Articles

Spectroscopy, July 2019

Click the title above to open the Spectroscopy July regular issue, Volume 34, Issue 7, in an interactive PDF format.

Vol 34 No 7 Spectroscopy July 2019 Regular Issue PDF

The nitrile functional group consists of a carbon-nitrogen triple bond with one substituent as such: -C≡N. Nitriles have an intense and sharp C≡N stretching peak near 2200 cm

. This peak for aromatic nitriles is lower than for saturated nitriles because of conjugation, like what we saw for C=O bonds. Nitriles are not that common but their presence is easy to spot because of the unusual intensity and position of the C≡N stretching peak.

	In the previous three installments of our examination of the infrared (IR) spectra of organic nitrogen containing functional groups, I introduced the topic and then we looked at primary, secondary, and tertiary amines. In the introductory column (1), we discussed how carbon and nitrogen can form single, double, and triple bonds, as seen in Figure 1.


	Figure 1: Examples of nitrogen-carbon bonds. From left to right a single bond, a double bond, and a triple bond.

	In that first column, I also bemoaned the fact that C-N bonds have weak stretching peaks and show up in the busy fingerprint region so are hard to spot, and that C=N bonds are chemically unstable and rare. Thus, there is not a good IR spectral signature for either of these bonds. The carbon-nitrogen triple bond gives rise to a functional group called 

, and contain a C≡N bond with a single substituent. If the carbon adjacent to the nitrile is saturated, the nitrile is said to be saturated. Similarly, if the carbon adjacent to the nitrile is aromatic, the nitrile is aromatic.

	Recall that one of the things that determines peak positions in IR spectroscopy is chemical bond strength, and that as chemical bond order goes up the force constant increases and peak positions move to higher wavenumbers (2). For example, we have learned that C-C stretches fall between 1600 and 1400 cm

 (3) (going forward, assume all peak positions are in cm

 or wavenumber units), that C=C stretches fall between 1680 and 1630 (4), and that C≡C stretches fall around 2200 (5).

	Because the nitrile group contains a triple bond, and because carbon and nitrogen have similar atomic masses (remember that atomic mass is another thing that determines peak positions [2]) we might expect C≡N stretching peaks to fall near where C≡C stretching peaks do, near 2200. This is, in fact, the case-as seen in the spectrum of the common solvent acetonitrile, shown in Figure 2. The peak at 2252 is the C≡N stretch of this molecule.


	Figure 2: The infrared spectrum of acetonitrile, C

	Note the intensity and sharpness of the C≡N stretching peak. Recall that one of the things that determines peak intensities in IR spectroscopy is the change in dipole moment with respect to distance, dµ/dx, during a vibration (2). The carbon-nitrogen triple bond is relatively polar, and stretching it gives a large dµ/dx value, hence the intensity of the C≡N stretching peak. For saturated nitriles, this peak falls from 2260 to 2240. For aromatic nitriles, this peak is found between 2240 and 2220.

	Note that the C≡N stretching peak for aromatic nitriles is lower than for saturated nitriles. This is because of a phenomenon called 

, where there is an electronic interaction between the C≡N bond and the aromatic ring that weakens the C≡N bond, lowers its force constant, and hence lowers the peak position. Conjugation also takes place when C=O bonds are attached to aromatic rings, and as we have seen this causes the C=O stretching peak position to also be lowered (6). Table I summarizes the C≡N stretching peak positions for nitriles.

	There are no other useful group wavenumbers for nitriles, but the unusual position and intensity of the C≡N stretching peak make nitriles easy to detect by IR spectroscopy.

	I am afraid that, besides the C≡N stretching peak, the spectrum of acetonitrile is not a good reference spectrum. Note that the molecule contains a methyl group, hence a saturated carbon, but that the C-H stretches are above 3000. This violates the rule that, in general, saturated C-H stretches fall below 3000. (3). The problem here is that the methyl group is attached directly to the nitrile, messing up its stretching force constants. Recall, however (3), that methyl groups have a bending umbrella mode at 1375±10. The peak at 1375 in the spectrum of acetonitrile is the methyl group umbrella mode, the best indicator in this spectrum that there is a methyl group present. There are very few absolute rules when interpreting IR spectra, but I can honestly say that I have never seen a CH

-C group umbrella mode peak ever fall outside the 1375±10 range.

	Nitriles contain a C≡N bond, and are either saturated or aromatic depending on the adjacent carbon. They have an intense C≡N stretching peak around 2200 that is sharp, intense, and easy to spot, because of its unusual position. Aromatic nitriles have a lower C≡N stretching peak position than saturated nitriles, because of conjugation.

Answers to Previous IR Spectral Interpretation Challenges

	As discussed in the last column (7), the two previous "IR Spectral Interpretation Challenges" were out of order, meaning you didn't have the right information at the right time to complete them properly. As a result, the last column contained two IR spectral interpretation challenges. The solutions to both are here.

Solution to IR Spectral Interpretation Challenge #1

	The spectrum in question is seen in Figure i of the Quiz Section, along with a table of its peak positions.

	As always, we begin looking at an unknown spectrum from left to right, like a sentence in a book. The first things we encounter in Figure i are two medium intensity and medium width peaks at 3346 and 3358. We have seen that both OH stretches and N-H stretches fall in this wavenumber range (1). However, recall that, because OH bonds are more polar than NH bonds, OH bonds form stronger hydrogen bonds, and hence have wider peaks than NH bonds. Given the medium width of the two peaks at 3346 and 3358, they should be assigned as NH stretches. The fact that there are two NH stretching peaks means the nitrogen has two hydrogens attached, or in other words, is a NH

	One of the functional groups that has an NH

 group in it is a primary amine. One of the beauties of NH stretches is that their position is sensitive to whether the amine is saturated or aromatic. Using what we have learned previously (8), the NH stretching peak positions here mean the molecule is an aromatic amine. Specifically, the 3436 peak is the asymmetric NH

 symmetric stretch. Thus, based just on the number and position of the NH

 stretches we now know our molecule is a primary aromatic amine.

 bending peaks, and we should track these down to confirm our primary amine assignment. Primary amines have an NH

 scissoring peak between 1650 and 1580. The peak at 1622 in Figure i can be assigned as this peak. Primary amine NH

 groups also have a wagging vibration, with the hydrogens bending above and below the plane of the page. This peak falls from 850 to 750, and is broadened like the NH

 stretches, due to hydrogen bonding. Seeing this peak in Figure i is a little tricky. There is a broad envelope of absorbance that extends from 800 to 600, but there are sharp peaks from another functional group riding on top of it. As a result, we cannot see the top of the NH

 stretching peak, but the properly placed broadened envelope means it is there.

	Our NH stretches told us we have a primary aromatic amine, but we still need to confirm the presence of an aromatic ring by looking for its peaks. Recall that aromatic rings are unsaturated, and hence have C-H stretching peaks between 3100 and 3000 (9). There are several peaks in this region in Figure i, confirming the presence of unsaturated rings. However, C=C bonds also have C-H stretches in this region, so we must look for other aromatic peaks to confirm their presence. Recall that benzene rings have "ring modes" from the stretching of the C-C bonds in the ring that give sharp peaks from 1620 to 1400 (9). The sharp peaks at 1603 and 1429 are ring modes. Lastly, we need to figure out the substitution pattern around the benzene ring. Recall that one can distinguish mono-, ortho-, meta-, and para-substituted benzene rings from each other using the presence or absence of the ring bend at 690±10, and the position of the aromatic C-H wagging peak (9). The ring bend is present in Figure i at 693 and the C-H wag is at 754, meaning the benzene ring is mono-substituted. This means we are pretty much done. We know we have a primary aromatic amine, and a mono-substituted benzene ring. The only structure that makes sense it put the amine on the ring, which gives us the molecule aniline, whose structure is seen in Figure ii in the Quiz Section, and whose peak assignments are seen in Table i in the same section.

Solution to IR Spectral Interpretation Challenge #2

	The spectrum for the second IR Spectral Interpretation challenge is seen in Figure ii of the Quiz Section. Looking from left to right, the first peak we encounter is at 3294. Note that it is somewhat broadened but not anywhere near as broad as an OH stretch (1), which indicates it is a N-H stretch. Note that there is one N-H stretching peak, so there is only one N-H bond. We have learned that secondary amines have a single N-H stretch, so there is likely a secondary amine present here (7). The position of this peak can distinguish saturated secondary from aromatic secondary amines. The peak at 3294 tells us this secondary amine is saturated. We have learned that secondary amines also have a N-H wagging peak at 750–700 (7). There is a peak in Figure ii at 739 that is somewhat broadened from hydrogen bonding, indicating this is the N-H wag we are looking for.

	For this to truly be a saturated amine, the C-H stretches must confirm the presence of saturated carbons. Recall (3) that saturated C-H bonds have C-H stretching peaks between 3000 and 2800. There are several peaks in Figure ii in that range, meaning there are saturated carbons present. The lack of C-H stretching peaks between 3100 and 3000 means that there are no unsaturated carbons present, and that all the carbons present are saturated. The fact that there are more than two saturated C-H stretching peaks means there are methyl and methylenes in the molecule. Specifically, 2961 is the CH

 symmetric stretch. The fourth peak in this region at 2793 is too low to be a normal CH

 stretch. Recall that the N-methyl group, where there is a CH

 attached to a nitrogen atom, has a low wavenumber symmetric methyl stretch (8). The peak at 2793 is properly placed to be an N-methyl symmetric stretch. This peak position is sensitive to whether the amine is saturated or aromatic. This peak's position indicates the amine is saturated, consistent with what the N-H and C-H stretches have told us.

	So far, then, we know we have a saturated secondary amine, and that one of the substituents is a methyl group. Our final job then is to determine the other substituent on the amine nitrogen. We know we have CH

 present, so the second amine substituent is likely an alkyl chain. To confirm the presence of the CH

 there should be an umbrella mode at 1375±10 (3). The peak at 1378 in Figure ii fits the bill. Recall that the CH

 rocking peak appears at 720±10 if there are four or more methylenes in a row (3). There is a peak in Figure ii at 739, but it is a little too high in wavenumber and too wide to be a CH

 rock, which is why we have already assigned this peak as being from the NH wag of our secondary amine. This all means is that there is no CH

 rocking peak present, and that the alkyl chain present has less than four methylenes in a row. This means the alkyl chain has either one, two, or three methylenes in it.

	Recall that the relative intensity of the methylene to methyl asymmetric stretching peaks tracks with the CH

 ratio for a molecule (3). In Figure iii in the Quiz Section, these two peaks are the same size, indicating the CH

 ratio is two or three. This fact eliminates the substituent being an ethyl group, since for an ethyl group the methyl peak would be bigger than the methylene peak.

	I have said before that one of the biggest challenges in determining an entire molecular structure from an IR spectrum by itself is getting the exact length of alkyl chains. This is the case here. We have narrowed down the chain to having two or three methylenes, but that is a far as we can push the analysis. The final answer could be determined by simply looking up the two appropriate reference spectra.

	The substituent is a butyl group with three methylenes in a row, and the molecule is n-methyl butylamine. Its structure is seen in Figure iv in the Quiz Section, and its peak assignments are seen in Table ii in the same section.

 is the founder and CEO of Big Sur Scientific, a maker of portable mid-infrared cannabis analyzers. He has over 30 years experience as an industrial infrared spectroscopist, has published numerous peer reviewed papers, and has written three books on spectroscopy. As a trainer, he has helped thousands of people around the world improve their infrared analyses. In addition to writing for 

, Dr. Smith writes a regular column for its sister publication 

 and sits on its editorial board. He earned his PhD in physical chemistry from Dartmouth College. He can be reached at: 

IR Spectral Interpretation Workshop

Nitriles are easy to spot because of the unusual intensity and position of the C≡N stretching peak.

Organic Nitrogen Compounds IV: Nitriles

This month's "Atomic Perspectives" column takes a look at interface cones in inductively coupled plasma–mass spectrometry (ICP-MS), which are probably the most critical components of the entire ICP-MS system. The interface region gave the early pioneers of the technique the most problems to overcome. Although we take all the benefits of ICP-MS for granted, the process of taking a liquid sample, generating an aerosol that is suitable for ionization in the plasma, and then sampling a representative number of analyte ions, transporting them through the interface, focusing them via the ion optics into the mass spectrometer for separation and detection, is not a trivial task. Each part of the journey has its own unique problems to overcome, but probably the most challenging is the extraction of the ions from the plasma into the mass spectrometer. This article therefore focuses on how best to maintain the interface region and in particular, understanding when the sampler and skimmer cones need to be cleaned or changed to optimize performance and maximize instrument uptime. 

	The role of the interface region in inductively coupled plasma–mass spectrometry, which is shown circled in Figure 1, is to transport the ions efficiently, consistently, and with electrical integrity from the plasma, which is at atmospheric pressure (760 torr), to the mass spectrometer analyzer region at approximately 10-6 torr (1,2).


	Figure 1: The interface region (circled in yellow) of an ICP-MS instrument.

	This is first achieved by directing the ions into the interface region. The interface consists of two or three metallic cones (depending on the design) with very small orifices, which are maintained at a vacuum of ~1–2 torr with a mechanical roughing pump. After the ions are generated in the plasma, they pass into the first cone, known as the 

. From there, they travel a short distance to the 

, which is generally smaller and more pointed than the sampler cone, and typically has a much smaller orifice. In some designs, there is a third cone called the 

, which is used to reduce the vacuum in smaller steps and provide less dispersion of the ion beam. Whether the system uses two cones or three, the cones are usually made of nickel (Ni) or copper (Cu), but can be made of other materials such as platinum (Pt), which is far more tolerant to corrosive liquids. To reduce the effects of high-temperature plasma on the cones, the interface housing is water cooled and made from a material that dissipates heat easily, such as Cu or aluminum. The ions then emerge from the skimmer cone, where they are directed through the ion optics and, finally, guided into the mass separation device. A more detailed view of the interface region is shown in Figure 2.


	Figure 2: A more detailed view of the interface region for ICP-MS.

The Importance of Routine Maintenance of the Interface Cones

	All sample and skimmer cones, regardless of ICP-MS manufacturer or design, serve the same function. They are the differential apertures separating the sample at atmosphere from the mass filter and detector(s) at high vacuum. Though the fundamental dimensions are all essentially the same (sample cone with shallow angle, skimmer cone with acute angle), each manufacturer has designed the cone's critical dimensions to their own unique specifications, based on the design of the mass spectrometer (including, but not limited to, orifice diameter, hole depth, and material composition). The actual cone design may vary significantly from one instrument manufacturer to another. The analytical zone, the area most in need of maintenance, displays the most significant differences. These differences, together with sample type (for example, ground water, soil digests, or sea water) contribute to the rate at which oxides and salts accumulate in and around the cone orifice. They also affect the rate of cone degradation, and influence the cleaning methodology. Ultimately, the quality of the data is the best diagnostic tool for determining when to clean or change the cones. However, if significant buildup on the cone is observed and appears to be occluding the cone orifice, cone removal and cleaning is warranted. All instrument manufacturers prescribe a daily routine to evaluate the status of the system prior to an analytical run. These fast and easy tests can also be used to qualitatively evaluate cone condition. Let's take a closer look at how to recognize when the cones are beginning to negatively impact instrument performance (3).

As oxides and salts build up around and within a cone orifice, they will change the gas flow dynamics and perhaps electrical characteristics of the cone interface. Symptoms may include reduced sensitivity and a degradation of signal stability. In severe cases, there may even be a change in signal response across the mass range due to pressure changes (higher vacuum) between the sample and skimmer cone.

Like other components of the sample introduction system, sample cones are prone to certain memory effects, especially from elements like lithium, boron, and magnesium. This is particularly true when attempting trace analysis of elements that were previously in solution at high concentration. These effects may inhibit quality analysis well before visual buildup occurs around the analytical zone of the cones. In many, if not most, cases, memory effects are due to the sample introduction system (tubing, spray chamber, torch); therefore, thorough evaluation of all potential sources of memory effects should be carried out.

 Whether from extensive use, improper use, aggressive matrices, or blunt force trauma, over time cones will degrade beyond acceptable limits. Careful handling and proper cleaning will typically enable about 500 h (several months) of operational use from each set of Ni cones and about 1500 h (18 months) from Pt cones. Some common symptoms of cone degradation are elevated oxide levels, doubly-charged species, and in severe cases, poor vacuum levels.

 Pt-tipped cones offer greater lifetime than Ni cones, and usually require less day-to-day maintenance. Aggressive matrices (high in oxidizing conditions) may require Pt-tipped cones. Unlike Ni cones, Pt-tipped cones can be cost-effectively refurbished. In addition, refurbishing processes can more than double the normal lifetime of Pt-tipped cones. Pt usually stays cleaner because it runs hotter, and has a tighter grain in structure than Ni. The tip can usually be reshaped several times. After extensive use, Pt (as well as Ni) will suffer from high temperature embrittlement, characterized by large grain crystals that will crack when the tip is being reformed. We strongly advise that users not try to mechanically alter the tip in any way, because the tolerances are tight and the shape is critical to good performance. Such alteration can lead to the unnecessary, early retirement of a good cone, which in the case of Pt, could be extremely expensive.

	Whenever possible, users should clean cones under mild, non-mechanical conditions. This is especially true for Pt cones, where accidental damage is very costly. However, with certain applications, a refractory buildup is inevitable, and will require mechanical cleaning. Following are three recommended cleaning methods. Choose the one that best suits your needs.

	Note: An alternative to the more aggressive mechanical cleaning of Pt cones is to use a professional Pt cone refurbishing service, where the most severe refractory buildup can be removed and damaged Pt cones can be reshaped to their original specifications (5).

	The light cleaning-detergent method is recommended for cleaning cones where there is no severe buildup, and what little buildup exists is easily removed. It is primarily carried out to reduce elemental memory, and improve detection limits.

It is worth emphasizing that today's cones are available in many different materials, including pure Ni, Pt, Al, and Cu, as well as varying combinations of these metals in the cone base and tip.

		Soak in 1:20 metal cleaner (such as Citranox) (4):H

0 for 2 h (sonicate in an ultrasonic bath for last 15 min, if available).

O, and repeat soak for 10 min (sonicate).

		Dry thoroughly using argon or nitrogen, if available.

	Note: If using a bath sonicator (recommended), place cones in metal cleaner solution and sonicate for the final 15 min of the 2 h soaking period. Rinse thoroughly in tap water, then place in deionized H

O, and sonicate for 10 min. Replace deionized H

O and repeat. Dry with clean, dust free cloth, let air dry, and then finally blow dry with argon or nitrogen, or place in glassware oven (40-50 °C). Always take care not to place cone tips in a precarious position where they can fall or tip. For most sample cones, this is not a significant problem, because the base is large relative to the height. Skimmer cones, however, are more prone to tipping over and deforming, since they tend to have thinner walls and sharper edges.

	Cleaning cones with a Cu base requires a slightly modified technique so as not to unduly wear the Cu part of the cone, especially where it meets the Ni or Pt tip. Over time, metal cleaner will attack the Cu, and may degrade the weld between the two metals.

		Invert cone into a dilute metal cleaner solution, so that only the tip of the cone is immersed.

O, and repeat rinse (sonicate for 15 min).

	Note: Soaking and cleaning times may be modified, depending on need. The cone fixture, easily assembled from a spare cone container, enables more aggressive cleaning of the analytical zone without degrading the bimetallic weld. However, more exotic and functional fixtures would be advisable.

Standard Cleaning–Acid/Detergent Method

	Mild accumulation of salts and oxides may build up around the analytical zone of your cones. Over time, this will affect the quality of your analyses. In this case, it is recommended that a slightly more aggressive cleaning protocol be used.

		Invert cone into fixture containing 2% HNO

		Wipe off residue with cotton swab, then rinse in tap water

		Dry with clean, dust free cloth, and then let air dry, blow dry with argon or nitrogen, or place in glassware oven (40-50 °C).

	Cleaning cones with a Cu base requires a slightly modified technique, so as not to unduly wear the Cu part of the cone, especially where it meets the Ni or Pt tip. Over time, metal cleaner and nitric acid will attack the Cu, and may degrade the weld between the two metals.

		Invert cone into fixture containing metal cleaner:H

		Wipe off residue with cotton swab, then rinse in tap water.

		Dry with clean, dust free cloth, and then let air dry, blow dry with argon or nitrogen, or place in glassware oven (40-50 °C)

 solution directly onto the Pt tip. It is recommended, however, that application be accomplished with an acid-soaked swab. This should be done with great care to ensure the Ni or Cu base is not in contact with acid, because it could become degraded.

		Mix alumina powder with water to make a loose paste.

Atomic Perspectives

A detailed look at the most critical component of an ICP-MS instrument: the interface cones

ICP-MS Interface Cones: Maintaining the Critical Interface between the Mass Spectrometer and the Plasma Discharge to Optimize Performance and Maximize Instrument Productivity

Mid-infrared spectroscopy is one of the key analytical techniques used for chemical sensing of organic compounds for myriad applications, including forensic detection, environmental monitoring, and medical diagnosis. The typical mid-infrared spectrometer consists of a broadband thermal emitter as the source, combined with a broadband detector. The source and detector are matched with a Fourier-transform infrared (FT-IR) interferometer and external transfer optics for sampling; these components comprise the complete benchtop spectrometer. There is an increasing demand for more compact and cost-effective sensing solutions to replace bulky laboratory systems. Such portable infrared devices ideally incorporate all necessary optical components and functionality onto a single chip. A key development to realize such monolithically integrated spectrometers is the combination of a bi-functional quantum cascade laser with detection capabilities. We recently interviewed Benedikt Schwarz, PhD, a principal investigator at TU Wien, in Vienna, Austria, about his work in this field. Schwarz won a FACSS Innovation award at the 2018 SciX conference for this work, and this interview is part of a series of interviews with the winners of awards presented at SciX.

One of your most cited papers to date (1) discusses the integration of a bifunctional quantum cascade laser with detection capabilities. This work represents a novel, monolithically integrated sensor for mid-infrared absorption spectroscopy. What prompted you to investigate this problem?

	The initial idea goes back to the beginning of my PhD work in 2011. While my main interest was initially theory and modeling, I was looking for a topic that had the potential to achieve a real impact for practical applications. Since then, I have been working on the challenging goal to build on-chip integrated mid-infrared sensors and spectrometers.

How is what you have done different from what others have done previously?

	Many other researchers focus on the integration of mid-infrared photonics on the silicon platform. One of the key benefits of silicon is its natural oxide that led to the famous complementary metal-oxide-semiconductor (CMOS) process. In photonics, the silicon platform is mainly limited by the technological challenge to integrate active optical components, such as lasers and detectors. Nevertheless, silicon is often used for telecom wavelengths, due to its excellent waveguiding properties and low-cost of fabrication. In the mid-infrared, silicon photonics loses nearly all of its attraction (or appeal), as the natural oxide is absorbing over a wide spectral range.

	We decided to follow a different strategy. We used the best available laser technology, and extended it step-by-step to include both detectors and waveguides that are integrated and matched to the laser system.

Your device switches between laser and detector operation. You have chosen to create a device that interacts with the sample via a dielectric-loaded surface plasmon polariton (SPP) waveguide. What are the advantages of this approach to sampling for absorption spectroscopy?

	The idea is to bring the sensor to the sample, instead of the sample to the sensor. As an example, the dielectric-loaded surface plasmon polariton waveguides can be combined with microfluidics to detect the sample in place. An in situ monitoring system can give feedback for real-time process optimization, even with small sample volumes. This in situ sampling approach is currently being brought to the next level in a new research project at TU Wien together with the chemistry department.

In your referenced paper (1), you describe the capability to reach parts per million (ppm) accuracy over a wide range of concentrations (0–60%). Have you been able to demonstrate such accuracy in a real-world application? What are your thoughts on the practicality of this sensor design for routine analysis of infrared absorbing molecules at the ppm concentration level?

	Extending our sensors with microfluidics will enable us to probe the chips in a more controlled way, and to implement them into real-world applications. In particular, we are targeting the detection of the secondary structure of proteins. I would also like to note that the presented sensor chips were based on the first generation of the bifunctional material, and a major performance improvement was achieved during the last few years (3).

	We also learned that the integration of all optical components is an advantage in terms of sensitivity. The lasers remain unchanged, the on-chip waveguides are much more efficient compared to free-space optics, and the on-chip detectors have an exceptionally low noise, due to their small size. The main remaining challenge is most probably the contamination of the sensor. Given these improvements, we will have the capability of low concentration detection.

You have described an elegant solution to the sampling or coupling issue using a surface plasmon polariton (SPP) waveguide. Using this waveguide, the electromagnetic wave propagates along a metal–dielectric interface demonstrating an evanescent wave decay on both sides (2,3). Would you describe this wave interaction phenomenon for our readers?

	Traditional dielectric waveguides, such as optical fibers, are composed of two or more different dielectric materials. Metals show a fundamentally different optical behavior compared to dielectrics. The "free" electrons make them highly conductive, and they appear shiny. A surface plasmon polariton can propagate along the interface between a dielectric and a metal. Their occurrence is a natural consequence of light propagation governed by Maxwell's equations, and their name originates from the coupling of photons and electrons in a quasi-particle picture. The loss in the metal is thereby crucial for its functionality. Hence, surface plasmon polaritons will always demonstrate a certain amount of absorption, due to the damped oscillation of the electrons.

	For us, surface plasmon polariton waveguides have two key features. First, nearly all of the light propagates outside of the metal and can interact with the analyte, and second, they can be easily integrated together with our lasers and detectors on the same chip. A drawback is the optical signal losses due to the metal. Although, for liquid sensing, the metallic signal absorption losses are much smaller compared to the absorption signal of the analyte, alternative concepts will be required when aiming for on-chip gas detection due to the smaller analyte signal of gases, as they are strongly diluted compared to liquids.

What would you consider to be the most useful contribution of your work?

	I believe it is the principal strategy to directly integrate all optical components on the III–V semiconductor platform. Building good lasers is the most difficult part in the mid-infrared technology. Hence, it always seemed logical to me to use the best available laser technology, and to rather take a compromise on the other optical components.

	About a year ago, I started with a small research team to target the remaining technical challenges for multi-species analysis. The main question is: How can we efficiently integrate a mid-infrared spectrometer on a single chip? Currently, we are focusing on dual-comb spectroscopy. I am convinced that it is the ideal technique for miniaturization, and it is compatible with our previous efforts. However, it does not come for free. Quantum cascade laser frequency combs are very sensitive, and advanced stabilization techniques are required. In our recent work, we showed that, by coherent electrical injection locking, we can make them much more robust against harsh real-world conditions (4). Our findings enable the downscaling of dual-comb spectroscopy to very small footprints. Now, we are working on the on-chip integration of dual-comb spectroscopy and the reduction of the power consumption to target mobile hand-held devices. Our newest unpublished proof-of-principle demonstration is already very promising, and as soon as all technical challenges are solved, the real-world applications will be in reach.

	(1) B. Schwarz, P. Reininger, D. RistaniÄ, H. Detz, A.M. Andrews, W. Schrenk, and G. Strasser, 

Plasmonics: Fundamentals and Applications

 (Springer Science & Business Media, May 16, 2007).

	(3) B. Schwarz, C. A. Wang, L. Missaggia, T. S. Mansuripur, P. Chevalier, M. K. Connors, D. McNulty, J. Cederberg, G. Strasser, and F. Capasso, 

	(4) J. Hillbrand, A. M. Andrews, H. Detz, G. Strasser, B. Schwarz, 

 is a visiting postdoctoral researcher at Harvard University, School of Engineering and Applied Sciences with Prof. Federico Capasso. He is also a principal investigator at Technical University Wien, Vienna, Austria, Institute of Solid State Electronics. He completed his PhD in Technical Sciences at TU Wien, with a thesis on monolithic integration of mid-infrared photonics with Prof. Gottfried Strasser. Currently, he is building up his own research team at TU Wien, focusing on the on-chip integration of mid-infrared spectrometers using frequency combs.

Spectroscopy Spotlight

SciX Award winner Benedikt Schwarz explains his groundbreaking work in portable infrared devices.

Integration of a Bifunctional Quantum Cascade Laser with Detection for Lab-on-a-Chip Mid-Infrared Absorption Spectroscopy

As the use of nanoparticles in a wide range of applications continues to increase, the need for techniques to analyze these particles also grows. Javier Laserna of the University of Málaga, in Spain, has developed an approach to measure individual nanoparticles in air using optical trapping (OT), combined with laser-induced breakdown spectroscopy (LIBS). Laserna received the 2018 Lester W. Strock Award for this development, as well as for his broader research on analytical methods and instrumentation using LIBS. He recently spoke to us about this work.

What led you to investigate LIBS for nanoparticles?

	Well, to start with, nanoparticles, as everybody knows, are of extreme interest because of the growing number of applications for these objects in many areas of technology. There are only a few analytical techniques for measuring the composition of nanoparticles, and, with most of those techniques, it is only possible to analyze nanoparticles in bulk, or when they are attached to a surface or suspended in solution. The ability to analyze individual particles is almost nonexistent; only a few techniques are available, and none of them provide the complete analytical composition of the nanoparticle, as far as I know.

	LIBS is well known for its ability to provide panoramic spectra-from the ultraviolet (UV) to the infrared (IR)-in a single shot, in bulk materials. But in the case of nanoparticles, we wanted to analyze these extremely light and extremely low-mass particles in a single shot.So, we tried it, and, to our surprise, it worked rather well. The signal-to-noise ratio we obtained was very reasonable considering the very low mass of the nanoparticles.

You developed a method for elemental identification of nanoparticles using LIBS that involves optical trapping. How are the optical trapping and LIBS measurement carried out?

	One of the most difficult tasks in this work is to isolate an individual nanoparticle for analysis. For this purpose, we used an optical trap. The concept behind optical trapping technology is rather simple. A continuous-wave (CW) laser beam, propagating upwards, is focused on sample nanoparticles in air or in solution (Figure 1a). The radiation pressure created in this nanoparticle medium is able to counteract the gravity forces that are acting on the particles. So, in principle, this optical field is able to trap the particles. However, the trapping is not always easy, particularly when the particles are very small and when the trapping is performed in air. The air currents in the laboratory, for example, tend to disrupt the particles and they fall out of the trap.


	Figure 1: (a) Schematic view of the optical catapulting and optical trapping processes developed by Javier Laserna and his team at the University of Málaga to trap individual nanoparticles for analysis by laser-induced breakdown spectroscopy (LIBS), and (b) close-up of the optical trap within a cuvette.

	So, we created a confined nanoparticle medium by using a cuvette that is commonly used for spectrophotometry. That way, it is possible to keep the particles inside the trap and free from the air currents. But before you can get to the optical trapping, you have to create an aerosol of the particles you want to analyze. We did that using optical catapulting. This is done by putting a powder containing nanoparticles on top of a microscope slide. Then, we shoot a laser pulse into the nanopowder from under the slide. The shockwave created from the laser pulse beneath the slide creates the nanoparticle aerosol. The aerosol then interacts with the CW laser, which is sent colinearly with the catapulting beam, propagating upwards through the microscope slide. This CW laser is tightly focused by a microscope objective, which is a 20x laser objective with a numerical aperture of 0.4. All of this is done inside the cuvette, to keep the nanoparticles from blowing away (Figure 1b).

	The aerosol created in the cell is highly dispersed, so we have to wait for the nanoparticle to deposit by gravity. Usually, after three to five minutes, the particle is trapped; the exact time depends on factors such as particle size, particle geometry, and the optical properties of the particle, but most particles are trapped within about five minutes.

	In the optical trap, it is possible to capture a cluster of particles rather than a single particle. So, we repeat measurements several times until the signal has a low range signal-to-noise ratio (S/N). This measurement response indicates that the particles creating this signal are individual particles.

	We can modify the trapping beam to make the particles go up or down in the trap by changing the power of the trapping laser. We can manipulate the particles so that particles that are too large fall outside the trap. Also, we sometimes observe strings of particles in the trap. We manipulate the laser system so that we end up with only one particle in the trap.

	The next thing that is important is to have the particle in a precise position in space, so that the nanoparticle can be effectively interacted with by the LIBS excitation beam, which is shot perpendicular to the particle trapping beam. This analyzing laser is a moderate-power laser (200 millijoules per pulse), with excitation wavelength set at 1064 nm. We send a single shot of this laser at the nanoparticle that is trapped. It is important that the nanoparticle targeted is precisely located where the laser is going to hit. That way, we can measure the spectrum of the particle. If the nanoparticle is not in that precise spot (at 0,0,0 coordinates), the S/N is nil, and we don't see any signal from the analyte. The only thing we see in that case is the spectrum of air. Then what we observe in the spectra is the mixture of the spectra of the nanoparticle plus the spectrum of air.

	So how do we get rid of the air? There are several strategies we can use. The first is to delay integration of the signal by several microseconds. In this way, many of the ionic spectral lines of nitrogen and oxygen in particular are reduced, virtually to the point of disappearing, and so then we have only the weak ionic lines of air-nitrogen of oxygen-and the spectral lines belonging to the nanoparticle. Some atomic lines of oxygen and nitrogen are still present, but they can be resolved from the analyte lines by our spectrometer.

	The second strategy is to focus the analytical beam a little ahead of the position of the particle.

	The most intense lines of air are a little bit off axis of where we are observing the signal. The plasma of air and the nanoparticle are both observed simultaneously. We then, through data analysis, discriminate between the air and particle signals.

	This is a complex process, but the important thing is that we are able to discriminate the plasma of air from the plasma of the particle.

Your method enabled detection of nanoparticles smaller than 100 nm-in other words, in the attogram (10–18 gram) regime. What challenges did you have to overcome to achieve detection of particles that small?

	At first, we were a bit surprised that this technology, LIBS, could detect such tiny particles. We had heard many times that one of the barriers to wider adoption of LIBS is low sensitivity. We have been working with LIBS for many years, so we knew that what is meant by "low sensitivity" is low sensitivity when analyzing trace elements in a matrix. For example, in solid materials, such as when we are analyzing an alloy with one element at trace level, LIBS can rarely achieve a limit of detection (LOD) below the low ppm level. However, the ability to analyze signals from monoelemental materials, like pure copper particles, is much improved, and lower LODs than ppm can be achieved.

	So, we started investigating this. We started with microparticles in the 2-micrometer size and achieved a reasonable S/N. Then, we went down to 100-nm particles and still achieved a very nice signal-not only with metal nanoparticles like nickel particles, but also with silica nanoparticles. We analyzed smaller and smaller particles, and still obtained an acceptable S/N. So, we learned that LIBS is able to detect extremely low masses when particles are single-phase or single-component particles.

	Then, we discovered that we are not limited to analyzing mono-elemental particles. We were able to analyze multi-elemental particles, such as ferrite particles composed of from three to four elements, and we were able to analyze all the components at same time.

	So, we kept investigating ever smaller particles, and we were able to analyze 25-nm-sized particles; this is close to the LOD. Ultimately, at this size of nanoparticle, we achieved an LOD on the order of 57 attograms of mass, which is a spectacular tiny size and mass for LIBS and for comparable technologies.

What novel measurements have you been able to achieve with this approach so far?

	For the moment, we are still doing research work on this approach. We are still working to understand the many processes involved, and how to optimize the many variables that affect the system, and trying to understand how the signals can be captured in the most efficient manner possible. The solid angle that we use for signal collection in LIBS is very low in comparison with the 4Π emitting geometry of a plasma.

	We are trying to improve the capture of the photons emitted by the plasma. Also, we are trying to avoid one of the most inefficient steps in spectroscopy. When analyzing emission from a light source with a traditional monochromator or spectrograph, a considerable amount of photons from the source are lost at the entrance slit. This is the very nature of these devices; it is what they are designed to do.

	We are trying to avoid this loss of photon signal from traditional monochromator-based measurements by applying a hyperspectral approach to collect the signal from the plasma. By using an appropriate set of filters that collects all the signals from the analyte of interest, we expect to capture all the photons and analyze them all.

	This research is ongoing, and we still don't have definitive results. But if this works, we will have much better LODs at a much lower attogram sample mass level.

How could this technique actually be used in a practical application?

	To think about this, we are reviewing the sampling efficiency of the approach. As I said before, we can capture one particle about every three minutes on average. So, if we need to analyze hundreds of thousands of particles, this rate would be prohibitive, and we could not use this approach as a screening technique for large numbers of particles

	So, what we believe is the main application of our technology is the analysis conditions where we have only very tiny amounts of sample, such as functionalized nanoparticles that are synthesized in very small amounts for biology research. These samples are hard to analyze by other technologies. Most technologies need relatively large amounts of sample. So, when the amount of sample available for analysis is on the order of micrograms, our technology provides a solution.

	At the same time, we also intend to increase our sample throughput by analyzing the nanoparticles in a flow. This is another approach that has nothing to do with the optical trapping or catapulting we have been discussing. In this approach, we would create a sample suspension and create an aerosol of microdroplets and a device to dry the sample, so that we can analyze them on the fly, with a kind of time-of-flight (TOF) analyzer using LIBS. We are only starting this work, but are hopeful that this will provide much better sample throughput for analyzing single nanoparticles.

	For now, this is a laboratory technique, but we think this can evolve into a commercial product in the future.

	Remember that the need for this technology is growing because there are more and more nanoparticles being used in many types of technologies. That means the need to characterize the chemical properties-not only the optical or magnetic properties-of nanoparticles is increasing. Those chemical properties are very important in many applications, and currently, very few technologies are available to analyze the chemical composition of nanoparticles, so we see opportunity here. So, we plan to protect our work through a patent, and we are preparing a patent application now.

	(1) P. Purohit, F.J. Fortes, and J.J. Laserna, 

 is a professor of analytical chemistry at the University of Málaga, in Málaga, Spain. His current research interests include the use of lasers in chemical analysis, laser-induced plasma spectroscopy; time-of-flight mass spectrometry; secondary ionization mass spectrometry; surface analysis using laser ablation with optical and ion detection, and laser remote chemical analysis. In the 1990s, he succeeded in demonstrating large-scale optics standoff laser induced breakdown spectroscopy for analysis of distant objects. Since then, this technique has been used for the analysis of explosives and in space exploration. Currently, he is a co-investigator of the SuperCam instrument for the NASA Mars 2020 mission. Laserna has participated in several research projects funded by the European Commission's Framework Programme as both a consortium member and coordinator. He has also been the principal investigator of more than 25 research projects funded by the Spanish central government and the regional government of Andalucía. He has published over 300 scientific papers and six books or book chapters and is a co-inventor of seven patents held by the University of Málaga. Laserna was awarded the Spanish Royal Society of Chemistry National Award for Research in Analytical Chemistry in 2009 and the Spanish Society for Applied Spectroscopy National Award for his research career in applied spectroscopy in 2010. He is also the 2018 recipient of the Lester W. Strock Award from the Society for Applied Spectroscopy.

Javier Laserna has developed a technique for analyzing the composition of individual nanoparticles that is capable of detecting nanoparticles as the attogram level, by combining LIBS with optical trapping.

Detection and Identification of Individual Nanoparticles Using Laser-Induced Breakdown Spectroscopy (LIBS)

Nutrient Management Planning with Portable X-ray Fluorescence (pXRF)

Yadav Sapkota, B. Lee Drake, Louis M. McDonald, Thomas C. Griggs, and Thomas J. Basden

	Water quality around the world is impaired by nutrients. Excess phosphorus (P) and nitrogen (N) in surface water is one cause of eutrophication and toxic algal blooms. Much of this P and N comes from agricultural operations, especially when animals are concentrated in one location, including concentrated animal feeding operations (CAFO), as defined by the United States Department of Agriculture (USDA). A nutrient management plan (NMP) is a whole-farm planning tool to identify and manage nutrient inputs and outputs, and within-farm transfers-essentially a nutrient mass balance.

	An NMP that accounts for P inputs, outputs, and transfers within a farm can help to minimize P losses to the environment, and improve farm profitability. The traditional approach to data collection for an NMP is to collect soil, plant (forage, grain, and supplements), and manure samples for analysis in a commercial analytical laboratory. In addition to being slow (up to three weeks to receive data) and expensive, this process makes it difficult to account for the spatial and temporal changes in P concentrations. Here, we report our initial efforts to develop portable X-ray fluorescence (pXRF) methods to determine P in heterogeneous forage samples and manure to facilitate the nutrient management planning process.

	We assumed that the limiting factor for pXRF determination of P in forage would be its heterogeneous sample composition. Forage, either grazed or made into hay, is a mixture of plant species (grasses, legumes, and other weeds), plant parts, and dimensions (stems, leaves, and seeds), all of which could influence pXRF results. We assumed the limiting factor for pXRF determination of P in manure would be sample water content, because it is well known that water attenuates X-rays.

 = 42) from local farms. Samples were oven dried (60 °C for 3 d), and ground into three particle sizes (≤0.5 mm, 0.25 to 0.5 mm, and 1 to 2 mm). Prepared samples were scanned by pXRF (Tracer III-SD; SN T3S2102; Bruker Elemental, Kennewick, Washington), equipped with a 4W rhodium tube and a Peltier-cooled, 10 mm

 silicon drift detector, using a vacuum (<10 torr) without a filter. Per the calibration method, samples of each size fraction were placed in cups over 0.4 µm prolene X-ray film and scanned for 180 s using the portable XRF instrument (Tracer III-SD; SN T3S2102; Bruker Elemental, Kennewick, Washington) in benchtop mode, equipped with a 4W rhodium tube, and a Peltier-cooled, 10 mm

 silicon drift detector, with a voltage of 15 KeV and anode current of 26 uA without a filter. A subset (

 = 29) was also scanned for 60 and 120 s. The resulting pXFR counts were compared to traditional wet-chemical digestions with inductively coupled plasma optical emission spectrometry (ICP-OES) determinations. Specifically, subsamples were further ground to pass a 2 mm screen, acid-digested in triplicate by microwave (MARS Xpress, CEM Inc., Matthews, North Carolina) and the elemental concentration was determined using ICP-OES (Optima DV 2100, Perkin Elmer, Norwalk, Connecticut) with a NIST Certified Reference Material (CRM) 1573a-tomato leaf. The pXRF counts for P were compared to ICP-OES determinations using regression models. Oven dried beef, dairy, and poultry manure samples (

 = 40) were ground and adjusted to the additional four moisture ranges (

 moisture): 10–20%, 20–30%, 40–50%, and 60–70%. Samples were scanned by pXRF for 180 s using a vacuum (<10 torr) and without a filter as described above. Wet chemical determinations were also performed as described above, and pXRF counts for P were compared to ICP-OES determinations using regression models.

	Additionally, we used the entire spectrum to develop calibration models using random forest regression instead of using photon counts (intensity) of individual elements to predict P concentrations (Figure 2), and to predict and correct for sample moisture concentration (Figure 3). The entire XRF spectrum not only contains information on the energy level and photon counts, but also the record of the detector channel counts. The use of multivariate regression techniques, like random forest regression, uses the individual channel counts as independent variables to predict elemental concentration (2). Random forest regression is an ensemble computational method that creates pseudorandomized subsets of the data (decision trees), which can be used to assign weights to variables (3). By applying this to the detector channel counts, it is possible to identify variables that influence a given dependent variable with reduced covariance effects (4). The approach used here employs the randomForest package in R (4.6–14) using regression trees (5,6). Random forests have been used in infrared spectrometry using the same framework (7). The approach outlined here uses individual channel counts as variables, rather than specific elemental lines (2). It identifies the X-ray energies associated with any independent variable. The silicon-drift detector of pXRF used for this research has 2048 channels to record energies from 1 to 40 KeV.

	The amount of P in heterogeneous forage samples could be determined with pXRF (

 = 0.93; Figure 1). Relationship strength increased with decreasing particle size; however, the relationship was still strong (

 ≥ 0.57) at the largest particle size. Scanning time did not affect the relationship with ICP concentration for any of the particle sizes evaluated (1).


	Figure 1: Regression plot (p < 0.001) for ICP and pXRF determined P for the ≤0.5-mm samples. The shaded portion represents the standard error of the regression line.

	For P in manure, dried and ground samples produced a stronger relationship with wet chemical techniques (

 = 0.98; Figure 2). The presence of moisture negatively affected elemental determination in manure samples. Calibrations (

 = 200) were prepared using random forest regression with detector channel counts as independent variables. The back end of the spectrum (14–15 keV) had strong predictive power (

 = 0.98) for moisture content (Figure 3). The random forest approach increased 

 between pXRF and wet-chemical methods from less than 0.66 to greater than 0.96 for P compared to linear, non-linear, and Lucas-Tooth equations. These results indicated that elemental concentration can accurately be measured in dried and moist manure samples using pXRF, and expands the potential applications of pXRF to in situ elemental determinations for agricultural and environmental samples.


	Figure 2: Predicted phosphorus (P) concentration using random forest regression. (a) Validation of the predicted P concentration (estimated %) with the ICP measured P concentration (concentration %) of the sample. There was strong relationship (r

 = 0.98) between estimated and actual P concentration in the sample. (b) Importance plot showing the region of the spectrum for P prediction including the influential energy levels.

	Although additional work is needed to extend this to in situ field analysis, specifically the pXRF determination of whole forage plants, this simplification of the elemental analytical method could expedite NMP planning, contributing to better whole-farm P management and improved surface water quality.


	Figure 3: Moisture content prediction using random forest regression. (a) Validation of the predicted moisture content (estimated %) from random forest regression with the actual moisture content (concentration %) in the sample. There was strong relationship (r

 = 0.98) between estimated and actual moisture concentration in the sample. (b) Importance plot showing the region of the spectrum for moisture prediction. The moisture was predicted with better strength (r

 = 0.98) at the back end of the spectrum (14-15 KeV), with some influences from front end of the spectrum (2-4 KeV).

	(1) Y. Sapkota, L.M. McDonald, T.C. Griggs, T. Basden, and B.L. Drake. 

	(2) B. Drake, CloudCal v3.0. GitHub (2019). doi:10.5281/zenodo.2596154.

 is with the Division of Plant and Soil Sciences at West Virginia University, in Morgantown, West Virginia, and the Wetland and Aquatic Biogeochemistry Laboratory of the College of Coast and Environment at Louisiana State University, in Baton Rouge, Louisiana. 

 are with the Division of Plant and Soil Sciences at West Virginia University, in Morgantown, West Virginia. 

 is with the Department of Anthrolpology at the University of New Mexico, in Albuquerque, New Mexico. 

 is with the Extension service of West Virginia University, in Morgantown, West Virginia. Direct correspondence to: 

Total Reflection Energy Dispersive X-ray Fluorescence for Analysis of Human Samples: from Cells to Tissues

Ana Pejovic´-Milic´ and Gabriella Mankovskii

	Trace metals play a crucial role in the human body and health. Measuring the accumulation of these metals in human cells and tissues is important for diagnosis and treatment of many diseases, as well as for measuring environmental, occupational, and medical exposures. For example, environmental exposure of children to lead is associated with intellectual deficiency (1), while exposure to aluminum (2) and manganese (3) in the workplace highlights the need to analyze human samples to regulate occupational exposures and improve occupational hygiene. Furthermore, modern ophthalmology is exploring novel ideas on the effects of metals on eye health-namely, the link between neonatal cataracts and lead exposure during pregnancy, and the use of cosmetics, potentially leading to the accumulation of metals in the eye, and larger lacrimal sac stones in women.

Several leading scientists describe their recent work to advance X-ray analysis techniques.

Spectroscopy-07-01-2019

Vol 34 No 7 Spectroscopy July 2019 Regular Issue PDF

Organic Nitrogen Compounds IV: Nitriles

ICP-MS Interface Cones: Maintaining the Critical Interface between the Mass Spectrometer and the Plasma Discharge to Optimize Performance and Maximize Instrument Productivity

Integration of a Bifunctional Quantum Cascade Laser with Detection for Lab-on-a-Chip Mid-Infrared Absorption Spectroscopy

Detection and Identification of Individual Nanoparticles Using Laser-Induced Breakdown Spectroscopy (LIBS)

Trends in X-ray Techniques

Determination of Adsorption of Cationic and Anionic Dyes onto Shungite by Raman Spectroscopy