The first step of our study was to acquire basic biological data regarding the taxonomic status of the main microbial colonizers of the biofilms. Following preparation, samples were examined, by one of the authors (IP), under a digital microscope (magnification: × 460) and the main microbial groups were identified. For objects B and C, the biological material was sampled from the areas indicated by the dark-coloured spots observed in Fig. 1 (middle column) and the presence of lichens was revealed. For objects A and D, the biological material was scraped off the surface of the samples (Fig. 1, middle column) and the main colonizers were found to be algae (green areas) and fungi (black or brown areas). This information was essential in order to properly validate the results of our fluorescence spectroscopic analysis.
The next step involved the evaluation of the operational and performance characteristics of the LED-IF-adapted portable spectrometer (LED, filters, optics, spectrometer-detector) as well as the development of a straightforward measurement methodology that are essential prerequisites for the design of an instrument, which will ensure acquisition of reproducible and reliable results both in the laboratory and in the field. The aspects investigated include among others: (a) evaluation of the optical components of the instrument, (b) selection of the proper LED source, (c) comparison of fluorescence spectra acquired with the portable system against those obtained by use of a conventional laboratory fluorometer (d) study of the signal evolution (stability) over time, (e) treatment of the samples. In the following sections, the above aspects are detailed and the results of our studies are presented and discussed.
Evaluation of instrument performance and measurement parameters
Given that excitation is a critical experimental parameter when performing fluorescence measurements, we focussed our attention on the emission characteristics of the LED sources in terms of both spectral output and emission intensity. As widely known, LEDs emit light with a considerably broader spectral bandwidth compared to lasers and even to laboratory fluorometers, which make use of an excitation monochromator to select a narrow band from the output of a high intensity Xe arc lamp. The wide emission peak of the LED could on the one hand compromise selectivity in the excitation and, on the other, mask or superimpose on possible fluorescence features appearing in the vicinity of the LED excitation wavelength range. Given that fluorescence emission measurements require excitation at a relatively narrow spectral bandwidth, typically 1–5 nm, we employed narrow band-pass filters to reduce the FWHM of the LED spectral output. To this end, a band-pass filter at 370 nm (FB370-10, Thorlabs, USA, FWHM = 10 nm) was placed in front of the LED emitting at 375 nm and likewise a laser line filter at 441.6 nm (FL441.6-10, Thorlabs, USA, FWHM = 10 nm) was used for the LED emitting at 440 nm. In Fig. 2, the emission spectra of the LED sources are shown both raw and following transmission through the band-pass filters used. In order to measure the emission FWHM, a relatively low LED power has been employed so that the LED intensity does not lead the detector to saturation. It was found that, by use of the filters, the FWHM is reduced from 15 nm down to 10 nm for the visible LED (440 nm) and, likewise, from 11 nm down to 6 nm for the ultraviolet LED source (375 nm). It is noted that during fluorescence measurements, the maximum available power of the LED is applied and thus one should be aware that non-negligible excitation is supplied even at the wings of the spectral profile.
Apart from narrowing down the FWHM, the use of filters resulted in shifts, in the emission maximum of the LED spectral output mainly because of the mismatch between the bandpass filter’s transmission maximum and the LED emission maximum. The ultraviolet LED shifted down to 374 nm (bandpass filter central wavelength: 370 nm), while the blue LED showed an up-shift to 440 nm (filter’s central wavelength: 441.6 nm). These features, although not crucial for our measurements, need to be known prior to analysis for a complete characterisation of the portable system.
Regarding the fluorescence signal acquisition, it is of concern the fact that the increased integration time applied (i.e. 300 ms), may lead to significant interference from ambient light. To avoid this complexity, spectra need to be acquired in the dark, which in many cases is difficult if not impossible, particularly when measurements are acquired outdoors. This problem is addressed by means of spectral background correction, a feature available by the spectrometer software. Figure 3 shows raw background spectra (LED off, no sample) acquired with the laboratory lights on (black line) and defined as “dark” in the spectrometer software. Then background-corrected spectra (red line) were acquired with the spectrometer operated at the “scope corrected for dark” mode (i.e. spectrum displayed after subtracting the “dark” spectrum), with the ambient light peaks successfully eliminated.
To further verify the applicability of the background subtraction feature, test measurements were conducted outdoors, in daylight conditions, and following acquisition of a “dark” spectrum (Fig. 3, blue line) it was confirmed that upon background subtraction the resulting “scope corrected for dark” spectrum was also free of any interference. This background subtraction feature is indeed convenient, if long integration times need to be used, especially when measurements are performed in-situ (archaeological/historical sites and monuments).
Following initial observations that evidenced changes with time of the fluorescence emission signals, collected from the biological encrustation, we decided to investigate the stability of the fluorescence emission signal over time. For this reason, a set of 10 consecutive LED-IF measurements (each corresponding to integration time of 300 ms, LED emitting at 440 nm) was carried out on the same spot of a Hibiscus leaf sample (Fig. 4a). The two bands at 685 and 735 nm are assigned to chlorophyll species (Chl) (see Table 1 and Introduction for further explanations) and it is quite evident that, with continuous excitation by the LED beam, the overall emission intensity drops, while also the relative intensity of the bands changes. By integrating the area under the two bands, one observes that the total fluorescence emission intensity decreases abruptly, until half-way in the spectral series and eventually approaches a rather stable regime (Fig. 4b). This observation can be interpreted either on the basis of chromophore photobleaching [46] or as a result of chlorophyll fluorescence quenching. The latter has been previously observed by Kautsky [47]; upon transferring photosynthetic material from darkness to light an increase in the chlorophyll fluorescence occurs over a time period of 1 s and then following this, the fluorescence intensity level typically starts falling again over a time scale of a few minutes, depending on the species [47].
Given that comparisons among different areas bearing biodeterioration are to be made, which might help one to estimate the quantity of the biological deposit, it is important to conduct measurements under conditions which warrant signal stability. As observed in Fig. 4b any measurement performed in the stabilised intensity regime can be considered as acceptable and, in the context of the present study, it was decided to keep the spectrum from the 10th measurement and this was consistently followed throughout the experiments.
Fluorescence characteristics of biodeteriogens
In the beginning of this study, a preliminary set of fluorescence measurements was conducted on the stone objects, utilizing the Fluoromax benchtop spectrofluorometer. This facilitated a better understanding of the fluorescence emission properties of the biological deposits investigated and moreover indicated the excitation wavelength range most suitable for obtaining efficient excitation of the chromophores and thus securing optimum detection of their fluorescence emission. These measurements also provided critical input needed in order to select the most suitable LED for the portable instrument and, in addition, served as a reference against which to compare the LED-IF data.
Fluorescence emission spectra were recorded upon excitation at wavelengths varying from 400 to 600 nm in increments of 25 nm (excitation wavelength scanning). The two limestone objects (A and B), the mortar object (D) and the Hibiscus leaf were studied. The archaeological samples were introduced in the sample chamber following wetting of their surface according to the procedure described in “Experimental setup—materials and methods” Section (see also “The effect of wetting the samples surface on the fluorescence signal” Section). The third limestone object, C, was not analysed because its large size made it impossible to position it properly in the sample chamber of the fluorometer. To ensure reproducibility, the excitation wavelength scanning was carried out three times for each sample at the same area (marked by a red circle in the images shown in Fig. 1). The reported spectra represent the average of the three different acquisitions. In the case of the leaf, it is noted that the three measurements were acquired on three different leaves. All spectra have been normalised with respect to the emission intensity at about 640 nm.
The fluorescence emission spectra obtained from the leaf are shown in Fig. 5 and as expected, two main bands are visible, at 685 nm and 735 nm, both attributed to chlorophyll species [12]. It is observed that the band at 685 nm exhibits its highest intensity for λexc at 475 and 500 nm. The highest intensity for the band at 735 nm, appears with λexc in the range of 500–575 nm. Clearly, the two fluorescence bands behave differently upon excitation of the leaf at different wavelengths and this is evident by observing the variation in the values of the ratio of their intensities at the maxima of the emission bands, I685/I735, as a function of λexc (Fig. 5b).
This sort of behaviour has been reported previously and is attributed to re-absorption of the emitted chlorophyll fluorescence at 685 nm by the chlorophyll absorption band at around 670 nm [48]. The green leaf pigments, chlorophylls and carotenoids, have broad absorption bands between 400 and 500 nm, so when the excitation wavelength is at that range, light does not penetrate deeply into the leaf tissue, thus chlorophyll fluorescence is mainly emitted from cells close to the leaf surface, and only little re-absorption takes place. With excitation light between 500 and 600 nm, that is not absorbed by carotenoids and penetrates more deeply into the green leaf mesophyll, the chlorophyll fluorescence is also emitted from the interior layers of the leaf and along its path to the surface, it likely suffers significant re-absorption compared to chlorophyll fluorescence induced by blue light excitation [48]. Although there exist differences on the biological characteristics between the leaf and the microorganisms investigated in the present study, the chlorophyll fluorescence is expected to show similar behaviour. This behaviour should be taken into account when interpreting fluorescence spectra of this type of samples, since factors such as the excitation wavelength, could affect the observed bands intensity and the relative intensity between bands. Moreover, a variability in intensity and intensity ratio is observed among different leaves or even on the same leaf if different areas are probed (the relative standard deviation of the band intensity at its spectral maximum for the three measurements can be as high as 50%; data not shown here). Especially for the intensity ratio I685/I735, this varies also depending on the content of the chlorophyll species [49], which should also be considered, while studying the fluorescence of such materials.
Figure 6 shows the fluorescence emission spectra collected from object A (see image in Fig. 1). The main band observed appears having its maximum at about 675 nm and reaches its highest intensity for λexc = 425 nm. The emission intensity is quite low upon excitation across the whole range of λexc; especially for λexc = 575 nm and λexc = 600 nm the band is hardly noticeable over the spectral background (spectra omitted from the graph). Observation of the fluorescence band at 675 nm is an indication that the biodeteriogen contains chlorophyll, confirming the findings of the classical taxonomy. The observed shift of the band emission maximum, compared to that of the leaf chlorophyll, is probably due to the reduced chlorophyll content of the biodeteriogen, which causes a shift of the chlorophyll fluorescence maximum to lower wavelengths [48, 50].
In the case of Οbject A, the excitation wavelength that leads to maximum fluorescence emission is λexc = 425 nm, a value close to the main absorption band of chlorophyll at 418 nm (Soret band), and this is also indicative of the presence of chlorophyll in this type of biodeterioration [19, 20]. The results obtained from object D are similar to the ones acquired for object A, therefore it was not deemed necessary to present them here.
Concerning object B, the fluorescence emission spectra are displayed in Fig. 7 (see also image in Fig. 1). The main emission band is observed to have its maximum at around 660 nm while the highest intensity is obtained at λexc = 575 nm. This observation suggests that the fluorophores in this biological crust are different compared to the ones on object A. Based on literature data, a potential chromophore is allophycocyanin that indeed fluoresces at 660 nm (see Table 1) [12]. Moreover, the fact that, for object B, the maximum intensity of fluorescence signal is observed at λexc = 575–600 nm could be correlated with the presence of cyanobacteria, which have an absorption maximum at about 570–650 nm (see Table 1) and this is actually in accordance with the classical taxonomy findings.
Based on these experimental results, it is apparent that there exist different fluorophores (biodeteriogens) on object B and objects A and D, given that their fluorescence bands appear in different wavelengths and the maximum fluorescence emission intensity is induced by different excitation wavelengths. These observations are in agreement with the results from the classical taxonomy analysis.
The effect of wetting the samples surface on the fluorescence signal
The surface of each sample has been wetted prior to analysis, as mentioned in “Experimental setup—materials and methods” Section, and two are the main reasons for this practice. First, it was observed in some preliminary experiments conducted in our laboratory that wetting the surface leads to an increase of the chlorophyll fluorescence intensity. This effect is not fully explained yet but there are two phenomena that could act synergistically. On the one hand, the application of water changes the optical properties of the sample surface, effectively the refractive index, reducing scattering and permitting incident light (here the LED light) to penetrate more efficiently in the rough biological surface thus resulting in higher fluorescence emission. On the other hand, photoautotroph microorganisms such as cyanobacteria are known to be able to resume their photosynthetic activity rapidly upon rehydration (called activation), which entails a ‘greening’ of the surface, obvious in our experiments soon after surface wetting, and a subsequent increase in fluorescence signal even within minutes [49]. It should be noted though that not all biodeteriogens are affected similarly by the presence of water. For example, fungi can survive in extreme environmental conditions, so the addition of water is not expected to affect their behaviour and subsequently their fluorescence signal. In order to further investigate this phenomenon, digital microscope images were acquired for the dry and wet surface of objects B and D (Fig. 8); a clear difference is noticed between the two surfaces, with the wet surface appearing greener.
The second reason for wetting the samples relates to an attempt to simulate a real case scenario. Meteorological phenomena, such as rainfall, snowfall, ice formation, morning dew etc. affect the exterior of buildings and monuments as well as sculpture displayed outdoors by increasing the humidity on their surface. In order to better study the effect of humidity and water on the fluorescence signal of the biological materials investigated in this work, the samples were analysed both in their original state, i.e. dry, as well as wet following the procedure described in “Experimental setup—materials and methods” Section.
In this section, the effect of wetting of the biofilms on their fluorescence emissions is detailed. As a control, it was deemed necessary to acquire first the fluorescence signal from the substrate of each archaeological object, namely an area on its surface, where no sign of biological growth is detected under visual observation with a microscope. The objects were analysed on their back side, where the bulk material was observed, and LED-IF spectra were collected with λexc = 375 nm (Fig. 9). A broad emission band in the range of 500–600 nm is observed for most of the substrates with a varying intensity. This broad band has been reported in the literature and its origin is interpreted as emission from minerals, such as calcite (the main mineral in the objects of the present study), which is caused by adsorbed cations and by defects in the mineralogical periodicity [30, 51, 52]. It is worth noticing that the substrates are quite heterogeneous, fact that is reflected in the varying intensity of this band across the surface of the same object but, in any case, the substrate spectra can serve as a background against which the biological material fluorescence spectra are compared.
Following acquisition of the substrate spectra, the biofilms (both in dry and wet state) for each one of the four objects were examined. In Fig. 10, the fluorescence emission spectra from two of the objects, A (Fig. 10a) and C (Fig. 10b), analysed with the ultraviolet LED source are depicted, with a spectrum of the Hibiscus leaf superimposed (green-coloured spectrum) serving as a reference for chlorophyll emission. As expected, a chlorophyll fluorescence band at around 680 nm was detected (Fig. 10a) and by comparing the spectra of the dry and wet samples, it is evident that upon wetting of the surface, the chlorophyll signal increases by a factor of nearly 3. To eliminate the possibility that this signal enhancement could happen also in non-biological material, the fluorescence spectra of the substrates, dry and following wetting, were recorded as well, and are shown in Fig. 10. The only signal observed for both the dry and the wet substrate is a low-intensity broad band emission at around 500 nm (as also seen in Fig. 9) which does not change in intensity at either the dry or the wet state. This implies that indeed, applying water on the sample surface leads to a noticeable increase in fluorescence emission within a few minutes, in the case of the biofilm, while it does not affect significantly the fluorescence signal from the substrate.
It is noteworthy that the band at 500 nm, although originating from the substrate shows lower intensity for the bare substrate than for the biofilm. This could be explained on the basis of a number of reasons. First, the biodeteriorated surface morphology and structure has been altered by the microorganisms colonising it, which may have changed the crystal structure of the substrate inducing defects and altering its intrinsic fluorescence signal. Second, the substrates have a quite heterogeneous macroscopic structure; even analysis at different points on the same object results in fluorescence intensity variations. Also, it is likely that the biodeteriorated surface presents other impurities apart from microorganisms (such as environmental pollution deposits) that may be responsible for additional background emissions. So, in reality, it is not possible to compare quantitatively, namely in terms of their fluorescence emission intensity, the substrate/bulk material to the biofilm, since the two surfaces are very different in nature. The analysis of the substrate can only give qualitative information as for the origin of specific bands.
For all samples analysed, a similar behaviour was observed; the chlorophyll signal was consistently higher on a wetted surface. In some cases, wetting of the sample even made possible the detection of distinct spectral features (see where arrow points in Fig. 10b), which were hardly noticeable in the dry sample spectra. These observations clearly point out to an important climatic parameter that needs to be considered when analysing biodeteriogens on monuments, and this relates to weather conditions and seasonality. If the season is characterised by reduced humidity, i.e. summer, some biodeteriogens that contain chlorophyll might not be detected or their quantity could be underestimated leading to wrong conclusions about the presence and extent of biological growth on a monument. A humid or rainy season would be preferred when photoautotroph organisms need to be detected in-situ by means of their fluorescence.
In the context of this study and given that our main objective was the investigation of the capabilities and limitations of the LED-IF instrument and the analysis methodology, it was decided that all the samples be wetted prior to analysis, in order to ensure that the photoautotroph biodeteriogens, if present, will be more easily detected.
Comparison of the two LED excitation sources
The experimental findings concerning the fluorescence properties of the biodeteriogens (“Fluorescence characteristics of biodeteriogens” Section) showed that objects A and D, as well as the leaf, exhibit maximum fluorescence signal with excitation wavelengths in the range of 425–475 nm. Based on the LED sources available in the market, it was decided to use a diode emitting in the range of 400–450 nm.
It should be noted here, that sample B exhibited maximum fluorescence signal intensity for excitation wavelengths over 550 nm due to the different fluorophores present in this sample (i.e. cyanobacteria) (see also “Fluorescence characteristics of biodeteriogens” Section). It was decided though not to select a LED source emitting near 550 nm, because this way chlorophyll species would not be efficiently excited and furthermore because such a choice would eliminate the possibility to record potential emission bands from other fluorophores below 550 nm. Therefore, in the present study, two LEDs were assessed based on their ability to excite the fluorophores present in the biodeteriogens: one emitting in the ultraviolet (UV), λexc = 375 nm, and another in the visible (Vis, blue), λexc = 440 nm.
As a test, the Hibiscus leaf was analysed with both LEDs to study the effect of different excitation wavelengths and power on the induced fluorescence and spectral data are shown in Fig. 11 (with no spectral pre-processing). With either excitation source the fluorescence emission bands of chlorophyll species in the leaf are clearly evident. Comparing the emission spectra of the leaf obtained with both LEDs delivering approximately the same power (P = 6 mW) on the sample surface, it is evident that visible excitation yields a clean spectrum with a higher value of signal-to-noise ratio (SNR = 43), compared to UV excitation (SNR = 25). The noise was calculated as the standard deviation of a baseline spectrum. It should be noted, however, that measurements were not acquired on the same spot of the leaf, so some deviations are expected in the intensity. Nevertheless, the blue LED source yielded consistently spectra with higher SNR values. By increasing the power of the visible LED, from P = 6 mW to P = 26 mW the intensity of the chlorophyll emission at 685 nm increased approximately by a factor of 4, following almost linearly the increase of the source power.
Discrimination of different types of biodeterioration
The major goal of this study has been to investigate whether the new portable LED-IF instrument is capable of detecting and discriminating the possible types of biodeterioration products that are present on monuments on the basis of their fluorescence profile. Figure 12 depicts fluorescence emission spectra from each one of the archaeological objects, collected with both LEDs operating at their maximum available power. The spectra have been normalised with respect to the intensity of the emission band at around 500 nm.
Object A exhibits a main emission band at 675 nm and a shoulder around 720 nm, both attributed to chlorophyll species. The spectra are similar to those acquired on the spectrofluorometer (see Fig. 6). The reason for the overall different spectral profile and the shift of the bands in comparison to the observed leaf spectra is explained in “Fluorescence characteristics of biodeteriogens” Section.
A comparison of the spectra in terms of SNR shows that the visible LED at 440 nm gives rise to cleaner spectra and this is attributed to its higher power output. It is, however, understood that other factors are critical as well, for example, the overall absorbance of the biofilm at the excitation wavelength.
The analysis of objects B and C was performed on the dark round spots (see Fig. 1) but since the area probed with the portable LED-IF instrument is larger, about 2 × 2 mm2, recording of emissions from the surrounding area cannot strictly be avoided. On object B, a predominant fluorescence band at 660 nm was detected, along with chlorophyll bands of lower intensity at about 675 and 720 nm. The band at 660 nm originates most probably from allophycocyanin (APhC), present in cyanobacteria, and this behaviour is in accordance with the fluorescence emission spectra acquired with the spectrofluorometer (see Fig. 7).
Fluorescence emission spectra, collected from object C, revealed the presence of chlorophyll bands at 678 and 720 nm, as well as a shoulder at 660 nm, observed better with visible excitation. This finding indicates that object C contains both chlorophyll and allophycocyanin with the former prevailing against the latter. Finally, after analysis of object D, a band appears at 685 along with a shoulder at 735 nm, which are both attributed to chlorophyll species and the fluorescence emission maxima coincide with those recorded on the leaf.
The fluorescence analysis results, obtained by use of the LED-IF probe, are in agreement with the classical taxonomy analysis. In objects B and C, fluorescence analysis revealed the presence of (a) allophycocyanin, which is a pigment characteristic of cyanobacteria that are contained in lichens and (b) chlorophyll, which is also contained in cyanobacteria. In objects A and D, fluorescence shows mainly the presence of chlorophyll, which is the main pigment for algae. It is worth noting that although based on the classical taxonomy, fungi were present in some samples, no relevant fluorescence signal was recorded. This can be explained by the fact that the fluorophores existing in fungi exhibit absorption and fluorescence maxima in the UV range. There are some fluorescence bands in the visible but these might be overlapping with the fluorescence coming from the substrate, which lies in the same spectral area (See Table 1 and references therein).
These four objects illustrate clearly different cases of biodeterioration. In object D, chlorophyll is detected with the emission band maxima appearing at the same wavelengths as those in the leaf emissions. In object A, chlorophyll emission was also detected but with a band shift compared to the leaf chlorophyll emission. In object B, the signal originates mainly from allophycocyanin with a small contribution from chlorophyll. Finally, in object C, chlorophyll emission bands prevail over the one from allophycocyanin. These results indicate that with the present portable LED-IF spectrometer, it is possible to detect different types of biodeteriogens that could (co-)exist on monument surfaces, opening new possibilities in the protection of heritage artefacts, historic buldings and monuments against biodeterioration.
It is finally noted that, in the context of the present study, major colonizers were identified to group level (black microcolonial fungi, eukaryotic algae and lichens) based on their morphological features available at the time of their study. Further characterization, within each group, down to family or genus level requires the presence of different diagnostic fluorophores and is currently a subject under investigation.
