is because of the difficulty in discriminating small and dark-coloured particles against a black background. The end result is that the area covered as estimated from the slides is about 20% greater than that estimated from the sticky pads.

The results for the percentage of the area covered may be compared directly with Stuart Adams' "loss of gloss" measurements on the same slides - his results are expressed in soiling units, where 1 SU corresponds to a 1% decrease in reflectance.

When we compare the figures for the percentage of area covered with the percentage decrease in reflectance, we again find a very good correlation, but curiously the number of soiling units is twice as large as the percentage of area covered. It seemed very hard to account for this, but following up a suggestion from Peter Brimblecombe, I looked at the geometry of the two measurement techniques. When the slides are examined with the microscope, the light is incident at 90° so the area of the shadow of the particle is the same as the its area projected onto the plane of the slide. In the reflectometer used to measure the decrease in reflectance, the light is incident at 45°, so the area of the shadow is not the same as the projected area of the particle. Instead, the area of the shadow depends on the shape of the particle. The relationship between the area of the shadow and the projected area of the particle can be worked out for a variety of shapes and orientations, but I will just look at the simplest case of a spherical particle.

The geometry is not quite as simple as you might think, and I hope the situation is clear from the diagram below. The light is coming from the right at 45° to the plane of the slide. Some of the light is blocked by the particle before it is reflected from the slide and some is blocked after it is reflected. Some easy trigonometry shows that while the shadow cast by a spherical particle of radius r also has a radius of r when illuminated vertically, when the particle is illuminated at 45° it casts an elliptical shadow of length (r tan 67.5°) and width 2r. We also need to consider the light blocked directly by the particle, and together with the shadow the total area covered can be approximated by an ellipse of length r(1 + tan 67.5°) and width 2r. The area of this ellipse is p x r x r/2(1 + tan 67.5°), while the area of the circular shadow when the particle is illuminated vertically is pr², so the area of the elliptical shadow is approximately 1.7 times the area of the circular one. Similar calculations can be carried out for other shapes of particle in different orientations. These show, for instance, that for flat plates the area of the shadow when seen at 45° is only 0.707 of the area when seen vertically. The actual measured ratio will depend on the average shape of the particles, and may therefore vary from location to location, depending on the nature of the dust.

Dave Howell kindly carried out some experiments for me by placing some lead shot on a horizontal microscope slide, illuminating them at 45°, and photographing the shadows. It can be seen that the shadows do indeed have the shapes shown in the diagram, but in addition, there is a very clear reflection from the underside of the slide. Because the thickness of the slide is much greater than the diameter of a dust particle, this second shadow will be considerably displaced from the position of the particle. If this shadow is indeed measured by the reflectometer, it would explain why the area covered appears to be twice as large as when measured with vertical illumination.

This is not to cast doubt on the value of the method, however, since it is quick and easy to carry out, and enables comparisons of the rate of dust deposition to be made between different locations. It can also be argued that viewing the slides at 45° is more closely related to the way in which dust is actually perceived in the real historic house environment. As every housewife knows, if you want to know how dusty a polished surface is, you look at it at an angle, not vertically!

I would particularly like to thank the Millennium Commission for giving me a 'Sharing Museum Skills' award so that I could carry out the work presented here, Amber Xavier-Rowe for encouraging me to apply for the award and English Heritage for giving me study leave, and Jenny Band of the Textile Conservation Studio at Hampton Court for agreeing to host me. I am also grateful to Dave Howell, Peter Brimblecombe, Young-Hun Yoon and Stuart Adams for their help and support throughout.

COMPARISON OF TECHNIQUES

	Sticky pads	Dust slides	Reflectance
Angle of incidence	90°	90°	45°
Field of view	Rectangular 1.26 x 0.96 mm	Rectangular 1.19 x 0.91 mm	Elliptical ca. 10 x 14 mm
Area analysed	1.21 mm2 x 50 = 60.5 mm2	1.08 mm2 x 50 = 54 mm2	110 mm2 x 3 = 330 mm2

Sticky Pads	Area covered (µm)2 = 143 x number of particles
Dust slides	Area covered (µm)2 = 94 x number of particles

Number of particles (dust slides) = 1.73 x number of particles (sticky pads)

Area covered (dust slides) = 1.20 x area covered (sticky pads)

% of area covered (dust slides) = 0.5 x Soiling Units

References:

Adams, S., Dust deposition and measurement - a modified approach, Environmental Technology 18 (1997) 345-350.

Eremin, K., Adams, S. and Tate, J., Monitoring of deposited particle levels within the Museum of Scotland: during and after construction, The Conservator 24 (2000) 15-23.

Ford, D., Dust monitoring, Victoria & Albert Museum, Museum Practice 4 (1997) 86-87.

Kibrya, R., Surveying dust levels, Museum Practice 12 (1999) 34-36.

Limpert, E., Stahel, W.A., and Abbt, M., Log-normal distributions across the sciences: keys and clues, BioScience 51 (2001) 341-352.

Barry Knight
English Heritage
23 Savile Row
London W1S 2ET
United Kingdom
E-mail: barry.knight@english-heritage.org.uk