The Lorentz oscillator model is the highly preferred formula for calculating the re-fractive index of binary mixtures, such as adulterated diesel oils. This is given in (Bar-anovic G, 2017), and is expressed as:
𝑛2−1 𝑛2+2= 𝑓𝐷
𝑛𝐷2−1
𝑛𝐷2+2+(1−𝑓𝐷)(𝑛𝐾2−1)
𝑛𝐾2+2 , (3.18)
where n is the refractive index of the resulting mixture, nD is the refractive index of diesel oils,and nK is for kerosene. We can define an ideal mixture using this model
20
because, it gives good estimate of volume fill fraction for ideal mixtures when there are no interactions between molecules of the participant liquids. However, when the interactions exist, this formula gives erroneous volume estimates (Paper IV).
3.8 MODIFIED IDEAL LAW OF BINARY MITURES
The ideal mixture equation (Eq. 3.16) can be re-written so that the total volume ex-pression includes the volumes of the individual binary mixture constituents, namely diesel oil as well as kerosene. This leads to a modified equation given as:
𝜀𝑖𝑑𝑒𝑎𝑙=𝑉𝐷
𝑉 𝜀𝐷+𝑉𝐾
𝑉 𝜀𝐾, (3.19)
where V = VD + VK is the total volume of the mixture.
Next, Eq. (3.16) is further modified to incorporate the novel concept of increase of volume of pure diesel oil in the suspected sample. This was achieved by introducing another variable V’, this leads to the modified formula given by:
𝜀𝑖𝑑𝑒𝑎𝑙(𝑉′) =𝑉𝐷+𝑉′
𝑉+𝑉′ 𝜀𝐷+ 𝑉𝐾
𝑉+𝑉′𝜀𝐾. (3.20)
The added volume V’ should be with respect to the magnitude of the initial vol-ume V of the suspected sample. If the volvol-ume V’ is continuously increased, the sam-ple approaches the case of ideal mixture as the concentration of kerosene continues to diminish. The limiting value for this mixing procedure is given by Eq. (3.21) (Paper IV).
𝑉′→∞lim 𝜀𝑖𝑑𝑒𝑎𝑙(𝑉′) = 𝜀𝐷. (3.21)
21
4 OPTICAL MEASUREMENTS
In this chapter we briefly describe the mechanism behind signal measurement method of the prototype sensor, this is followed by description of the samples that were utilized in this thesis. Thereafter, the refractive index data which were ured both in Finland and Tanzania are presented. Finally, the transmittance meas-urements for the samples are presented. These were necessary for calculations and computations in the next parts.
4.1 OPTICAL SIGNAL MEASUREMENT (PROTOTYPE)
Optical sensors provide quick means to solve complex measurement problems, such as the case of diesel oil adulterated by kerosene. Herein, a prototype sensor for de-tection of adulteration, which is a modified version of a handheld gloss meter (Kui-valainen K., 2010), is presented (Paper I). The theoretical background of light inter-action with rough glass-liquid interface, which was adopted from (Niskanen I., 2012;
Beckmann P., 1963; Nussbaumer R. J., 2005), was presented in chapter 3. This is fur-ther complemented by the theory of wetting process and refractive index mismatch which are well addressed in several studies (Nussbaumer R. J., 2005; Quere D, 2008;
Furmidge C. G. L, 1962; Raltson J., 2008; Cazabat A. M. & Cohen S. M. A, 1987).
For mixtures of less problematic samples such as sample A and D, the table model Abbe refractometer give accurate results which makes it easy to separate the samples.
However, for field measurement conditions such a bulky device is not practical. To the contrary there do exist a handheld Abbe refractometer (Atago, H-50), which was tested for measurement but offers poor accuracy. However, in the later study a method that can make use of a handheld refractometer to predict and separate highly adulterated fuels from low adulterated fuels was developed (Paper IV). The pre-sented prototype relies on combined effect of roughness, contact angle, wetting, and refractive index mismatch. These effects cause different liquids to behave differently on the rough glass surface, enabling the identification and separation of signals rec-orded from different samples with high accuracy.
The other stimulus for this work, was a recent article wherein the handheld gloss meter was utilized for screening fake antimalarial tablets (Bawuah P., 2017), thanks to the diffractive optical mechanism incorporated with the sensor. Similar dif-fractive optical mechanism together with laser transmission have also been utilized in other related works (Silvennoinen R., 1999; Jääskeläinen A., 2000). The sensor lay-out is presented in in Fig. 4.1, and the light source of the sensor is a semiconductor laser with an output power of 0.8 mW and which is lasing at 635 nm. Both the DOE as well as the laser are inside the device depicted in Fig. 4.1. The rough quartz glass which was adopted for this work is (VWR microscope slide ECN 631-1550) whose refractive index value at 635 nm is 1.4570. For more detailed description of the pro-totype refer (Paper I).
22
Figure 4.1. Schematic diagram of a handheld sensor for fake diesel oil screening. DOE is a diffractive optical element. The dashed lines describe scattering of light due to the surface roughness and refractive index mismatch between the fuel and the glass (Paper I).
4.2 DESCRIPTION OF THE TRAINING SET
One of the challenges faced by researchers in various fields is the issue of sampling, namely choosing acceptable samples whose results can be representative of the pop-ulation of the study. In this work, the issue was considered critically and as a result both summer and winter categories were included, to represent the varying climatic conditions across the globe. We utilized one summer diesel oil grade (sample A), and two winter diesel oils (sample B, and sample C). The difference between winter diesel oil samples is based on the lowest temperature that is specified for the car to operate, namely the lowest temperature at which the engine should start. For sample B it is (- 20˚C) and for sample C it is (- 32˚C). The origin of crude oil for both summer and winter diesel oil samples is Russia. Moreover, we utilized a well-known brand of kerosene as an adulterant (Alfa Aesar, Haverhill, MA, USA), this is available for sci-entists across the globe.
Adulterations were prepared by blending pure diesel oils (sample A, B and C) with kerosene (sample D). Typical adulteration level across the globe is 20%-30%
(Mishra V., 2008), therefore, in this work lower percentages of 5%, 10% and 15% were selected and utilized for analytical purposes. These samples were measured in the laboratory in Finland.
The laboratory measurements in Finland alone were not enough, because fuel adulteration is a global challenge which highly affect third world and developing countries, more than developed countries. Based on this reason, we performed field measurements in Tanzania. The samples for field measurements were provided by fuel regulatory authority of Tanzania (EWURA). This was also necessary to enable
23 the study of variations in diesel oil and kerosene samples. Usually these samples have varying properties based on oil fields of origin, because these liquids constitute hundreds of hydrocarbons with different variabilities across the globe (Szymkowice P. G & Benoges J, 2008). Moreover, these variations are also depicted in the refractive index values of these liquids, which highly relies on the crude oil of origin.
Differentiating diesel oils and kerosene mixtures is difficult, since both fuels have overlapping fingerprints in NIR spectral region. The sets A-D serve us as a “training set” for exploring optical properties of diesel oil grades and kerosene, and thereafter the data obtained with the aid of the training set can be used for designing practical sensors for field conditions, and relevant software, to identify any adulterated diesel oil product.
4.3 REFRACTIVE INDEX MEASUREMENTS IN FINLAND
Refractive index is a constant which describes or depicts the way in which light in-teracts with the medium. It considers other external factors such as temperature as well as pressure, which are not captured by the density measurements. This quantity is very useful for characterization, namely in cases where two samples have varying refractive index values, it is possible to separate them. Refractive index has been uti-lized for characterizing samples in different studies (Payri R., 2013; Geacai S., 2012;
Polynkin P., 2005; Kim C. -B. & Su C. B, 2004; Magnusson R., 2010; Fernandes V. H, 2008; Mishra V., 2008; Ariponnammal S., 2012). However, there are circumstances when the mixtures are problematic such as adulterated diesel oils, and the difference is not obvious. Nevertheless, the refractive index at one wavelength (anchor point) can be utilized, to indirectly extrapolate optical properties of the same material at other wavelengths. This provide more possibilities for differentiation of the samples.
In this thesis Abbe refractometer (Atago RX5000) was used for measuring the re-fractive index of different fuel samples at 589 nm, the device measures the rere-fractive index to an accuracy of ±0.00004. Table 1 shows the measured values for both authen-tic and adulterated samples under room conditions (23˚C). From the third column of Table 1, the magnitude of refractive index for summer diesel oil sample A is different from those of winter diesel oil samples B and C. Moreover, for samples B and C there is only slight difference in the third decimal, whereas the values for samples B and C, are very close to that of sample D, making a mixture of these samples and kerosene a problematic case.
24
Table 1: Refractive index of different authentic and adulterated fuel samples.
Sample Volume
Winter diesel oil (B) 1.44775
5% 1.44894
10% 1.44863
15% 1.44791
Winter diesel oil (C) 1.44653
5% 1.44742
10% 1.4472
15% 1.44702
Kerosene (D) 1.44230
In Table 1 (fourth column) also are shown the refractive index data for samples A, B and C, adulterated by different volume levels of kerosene, as measured by Abbe re-fractometer. For sample A there is decrease in refractive index as the volume of adul-terant (kerosene) increases, also the refractive index difference between 5% and 10%
adulteration is in the fourth decimal, while the difference between 15% adulteration and the former is in the third decimal. Moreover, the values of refractive index for the mixtures lies between the refractive index of authentic samples A and D, this agrees with the theory of conventional binary mixtures.
Likewise, for the case of samples B and C from Table 1 (fourth column) there is a similar trend as for sample A. However, the mixtures have higher values than au-thentic samples. This is not expected conventionally, usually the mixture refractive index lies in between the two mixture constituents. If only refractive index is consid-ered one might confuse adulterated samples with authentic ones. These measure-ments were performed again after a few months to assess if the values will be differ-ent, but no substantial changes were noticed. This led to the belief that, the reason for the abnormal refractive index value of complex winter diesel oil and kerosene mix-tures may result from chemical interactions, this we demonstrated in (Papers II and III). There might be even more difficult circumstances of adulteration, for example
25 when adulteration is done by more than one substance. In such a scenario the refrac-tometer might fail to differentiate the samples, the feasible approach for such a prob-lem is presented in (Paper IV).
4.4 REFRACTIVE INDEX MEASUREMENTS IN TANZANIA
The field samples in Tanzania were directly measured by Abbe refractometer Atago H-50. This was necessary to confirm the possibility for detection of adulteration by onetime measurements. It is also worthy to point out that, the application of this de-vice to detect adulteration of liquid fuels has never been reported anywhere in liter-ature. In Table 2 are shown the refractive index measurements for adulterated sam-ples which were performed at two different temperatures. This was necessary to as-sess the effect of temperature.
From Table 2, it is obvious that unlike the training set measurements of Table 1 where diesel oils have higher refractive index values than kerosene, here kerosene has a higher value. Moreover, the variation of refractive index (diesel oil as compared to kerosene) is large enough. For the case of adulterated samples, the situation is in-teresting, namely the refractive index value for 5% adulteration is above that of pure diesel oil. However, for 10% and 15% the value is same. The handheld device is inca-pable to differentiate between these samples. This is still not a weakness if one is interested to reveal adulteration because still the value is higher compared to that of authentic diesel oil, and even higher compared to that of 5%. The adulterated sam-ples behave normally since the refractive index for both 5% 10% and 15% falls in between thus, nicely in line with binary mixture rules. For more details including the volume fill fraction calculations for Tanzanian samples refer (Paper IV).
Table 2: Refractive index data for authentic diesel oil, kerosene and their mixtures measured in Tanzania.
Sample Adulteration percent Refractive index n (25˚C)
Refractive index n (27˚C)
Diesel 0% 1.4440 1.4433
5% 1.4451 1.4444
10% 1.4463 1.4456
15% 1.4463 1.4456
Kerosene 0% 1.4640 1.4644
26
The refractive index values that were measured in Finland (Table 1), were com-bined with the measured transmittance spectral data for further analysis.
4.5 TRANSMITTANCE MEASUREMENTS
In this section we briefly consider the transmittance measurements for both authentic and adulterated samples in Vis-NIR spectral range. Moreover, the cuvette reflection issues are addressed to ensure accurate data for further processing and analysis.
4.5.1 Double optical path length method (authentic samples)
The NIR transmittance spectra of the fuel samples A-D, were measured at room tem-perature with the aid of a spectrophotometer (Perkin Elmer Lambda 9), and a 5mm as well as 10mm thick quartz cuvette, the two cuvettes were utilized to try to get rid of cuvette reflections by double optical path length method.Data pretreatment was twofold, namely initial baseline correction was done at the beginning by the instru-ment with empty cuvette, later multiplicative scatter correction (MSC) which takes care of pathlength errors, baseline shift and interference was done using (PLS-Toolbox, Eigen vector Research, INC, USA) in Matlab software package. The accu-racy of spectrophotometer in UV-Vis-NIR range was 0.07% for 1 unit of absorbance, and the target spectral range for this work was Vis-NIR spectrum. The spectral range below 2000 nm was chosen, to facilitate the possibility of utilizing the cheap commer-cial spectrometers operating at this range, for field condition measurements.
Usually during spectroscopic measurements, the reflections form outer and inner surfaces of the cuvette affects the measurement accuracy of both absorption coeffi-cient and extinction coefficoeffi-cient of fuel samples. The refractive index value of the quartz cuvette is not so much different from the fuel samples of this study, and there-fore reflection losses are rather low. The two- optical path length method was ex-ploited to get the best estimate of the properties of authentic diesels, it is highly im-portant to ensure that, the data used is highly immune to external influences, this is imperative especially for liquids with closer refractive index values. To cancel reflec-tions by double optical path length method, the Beer Lambert’s law from (Eq. (3.7)) was applied. The path length for 10 mm cuvette is represented by 2d, while the one for 5 mm cuvette is represented by d. The measured transmittances are denoted here by T1 = exp(-α(λ)d) and T2 = exp(-α(λ)2d). First, the transmittance ratio r was calculated by r = T2/T1 = exp(-α(λ)d), and next the estimate of the absorption coefficient was cal-culated by equation α = (−1/d)ln(r) where d = 5 mm. The calcal-culated absorption coeffi-cients were further utilized for analysis.
In Fig. 4.2 are shown the Vis-NIR transmittance curves for different fuel samples at ca. 431 nm to 1600 nm, the double optical pathlength method was utilized to obtain the transmittance data with less effects from cuvette reflections. It is evident from (Fig. 4.2) that, in the NIR region at ca. 1200 nm to 1400 nm, the transmission is weak due to strong absorption of NIR radiation which is due to hydro-carbons of the fuels
27 (C-H stretch). There is overlap in the spectra and there is decrease in transmittance for sample A as compared to other samples, which is caused by additives which are present in winter diesel oils. Therefore, the region below 600 nm is useful for differ-entiating summer from winter diesel oils. Moreover, the transmittance of kerosene is very close to that of diesel oils, this is one of the reasons why it is difficult to separate and differentiate adulterated samples. Therefore, a combination of refractive index and transmission spectrum is considered to address this problem.
Figure 4.2. The transmittance curves for diesel oils samples (A-C), and kerosene sample (D) (Paper II).
Next, the usefulness of double optical pathlength method is demonstrated in the ab-sorption coefficient curve in Figure 4.3. From (Fig. 4.3) we can see the abab-sorption coefficient resulting from the ratio of the measurements by two cuvettes (estimated absorption coefficient for sample A), as well as the measurements performed by 5 mm and 10 mm cuvettes. Based on the curves obtained, the absorption coefficient resulting from the ratio is above zero line, while the absorption coefficients resulting from other measurements have values of zero. This is in perfect agreement with the-ory which suggests that the absorption coefficient for these samples always have a positive value above zero. Similar case holds for samples B, C, D, and E. The absorp-tion coefficients resulting from transmittance ratios were utilized in further calcula-tions.
28
Figure 4. 3. The measured and estimated absorption coefficient curves for sample A
4.5.2 Adulterated samples
Here we briefly present the transmittance curves for samples A and C, the case of sample B is similar to sample C. In Fig. 4.4 are shown the transmittance curves for diesel oil (sample A) adulterated by three varying volumes of kerosene (sample D), at ca. 431 nm to 1600 nm. It is obvious that the transmission is relatively week in the vicinity of 1200 nm to 1600 nm, this is caused by hydrocarbons of the fuel (C-H stretch) which strongly absorb light at 1200 nm and 1400 nm. This is the reason why NIR range is useful for characterizing materials, because materials which have or-ganic functional groups absorb NIR light. Moreover, there is a clear spectral distinc-tion between the authentic and adulterated samples, but authentic and kerosene sam-ple almost overlapp. On the other hand, the curves resulting from 10% and 15% adul-teration are overlapping in most areas while, the curve resulting from 5% adultera-tion has slight spectral distincadultera-tion in certain areas. For the case of 10% and 15% adul-teration it is difficult to separate them based on spectral features.
The nature of transmission spectrum for sample B and C is pretty much like that of sample A, but only differs in the Vis region, only the curves for sample C are shown in Fig. 4.5 the one for sample B deserve similar treatment. Both curves are overlapping in most areas of Vis range. Furthermore, in NIR region the adulterated samples are slightly distinguishable, namely the curve for 15% adulteration is slightly separated from 10% and 15% adulteration. If one exclusively considers the case of sample C, the situation of screening of adulterated diesel oils is much worse with respect to the interpretation of the spectral data, moreover, samples C and D have very similar values. This is one more indicator of the issue of screening fake diesel oils by their Vis-NIR spectra or the refractive index data. In the next part, we
29 explore other advanced data analysis methods that were used to characterize the adulterated samples. These were utilized in (Papers II and III)
Figure 4.4. Transmission curves for authentic and adulterated sample A with kerosene at Vis-NIR (Paper III).
Figure 4.5. Transmission curves for authentic and adulterated sample C with kerosene, at Vis-NIR.
30
5 DATA ANALYSIS METHODS
In this part we briefly and theoretically describe several computational methods that were utilized in this thesis. The methods were applied to calculate several optical constants both in broader and narrow spectral ranges, the results were further uti-lized to characterize different samples. These methods include the imaginary refrac-tive index (extinction coefficient), real refracrefrac-tive index, excess permittivity, imagi-nary excess permittivity, and finally the method of increase of volume. Except for the later, the remaining quantities were calculated as a function of wavelength in broader Vis-NIR spectral range and the results were utilized for sample characterization.
5.1 EXTINCTION COEFFICIENT
The extinction coefficient is the fundamental optical property of a material, which describe and quantify the decay of the amplitude of incident electric field as light
The extinction coefficient is the fundamental optical property of a material, which describe and quantify the decay of the amplitude of incident electric field as light