Calibration and Use of Photosensitive Materials for Light Monitoring in Museums

НазваниеCalibration and Use of Photosensitive Materials for Light Monitoring in Museums
Дата конвертации27.10.2012
Размер86.6 Kb.
  1   2
Calibration and Use of Photosensitive Materials for Light Monitoring in Museums


Mauro Bacci, Costanza Cucci, Andrea Azelio Mencaglia, Anna Grazia Mignani and Simone Porcinai

Methodological aspects of the use of photosensitive materials as light dosimeters in museum monitoring are investigated. As a case study, a spectroscopic investigation in the 400-700 nm range is developed for Blue Wool Standard No. i (BWS1), a colorimetric indicator traditionally used by conservators to obtain a qualitative indication if the risk associated with lighting. Experiments using both artificial and natural light-aging were performed on BWS1 samples and a set of reflectance spectra was non-invasively collected during the aging process at increasing light-doses. A portable device based on fibre-optics technology, operating in real time, was used for the spectral acquisition. The problem of calibrating the BWS1 response to the light was investigated using the data-set of artificially light-aged samples, and different calibration models were compared. The calibration curves were then used to estimate the alteration, due to the possible synergistic action of light and other factors, which occurred in BWS1 samples naturally aged by exposure in situ. A method to account semi-quantitativcly for the cumulative action of light and other environmental agents is proposed, by introducing the concept of 'equivalent light-dose', in terms of which the overall risk factor can be evaluated.


The environmental control of museums and art galleries is one of the most important subjects which curators and conservators are faced with [1-3]. Among the various environmental factors that are potentially hazardous for works of art, light is certainly one of the most dangerous [4]. On the other hand, adequate lighting is of the utmost importance for guaranteeing the best enjoyment of works of art. Consequently, a continuous com­promise between care of the objects and display requirements has to be achieved in setting up exhibition rooms. The guidelines for a correct policy with regard to this topic of conservation are becoming more and more precise [5]: they are beginning to take into account a wide variety of considerations related to light deterioration, such as the different lightfastness of materials, various photochemical mechanisms, depend­ence on the radiation range and hence on the types of light source, etc.

Within this context, a general recommendation, which holds irrespective of the type of collection, suggests minimizing the total exposure of objects to light, assuming the validity of the 'law of reciprocity'. This law states that 'the net photochemical effect is the result of the total exposure which an object receives, which is the product of the light-intensity multiplied by the time' [6]. Following this principle, new criteria in light monitoring have been proposed, based on the use of photosensitive materials as indicators of the cumulative exposure, or light-dose, received. Although such an approach cannot constitute an alternative to conventional instrumentation in measuring light levels, it nevertheless appears very attractive to the conservation community, since it focuses on revealing the cumulative damage that is induced by light together with other environmental factors (temperature, relative humidity, pollutants, etc.). Indeed, it is by now widely recognized that several photo-induced deterioration mechanisms are enhanced, or accelerated, by different physico-chemical factors, and such a co-operative action cannot be correctly evaluated by a separate measurement of each environmental parameter.

Hence, a new generation of sensors is at present under investigation. Such sensors are passive; they have to be placed close to the artworks during a defined period of time. After this time, alterations due to the environmental impact can be detected, e.g., with spectroscopic measurements, and then compared with those that have occurred in 'twin' specimens, artificially aged under known lighting conditions. Different types of these non-traditional light sensors have been developed. For example, sensors designed to monitor the different categories of light sensitivity into which artworks are divided have been proposed [7, 8]. They can be prepared with different combinations of dyes in a polymeric matrix, so that active layers with the appropriate lightfastness can be obtained [9]. Alter­natively, different types of sensors, aimed at investigating the photochemical reactions involved in the alteration of polychrome surfaces, have been proposed. These art-prepared following specific recipes, in order to reproduce typical artists' materials [10].

However, whichever type of sensor is selected, a characterization of its response to light — that is, a calibration — is a fundamental requirement for using it as a 'light dosimeter'. Moreover, in order to obtain a quantitative indication of the exposure received, the calibration curve must be constructed by providing a suitably dense set of data points in the light-dose range of practical use for that sensor.

In this work, possible calibration procedures and methodological aspects, as well as practical questions related to applications in situ, are investigated in order to explore the practical applicability of this new class of light sensor. Therefore, a pilot study has been carried out by adopting the Blue Wool Standard (BWS) as test material. The BWS is a well-known photosensitive material, introduced in conservation in the 1950s [11] and now commercially available as a standard. According to the International Organization for Standardization (ISO), it consists of a series of eight blue-dyed wool cloths that present increasing lightfastness and fade at different rates when exposed to light. Therefore, the BWS can be viewed as a precursor of this new generation of light sensors: it is generally used to achieve a qualitative indication of the light exposure received by an object or, alternatively, to classify its lightfastness level. This is done by a visual evaluation of the change of colour that occurs in the BWS sample. As for the quantitative applications of BWS used as a colorimetric indicator, several studies are available in the literature [12-15]. Nevertheless, they are mainly devoted to exploring the BWS colour changes over very large light-dose ranges (up to several Mlux.h) to attain a quite genera] characterization, which is useful for monitoring long-term exposures or stable artifacts.

The present study focuses instead on exploring the behaviour of the most sensitive BWS, identified as No. 1 (BWS1), in the low light-dose region (0-500 klux.h) where, to the best of our knowledge, there is a lack of detailed information. Taking into account the recommended light levels [1, 16], the above range can be considered of primary interest in applications where photosensitive materials are used as tools for early warning of potentially unsafe situations.

Another goal of the analysis is to explore to what extent the BWS, as well as the other light sensors based on photosensitive materials, could be used to evaluate the influence of the micro-environment in its own fading process. To this end, the characterization of BWS1 has been extended beyond the colorimetric approach, by considering the spectral features of the material and their evolution induced by successive light-aging. The calibration problem has been examined by testing different approaches, in order to provide some guidelines on their applicability to other light sensors. Three different calibration curves were constructed for BWS1, using the spectral data acquired on artificially-aged samples. Subsequendy, these curves were used to estimate the damage that occurred on analogous samples exposed for two months under typical display conditions in a museum gallery.

A point carefully examined concerns the inter­pretation of the results obtained by the exposures in situ. The samples of BWS1 exposed in situ recorded the integrated damaging effect of light and the total microclimate. In order to give a semi-quantitative indication of the amount of this cumulative damage, the concept of 'equivalent light-dose' (ELD) is introduced. The ELD can be defined as the light-dose which is capable of producing, in a given material in an uncontrolled environment, the same spectral variation as that measured in the same material exposed under well-defined and controlled environmental conditions. Actually, for example, the colour change measurements can only prove that some damage has occurred; they cannot distinguish the contribution of light from that of other environmental factors.

At this early stage of the research, the ELD values obtained for BWS1 exposed in situ have been compared with the actual light-doses, measured with traditional instrumentation. In this way, the ELD has been interpreted as an indicator of the aggressiveness of the specific micro-environment which causes an increase in fading of the test material.



The experiments involved two steps:

• Accelerated light-aging of the BWS1 samples, performed in the laboratory and aimed at attaining a spectral data-set for calibration.

In situ exposure aimed at producing a set of naturally-aged samples, to be evaluated in terms of ELD on the basis of the calibration curve constructed with the artificial aging data.

The instrumentation used to collect spectral data for both steps of the experiment was a miniaturized fibre-optic spectrophotometer (see Figure 1), specifically designed for museum monitoring [17]. This device makes it possible to follow the spectral evolution of a specimen during a prolonged period, by means of an automatic mechanism of periodic spectral acquisition. The holder in which the specimens are housed is a wheel. One half of the wheel surface is exposed to the external environment, and the other half faces the fibre-optical acquisition system within a darkened case. Although completely obscured, the case is designed so that a continuous air-exchange with the external

Figure 1 Experimental set-up scheme. The miniaturized, fibre-optics-based spectrophotometer is contained in the case on which the wheel sample-holder is mounted. By means of a periodic rotation of the wheel, the BWS samples are carried into the box, where their reflectance spectrum is acquired. A light source placed at a fixed distance is used to perform progressive artificial light-aging. For the in situ natural aging, the apparatus is left in the museum, exposed to environmental light. The instrumentation is connected to a portable PC, which allows the experimental settings to be controlled in real time.

environment is permitted through the bottom of the box, which is not airtight.

The samples mounted on the higher part of the disc remain exposed to the light and the surrounding environment during almost the entire exposure time. Through the periodic rotation of the wheel, these samples are temporarily carried to the lower position in the box, in front of a fibre-optic probe, for the acquisition of their reflectance spectrum. The spectral acquisition takes a few seconds, after which the samples are returned to their original position, to be exposed again to the light. Furthermore, a reference sample, placed in the lower half of the wheel, is maintained in the dark and measured at the same intervals as the exposed samples, in order to record the possible spectral variation induced by factors other than light. The box guarantees darkness but, since it is not completely closed, the microclimate conditions inside can be assumed to be the same as those outside. A Spectralon® sample is also housed on the wheel and is used as reference white for calculating the spectral reflectance. The reference white reflectance is recorded after each single acquisition on the samples, so that the calibration of the signal is assured throughout long-term experi­ments.

During the experiment, the instrumentation was also equipped with a pair of dataloggers (IrLog, Elsec), compact devices which can record light (visible and UV), temperature and relative humidity at specified time intervals. The dataloggers were placed beside the wheel, close to the exposed samples, in order to monitor the light conditions during the experiment. Since the illumination field on the wheel surface may not be homogeneous, especially in in situ exposure, two dataloggers were used in different positions, in order to have a mean value that would effectively represent the light flux impinging on the sample area.


All the samples employed, each about 2 cnr in size, were obtained from the same BWS1 card, so that any possible error due to variation among the BWS1 batches was avoided. In each sample, the area investigated by means of the fibre-optic probe is a spot about 5 mm in diameter. For each exposure run — the artificial and the natural light-aging — three specimens of BWS 1 were used: two identical samples, housed in the upper half of the wheel, for the light exposure, and one reference sample, used as a control, housed in the opposite half of the wheel and maintained in the dark.

Exposure conditions and spectral measurements

The artificial light-aging was performed using a quartz tungsten halogen lamp (150 watt) as the source, placed at a distance of 1 m from the holder. This set-up guaranteed a stable luminous flux on the samples of about 7000 lux (the actual value and the stability of the light source were monitored with the datalogger during the entire exposure period), without heating effects due to the lamp. The exposure was performed during a period of about 70 hours, at a spectral acquisition rate of 10 minutes, in order to obtain frequent sampling in the 0-500 klux.h range of the total light-dose, which is the region of interest for practical applications in the conservation field. Two interruptions in the light exposure, lasting 67 hours and seven hours respectively, were effected during the experiment, in order to test the behaviour of the BWSl in the sequence 'dark after light', since phenomena such as colour regeneration or spectral recovery could occur [8j.

The artificial light-aging process was performed under standard environmental conditions, in order to simulate typical exposure conditions (see Appendix). At the same time, the experiment was performed in an accelerated way, in order to render negligible all the aging effects arising from factors other than light, such as temperature or relative humidity. Even though performed under room conditions, the experiment was completed within a reduced period of time, during which typical microclimate flucatations could safely be assumed to be insignificant. On the other hand, a further compression of the exposure time combined with the use of a higher intensity light source could have resulted in a possible failure of the law of reciprocity, which is the case when extreme lighting conditions are adopted [8, 18].

Spectral characterization was performed in the 400-700 nm range, using 10 nm steps. As a result of the accelerated light-aging experiment, a data-set of 268 reflectance spectra for each BWSl sample (two light-aged and one reference) was obtained.

The natural aging of BWSl was attained by placing the instrumentation, equipped with the BWSl samples, in the Geographical Maps Room of the Uffizi Gallery in Florence (see Appendix). The reflectance spectra, measured between 400 and 700 nm in 10 nm steps, were collected every two hours during the period from 3 June to 3 August 2002, employing the same instrumentation as that used for artificial aging. Two BWSl samples were exposed to natural environmental conditions (light plus other factors), while a third BWSl, housed in the lower part of the holder wheel, was

maintained in the dark as a control for factors other than light. As described above, simultaneous monitoring of the overall environmental conditions was performed by means of the two dataloggers placed beside the BWSl samples: light (visible and UV), temperature and relative humidity were recorded with a sampling interval of 10 minutes during the entire time of the exposure. A data-set of 733 reflectance spectra for each BWS 1 sample resulted from this in situ measurement campaign.


Spectral analysis

The typical reflectance spectrum of BWSl and its variations, induced by exposure to light, were investigated in the visible range (400—700 run). The goal was to describe the photo-induced spectral evolution in terms of a suitable parameter, P, which could be expressed as a function of the variable light-dose D, i.e.. the product of the illuminance for the duration of exposure, measured in lux.h, and corresponding to the exposure. Hence, a calibration curve, P = P(D), could be attained and used to deduce the light-dose received by BWSl samples exposed under unknown or un­controlled conditions.

Results of artificial light-aging

The set of reflectance spectra acquired on artificially aged BWSl shows a progressive variation as the light-dose is increased (Figure 2). The results from light-aging display an intensity variation across the visible region, except at two points where the spectra remain almost unaffected. Because of the presence of these two isosbestic points, located at about 510 nm and at 650 nm, three well-defined regions, characterized by typical spectral variations, are identifiable: the A1 region, which spans the range below 510 nm; the central region A2, between the two nodes; and a third region, A3, above 650 nm. Region A2, where each spectrum shows a reflectance minimum, corresponds to the well-known absorption band of BWSl. Nevertheless, the photo-sensitivity of BWSl is appreciable over the entire visible range. Indeed, as can be seen in Figure 2, the spectral variations in the A1 and A3 regions are comparable to, or higher than, those characterizing the main absorption band in A2.

In order to examine these features thoroughly, a normalization was adopted that emphasizes the spectral variations induced by the aging process. Thus, the

Figure 2 Artificial light-aging: BWS1 reflectance spectra collected during the entire exposure. The arrows indicate the direction of the spectral evolution induced by increasing the light-dose.

spectra were processed by calculating the 'relative reflectance' (RR) with respect to the initial reflectance, that is, the ratio between the reflectance spectra of the aged and the imaged BWS1 samples. Then, the RR spectra were obtained by normalizing each reflectance spectrum to the one collected at time zero on the same sample.

In Figure 3, the RR spectra for both light-exposed and unexposed BWSl samples are reported. The two sets of spectra, relative to the complete exposure, are sequenced temporally. It can also be observed that the temporal sequence shows a gap, which corresponds to the first period of 67 hours when the lamp was switched off.

Figure 3 Artificial light-aging: BWSl relative reflectance (RR) spectra of the exposed sample (grey curves) and the unexposed reference sample (black curves). The arrows indicate the direction of the spectral evolution due to the progressive stages of exposure.

Opposite trends in the temporal evolution char­acterize the different spectral regions. The A2 region, located between the two isosbestic points, is char­acterized by an increase in reflectance with respect to the imaged BWSl spectrum, while in the Al and A3 regions, reflectance decreases. This fact indicates that the effects of aging involve the whole spectrum but, depending on the spectral region, different mechanisms, probably due to absorption and scattering effects, seem to prevail in the spectral alterations. This aspect has to be taken into account in processing the spectra for calibration purposes.

In order to investigate the different role of light with respect to other possible agents of aging, the spectra collected on the reference sample, that is, on the BWSl kept in the dark during the exposure, have been analysed. In Figure 3 the two sets of spectra, relative to the exposed and unexposed BWSl, are reported together. It is evident that the spectral variations in the unexposed BWSl sample are much lower than those in the exposed samples. Nevertheless, even the RR data-set of unexposed BWSl shows a temporal progression, analogous to that of the exposed samples. This fact suggests that some aging process acts even on the unexposed BWSl, and that environmental factors other than light can interact with the BWSl dye, causing spectral changes similar to the photo-induced ones.

Moreover, as can be observed in Figure 3, the highest variation for the unexposed sample involves the spectral region A3, which seems to be the most sensitive to the action of factors other than light. This different sensitivity to light for each spectral region has been investigated by comparing the spectral evolution IN A1, A2 and A3. To this end, the behaviour of the stationary points, i.e., the maximum or minimum intensity in the RR spectrum, has been reported against the light-dose for the A2, A1 and A3 region in Figures 4. 5 and 6, respectively. For comparison purposes,1 the same curve has been plotted also for the unexposed BWS1.

It can be observed that in both the Al and the A2 regions, at the end of the experiment, the global variation for the unexposed BWSl represents about 5% of the exposed one, while in the A3 region it reaches about 20%. This confirms that in the A3 interval the

1It should be noted that the independent variable in the plot relative to the unexposed sample is the time at which the measurement was done, and no longer the light-dose. However, since the artificial aging experiment was performed using a stable source of light, these quantities can be considered equivalent apart from a constant.

Figure 4 Artificial light-aging: behaviour of the relative reflectance (RR) maximum intensity, in the A2 spectral region, for the exposed (*) and unexposed (•) BWS1 sample, for the progressive stages of exposure.

BWS1 is particularly sensitive to factors other than light, such as, for example, thermo-hygrometric conditions.

Let us consider now the dependence on the light-dose of the exposed samples. It can be seen that, in the A2 region, the RR intensity maximum follows an almost linear behaviour, while in the other two regions (Al and A3) the curves show non-linear behaviour, characterized by a decreasing slope as the light-dose increases. Moreover, among the three graphs, only the curve relative to the A3 region (Figure 6) clearly shows two discontinuities, which correspond to the two dark periods during the experiment. It should be noted that some recovery effect influences the spectral evolution in

Figure 6 Artificial light-aging: behaviour of the relative reflectance (RR) minimum intensity, in the A3 spectral region, for the exposed (*) and unexposed (■) BWS1 sample, for the progressive stages of exposure.

Figure 5 Artificial light-aging: behaviour of the relative reflectance (RR) minimum intensity, in the A1 spectral region, for the exposed (*) and unexposed (■) BWS1 sample, for the progressive stages of exposure.

the A3 interval, since, during each dark period (67 and seven hours), the RR minimum is found to return to higher values. This fact, which indicates the presence of chemical reactions that are not light-dependent, should he carefully considered in constructing a calibration curve of the BWS1 response to light.

Summarizing, the active material constituting BWS1 shows, above 650 nm, a spectral variation induced by (unspecified) factors other than light and. therefore, the inclusion of the A3 region in the calibration analysis can affect the accuracy of the final result.

On the basis of the previous analysis, the A2 region appears to be the most suitable for calibrating BWS1 as a light sensor, since this region shows the highest sensitivity to the light. Moreover, unlike the A1 region, the A2 region of BWS1 shows a uniform response to the light-dose received (almost constant derivative in the plot in Figure 4), which makes it possible to represent the spectral alterations by adopting a simpler model.


From a quite general point of view, by means of non-invasive spectral measurements it is possible to characterize the different aging stages of a given photosensitive material treated with increasing light-doses. In this way a calibration scale is obtained that is usable for assessing the light-dose received by other specimens exposed under unknown conditions.

The calibration scale consists of a curve P = P(D) that expresses the spectral dependence on the light-dose, and that can be inverted to determine the received dose as D = D-1(P). In constructing this scale, different approaches can be followed, depending on the choice of spectral parameter P. P can be either a local parameter, related for example to the absorption band, or a global parameter, related to the whole variation, such as those obtained by principal component analysis (PCA) |19]. In general, no universal criterion exists for determining the most suitable choice, which depends on many factors, such as the complexity of the spectral shape, the number of variables which can affect the spectrum, the kind of law which connects P to them, etc.

The calibration problem for BWS1 was investigated by following different approaches. Thereafter, a critical comparison between the resultant calibration scales was made in an attempt to identify the main methodological aspects.

Three calibration curves were obtained. The first curve was obtained by considering an integral on a suitable spectral interval, the A2 region, in which the material exhibits a high light-sensitivity. The second calibration curve was obtained by applying the PCA, and the third by means of colorimetric analysis.

Before presenting the results obtained from the different calibrations, a general problem must be mentioned. This concerns the question of comparability between sets of spectra collected on different specimens. Usually, the data-set relative to samples exposed in situ has to be projected on the calibration curve, which is obtained by using data collected on different samples in laboratory tests. Therefore, the correctness of the calibration should be checked by considering the possible sources of variability in the experimental conditions (such as change of specimens, spectral acquisition procedure, etc.).

In the present study, it was recognized that even a slight variability in the initial experimental conditions could affect the absolute values of the spectral intensity, and then the raw reflectance data could not be directly compared without a suitable normalization. As an example, in Figure 7 the Kubelka-Munk (KM) peak is reported versus the light-dose for the two 'twin' data-sets, collected simultaneously on two BWS1 samples exposed under the same nominal lighting conditions. The two curves do not overlap or cross each other and their displacement is not significantly varied over the range investigated. Therefore, the variability is mainly related to some intrinsic difference in the initial reflectance measured, rather than to some difference in the fading process on the samples.

The difference in the intensities has been considered as an estimate of the uncertainty in the final light-dose

Figure 7 Artificial light-aging: comparison between the data-sets collected on the two exposed BWS1 samples. Behaviour of the Kubelka-Munk peaks in relation to the light-dose.

values in the calibration curves. As regards the possible reasons for such variability, various aspects should be considered. The first is the non-homogeneity of the illumination field: it follows that displaced samples, differently lighted, show a different spectral response. Actually, this is an intrinsic problem which is also relevant to artworks on display, where lighting conditions are spatially variable and even affected by outdoor environmental conditions.

Another point that has to be considered concerns the unavoidable, even if slight, differences between the surfaces of the two samples, that can be more or less stretched when fixed in their support. This point becomes particularly important when the samples consist of textiles, like BWS.

All these factors are intrinsic to the experimental procedure adopted and they give rise to some difference in the initial reflectance of nominally identical samples. Next, a suitable normalization has to be adopted to make possible a quantitative comparison among different spectral data-sets. In the case under examination, the RR spectra were calculated, since this normalization eliminates the discrepancy in the initial reflectance spectra. Moreover, the RR spectra demonstrate light-induced spectral evolution, rather than spectral features related to the immediate conditions of the sample.

On the basis of the previous analysis, the simplest way of constructing the calibration scale of BWS1 appeared to be to integrate the RR spectra over the central region A2, which corresponds to the absorption band. As already mentioned, variations over the whole central region can be attributed almost exclusively to the effect of lighting. The integral, rather than the peak intensity, was selected as the calibration parameter, since it is a quantity less dependent on the instrumentation. Next, the integral of the RR between 510 and 650 nm, referred to as I2, was evaluated for the two samples of BWS1 exposed to the light. The two 12 data-sets were averaged and fitted by using the following law:

where D is the light-dose and a and b are constants (Figure S). Such a curve provided an excellent fit of the experimental data over the range examined (R = 0.999). The value obtained with the best fit routine for the exponent b is very close to unity (b = 0.871 ± 0.004), indicating a pseudo-linear behaviour of the experimental data-set in the light-dose range investigated. As can be seen in Figure 8, the non-linearity mainly concerns the very low-dose range (below approximately 50 klux.h), where the BWS can exhibit a noisier response.

The second approach considered for calibration was a multivariate method, based on PCA. This approach can be considered a more general way to address the calibration problem, and makes it possible to extract a few meaningful parameters that summarize the spectral changes over the range of wavelengths examined (400-700 nm). Again, PCA was applied to the normalized spectra (RR data). For the data-set considered, it was found that the first two principal components covered 99.9% of the cumulative variance. This means that all the spectral variations in the data-set are represented almost entirely by the two variables PC1 (98.1%) and PC2

Figure 8 12-based calibration curve for the artificially light-aged BWS1, obtained by integrating the RR spectra over the A2 region. Experimental data (O); non-linear (power law) best fit (—); linear fit (...)

(1.9%). In Figure 9, the loadings of the PCs are shown. As expected, the most important contribution to PC1 arises from the central spectral region corresponding to the absorption band, whereas major contribution to PC2 conies from the A3 region (above 650 nm).

The hypothesis that the A3 region may be not suitable for calibrating BWS1, owing to its higher sensitivity to environmental factors other than light, was re-investigated in the framework of the PCA approach. To this end, the correlation between the PCs and the light-dose was analysed. The values obtained for the correlation coefficients r indicated that the PC1 scores are highly correlated with the light-dose (r = 0.99), whereas the PC2 showed a very poor con-elation (r = 0.11). As a consequence of this, a more suitable spectral range was selected for the calibration, which was developed by considering the spectra on a limited interval (400-640 nm), from which the A3 region was excluded. The data-set was reprocessed with PCA in the restricted range and, in this case, almost all the variance that resulted could be represented by one variable — the first principal component (representing 99.3% of the total variance). This indicates that all the variables (wavelengths) are highly correlated to each other and that their behaviour can be represented by only one parameter (PC1). The PC1 was then adopted for constructing a different calibration curve of the BWS1, as reported in Figure 10.

For the sake of completeness, it was appropriate to end this study of possible approaches to calibration by evaluating also the colour variation of BWS1.

Figure 9 Spectral distribution of the PC1 (■) and PC2 (O) loadings, calculated for the RR data-set of the artificially light-aged BWS1. For comparison purposes, the plot of a RR spectrum (—), relative to the 400 klux.h light-aged stage, is superimposed.

Figure 10 PCA-based calibration curve for the artificially light-aged BWS1. Plot of the PC1 score in relation to the light-dose.

As already mentioned, the typical usage of the BWS in the conservation field is based on a visual assessment of the exposed sample, which is compared with a reference grey-scale. Within this framework, an obvious approach for quantifying fading is to adopt a colorimetric analysis. Indeed, a widely used parameter for char­acterizing the BWS response to different lighting conditions is the colour change ΔE* [ 14. 20], as defined in the standard CIE colorimetric space [21]. In other words, the variable ΔE* can be considered as a possible parameter for a calibration curve, more readily related to the phenomenology of the fading.

As shown in Figure 11. the colour change ΔE* has been calculated, using the CIELAB76 Colour System, for the BWS1 spectral data and it was plotted against the light-dose. The experimental data have been fitted by using the following curve:

where D is the light-dose. This law shows the expected asymptotic behaviour, i.e., a ΔE* saturation within the limit of high light-dose values. The curve fit is also reported in Figure 11.

Assessment of damage on naturally-aged samples

The three procedures for treating spectral data presented above lead independently to a calibration curve, P = P(D), that can be used to assess the light-dose D received by samples exposed in situ. Therefore the data-set collected on the BWS1 exposed in the Uffizi Gallery was projected on the three calibration curves.

Figure 11 Calibration curve based on the colour change (ΔE) for the artificially light-aged BWS1. Experimental data (O) and best-fit curve (—) (see equation in the text). The error bars take into account the spectral differences between the two BWS1 samples exposed together under the same aging conditions. The colour variation (ΔE) has been calculated, for each stage of progressive aging, with respect to the unaged BWS1 (measured before starting the exposure), according to the CIELAB76 Colour System.

First, it was observed that the spectral features and the temporal evolution of the naturally aged samples were similar to those obtained in the laboratory tests, where the samples were aged only by light. In other words, the spectral alterations which occurred during in situ exposure appeared to be analogous to those obtained in the laboratory experiment.

Therefore, it makes sense to use the calibration curves to estimate the damage induced on the BWS1 by the overall effect of light and microclimate conditions. The overall exposure of the samples exposed in the museum has been expressed in terms of 'equivalent light-dose' (ELD). The ELD for the two BWS Is exposed in the Uffizi Gallery was assessed by using all the three different calibration curves.

In Figure 12 the ELD values predicted are reported as a function of time and compared to the actual values of the light-dose independently measured by means of the dataloggers placed close to the samples. First of all, it should be noted that all three prediction curves lie above the one reporting the measured light-doses. Hence, whichever calibration curve is adopted, the estimated dose significantly exceeds the value actually measured. This fact indicates that, beside light, some other factor contributes to the BWS1 fading and such cumulative effect can be expressed by the ELD value.

The second aspect to be discussed concerns the comparison between the different models and the selection of the most suitable one to treat the system.

Figure 12 Natural light-aging in the Uffizi Gallery (Florence): equivalent light-doses calculated from the spectra acquired in situ on BWSl exposed from 3 June to 3 August 2003. The ELD values calculated using different calibration curves are compared: 12-based model (O), PCA-based model (0) and ΔE-based model (D). For comparison, the actual light-dose (x), measured in real time by the datalogger, is also reported. The error bars were obtained by considering the difference between the two data-sets of the 'twin' BWS1 samples simultaneously exposed to the light.

Indeed, as can be seen in Figure 12, the ELDs predicted by the colorimctric calibration do not coincide with those obtained by the other models. As can be observed, the differences between the ELD values predicted by the 12 and the PCA models fall within the error bars and, therefore, these two approaches can be considered to be reciprocally validated and appear equivalent for practical applications. It should also be remarked here that a PCA-based model would be more appropriate for handling highly structured (multi-band) spectral data, where the light-induced variations involve several absorption bands. This can be the case, for example, if a mixture of different dyes is used to build up the photosensitive layer of a light sensor (8, 9]. In the present study, the multivariate analysis was considered merely for the sake of generality, as a possible methodology to be extended to systems more complex than the BWS.

As for the calibration curve based on ΔE*, as observed, it provides lower ELD values with respect to the other two curves although, even in this case, the predicted ELD is greater than the actual measured light-dose. Such a discrepancy can be interpreted by considering the definition of the parameter ΔE* and its relationship with the reflectance spectra of the samples. According to the mathematical definition:

the colour variation in the CIELAB76 System is expressed in terms of the co-ordinates L*, a* and b*, which are related to the lightness (L*), the green-red (a*) and the blue-yellow (b*) chromatic stimuli, respectively. The spectral data are contained in a very complex way withm the L*, a*, b* values, which are defined in terms of the tristimulus functions X, Y and Z. In particular, the definition of X, Y and Z involves an integral of the product of the reflectance spectrum, R(λ), the colour matching functions, x{λ), y(λ), z(λ) and the spectral power distribution of the illuminant, P(λ) [21]. As a con­sequence of this involved relationship, the original spectral data are somehow averaged within the colonmetnc co-ordinates, due to the convolution of R(λ) with other functions that are included to take into account the response of the human eye. The better to clarify this point, in Figure 13 the relative reflectance of the artificially aged BWSl has been plotted together with the colour matching functions, x{λ), y(λ), z(λ). Since the colorimetric co-ordinates are obtained by multi­plying the whole reflectance spectrum by the colour matching functions, the various spectral regions are differently weighted. Moreover, in Figure 2, it can be seen that the light-induced spectral variation in the BWSl is characterized by opposite trends in the 400-500 nm and in the 500-650 nm regions, where the absorbance of the material increases and decreases, respectively. Consequently, a compensating effect, or an attenuation of the actual overall variation, is expected to occur in the final colorimetric values, when the product of the spectrum R(λ) by x{λ), y(λ) or z(λ) is calculated. In other words, the colorimetric co-ordinates, which are suitable for expressing quantitatively the human visual

Figure 13 Plots of the x(λ), y(λ) and z(λ) colour matching functions. For comparison purposes, the typical RR data-set of progressively light-aged BWS1 has been superimposed on the plots.

response to a certain colorimetric stimulus, average out the actual spectral response of the material to light-aging. As a consequence, the calibration curve based on the ΔE* parameter is likely to be affected by an underestimate of the actual spectral variation induced by the light.

To summarize, it can be stated that, among the calibration models considered, the one based on AE* appears less suitable for assessing the cumulative damage in terms of ELD. The ELD is, indeed, conceived to be related as closely as possible to the overall spectral variation occurring in a sample.

Instead, the A2- and PCA-based models agree with each other, and appear to be a tool for semi-quantitative predictions about the overall environmental action in accelerating BWSl fading. In Figure 12 it can be seen that, according to both the A2 and the PCA curves, the damage occurring on the BWS1 sensors after an exposure of 15 days was equivalent to a light exposure of about 100 klux.h, while the datalogger, after the same time, recorded a dose of about 60 klux.h. This result can be attributed to the action of the microclimate, which amplified the light effects on the BWSl exposed in the Geographic Maps Room; as a consequence, the ELD revealed by the BWSl sensor was found to be more than one and a half times the actual measured light-dose. To explore this hypothesis better, let us now consider the effect of all microclimatic parameters, other than light, on the non-exposed BWSl sensor, which was main­tained in the dark throughout the experiment. Since this non-exposed sample showed spectral change, an assessment of the ELD based on the calibration curve could be made. In Figure 14, the behaviour of the ELD estimated for the non-exposed sample is displayed, with the calculated linear regression superimposed. As expected, the spectral variation found in the BWSl reference sensor is very small and therefore the ELD data are widely scattered. Nevertheless, an increasing trend with the time of exposure can be seen, and the total ELD, which was measured on the BWSl reference after 60 days, amounts to about 15 klux.h. This value can be explained only in terms of aging due to micro-environmental parameters other than light. One could object that such an ELD value could be due to the partial exposure to light which the reference sample undergoes at every rotation of the wheel sample-holder. This hypothesis can be excluded on the basis of a simple calculation, which provides a rough estimate of the total light-dose received, owing to the periodic rotation, during the in situ experiment. Considering one rotation every two hours for two months, with a permanence of approximately 10 seconds of the BWSl sample at the

Figure 14 Equivalent light-dose values calculated from the non-exposed BWS1 sample, using the I2 model. The experimental data are plotted together with the linear best fit (—).

maximum daylight value recorded (see Appendix), the light-dose obtained is:

(10 seconds) x (1400 lux) x (360 passages) = 1400 lux.h

This figure is considerably lower than the value predicted (ELD) on the basis of the spectral variations that occurred in the sample, even though the above calculation clearly overestimates the actual level of the light because, as an extreme situation, the maximum value was assumed.

It should be pointed out that all the above results stress the cooperative, and not simply additive, character of the cumulative damage that occurred in the sensor, which faded under the action of light and other environmental factors. Indeed, the reference BWSl, maintained in the dark, cannot be useful tor 'measuring' the separate contribution of factors other than light. For example, the damage that occurred in the non-exposed BWSl after 30 days can be evaluated as a 10 klux.h exposure (Figure 14). By adding to this value the actual light-dose measured by the datalogger after 30 days (that is, about 130 klux.h), we obtain a total light-dose of 140 klux.h, which is lower than the corresponding ELD recorded by the exposed BWS1 (more than 200 klux.h after 30 days, Figure 12).

  1   2

Добавить в свой блог или на сайт


Calibration and Use of Photosensitive Materials for Light Monitoring in Museums iconPhotonic Band Gap Materials: Light Control at Will

Calibration and Use of Photosensitive Materials for Light Monitoring in Museums iconThe light is dim as you enter the room after receiving no answer from repeated knocking. The lack of light is due to aluminum foil and newsprint taped to the

Calibration and Use of Photosensitive Materials for Light Monitoring in Museums iconThe laser is a device that a beam of light that is both scientifically and practically of great use because it is coherent light. The beam is produced by a

Calibration and Use of Photosensitive Materials for Light Monitoring in Museums iconLanguage: Most of the materials in this collection are in English. Occasionally, there are materials in German

Calibration and Use of Photosensitive Materials for Light Monitoring in Museums iconGalleries/Museums/Collections [A. 8]

Calibration and Use of Photosensitive Materials for Light Monitoring in Museums iconD irector of Oxford Centre for Advanced Materials and Composites (ocamac) and Faraday Partnership In Automotive And Aerospace Materials

Calibration and Use of Photosensitive Materials for Light Monitoring in Museums iconIn short: when it comes to problems solving in practice sessions, nothing beats above mentioned materials. But for strategies (read shortcuts ) you need to look for some extra materials

Calibration and Use of Photosensitive Materials for Light Monitoring in Museums iconMuseums as ‘ Flagships ’ of Urban Development

Calibration and Use of Photosensitive Materials for Light Monitoring in Museums iconStandards & calibration

Calibration and Use of Photosensitive Materials for Light Monitoring in Museums icon5 Seismic Sensors and their Calibration

Разместите кнопку на своём сайте:

База данных защищена авторским правом © 2012
обратиться к администрации
Главная страница