skip to main content

Title: Oxidation of Sulfonamide Antibiotics of Six-membered Heterocyclic Moiety by Ferrate(VI): Kinetics and Mechanistic Insight into SO2 Extrusion
This paper presents ferrate(VI) (FeVIO42-, FeVI) oxidation of a wide range of sulfonamide antibiotics (SAs) containing five- and six-membered heterocyclic moieties (R) in their molecular structures. Kinetics measurements of the reactions between FeVI and SAs at different pH (6.5 – 10.0) give species-specific second-order rate constants, k5 and k6 of the reactions of protonated FeVI (HFeO4-) and unprotonated FeVI (FeVIO42-) with protonated SAs (HX), respectively. The values of k5 varied from (1.2 ± 0.1) × 103 to (2.2 ± 0.2) × 104 M-1 s-1, while the range of k6 was from (1.1 ± 0.1) × 102 to (1.0 ± 0.1) × 103 M-1 s-1 for different SAs. The transformation products of reaction between FeVI and sulfadiazine (SDZ, contains a six-membered R) include SO2 extrusion oxidized products (OPs) and aniline hydroxylated products. Comparatively, oxidation of sulfisoxazole (SIZ, a five-membered R) by FeVI has OPs that have no SO2 extrusion in their structures. Density functional theory (DFT) calculations are performed to demonstrate SO2 extrusion in oxidation of SDZ by FeVI. The detailed mechanisms of oxidation are proposed to describe the differences in the oxidation of six- and five-membered heterocyclic moieties (R) containing SAs (i.e., SDZ versus SIZ) by FeVI.  more » « less
Award ID(s):
Author(s) / Creator(s):
; ; ; ; ;
Date Published:
Journal Name:
Environmental science & technology
Page Range / eLocation ID:
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. Obeid, Iyad Selesnick (Ed.)
    Electroencephalography (EEG) is a popular clinical monitoring tool used for diagnosing brain-related disorders such as epilepsy [1]. As monitoring EEGs in a critical-care setting is an expensive and tedious task, there is a great interest in developing real-time EEG monitoring tools to improve patient care quality and efficiency [2]. However, clinicians require automatic seizure detection tools that provide decisions with at least 75% sensitivity and less than 1 false alarm (FA) per 24 hours [3]. Some commercial tools recently claim to reach such performance levels, including the Olympic Brainz Monitor [4] and Persyst 14 [5]. In this abstract, we describe our efforts to transform a high-performance offline seizure detection system [3] into a low latency real-time or online seizure detection system. An overview of the system is shown in Figure 1. The main difference between an online versus offline system is that an online system should always be causal and has minimum latency which is often defined by domain experts. The offline system, shown in Figure 2, uses two phases of deep learning models with postprocessing [3]. The channel-based long short term memory (LSTM) model (Phase 1 or P1) processes linear frequency cepstral coefficients (LFCC) [6] features from each EEG channel separately. We use the hypotheses generated by the P1 model and create additional features that carry information about the detected events and their confidence. The P2 model uses these additional features and the LFCC features to learn the temporal and spatial aspects of the EEG signals using a hybrid convolutional neural network (CNN) and LSTM model. Finally, Phase 3 aggregates the results from both P1 and P2 before applying a final postprocessing step. The online system implements Phase 1 by taking advantage of the Linux piping mechanism, multithreading techniques, and multi-core processors. To convert Phase 1 into an online system, we divide the system into five major modules: signal preprocessor, feature extractor, event decoder, postprocessor, and visualizer. The system reads 0.1-second frames from each EEG channel and sends them to the feature extractor and the visualizer. The feature extractor generates LFCC features in real time from the streaming EEG signal. Next, the system computes seizure and background probabilities using a channel-based LSTM model and applies a postprocessor to aggregate the detected events across channels. The system then displays the EEG signal and the decisions simultaneously using a visualization module. The online system uses C++, Python, TensorFlow, and PyQtGraph in its implementation. The online system accepts streamed EEG data sampled at 250 Hz as input. The system begins processing the EEG signal by applying a TCP montage [8]. Depending on the type of the montage, the EEG signal can have either 22 or 20 channels. To enable the online operation, we send 0.1-second (25 samples) length frames from each channel of the streamed EEG signal to the feature extractor and the visualizer. Feature extraction is performed sequentially on each channel. The signal preprocessor writes the sample frames into two streams to facilitate these modules. In the first stream, the feature extractor receives the signals using stdin. In parallel, as a second stream, the visualizer shares a user-defined file with the signal preprocessor. This user-defined file holds raw signal information as a buffer for the visualizer. The signal preprocessor writes into the file while the visualizer reads from it. Reading and writing into the same file poses a challenge. The visualizer can start reading while the signal preprocessor is writing into it. To resolve this issue, we utilize a file locking mechanism in the signal preprocessor and visualizer. Each of the processes temporarily locks the file, performs its operation, releases the lock, and tries to obtain the lock after a waiting period. The file locking mechanism ensures that only one process can access the file by prohibiting other processes from reading or writing while one process is modifying the file [9]. The feature extractor uses circular buffers to save 0.3 seconds or 75 samples from each channel for extracting 0.2-second or 50-sample long center-aligned windows. The module generates 8 absolute LFCC features where the zeroth cepstral coefficient is replaced by a temporal domain energy term. For extracting the rest of the features, three pipelines are used. The differential energy feature is calculated in a 0.9-second absolute feature window with a frame size of 0.1 seconds. The difference between the maximum and minimum temporal energy terms is calculated in this range. Then, the first derivative or the delta features are calculated using another 0.9-second window. Finally, the second derivative or delta-delta features are calculated using a 0.3-second window [6]. The differential energy for the delta-delta features is not included. In total, we extract 26 features from the raw sample windows which add 1.1 seconds of delay to the system. We used the Temple University Hospital Seizure Database (TUSZ) v1.2.1 for developing the online system [10]. The statistics for this dataset are shown in Table 1. A channel-based LSTM model was trained using the features derived from the train set using the online feature extractor module. A window-based normalization technique was applied to those features. In the offline model, we scale features by normalizing using the maximum absolute value of a channel [11] before applying a sliding window approach. Since the online system has access to a limited amount of data, we normalize based on the observed window. The model uses the feature vectors with a frame size of 1 second and a window size of 7 seconds. We evaluated the model using the offline P1 postprocessor to determine the efficacy of the delayed features and the window-based normalization technique. As shown by the results of experiments 1 and 4 in Table 2, these changes give us a comparable performance to the offline model. The online event decoder module utilizes this trained model for computing probabilities for the seizure and background classes. These posteriors are then postprocessed to remove spurious detections. The online postprocessor receives and saves 8 seconds of class posteriors in a buffer for further processing. It applies multiple heuristic filters (e.g., probability threshold) to make an overall decision by combining events across the channels. These filters evaluate the average confidence, the duration of a seizure, and the channels where the seizures were observed. The postprocessor delivers the label and confidence to the visualizer. The visualizer starts to display the signal as soon as it gets access to the signal file, as shown in Figure 1 using the “Signal File” and “Visualizer” blocks. Once the visualizer receives the label and confidence for the latest epoch from the postprocessor, it overlays the decision and color codes that epoch. The visualizer uses red for seizure with the label SEIZ and green for the background class with the label BCKG. Once the streaming finishes, the system saves three files: a signal file in which the sample frames are saved in the order they were streamed, a time segmented event (TSE) file with the overall decisions and confidences, and a hypotheses (HYP) file that saves the label and confidence for each epoch. The user can plot the signal and decisions using the signal and HYP files with only the visualizer by enabling appropriate options. For comparing the performance of different stages of development, we used the test set of TUSZ v1.2.1 database. It contains 1015 EEG records of varying duration. The any-overlap performance [12] of the overall system shown in Figure 2 is 40.29% sensitivity with 5.77 FAs per 24 hours. For comparison, the previous state-of-the-art model developed on this database performed at 30.71% sensitivity with 6.77 FAs per 24 hours [3]. The individual performances of the deep learning phases are as follows: Phase 1’s (P1) performance is 39.46% sensitivity and 11.62 FAs per 24 hours, and Phase 2 detects seizures with 41.16% sensitivity and 11.69 FAs per 24 hours. We trained an LSTM model with the delayed features and the window-based normalization technique for developing the online system. Using the offline decoder and postprocessor, the model performed at 36.23% sensitivity with 9.52 FAs per 24 hours. The trained model was then evaluated with the online modules. The current performance of the overall online system is 45.80% sensitivity with 28.14 FAs per 24 hours. Table 2 summarizes the performances of these systems. The performance of the online system deviates from the offline P1 model because the online postprocessor fails to combine the events as the seizure probability fluctuates during an event. The modules in the online system add a total of 11.1 seconds of delay for processing each second of the data, as shown in Figure 3. In practice, we also count the time for loading the model and starting the visualizer block. When we consider these facts, the system consumes 15 seconds to display the first hypothesis. The system detects seizure onsets with an average latency of 15 seconds. Implementing an automatic seizure detection model in real time is not trivial. We used a variety of techniques such as the file locking mechanism, multithreading, circular buffers, real-time event decoding, and signal-decision plotting to realize the system. A video demonstrating the system is available at: The final conference submission will include a more detailed analysis of the online performance of each module. ACKNOWLEDGMENTS Research reported in this publication was most recently supported by the National Science Foundation Partnership for Innovation award number IIP-1827565 and the Pennsylvania Commonwealth Universal Research Enhancement Program (PA CURE). Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the official views of any of these organizations. REFERENCES [1] A. Craik, Y. He, and J. L. Contreras-Vidal, “Deep learning for electroencephalogram (EEG) classification tasks: a review,” J. Neural Eng., vol. 16, no. 3, p. 031001, 2019. [2] A. C. Bridi, T. Q. Louro, and R. C. L. Da Silva, “Clinical Alarms in intensive care: implications of alarm fatigue for the safety of patients,” Rev. Lat. Am. Enfermagem, vol. 22, no. 6, p. 1034, 2014. [3] M. Golmohammadi, V. Shah, I. Obeid, and J. Picone, “Deep Learning Approaches for Automatic Seizure Detection from Scalp Electroencephalograms,” in Signal Processing in Medicine and Biology: Emerging Trends in Research and Applications, 1st ed., I. Obeid, I. Selesnick, and J. Picone, Eds. New York, New York, USA: Springer, 2020, pp. 233–274. [4] “CFM Olympic Brainz Monitor.” [Online]. Available: [Accessed: 17-Jul-2020]. [5] M. L. Scheuer, S. B. Wilson, A. Antony, G. Ghearing, A. Urban, and A. I. Bagic, “Seizure Detection: Interreader Agreement and Detection Algorithm Assessments Using a Large Dataset,” J. Clin. Neurophysiol., 2020. [6] A. Harati, M. Golmohammadi, S. Lopez, I. Obeid, and J. Picone, “Improved EEG Event Classification Using Differential Energy,” in Proceedings of the IEEE Signal Processing in Medicine and Biology Symposium, 2015, pp. 1–4. [7] V. Shah, C. Campbell, I. Obeid, and J. Picone, “Improved Spatio-Temporal Modeling in Automated Seizure Detection using Channel-Dependent Posteriors,” Neurocomputing, 2021. [8] W. Tatum, A. Husain, S. Benbadis, and P. Kaplan, Handbook of EEG Interpretation. New York City, New York, USA: Demos Medical Publishing, 2007. [9] D. P. Bovet and C. Marco, Understanding the Linux Kernel, 3rd ed. O’Reilly Media, Inc., 2005. [10] V. Shah et al., “The Temple University Hospital Seizure Detection Corpus,” Front. Neuroinform., vol. 12, pp. 1–6, 2018. [11] F. Pedregosa et al., “Scikit-learn: Machine Learning in Python,” J. Mach. Learn. Res., vol. 12, pp. 2825–2830, 2011. [12] J. Gotman, D. Flanagan, J. Zhang, and B. Rosenblatt, “Automatic seizure detection in the newborn: Methods and initial evaluation,” Electroencephalogr. Clin. Neurophysiol., vol. 103, no. 3, pp. 356–362, 1997. 
    more » « less
  2. 1-Amino-1-hydrazino-2,2-dinitroethylene (HFOX) is a potential reactive intermediate for a new class of energetic materials. Now we describe its condensation with various carbonyl compounds in the presence of acidic and basic catalysts. Condensation of HFOX with α-diones and β-diones gives products of much interest. α-Diones undergo cyclization in the presence of base to form six-membered ring products, while β-diones cyclize to five-membered ring products in the presence of acid. One of the exciting reactions is the formation of ammonium (5,6-dimethyl-1,2,4-triazin-3-yl)dinitromethanide salt, 5c, which was isolated by using aqueous ammonia as a nucleophilic base. All new compounds were fully characterized by advanced spectroscopic techniques. The structures of 5, 5c, 5e, 9, 11, and 12a are supported by single crystal X-ray diffraction analysis. Most of the new six membered ring compounds have good thermostabilities (>200 °C), while the fluorinated five membered ring compound, 12b, has the highest density of 2.04 g cm −3 at 25 °C. Heats of formation and detonation properties were calculated by using Gaussian 03 and EXPLO5 software programs. Nearly all new compounds have very good detonation properties, especially, triazine salts, 5e ( D v = 7513 m s −1 ; P = 24.45 G P a), and 5f ( D v = 7948 m s −1 ; P = 26.27 G P a) as well as azide derivative 11 ( D v = 8166 m s −1 ; P = 25.48 G P a), which are superior to TNT ( D v = 6824 m s −1 ; P = 19.40 G P a). These findings provide a new perspective for the synthesis of novel high performing energetic materials. 
    more » « less
  3. Secondary organic aerosol (SOA) is a significant component of atmospheric fine particulate matter (PM2.5) globally that can form through multiphase chemistry of oxidized volatile organic compounds (VOC) leading to lower‐volatility particulate species. Condensed phase reactions of certain SOA constituents with inorganic sulfate derived from SO2 oxidation will lead to the formation of organosulfates, which can account for up to 10 – 15% of the organic mass within PM2.5. Despite the ubiquitous presence of atmospheric fine particulate organosulfates, our fundamental understanding of the molecular structure of organosulfates is limited, including for 2‐methyltetrol organosulfates (2‐MTSs), which are typically the single most abundant organosulfates measured in PM2.5, formed from isoprene oxidation products. As atmospheric aerosol pH varies widely (0 – 6), it is important to know whether organosulfates exist primarily in their protonated (ROSO3H) or deprotonated (ROSO3 ‐) forms. In this study, vibrational modes of synthetically‐pure 2‐MTSs were spectroscopically probed using Raman and infrared (IR) spectroscopies, supported by density functional theory (DFT) of the protonated and deprotonated structures. Vibrational bands at 1035 and 1059 cm‐1 were seen in both the IR and Raman spectra, and were associated with the ROSO3 ‐ anion by comparison to DFT calculations. Analysis of Raman spectra across a range of acidities (pH = 0 – 10) shows that 2‐MTSs are deprotonated (ROSO3 ‐) at those pH values. Additional DFT calculations for organosulfates derived from isoprene, α‐pinene, β‐caryophyllene, and toluene suggest that most organosulfates exist in their deprotonated form (ROSO3 ‐) in atmospheric particles. These charged species may have significant implications for our understanding of aerosol acidity and should be considered in thermodynamic model calculations. 
    more » « less
  4. Abstract. Triplet excited states of organic matter are formed when colored organicmatter (i.e., brown carbon) absorbs light. While these “triplets” can beimportant photooxidants in atmospheric drops and particles (e.g., theyrapidly oxidize phenols), very little is known about their reactivity towardmany classes of organic compounds in the atmosphere. Here we measure thebimolecular rate constants of the triplet excited state of benzophenone(3BP), a model species, with 17 water-solubleC3C6 alkenes that have either been found in theatmosphere or are reasonable surrogates for identified species. Measured rateconstants (kALK+3BP) vary by a factor of 30 and are in therange of (0.24–7.5) ×109 M−1 s−1. Biogenic alkenesfound in the atmosphere – e.g., cis-3-hexen-1-ol, cis-3-hexenyl acetate, andmethyl jasmonate – react rapidly, with rate constants above 1×109 M−1 s−1. Rate constants depend on alkene characteristicssuch as the location of the double bond, stereochemistry, and alkylsubstitution on the double bond. There is a reasonable correlation betweenkALK+3BP and the calculated one-electron oxidation potential(OP) of the alkenes (R2=0.58); in contrast, rate constants are notcorrelated with bond dissociation enthalpies, bond dissociation freeenergies, or computed energy barriers for hydrogen abstraction. Using the OPrelationship, we estimate aqueous rate constants for a number of unsaturatedisoprene and limonene oxidation products with 3BP: values are inthe range of (0.080–1.7) ×109 M−1 s−1, withgenerally faster values for limonene products. Rate constants with lessreactive triplets, which are probably more environmentally relevant, arelikely roughly 25 times slower. Using our predicted rate constants, alongwith values for other reactions from the literature, we conclude thattriplets are probably minor oxidants for isoprene- and limonene-relatedcompounds in cloudy or foggy atmospheres, except in cases in which the tripletsare very reactive.

    more » « less
  5. Abstract. Oceanic emissions of dimethyl sulfide (CH3SCH3,DMS) have long been recognized to impact aerosol particle composition andsize, the concentration of cloud condensation nuclei (CCN), and Earth'sradiation balance. The impact of oceanic emissions of methanethiol(CH3SH, MeSH), which is produced by the same oceanic precursor as DMS,on the volatile sulfur budget of the marine atmosphere is largelyunconstrained. Here we present direct flux measurements of MeSH oceanicemissions using the eddy covariance (EC) method with a high-resolutionproton-transfer-reaction time-of-flight mass spectrometer (PTR-ToFMS)detector and compare them to simultaneous flux measurements of DMS emissionsfrom a coastal ocean site. Campaign mean mixing ratios of DMS and MeSH were72 ppt (28–90 ppt interquartile range) and 19.1 ppt (7.6–24.5 pptinterquartile range), respectively. Campaign mean emission fluxes of DMS (FDMS) and MeSH (FMeSH) were 1.13 ppt m s−1 (0.53–1.61 ppt m s−1 interquartile range) and 0.21 ppt m s−1 (0.10–0.31 ppt m s−1 interquartile range), respectively. Linear least squares regression of observed MeSH and DMS flux indicates the emissions are highly correlatedwith each other (R2=0.65) over the course of the campaign,consistent with a shared oceanic source. The campaign mean DMS to MeSH fluxratio (FDMS:FMeSH) was 5.5 ± 3.0, calculated from the ratio of 304 individual coincident measurements of FDMS and FMeSH. Measured FDMS:FMeSH was weakly correlated (R2=0.15) withocean chlorophyll concentrations, with FDMS:FMeSH reaching a maximumof 10.8 ± 4.4 during a phytoplankton bloom period. No other volatilesulfur compounds were observed by PTR-ToFMS to have a resolvable emissionflux above their flux limit of detection or to have a gas-phase mixing ratio consistently above their limit of detection during the study period,suggesting DMS and MeSH are the dominant volatile organic sulfur compoundsemitted from the ocean at this site. The impact of this MeSH emission source on atmospheric budgets of sulfurdioxide (SO2) was evaluated by implementing observed emissions in a coupled ocean–atmosphere chemical box model using a newly compiled MeSHoxidation mechanism. Model results suggest that MeSH emissions lead toafternoon instantaneous SO2 production of 2.5 ppt h−1, which results in a 43 % increase in total SO2 production compared to a casewhere only DMS emissions are considered and accounts for 30% of theinstantaneous SO2 production in the marine boundary layer at the meanmeasured FDMS and FMeSH. This contribution of MeSH to SO2production is driven by a higher effective yield of SO2 from MeSHoxidation and the shorter oxidation lifetime of MeSH compared to DMS. Thislarge additional source of marine SO2 has not been previouslyconsidered in global models of marine sulfur cycling. The field measurementsand modeling results presented here demonstrate that MeSH is an importantcontributor to volatile sulfur budgets in the marine atmosphere and must be measured along with DMS in order to constrain marine sulfur budgets. Thislarge additional source of marine–reduced sulfur from MeSH will contribute to particle formation and growth and CCN abundance in the marine atmosphere, with subsequent impacts on climate. 
    more » « less