Laser-Induced Breakdown Spectroscopy for Analysis of Ambient Air Particles and Lunar Dust Simulants: Consideration of Statistical Methods and Pressure Effects

Material Information

Laser-Induced Breakdown Spectroscopy for Analysis of Ambient Air Particles and Lunar Dust Simulants: Consideration of Statistical Methods and Pressure Effects
Loper, Kristofer
Publication Date:


Subjects / Keywords:
Aerosols ( jstor )
Base composition ( jstor )
Cadmium ( jstor )
Calcium ( jstor )
Emission spectra ( jstor )
Particle emission ( jstor )
Plasmas ( jstor )
Signals ( jstor )
Standard deviation ( jstor )
Vacuum ( jstor )
Laser-induced breakdown spectroscopy
Lunar dust
Undergraduate Honors Thesis


This report focuses on Laser-Induced Breakdown Spectroscopy (LIBS) as an analysis tool for detection of harmful particulate matter. It also focuses on different statistical methods for processing aerosol data and the effects of pressure on the analysis of aerosol emission signals. LIBS was proven to be a successful analytical tool for detecting lead, titanium, chromium, iron, and cadmium in atmospheric air during this experiment. It would prove as a useful tool when trying to determine composition and concentration of expelled gases from energy production facilities, which are responsible for a majority of harmful particulate matter. Two analytical methods were compared during this report: conditional analysis and standard deviation analysis. Standard deviation analysis works particularly well for identifying the composition of an unknown aerosol sample, whereas, conditional analysis works well for determining the concentrations of elements in an aerosol sample. Two pressures were compared during this report: atmospheric pressure (0 in Hg gage pressure) and near vacuum pressure (-28 in Hg gage pressure). The near vacuum spectral signal provided stronger and sharper emission peaks than the atmospheric spectral signal resulting in more sensitivity and accuracy. However, the construction of a vacuum environment is time consuming and difficult. The gains acquired from vacuum pressure analysis don’t outweigh the losses experienced when deviating from the original atmospheric pressure analysis. ( en )
General Note:
Awarded Bachelor of Science in Mechanical Engineering; Graduated May 4, 2010 summa cum laude. Major: Mechanical Engineering
General Note:
Awarded Bachelor of Science in Aerospace Engineering; Graduated May 4, 2010 cum laude. Major: Aerospace Engineering
General Note:
Advisor: David W. Hahn
General Note:
College/School: College of Engineering
General Note:
Legacy honors title: Only abstract available from former Honors Program sponsored database.

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
Copyright Kristofer Loper. Permission granted to the University of Florida to digitize, archive and distribute this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.


This item is only available as the following downloads:

Full Text


Laser Induced Breakdown Spectroscopy for Analysis of Ambient Air Particles and Lunar Dust Simulants: Consideration of Statistical Methods and Pressure Effects KRISTOFER HOWELL LOPER SPRING 2010 SUMMA CUM LAUDE BACHELOR OF SCIENCE IN MECHANICAL ENGINEERING


1. Introduction Laser Induced Breakdown Spec troscopy, or LIBS, is an atomic emission analysis technique which may be used for aerosol analysis and which ha s been explained in g reat detail in previous reports. 1 5 LIBS uses optics to focus a pulsed laser beam to produce laser induced plasma. The volume of this plasma envelopes nearby molecules and fine particles and excites them. The excited molecules and particles then emit emission lines corresponding to the elements present in the molecules and particles. This emission can be collected spectrally and an alyzed to determine the composition and concentrations of elements in the sampled aeroso l Over the last decade, the discussion about airborne particulate matter as it relates to human health has intensified. Some of the se discussions focus on the ill effects that particulate matter less than 2.5 m in diameter (PM 2.5 ) have on human health According to the report 2.5 emissions are from human created sources while the other half are from natural sources. Most anthropogenic PM 2.5 emissions are the result from energy production. Since people are exposed to PM 2.5 on a continual basis, it is important to be able to detect harmful PM 2.5 6 This type of anal ysis tool could be used at energy production facilities, such as power plants to measure the amount of harmful PM 2.5 released into the atmosphere T his knowledge would aide greatly in the management of air quality. For this report, six toxic elements were targeted for analysis : chromium, titanium, cadmium, lead, manganese, and v anadium. These elements were chosen based on their toxicity and location of atomic emission wavelengths, which will be discussed in later sections. Chromium uptake is possible thro ugh inhalation or ingestion Its toxicity is dependent on its oxidation state. In it s most toxic state chromium can cause respiratory problems, kidney and liver damage, and in rare cases, lung cancer and even death. Titanium has a low order of toxicity and is clas sified within Group 3, an agent that is not classifiable as to its carcinogenicity to humans, by the International Agency for Research on Cancer. However, overexposure to titanium may cause coughing, tightness in chest, and difficulty in breathing. Cadmium is prevalent in tobacco smoke which transports the cadmium directly to the lungs. Cadmium can severely damage the lungs as well as t he central nervous system and immune system. Lead is one out of four specific elements that cause the most damage to human health. As with cadmium, lead is present in small amounts in cigarette smoke. Lead can cause many ill effects even severe damage to the brain and nervo us system of an unborn child. Manganese is an element that is essential to humans in v ery small amounts. W hen the uptake of manganese is too high however, it becomes toxic and causes serious health issues. These health issues include hallucinations, nerve damage, Parkinson, lung embolism and bronchitis. Workers exposed to dust particles of manganese compounds ov er extended periods of time showed


incidences of pneumonia and upper respiratory infections. V like chromium, is dependent on its oxidation state. Vanadium is not regarded as a severely toxic element; however, overexposure to vanadium dust can cause many ailments including eye, nose, and throat irritation, pneumonia, bronchitis, damage to the nervous system and bleeding of the liver and kidneys 7 The second part of this report compares two statistical methods for data p rocessing. In previous reports the two methods used to analyze LIBS aerosol data are the conditional analysis method and the standard deviation analysis method 2 4 Con ditional analysis is explained in shot spectral analysis by use of laser explained below. For conditio nal analysis each re corded spectrum is analyzed by comparing the emission intensity of the expected analyte peak to the emission intensity of a spectral region (peak to base ratio ) is greater than a defined threshold than that shot is considered a particle hit. In earlier reports this threshold was typically set to provide 0 .1% to 0.5 % false hits. A false hit is when the peak to base ratio is greater than the threshold value b ut there is no targeted analyte. This is done so that weak, but true analyte signals, which can be compared to fluctuating noise peaks, are not eliminated; however, this does result in a significant amount of false hits. Once every shot is processed throug h this step every particle hit is analyzed further to determine if it is a true hit or a false hit. This process uses additional nearby emission lines of the same analyte to compare to the particle hit spectrum. If a true hit is present, the particle spec trum will exhibit peaks at both emission lines of the analyte; whereas a false hit is statistically unlikely to exhibit a peak at the second emission line of the analyte 2 In 2008, L. A. Alvarez Trujillo et al. wrote a report about another method for proce ssing LIBS data called Induced implements the standard deviation of a set of data. For a sp ecified number of shots the s tandard deviation is calculated for each pixel, or wavelength division. If all shots at a specified pixel have relatively the same emission intensity then the standard deviation at this pixel would be small. However, if one or two shots have a much greater emission intensity (a.k.a. a particle hit) while all the other shots are relatively the same, the standard deviation at this pixel would i ncrease significantly. Once the standard deviations for each pixel are calculated they are plotted against the corresponding wavelength s for analysis This results in something that resembles an actual spectrum, but the emission intensity is replaced by the variation in intensity. 5 Examples of both analytical methods discussed above are shown and discussed in later sections. The third part of this report looks into the effects of different pressures on the analysis of a moon dust s imulant. Obviously, m oon dust is located on the moon and the moon exhibits an


atmospheric pressure of 3 10 15 bar which is essentially zero, or in other words, the moon dust is located in a vacuum. The purpose of this part of the report was to determine if an analysis using LIBS conducted in a vacuum environment differs from the analysis conducted at atmospheric pressure for a moon dust simulant 2. Experimental Methods Laser Induced Breakdown Spectroscopy Analysis. The LIBS syst em has been described in great detail in earlier work s 1 5 so only a brief explanation is supplied here. The laser induced plasma was produced by a Nd: YAG Las er with a 10 ns pulse width, 375 mJ pulse energy and a 5 Hz pulse frequency The beam is focused into a point using a 75 mm focal length UV grade lens resulting in a hig h energy plasma The plasma emission was collected using a pierced mirror and a fiber coupled to a 0.275 m spectrometer with a 2400 groove/mm grating and recorded with an intensified, charge coupled device (iCCD). The integration delay and width ti mes used were both 40 s with respect to the laser pulse (t = 0 s) Sampling System. For the first part of this project, atmospheric a ir was drawn in through a PM 2.5 inlet that removed all particles greater than 2.5 m in diameter Then the air flows through the sampling chamber at 16.7 liters/min (~ 1 m 3 /hr) The schematic for this system is shown in Figure 1. T his experiment was conducted in Gainesville, Florida on the University of The sampling inlet was located on a second story window sill of a three story building. The location below the inlet is a popular smoking section for smokers and t he adjacent lawn underwent landscaping proje cts, which utilized some mechanical machines, during the data acquisition. Data acquisition took place between January 19, 2010 and March 23, 2010 for a total of 7 individual 2 hour sampling period s The conditional analysis method collected 9 sets of 1500 shots while the standa rd de viation method collected 40 set s of 250 shot s This resulted in 23 ,500 spectra per sampling per iod. Since there were 7 sampl ing periods a total of 164 ,500 spectra were collected and analyzed. For the moon dust part of the project, the simulant was analyzed in a vacuum chamber, so overcome this roadblock a small motor wa s used to rotate the sampling chamber about the laser beam axis so that gravity would stir the moon dust simulant into a fine dust cloud. This rotation allowed the fine dust cloud to interact with the plasma resulting from the focused las er beam.


Figure 1 : Diagram of the ambient air sampling system. This system includes a PM2.5 inlet that removes all particles greater than 2.5 m in diameter. It also shows the sampling chamber where the ambient air interacts with the laser induced plasma. Detection of Toxic Elements in Ambient Air The following experimental methods are specific to the analysis of toxic elements. The first step for this experiment was to c hoose appropriate elements to analyze. In previous experiments 1 3 only one e lement was analyzed at any one time This process would result in a very long acquisition time if multiple elements were being analyzed. T herefore, mult iple elements that have emiss ion lines close to one another were chosen so that the spectrometer window range could remain stationary This process led to a spectral sampling window cent ered at 371.5 nm and a r ange between 354.2 nm and 387.6 nm. The six toxic element s that fell into t his spectral range were mentioned before as well as some of their characteristics i n the introduction. Chromium had three strong peaks at 357.9, 3 59.3, and 360.5 nm. Titanium had its strongest pe a k at 368.5 nm and two additional peaks at 375.9 and 376.1 nm. Cadmium, unfortunately had only one peak within thi s range at 361.1 nm. This proved troublesome when dete rmining whether a particle hit for cadmium wa s a t rue hit or a false hi t. Lead had its strongest peak at 368.3 nm and a n additional peak at 364.0 nm. Manganese had its stronges t peak at 356.9 n m and additional peaks at 381.0, 382.4, and 383.4 nm. Vanadium had its strongest peak at 370.4 nm and a strong second peak at 385.6 nm All of the reference spectra were found using liquid solutions of each toxic element These liquid solutions were tested in a concentration of ~ 5000 g/mL of de ionized water. This liquid solution was nebulized an d pumped directly into the sampling chambe r Figure 2 t hrough Figure 7 show the spectrum signal of each element versus the spectrum signal of de ionized water For the measurements, the element concentration was about 15 g/L of gas, or about 15 ppm (mass), at the point of the LIBS plasma.


Figure s 2 7: Resultant spectra for each toxic element. Each spectrum is compared to the spectr um of pure de ionized water. The liquid solution of the toxic element s had a conc entration of ~ 5000 g/mL of de ionized water.


Based on the figures target emission peaks were chosen for each element to compare to two baseline wavelengths of 362.5 and 366.9 nm These baseline wavelengths are located where none of the element emission peaks will affect the signal. The next step was setting the threshold values for each element. This wa s an iterative process and for each iteration there were 1000 shots collected. From these shots the thresholds were adjusted and then 1000 more shots were collected and the process repeated The goal was to obtain 1 to 5 hits for each element for every 1000 shots which results in 0.1% to 0.5% hit rate After sixteen iterations these goals were met and the thresholds were set. Finally, the data was collected Unfo rtunately, the wavele ngths each analyte emission peak was located on in from 8 In order to correct for this mismatch a calibration was applied to the wa velengths as func tions of pixel numbers For this calibration the pixel that each emission peak was located on in the spectral window wa s recorded. Then the 8 were also recorded. The wavelengths were then plotted as a function of pi xel number. A second order curve was fit to the data and the coefficients for this curve are displayed in Figure 8. The correlation coefficient, R, of 0.999978 suggests that the error between the calculated wavelengths as functio ns of pixel number and the NIST Atomic Database wavelength s is no more than 0.05 nm. Figure 8: Calibration curve relating wavelengths, acquired from NIST Atomic Database, to the pixel number, acquired from the spectral window of the analysis program. 8


As mentioned before, seven sampling periods were taken between the dates of January 19, 2010 and March 23, 2010. The condition al analysis method collected 9 sets of 1500 shots and th e standard deviation method collected 40 sets of 25 0 shots. The results from all of these Comparison of Statistical Methods for Data Processing. The experimental methods for this part of the project are similar to the experimental methods used in the t of the project but with a few differences. Instead of using toxic elements as the target analytes, calcium was used as the target analyte. The location and strength of two of strongest emission line s have been well documented in previous reports to occur at 39 3. 37 and 396.85 nm 3 5 T hese reports also made it apparent that calcium was a relatively common element found in atmospheric air, about a 0.3% hit rate. The spectral window was set at a center wavelength of 406.00 nm. 100 sets of data, containi ng 250 shots each, were collected for this part of the project, totaling 25 000 shots. Assuming the hit rate of 0.3% would hold we expected to see about 75 calcium hits. Unlike the first part of this project, each set of data underwent a standard deviation analysis as well as a conditional analysis. This allowed for a direct comparison between the two analysis methods since the same data was used for both methods. For con zero value. This was calculated by averaging the first and last pixels of data. The next step was the difference between a region without an analyte emission signal and the offset of the spectral data. The location of the non analyte emission signals were chosen at 401.25 and 406.83 nm. This location was selected because it is una ffected by any expecte d analyte emission signal s represents the difference between a n analyte emission signal and the offset of the spectral data. The location of the analyte signals are located on the expec ted calcium emission lines of 39 3.37 and 396.85 nm. These v alues are rep resented graphically in Figure 9 to base ratios for each calcium peak were calculated. possi bility of it messing up the peak to base ratio calculations. Threshold values were then set base d on these peak to base ratios. The results of this analysis will be discussed further in the


Figure 9 : Graphical pixels used to calculate the averag e value of the respective variables They are color coordinated for clarity as to which lines apply to which value. The above plot is that of the reference calcium emission line. Pressure Effects on the Analysis of Lunar Dust Simulant. The exper imental comditio ns for this part of the project are significantly different than the other two. The biggest difference is that the medium we are analyzing is no longer atmospheric air. Inst ead, our medium i s the lunar dust simulant JSC 1 under reduced pressure The composition of t he simulant and actua l lunar dust obtain ed from an Apollo mission, are shown in Table 1 9 The most prevalent elements seen in these samples are silicon, aluminum, calcium, iron, and magnesium.


Table 1 : Composition of Lunar Soi l collected during an Apollo mission and a Lunar Soil Simulant 9 The delay time used for this experiment was 7 s, the integration time was 4 s and the laser emitted 200 mJ per pulse of energy. Two spectral windows were used for this analysis: one centered at 325 nm and the other centered at 450 nm. At each of these w indows 3 sets of data were collected, containing 50 shots each, for two different pressures. The first pressure pressure. T he second pressure 28 in Hg gage pressure. The problem of dispersing the lunar dust simulant in a vacuum environment was solved by simply rotating the testing vessel and allowing gravitational forces to disperse the simu lant. The sampling chamber was a six way stainless steel vacuum cross, as seen in Figure 1. The chamber was attached to a motor that rotated it about the axis of the laser beam. To make sure the simulant was properly dispersed mess screens were installed into the four cross paths perpendicular to the laser beam. At th e start of the experiment 280 mg of the JSC 1 Lunar Dust Simulant were placed into the sampling chamber and then the analyzing began. The results of 3. Results and Discussions Dete ction of Toxic Elements. As mentioned previously, the conditional analysis method inc luded 9 se ts of 1500 shots resulting in 13500 total shots. Multiply this by the 7 sampling periods and a total of 94500 shots were taken using the conditional analysis met hod. Out of these shots only six confirmed particle hits were found. Five of these confirmed hits were cadmium hits. Since cadmium only has the one e mission peak within our spectrum window it was difficult to determine whether or not a cadmium hit were a t rue hit or a false hit. To distinguish between the two, a simple method was developed Figure 10 shows a true cadmium hit and Figure 11 shows a false cadmium hit. The difference between the two is that three of JSC 1 Simulant Lunar Soil Oxide Conc. (Wt %) Conc. (Wt %) SiO 2 47.71 47.30 TiO 2 1.59 1.60 Al 2 O 3 15.02 17.80 Fe 2 O 3 3.44 0.00 FeO 7.35 10.50 MgO 9.01 9.60 CaO 10.42 11.40


the data points for the true cadmium hit lie significantly above the noisy baseline signal; whereas, the false cadmium hit only has one data point that lies above the noisy baseline signal. In short, this method states that if three of the data points from the cadmium emission peak lie above a ce rtain threshold, that peak is considered a true cadmium hit. Or in other words, if the cadmium hit peak is comparable to other noisy peaks in the data (Figure 11) then that cadmium peak is a false hit. This method is not based on numerical calculations, bu t rather on a comparison between the emission peak and the noisy baseline signal. Figure 10: This is a plot of a true cadmium hit. The red line is the cadmium reference spectrum and the blue line is the cadmium hit spectrum. The thick black line repres ents the threshold that three of the data points must be above for this to be considered a true hit. This cadmium hit meets that requirement. Figure 11: This is a plot of a false cadmium hit. The red line is the cadmium reference spectrum and the b lue line is the cadmium hit spectrum. The thick black line represents the threshold that three of the data points must be above for this to b e considered a true hit. This cadmium hit does not meet that requirement.


Using this method, five true cadmium hits were observed. This results in a cadmium hit rate of ~ 0.0 05 %. Comparing to calcium, which has a hit rate of 0.3% in atmospheric air 3 5 cadmium is not present in dangerous concentrations in the atmospheric air. Cadmium did show up more that the other toxic element s however. This may be due to two reasons: One, the location of the sampling inlet is situated above a popular smoking area and, as was mentioned in the introduction, cadmium is present in tobacco smoke ; or two, the method used to classify the cadmium hits as true hits is insufficient These results were based on a non numerical analysis. In order to obtain clear cut conclusions about the prevalence of cadmium in atmospheric air an analysis using a spectral window with two cadmium emission peaks would be appropriate. The other hit observed during the conditional analysis method was a manganese hit. However, after analyzing this shot it was apparent that it was a false hit for manganese but a confir med hit for iron Figure 1 2 shows the shot that was collected as a hit for manganese. The blue dots represent scaled intensity peaks of the iron emission spectrum acquired from the NIST Database 8 The reason t his came across as a manganese hit is because the first iron emission peak is very near the targeted manganese emission peak. Since they were so close together the iron peak brought the peak to base ratio above the necessary threshold and it is was collec ted as a manganese hit. Figure 12: This is the shot of a false manganese hit, but a confirmed iron hit. The blue dots represent the scaled emission peaks of the iron spectrum. This was confirmed an iron hit because an overwhelming number of emission peaks match up with the emission peaks of iron, which were acquired from the NIST Database.


The standard deviation analysis method collected 40 loops of 250 shots each resulting in 10000 total shots. Multiply this by 7 sampling periods and a total of 70000 shots were taken using the standard deviation analysis method. Out of these shots only one particle was found. This particle included elements such as iron, magnesium, chromium, lead, and titanium. Each element was proven to exist inside this particle by matching up the emission peaks of the particle shot to emission peaks of the aforementioned elements. These plots are shown in Figures 13 18. This particle is concluded to be some type of metal alloy that was possibly released by the small mechanical landscaping machines that were active near the sampling inlet during the data acquisition. Had this shot have been analyzed by the conditional analysis method, there is no doubt that it would have triggered a hit for chromium, titanium, lead, and maybe even a false manganese hit again. This single particle hit pr oves that LIBS is a successful analy sis tool when it comes to detecting four out of the six targeted toxic element s in atmospheric air Manganese and vanadium were ne ver observed during the experiment It also proves that this local area of Gainesville, Florida does not have dangerous concen trations of the six targeted toxic elements Figure 1 3 : This plot shows the particle hit spectrum versus five elements that make up part of the composition of this particle.


Figures 1 4 18: Various plots of the particle shot versus individual elementspectrums expected to be in the particle.


Comparison of Statistical Methods for Data Processing. During the dat a acquisition three sets of data were insufficient to analysis. Therefore, a total of 97 sets of data with 250 shots each were collected. This is a total of 24250 shots. 32 calcium hits were observed out of these shots, which correspond to a 0.132% hit rate. Just by reviewing the calculations required for each analysis method displays a significant difference. Conditional analysis (CA) takes every single shot, calculates the peak to base ratio, determines if it classifies as a hit, and then stores that information. Standard deviati on analysis (SDA) requires all 250 shots to calculate the standard deviation. Just from this it is apparent that CA can determine how many hits have been collected. It would take a considerable amount of extra analysis and time to figure out how many hits were collected using SDA. Sensitivity is also affected by the calculations required for each method. CA is a lot more sensitive to particle hits than S DA because its calculations depend on one singular shot. It was mentioned above that peak to base ratios were calculated, for CA, to determine whether or not a shot was considered a hit based on a set threshold. This threshold turned out to be around 1.5. In other words, if the peak to base ratios for both calcium emission peaks were above 1.5, that shot wou ld be considered a calcium hit. This threshold allows even weak calcium shots to be detected, as in Figure 19 where the peak to base rat ios are only 1.54 and 1.59. I t is still apparent that the calcium emission peaks are there. Fi gure 19: Plot of a particle hit versus the calcium reference spectrum. The red line is the reference spectrum and the blue line is the true calcium hit. The peak to base ratios of the first and second peaks were 1.54 and 1.59, respectively.


The sensitiv ity of SDA is very poor compared to CA. If the peak to base ratios are very big for a particle hit, then SDA does pick up on the particle hit, as seen with the peak to base ratios of 16.1 and 11.1 in Figure 20. As the peak to base ratios decrease, however, the sensitivity of SDA decreases until it no longer even registers weak particle hits. This is displayed in Figures 21 and 22. With peak to base ratios of 3.5 and 2.9 on the calcium hit the SDA barely shows the peaks. When the peak to base ratios drop to 2 .7 and 2.4 on the calcium hit the SDA even register the peaks. This occurs because the other calcium hit s to stand out To ob tain better sensitivity with SDA the sample size woul overcome by the non hit spectrum shots Figure 20: Plot of a true calcium hit versus the stan dard deviation spectrum for the corresponding data set. The peak to base ratios of the calcium hit emission peaks were 16.1 and 11.1


Figure 21: Plot of a true calcium hit versus the standard deviation spectrum for the corresponding data set. The peak to base ratios of the calcium hit emission peaks were 3.5 and 2.9. The standard deviation barely shows the calcium peaks resulting from the calcium hit. Figure 22: Plot of a true calcium hit versus the standard deviation spectrum for the c orresponding data set. The peak to base ng from the calcium hit.


Another big difference between the two analytical methods is the fact that for CA the SDA the location of the analyte is not necessary. If the composition of an aerosol is unknown SDA would For a lot of applications the concentration of an element that is present in an aer osol is desired. CA has been proven to supply adequate concentration c alculations of a calcium sample 3 A calibration analysis was conducted to supply a calcium concentration based solely on the peak to base ratios of the calcium hits. These calibration curves resulted in a very linear distribution, meaning the calcium conce ntrations were directly related to the strength of the peak to base ratios of the calcium hits. A calibration analysis was not conducted for this experiment. Instead, both methods underwent peak to base ratio calculations about the expected calcium emissio n peaks. These calculations were then compared to one another. The desired relationship between the two analytical methods is linear. A highly linear relationship would mean that the SDA is just as effective cal culating concentrations than CA. Figure 23 sh ows the relationships between the two methods. The plot shows a general linear trend, however, it is not highly linear with a c orrelation constant, R, of 0.95992 This relationship suggests the SDA is not accurate enough to predict analyte concentrations. When CA collected zero hits a peak to base ratio of zero was calculated. For the same zero hit peak to base ratios ranging from 0.4 to 0.05. This broad range of ratios would result in a broad range of false analyte concentrations. SDA also calculated negative peak to base ratios for some weak calcium hits. The conclusion from this comparison is that CA is a better tool when quantifying the concentration of an analyte. The algorithm used for SDA is not accurate or sensitive enough to pro vide acceptable peak to base ratio calculations. Figure 23: Relationship between the peak to base ratios of CA and SDA. There is a general linear trend in the relationship resulting in a correlation coefficient of 0.95992 However, SDA loses sensitivity for very weak calcium hits and shows a lot of variability when the expected peak to base ratio is zero.


Pressure Effects on the Analysis of Lunar Dust Simulant. In a vacuum it was expected that the laser induced plasma wouldn sampling chamber. A true vacuum is equivalent to 29.9 in Hg gage pressure. Our system was able to obtain a m inimum gage pressure of 28 in Hg. Therefore, the laser induced plasma was still visible. Initially, the lunar dust simulant was left out of the sampling chamber to analyze differences in the emission spectrum. One of the biggest differences between the t wo pressures was audible. As this plasma is created it emit brought d own to a near vacuum The emission intensities showed the same trend when analyzing the spectral emi ssion signal. Figure 24 compares the spectral emission signal between the two pressures. These signals represent the sum of the intensities of 50 shots of data. The intensity sums for the near vacuum pressure are two orders of magnitude less than the atmos pheric pressure sample. These two observations suggest that the energy of the laser induced plasma is significantly smaller in the vacuum environment due to the reduced contribution of gaseous species to the plasma The near vacuum signal also had sharper/steeper peaks compared to the atmospheric pressure signal. When the lunar dust simulant was introduced into the system, several emission peaks appeared. These peaks are resultant from the common elements found in the simulant: silicon, aluminum, calcium, iron and magnesium. These peaks were apparent for both press ure settings; however, there were differences in the geometry of the peaks The near vacuum signal produced sharper, clearer, and stronger peaks than the atmospheric pressure signal (Figure 25) This observation suggests that the near vacuum environment provides better emission peak s ignal s than the atmospheric pressure signal s Better signals, in turn, would lead to better sensitivity to weak emission peaks and better concentration calculations.


Figure 24: Plots comparing the pressure effects on the spe ctral emission signal. The dust simulant is absent in this analysis The spectral window s were centered at 325 and 450 nm.


Figure 25: Plots comparing the pressure e ffects on the spectral emission signal of the lunar dust simulant. The spectral window s were centered at 325 and 450 nm.


4. Conclusions LIBS analysis has been proven as a successful detector of calcium particulate matter in aerosol samples. 1 3 During this project LIBS also proved as a successful detector of iron, lead, titanium, chromium, and cadmium particulate matter in ambient air LIBS would be a successful tool for energy production facilities for analyzing the air quality that is expelled at these plant s. It would be able to determine the composition and c oncentration of expelled gases, a knowledge that would be valuable to the Environmental Protection Agency and their air quality management objectives. This experiment also proved that the targeted toxic elements are not Conditional analysis is proven as a successful analytical technique. 1 4 It is able to accurately determine the concentration of aerosol samples usin g peak to base ratios and calibrated data. Conditional analysis, however, requires the knowledge of the composition of an aerosol system unknown. Standard deviation a nalysis does not need such information, so it could be used for such unknown aerosol samples. It is able to identify any elements that have emission peaks within the spectral window of the analysis. However, standard deviation analysis is not a good tool f or determining the concentration of aerosol samples. These two analytical methods could be applied in a one two punch sequence. First, standard deviation analysis could be used on an unknown aerosol to determine the composition of this aerosol. Basically, it could be used as a rapid screening tool. Then condit ional analysis could be used to determine the concentrations of the elements found using the standard deviation analysis. A near vacuum environment provided stronger and sharper emission peaks than an atmospheric pressure environment. This is presumed due to the reduction of plas ma emission due to components o f the gas. Therefore, it is conc luded that analyzing samples in a vacuum is better because it is more sensitive to weak particle hits and it provides more accurate calculations of concentrations. However, the construction of a vacuum environment is time consuming and difficult. It also d and rotating vessels are necessary to disperse the aerosol sample inside the sampling chamber. The added sensitivity and accuracy of analysis in a vacuum environment is not sufficient enough to depart from the regular analysis of an aerosol at atmospheric pressure. Calibration will also most likely be difficult, and the influence of the laser interacting with different particle types is difficult to predict.


References 1. J. E. Carranza, B. T. Fisher, G. D. Yoder, and D. W. Hahn, Spectrochim. Acta, Part B 56 851 (2001). 2. J. E. Carranza, K. Iida, and D. W. Hahn, Appl. Opt. 42 6022 (2003). 3. P. B. Dixon and D. W. Hahn, Anal. Chem. 77, 631 (2005). 4. B. Hettinger, V. Hohreiter, M. Swingle, and D. W. Hahn, Appl. Spectrosc. 60, 237 (2006). 5. L. A. Alvarez Trujillo, A. Ferrero, J. J. Laserna, and D. W. Hahn Appl. Spectrosc. 62, 1144 (2008). 6. Cliff I. Davidson, Robert F. Phalen, and Paul A. Solomon, Aerosol Science and Technology. 39, 737 (2005) 7. Periodic Table. 15 April 2010. Lenntech Water Treatment Solution. 2009. < chart.htm > 8. Ralchenko, Yu., Kramida, A.E., Reader, J., and NIST ASD Team (2008). NIST Atomic Spectra Database (version 3.1.5), [Online]. Available: [2010, April 14]. National Institute of Standards and Technology, Gaithersburg, MD. 9. D. S. McKay, J. L. Carter, W. W. Boles, C. C. Allen, and J. H. Allton, Engineering, Construction, and Operations in Space IV. 857 (1994). < >