Oil testing kit Blog



Preliminary LED and isopropyl alcohol solvent tests for oil fluorescence analysis

by warren | about 8 years ago | 6 | 5

Just wanted to post a followup to @matej's post about the Sunview Luncheonette event -- we had a couple important successes which merit some more detail.

We'd long been meaning to test dissolving petroleum pollutants like crude oil in a different solvent than the mineral oil we'd been using, but things like methanol or especially hexane are quite toxic, and we wanted to avoid them. But we were hoping for a solvent that was as ubiquitous as mineral oil, yet possible to evaporatively concentrate. On @mattp's suggestion from the 2015 Barnraising, we tried isopropyl alcohol, at 91% from a pharmacy.

Screenshot_2016-02-19_at_6.09.21_PM.png

Evaporative concentration would be basically leaving it out to evaporate away the alcohol, leaving whatever's left to become more and more concentrated. Even if we had 100:1 alcohol to oil, we could evaporate out 90% of the alcohol to achieve a 10:1 ratio. Some of the oil could evaporate as well, but this is still a standard technique in lab methods -- for example, read this note by @silverhammer about evaporating off solvents.

In our quick test, a few drops of crude into a few ounces of 91% isopropyl dissolved quickly as I shook up the nalgene bottle, and when scanned, produced strong fluorescence:

Another test where I soaked a crude-soaked piece of PIG mat in isopropyl for 15 minutes appeared to yellow the alcohol, but when I removed the PIG mat, it looked almost clear again. But when visually compared to pure isopropyl, there was an obvious yellow tint.

IMG_20160213_150057.jpg

In any case, not enough crude dissolved into that isopropyl to get good fluorescence (see combined graph below). I wonder if soaking longer would be helpful -- it seems like it would, along with evaporatively concentrating it, since drops of crude in isopropyl did work.

Acrylic/PMMA cuvettes

A big lesson learned was that alcohol apparently melts acrylic -- the cuvettes shattered from the pressure of putting their caps in. Luckily, I had some polystyrene cuvettes, which are actually more common. We'll have to see if some of the oily samples dissolve the polystyrene... they didn't during the tests we did, but we only tested crude.

IMG_20160213_153422.jpg

I'm not sure what we'd use if we dissolved diesel in isopropyl -- the diesel would melt the polystyrene, but the isopropyl would melt the acrylic. Maybe we'd need to use glass? Anyhow, we're currently more interested in heavier residues anyhow, so we'll set that question aside for later. Here are the second set of isopropyl samples in polystyrene cuvettes (mixing was done in nalgenes):

IMG_20160213_154651.jpg

Data

Finally, we scanned some of the crude-in-isopropyl samples and got strong fluorescence. We were testing with 405nm LEDs, and this bears separate testing with lasers, but the results seem very positive. Here's an assorted set compared to one which used a laser (by @ethanbass, actually):

405nm LED input light

As mentioned, we were also testing 405nm LEDs as a possible replacement for 405nm lasers for fluorescence testing. The advantages have been laid out, so the question is whether we can observe the same red shift in heavier oil samples vs. lighter ones. And a valid concern by @stoft has been that a broader peak would lower the resolution (or detail) of any fluorescent curves we get -- we'll have to see if we can distinguish oils by weight in the way we've been able to do with lasers.

But before we get there, we just wanted to try getting decent fluorescence with LEDs at all -- just seeing if they're bright and repeatable enough in basic use. The answer to this was yes -- they're easy to use, running just off 3.3v power from an Arduino-compatible Trinket board. We didn't have to dim them much, but as @stoft has pointed out, it's best to dim the fluorescent light, not the input light, so one thing we did a bit of prototyping on was a wheel-based attenuator, which could be easier to set and use than a strip-based one. More on this soon, but as the cuvette frame project progresses, we'll try to integrate this as a feature.

IMG_20160216_140916_2.jpg

See the sample scans above -- as mentioned, we'll want to be testing LED lighting with known samples identically to the laser tests we've done, for comparison, and to test isopropyl with lasers, separating these two experiments in some more controlled tests.

I'm happy to ship folks these UV LEDs for testing -- they're easy to set up without soldering by plugging into the 3v source of a standard Arduino. And while you can find the latest cuvette frame design here (latest version in the comments), I'll be posting an update with a redesigned cuvette-holding interior soon.

Spectrum alignment in spectrometer

I need to reproduce this and document clearly, but one more note -- we noticed that in my v3 PL spectrometer, the spectrum was slightly diagonal, which isn't a problem except that since brightness varied, we'd sometimes get the bright region of the blue and green, but the dim region of the red, as the cross-section line intersected the spectrum at an angle. There are a few things we could try to do about this, but it will be an important thing to resolve if we hope to do cross-instrument comparisons of any kind. It may not have a big effect on analyses in the same device, as long as it's not affected by cuvette position -- the cuvette should be pushed all the way to the bottom of the frame, firmly.

Thanks to everyone who came to the event -- it was fun, engaging, and we made some really important steps and discoveries. Let's follow up with some rigorous replications!

Read more Follow

new-york-city spectrometer oil workshop


Report: Public Lab Workshop at Sunview

by Matej | about 8 years ago | 2 | 3

PL_Workshop_Sunview-3356.jpg PL_Workshop_Sunview-3431.jpg

What Did We Do

Saturday February February 13th we held an oil testing and oil collection workshop at the beautiful old diner, Sunview Luncheonette, in Greenpoint Brooklyn right near Newtown Creek. Adjacent to the area of one of the largest (but not widely known) oil spills in the US history. As mentioned in the invitiation. We did this in collaboration with the Newtown Creek Alliance (NCA), Sunview Luncheonette (obviously) and CLAKULA Productions (as a PR partner). Prior to the workshop we did an oil collecting field trip in Dutch Kills area of Newtown Creek, together with Willis Elkins from the North Brooklyn Boat Club and NCA. You can read more about the field trip here. For this collection we first used the Coastguard Oil Collection Kit from General Oceanic and PIG Oil Absorbent Only Mats for a larger spill. We took also a water sample from both places. Two days after we gathered at Sunview to analyze the Samples.

Our Attempt and Results

For the oil extraction we used the Freezing Method, Alcohol extraction, and mineral oil extraction both for the PIG Mats.

PL_Workshop_Sunview-3361.jpg

We used the mineral oil extraction also for the Coastguard Oil Collection Kit. Except the alcohol test all of these methods were tested before only in a “laboratory setting” with “ideal” ratios of oil and water mixture. This was the first time we tested them actually in the field.

PL_Workshop_Sunview-7100.jpg PL_Workshop_Sunview-7099.jpg PL_Workshop_Sunview-7101.jpg

Coastguard Oil Spill Kit Extraction

The freezing Method: Water samples was collected from the surface and 2” to 3” deep. this water sample contained oil sheen. The night before the workshop I split part of the sample into two 1oz plastic jars and froze them. During the workshop we use cotton swabs to collect the residuum from the surface and diluted it in 1ml of mineral oil in the cuvette. We detected changes in the measured spectra compared to just pure mineral oil. (I will post the spectra soon)

PL_Workshop_Sunview-3376.jpg PL_Workshop_Sunview-3375.jpg PL_Workshop_Sunview-3366.jpg PL_Workshop_Sunview-3367.jpg PL_Workshop_Sunview-3369.jpg PL_Workshop_Sunview-3373.jpg

Jeff tested oil dilution using 91% isopropyl alcohol on a crude oil. Please see his spectra in the comment section of this research note. This technique is very promising because it allows for concentration of the sample using evaporation techniques.

PL_Workshop_Sunview-3384.jpg PL_Workshop_Sunview-3388.jpg PL_Workshop_Sunview-3391.jpg PL_Workshop_Sunview-3392.jpg PL_Workshop_Sunview-3393.jpg PL_Workshop_Sunview-3394.jpg PL_Workshop_Sunview-3395.jpg PL_Workshop_Sunview-3396.jpg

oil dilution using 91% isopropyl alcohol

We did our spectral measurements on a new prototype of the LED cuvette frame. Presented by Jeff.

PL_Workshop_Sunview-7104.jpg PL_Workshop_Sunview-3409.jpg PL_Workshop_Sunview-3410.jpg PL_Workshop_Sunview-3414.jpg PL_Workshop_Sunview-3415.jpg PL_Workshop_Sunview-3426.jpg PL_Workshop_Sunview-3427.jpg

Questions and next steps

This is just a beginning. As a matter of fact all of the above mentioned techniques were tested for the first time in the field, and some first time ever. I kind of wish we will have more time and warmer, well ventilated environment. 4 hours was certainly not enough :-) All things we tested look very promising and exciting! We should do this productive gathering again.

Special thanks to Liz Barry for helping us organize everything and ordering testing kits!!!

PL_Workshop_Sunview-7111.jpg PL_Workshop_Sunview-7107.jpg PL_Workshop_Sunview-7110.jpg PL_Workshop_Sunview-7106.jpg PL_Workshop_Sunview-7109.jpg PL_Workshop_Sunview-7105.jpg PL_Workshop_Sunview-7103.jpg PL_Workshop_Sunview-3429.jpg PL_Workshop_Sunview-3420.jpg PL_Workshop_Sunview-140559280.jpg PL_Workshop_Sunview-124621231.jpg PL_Workshop_Sunview-3425.jpg PL_Workshop_Sunview-143854259.jpg PL_Workshop_Sunview-3431.jpg PL_Workshop_Sunview-3383.jpg PL_Workshop_Sunview-3378.jpg PL_Workshop_Sunview-3381.jpg PL_Workshop_Sunview-3382.jpg PL_Workshop_Sunview-3363.jpg PL_Workshop_Sunview-3362.jpg PL_Workshop_Sunview-3359.jpg PL_Workshop_Sunview-3358.jpg PL_Workshop_Sunview--3.jpg PL_Workshop_Sunview-.jpg PL_Workshop_Sunview--2.jpg PL_Workshop_Sunview--4.jpg

Read more Follow

new-york-city spectrometer uv oil


Collecting Oil Sheen on Newtown Creek, Dutch Kills, Queens, NY

by Matej with nameiswillis | about 8 years ago | 4 | 6

What did I do

On a cold day of .... Willis Elkins (North Brooklyn Boat Club and Newtown Creek Alliance) took me on a boat ride across Newtown Creek to look for some random oil sheens which are sometimes floating on Newtown Creek. We planned visit several places where we would have a higher chance to find them. First we decided to focus on the Dutch Kills area in Queens and than on other spots along the creek.

Shortly after we entered the Dutch Kills, we noticed several 2" - 5" thick strings of sheens which reminded us coal tar rather than oil. We collected them with our newly acquired Coast Guard Oil Collection Kit. (small net on the picture below). Than we proceeded deeper into the Dutch Kills looking for more sheens. We could not find any until we came by the 49th Ave Bridge (marked on the Google map above). Than we realized that we are floating in the middle of a huge oil sheen which spans most of that area of the creek. We did not noticed it first, because the whole water surface was a sheen so there was virtually no clean spot of water to compare and see the difference. We collected samples and returned home because we already had enough of material to test.

My attempt and results

We collected the first sheen (suspected coal tar) with Coast Guard Oil Collection Kit. (small net on the picture)

DSC08981.JPG

As the kit manual suggest we collected also a water spec from the same area to compare. (into a small bottle which came with the kit)

For the second - large sheen we used two PIG Mats and also larger glass bottle to collect water and the sheen.

DSC09006.JPG

Questions and next steps

We will attempt to analyze the collected sheens during our workshop at the Sunview Luncheonette using (also testing) the new LED Cuvette frame for our OTK, Freezing Method, and pressure methods for the PIG mats.

More pics from the trip

IMG_7020.JPG

IMG_7021.JPG

IMG_7019.JPG

IMG_7018.JPG

IMG_7017.JPG

IMG_7016.JPG

Read more Follow

new-york-city newtown-creek oil-testing-kit sheen


Announcing Spectral Workbench 2

by warren | about 8 years ago | 0 | 4

A couple weeks ago, we soft-launched Spectral Workbench 2, and today we've turned on these interfaces as the default for all users. This version has been in development for many months, and we've integrated much of the feedback you've provided starting exactly one year ago.

This release includes many exciting new features and upgrades, including:


What does this mean for you?

Most of the changes are in the Analysis interface; new ways to calibrate, to extract data from images, to manipulate, subtract, and compare spectra. Our hope is that many basic tasks will be easier, more intuitive, more flexible, and more powerful. Many new tasks are now possible as well, including wavelength range limiting, (reversible) smoothing, and as mentioned, subtraction of spectra from other spectra. These tools are built in a new, standardized user interface which is extensible -- we plan to continue adding new features as we go, building on this foundation.

So a few notes on the biggest new features -- ones which will fundamentally change how you interact with data on Spectral Workbench:

Snapshots screenshot

What are Snapshots?

Snapshots freeze a specific state of your spectrum so that the data as it exists at that moment can be used in other contexts, for example by other users. It's like version control for your data -- if you use a spectrum by reference, such as to calibrate, or in a comparison, you can expect that you'll see the snapshot of that spectrum just as it was when you made that reference (see References vs. Snapshots below). No later changes to the spectrum will affect referring work. Snapshots appear as a "thumbtack" icon in the Operations table, which we'll get to in a moment.

This is a complex feature, so we've created a thorough overview of Snapshots to help you work with them.


Operations screenshot

What are Operations?

A major change in Spectral Workbench 2 is a system called Operations (previously known as PowerTags). When you save a spectrum, you usually have to perform additional operations, sometimes setting a cross section, certainly calibrating, and sometimes then smoothing or otherwise manipulating the data. But for various reasons, you often want to record and even revert some of those changes. Maybe you want to do them differently, or to recalibrate.

Operations are the new way to do this. Each time you apply an operation, the resulting dataset is saved in a snapshot, so that a copy is archived and available in the database without having to run each calculation every time. operations typically feature a key:value pair. Creating operations is easy; the Tools pane below each spectrum guides you through creating them with a simple form, and adds your operation when you click Apply. Some operations reference other spectra, for comparison or subtraction. You'll be presented with a menu and search form to find the right spectrum to use.

Read more about Operations here


Videos

In general, I've spent a lot of time rewriting and revising the documentation for all of Spectral Workbench, and new tutorial videos are being posted:

See tutorial videos


Troubleshooting

As with any new, complex system, we expect some hiccups as this rolls out. We've massively expanded the test suite for both server-side and client-side code, which has already done much to improve reliability and catch problems before they're published.

But if you have trouble, please help us fix and improve the software by reporting issues -- we rely on your feedback to address problems!

Screenshot_2016-01-26_at_3.06.14_PM.png

Where's the code?

Do you code? The latest version of Spectral Workbench can always be found on Github at https://github.com/publiclab/spectral-workbench/

If you're interested in contributing, please check out our Developers page and our guide to contributing to Public Lab software. We need your help!


Anyhow, thanks to everyone who's helped to make this release possible; especially to @liz, @stevie, @stoft, @ygstcu, @cfastie, @justinmanley, @icarito, and the many others who dreamed up features, designed and tested them, found and solved problems, and generally helped to make Public Lab software excellent. And thanks to everyone who was patient and supportive of me and the rest of the SWB team for the past months!

Read more Follow

spectrometer software spectral-workbench developers


Oil Testing Kit Initial Electronics

by cbreuer | about 8 years ago | 7 | 4

What I Want To Do

I'm currently working on designing the electronics and code, and eventually working on Jeff's design of the casing, for the florescence oil testing kit.

My Attempt and Results

Using Autodesk's 123D Circuits, I modeled the circuit and simulated the code running on it, which all seemed fine. The circuit currently uses a full R3 Arduino, though switching to a 3V Trinket from Adafruit should have no problems. Essentially, the circuit is just a potentiometer sending values to the arduino, where the resistance value gets scaled down to being out of 255, the max brightness value. The value is then sent to the LED, which has a hardwired switch to turn the LED on and off. The circuit/code/simulation can be found here.

Code:

int potPin= A2;  
int ledPin= 9;  
int readValue;  
int writeValue; 

  void setup() {
  pinMode(potPin, INPUT);  
  pinMode(ledPin, OUTPUT); 
  Serial.begin(9600);      
}

void loop() {
 delay(1000);
 readValue = analogRead(potPin);  
 writeValue = (255./1023.) * readValue; 
 analogWrite(ledPin, writeValue);      
}

Next Steps

I just ordered a 3V Trinket and the other parts so they should be here soon, at which point I can actually wire it up and make sure it works. One possibility for the code, which I've discussed with Jeff is the option to send the brightness value on the Trinket to Spectral Workbench, in case a use for it is found. I'll probably start looking at Jeff's design for a casing and modify it to fit the electronics.

Read more Follow

spectrometer fluorescence cuvette oil-testing-kit


OTK Proto3 Analysis

by stoft | about 8 years ago | 5 | 2

Preface

In my previous Research Note 'OTK Proto3' ( https://publiclab.org/notes/stoft/01-19-2016/otk-proto3 ) I described the construction, validation, measurements and related observations for a new Proto3 OTK. With the device and matlab interface running, I proceeded with some measurements looking at hardware performance. As I continued updating the note, it became obvious that the OTK Proto3 measurement topic required it's own space simply due to the length of the note original Note. So, some of the relevant material contained in that note has been moved here.

Abstract

The OTK, which is based on the detection technique of using visual-spectrum UV fluorescence, has limitations because the fluorescence spectra of many PAH compounds are broad and so overlap. The OTK is also limited in PAH detection sensitivity by the 5 mW intensity of the 405nm UV pocket laser light and is limited in differentiation of materials containing PAH compounds by the webcam's 400-650nm detection bandwidth. This note explores the boundaries of those limits and describes one method of extracting and displaying information contained within OTK Proto3 measurement data from a small set of samples. This technique might be improved by increasing the number of reference samples. Note; Additional testing below has shown that EVO is not an ideal carrier, especially for low concentrations.

Disclaimers

[ Disclaimer: I'm an engineer, not a chemist, so I welcome additional observations and knowledge from those with a deeper understanding of the chemistry. The following simply represents my technical interpretation of OTK experimental results and the variety of information available online. ]

[ Disclaimer: The Public Lab "ND Crude" sample provided with the OTK may, or may not, actually be North Dakota crude oil. Some OTK 'crude oil' samples come from a "hair preparation product" which is claimed by that site to be "pure Pennsylvania crude oil without processing" but the provenance is unknown. The non-scientific evidence is only that it does smell like oil. If those samples are Pennsylvanian crude, it is certainly not North Dakota crude as some sample bottles are labeled in Public Lab kits, and the 'hair treatment' scam is selling the substance at $7,200 / barrel. Therefore, ALL references to ND Crude in this note are references ONLY to the label on the bottle provided with the OTK. At this time, I have no further information on its provenance. ]

Extracting Measurement Parameters

1) Curve Peaks

Considering the possibility of identifying the type of oil contaminant suggests the need to find spectral curve features to measure. Since the curve "peaks" are often noisy, the curve data is averaged and then equal-value nm points on that curve above and below the numerical 'peak' are found. The "peak" is taken as the nm point mid-way between these two values.

OTKP3_FIND_PEAK.gif

All peak-detection methods have errors based on the assumptions about the meaning of the peak's noise and distortion. For broad peaks, this method is easy to code, has been reasonably consistent and is less a function of noise than simply finding the max value. Yes, CFL spectral curve peaks are easier to acquire. Sometimes using curve averaged data is more consistent.

2) Curve Centroids

Another simple measure is to calculate the centroid of each curve. The 2D centroid can be thought of in terms of "weighing" the curve on a balance where the centroid is the (nm) point of the curve where each side has the same "weight". This wavelength might give a measure of the predominant fluorescence which is not "distorted" by the visually subjective measure of where the curve peak appears to be located. Below is a plot of the centroids of the five (5) sample oil types:

OTKP3_5-OILS_CENTROIDS.gif

Notice that there does appear to be some correlation. The list of oils is roughly in order of decreasing density and the vertical lines (of the same color) show an associated monotonic decrease in wavelength. It appears this means that as the oil molecules become "lighter" the majority fluorescence wavelength moves toward the UV. However, the actual mechanism is likely more complex -- eg. lighter oils may contain more related chemical compounds of types that fluoresce. Even so, this is an interesting correlation worth more observation.

However, an obvious cross-check is to observe the effects of dilution on the centroid which is easy by using various data from the 80W-90 dilution tests above:

OTKP3_80W_DIL_CENT_1.gif

Note that the fluorescence intensity varies with the dilution of 80W-90 in EVO but so does the centroid of the spectral peak. Since a range of dilutions from 1/1000 to 1/2 were observed (and 1/200 to 1/2 plotted above), that dilution as a percentage can be plotted against the peak center and centroid wavelengths:

OTKP3_80W_DIL_CENT_2.gif

The above plot contains an anomaly (probably just noise) and the curve is not linear but the relationship between 80W concentration in EVO and the associated fluorescence peak wavelength is clear.

Unfortunately, it appears that both oil "weight" and "concentration" affect the fluorescence peak wavelength. Some additional relationship needs to be uncovered to provide better correlation to oil type which is independent of dilution.

[ Note: The asymmetry of the broad fluorescence peak is an artifact of the peak's center location (~470nm) which is closer to the 400nm end than the 650nm end of the usable webcam bandwidth. This causes the "tail-end" of the curve at the 400nm end to effectively be cut off. As a result, the centroid is found by limiting the bounds of the calculation to the -6dB (25% of peak) either side of a curve peak's maximum level. ]

3) Oil Type vs Spectral Peak

Continuing this thought experiment, the following graphic was found on the SEOS web site. I shows the fluorescence spectra of a number of oil types when stimulated by 308nm UV laser light. The first thing to realize is that the OTK is only able to detect the central 400-650nm band -- which is primarily why the spectral curve shapes shown here look so different.

OTKP3_OIL_IN_WATER.gif

Notice that the UV-side (shorter wavelengths) of the curves will be "rolled-off" by the webcam's 400nm limit. However, the webcam's 650nm limit may allow the extraction of more information about oil type; i.e. perhaps the concentration of PAHs, and thereby, more information about oil 'weight' or type.

Another interesting plot from the web [Source: The Milan Unit of IENI (Institute for ENergetics and Interphases)] is shown below. Note that the 400-650nm detection band of the webcam indicates that the OTK will primarily be detecting fluorescence from PAH compounds with molecular weights greater than the common solvent benzene.

OTKP3_PAH_VS_WAVELENGTH.gif

Looking closer, one of the higher concentration (4.36 ug/ml) PAH compounds listed for used motor oil is Fluoranthene. With a density of 1.25x water we might guess at a concentration of ~3.5 ppm. While low, the published spectrum from a 288nm UV excitation source provides an interesting example:

OTKP3_FLUORANTHENE.gif

The plot results from analysis by a gas chromatography so there is obviously much more detail than provided by an OTK. What the plot shows is that the fluorescence spectra of individual PAH compounds found in oil (and there is a long list of them) will easily obscure the details of each other. This means that the only data which can be extracted from the OTK spectra will be the average of many PAH compounds (most of which will be large molecules) and can only suggest a general shift between lighter and heavier PAH compounds as the predominant indicator of the primary contaminant type.

4) Crude Oil Spectra

Each oil type appears to have its own unique "spectral signature" features yet since different oil types (including motor oils, etc.) can have similar features. This means that while each oil has its own spectra, it is not so easy to translate a spectra back to an oil type. The following plot from HyperVolatility.com shows the overall scale of crude oil in terms of weight vs sulfur content. It does not relate the data to fluorescence spectra but it does put measuring "crude oil" in perspective.

OTKP3_CRUDE_OIL_GRADES.gif

Going back to the fluorescence measurement, the following plot shows fluorescence spectra for various weight for Light (left) and Medium(right) North Sea crude oil:

OTKP3_CRUDE_OIL_LT_VS_MED.gif

While the differences are relatively small, there are some visible correlations. 1) Medium crude tends to have more high-carbon PAH compounds (the "bulge" on the right side of the curve on the right) and 2) Light crude tends to have relatively few high-carbon PAH compounds no matter the oil weight. While these are generalizations, they may be useful.

OTK Proto3 Spectra Analysis

Now, understanding just a bit more about the many relationships between oils and UV fluorescence spectra, can any of these correlations be used as a basis for indicating the chemical type from the OTK spectral data?

First, consider the dynamic range of "signal intensity" associated with each sample type; 90 weight motor oil has a very strong signature relative to auto diesel.

OTKP3_10PCT_6TYPES_FIXED_SCALED.gif

In the above plot, all chemicals have been diluted to 10% but to be able to display a measurable signature, the fluorescence light was attenuated at #160 for 90 weight oil and #00 for diesel. (The attenuation numbers fall roughly in the ramge from 0-255 as a range from clear to black as inkjet sheet film.) As this demonstrates, the dynamic range is very high; which also means the detect ability will also be compound type-dependent.

So, leaving this sensitivity issue aside so as to reduce the variables, the focus becomes analyzing spectral curve shapes. To do this, each of these samples, and the selectable attenuation, were adjusted to minimize the measurement dynamic range. ( This means that large signals are attenuated to better match with weak signals. There is no dilution so as to avoid that orthogonal issue [see below]) In addition, the resulting curves were all normalized to the same peak intensity -- which was declared to be a unitless value of 100 for plotting and analysis. The result is shown below:

OTKP3_6-SAMPLE_SPECTRA.gif

By adjusting concentrations and attenuation, the noise has been reduced sufficiently to perform run-length averaging before extracting curve parameters.

The chosen parameters of interest include: curve peak nm, curve centroid nm and curve -6dB bandwidth (wavelengths at ~25% of peak). Now the challenge is to find a way to easily plot the combined effect and give meaning to the results. After a number of data plotting experiments, the following plot developed:

OTKP3_6-SAMPLE_ANALYSIS.gif

All plot values are based on using the 80W-90 spectra as a 'center reference'. This oil sample was selected because 1) it has a strong signature, 2) it's spectra is roughly in the middle of the spectra of the other samples and 3) its spectra is relatively symmetrical about the peak. Also, all plot values are difference values from this reference, not absolute values. (Note: The colors of this plot and the previous plot both reference the same oil types.)

  • The 'X' axis plots the curve peak nm, relative to the 80W-90 reference

  • The 'Y' axis plots the curve centroid nm, relative to the 80W-90 reference

  • The relative size of each color square is made linearly related to the -6dB bandwidth of the associated curve

  • The lower-left corner of the plot will tend to have "lighter" oils and the upper-right corner of the plot will tend to have "heavier" oils

[ NOTE: The Yellow plot is not an oil -- it is a diluted sample of TideHD laundry detergent. It has a very strong signature and although it it appears at the "Light" end of the plot, its OTK Proto3 UV fluorescence spectra is not very different from Diesel oil. It is a valuable sample to include simply because it provides a sanity check on assumptions when looking at the spectra. ]

The thought behind this final plot was to demonstrates two ideas:

1) Oil type analysis is very, very difficult

2) There might be some value in measuring a wider range of UV-detectable contaminants which might "fill-out" this plot and provide some basis for allowing OTK spectra to "suggest" possible candidates for the contaminant which has been detected.

Dilution Effects

While the above plots show some promise, it is always necessary to keep pushing the boundaries of a theory and look for exceptions. In the prior Research Note 'OTK Proto3' ( https://publiclab.org/notes/stoft/01-19-2016/otk-proto3 ) I showed an example of the effects of dilution on the Fluorescence spectra. That same dilution data is shown below, except that all of the curves are auto-scaled to normalize the peak value:

OTKP3_80W-90-DIL_SPECTRA.gif

That data is then re-plotted in the same analysis format to observe the Peak vs Centroid vs Bandwidth effects of dilution:

OTKP3_80W-90-DIL_ANALYSIS.gif

The hopeful trend is that the four points are monotonic; as the 80W-90 oils is diluted, the spectra always shifts in the same direction. The difficulty is that this data suggests a strong dependency on concentration which is not well understood and which makes it harder to identify the difference between the molecular weight of the PAH compounds present in the sample and the oil type. One thought is that dilution of 80W-90 in EVO actually affects the chemistry which results in a shift in the PAH concentrations; but that's just a guess.

Opacity

It's always good to keep testing the boundaries so here's a look at the light filtration effects through the cuvettes and EVO as a carrier. (This uses a broadband Solux 4700K halogen bulb passing light through the cuvette; not using the OTK's UV laser as with the other measurements shown here.) The plot below shows (Black) a Solux source with no cuvette, (Red) shows a cuvette with water, and (Lt Blue) a cuvette with EVO:

OTKP3_SOLUX_EVO_FILTER.gif

Clearly, the cuvette and water have relatively no effect on the light path but the EVO does filter the spectrum, resulting in ~3dB attenuation at 500nm. This was not anticipated and so it appears that EVO is not an ideal medium -- but it was an easy material to use. However, it is likely that for observing UV fluorescence glow from within the cuvette, this effect would not be as serious.

The next plot looks at the effects of dilution for the purpose of observing any wavelength shifts which might be related to concentration. The biggest effect is simply attenuation when the concentration is high. ( Yes, there is a testing gap from 2% to 100% as those samples were not pre-mixed at the time. The only impact is knowing where the attenuation becomes significant; which is of less importance to this observation. )

OTKP3_SOLUX_EVO_DILUTE-80W-90.gif

The interesting thing to notice is that the 80W-90 oil, even at low concentrations, seems to provide some improvement in terms of the high-pass filtering effect as the attenuation is not so pronounced at 500nm.

[ Note: A separate project to identify the best candidate for a carrier fluid for testing low concentrations, where the carrier has no to minimal effect on the transmission spectra should be done. While observing fluorescence from within the fluid in the cuvette is not the same as passing a light source through, there is likely still some filtering effects from the fluid composition. They should be identified to ensure their impacts are limited or characterized. ]

Detection Threshold

The general issue of measuring oil as a contaminant begs the question of measurement sensitivity; the limits of detectability. It was shown above that EVO spectra does not display any measurable fluorescence so is used for dilution of the 80W-90 oil. One drop of the 80W-90 "contaminant oil" was added to in increasing volume of EVO. The results are plotted below:

OTKP3_80W_DETECT_LIMIT_NO-ATTEN.gif

Note that with all of the plotted spectra, no attenuation was used so these observations are intended to show the threshold of detectability of the 80W-90 sample, given the Beta OTK's 5mW 405nm UV pocket laser. The last plot, in Green, is visually just above the noise floor in the 425-475nm band which is the approximate peak of the 80W-90 spectral response for higher concentrations. Clearly, a 1:1000 dilution ratio for 80W-90 represents the detectability threshold under these ideal conditions.

[ NOTE: It should be clear from the previous measurements designed to analyze the type of compound being detected, that the detection threshold should be expected to be quite different for each of the oil and other contaminant types.]

Conclusions

  • A simple 1:1 relationship between any single measurement parameter (like spectral peak wavelength) and oil type does not appear to exist -- because other characteristics of the fluorescent "peak" also change.

  • From both online information, and OTK measurements, there does appear to be a rough correlation between the molecular size of PAH compounds and the wavelength of that compound's fluorescent spectral energy. ( Very generally, "narrow" peaks toward the UV tend to correlates with "lighter" oils and "more broad" peaks toward longer wavelengths tend to correlate with "heavier" oils.

  • The OTK detectable range of wavelengths very roughly correlates with the PAH compounds classified by "carbon weight": C22 - C40.

  • Combining observable curve parameters can shed a little light on some of the relationships between oil type, PAH content and UV fluorescence spectra.

  • There does appear to be a shift in the peak center fluorescence wavelength which is related to the dilution ratio (in addition to the strength of the signal). If it were exclusively attenuation, I'd expect the peak nm to remain the same. Perhaps it is related to optical filtering of the spectra ( assuming such an effect is essentially "rolling-off" the lower wavelengths [a High-Pass filter])?

  • The use of EVO as a carrier medium appears less than ideal as it causes some high-pass filtering effects on the signals which, especially for low concentrations, will likely have a negative effect on accurate observation of PAH fluorescence spectral data at the shorter wavelengths. However, this effect may be limited for observing fluorescent glow coming from within the cuvette.

  • While the Solux gain-correction is useful, more work is needed to make it easier to provide accurate correction data.

  • Adding more "reference" sample spectral data might provide better guesses about the nature of an unknown sample with detectable UV fluorescence.

Read more Follow

spectrometer oil laser oil-testing-kit