skip to main content

Title: Relevance of the Nuclear Structure of the Stable Ge Isotopes to the Neutrino-less Double-Beta Decay of 76 Ge
Gamma-ray detection following the inelastic neutron scattering reaction on isotopically enriched material was used to study the nuclear structure of 74 Ge. From these measurements, low-lying, low-spin excited states were characterized, new states and their decays were identified, level lifetimes were measured with the Doppler-shift attenuation method (DSAM), multipole mixing ratios were established, and transition probabilities were determined. New structural features in 74 Ge were identified, and the reanalysis of older 76 Ge data led to the placement of the 2 + member of the intruder band. In addition, a number of previously placed states in 74 Ge were shown not to exist. A procedure for future work, which will lead to meaningful data for constraining calculations of the neutrinoless double-beta decay matrix element, is suggested.
Authors:
; ; ; ; ; ; ;
Award ID(s):
1913028
Publication Date:
NSF-PAR ID:
10150282
Journal Name:
EPJ Web of Conferences
Volume:
232
Page Range or eLocation-ID:
04011
ISSN:
2100-014X
Sponsoring Org:
National Science Foundation
More Like this
  1. Obeid, I. (Ed.)
    The Neural Engineering Data Consortium (NEDC) is developing the Temple University Digital Pathology Corpus (TUDP), an open source database of high-resolution images from scanned pathology samples [1], as part of its National Science Foundation-funded Major Research Instrumentation grant titled “MRI: High Performance Digital Pathology Using Big Data and Machine Learning” [2]. The long-term goal of this project is to release one million images. We have currently scanned over 100,000 images and are in the process of annotating breast tissue data for our first official corpus release, v1.0.0. This release contains 3,505 annotated images of breast tissue including 74 patients with cancerous diagnoses (out of a total of 296 patients). In this poster, we will present an analysis of this corpus and discuss the challenges we have faced in efficiently producing high quality annotations of breast tissue. It is well known that state of the art algorithms in machine learning require vast amounts of data. Fields such as speech recognition [3], image recognition [4] and text processing [5] are able to deliver impressive performance with complex deep learning models because they have developed large corpora to support training of extremely high-dimensional models (e.g., billions of parameters). Other fields that do notmore »have access to such data resources must rely on techniques in which existing models can be adapted to new datasets [6]. A preliminary version of this breast corpus release was tested in a pilot study using a baseline machine learning system, ResNet18 [7], that leverages several open-source Python tools. The pilot corpus was divided into three sets: train, development, and evaluation. Portions of these slides were manually annotated [1] using the nine labels in Table 1 [8] to identify five to ten examples of pathological features on each slide. Not every pathological feature is annotated, meaning excluded areas can include focuses particular to these labels that are not used for training. A summary of the number of patches within each label is given in Table 2. To maintain a balanced training set, 1,000 patches of each label were used to train the machine learning model. Throughout all sets, only annotated patches were involved in model development. The performance of this model in identifying all the patches in the evaluation set can be seen in the confusion matrix of classification accuracy in Table 3. The highest performing labels were background, 97% correct identification, and artifact, 76% correct identification. A correlation exists between labels with more than 6,000 development patches and accurate performance on the evaluation set. Additionally, these results indicated a need to further refine the annotation of invasive ductal carcinoma (“indc”), inflammation (“infl”), nonneoplastic features (“nneo”), normal (“norm”) and suspicious (“susp”). This pilot experiment motivated changes to the corpus that will be discussed in detail in this poster presentation. To increase the accuracy of the machine learning model, we modified how we addressed underperforming labels. One common source of error arose with how non-background labels were converted into patches. Large areas of background within other labels were isolated within a patch resulting in connective tissue misrepresenting a non-background label. In response, the annotation overlay margins were revised to exclude benign connective tissue in non-background labels. Corresponding patient reports and supporting immunohistochemical stains further guided annotation reviews. The microscopic diagnoses given by the primary pathologist in these reports detail the pathological findings within each tissue site, but not within each specific slide. The microscopic diagnoses informed revisions specifically targeting annotated regions classified as cancerous, ensuring that the labels “indc” and “dcis” were used only in situations where a micropathologist diagnosed it as such. Further differentiation of cancerous and precancerous labels, as well as the location of their focus on a slide, could be accomplished with supplemental immunohistochemically (IHC) stained slides. When distinguishing whether a focus is a nonneoplastic feature versus a cancerous growth, pathologists employ antigen targeting stains to the tissue in question to confirm the diagnosis. For example, a nonneoplastic feature of usual ductal hyperplasia will display diffuse staining for cytokeratin 5 (CK5) and no diffuse staining for estrogen receptor (ER), while a cancerous growth of ductal carcinoma in situ will have negative or focally positive staining for CK5 and diffuse staining for ER [9]. Many tissue samples contain cancerous and non-cancerous features with morphological overlaps that cause variability between annotators. The informative fields IHC slides provide could play an integral role in machine model pathology diagnostics. Following the revisions made on all the annotations, a second experiment was run using ResNet18. Compared to the pilot study, an increase of model prediction accuracy was seen for the labels indc, infl, nneo, norm, and null. This increase is correlated with an increase in annotated area and annotation accuracy. Model performance in identifying the suspicious label decreased by 25% due to the decrease of 57% in the total annotated area described by this label. A summary of the model performance is given in Table 4, which shows the new prediction accuracy and the absolute change in error rate compared to Table 3. The breast tissue subset we are developing includes 3,505 annotated breast pathology slides from 296 patients. The average size of a scanned SVS file is 363 MB. The annotations are stored in an XML format. A CSV version of the annotation file is also available which provides a flat, or simple, annotation that is easy for machine learning researchers to access and interface to their systems. Each patient is identified by an anonymized medical reference number. Within each patient’s directory, one or more sessions are identified, also anonymized to the first of the month in which the sample was taken. These sessions are broken into groupings of tissue taken on that date (in this case, breast tissue). A deidentified patient report stored as a flat text file is also available. Within these slides there are a total of 16,971 total annotated regions with an average of 4.84 annotations per slide. Among those annotations, 8,035 are non-cancerous (normal, background, null, and artifact,) 6,222 are carcinogenic signs (inflammation, nonneoplastic and suspicious,) and 2,714 are cancerous labels (ductal carcinoma in situ and invasive ductal carcinoma in situ.) The individual patients are split up into three sets: train, development, and evaluation. Of the 74 cancerous patients, 20 were allotted for both the development and evaluation sets, while the remain 34 were allotted for train. The remaining 222 patients were split up to preserve the overall distribution of labels within the corpus. This was done in hope of creating control sets for comparable studies. Overall, the development and evaluation sets each have 80 patients, while the training set has 136 patients. In a related component of this project, slides from the Fox Chase Cancer Center (FCCC) Biosample Repository (https://www.foxchase.org/research/facilities/genetic-research-facilities/biosample-repository -facility) are being digitized in addition to slides provided by Temple University Hospital. This data includes 18 different types of tissue including approximately 38.5% urinary tissue and 16.5% gynecological tissue. These slides and the metadata provided with them are already anonymized and include diagnoses in a spreadsheet with sample and patient ID. We plan to release over 13,000 unannotated slides from the FCCC Corpus simultaneously with v1.0.0 of TUDP. Details of this release will also be discussed in this poster. Few digitally annotated databases of pathology samples like TUDP exist due to the extensive data collection and processing required. The breast corpus subset should be released by November 2021. By December 2021 we should also release the unannotated FCCC data. We are currently annotating urinary tract data as well. We expect to release about 5,600 processed TUH slides in this subset. We have an additional 53,000 unprocessed TUH slides digitized. Corpora of this size will stimulate the development of a new generation of deep learning technology. In clinical settings where resources are limited, an assistive diagnoses model could support pathologists’ workload and even help prioritize suspected cancerous cases. ACKNOWLEDGMENTS This material is supported by the National Science Foundation under grants nos. CNS-1726188 and 1925494. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation. REFERENCES [1] N. Shawki et al., “The Temple University Digital Pathology Corpus,” 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 City, New York, USA: Springer, 2020, pp. 67 104. https://www.springer.com/gp/book/9783030368432. [2] J. Picone, T. Farkas, I. Obeid, and Y. Persidsky, “MRI: High Performance Digital Pathology Using Big Data and Machine Learning.” Major Research Instrumentation (MRI), Division of Computer and Network Systems, Award No. 1726188, January 1, 2018 – December 31, 2021. https://www. isip.piconepress.com/projects/nsf_dpath/. [3] A. Gulati et al., “Conformer: Convolution-augmented Transformer for Speech Recognition,” in Proceedings of the Annual Conference of the International Speech Communication Association (INTERSPEECH), 2020, pp. 5036-5040. https://doi.org/10.21437/interspeech.2020-3015. [4] C.-J. Wu et al., “Machine Learning at Facebook: Understanding Inference at the Edge,” in Proceedings of the IEEE International Symposium on High Performance Computer Architecture (HPCA), 2019, pp. 331–344. https://ieeexplore.ieee.org/document/8675201. [5] I. Caswell and B. Liang, “Recent Advances in Google Translate,” Google AI Blog: The latest from Google Research, 2020. [Online]. Available: https://ai.googleblog.com/2020/06/recent-advances-in-google-translate.html. [Accessed: 01-Aug-2021]. [6] V. Khalkhali, N. Shawki, V. Shah, M. Golmohammadi, I. Obeid, and J. Picone, “Low Latency Real-Time Seizure Detection Using Transfer Deep Learning,” in Proceedings of the IEEE Signal Processing in Medicine and Biology Symposium (SPMB), 2021, pp. 1 7. https://www.isip. piconepress.com/publications/conference_proceedings/2021/ieee_spmb/eeg_transfer_learning/. [7] J. Picone, T. Farkas, I. Obeid, and Y. Persidsky, “MRI: High Performance Digital Pathology Using Big Data and Machine Learning,” Philadelphia, Pennsylvania, USA, 2020. https://www.isip.piconepress.com/publications/reports/2020/nsf/mri_dpath/. [8] I. Hunt, S. Husain, J. Simons, I. Obeid, and J. Picone, “Recent Advances in the Temple University Digital Pathology Corpus,” in Proceedings of the IEEE Signal Processing in Medicine and Biology Symposium (SPMB), 2019, pp. 1–4. https://ieeexplore.ieee.org/document/9037859. [9] A. P. Martinez, C. Cohen, K. Z. Hanley, and X. (Bill) Li, “Estrogen Receptor and Cytokeratin 5 Are Reliable Markers to Separate Usual Ductal Hyperplasia From Atypical Ductal Hyperplasia and Low-Grade Ductal Carcinoma In Situ,” Arch. Pathol. Lab. Med., vol. 140, no. 7, pp. 686–689, Apr. 2016. https://doi.org/10.5858/arpa.2015-0238-OA.« less
  2. In situ electrochemical cells were assembled with an amorphous germanium (a-Ge) film as working electrode and sodium foil as reference and counter electrode. The stresses generated in a-Ge electrodes due to electrochemical reaction with sodium were measured in real-time during the galvanostatic cycling. A specially designed patterned a-Ge electrode was cycled against sodium and the corresponding volume changes were measured using an AFM; it was observed that sodiation/desodiation of a-Ge results in more than 300% volume change, consistent with literature. The potential and stress response showed that the a-Ge film undergoes irreversible changes during the first sodiation process, but the subsequent desodiation/sodiation cycles are reversible. The stress response of the film reached steady-state after the initial sodiation and is qualitatively similar to the response of Ge during lithiation, i.e., initial linear elastic response followed by extensive plastic deformation of the film to accommodate large volume changes. However, despite being bigger ion, sodiation of Ge generated lower stress levels compared to lithiation. Consequently, the mechanical dissipation losses associated with plastic deformation are lower during sodiation process than it is for lithiation.

  3. Abstract We reassess the 65 As(p, γ ) 66 Se reaction rates based on a set of proton thresholds of 66 Se, S p ( 66 Se), estimated from the experimental mirror nuclear masses, theoretical mirror displacement energies, and full p f -model space shell-model calculation. The self-consistent relativistic Hartree–Bogoliubov theory is employed to obtain the mirror displacement energies with much reduced uncertainty, and thus reducing the proton-threshold uncertainty up to 161 keV compared to the AME2020 evaluation. Using the simulation instantiated by the one-dimensional multi-zone hydrodynamic code, K epler , which closely reproduces the observed GS 1826−24 clocked bursts, the present forward and reverse 65 As(p, γ ) 66 Se reaction rates based on a selected S p ( 66 Se) = 2.469 ± 0.054 MeV, and the latest 22 Mg( α ,p) 25 Al, 56 Ni(p, γ ) 57 Cu, 57 Cu(p, γ ) 58 Zn, 55 Ni(p, γ ) 56 Cu, and 64 Ge(p, γ ) 65 As reaction rates, we find that though the GeAs cycles are weakly established in the rapid-proton capture process path, the 65 As(p, γ ) 66 Se reaction still strongly characterizes the burst tail end due to the two-proton sequential capturemore »on 64 Ge, not found by the Cyburt et al. sensitivity study. The 65 As(p, γ ) 66 Se reaction influences the abundances of nuclei A = 64, 68, 72, 76, and 80 up to a factor of 1.4. The new S p ( 66 Se) and the inclusion of the updated 22 Mg( α ,p) 25 Al reaction rate increases the production of 12 C up to a factor of 4.5, which is not observable and could be the main fuel for a superburst. The enhancement of the 12 C mass fraction alleviates the discrepancy in explaining the origin of the superburst. The waiting point status of and two-proton sequential capture on 64 Ge, the weak-cycle feature of GeAs at a region heavier than 64 Ge, and the impact of other possible S p ( 66 Se) are also discussed.« less
  4. Liu, W. ; Wang, Y. ; Guo, B. ; Tang, X. ; Zeng, S. (Ed.)
    In Type-I X-ray bursts (XRBs), the rapid-proton capture (rp-) process passes through the NiCu and ZnGa cycles before reaching the region above Ge and Se isotopes that hydrogen burning actively powers the XRBs. The sensitivity study performed by Cyburt et al . [1] shows that the 57 Cu(p, γ ) 58 Zn reaction in the NiCu cycles is the fifth most important rp-reaction influencing the burst light curves. Langer et al . [2] precisely measured some low-lying energy levels of 58 Zn to deduce the 57 Cu(p, γ ) 58 Zn reaction rate. Nevertheless, the order of the 1 + 1 and 2 + 3 resonance states that dominate at 0:2 ≲ T (GK) ≲ 0:8 is not confirmed. The 1 + 2 resonance state, which dominates at the XRB sensitive temperature regime 0:8 ≲ T (GK) ≲ 2 was not detected. Using isobaric-multipletmass equation (IMME), we estimate the order of the 1 + 1 and 2 + 3 resonance states and estimate the lower limit of the 1 + 2 resonance energy. We then determine the 57 Cu(p, γ ) 58 Zn reaction rate using the full pf -model space shell model calculations. The new rate is up tomore »a factor of four lower than the Forstner et al . [3] rate recommended by JINA REACLIBv2.2. Using the present 57 Cu(p, γ ) 58 Zn, the latest 56 Ni(p, γ ) 57 Cu and 55 Ni(p, γ ) 56 Cu reaction rates, and 1D implicit hydrodynamic K epler code, we model the thermonuclear XRBs of the clocked burster GS 1826–24. We find that the new rates regulate the reaction flow in the NiCu cycles and strongly influence the burst-ash composition. The 59 Cu(p, γ ) 56 Ni and 59 Cu(p, α ) 60 Zn reactions suppress the influence of the 57 Cu(p, γ ) 58 Zn reaction. They strongly diminish the impact of the nuclear reaction flow that bypasses the 56 Ni waiting point induced by the 55 Ni(p, γ ) 56 Cu reaction on burst light curve.« less
  5. Single crystals of two new germanates, [(Rb 6 F)(Rb 4 F)][Ge 14 O 32 ] and [(Rb 6 F)(Rb 3.1 Co 0.9 F 0.96 )][Co 3.8 Ge 10.2 O 30 F 2 ], were synthesized via high temperature RbCl/RbF flux growth. Both compounds crystallize in the cubic space group F 4̄3 m and possess the germanium framework of the previously reported salt inclusion material (SIM), [(Cs 6 F)(Cs 3 AgF)][Ge 14 O 32 ], related to the Ge 7 O 16 zeolitic family. These materials demonstrate the ability to accommodate a variety of salt-inclusions, and exhibit chemical flexibility enabling modifications of the framework through incorporation of Co. Alteration of the salt-inclusion led to intrinsic luminescence of [(Rb 6 F)(Rb 4 F)][Ge 14 O 32 ] while modification of the framework resulted in an unanticipated Rb/Co salt/inclusion in [(Rb 6 F)(Rb 3.1 Co 0.9 F 0.96 )][Co 3.8 Ge 10.2 O 30 F 2 ]. Fluorescence measurements were performed on [(Rb 6 F)(Rb 4 F)][Ge 14 O 32 ]. First-principles calculations in the form of density functional theory (DFT) were performed for [(Rb 6 F)(Rb 3.1 Co 0.9 F 0.96 )][Co 3.8 Ge 10.2 O 30 F 2 ] to elucidatemore »its electronic and magnetic properties, and stability at 0 K.« less