搜

... Atomic composition/configuration dependent bulk moduli of AlC composites Cite as: AIP Advances 12, 115008 (2022); https://doi.org/10.1063/5.0117900 Submitted: 08 August 2022 Accepted: 11 October 2022 Published Online: 04 November 2022 Hansika I. Sirikumara, Wilson Rativa-Parada, Robinson Karunanithy, et al. AIP Advances 12, 115008 (2022); https://doi.org/10.1063/5.0117900 2022 Author(s). 12, 115008 AIP Advances ARTICLE scitation.org/journal/adv Atomic composition/configuration dependent bulk moduli of AlC composites Cite as: AIP Advances 12, 115008 (2022); doi: 10.1063/5.0117900 Submitted: 8 August 2022 Accepted: 11 October 2022 Published Online: 4 November 2022 Hansika I. Sirikumara,1,a) Wilson Rativa-Parada,2 Robinson Karunanithy,3 Poopalasingam Sivakumar,3 Sabrina Nilufar,2,b) and Thushari Jayasekera3 AFFILIATIONS 1 E. S. Witchger School of Engineering, Marian University Indianapolis, Indianapolis, Indiana 46222, USA School of Mechanical, Aerospace, and Materials Engineering, Southern Illinois University Carbondale, Carbondale, Illinois 62901, USA 3 School of Physics and Applied Physics, Southern Illinois University Carbondale, Carbondale, Illinois 62901, USA 2 a) Author to whom correspondence should be addressed: hsirikumara@marian.edu Electronic mail: sabrina.nilufar@siu.edu b) ABSTRACT Embedding carbon in metals has long been known to enhance the mechanical properties of metal carbon composites. We report the possibility of growing AlC composites by the hot isostatic pressing method, with carbon embedded into an Al lattice in graphitic form without the formation of Al4 C3 . Raman spectroscopy confirms the formation of sp2 -hybridized carbon clusters in the aluminum lattice. The bulk moduli of the samples were measured to be between 60 and 100 GPa. From the results of first principles density functional theory calculations, we show that the formation of sp2 -hybridized carbon clusters is more stable than having isolated C scatterers in aluminum. Our results show that the extended network of C clusters shows a higher bulk modulus while isolated scattering centers could lower the bulk modulus. We explain this behavior with the analysis of total charge distribution. Localization of charge density decreases materials ability to respond to external stress, thus showing a reduced bulk modulus. Some defect configuration may reduce the symmetry while others keep the symmetry of the host configuration even for the same chemical composition of AlC composites. 2022 Author(s). All article content, except where otherwise noted, is licensed under a Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/). https://doi.org/10.1063/5.0117900 Aluminum (Al) and its alloys are of interest because of their significant role in applications requiring lightweight yet strong materials utilized in the building of infrastructure such as transmission of electricity in high-voltage power lines, fabrication of high performance electronics, and construction of lightweight aerospace, naval, and automotive vehicles. Several strengthening techniques have been employed in the development of Al alloys and composites, such as solid solutions, precipitation strengthening, etc.1 Aluminum matrix composites based on particulate reinforcement provide an excellent approach to improve the mechanical properties, with a combination of high specific strength, high stiffness, and lightweight.2,3 Various ceramic reinforcements, such as SiC, B4 C, TiC, Al2 O3 , and different carbon allotropes, are commonly used in the manufacturing of discontinuously reinforced Al matrix composites.4 Graphene, a carbon allotrope, is used as the reinforcement in this study. In the process of incorporating carbon into the Al matrix, AIP Advances 12, 115008 (2022); doi: 10.1063/5.0117900 Author(s) 2022 it is important to limit the formation of Al4 C3 while preserving the sp2 bonding network of carbon. In this work, we report the fabrication of the AlC composite with a graphitic network of carbon by the hot isostatic pressing method. Raman spectroscopy confirms the formation of sp2 hybridized graphitic networks in Al. Our mechanical strength measurements show a bulk modulus in the range of 60100 GPa in AlC composites. Using the results from first principles Density Functional Theory (DFT) calculations, we explore the reasoning for such variation in the bulk modulus of AlC composites at the atomistic level. Formation of a graphene sheet on the Al(111) surface has been previously studied by various groups.57 However, the formation of C islands within the Al matrix has not been studied yet. In this work, we will discuss the bulk modulus of AlC composites with C scatterers and sp2 -hybridized graphitic C islands in the Al matrix. Our results suggest that C scatterers in the Al matrix will 12, 115008-1 AIP Advances ARTICLE result in localization of the charge density, thus decreasing the bulk modulus. An extended network of C in Al will preserve the uniform charge distribution, which will result in an increase in the bulk modulus. Here, we explain the methodology for sample preparation, characterization of samples with Raman spectroscopy, measurement of the bulk modulus, and the detailed computational approach for calculating the bulk modulus of AlC composites using first principles Density Functional Theory (DFT). Al 6061 powder used as the base material (mesh size: 140/+325) was obtained from READE Advanced Materials. Graphene nanoplatelets with an average thickness of 28 nm, an average number of layers of 36, and 99.5+% purity were obtained from US Research Nanomaterials Inc. Both Al 6061 neat- and Al 6061-graphene composites were produced by powder metallurgy. The composites were mechanically milled using a SPEX 8000 mixer (SPEX corp.) for 4 h, with 1 wt. % of stearic acid [CH3 (CH2 )16COOH] obtained from Sigma-Aldrich (95%) used as a lubricant agent. The stearic acid was evaporated for 30 min at 450 C before pressing. Hot isostatic pressing was used to consolidate the mechanically milled powder composites with 0.5 and 1 volume fractions of graphene reinforcement. The mixed powder was placed in a graphite die with a diameter of 25.4 mm and pressed for 1 h at 500 C and 70 MPa under vacuum conditions in a frontloading hot press furnace (Materials Research Furnaces Inc.). A Horiba iHR550 imaging spectrometer with a near-infrared (NIR) excitation light source at a wavelength of 785 nm (iBeamSmart-785-S-WS, TOPTICA Photonics) was utilized for the Raman study. The system is equipped with an Olympus BX 41 microscope with 10, 20, 50, and 100 magnification objectives. A grating of 600 gr/mm was used in the spectrometer. Except for carbon power samples, the spectra were collected at 10 magnification at 120 mW laser power with 15-s acquisition time, and a total of ten scans were obtained over the desired range. The 100 objective with 10 mW laser power was used for carbon powder samples to avoid burning the samples. Compressive tests were performed according to the standard ASTM E9 at room temperature in an MTS Insight 30 kN testing machine with a constant crosshead speed of 0.05 mm/min. The samples were machined to rectangular shapes with a length/width ratio of 2:1. Three samples of each composite material (neat, 0.5 and 1 vol. % of graphene) were tested in accordance with the ASTM E9 standard. The bulk modulus of a system is defined as B = V scitation.org/journal/adv of the bulk modulus. The corresponding pressure (P) can also be dE written as P = dV . In order to extract the value of the bulk modulus, B, we calculated the energy of the AlC composites for a series of lattice constants around the equilibrium lattice constant and then fitted it to the Murnaghan equation of state through implementation in the ev.x code in the Quantum Espresso (QE) package.10 Aluminum (Al) crystallizes in an FCC structure with a lattice constant of 4.04 . There are four atoms in a non-primitive cubic unit cell. In the present work, we considered a 2 2 2 supercell of aluminum with 32 atomic sites. This super cell allows us to study the 3.125% of C in aluminum, where only one out of 32 Al atoms is substituted with C. We considered three types of C dopants in Al: pure substitutional, pure interstitial, and mixture of substitutional/interstitial dopants. Our calculation cell with the composition Al(32n) Cm is denoted as Cnm , for which n = m is a purely substitutional configuration. All energy values for this study were calculated using first principles Density Functional Theory (DFT) through implementation in the Quantum Espresso (QE) package.11 The Generalized Gradient Approximation of Perdew, Burke, and Ernzerhof (GGA-PBE) was used for the exchange and correlation functionals with a 40 Ry energy cut-off for plane wave expansion.12 A 12 12 12 MonkhorstPack grid was used to sample the Brillouin zone, and all the structures were optimized to forces less than 0.025 eV/. dP , dV which can be calculated with the equation of state, the relationship between the volume (V) of a material and its pressure (P).8 In this work, B is calculated using by fitting the energyvolume curve to the Murnaghan equation of state, P(V) = ( B V0 B )[( ) 1], B V which assumes a linear behavior for the bulk modulus with respect to pressure.9 Here, B is the bulk modulus, and B is the first derivative AIP Advances 12, 115008 (2022); doi: 10.1063/5.0117900 Author(s) 2022 FIG. 1. (a) Upper panel shows the experimentally measured bulk modulus of nine samples of Al and AlC composites, and (b) the lower panel shows the Raman spectra obtained for Al with no carbon in comparison with the two samples of the AlC composite. 12, 115008-2 AIP Advances In order to get a deeper understanding on the relation between the change in the bulk modulus and the atomic configuration, we analyzed the charge density distribution of each system. Charge density is an indicator of the symmetry of the system. Once the symmetry is broken, it has the ability to respond to the reduction in the external force, thus reducing the bulk modulus. We also calculated gamma-point phonon frequencies of the selected AlC composites to further understand the nature of bonding, which was also observed through Raman spectroscopy. The phonon frequencies were calculated using density functional perturbation theory through implementation in the Quantum Espresso package. Measurements of the bulk modulus for nine samples show a variation from 60 to 100 GPa, as shown in the upper panel of Fig. 1. These nine samples have three different chemical composites. The compressive modulus is determined for three samples of each kind, namely, neat Al and 0.5 and 1 vol. % graphene reinforced Al. The experimental modulus shows scattered results for different samples for the same chemical composition. Raman spectra (lower panel of Fig. 1) shows a distinct G-peak (1600 cm1 ) and D-peak (1290 cm1 ) compared to the Raman spectra of the control sample with no carbon. This is a clear indication of the presence of graphitic carbon in Al. Iftekhar Jaim et al. studied the presence of epitaxial graphene on the Al surface.5 In this work, we study the effect of isolated carbon scatterers and planar carbon networks on the structure and bulk modulus of AlC composites. Figure 2 shows the total energy as a function of volume for (a) pristine Al and (b) C01 (purely substitutional doped) and ARTICLE scitation.org/journal/adv TABLE I. Structural information and the bulk modulus of AlC composites (a: lattice constant; B: bulk modulus). Configuration C00 C01 C10 C20 C02 (a) C02 (b) C11 C60 (a) C60 (b) C18 0 Composition a () B (GPa) Al32 C0 Al31 C1 Al32 C1 Al32 C2 Al30 C2 Al30 C2 Al31 C2 Al32 C6 Al32 C6 Al32 C6 8.08 8.07 8.106 8.13 8.06 8.11 8.04 8.22 8.42 8.38 77.3 70.6 79.9 84.2 61.2 71.2 78.9 93.8 63.4 94.4 (c) C10 (purely interstitial doped) configurations. In the substitutional doped configuration, C01 [shown in Fig. 2(b)], one out of 32 C atoms in the supercell is doped with a C atom, whereas an additional C atom takes one out of eight empty body centered positions in the purely interstitial configuration, C10 [shown in Fig. 2(c)]. The lattice constant and the bulk modulus of pristine Al are 4.04 and 77.3 GPa, respectively, which are in agreement with the previous theoretical and experimental results.13,14 The change in equilibrium lattice constants of the C01 and C10 configurations is 0.12% and FIG. 2. Upper panels show the total energy as a function of volume for (a) pristine Al and (b) C01 (purely substitutional doped) and (c) C10 (purely interstitial doped) configurations, with their charge density shown in the corresponding lower panels. AIP Advances 12, 115008 (2022); doi: 10.1063/5.0117900 Author(s) 2022 12, 115008-3 AIP Advances FIG. 3. Theoretically calculated bulk modulus of AlC configurations [1: C10 , 2: C01 , 1 3: C02 (a), 4: C02 (b), 6: C20 , 7: C60 (a), 8: C18 0 , and 9: C1 ]. 0.32%, respectively (Table I). In the energyvolume curves shown in Fig. 2, the volume-axis is centered at the equilibrium volume, and the zero of the energy-axis is the equilibrium energy of each configuration. ARTICLE scitation.org/journal/adv The lower panel of Fig. 2 shows the charge density of each configuration. The pristine configuration is a uniformly distributed delocalized charge configuration. When a single Al atom is substituted by a C atom, localization of charge density is observed as shown in panel (b) of Fig. 2, which results in an 8.6% decrease in the bulk modulus, to 70.6 GPa. Purely interstitial substitutions to the Al crystal cause only a slight disturbance to the charge density; thus, the bulk modulus increases by 3.3%, to 79.9 GPa. The localization of charge in the substitutional doped configuration ( C01 ) and delocalized charge distribution of the interstitial doped configuration explain this behavior, which is in agreement with the previously reported studies of elastic properties in other materials.15 In order to understand the variation in the bulk modulus from the experimental study [shown in Fig. 1(a)], we calculated the bulk moduli of seven other AlC composite configurations. The chemical compositions, lattice constants, and the bulk moduli of these configurations are shown in Table I. As shown in Table I, the bulk moduli of AlC composites vary between 60 and 100 GPa for those considered configurations (Fig. 3), which are described in detail below. C20 has two substitutional atoms, FIG. 4. Charge density distribution of (a) C20 , (b) C11 , (c) C02 (a), and (d) C02 (b) configurations. AIP Advances 12, 115008 (2022); doi: 10.1063/5.0117900 Author(s) 2022 12, 115008-4 AIP Advances ARTICLE scitation.org/journal/adv FIG. 5. Charge density distribution of (a) C60 (a), (b) C60 (b), and (c) C18 0 configurations. with the chemical composition Al32 C2 showing B = 84.2 GPa. This structure shows a delocalized charge distribution. The structure C11 with the chemical composition Al31 C1 has one substituted C atom and one interstitial C atom. This structure shows slight localization of charge density compared to C20 . The C11 structure shows B = 78.9 GPa. Both C02 (a) and C02 (b) have the same chemical composition Al30 C2 , that is, with two substituted C atoms. The relative positions of the defects are different in the two structures. Both structures show localized charge distribution, as shown in Figs. 4(c) and 4(d) with B = 61.2 and 71.2 GPa, respectively. We have performed calculations on three additional AlC composites to understand the behavior of C rings in the Al lattice. The first two structures have the chemical composition Al32 C6 , both with the configuration C60 , i.e., all six C atoms take interstitial positions. Our results show B = 63.4 GPa for C60 (a) with a ring-shaped carbon cluster and B = 93.8 GPa for C60 (b) with a pattern of scatterers, which causes less disturbance to the charge distribution, as shown in Fig. 5. Our calculations show that the ring-shaped cluster is more energetically favorable than the isolated scatterers. A ring-shaped carbon cluster in the Al matrix with the composition Al26 C18 shows a higher bulk modulus of 94.4 GPa. The extended C-network supports the localized charge distribution less, which results in a higher bulk modulus. Atomic displacements corresponding to the highest frequency phonons at the gamma point for the three configurations are shown in Fig. 6. When there are carbon scatterers [such as in configurations (a) C10 and (b) C60 (a)], the highest vibrational frequency is around 600 cm1 , and the corresponding atomic displacement shows a vibration of C atoms along with the Al-matrix, as shown in Figs. 6(a) and 6(b). In the C18 0 configuration, the highest vibrational frequency is shown at 1357 cm1 , and the corresponding atomic displacements are solely restricted to the plane of honey- FIG. 6. Atomic displacement pattern corresponding to the highest frequency of gamma-point phonons: (a) C10 , (b) C60 (a), and (c) C18 0 configurations. AIP Advances 12, 115008 (2022); doi: 10.1063/5.0117900 Author(s) 2022 12, 115008-5 AIP Advances combs in the C-cluster, in addition, it has similarities to the D-peak observed at 1290 cm1 for graphene,16 which further confirms our conclusion on formation of sp2 hybridized carbon clusters in the Al-matrix. Our work shows that graphitic carbon can form in the Al matrix by the hot isostatic process growth approach. Our calculations also show that ring-shaped structures are energetically more favorable than isolated scatterers. However, even the small rings or isolated scatterers can reduce the bulk modulus if defects are formed at the localized charge centers in the Al matrix. Extended ring-shaped graphitic carbon clusters can increase the bulk modulus of Al. This work was partially supported by an NSF grant (Award No. 2138459) and the NSF DMR and DoD Assure REU Program (Award No. 1757954). The authors would like to thank Southern Illinois University for providing the computer facilities. AUTHOR DECLARATIONS Conflict of Interest The authors have no conflicts to disclose. Author Contributions Hansika I. Sirikumara: Writing original draft (equal); Writing review & editing (equal). Wilson Rativa-Parada: Writing review & editing (supporting). Robinson Karunanithy: Methodology (supporting). Poopalasingam Sivakumar: Methodology (supporting). Sabrina Nilufar: Methodology (supporting); Writing review & editing (supporting). Thushari Jayasekera: Writing review & editing (supporting). DATA AVAILABILITY The data that support the findings of this study are available within the article. AIP Advances 12, 115008 (2022); doi: 10.1063/5.0117900 Author(s) 2022 ARTICLE scitation.org/journal/adv REFERENCES 1 Z. Wang, R. T. Qu, S. Scudino, B. A. Sun, K. G. Prashanth, D. V. LouzguineLuzgin, M. W. Chen, Z. F. Zhang, and J. Eckert, Hybrid nanostructured aluminum alloy with super-high strength, NPG Asia Mater. 7, e229 (2015). 2 N. Chawla and Y.-L. Shen, Mechanical behavior of particle reinforced metal matrix composites, Adv. Eng. Mater. 3, 357370 (2001). 3 C. Suryanarayana and N. Al-Aqeeli, Mechanically alloyed nanocomposites, Prog. Mater. Sci. 58, 383502 (2013). 4 C. A. Smith, Discontinuous Reinforcements for Metal-Matrix Composites (ASM International, Materials Park, OH, 2001), pp. 5155. 5 H. M. I. Jaim, R. A. Isaacs, S. N. Rashkeev, M. Kuklja, D. P. Cole, M. C. LeMieux, I. Jasiuk, S. Nilufar, and L. G. Salamanca-Riba, Sp2 carbon embedded in Al-6061 and Al-7075 alloys in the form of crystalline graphene nanoribbons, Carbon 107, 5666 (2016). 6 B. Sahoo, J. Joseph, A. Sharma, and J. Paul, Surface modification of aluminium by graphene impregnation, Mater. Des. 116, 5164 (2017). 7 S. Zhang, D. He, P. Huang, and F. Wang, Moir pattern at graphene/Al (111) interface: Experiment and simulation, Mater. Des. 201, 109509 (2021). 8 A. J. Jackson, J. M. Skelton, C. H. Hendon, K. T. Butler, and A. Walsh, Crystal structure optimisation using an auxiliary equation of state, J. Chem. Phys. 143, 184101 (2015). 9 F. D. Murnaghan, The compressibility of media under extreme pressures, Proc. Natl. Acad. Sci. U. S. A. 30, 244247 (1944). 10 F. Birch, Finite elastic strain of cubic crystals, Phys. Rev. 71, 809 (1947). 11 P. Giannozzi, S. Baroni, N. Bonini, M. Calandra, R. Car, C. Cavazzoni, D. Ceresoli, G. L. Chiarotti, M. Cococcioni, I. Dabo, A. Dal Corso, S. de Gironcoli, S. Fabris, G. Fratesi, R. Gebauer, U. Gerstmann, C. Gougoussis, A. Kokalj, M. Lazzeri, L. Martin-Samos, N. Marzari, F. Mauri, R. Mazzarello, S. Paolini, A. Pasquarello, L. Paulatto, C. Sbraccia, S. Scandolo, G. Sclauzero, A. P. Seitsonen, A. Smogunov, P. Umari, and R. M. Wentzcovitch, QUANTUM ESPRESSO: A modular and open-source software project for quantum simulations of materials, J. Phys.: Condens. Matter 21, 395502 (2009). 12 J. P. Perdew, K. Burke, and M. Ernzerhof, Generalized gradient approximation made simple, Phys. Rev. Lett. 77, 3865 (1996). 13 P. Jacobs, Y. F. Zhukovskii, Y. Mastrikov, and Y. N. Shunin, Bulk and surface properties of metallic aluminium: DFT simulations, Comput. Modell. New Technol. 6, 728 (2002). 14 W. Li and T. Wang, Ab initio investigation of the elasticity and stability of aluminium, J. Phys.: Condens. Matter 10, 9889 (1998). 15 M. Woodcox, J. Young, and M. Smeu, Ab initio investigation of the elastic properties of bismuth-based alloys, Phys. Rev. B 100, 104105 (2019). 16 J. Pesic, V. Damljanovic, R. Gajic, K. Hingerl, and M. Belic, Density functional theory study of phonons in graphene doped with Li, Ca and Ba, Europhys. Lett. 112, 67006 (2016). 12, 115008-6  ...

... The item referenced in this repository content can be found by following the link on the descriptive page.  ...

... The item referenced in this repository content can be found by following the link on the descriptive page.  ...

... bioRxiv preprint doi: https://doi.org/10.1101/2022.01.27.478095; this version posted January 28, 2022. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Greenwood, et al. 1 Article type: Research article 2 Short title: OPS impacts flg22-induced responses 3 Corresponding author: Carina Collins, Marian University, Department of Biology, 3200 Cold 4 Spring Road, Indianapolis, IN 46222, USA. 5 Email: ccollins@marian.edu 6 7 The phloem-resident OCTOPUS protein is a 8 novel regulator of flg22-induced responses in Arabidopsis thaliana 9 10 11 Kaitlyn N. Greenwood a,1#, Courtney L. King a,2#, Isabella Melena a,3#, Katherine A. Stegemann b, 12 Carina A. Collins a,b,4,5 13 14 a Department of Chemistry and Physics, Drury University, Springfield, MO 65802 15 b Department of Biology, Marian University, Indianapolis, IN 46222 16 17 ORCIDS: 0000-0003-1081-3518 (K.N.G); 0000-0003-0368-4324 (C.L.K.); 0000-0001-7894- 18 0682 (I.M.); 0000-0002-5584-5890 (K.A.S.); 0000-0001-9150-1924 (C.A.C) 19 20 bioRxiv preprint doi: https://doi.org/10.1101/2022.01.27.478095; this version posted January 28, 2022. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Greenwood, et al. 21 Footnotes 22 # 23 Author contributions: CAC supervised the experiments; KNG, CLK, IM, and KAS performed 24 experiments with assistance of CAC; CAC, KNG, CLK, IM, and KAS designed experiments and 25 analyzed the data; CAC, KNG, CLK, IM, and KAS wrote the article; CAC agrees to serve as the 26 author responsible for contact and ensures communication. These authors contributed equally to this work. 27 28 Funding information: This research was funded by start-up funds from Drury University (CAC) 29 and Marian University (CAC), Drury University Research Experience in the Natural Sciences 30 Undergraduate Summer Fellowship (KNG and CLK), and American Society of Plant Biologists 31 Summer Undergraduate Research Fellowship (KAS). 32 33 Present addresses: 34 KNG: DaVita Dialysis, Overland Park, KS 66210, US 35 CLK: University of Notre Dame, Department of Chemistry and Biochemistry, South Bend, IN 36 46556, USA 37 IM: Washington University in St. Louis, School of Medicine, St. Louis, MO 63130, USA 38 CAC: Marian University, Department of Biology, Indianapolis, IN 46222, USA. 39 40 41 42 Email address of Author for Contact: Carina Collins, ccollins@marian.edu bioRxiv preprint doi: https://doi.org/10.1101/2022.01.27.478095; this version posted January 28, 2022. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Greenwood, et al. 43 Abstract 44 Phloem is a critical tissue that transports photosynthates and extracellular signals in 45 vascular plants. Although a functional phloem is necessary for plant health, it is also an ideal 46 environment for pathogens to access host nutrients to promote pathogenesis. Even though many 47 vascular pathogens induce economically relevant crop damage, very little is known about the 48 mechanism(s) by which phloem cells detect potential pathogens and signal to minimize damage. 49 Our lab searched existing phosphoproteomic databases, mining for proteins that were 50 phosphorylated in response to the defense-elicitor flagellin, or flg22, AND were expressed in 51 vascular cells, and we identified Octopus (OPS). OPS is polarly associated with the plasma 52 membrane (PM) of sieve element cells and promotes their differentiation from procambial 53 precursor cells by inhibiting the function of BIN2 in brassinosteroid-related signaling. The 54 observation that OPS is differentially phosphorylated in response to flg22 led us to the examine 55 whether OPS may function in flg22-induced signaling using Arabidopsis T-DNA insertion 56 mutants lacking a functional OPS. In wild-type (WT) seedlings, flg22 binds to the PM receptor 57 flagellin sensing 2 (FLS2) to initiate three branches of a signaling cascade that culminates in 58 increased expression of distinct marker genes. Ultimately these signaling pathways lead to the 59 restriction of pathogen growth. Two independent alleles of ops were treated with 100 !M flg22 60 and marker genes from all three branches of FLS2 signaling exhibited higher expression than WT. 61 We also found that in the absence of any flg22, ops mutants displayed increased flg22 signaling 62 responses. Our results indicate that OPS may function as a negative regulator of flg22-induced 63 signaling events and is one of very few phloem-resident proteins with a documented role in flg22 64 signaling. These results indicate that the phloem may be able to sense and response to the threat 65 of bacterial pathogens in a unique way. 66 67 68 Key words: plant immunity, FLS2, BRI1, phloem, gene expression, Arabidopsis bioRxiv preprint doi: https://doi.org/10.1101/2022.01.27.478095; this version posted January 28, 2022. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Greenwood, et al. 69 Introduction 70 The phloem is a critical tissue in vascular plants that transports photosynthetic sugars from 71 source tissues (leaves) to sink tissues (roots, flowers, and fruit); but this carbon-rich nutrient 72 content also makes the phloem an ideal tissue for pathogenic microbes to colonize (1). There are 73 many examples of phloem-limited bacterial pathogens, including the devastating citrus pathogen 74 Candidatus Liberibacter asiaticus (CLas), the causal agent of citrus-greening disease (or 75 Huanglongbing). CLas gains access to the phloem of citrus trees after being deposited by the 76 brown planthopper or related insect vectors (2, 3). The phloem is such a rich source of nutrients 77 that some pathogens that do not colonize the phloem directly will stimulate the aberrant 78 development of phloem cells in other tissues to tap its resources (4). Two distinct cell types 79 comprise the phloem transport system in plants: sieve elements and companion cells. Sieve 80 elements are elongated cells whose primary function is to transport sap contents (photosynthetic 81 sugars, RNA, peptides, and other small organic molecules) from source to sink. Adjacent to sieve 82 elements are the companion cells, which are responsible for loading sap contents into sieve 83 elements (5). 84 Current models of immune signaling in response to a pathogen generally assume that all 85 host cells must individually recognize extracellular pathogen-associated molecular patterns 86 (PAMPs). PAMPs are detected by plasma membrane (PM)- localized pattern recognition receptor 87 (PRR) proteins that, after PAMP binding, can initiate intracellular signaling and activate robust 88 defenses (6, 7). The classic example of a PRR is FLAGELLIN SENSING 2 (FLS2), a member of 89 the leucine-rich repeat receptor-like kinase (LRR-RLK) family of receptors. FLS2 binds to the 90 bacterial motor protein flagellin or to a conserved 22-amino acid peptide, flg22, derived from 91 bacterial flagellin, to initiate a cascade of intracellular signaling events that contribute to the host 92 immune response (8). Interestingly, FLS2 signaling events fall into one of three branches, each of 93 which culminates in expression of marker genes (9, 10). For example, the calcium-dependent 94 branch of FLS2 signaling induces expression of PHOSPHATE-INDUCIBLE 1 (PHI1), the 95 mitogen-activated protein kinase (MAPK) pathway activates FLG22-INDUCED RECEPTOR 96 KINASE (FRK1), and activation of the salicylic acid (SA) pathway causes increased expression of 97 PATHOGENESIS-RELATED1 (PR1) (7, 11, 12). In addition to changes in gene expression, FLS2 98 also initiates the production of extracellular reactive oxygen species (ROS) and the deposition of bioRxiv preprint doi: https://doi.org/10.1101/2022.01.27.478095; this version posted January 28, 2022. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Greenwood, et al. 99 100 callose at the cell wall (7, 11, 12). Combined, these independent signaling events promote the restriction of pathogen growth and promote host immunity (12). 101 Largescale phosphoproteomic screens have been used to detect proteins that are 102 differentially phosphorylated in response to flg22, potentially identifying proteins involved in the 103 regulation of flg22-induced signaling (9, 13, 14). To find potential regulators of flg22-induced 104 signaling in the phloem, we mined existing flg22 phosphoproteomic datasets and identified 105 OCTOPUS (OPS), a protein that is differentially phosphorylated in response to flg22 in 106 Arabidopsis suspension cell cultures (15). This finding was interesting because to the extent that 107 its function is known, previous data indicates that OPS has a role in phloem development, not 108 immune signaling (16-20). Arabidopsis ops mutant plants, which lack OPS, display decreased 109 phloem pattern complexity, and contain undifferentiated sieve element cells, resulting in gaps in 110 phloem strands of the root (16-18, 21). Further investigations indicated that OPS may regulate 111 phloem developmental processes by promoting signaling events after perception of the hormone 112 brassinolide (BL), a member of the brassinosteriod (BR) class (22). BL is an endogenous steroid 113 hormone detected by the BRassinosteroid Insensitive-1 (BRI1) receptor kinase. When BRI1 binds 114 BL, a series of downstream signaling events occurs, leading to cell elongation and growth. OPS 115 interacts with a member of the GLYGOGEN SYNTHASE KINASE3 (GSK3) family, 116 Brassinosteroid Insensitive-2 (BIN2), at the PM (22). The retention of BIN2 at the PM prevents 117 BIN2 from inhibiting transcriptional changes needed to induce cell growth. Using #-glucuronidase 118 (GUS) and GFP to visualize expression of OPS in Arabidopsis, these previous OPS studies found 119 that OPS is present only in the phloem (16). Combined, this evidence indicates that OPS is an ideal 120 candidate for studying flg22-signaling in the phloem. 121 Despite the devastating effects of phloem-dwelling pathogens on crops, it is not yet known 122 in detail how the cells of the phloem detect the presence of bacterial pathogens or how they respond 123 to them (1). Identifying the contributions of the phloem to pathogen detection and response will 124 increase our knowledge of the tissue-specific mechanisms immune signaling. Using loss-of- 125 function T-DNA mutants and gene expression analysis, we identify the sieve element protein 126 OCTOPUS (OPS) as a regulator of flg22-induced signaling events in Arabidopsis seedlings. To 127 our knowledge, this is the first example of a protein localized in sieve elements with such a role in 128 flg22 signaling. 129 bioRxiv preprint doi: https://doi.org/10.1101/2022.01.27.478095; this version posted January 28, 2022. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Greenwood, et al. 130 Results 131 Root length is reduced in ops-3 and ops-4 seedlings 132 To explore a potential role for OPS in flg22-elicited immune signaling, we obtained 133 previously described T-DNA insertion mutant lines ops-3 and ops-4 (20). Using qPCR, we 134 confirmed that both lines exhibit dramatically reduced OPS transcript levels (Fig. 1A-B). 135 Consistent with previous studies (16), seedlings from both ops T-DNA lines displayed shorter 136 roots with an increased amount of branching (Fig. 1 C-D). Despite the root growth defects, no 137 gross morphological phenotypes, particularly in aerial tissue, were observed in either mutant as 138 compared to Col-0. 139 140 flg22-induced responses are enhanced in ops-3 and ops-4 mutants 141 We hypothesized that because OPS was identified as a protein that was differentially 142 phosphorylated after flg22 treatment in a phosphoproteomic screen (15), OPS might have a role in 143 the regulation of flg22-induced signaling events in Arabidopsis. We first sought to determine how 144 plants lacking OPS would respond to flg22 using a seedling growth inhibition assay. When Col-0 145 plants are grown in the presence of flg22 for an extended period (2 weeks), the seedlings 146 experience growth stunting as a result of flg22 detection (8, 23). Col-0 and both independent ops- 147 3 and ops-4 mutant alleles were grown in liquid media supplemented with either 0 or 1 !M flg22 148 for two weeks before measuring their fresh weight. Growth of the Col-0 seedlings was inhibited 149 by 74% (Fig. 2). Interestingly, both ops mutants showed significantly increased percent growth 150 inhibition; 80% and 82%, respectively, when compared to Col-0 (Fig. 2). Importantly, the seedling 151 growth inhibition was similar between ops-3 and ops-4, indicating that this phenotype is indeed 152 the result of loss of OPS. 153 Because we observed an increase of flg22-induced growth inhibition in ops mutants, which 154 may be indicative of a defect in flg22 perceptions, we sought to investigate whether the lack of 155 OPS led to a more specific flg22 signaling defect. In Arabidopsis when FLS2 binds flg22, a 156 network of multiple signaling pathway branches are initiated, and each independent branch results 157 in the expression of a pathway-specific marker gene (9, 10). To test if OPS has a role in one or 158 more of these specific pathways, we measured induction of the flg22-induced and Ca2+ pathway- 159 dependent expression of the marker gene PHI1. In Col-0 seedlings treated with 100 nM flg22, 160 PHI1 expression remains low prior to flg22 exposure (0 minutes), peaks within 30 minutes, and bioRxiv preprint doi: https://doi.org/10.1101/2022.01.27.478095; this version posted January 28, 2022. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Greenwood, et al. 161 then falls to near base levels for 1-3 hours (Fig. 3A). When ops-3 and ops-4 seedlings were treated 162 with 100 nM flg22 for the same time course, the expression pattern of PHI1 was observably 163 different. Even without the addition of 100 nM flg22, both ops-3 and ops-4 mutants showed 164 significantly higher PHI1 expression than Col-0 and while PHI1 peaked in both mutants 30 165 minutes after treatment, expression levels were significantly higher than Col-0 (Fig. 3A). After 166 observing that even in the absence of flg22, ops mutants had increased PHI1 expression, we 167 decided to examine subsequent marker genes in the absence of flg22. 168 MAPK activation after flg22 elicitation occurs independently of Ca2+ induction and leads 169 to expression of WRKY33 and FRK1, so we next measured expression of the marker gene FRK1. 170 Similarly to PHI1, FRK1 expression was significantly higher in both ops-3 and ops-4 seedlings, 171 even in the absence of flg22 (Fig. 3B-C). The final pathway we examined was the SA-dependent 172 pathway of FLS2 signaling by measuring expression of the marker gene PR1. Consistent with our 173 previous qPCR results, PR1 expression was significantly increased in both ops-3 and ops-4 174 mutants. Together, these data suggest that OPS may be a negative regulator of flg22-induced 175 immune responses. 176 177 Brassinosteroid-induced hypocotyl growth is impaired in ops-3 and ops-4 178 Previous investigations of OPS function showed that OPS interacts with the kinase 179 BRASSINOSTEROID-INSENSITIVE 2 (BIN2) at the PM, preventing BIN2 from inhibiting 180 transcriptional changes needed to induce cell growth (22). These data indicate that OPS plays a 181 role in promoting BL-induced cellular responses which may explain some of the phloem 182 developmental defects observed in ops mutants (16, 24, 25); however, these observations were 183 made by studying Arabidopsis plants expressing OPS under the control of constitutive promoter 184 which may not accurately reflect in planta conditions. 185 To gain further evidence of a role for OPS in BL signaling, we treated Col-0 and ops 186 mutants with 1 !M epibrassinolide and measured hypocotyl elongation as a part of a standard BL- 187 response assay. Col-0 seedlings exhibit elongated hypocotyls when exposed to BL, whereas 188 elongation in both ops-3 and op-4 mutants was comparatively reduced. (Fig. 4A). When 189 Arabidopsis seedlings are grown in the dark, their hypocotyls elongate to produce a BL-dependent 190 etiolated phenotype (26). To gain further insights into the role of OPS in BR-mediated signaling, 191 we measured the hypocotyls of Col-0, ops-3 and ops-4 seedlings grown in the light and the dark. bioRxiv preprint doi: https://doi.org/10.1101/2022.01.27.478095; this version posted January 28, 2022. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Greenwood, et al. 192 Col-0 hypocotyls elongate significantly in the dark, and while the length of hypocotyls of both 193 ops-3 and ops-4 seedlings do elongate, they are significantly shorter than those of Col-0 (Fig. 4B). 194 These results provide additional evidence that OPS is required for BR-induced hypocotyl 195 elongation. 196 197 198 199 Discussion 200 Phloem-limited pathogens cause devastating crop losses, but our understanding of how 201 vasculature tissue responds to pathogen invasion is incomplete, making it imperative to study the 202 contributions of tissue-specific immune signaling to overall plant defenses. A new study 203 demonstrated for the first time that when Citrus sinesis or Valencia trees are infected with the 204 phloem-limited pathogen CLas they exhibit increased ROS production and callose deposition, 205 indicative of a PAMP-triggered immune response (27). Moreover, authors showed that both sieve 206 elements and companion cells underwent programmed cell death in response to prolonged CLas 207 infection (27). Here, we have identified a protein exclusively expressed in sieve elements of the 208 phloem that also has a role in flg22-induced immune responses, which we believe may be the first 209 of its kind. These recent advances highlight the likelihood that phloem cells can detect the presence 210 of bacterial pathogens and are active in signaling a response. 211 In this work, we demonstrate that two independent Arabidopsis lines harboring T-DNA 212 insertions in the OPS gene lack OPS expression and display constitutive and increased flg22- 213 induced expression of several immunity-related marker genes. Furthermore, we found that 214 expression of these genes can be detected even in the absence of flg22. This combination of results 215 indicates that OPS functions as a negative regulator of these flg22-induced responses and therefore 216 OPS exhibits a suppressive effect on this pathway (Fig. 5). While the specific mechanism of OPS 217 function in flg22 signaling remains unclear, OPS may suppress activation of PHI1, FRK1, 218 WRKY33, and PR1 expression by directly inhibiting a signaling event in the flg22 pathway. One 219 possibility is OPS may be directly targeting another member of the GSK3 protein family that is 220 instead known to regulate flg22-induced signaling, such as Arabidopsis Protein Kinase $ (ASK $) 221 (28). Because genes from three independent branches of FLS2-flg22 signaling show the same 222 increased expression in the ops mutant backgrounds, it is likely that if OPS is acting directly to bioRxiv preprint doi: https://doi.org/10.1101/2022.01.27.478095; this version posted January 28, 2022. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Greenwood, et al. 223 regulate these signals, it must function early in flg22 signaling. A second possibility remains that 224 OPS has an indirect role in flg22 signaling. OPS is expressed in the sieve elements of the phloem 225 in Arabidopsis and previous studies identified that ops mutants display in incomplete sieve element 226 differentiation which results in a discontinuous protophloem and metaphloem cell file in the roots 227 (16, 17, 24). Therefore, our observation that ops mutants exhibit increased flg22 marker gene 228 expression could be the result of incomplete phloem transport, resulting in the loss of transport of 229 a yet unknown phloem-mobile regulator of flg22 responses. 230 Our results are consistent with reports that FLS2 expression is detected in the vasculature 231 of Arabidopsis cotyledons in roots (29, 30), however detection of FLS2 could only be confirmed 232 in the stele, which comprises both sieve elements and companion cells. FLS2 expression in the 233 stele suggests that some cells of the vasculature are capable of detecting and responding to flg22, 234 and while it has yet to be determined whether FLS2 expressed in the vasculature induces flg22 235 signaling events, our results indicate that some element of flg22 signaling does occur in the 236 phloem. 237 In addition to our data demonstrating a role for OPS in flg22 signaling, we found that 238 Arabidopsis loss-of-function ops mutants have shorter hypocotyls in response to BL treatments 239 and in the dark. This is consistent with other work showing that OPS overexpression lines display 240 increased BR responses (22), and confirms that OPS promotes BR-induced signaling for hypocotyl 241 elongation. Finding that OPS functions in both the BRI1 and FLS2 signaling pathways is 242 particularly interesting because these signaling pathways intersect downstream of initial receptor- 243 ligand perception. Initiation of the BRI1 signaling pathway induces expression of several WRKY 244 family transcription factors that inhibit the activation of gene expression of some flg22-inudced 245 genes (none that were tested in this work) resulting in a suppression of defense responses (31, 32). 246 That OPS functions as a positive regulator of BL signaling may explain why in ops mutants, we 247 observe increased flg22-induced immune responses. While there is much left to understand about 248 the function of OPS, it remains one of the few phloem-localized proteins identified with a role in 249 immunity-related signaling events. Because it remains unknown whether sieve elements or 250 companion cells can directly test and respond to PAMPs, future studies could use OPS as a model 251 to begin answering these questions. 252 bioRxiv preprint doi: https://doi.org/10.1101/2022.01.27.478095; this version posted January 28, 2022. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Greenwood, et al. 253 Methods 254 Plant Material and Growth Conditions 255 T-DNA insertion lines of Arabidopsis (Arabidopsis thaliana) ops-3 SALK_089722 (20) and ops- 256 4 SALK_042563 (20) were obtained from the Arabidopsis Biological Resource Center at The Ohio 257 State University (https://abrc.osu.edu/). Genotypes were confirmed using PCR (primers listed in 258 Supplemental Table 1). Surface-sterilized seeds were sown on 0.5X Murashige and Skoog 259 medium + 1% (w/v) sucrose solidified with 0.6% (w/v) agar as described. After 2 days 260 stratification at 4C, seedlings were germinated and grown at 22C with an 10-h-light/14-h-dark 261 photoperiod at 82 mmol m22 s21. Unless otherwise noted, seedlings were grown under these 262 conditions for 10 days before sample analysis. 263 264 Peptides and Hormones 265 The peptide flg22 (QRLSTGSRINSAKDDAAGLQIA) was purchased from Genscript and used 266 for elicitation at the indicated concentrations and for the indicated times as described. 24- 267 epibrassinolide (BL) was purchased from MilliporeSigma and used for the indicated times and 268 concentrations described. 269 270 Root and Hypocotyl Measurements 271 Root length measurements were performed as described (33). 10-day-old seedlings grown under 272 the conditions described, were traced using Fiji Free-hand tool. For hypocotyl measurements, 4 273 day-old-seedlings were treated with the indicated concentration of BL in liquid media and placed 274 at 22C for an additional 7 days. Seedlings were removed and hypocotyls were traced using the 275 Fiji Free-hand tool. 276 277 Flg22 Seedling Growth Inhibition 278 Seedling growth inhibition was measured as previously described (23). Briefly, four-day-old 279 seedlings were aseptically transferred from MS agar to wells of a 12-well microtiter plate (three 280 seedlings per well) containing 1 ml of liquid MS medium with or without 1 !M flg22. After 14 281 days, seedling fresh weights were recorded. 282 283 RNA Isolation and RT-qPCR bioRxiv preprint doi: https://doi.org/10.1101/2022.01.27.478095; this version posted January 28, 2022. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Greenwood, et al. 284 RNA isolation, cDNA synthesis, and RT-qPCR were performed as previously described (34). 285 Unless stated otherwise, for each sample, three to five seedlings were elicited with the indicated 286 concentration of flg22 peptide and placed at 22C for the time indicated. Tissue was flash frozen 287 in liquid nitrogen at the indicated time points. Total RNA was isolated from tissue using Trizol 288 reagent (Sigma-Aldrich) according to the manufacturers protocol. RT-qPCR was performed on 289 cDNA using a Rotor-Q Real-Time PCR Cycler from Qiagen using gene-specific primers and 290 normalized to the ACTIN gene (listed in Supplemental Table 1). 291 bioRxiv preprint doi: https://doi.org/10.1101/2022.01.27.478095; this version posted January 28, 2022. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Greenwood, et al. 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 References 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. Y. Jiang, C. X. Zhang, R. Chen, S. Y. He, Challenging battles of plants with phloemfeeding insects and prokaryotic pathogens. Proc Natl Acad Sci U S A 116, 23390-23397 (2019). E. D. Ammar, J. George, K. Sturgeon, L. L. Stelinski, R. G. Shatters, Asian citrus psyllid adults inoculate huanglongbing bacterium more efficiently than nymphs when this bacterium is acquired by early instar nymphs. Sci Rep 10, 18244 (2020). M. Ghanim et al., 'Candidatus Liberibacter asiaticus' Accumulates inside Endoplasmic Reticulum Associated Vacuoles in the Gut Cells of Diaphorina citri. Sci Rep 7, 16945 (2017). P. Walerowski et al., Clubroot Disease Stimulates Early Steps of Phloem Differentiation and Recruits SWEET Sucrose Transporters within Developing Galls. Plant Cell 30, 3058-3073 (2018). R. Breia et al., Plant SWEETs: from sugar transport to plant-pathogen interaction and more unexpected physiological roles. Plant Physiol 186, 836-852 (2021). D. Couto, C. Zipfel, Regulation of pattern recognition receptor signalling in plants. Nat Rev Immunol 16, 537-552 (2016). V. Nicaise, M. Roux, C. Zipfel, Recent advances in PAMP-triggered immunity against bacteria: pattern recognition receptors watch over and raise the alarm. Plant Physiol 150, 1638-1647 (2009). L. Gomez-Gomez, T. Boller, FLS2: an LRR receptor-like kinase involved in the perception of the bacterial elicitor flagellin in Arabidopsis. Mol Cell 5, 1003-1011 (2000). J. M. Smith et al., Loss of Arabidopsis thaliana Dynamin-Related Protein 2B reveals separation of innate immune signaling pathways. PLoS Pathog 10, e1004578 (2014). D. A. Korasick et al., Novel functions of Stomatal Cytokinesis-Defective 1 (SCD1) in innate immune responses against bacteria. J Biol Chem 285, 23342-23350 (2010). J. Monaghan, C. Zipfel, Plant pattern recognition receptor complexes at the plasma membrane. Curr Opin Plant Biol 15, 349-357 (2012). T. A. DeFalco, C. Zipfel, Molecular mechanisms of early plant pattern-triggered immune signaling. Mol Cell 81, 4346 (2021). M. Kalde, T. S. Nuhse, K. Findlay, S. C. Peck, The syntaxin SYP132 contributes to plant resistance against bacteria and secretion of pathogenesis-related protein 1. Proc Natl Acad Sci U S A 104, 11850-11855 (2007). T. S. Nuhse, A. R. Bottrill, A. M. Jones, S. C. Peck, Quantitative phosphoproteomic analysis of plasma membrane proteins reveals regulatory mechanisms of plant innate immune responses. Plant J 51, 931-940 (2007). J. J. Benschop et al., Quantitative phosphoproteomics of early elicitor signaling in Arabidopsis. Mol Cell Proteomics 6, 1198-1214 (2007). E. Truernit, H. Bauby, K. Belcram, J. Barthelemy, J. C. Palauqui, OCTOPUS, a polarly localised membrane-associated protein, regulates phloem differentiation entry in Arabidopsis thaliana. Development 139, 1306-1315 (2012). M. A. Ruiz Sola et al., OCTOPUS-LIKE 2, a novel player in Arabidopsis root and vascular development, reveals a key role for OCTOPUS family genes in root metaphloem sieve tube differentiation. New Phytol 216, 1191-1204 (2017). bioRxiv preprint doi: https://doi.org/10.1101/2022.01.27.478095; this version posted January 28, 2022. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. Greenwood, et al. 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. A. S. Breda, O. Hazak, C. S. Hardtke, Phosphosite charge rather than shootward localization determines OCTOPUS activity in root protophloem. Proc Natl Acad Sci U S A 114, E5721-E5730 (2017). Y. H. Kang, A. Breda, C. S. Hardtke, Brassinosteroid signaling directs formative cell divisions and protophloem differentiation in Arabidopsis root meristems. Development 144, 272-280 (2017). H. Roschzttardtz et al., The VASCULATURE COMPLEXITY AND CONNECTIVITY gene encodes a plant-specific protein required for embryo provasculature development. Plant Physiol 166, 889-902 (2014). B. Gujas et al., A Reservoir of Pluripotent Phloem Cells Safeguards the Linear Developmental Trajectory of Protophloem Sieve Elements. Curr Biol 30, 755-766 e754 (2020). P. Anne et al., OCTOPUS Negatively Regulates BIN2 to Control Phloem Differentiation in Arabidopsis thaliana. Curr Biol 25, 2584-2590 (2015). J. C. Anderson et al., Arabidopsis MAP Kinase Phosphatase 1 (AtMKP1) negatively regulates MPK6-mediated PAMP responses and resistance against bacteria. Plant J 67, 258-268 (2011). A. Rodriguez-Villalon et al., Molecular genetic framework for protophloem formation. Proc Natl Acad Sci U S A 111, 11551-11556 (2014). A. Rodriguez-Villalon, B. Gujas, R. van Wijk, T. Munnik, C. S. Hardtke, Primary root protophloem differentiation requires balanced phosphatidylinositol-4,5-biphosphate levels and systemically affects root branching. Development 142, 1437-1446 (2015). S. Fujioka et al., The Arabidopsis deetiolated2 mutant is blocked early in brassinosteroid biosynthesis. Plant Cell 9, 1951-1962 (1997). W. Ma et al., Citrus Huanglongbing is a pathogen-triggered immune disease that can be mitigated with antioxidants and gibberellin. Nature Communications 13, 529 (2022). H. Stampfl, M. Fritz, S. Dal Santo, C. Jonak, The GSK3/Shaggy-Like Kinase ASKalpha Contributes to Pattern-Triggered Immunity. Plant Physiol 171, 1366-1377 (2016). M. Beck et al., Expression patterns of flagellin sensing 2 map to bacterial entry sites in plant shoots and roots. J Exp Bot 65, 6487-6498 (2014). I. Wyrsch, A. Dominguez-Ferreras, N. Geldner, T. Boller, Tissue-specific FLAGELLINSENSING 2 (FLS2) expression in roots restores immune responses in Arabidopsis fls2 mutants. New Phytol 206, 774-784 (2015). R. Lozano-Duran et al., The transcriptional regulator BZR1 mediates trade-off between plant innate immunity and growth. Elife 2, e00983 (2013). Y. Belkhadir, L. Yang, J. Hetzel, J. L. Dangl, J. Chory, The growth-defense pivot: crisis management in plants mediated by LRR-RK surface receptors. Trends Biochem Sci 39, 447-456 (2014). G. Ekanayake et al., DYNAMIN-RELATED PROTEIN DRP1A functions with DRP2B in plant growth, flg22-immune responses, and endocytosis. Plant Physiol 185, 1986-2002 (2021). C. A. Collins et al., EPSIN1 Modulates the Plasma Membrane Abundance of FLAGELLIN SENSING2 for Effective Immune Responses. Plant Physiol 182, 17621775 (2020). C A ops-4 (SALK_042563) OPS 5 1 bp 3 2058 bp ops-3 ops-3 (SALK_089722) a 50 OPS Expression Relative to Actin b b 20 a 15 D 40 30 5-23-19 Primary Root Growth OPS Col-0 Root Length (cm) B ops-4 10 b b 5 10 0 0 Col-0 ops-3 ops-4 Col-0 ops-3 ops-4 4 3 0 4 3 ol -0 Figure 1. Characterization of two independent OPS T-DNA insertion lines op s- op s- ol C op s- op s- C A. Location of T-DNA insertions in OPS and mutants used in this study. 20 B. Expression of OPS mRNA in Col0 and ops mutants measured using qPCR. OPS expression is shown relative to Actin, n=6. C. Comparison of root length between Col-0 WT, ops-3 and ops-4 mutants from vertically grown, day 7 old seedlings. Shown is a representative image. D. Analysis of root length in Col-0 WT, ops-3, and ops-4 mutants. Measurements were made from images (as in C) using ImageJ, n = 20. Letters above all bars represent statistical differences as measured by a one-way ANOVA with Tukey multiple comparison test where P < 0.05. All error bars represent SE. A % Growth Inhibition 85 b b ops-3 ops-4 80 a 75 70 Col-0 op s3 op s4 0 Figure 2. Lack of OPS increases flg22induced growth inhibition C ol A. Four-day old seedlings were grown in the presence of 0 or 1 M flg22 for two weeks before measuring their fresh weights and calculating the percent growth inhibition between growth conditions. Shown are percent change calculations pooled from 4 independent experiments, n = 40-60 seedlings per genotype per experiment. ANOVA with Tukey multiple comparison test with a P <0.05. All error bars are SE A Expression Relative to ACTIN 80 PHI1 *** Col0 ** 60 ops-3 *** ops-4 *** 40 20 0 0 0.5 1 3 0 Col-0 B 1 3 0 0.5 C D FRK1 5 Expression Relative to C ACTIN a a 4 60 PR1 a a 2 b 20 b 10 5 c 1 b 0 15 3 a 40 Hours 3 ops-4 ops-3 WRKY33 1 ol C 00 ol -0 min C 30 ol m i C 06 n ol 0 m -0 18 in op 0 m sin op 3 0 m sin 3 3 op 0 m sin 3 Expression Relativeoto ACTIN 6 ps 0 m -3 18 in op 0 m sin op 4 0 m sin 4 3 op 0 m sop 4 6 in s- 0 m 4 18 in 0 m in Expression Relative to ACTIN 80 0.5 0 0 min 0 0 min 0 min 3 4 op s- op s- ol C 3 4 op s- op s- C ol 0 4 op s- 3 op s- C ol -0 Figure 3. OPS functions as a negative regulator of flg22-induced immune signaling A. Ten-day old old seedlings (Col-0 WT, ops-3, and ops-4) were treated with 100 nM flg22 and qPCR was used to measure expression of marker gene PHI1 over a 3-hour time-course. Shown are means of pooled seedling samples (n = 4). Asterisks indicate significant differences compared with Col-0 0 min (lines) and Col-0 30 min (brackets) using a two-tailed Students t-test ** P > 0.01, *** P > 0.0001. B-D. qPCR analysis of untreated, ten-day old seedlings (Col-0 WT, ops-3, and ops-4). Expression of the marker genes WRKY33 (B), FRK1 (C), and PR1 (D) was examined. Letters represent statistical differences as measured by one-way ANOVA with Tukey multiple comparison at P <0.01. All error bars are SE. Experiments were repeated at least three times with similar results. Hypocotyl Length (cm) 1.5 b 1.0 0.5 0.0 B 1.5 BR-induced a b Hypocotyl Length (cm) A b a b 1.0 a 0.5 c c c + Col-0 - + 1 M ops-4 epiBL + ops-3 Dark-induced 0.0 c c + Col-0 + ops-3 c + ops-4 Dark Figure 4. OPS is necessary for hypocotyl elongation ol C -0 L ol op 0 D s op -3 L sop 3 D s op -4 L s4 D C C ol 0 C ol op 0 + sop 3 -3 op + -4 op -4 + A. Hypocotyl length measurements taken of ten-day old seedlings (Col-0, ops-3, and ops-4) treated with 0 (-) or 1 (+) M epibrassinolide. Values shown are pooled from three independent expriments, n = 60-90. B. Length of hypocotyls of ten-day old seedlings grown in the light (-) or the dark (+) for Col-0 WT, ops-3, and ops-4. Measurements are pooled from three independent experiments, n = 30-60. Letters represent statistical differences as measured by two-way ANOVA with Tukey multiple comparison test at P <0.05. Error bars are SE and all experiments were repeated at least three times with similar results. ops Col-0 FLS2 FLS2 X OPS OPS Ca2+dependent PHI1 MAPKdependent WRKY33 FRK1 Ca2+dependent SAdependent PHI1 PR1 MAPKdependent SAdependent PR1 WRKY33 FRK1 Figure 5. Summary of OPS flg22 signaling phenotype Loss of the protein OPS (ops; right panel) results in increased expression of marker genes from three independent branches of the FLS2 signaling network (red arrow). This indicates that OPS plays a role in suppressing flg22-induced signaling (left panel).  ...

... The item referenced in this repository content can be found by following the link on the descriptive page.  ...

... International Journal of Financial Studies Article Stock Market Contagion during the Global Financial Crises: Evidence from the Chilean Stock Market Sakthi Mahenthiran 1, * , Tom Gjerde 2 and Berta Silva 3 1 2 3 * Lacy School of Business, Butler University, Indianapolis, IN 46208, USA Clark H. Byrum School of Business, Marian University, Indianapolis, IN 46222, USA; tgjerde@marian.edu Escuela de Comercio, Pontificia Universidad Catlica de Valparaso, Valparaso 2340031, Chile; berta.silva@pucv.cl Correspondence: smahenth@butler.edu; Tel.: +1-317-940-8024 Received: 10 March 2020; Accepted: 14 April 2020; Published: 20 April 2020



 Abstract: The study examines evidence for the transmission of the US and EU financial crises via investor holdings into the Chilean stock market following two global financial crises, in 2008 and 2011. The study modified the models of Bekaert et al. (2014), and Dungey and Gajurel (2015) on the 20072009 global financial crisis and extends the period to include the European debt crisis of 20102011. The study produced three main contributions. First, changes in the equity holdings of retail investors were a key source of contagion following the 2008 US financial crisis. Second, investor herding during the 2011 financial crisis is shown to be low based on the co-movement of equity holdings between the four investor groups studied. Third, investor behavior during the 2011 EU crisis differs from that of the 2008 US financial crisis, which we attribute to firms in Chile adopting international financial reporting standards (IFRS) and improving their corporate governance. We compared the findings to the prior contagion studies that rely on Chilean return data to highlight the contributions to international financial research, particularly as it relates to the functioning of emerging capital markets during financial crises. Keywords: stock market contagion; global financial crises; Chilean stock market; investor herding; investor holdings; emerging capital markets; corporate governance JEL Classification: G11; G15; G40 1. Introduction Investor herd behavior is defined as an intention of investors to mimic the behavior of other investors. This is a psychological behavior of market participants, and finance researchers have detected the presence of herding by looking at the relationship between individual firm stock returns and the average market returns. It is argued that investors ignore the fundamental analysis to explain stock prices movements and instead base their decisions on aggregate market behavior, which has been found to be the case during the periods of large market movements (Chang et al. 2000; Bui et al. 2018). This may be the case with the coronavirus-related financial shock that is rattling the global financial markets in 2020. Hence, to study the 2020 global financial crises, it would be important to gain a better understanding of the reasons for the 2008 and 2011 crises. Bui et al. (2018) found that investor herd behavior is driven by both up and down overall market returns and showed that the variations in US stock market returns (not the Hong Kong stock market) were responsible for herd behavior in the frontier market of the Vietnamese stock exchange. Bekaert et al. (2014) and Dungey and Gajurel (2015) addressed the shortcoming in the literature on the co-movement of equity market returns surrounding the 2008 US mortgage crisis. The term spillover effect refers to the historical or expected cross-border Int. J. Financial Stud. 2020, 8, 26; doi:10.3390/ijfs8020026 www.mdpi.com/journal/ijfs Int. J. Financial Stud. 2020, 8, 26 2 of 22 co-movement in asset prices, while the term contagion is reserved for unexpected or excessive spillover and is typically associated with negative shocks to the global financial system, for example, the 2020 global shock related to the coronavirus pandemic. The tendency of investors to sell in unison during a global financial market panic is termed investor herding and provides a potential explanation for a portion of the observed global co-movement in asset prices. Negative aspects of this type of contagion can be severe and include a reduction in the benefits from portfolio diversification, adverse effects on wealth and economic growth, and greater risk management issues for investors and policy makers (Connolly and Wang 2000; Kyle and Xiong 2001). This study differs from previous contagion studies by examining changes in investors equity holdings instead of equity returns. The objective is to determine if the rebalancing of equity portfolios by insiders, institutional investors, and retail investors is useful for studying the transmission of spillover, contagion, and investor herding during a global financial crisis (referred to as a GFC) into an emerging capital market. Although the methodology can be applied to any country, the current study examines the Chilean equity market during the period 20072013, a period that includes two GFCsthe US mortgage crisis and the EU debt crisis. Studies by Bekaert et al. (2014) and Dungey and Gajurel (2015) spanned the period 20072009 and did not include an analysis of the EU debt crisis. A synthesis of those two studies suggests that the Chilean equity market response to the US mortgage crisis was driven by excess movement in Chilean equity returns relative to expected returns based on company fundamentals. This type of contagion is referred to as domestic contagion by Bekaert et al. (2014) and as idiosyncratic contagion by Dungey and Gajurel (2015). The additional feature of the 20072013 study period is that it includes the adoption of international financial reporting standards (IFRS) by Chilean firms in 2009, which required the voluntary adoption of fair value accounting standards during the period 20092013. Firms in the study began voluntarily adopting fair value standards in 2009, introducing the possibility that changes in the reporting environment may explain changes in equity holdings, given that the US mortgage crisis occurred pre-adoption and the EU debt crisis occurred post-adoption. In emerging markets, insiders are long-term investors with superior information on the value of the firm relative to outside investors. Hence, if a GFC causes firm value to deviate from intrinsic value, then insiders might see it as an opportunity to increase their equity investment positions during a period in which outsiders are reducing their equity holdings. On the other hand, since insiders already control the majority of the equity shares, they may not feel the urge to interfere and catch a falling knife when their own company share prices are tumbling. Therefore, a critical tradeoff that arises during a GFC is the degree to which informed insiders are willing to ignore a temporary increase in risk and uncertainty and step-in to purchase shares they view as temporarily undervalued in an emerging market. A reason this tradeoff is crucial is that without insider buying, emerging markets with highly concentrated ownership may become illiquid during episodes of panic selling during a GFC. However, if informed insiders view these departures from intrinsic value as a buying opportunity, then they serve an invaluable role as liquidity providers to the market during periods of contagion. Additionally, this role of insiders would be more important in an emerging capital market than in a developed capital market because of the limited flexibility of the central bank and security regulators to avert the effects of a GFC. The US is the top foreign source of investment in Chilean equity and investment fund shares, by value of holdings. Following the US, the EU countries represent seven of the next nine sources of foreign investment in Chilean equity and investment fund shares1 . Therefore, international investors in Chilean securities introduce a mechanism through which foreign shocks can enter Chiles financial markets. Aguiar and Gopinath (2005) suggest that financial panics create illiquid markets when foreign 1 From Coordinated Portfolio Investment Survey 2013, IMF Table 13: Portfolio Investment Liabilities: Top Ten Economies by Size of Liabilities/ equity and investment fund shares, for Chile. http://data.imf.org/regular.aspx?key=32986. Int. J. Financial Stud. 2020, 8, 26 3 of 22 sellers flood the market with sell orders at fire sale prices and flee less developed emerging markets for safe havens in developed countries. Hence, an adverse economic shock in the US or EU may force international retail investors and mutual fund managers to sell their Chilean equity holdings to meet margin calls and redemption requests. In this manner, changes in the Chilean equity holdings of international investors could play an important role in the transmission of a foreign-sourced financial crisis into the Chilean stock market. A source of significant local institutional investment in Chile is its system of private pension funds. Chilean law limits the level of foreign securities that pension funds can hold. Thus, the exposure of local pension funds to foreign equity markets is limited, thereby reducing the ability of pension funds to serve as a transmission mechanism for foreign-sourced crises and contagions. According to Gillian and Starks (2003), the typical retail investor is less sophisticated than institutional fund managers are, even though retail investors may be rational in the short-term during a GFC. Furthermore, retail investors include both domestic and foreign investors, and may be more prone to panic selling and herding behavior during a GFC relative to local institutional investors. Therefore, we explore the possibility that retail investors reduce their holdings of Chilean companies during a GFC, which exacerbate any downward pressure on the value of equity shares traded on the Chilean stock exchange, now referred to as the Santiago stock exchange. Given the potential for insiders, institutional investors, and retail investors to magnify or mitigate spillovers and contagion, we explore the relationship between the two recent past GFCs, contagion effects, and creditor monitoring as a corporate governance mechanism on the equity holdings during the 20072013 study period. Our study departs from the existing literature on herding behavior and contagion in a number of ways. First, we examine the relationship between equity market holdings of investors instead of equity market returns. Our focus on equity holdings, in contrast to equity returns, introduces the possibility of identifying the specific groups of investors who serve as a mechanism for the transmission of foreign-sourced financial shocks into an emerging capital market. Second, we introduce EU equity returns as an additional source of spillover and contagion allowing us to study two crises, the US mortgage crisis and the EU debt crisis. Interestingly, the adoption of IFRS by Chilean companies occurred between the two GFCs and provides an opportunity to consider whether differing responses of the investor groups to the GFCs are attributable to IFRS adoption. Third, the use of investor holdings rather than return models introduces a novel method of studying contagion and herding among investor groups that can be generalized to studies of other countries and future global financial crises. In the current study, we view retail investors as noise traders, pension funds and insiders as long-term value-based traders, and mutual funds as convergence traders, who may or may not be able to distinguish liquidity shocks from shocks to company fundamentals. Hence, we posit that in emerging capital markets, convergence traders exaggerate contagion effects if long-term investors, such as insiders or pension funds, fail to provide the same level of liquidity that they normally provide during a non-crisis period. Further, the link between equity market returns and aggregate household wealth creates a means for contagion to have negative and persistent effects on the real economy due to the behavior of retail investors. For example, Yuan (2005) found that wealth effects associated with investors can persist even if only a small fraction of investors experience borrowing constraints. The studys findings support the idea that retail investors play a key role in the spread of contagion into the Chilean stock market. We measure changes in equity holdings across investor groups to show that herding behavior is not an important contributor to the lagged effects of the contagion. The result demonstrates the potential for our holdings-based methodology that can complement the traditional return-based methodology of Bekaert et al. (2014), Dungey and Gajurel (2015) and Bui et al. (2018). The study results highlight the importance of effective stock market regulation, banking regulation, and creditor monitoring for mitigating the effects of global financial crises. The remainder of the study is organized as follows. The second section provides an outline of the Chilean institutional context exploring investor behaviors, the third section reviews the literature, Int. J. Financial Stud. 2020, 8, 26 4 of 22 and the fourth section describes the methodology leading to the development of five hypotheses, and describes the sample, variable measurements, and the models tested. Section 5 presents the results, and Section 6 discusses investor behaviors and concludes the study by highlighting its limitations and future research directions. Int. J. Financial Stud. 2020, 8, x FOR PEER REVIEW 4 of 22 2. Chilean Institutional Context 2. Chilean Institutional Context Chiles private pension system is characterized by six pension funds, referred to as Chiles private pension system is characterized by six pension funds, referred to as Adminstradoras de Fondos de Pensiones, or AFP. They have restrictions on their investments, Adminstradoras de Fondos de Pensiones, or AFP. They have restrictions on their investments, including the size of their investment holdings in foreign securities22. They manage the pension including the size of their investment holdings in foreign securities . They manage the pension accounts of the citizenry who invest in them through payroll deduction made by employers. According accounts of the citizenry who invest in them through payroll deduction made by employers. to Morales et al. (2013, p. 181), the AFP system had a positive effect in relation to, (i) the emergence of According to Morales et al. (2013, p. 181), the AFP system had a positive effect in relation to, (i) the reform in the legal system and the improvement of the oversight under which firms operate, which emergence of reform in the legal system and the improvement of the oversight under which firms have influenced the quality of the external mechanisms regulating corporate governance, (ii) the operate, which have influenced the quality of the external mechanisms regulating corporate emergence of greater capital market liquidity and trading volumes, and (iii) the professionalization of governance, (ii) the emergence of greater capital market liquidity and trading volumes, and (iii) the financial intermediaries. For these reasons, the Chilean stock market is considered a highly integrated professionalization of financial intermediaries. For these reasons, the Chilean stock market is Latin American market compared to most emerging country markets, such as the Vietnamese stock considered a highly integrated Latin American market compared to most emerging country markets, market studied by Bui et al. (2018). This fact makes the Chilean stock market an ideal emerging capital such as the Vietnamese stock market studied by Bui et al. (2018). This fact makes the Chilean stock market to study the effects of GFC on investor holdings. market an ideal emerging capital market to study the effects of GFC on investor holdings. Figure 1 depicts Chilean, US, and EU annual GDP growth, and highlights the cross-country Figure 1 depicts Chilean, US, and EU annual GDP growth, and highlights the cross-country variation of GDP during the GFC. Figure 2 displays the change in selected components of Chilean GDP. variation of GDP during the GFC. Figure 2 displays the change in selected components of Chilean It is noteworthy that the growth rate of Chilean total private consumption and GDP were impacted GDP. It is noteworthy that the growth rate of Chilean total private consumption and GDP were less by the GFCs relative to the substantial and negative growth in fixed capital formation and total impacted less by the GFCs relative to the substantial and negative growth in fixed capital formation domestic spending (i.e., GDP less imports). Hence, the impact of the GFC on Chiles economy were and total domestic spending (i.e., GDP less imports). Hence, the impact of the GFC on Chiles varied, yet substantial, on certain components of the GDP. economy were varied, yet substantial, on certain components of the GDP. 8% 6% 4% 2% 0% -2% -4% -6% 2007 2008 EU 2009 US 2010 2011 2012 2013 Annual % change Chile GDP per capita PPP (US$) Figure 1. Annual Annual percentage change in Chile, EU, and US GDP. GDP. 2 Foreign equity holdings was about 33% of total AFPs equity securities during 20092012. See https://www.spensiones.cl/ apps/boletinEstadistico/. A sixth AFP was created only in 2015. Int. J. Financial Stud. 2020, 8, 26 Int. J. Financial Stud. 2020, 8, x FOR PEER REVIEW 30% 25% 20% 15% 10% 5% 0% -5% -10% -15% -20% 2007 2008 2009 5 of 22 5 of 22 2010 2011 2012 2013 GDP Private Consumption Gross Fixed Capital Formation Exports of Goods and Services GDP Less Imports of Goods and Services Figure 2. Annual percentage change in GDP by selected expenditures. Figure 2. Annual percentage change in GDP by selected expenditures. Figure 3 below charts the annual returns for the Chilean, US, and EU equity markets across the Figure 3 below charts the annual returns for the Chilean, US, and EU equity markets across the 20072013 study period. Returns declined in all three equity markets during the 2008 US mortgage 20072013 study period. Returns declined in all three equity markets during the 2008 US mortgage crisis but equity market returns were more varied surrounding the EU debt crisis of 2011. Note that crisis but equity market returns were more varied surrounding the EU debt crisis of 2011. Note that the decline in the Chilean equity market return associated with the EU debt crisis was large relative to the decline in the Chilean equity market return associated with the EU debt crisis was large relative the US crisis. Figure 4 displays household financial net worth, and shows that changes in the value of to the US crisis. Figure 4 displays household financial net worth, and shows that changes in the value equity shares are closely related to changes in the value of household net worth. Chilean households of equity shares are closely related to changes in the value of household net worth. Chilean experienced a twenty-percent decline in net worth during the 2008 crisis, which was comparable to households experienced a twenty-percent decline in net worth during the 2008 crisis, which was the decline in the net worth of US households. But the effect on Chilean retail investors would be comparable to the decline in the net worth of US households. But the effect on Chilean retail investors substantial. Hence, if changes in household net worth have real effects on spending and investment, would be substantial. Hence, if changes in household net worth have real effects on spending and and given that changes in household net worth are associated with market returns, then it is important investment, and given that changes in household net worth are associated with market returns, then to improve our understanding of the various mechanisms and policies that magnify and mitigate it is important to improve our understanding of the various mechanisms and policies that magnify contagion from GFCs into the real economy. Figures 3 and 4 show that the recovery of Chilean and mitigate contagion from GFCs into the real economy. Figures 3 and 4 show that the recovery of investors is slower than that of US and EU equity investors. Chilean investors is slower than that of US and EU equity investors. A key feature of Chilean equity ownership displayed in Figure 5 below is the high concentration of shares held by insiders. Insiders are the largest shareholder group in Chile, holding over 50% of outstanding shares. Many Chilean publicly listed companies belong to a business group, and family insiders typically control the holding company (Lefort and Walker 2007). Further, highly concentrated ownership in Chilean companies has been shown to affect the liquidity of shares and their governance (Morales et al. 2013; Gjerde et al. 2013). Int. J. Financial Stud. 2020, 8, 26 Int. J. J. Financial Financial Stud. Stud. 2020, 2020, 8, 8, xx FOR FOR PEER PEER REVIEW REVIEW Int. 6 of 22 of 22 22 66 of 60% 50% 40% 30% 20% 10% 0% -10% -20% -30% -40% -50% 2007 2008 2009 2010 Chile 2011 EU 2012 2013 US Figure 3. 3. Equity Equity market market returns returns for for Chile, Chile, EU, EU, and and US. US. Figure 20% 20% 15% 15% 10% 10% 5% 5% 0% 0% -5% -5% -10% -10% -15% -15% -20% -20% -25% -25% 2008 2008 2009 2009 2010 2010 Chile Chile 2011 2011 United States States United 2012 2012 2013 2013 Euro area area Euro Figure 4. Change in in households households net percentage of of Gross Gross Domestic Domestic Investment Investment for 4. Change net worth worth as as aa percentage for Chile, Chile, Figure 4. EU, and US. EU, and US. A key key feature feature of of Chilean Chilean equity equity ownership ownership displayed displayed in in Figure Figure 55 below below is is the the high high concentration concentration A of shares shares held held by by insiders. insiders. Insiders Insiders are are the the largest largest shareholder shareholder group group in in Chile, Chile, holding holding over over 50% 50% of of of outstanding shares. Many Chilean publicly listed companies belong to a business group, and family outstanding shares. Many Chilean publicly listed companies belong to a business group, and family insiders typically typically control control the the holding holding company company (Lefort (Lefort and and Walker Walker 2007). 2007). Further, Further, highly highly insiders concentrated ownership in Chilean companies has been shown to affect the liquidity of shares and concentrated ownership in Chilean companies has been shown to affect the liquidity of shares and their governance (Morales et al. 2013; Gjerde et al. 2013). their governance (Morales et al. 2013; Gjerde et al. 2013). funds increased their equity holdings by over a third before falling back towards 7% in 2013. AFPs have a steady inflow of funds from payroll deduction associated with Chiles privatized pension system. Given that AFPs are relatively passive portfolio managers, a potential explanation for their behavior shown in Figure 5 is that an increase in redemption orders and asset reallocation driven by two forced AFP managers to sell equity holdings at a rate that exceeded equity Int. J. Financial Stud. GFCs 2020, 8, 26 7 of 22 purchases associated with payroll contributions. 60% 10.0% 50% 8.0% 40% 6.0% 30% 4.0% 20% 2.0% 10% 2007 0.0% 2008 2009 informed(insiders) 2010 retail 2011 2012 pension_funds 2013 mutual_funds Figure Figure5.5.Percentage Percentageof ofoutstanding outstandingshares sharesheld heldby byinsiders, insiders,retail, retail,pension pensionfunds, funds,and andmutual mutualfunds. funds. Holdings Holdingsofofinsiders insidersand andretail retailisismeasured measuredon onthe theleft leftaxis, axis,AFPs AFPs(Adminstradoras (Adminstradorasde deFondos Fondosde de Pensiones) and mutual fund ownership is measured on the right axis. Pensiones) and mutual fund ownership is measured on the right axis. Retail investors up a second shareholder group,to controlling the not held Another featuremake apparent in Figure 5 is that relative the change inoutstanding the holdingsshares of insiders, the by insiders and institutional investors. Inspection of significantly Figure 5 reveals a noticeable decline thetrended equity equity holdings of mutual funds and AFPs varied during the study periodinbut holdings of retail investors 2008 a simultaneous in the equity holdingsterms, of insiders in opposite directions. Theduring changes in and insider holdings areincrease much smaller in percentage but a and the two groups of institutional investors. The decline in the equity holdings of retail investors in small change in insider holdings can have a relatively large effect given that insiders own over half 2008 that selling investors maymarket. be responsible for the Chilean contagion results cited of allsuggests outstanding sharesby onretail the Chilean stock Compared to insiders, mutual fund holdings inare themuch Dungey and Gajurel (2015) and Bekaert et al. (2014) studies. smaller. Nevertheless, mutual fund holdings of Chilean equities may represent only a Institutional investorsdiversified are anotherportfolio, important group shareholders in Chile and consistobjectives. of pension portion of their globally and theyofcan have a variety of investment fundsAlthough and mutual funds. Institutional control one ingoals, seventhe shares. In Figure 5, mutual mutual funds and AFPsinvestors may share investment regulatory burden is much fund and AFP holdings were almost equal at 6% in 2009, the year that most publicly held companies less for mutual funds. The AFPs have limitations on the amount of funds they can invest in foreign adopted IFRS andmutual global financial markets fromand the substantial US mortgage crises. Over the securities, while funds may have began foreignrecovering management exposure to global next four years, AFPs reduced their equity holdings by approximately a third, while mutual funds markets. It follows that changes in mutual fund holdings relative to changes in AFP holdings may be increased their equity holdings by over a third before falling back towards 7% in 2013. AFPs have a steady inflow of funds from payroll deduction associated with Chiles privatized pension system. Given that AFPs are relatively passive portfolio managers, a potential explanation for their behavior shown in Figure 5 is that an increase in redemption orders and asset reallocation driven by two GFCs forced AFP managers to sell equity holdings at a rate that exceeded equity purchases associated with payroll contributions. Another feature apparent in Figure 5 is that relative to the change in the holdings of insiders, the equity holdings of mutual funds and AFPs varied significantly during the study period but trended in opposite directions. The changes in insider holdings are much smaller in percentage terms, but a small change in insider holdings can have a relatively large effect given that insiders own over half of all outstanding shares on the Chilean stock market. Compared to insiders, mutual fund holdings are much smaller. Nevertheless, mutual fund holdings of Chilean equities may represent only a portion of their globally diversified portfolio, and they can have a variety of investment objectives. Although mutual funds and AFPs may share investment goals, the regulatory burden is much less for mutual funds. The AFPs have limitations on the amount of funds they can invest in foreign securities, while mutual funds may have foreign management and substantial exposure to global Int. J. Financial Stud. 2020, 8, 26 8 of 22 markets. It follows that changes in mutual fund holdings relative to changes in AFP holdings may be less sensitive to changes in Chilean economic conditions and more sensitive to changes in global economic conditions. However, changes in the holdings of mutual funds and AFPs in Figure 5 suggest that both AFPs and mutual funds are relatively sensitive to changes in the local market conditions, although their incentives for holding Chilean equity shares may differ. Of note is the sharp increase in shares held by mutual funds in 2008, perhaps due to mutual fund managers viewing Chile as a relatively less risky equity market (compared to other emerging markets) during the 2008 GFC. Thus, mutual fund purchases of Chilean equity shares may have mitigated the transmission of contagion from the 2008 GFC into the Chilean equity market. This understanding helps focus our literature review and hypotheses development. 3. Literature Review Bekaert et al. (2014) used a three-factor asset-pricing model of country-sector equity returns to distinguish between equity-market co-movements due to a US-specific factor, a global financial factor, and a domestic factor. Accordingly, they state (Bekaert et al. 2014, p. 2602) the inclusion of three different factors in our model enables us to distinguish between three types of contagion. Contagion may stem from the US or from the global financial sector, implying a high co-movement of domestic sector portfolios with the US or the global factors. We will label these US contagion and global contagion, respectively. Alternatively, while investors may continue to discriminate across countries in response to global or US-specific shocks during crises, they may discriminate less across stocks within countries in response to idiosyncratic, country-specific shocks, thus giving rise to what we call domestic contagion. Among the authors conclusions was the finding that domestic contagion was the major source of contagion for Chile during the 2008 GFC. Dungey and Gajurel (2015) refer to the global contagion as systematic contagion. Systemic contagion is typically associated with a common global shock experienced simultaneously across markets that is similar to the contagion caused by the 2020 coronavirus-related financial shock. They identify a second type of contagion as idiosyncratic contagion, which occurs when a shock from a crisis originating country spreads to another economy. For example, the US-based mortgage crisis causing price movements in the Latin American stock markets due to fear of global debt defaults. They define a third type of contagion as volatility contagion, which occurs when a shock causes an increase in volatility in one market and then the effects spread to foreign markets. The authors conclude that the 2008 GFC resulted in about one-fifth of the countries studied, including Chile, experiencing the US-sourced idiosyncratic contagion. This result is consistent with Bekaert et al. (2014) given that the definition of domestic contagion in that study is consistent with the definition of idiosyncratic contagion in Dungey and Gajurel (2015). Moreover, idiosyncratic contagion was the main source of contagion for several advanced economies including Japan and France, partly because of these markets greater level of integration with global markets relative to frontier markets like Vietnam, which was studied by Bui et al. (2018). Ma et al. (2018) noted that during crisis periods, the unavailability of liquidity is an important channel through which market volatility affects stock returns in equity markets. These authors suggest that market makers faced with credit constraints and heightened uncertainty pull back from providing liquidity. Consequently, they argue that illiquidity drives firm value further from intrinsic value, which is thought to be a feature of the 2020 coronavirus shock. Further inspection of Figure 5 above reveals that, during our study period, the ownership level of insiders nearly mirrored that of retail investors. Insider holdings rose (fell) when the ownership of retail investors fell (rose). Thus, it is possible that insiders serve as liquidity providers, a role that takes on greater importance during a GFC given the finding in Ma et al. (2018) that market makers withdraw from providing liquidity during a crisis. Given that we use equity holdings instead of equity returns to study contagion, we review the literature on portfolio rebalancing as potential drivers of contagion. Kodres and Pritsker (2002) show that through cross-market portfolio rebalancing, investors transmit idiosyncratic shocks from one market to another when they adjust their portfolios exposures to shared macroeconomic risks. Int. J. Financial Stud. 2020, 8, 26 9 of 22 Their model shows that portfolio rebalancing can generate contagion between markets that do not directly share common economic risks. Moreover, they show that contagion among emerging markets can occur indirectly through a third developed country (like the US) without significantly affecting prices in the developed country. The authors identify correlated information channels and correlated liquidity shock channels as potential mechanisms to transmit the contagion. Particularly, Kodres and Pritsker (2002, p. 770) state that, Under the correlated information channel, price changes in one market are perceived as having implications for the values of assets in other markets that causes their prices to change as well. The correlated liquidity shock channel posits that when some market participants need to liquidate some of their assets to obtain cash, they choose to liquidate assets in a number of markets, effectively transmitting the shock to other markets (Calvo 1999). Additionally, when portfolio rebalancing occurs in markets with information asymmetries, the resulting price co-movements are exaggerated because the order flows are misconstrued as being information-based flows. This led them to conclude that price co-movements are exaggerated in emerging markets that have a significant proportion of uninformed investors. Boyer et al. (2006) provide evidence that stock market crises spread globally through the asset holdings of international retail investors. These studies argue that uninformed but rational investors are unable to distinguish between selling based on liquidity shocks and selling based on fundamental economic shocks in the presence of recursive relationships between a countrys fiscal policies, monetary policies, and market liquidity (see Chowdhury et al. 2018). This may be the case with the 2020 coronavirus shock contagion that is rattling global capital markets. Kyle and Xiong (2001) also consider the relationship between contagion and imperfect information. The authors separated investors into three groups, noise traders trading randomly in only one market, long-term value-based investors trading on fundamentals and providing liquidity, and convergence traders who trade optimally in multiple markets. Convergence traders are perfect competitors, and rational in the sense that their trading strategies correctly take into account the effect of all market participants on the price dynamics in more than one market. Thus, convergence traders, unlike long-term investors, aggressively exploit short-term opportunities by taking the other side of noise trading. On the other hand, long-term investors are not fully rational in the sense that they tend to ignore the short-term opportunities caused by noise traders. Moreover, Kyle and Xiong (2001) argue that it is possible for contagion to result from confused convergent traders if convergence traders cannot distinguish between liquidity shocks and shocks to company fundamentals. Gromb and Vayanos (2010) refer to the risk that stems from noise traders as non-fundamental risk to emphasize the possibility that demand shocks may be unrelated to asset payoffs, which can arise from rational short-term trading behavior by convergence traders. A consequence is that price declines are exaggerated when arbitrageurs, such as convergence traders, are forced to liquidate their position and long-term investors like insiders fail to provide liquidity to the market. Thus, if convergence traders are arbitrageurs who cannot exit their positions in an illiquid market during a crisis, then they can have a destabilizing effect on emerging markets that can lead to a market crash. 4. Methodology The analysis attempts to measure the amount of variation in equity holdings explained by lagged equity holdings, spillover, contagion, and control variables during the 20072013 study period. Total outstanding shares in the Chilean equity market are allocated among four groups of investors according to ownership data obtained from the Superintendcia Valores y Seguros that is referred to as the SVS, which is the equivalent of the United States Securities and Exchange Commission in Chile. The four ownership groups are pension funds, mutual funds, retail investors, and insiders. A goal of modeling equity holdings instead of equity returns is to determine if any of the four investor groups made substantial changes to their equity holdings in response to the two GFCs within the study period. If they did, then we can conclude that portfolio rebalancing may have served as a mechanism to transmit contagion into the domestic economy. Furthermore, with our methodology, Int. J. Financial Stud. 2020, 8, 26 10 of 22 it may be possible to identify which investor groups magnified (or mitigated) contagion effects from the GFCs. The spillover markets proxies are the annual return on three foreign equity indices consisting of an MSCI US equity index (referred to as US Spillover), an MSCI EU equity index (referred to as EU Spillover), and an MSCI global equity index (referred to as Global Spillover)3 . The annual return on the Chilean equity market measures domestic market influence on equity holdings. As a first step, the domestic (Chilean) equity return is orthogonalized by regressing it against the other three returns. Hence, the residuals represent the portion of variation in the domestic index returns that is not explained by variation in the other three external indices. The residuals become the domestic factors within the regression model. Dungey and Gajurel (2015) and Bekaert et al. (2014) employ a similar methodology to measure spillover effects on the equity returns in their studies. Contagion is defined as the additional or unexpected spillover that may occur during a GFC. We applied the methodology employed by Dungey and Gajurel (2015) and Bekaert et al. (2014) to create crisis dummies intended to capture the additional spillover, i.e., contagion, which occurs during a GFC. We identified 2008 as the crisis year associated with the US mortgage crisis, and the year 2011 as the crisis year associated with the EU debt crisis. Each GFC dummy variables equals zero in non-crisis years. In 2008, the GFC dummy Global_crisis_08 equals the 2008 annual return for one of the three equity market indices, and similarly for the 2011 GFC dummy Global_crisis_11. We estimated a separate model for each of the three foreign spillover markets and a fourth model with no spillover market. Each model contains both GFC dummies and the model without a spillover market isolates the impact of variation in the orthogonalized domestic index return on equity holdings. The relationship between equity holdings and several of the modeled control variables are also of interest. For example, we controlled for the impact of long-term debt (Percent_bonds) on equity holdings and interpreted the results in terms of creditor monitoring that is a proxy for external corporate governance. The liquidity control variables include the annual number of trading days per company with no change in return (Zero_return_days), and the effective spread (Effective_spread). Firms operating in the financial sector are identified by a dummy variable (Financial_sector) equal to 1 for financial sector firms and 0 otherwise. Note that Dungey and Gajurel (2015) restrict their analysis to firms in the financial sector based on the observation that the banking sector represent a channel through which contagion may be transmitted across borders. Chilean firms adopted IFRS and fair value accounting standards during the study period, so we included a control variable based on the year of adoption, and we expect the requirements for fair market valuation requirements in IFRS to significantly affect the financial sector. Many firms voluntarily adopted fair value standards early, and we controlled for early adoption through a dummy variable (Fair_Value) that equals 1 upon adoption and each year thereafter, and 0 otherwise. Dungey and Gajurel (2015) and Bekaert et al. (2014) both address the impact of portfolio return volatility in their multi-country studies of contagion. Bekaert et al. (2014) implicitly control for volatility by introducing a volatility ratio of market volatility to factor volatility. Dungey and Gajurel (2015) explicitly model volatility contagion and find that while volatility spillover is common to most countries, roughly 40% of them experience volatility-driven contagion. Furthermore, they note that volatility contagion is rarely the only driver of contagion effects and is associated with policy uncertainty too. Given that the focus of the current study is on a single country, we modeled variation in equity holdings instead of equity returns, and we defined the average standard deviation of daily returns per company as our control for volatility. Portfolio turnover and investor herding are measured by the inclusion of lagged dependent variables on the right-hand side each model. For any given shareholder group, we interpreted a 3 MSCI Inc., is an American Finance Company headquartered in New York City, and it is a global provider of equity, fixed income, and stock market indexes and multi-asset portfolio analysis tools. Its URL is: https://www.msci.com/. Int. J. Financial Stud. 2020, 8, 26 11 of 22 statistically significant parameter estimate on lagged equity holdings as a measure of portfolio turnover. As an extreme example, if equity holdings do not change form one period to another, then this periods equity holdings equal last periods equity holdings, so that the parameter estimate is 1 and the lagged holdings can explain all the variation in current period equity holdings. Similarly, herding is defined for shareholder group i as a statistically significant parameter estimate on the lagged holdings of shareholder group j. In other words, modeled herding occurs when the trades of one investor group are highly correlated with the lagged trades of a second investor group4 . The equity holdings of AFPs, mutual funds, and insiders are modeled in Equation (1) as a function of the two institutional investors and insiders. The three groups of investors are superscripted with i, j, and k in Equation (1) below. Thus, for each investor group i, and given two other investor groups, j and k, the general holdings model is expressed in Equation (1) as: hi = 0 + 1 Lag_hi + 2 Lag_hj + 3 Lag_hk + 4 Effective_spread + 5 Zero_return _days +6 Percentage_bonds + 7 Fair_value + 8 Volatility + 9 Financial _sector +10 Domestic_contagion + 11 Spillover_market + 12 Global_crisis_08 +11 Global_crisis_11 + i . The retail investor model is identical to Equation (1), but excludes lagged pension fund and mutual fund holdings5 . 4.1. Hypotheses Our models separate the equity market response to the two GFCs into changes in the equity holdings of four distinct shareholder groups: insiders, local pension funds or AFPs, mutual funds, and retail investors. In this context, to provide directional hypotheses, we state our hypotheses in terms of the type of investor whose holdings are likely to be a mechanism for transmitting the contagion during a GFC, and those that are not likely to be mechanisms for transmitting the contagion. Hypothesis 1 (H1). Change in the level of local pension fund holdings is not a mechanism for the transmission of GFC effects to the Chilean financial markets. Hypothesis 2 (H2). Change in the level of mutual fund holdings is not a mechanism for the transmission of GFC effects to the Chilean financial markets. Hypothesis 3 (H3). Change in the level of retail investor holdings is a mechanism for the transmission of GFC effects to the Chilean financial markets. Hypothesis 4 (H4). Change in the level of insider holdings is not a mechanism for the transmission of GFC effects to the Chilean financial markets. To test the hypotheses, four models were estimated for insiders, mutual funds, pension funds, and retail investors that complement the studies of Dungey and Gajurel (2015) and Bekaert et al. (2014). However, as shown above, Equation (1) replaces their dependent variable based on equity returns with the equity holdings of insiders, mutual funds, pension funds, and retail investors. An analysis based on equity holdings facilitates an examination of the possibility that institutional managers, retail investors, 4 5 Durbin-h test indicates the presence of autocorrelation in all models except the insider-holding model. Estimation of the models with auto regression (AR1) or ordinary least square (OLS) does not have a qualitatively different impact on the results in Tables 36. Further, all continuous variables are standardized with mean of 0 and standard deviation of 1. We do not include all four investor groups in a single model because retail investor holdings equal the residual of total outstanding shares minus the holdings of the other three investor groups. Thus, the holdings of any single group are a linear combination of the other 3. The current specification provides the best fit to the data, and facilitates a discussion of the relationship observed in Figure 5 between retail investors and insiders. Int. J. Financial Stud. 2020, 8, 26 12 of 22 and insiders provide liquidity when market makers pull back from the provision of providing liquidity during GFCs. In summary, AFP managers are relatively passive investors given their mandate to invest the payroll contributions of Chilean civilians. Mutual funds may have more exposure to global markets and may have viewed Chile as a relatively safe haven during the 2008 and 2011 GFCs. Retail investors are perhaps the least informed group of Chilean investors and this group is most likely to misinterpret the impact of foreign shocks on domestic fundamentals, and may also be prone to herd behavior during panics. Insiders are informed traders with superior information on the value of the firm, and the assumption is that they do not panic during a GFC. As such, insiders should view GFC-driven panic selling as buying opportunities and add liquidity to the emerging market during sell-offs. The tradeoff between risk reduction associated with creditor monitoring and increased risk of default associated with two GFCs during the study period motivates a fifth hypothesis. Hypothesis 5 tests for the possibility that a higher level of publicly issued debt in the capital structure may motivate investors to increase equity holdings to take advantage of credit monitoring. The converse of this hypothesis is that investors held fewer shares in companies with more publicly issued debt to avoid exposure to a heightened probability of default that generally existed during the study period. The tradeoffs are particularly interesting in light of the fact that the Chilean banking model is a traditional deposit-loan model rather than an investment banking/securitization model, which is also the case in most emerging markets. During normal times, debt is a significant source of funds for Chilean firms. Total debt represents just over 45% of our sample firms capital structure, while long-term debt represents nearly 25% of our sample firms capital structure. Publicly issued long-term bonds represent close to 12% of the capital structure. The relatively high level of debt in the capital structure of Chilean companies is explained in part by a high concentration of shares held by insiders who seek to maintain majority control by issuing debt instead of equity shares (Lefort and Walker 2007). In emerging markets, Denis and McConnell (2003) note the importance of the ability of creditors to monitor insiders. Additionally, studies have shown that creditor monitoring is present except when a country has an easy monetary policy (Lopez-de-Foronda et al. 2018). Thus, by including the percentage of bond holdings in our models, we allow for the conflicting effects of it on equity holdings, a heightened probability of default during the study period, potential value of creditor monitoring, and the possibility of flight to quality (Optiz and Szimayer 2018). Moreover, we allow the data to speak on the direction of the correlation of this variable with different equity holdings. Thus, we hypothesize that equity investors recognize a relationship between equity value and the level of debt in the capital structure during a crisis. Hypothesis 5 (H5). The equity holdings of mutual funds, local pension funds, and retail investors are sensitive to the presence of more debt in the capital structure. 4.2. Sample For hypothesis testing, we used price and volume data provided by the regulators of the Santiago Stock Exchange, for the period January 2007 through December 2013. Equity holdings were obtained from the Superintendencia de Valores y Seguros (SVS) that is also referred to as the Commission for the Financial Market. Additionally, we collected annual measures of financial statement items from the Economatica database. Annual equity market returns were obtained from MSCI for Chile, US, EU, and global stock market indices. Annual equity returns were derived using local currencies. Real GDP growth rates and components for Chile, EU, and US were obtained from the Statistics Database Banco Central De Chile. Total portfolio investment was from the International Monetary Fund Coordinated Portfolio Investment Survey (CPISTables 12 and 13). Household net worth was from the Organization for Economic Corporation and Development (OECD) Stat, household dashboard. Further, all firms were required to adopt fair value measurement standards by December Int. J. Financial Stud. 2020, 8, 26 13 of 22 2013, as prescribed by International Accounting Standards Board6 . The final sample includes 63 firms, which we pooled across the period 2007 through 2013, producing a total of 339-pooled firm years after removing observations with missing data. Appendix A lists the variables and their definitions. 5. Results Table 1 provides descriptive statistics that are based on 339 firm years. The high degree of insider ownership in Chile is evidenced by an average insider ownership level of 53.12%. In contrast, the average local pension fund holdings were 5.88%, and mutual funds held 7.8% of the outstanding shares. Retail investors, who control 33.2% of shares, are the second largest shareholders. Table 1. Mean, standard deviation (SD), minimum (min), and maximum (max) values for the sample. Variable Mean SD Min Max Insiders Pension funds Mutual funds Retail Lag_retail Lag_Insiders Lag_Pension funds Lag_Mutual funds Effective_spread Zero_return_days Percent_bonds Volatility Financial_sector Fair_value Domestic_contagion Global spillover US spillover EU spillover 53.12% 5.88% 7.80% 33.20% 34.32% 52.60% 6.20% 6.89% 1.52 11.28 11.96% 2.67 0.31 30.68% 0.00% 3.30% 5.66% 1.03% 21.30% 6.30% 9.88% 23.65% 23.22% 20.70% 6.64% 9.58% 1.72 8.41 16.14% 1.83 0.46 46.18% 8.67% 12.84% 11.88% 16.58% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.00% 0.18 1 0.00% 0.06 0 0.00% 10.40% 16.86% 12.98% 25.39% 99.28% 25.11% 39.00% 97.09% 97.09% 97.80% 25.11% 42.49% 14.41 43 83.52% 15.34 1 100.00% 11.61% 24.06% 27.46% 23.48% Table 2 displays correlations for the study variables. Insider ownership is significant and negatively correlated with both contemporaneous and lagged values of pension fund and mutual fund holdings, at p < 0.01 and p < 0.10 levels. As expected, insiders ownership is not correlated with any of the external spillover proxies or the domestic contagion factor, EU Spillover, Global Spillover, and US Spillover. Further, insiders ownership is positively related to the liquidity proxy effective_spread at the p < 0.05 level. This is consistent with the idea that higher levels of insider ownership suggest less liquidity and wider effective spreads. The creditor monitoring proxy Percent_bonds is positively and significantly correlated with the holdings of Insiders, Pension_funds, Lag_insiders, Lag_Pension_funds, and Lag_mutual_funds at the p < 0.01 level. It is also significant and negatively correlated with holdings of Retail investors at the p < 0.01 level. Fair_value is positively correlated with Pension_funds and Mutual_funds at the p < 0.10 l and p < 0.01 levels, but negative and significantly correlated with Retail, Zero_ret_days, and the Financial_sector dummy at the p < 0.05 level. These results suggest that retail investor behavior is different from the other groups, and the financial sector is less likely to have adopted fair value reporting of assets and liabilities. 6 All but 7 firms adopted IFRS in 2009, so we omitted those 7 firms. Early adopters of IFRS 13 provided levels 13 fair value measurements in their financial statements, thus, they had recognized (not simply disclosed) fair value measurement before the mandated date of December 2013. Int. J. Financial Stud. 2020, 8, 26 14 of 22 Table 2. Pearson correlation matrix. 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. Variable 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 1. Insiders 2. Pension funds 3. Mutual funds 4. Retail 5. Lag_retail 6. Lag_Insiders 7. Lag_Pension funds 8. Lag_Mutual funds 9. Effective_spread 10. Zero_return_days 11. Percent_bonds 12. Volatility 13. Financial_sector 14. Fair_value 15. Domestic_contagion 16. Global spillover 17. US spillover 18. EU spillover 1 0.164 *** 0.102 * 0.814 *** 0.738 *** 0.918 0.141 *** 0.098 * 0.138 ** 0.102 * 0.150 *** 0.005 0.145 *** 0.088 0.102 0.021 0.024 0.021 1 0.448 0.305 0.288 0.146 *** 0.922 *** 0.376 0.268 0.131 ** 0.307 *** 0.155 *** 0.426 0.095 * 0.037 0.055 0.067 0.050 1 0.444 0.395 0.066 0.406 0.820 *** 0.435 0.255 0.281 *** 0.269 0.313 0.196 *** 0.002 0.038 0.044 0.038 1 0.906 *** 0.761 *** 0.288 0.354 0.128 ** 0.233 0.334 *** 0.149 *** 0.113 ** 0.187 *** 0.000 0.018 0.021 0.016 1 0.818 *** 0.297 0.451 0.131 ** 0.250 0.289 *** 0.168 *** 0.089 0.238 0.023 0.060 0.053 0.064 1 0.154 *** 0.071 0.123 ** 0.126 ** 0.145 *** 0.012 0.148 0.111 0.008 0.010 0.022 0.003 1 0.361 0.261 0.133 ** 0.286 *** 0.146 *** 0.448 0.135 ** 0.038 0.040 0.048 0.036 1 0.405 0.240 0.188 *** 0.281 0.225 0.242 0.012 0.198 *** 0.212 0.190 *** 1 0.062 0.250 0.440 0.087 0.264 0.083 0.192 *** 0.214 0.174 *** 1 0.055 0.176 *** 0.168 *** 0.135 ** 0.166 *** 0.043 0.094 * 0.010 1 0.226 0.119 ** 0.105 * 0.037 0.066 0.084 0.053 1 0.019 0.233 0.006 0.138** 0.135** 0.123** 1 0.113 ** 0.015 0.003 0.008 0.001 1 0.112 ** 0.202 *** 0.223 0.188 *** 1 0.00 0.00 0.00 1 0.974 *** 0.983 *** 1 0.924 *** In all the tables, *** Significant at the 1% level, ** significant at the 5% level, and * significant at 10%. Int. J. Financial Stud. 2020, 8, 26 15 of 22 Note that Domestic_contagion in column 15 of Table 2 is uncorrelated with the spillover proxies EU Spillover, Global Spillover, and US Spillover. This is the result of orthogonalizing Domestic_contagion against the spillover proxies. However, the spillover proxies are highly correlated with each other, so we avoided collinearity issues within regressions by running a separate regression for each spillover market in addition to a baseline regression with no spillover markets. The outcome of this estimation strategy is that it allows Domestic_contagion to appear in all regressions to isolate the impact of the orthogonalized domestic market effect after controlling for external market effects. Regression results for the equity holdings of AFPs, mutual funds, retail investors, and insiders are summarized in Tables 36, respectively. Each table contains four models with each model identified by the variable Spillover_market. Model 1 has no spillover market and serves as the baseline model. Spillover_Market equals EU Spillover in Model 2, Global Spillover in Model 3, and US Spillover in Model 4. Recall from our discussion that spillover refers to the expected cross-border co-movement in asset prices, while the term contagion refers to unexpected or excessive spillover. With these definitions in mind, the variables Global_crisis_08 and Global_crisis_11 measure contagion effects associated with the US mortgage crisis in 2008 and the EU debt crisis in 2011. Each of those variables is 0 during non-crisis years and take on the value of the spillover variable in crisis years. The same methodology was employed by Bekaert et al. (2014) and Dungey and Gajurel (2015) to measure contagion effects. Table 3. Regression AR1 coefficient estimates of pension fund holdings on contagion models for the 2008 and 2011 global financial crises (t-stats). Variable Model (1) Model (2) Model (3) Model (4) Intercept 0.019 (0.307) 0.007 (0.298) 0.856 *** (30.808) 0.051 * (1.888) 0.020 (0.831) 0.038 (1.462) 0.036 (1.475) 0.003 (0.121) 0.079 (1.461) 0.043 (0.750) 0.035 (1.351) 0.028 *** (2.613) 0.0036 (0.459) 0.056 (1.149) 0.012 (0.486) 0.851 *** (30.621) 0.048 * (1.794) 0.032 (1.286) 0.045 * (1.709) 0.043 * (1.770) 0.002 (0.070) 0.066 (1.213) 0.040 (0.700) 0.064 ** (2.236) 0.018 *** (3.233) 0.0062 (0.843) 0.103 ** (2.268) 0.035 (0.742) 0.009 (0.377) 0.854 *** (30.737) 0.049 * (1.831) 0.024 (0.992) 0.041 (1.557) 0.043 * (1.738) 0.002 (0.086) 0.073 (1.344) 0.041 (0.729) 0.055 * (1.950) 0.022 *** (2.853) 0.0076 (0.528) 0.075 * (1.760) 0.011 (0.249) 0.008 (0.309) 0.856 *** (30.774) 0.050 * (1.861) 0.020 (0.818) 0.038 (1.465) 0.041 (1.645) 0.001 (0.034) 0.077 (1.430) 0.042 (0.747) 0.048 * (1.707) 0.021 ** (2.483) 0.0039 (0.107) 0.043 (1.174) 0.832 0.832 0.832 0.832 Lag_insiders Lag_pension_funds Lag_mutual_funds Zero_return_days Effective_spread Percent_bonds Volatility Fair_value Financial_sector Domestic_contagion Global_crisis_08 Global_crisis_11 Spillover_market Adj. R2 In all the tables, *** Significant at the 1% level, ** significant at the 5% level, and * significant at 10%. Int. J. Financial Stud. 2020, 8, 26 16 of 22 Table 4. Regression AR1 coefficient estimates of mutual fund holdings on contagion models for the 2008 and 2011 global financial crises (t-stats). Variable Model (1) Model (2) Model (3) Model (4) Intercept 0.010 (0.186) 0.010 (0.287) 0.022 (0.587) 0.682 *** (18.624) 0.110 *** (3.325) 0.152 *** (4.256) 0.102 *** (3.102) 0.010 (0.308) 0.032 (0.436) 0.235 *** (3.073) 0.014 (0.405) 0.060 *** (4.134) 0.006 (0.574) 0.005 (0.074) 0.010 (0.295) 0.022 (0.580) 0.681 *** (18.577) 0.110 *** (3.275) 0.153 *** (4.240) 0.103 *** (3.083) 0.010 (0.297) 0.031 (0.421) 0.235 *** (3.064) 0.016 (0.412) 0.015 ** (2.036) 0.006 (0.542) 0.007 (0.114) 0.041 (0.650) 0.008 (0.249) 0.023 (0.602) 0.683 *** (18.638) 0.107 *** (3.231) 0.151 *** (4.212) 0.098 *** (2.920) 0.014 (0.404) 0.036 (0.492) 0.236 *** (3.086) 0.001 (0.012) 0.014 (1.376) 0.001 (0.032) 0.051 (0.865) 0.051 (0.876) 0.009 (0.274) 0.021 (0.571) 0.682 *** (18.681) 0.110 *** (3.349) 0.153 *** (4.271) 0.094 *** (2.812) 0.017 (0.505) 0.035 (0.478) 0.235 *** (3.087) 0.010 (0.266) 0.014 (1.165) 0.005 (0.105) 0.078 (1.558) 0.696 0.695 0.697 0.698 Lag_insiders Lag_pension_funds Lag_mutual_funds Zero_return_days Effective_spread Percent_bonds Volatility Fair_value Financial_sector Domestic_contagion Global_crisis_08 Global_crisis_11 Spillover_market Adj. R2 In all the tables, *** Significant at the 1% level, ** significant at the 5% level. Hypothesis H1 states that change in the level of pension fund or AFP holdings is not a mechanism for the transmission of GFC effects into the Chilean financial market. The results in Table 3 below contain no evidence to reject H1. The parameters on the contagion variable Global_crisis_08 are negative and significant at p < 0.05 across all spillover markets (i.e., = 0.028 for the baseline Model 1, = 0.018 for the EU Model 2, = 0.022 for the Global Model 3, and = 0.021 for the US Model 4). Additionally, the only spillover market with a p-value below 0.05 is the EU market with a positive relationship to AFP equity holdings. An interpretation is that the equity holdings of AFPs are positively correlated with EU equity market returns during normal times and increase along with increases in the return to EU equity. However, during the 2008 GFC, the equity holdings of AFPs increased as the return in all foreign equity markets declined. Thus, AFPs do not appear to be a source of the transmission for the contagion Chile experienced during the 2008 GFC as identified by Bekaert et al. (2014) and Dungey and Gajurel (2015). Furthermore, the lack of significance between the equity holdings of AFPs and the crisis variable Global_crisis_11 indicates that pension fund holdings were largely unaffected by the 2011 GFC. Therefore, our results support hypothesis H1 that AFPs were not a source of transmission of the 2008 and 2011 crises. Hypothesis H2 states that change in the level of mutual fund holdings is not a mechanism for the transmission of GFC effects. Table 4 below shows that none of the Spillover_market effects are significant. The measure of contagion, Global_crisis_08, is negative and significant for the Chilean market ( = 0.060, p < 0.01 in Model 1) and the EU market ( = 0.015, p < 0.05 in Model 2). Int. J. Financial Stud. 2020, 8, 26 17 of 22 These results suggest that the equity holdings of mutual funds increased as the returns on the Chilean equity market and the EU equity market declined during the 2008 GFC. Thus, mutual funds were not a mechanism for the transmission of contagion during the 2008 GFC. Additionally, the Global_crisis_11 estimates are not significant across all markets. Further, the lack of significance on the estimate for Domestic_contagion implies there is no evidence of a domestic effect on mutual fund holdings either. Therefore, we accept hypothesis H2: the holdings of mutual funds did not serve as a transmission mechanism for contagion during the 2008 GFC. Table 5. Regression AR1 coefficient estimates of retail investor holdings on contagion models for the 2008 and 2011 global financial crises (t-stats). Variable Model (1) Model (2) Model (3) Model (4) Intercept 0.005 (0.119) 0.136 *** (2.836) 0.751 *** (14.963) 0.043 * (1.691) 0.053 * (1.876) 0.083 *** (3.293) 0.004 (0.148) 0.035 (0.632) 0.100 * (1.808) 0.002 (0.069) 0.037 *** (3.252) 0.005 (0.595) 0.011 (0.213) 0.136 *** (2.835) 0.751 *** (14.899) 0.044 * (1.694) 0.053 * (1.882) 0.083 *** (3.285) 0.004 (0.161) 0.034 (0.609) 0.099 * (1.799) 0.005 (0.152) 0.010 * (1.685) 0.005 (0.595) 0.010 (0.189) 0.013 (0.267) 0.136 *** (2.835) 0.751 *** (14.928) 0.044 * (1.707) 0.053 * (1.883) 0.083 *** (3.293) 0.005 (0.177) 0.034 (0.612) 0.100 * (1.802) 0.006 (0.188) 0.015 * (1.895) 0.011 (0.653) 0.014 (0.286) 0.018 0.264 0.136 *** (2.833) 0.751 *** (14.943) 0.043 * 1.685 0.052 * 1.871 0.083 *** (3.300) 0.005 (0.187) 0.034 0.622 0.100 * 1.807 0.006 (0.210) 0.020 ** 2.147 0.027 (0.668) 0.014 (0.333) 0.827 0.828 0.828 0.827 Lag_insiders Lag_retail Zero_ret_days Effective_spread Percent_bonds Volatility Fair_value Financial_sector Domestic_contagion Global_crisis_08 Global_crisis_11 Spillover_market Adj. R2 In all the tables, *** Significant at the 1% level, ** significant at the 5% level, and * significant at 10%. Hypothesis H3 states that change in the level of retail investor holding is a mechanism for the transmission of GFC effects. Table 5 below shows that in contrast to the negative sign on the set of crisis dummies for mutual fund and pension fund holdings, the signs on the parameter estimates for Global_crisis_08 are positive and significant for the Chilean market ( = 0.037, p < 0.01 in Model 1) and the US market ( = 0.020, p < 0.05 in Model 4). Significance is marginal for the relationship with the EU and Global equity markets in Models 2 and 3, with p < 0.10. These results indicate that lower equity returns in the local Chilean market and the US market during the 2008 GFC are associated with lower retail investor holdings, and allowing for marginal significance on the EU and global markets, the result holds for all three foreign markets. Hence, declines in the equity holdings of retail investors during the 2008 GFC may have been a source of transmission of the domestic contagion in Chile as identified by Bekaert et al. (2014) and the idiosyncratic contagion in Chile identified by Dungey and Gajurel (2015). The lack of significance between the equity holdings of retail investors and the proxy for the Global_crisis_11 contagion suggests that retail investors did not facilitate the transmission of the contagion during the 2011 GFC. Since we identified retail investors as a source of Int. J. Financial Stud. 2020, 8, 26 18 of 22 transmission for the contagion associated with the 2008 crisis, but not the 2011 GFC, we conclude that there is mixed evidence to support hypothesis H3. Hypothesis H4 states that change in the level of insider holdings is not a mechanism for the transmission of GFC effects to the Chilean financial markets. Table 6 below show that across all models, there is no significant evidence for insiders being affected by the Chilean equity market or the foreign markets. In fact, the only parameter estimate of significance is on lagged insider holdings ( = 0.904, p < 0.01) across all the models. It appears that the trades of insiders are driven by variables outside the scope of the study. Given this result and the failure to find any significant estimates on the crisis dummies, we conclude that there is support for hypothesis H4, suggesting that insiders do not change their equity holdings during crisis periods. Hence, as suspected, insiders may be failing to provide liquidity during a GFC that they normally do during a non-crisis period. Table 6. Regression AR1 coefficient estimates of insider holdings on contagion models for the 2008 and 2011 global financial crises (t-stats). Variable Model (1) Model (2) Model (3) Model 4) Intercept 0.012 (0.321) 0.904 *** (37.245) 0.014 (0.503) 0.031 (1.113) 0.005 (0.217) 0.020 (0.732) 0.028 (1.141) 0.006 (0.237) 0.003 (0.048) 0.026 (0.472) 0.025 (0.950) 0.001 (0.107) 0.005 (0.610) 0.003 (0.070) 0.905 *** (37.073) 0.015 (0.524) 0.030 (1.092) 0.008 (0.306) 0.021 (0.775) 0.027 (1.067) 0.005 (0.201) 0.005 (0.097) 0.026 (0.460) 0.032 (1.065) 0.003 (0.481) 0.002 (0.217) 0.024 (0.489) 0.023 (0.478) 0.904 *** (37.188) 0.014 (0.498) 0.031 (1.122) 0.005 (0.184) 0.020 (0.715) 0.029 (1.182) 0.007 (0.274) 0.001 (0.025) 0.027 (0.475) 0.020 (0.693) 0.002 (0.261) 0.011 (0.712) 0.017 (0.369) 0.034 (0.769) 0.904 *** (37.281) 0.014 (0.515) 0.031 (1.130) 0.006 (0.227) 0.020 (0.737) 0.032 (1.283) 0.009 (0.354) 0.001 (0.020) 0.026 (0.471) 0.012 (0.411) 0.007 (0.760) 0.036 (0.899) 0.040 (1.016) 0.835 0.835 0.835 0.836 Lag_insiders Lag_pension_funds Lag_mutual_funds Zero_return_days Effective_spread Percent_bonds Volatility Fair_value financial_sector Domestic_contagion Global_crisis_08 Global_crisis_11 Spillover_market Adj. R2 In all the tables, *** Significant at the 1% level. With regard to the effect of creditor monitoring, hypothesis H5 states that the holdings of local pension funds, mutual funds, and retail investors are sensitive to the level of public debt issued by Chilean firms. The creditor monitoring proxy Percent_bonds is positive and significant in Table 3 above for pension funds. The mutual fund models in Table 4 contains estimates on Percent_bonds that are positive and significant across all models (at = 0.094 to 0.103, p < 0.05). However, the results for retail investor holdings in Table 5 reveal that the estimate on Percent_bonds is negative and significant (at = 0.083, p < 0.01 across all models). Thus, we conclude that the results provide support for hypothesis H5. It seems that retail investors held fewer shares in companies with more debt, perhaps Int. J. Financial Stud. 2020, 8, 26 19 of 22 due to imperfect information that led them to infer a higher probability of default. In contrast, mutual funds and pension funds (marginally significant) held more shares in firms with more debt in the capital structure, a result perhaps related to the value they place on the monitoring of management by creditors, an important external corporate governance mechanism. 6. Discussion In Table 4, for mutual fund holdings, the Financial_sector dummy enters all the models as negative and significant (i.e., = 0.235, p < 0.01), but positive and significant across all models for retail investors in Table 5 (i.e., = 0.10, p < 0.10). A study of the banking sector firms market reaction to the fair value accounting (FVA) and impairment rules during the 2008 crisis by Bowen and Khan (2014, p. 233) found that investors acted as if the potential negative effects of then-existing FVA and impairment rules outweighed any benefits associated with having more timely and transparent mark-to-market data for decision making. Thus, it may be that risk-averse mutual fund managers avoided shares of financial institutions during these two crisis periods. This could be related to contagion if mutual fund managers were biased away from investing in financial institutions, given the uncertainty surrounding their exposure to US mortgage-backed securities and EU sovereign debt defaults. Thus, while creditor monitoring may have motivated institutional investors to increase equity holdings during a period highlighted by two severe GFCs, uncertainty regarding the precise nature of assets on the balance sheets of financial institutions and a failure to transparently recognize impairment losses on their investments may have had a negative and offsetting effect. Therefore, weaker regulation of firms within the financial sector may have hastened the outflow of funds through a reduction in mutual fund holdings across the study period. Table 7 below compares our generalized spillover and contagion results to the Chilean results in studies by Bekaert et al. (2014) and Dungey and Gajurel (2015). Table 7 shows that our results for retail investors provide a possible source for the Bekaert et al. (2014) and Dungey and Gajurel (2015) findings for Chilean contagion during the 2008 crisis. Moreover, there is no evidence that the 2011 GFC had any contagion effects on Chiles equity market, which we attribute to improved regulations, creditor monitoring, and fair value measurement following IFRS adoption. We provided insights into potential herding behavior among the four investor groups, by considering the lagged values of ownership holdings. Table 3 indicates that across all models, the current period holdings of pension funds are a positive and a marginally significant function of the prior periods mutual fund holdings (e.g., = 0.051, p < 0.10 on Lag_Mutual_funds). This result supports the possibility that local pension funds or AFP managers base at least a portion of their trading decisions on the prior periods mutual fund trades, which we note in Table 7 as pension funds herding behavior. Brown et al. (2013) attribute institutional herding to reputational effects. However, in the Chilean context, it is unlikely that the AFP managers with perhaps superior information on the Chilean economy were motivated by reputational effects to follow mutual funds. Thus, our results are probably not consistent with Brown et al.s (2013) study. For mutual funds, the Chilean equity market may have provided a relatively less risky and fundamentally sound investment opportunity compared to global equity markets in the US and EU. Kabir (2018) studied herding in the context of the US financial sector and found evidence for spurious herding, which he defined as unintentional herding driven by fundamental factors. Given the Chilean institutional context, spurious herding, as suggested by Kabir (2018), rather than a reputation-based explanation, probably drives our result for AFP managers herding behavior. In Table 5, a relationship exists between retail investors and lagged insiders, but it is a negative function of Lag_insider holdings ( = 0.136, p < 0.01 across all models). A possible explanation is that when insiders trade shares, retail investors are on the other side of the trade, hence the negative sign on Lag_Insiders. However, the result is not contemporaneous and warrants further study in future emerging market contagion research. Int. J. Financial Stud. 2020, 8, 26 20 of 22 Table 7. Comparing this studys result coefficients for spillover and contagion with other studies (pos = positive, neg = negative). Summary of Ownership and Contagion Results of This Study Pension Funds Mutual Funds d_factor pos ** g_ret neg ** Factor Returns and Contagion Studies Retail Investor Insiders 0 0 0 0 0 0 e_ret neg ** 0 0 0 u_ret 0 0 0 0 d_gfc_2008 neg ** neg ** pos *** 0 d_gfc_2011 0 0 0 0 g_gfc_2008 neg *** 0 pos * 0 g_gfc_2011 0 0 0 0 e_gfc_2008 neg *** neg ** pos * 0 e_gfc_2011 0 0 0 0 u_gfc_2008 neg ** 0 pos * 0 u_gfc_2011 0 0 0 0 herding pos *, with Mutual funds no neg ***, with Insiders no Bekaert et al. (2014) Dungey and Gajurel (2015) pos neg pos pos pos pos pos neg yes In all the tables, *** Significant at the 1% level, ** significant at the 5% level, and * significant at 10%. 7. Conclusions In this study, the methodology of Bekaert et al. (2014) and Dungey and Gajurel (2015) was modified to facilitate an analysis of contagion, creditor monitoring, and herding in terms of equity holdings instead of equity returns during a period that includes the 2008 US mortgage crisis and the 2011 EU debt crisis. One conclusion was that publicly traded shares of Chilean firms became more concentrated in the hands of mutual funds and pension funds during the 2008 US mortgage crisis, and less concentrated in the hands of retail investors. Thus, retail investors served as a mechanism to transmit contagion to Chiles stock market during the 2008 global financial crisis, while changes in the equity holdings of institutional investors tended to mitigate the transmission of the contagion. However, the same result does not apply to the 2011 EU debt crisis. None of the four investor groups equity holdings served to mitigate or magnify contagion effects during the 2011 crisis, and we suggested possible reasons for it. A second finding was of potential herding behavior in the Chilean equity market. The results indicate that variation in lagged mutual fund holdings explain a statistically significant portion of the variation in current period local pension fund or AFPs holdings during this period. Another result pertains to the relationship between creditor monitoring and the equity holdings of institutional investors. Results show that high levels of publicly issued debt in the capital structure of Chilean firms are associated with higher pension fund and mutual fund equity holdings. In contrast, the equity holdings of retail investors are lower in companies with higher levels of publicly issued debt. It appears that during a period defined by two substantial financial crises, institutional investors favored firms with more public debt in their capital structure, while retail investors may have associated a higher probability of default with more debt. The result suggests that in emerging markets during crisis periods, institutional investors prefer to invest in companies with greater potential for creditor monitoring of management, while retail investors may fear a greater risk of default. Thus, it does not automatically follow that firms with greater debt are more likely to have greater outflows. Some local institutional investors have a relatively steady flow of funds regardless of market conditions. This is particularly true of Chilean pension funds, given that payroll deduction is the source of their investable funds. It is plausible that greater opaqueness prior to IFRS adoption in 2009 led institutional investors to continue to invest in Chilean equities at historical rates during the Int. J. Financial Stud. 2020, 8, 26 21 of 22 2008 crisis, but not during the post IFRS adoption period containing the 2011 EU crisis. In other words, it is possible that an increase in transparency associated with fair value reporting allowed fund managers to differentiate companies by risk and adjust their investment strategies accordingly, particularly in the financial sector. If true, then IFRS adoption during the study period may explain the finding that institutional investors mitigated contagion during the 2008 crisis, but not during the 2011 crisis. A major limitation of the study is that it covers only the Chilean stock market that has less than 70 actively traded firms. Nevertheless, the conjectures that we raised warrant further analysis, and we encourage future research to pursue this line of investigation to identify the sources of contagion transmission into other emerging capital markets by extending Bekaert et al.s (2014) model and studying the benefits of having an active bond market. Author Contributions: All authors contributed equally. Conceptualization, S.M.; Data curation, T.G.; Formal analysis, S.M.; Investigation, B.S.-P.; Methodology, T.G.; Project administration, B.S.-P. All authors have read and agreed to the published version of the manuscript. Funding: This research received no external funding. Acknowledgments: The authors would like to thank the Chilean Stock Exchange for providing us the data. We would also like to thank the regulators and research staff of the Chilean Santiago Stock Exchange and Superintendecia Valores y Seguros that is now referred to as the Commission for the Financial Markets for their valuable insight during the presentation at their offices in Santiago, Chile. Conflicts of Interest: The authors declare no conflict of interest. Appendix A Table A1. Study Variables and Definitions. Variables Definitions insiders pension_funds mutual_funds retail pct_bonds spillover_market The percentage of outstanding shares held by the top 10 shareholders. The investment of AFPs measured as the percentage of shares held by local pension funds. The percentage of outstanding shares held by all other institutional investors. The percentage of outstanding shares not held by insiders and institutional investors. The percentage value of publicly issued bonds to total equity. The annual return for each of three foreign equity markets: EU Spilloverthe annual return for the MSCI European stock market Index, Global Spilloverthe annual return for the MSCI World Index, and US Spilloverthe annual return for the MSCI US stock market Index. Dummy variable that equals spillover_market in 2008, 0 otherwise. Dummy variable that equals spillover_market in 2011, 0 otherwise. domestic factorthe orthogonalized annual return of the Chilean equity market against EU Spillover, Global Spillover, and US Spillover. Global_crisis_08 Global_crisis_11 Domestic_contagion Control Variables volatility zero_return_days fair_value fin_dum es The annual average standard deviation of the daily stock return for each company. The number of each companys trading days per year with a 0% return. 1 if the firm uses fair value reporting, 0 otherwise. 1 if the firm is in the financial sector, 0 otherwise. Effective spreadtrading cost defined as: Trade execution price-midpoint of quoted spread divided quoted spread. References Aguiar, Mark, and Gita Gopinath. 2005. Fire-sale foreign direct investment and liquidity crises. Review of Economics and Statistics 87: 43952. [CrossRef] Bekaert, Geert, Michael Ehrmann, Marcel Fratzscher, and Arnaud Mehl. 2014. The global crisis and equity market contagion. The Journal of Finance LXIX: 2597655. [CrossRef] Bowen, Robert M., and Urooj Khan. 2014. Market Reactions to policy deliberations on fair value accounting and impairment rules during the financial crisis of 20082009. Journal of Accounting and Public Policy 33: 23359. [CrossRef] Int. J. Financial Stud. 2020, 8, 26 22 of 22 Boyer, Brian, Tomomi Kumagai, and Kathy Yuan. 2006. How do crisis spread? Evidence form accessible and inaccessible stock indices. Journal of Finance LXI: 9571003. [CrossRef] Brown, Nerissa C., Kelsey Wei, and Russ Wermers. 2013. Analyst recommendations, mutual fund herding, and overreaction in stock prices. Management Science 60: 120. [CrossRef] Bui, Nha Duc, Loan Thi Bich Nguyen, Nhung Thi Tuyet Nguyen, and Gordon Frederick Titman. 2018. Herding in frontier markets: Evidence from the Vietnamese stock market. Accounting and Finance 58: 5981. [CrossRef] Calvo, Sara. 1999. Capital flows to Latin America: Is there evidence of contagion effects? Available online: https://elibrary.worldbank.org/doi/pdf/10.1596/1813-9450-1619 (accessed on 3 March 2020). Chang, Eric C., Joseph W. Cheng, and Ajay Khorana. 2000. An examination of herd behavior in equity markets: An international perspective. Journal of Banking and Finance 24: 165179. [CrossRef] Chowdhury, Anup, Moshfique Uddin, and Keith Anderson. 2018. Liquidity and macroeconomic management in emerging markets. Emerging Markets Review 34: 124. [CrossRef] Connolly, Robert A., and F. Albert Wang. 2000. On Stock Market Return Co-Movements: Macroeconomics News, Dispersion of Beliefs, and Contagion. Available online: https://papers.ssrn.com/sol3/papers.cfm?abstract_ id=233924 (accessed on 3 March 2020). Denis, Diane K., and John J. McConnell. 2003. International Corporate Governance. Journal of Financial and Quantitative Analysis 38: 136. [CrossRef] Dungey, Mardi, and Dinesh Gajurel. 2015. Contagion and banking crisisInternational evidence for 20072009. Journal of Banking & Finance 60: 27183. Gillian, Stuart, and Laura Starks. 2003. Corporate governance, corporate ownership, and the role of institutional Investors: A global perspective. Journal of Applied Finance 13: 422. [CrossRef] Gjerde, Tom, Sakthi Mahenthiran, and David Cademartori. 2013. Effect of ownership, governance, and transparency on LiquidityChilean Evidence. Journal of Contemporary Accounting and Economics 9: 183202. [CrossRef] Gromb, Denis, and Dimitri Vayanos. 2010. The limits of arbitrage. Annual Review of Financial Economics 2: 25175. [CrossRef] Kabir, M. Humayun. 2018. Did investors herd during the financial crisis? Evidence from the US financial industry. International Review of Finance 18: 5990. [CrossRef] Kodres, Laura E., and Matthew Pritsker. 2002. A rational expectation model of financial contagion. Journal of Finance 57: 76999. [CrossRef] Kyle, Albert S., and Wei Xiong. 2001. Contagion as a wealth effect. Journal of Finance 56: 140140. [CrossRef] Lefort, Fernando, and Eduardo Walker. 2007. Do markets penalize agency conflicts between controlling and minority Shareholders? Development Economics XLV: 283314. [CrossRef] Lopez-de-Foronda, scar, Florencio Lopez-de-Silanes, Flix Lopez-Iturriaga, and Marcos Santamaria-Mariscal. 2018. Overinvestment, leverage and financial system liquidity: A challenging approach. Business Research Quarterly 98: 19. [CrossRef] Ma, Rui, Hamish Anderson, and Ben Marshall. 2018. Market volatility, liquidity shocks, and stock returns: Worldwide evidence. Pacific-Basin Finance Journal 49: 16499. [CrossRef] Morales, Marco, Mara Jos Melendez, and Vanessa Ramirez. 2013. Determinants of ownership concentration in the Chilean Stock Market. CEPAL Review 110: 17588. Optiz, Sebastian, and Alexander Szimayer. 2018. What drives flight to quality? Accounting and Finance 58: 52971. [CrossRef] Yuan, Kathy. 2005. Asymmetric price movements and borrowing constraints: A rational expectation equilibrium model of crises, contagion and confusion. The Journal of Finance LX: 379409. [CrossRef] 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).  ...

... The item referenced in this repository content can be found by following the link on the descriptive page.  ...

... Paper ID #36905 Works-in-Progress: Introducing Active Learning in Semiconductor Device Course Hansika Sirikumara Hansika Sirikumara, Ph.D., is an Assistant professor of Physics and Engineering at E. S. Witchger School of Engineering, Marian University Indianapolis. She completed her MS and PhD degrees from Southern Illinois University Carbondale. Her research expertise/interests are in engineering material properties for semiconductor device applications using computational methods. American Society for Engineering Education, 2022 Powered by www.slayte.com (Works-in-progress): Introducing Active Learning in Semiconductor Device Course Abstract A study of Semiconductor device concepts is a core area in the electrical and computer engineering curriculum, which introduces the principles and operation of basic semiconductor devices, and device characterization. The primary goal of the course is to develop a solid understanding of the semiconductor concepts and mechanism. This semiconductor device course knowledge is the foundation of many other electrical, electronic circuit courses in the engineering curriculum such as analog and digital electronics and very-large scale integrated (VLSI) devices. To achieve the main goals, instructors should choose various teaching strategies that accomplish a particular course objective. Active engagement of students is a key factor for effective delivery of the subject matters in engineering/technical subjects. As teaching tools in semiconductor devices, incorporation of structure visualization techniques, band engineering and device simulation tools such as Xcrysden, Quantum espresso and nanohub could vastly improve the understanding of the key concepts of semiconductor device. Also, these visualization techniques and simulation tools helps to build up the students active engagement in the classroom. From this perspective, this article presents how to successfully achieve the course learning outcomes and reinforces the understanding of semiconductor devices by implementing active learning environment using the available models and computer tools in the classroom. Introduction Semiconductors can be found everywhere in our modern lifestyle. Any electronic device you can think of made of semiconductors. Evolution of semiconductor industry can be traced back to the invention of the transistor and which could consider as the birth of modern technology era. A decade after the invention of transistor, integrated circuits (ICs) were invented and which initiated the development of faster, smaller, light weighted and cheaper electronic and electric equipment [1]. During the last few decades, ICs integration advanced further to develop high performance, reliable, multi-functional, energy efficient large-scale integrated circuits (LSI), very large-scale integrated circuits (VLSI) and ultra-large scale integration devices (ULSI) [2,3]. Now, almost everything in our everyday life supported by semiconductor-based devices and appliances. As a science/engineering major student, learning the semiconductor device course is essential to understand the fundamentals of semiconductor devices and the semiconductor technology as well as for the future development of the semiconductor field. Semiconductor device course is one of a professional courses for the electrical and computer engineering curriculum. This course introduces basic concepts and principles, operation of basic semiconductor devices, and device characterization. It provides the foundation required to pursue a career in an electrical / computer engineering profession or higher studies. At the end of the semiconductor device course, the students will get the solid foundation about fundamental theories, concepts, and methods of semiconductor devices, which can apply for real world problems in semiconductor devices and upon successful completion of this course, students are expected to:  Describe the fundamental semiconductor properties. Model and analyze the energy band diagram for semiconductor materials. Describe the principle and analyze the operation of pn-junction diode and the Schottky diode. Describe the principle and analyze the operation of Metal-Oxide-Semiconductor field Effect Transistor (MOSFET) and the Bipolar Junction Transistor (BJT). This paper will explain how to achieve the learning outcomes and reinforces the understanding by active learning methods of the semiconductor device course. Course outline Following discussion is based on the semiconductor device course offered in the fall 2021 semester as a three-credit hour course. The content of the course mainly divided into two sections: Semiconductor Physics and semiconductor devices. Semiconductor physics section includes Crystal structure, Semiconductor in equilibrium, Carrier Transport and Band structures. Semiconductor device section includes pn junction & Diode, Transistors, Bipolar Junction Transistors (BJT) and Field effect transistor (FET) & metal-oxide semiconductor field-effect transistor MOSFET. [4] A detailed outline of the sections covered in the course and allocated time for each section are listed here Material properties, crystal structure, crystal growth Carrier modeling (3 lectures) Semiconductor models, carrier properties, distribution, and concentration Carrier Action (3 lectures) Drift, diffusion, recombination-generation pn junction electrostatics (4 lectures) Fabrication, quantitative electrostatic relationships pn junction diode (4 lectures) Ideal diode equation, Deviations from ideal behavior, avalanche, Zener and tunnel diodes. Bipolar junction transistors (5 lectures) BJT fundamentals, BJT static characterization, BJT dynamic response modeling Field effect transistors (4 lectures) General introduction, fundamentals, electrostatics MOSFET devices (4 lectures) Active learning approaches Traditional teaching methods typically rely on students learning class material passively, which involves listening to lectures and taking notes. But in general, student attention span during a lecture is limited. To keep the healthy learning environment throughout the lecture period, different active learning methods were introduced as in class activities. These activities help students to refresh their minds and actively engage in the learning process. Many studies show that the active learning approaches increase the students performance in the classroom than the traditional teaching methods [5,6]. Freeman et.al reported that the STEM degree holders are more favorable for active learning-based lectures than traditional lectures [7]. Brooks DC reported that technological based learning environment improved the students conceptual understanding about the subjects [8]. There are several active learning methods could be implemented during a classroom environment such as problem/project based learning, small group discussion, case studies, peer teaching [9]. Most students are visual learners and many of them also experience improved learning through the use of technology [10,11]. Therefore, in order to promote active learning environment in the semiconductor device course, graphical presentations, video presentations and hands-on learning activities were implemented. Also, inclass experiments, demonstrations and simulations were conducted to illustrate complex concepts in the class. At the introduction of a novel concept or a complex case related to the subject matter or an advanced mathematics related section, discussions were initiated with realworld situations and real-world examples. Once students understand the concepts qualitatively, mathematical equations were incorporated to quantify the concepts and to strengthen their understanding. Ultimately, one of the primary goals is prepare the students for future opportunities by improving their problem solving and critical thinking skills. Following section discuss the a few active learning tools implemented during the course. Hands-on learning activities According the textbook used for the semiconductor device course [4], the first chapter introduces the fundamentals of semiconductors, i.e., atoms and crystal structures of semiconductors. In this chapter it is essential to identify how students visualize and analyze these concepts. Couple of hands-on learning activities could lead the way to visualize the complex concepts. To visualize the atomic structure of semiconductors, students were guided to conduct small group activities using a molecular structure model. During the class time, following the theory section of crystal structures, students were encouraged to constructed different types of crystal models. The goal of this activity was to picture the 2D and 3D visualization of the crystal structure. Using these models, students were able to understand the lattice points, lattice constant, unit cell, nearest neighbors, coordination number, unit cell, miller indices etc. Figure 01 shows the selected atomic structure models, which were constructed during a classroom activity. Figure 01: Selected atomic structure models constructed by the students Problem based activities In-class worksheet problems were found to be a great addition to improve the active learning environment of the classroom. Following a short theory section or a concept, in-class worksheet problems were given to the students. Students were encouraged to engage and discuss the problems and answer with peers and share their knowledge. Also, they were motivated to express their understanding of the concepts or explain their solutions to the fellow students. The goal of the worksheet problems was to engage all the students in the learning process. As an example, in the chapter 1, the worksheet was designed for understanding the concept of lattice, unit cell and crystal structure. In this activity, students were constructed the crystal structure using a crystal structure model and then completed the in-class worksheet by looking at their crystal models. Computer Modeling Computer simulations and incorporation of computer programs in teaching activities are also an effective way of improving the learning outcomes. After students were familiar with the crystal structure modelling, crystal structure visualizing softwares were introduced. In semiconductor device course, students were trained to use Xcrysden software to visualize the crystal structures. Xcrysden is a molecular modeling and visualization software and it can be run in the UNIX platform [12]. Sample Activity: Understand the unit cell and electronic band structure STEP-01: Create an input file for the unit cell: (Unix platform) Bravais lattice type The lattice constant Number of atoms and types of atoms in the unit cell Atomic masses Atomic positions of each atom in the unit cell STEP-02: Visualize the structure (Using Xcrysden) Open the Xcrysden Go to open pwscf file STEP-03: Band Structure calculation (Quantum espresso package) Generate k points in high symmetric points Optimization of the atomic coordinates to minimize the forces Calculate the electronic bands for high symmetric points in the crystal cell. The goal of this activity was to introduce the concepts about crystal structure and calculating the electronic band structures. During this classroom activity, students were guided to create an input file for relevant crystal structure using a text editor such as Vi-editor (STEP-01) in the above chart). The following chart shows the sample input file for Silicon crystal structure. Sample input file for silicon crystal structure &CONTROL prefix=silicon Outdir=./ pseudo_dir = './' / &SYSTEM ibrav = 2 celldm(1) = 5.431 nat = 2 ntyp = 1 ecutwfc = 18.0 / &ELECTRONS / ATOMIC_SPECIES Si 28.085 Si.pz-vbc.UPF ATOMIC_POSITIONS crystal Si 0.00 0.00 0.00 Si 0.25 0.25 0.25 K_POINTS automatics 444000 Students use this input file to visualize the crystal structure during the lecture (STEP-02). Moreover, they can understand and visualize the lattice constant, number of atoms in the unit cell, bonding length and angles and bravais lattice types for different semiconductor materials such as Graphene, Al, Ge etc. Visualization angles created by Xcrysden software were used to identify the pure and doped crystal structures, which is shown in Figure 02-(a). Further, students were guided to use the created input file to calculate the electronic band structures for given materials as shown in Figure 02-(b), which will discuss in the next section. Figure 02: (a) Si Crystal structure modeled from Xcrysden and (b) electronic band structure for Si Computer Simulation Computer simulations could use to visualize physical concepts that are hidden in the abstract mathematical language. In Semiconductor device course, semiconductor band engineering is one of the advanced topics, which involves numerous mathematical equations and mathematical concepts. Quantum Espresso (QE) program was incorporated to introduce the band structure theory to the students [13]. Quantum Espresso is an open-source software which can be used for calculating electronic band structures using first principle calculations. Students were trained to use the QE to calculate the simple band structures such as Si, GaAs, Graphene. These calculations were beneficial to understand the concepts of energy band levels, Fermi energy, carrier concentration, band gap, density of states and effective mass. Figure 03: Generated high symmetric points for FCC lattice structure using Xcrysden According to STEP 03, students were guided to use the created input file as shown in above to identify and generate the high symmetric k points in the crystal structure using the Xcrysden software. The generated k points can be used to calculate the band structure of selected crystal structures. The Figure 03 shows the generated high symmetric k points for silicon crystal using Xcrysden software. Figure 02-(b) shows the calculated band structure for pure Si crystal structure using the QE software as a classroom activity. After students gained firm knowledge about these concepts, it would be effortless to introduce the working principle of semiconductor devices such as pn junction and fields effect transistors (FET). Construction (modeling, designing, and simulating) of pn junction diode is a great example to demonstrate the knowledge of semiconductor concepts learned so far in the course. During this section of the semiconductor device course, the working principle of pn junction theory and pn junction device properties were introduced to the students. To buildup the understanding of this section, nanohub simulations were incorporated as an active learning tool. nanohub is a collection of simulation tools which can be used for material related computational simulations [14,15]. In this course, students were directed to use a few nanohub tools such as pn junction lab, 2DFET [16,17]. As a classroom activity, students were instructed to modify the pn junction diode parameters to using nanohub tools to observe the effect of various parameters. This class activity guide students to creatively understand the future developments of the devices. Sample Activity: Analyze and understand the concepts of semiconductor devices using the depletion approximation STEP-01: Create a pn junction for given parameters using nanohub tool. Example: A Si pn junction is at room temperature under equilibrium conditions. It consists with a p-doping density of NA = 2 x 1015 /cm3 and an n-doping of ND = 1x1015/cm3. STEP-02: Compute and analyze the several characteristic curves and parameters for the created device. Example: Depletion layer width, boundaries and potentials Plot charge density, electric field, and electrostatic potential as a function of position STEP-03: Compare and analyze the results by changing the doping density , temperature and width of the device. Figure 04 shows the selected characterizations curves for pn junction obtained from the nanohub tool during a classroom activity. Figure 04: Selected characterizations curves for Si pn junction obtained from the pn junction lab nanohub tool; (a). pn junction parameters (b). IV-curve (c) CV-curve (d) Energy band diagram (e). Charge density diagram (f) Electric field diagram Assessment Evaluation of students performance through continuous assessments is important for the students as well as for the future development of the course. Also, the assessment results indicate the evidence for the effectiveness of active learning methods implemented in the course. To collect the students feedbacks about the active learning methods used in the class, the informal survey was done in the mid of the semester and the following chart shows the evaluation results for informal survey. The following questions were used to get the students feedbacks. Q1: Multiple instructional methods were used in the course (e.g. lectures, problem solving, case studies, hands-on-activities, in class activities, computer tools, discussions, etc.). Q2: The readings, discussions, lectures, in class activities, and/or projects helped me attain the stated learning outcomes of this course. Q3: The instructional activities and assignments supported the course learning outcomes. Q4: The activities and assignments challenged me to think more deeply/critically about the course subject matter. Q5: The in-class activities/simulation/computational tools/small group sessions in this course furthered my understanding of the subject matter. Figure 05: The students feedback for informal survey. The students feedback shows that the activities and assignments done during the semiconductor course were helpful to understand and think more critically about the semiconductor device concepts. In the current semiconductor devices course, the assessments were based on weekly homework (40%), Two Midterm Exams (15% each), Quizzes (at the end of each sections-10%) and a Final Exam (20%). The simulations activities are included in the homework assessment. Following table shows the direct assessment averages of each category for fall 2021 semiconductor devices course at the Marion University. [Note: this course offered by the author for the first time] Table 1: Course direct assessment averages for each assessment category Assessments Homework Quiz Mid Term Final Class average 89.5 85.4 79.4 82.25 According to these results, the class average for each assessment is above 75%. It indicates that the active learning methods used in the semiconductor device course have been successful and the students meet their learning goals. Based on the survey responses collected from Marian University at the end of the fall 2021 semester and informal survey done by the instructor (author), students acknowledged the instructors attempts to create active learning approaches during the Semiconductor device course. Conclusion Most of the industries are looking for engineering graduates with solid knowledge as well as excellent technical skills. In order to fulfill the current demand, instructor must train students for defining, understanding, and solving problems with an organized critical way of thinking. Active learning approaches are one of the predominant pathways to train students towards these goals. This article provided the detailed explanations of active learning approaches, which can be used in the semiconductor device course. According to the preliminary results, the students valued the active learning activities implemented during the teaching/leaning process. I believe that active leaning environment is essential for successfully achieve the teaching goals as well as it is beneficial for students development and society at large References 1. Saxena, A.N., 2009. Invention of integrated circuits: untold important facts. World Scientific. 2. Zhao, B., 1998, October. Advanced interconnect systems for ULSI technology. In 1998 5th International Conference on Solid-State and Integrated Circuit Technology. Proceedings (Cat. No. 98EX105) (pp. 43-46). IEEE. 3. Schaper, L., Burkett, S., Gordon, M., Cai, L., Liu, Y., Jampana, G. and Abhulimen, I.U., 2007, May. Integrated system development for 3-D VLSI. In 2007 Proceedings 57th Electronic Components and Technology Conference (pp. 853-857). IEEE. 4. Pierret, R.F., 1996. Semiconductor device fundamentals. Pearson Education India. 5. McCarthy JP, Anderson L. Active learning techniques versus traditional teaching styles: Two experiments from history and political science. Innovative higher education. 2000 Jun;24(4):279-94. 6. Hamann K, Pollock PH, Wilson BM. Assessing student perceptions of the benefits of discussions in small-group, large-class, and online learning contexts. College Teaching. 2012 Apr 1;60(2):65-75. 7. Freeman, S., Eddy, S.L., McDonough, M., Smith, M.K., Okoroafor, N., Jordt, H. and Wenderoth, M.P., 2014. Active learning increases student performance in science, engineering, and mathematics. Proceedings of the national academy of sciences, 111(23), pp.8410-8415. 8. Brooks DC. Space matters: The impact of formal learning environments on student learning. British Journal of Educational Technology. 2011 Sep;42(5):719-26. 9. Hernndez-de-Menndez M, Vallejo Guevara A, Tudn Martnez JC, Hernndez Alcntara D, Morales-Menendez R. Active learning in engineering education. A review of fundamentals, best practices and experiences. International Journal on Interactive Design and Manufacturing (IJIDeM). 2019 Sep;13(3):909-22. 10. Zopf, R., Giabbiconi, C.M., Gruber, T. and Mller, M.M., 2004. Attentional modulation of the human somatosensory evoked potential in a trial-by-trial spatial cueing and sustained spatial attention task measured with high density 128 channels EEG. Cognitive brain research, 20(3), pp.491-509. 11. Jawed, S., Amin, H.U., Malik, A.S. and Faye, I., 2019. Classification of visual and nonvisual learners using electroencephalographic alpha and gamma activities. Frontiers in behavioral neuroscience, 13, p.86. 12. Kokalj, A., 1999. XCrySDena new program for displaying crystalline structures and electron densities. Journal of Molecular Graphics and Modelling, 17(3-4), pp.176-179. 13. Giannozzi, P., Baroni, S., Bonini, N., Calandra, M., Car, R., Cavazzoni, C., Ceresoli, D., Chiarotti, G.L., Cococcioni, M., Dabo, I. and Dal Corso, A., 2009. QUANTUM ESPRESSO: a modular and open-source software project for quantum simulations of materials. Journal of physics: Condensed matter, 21(39), p.395502. 14. Madhavan, K., Zentner, L., Farnsworth, V., Shivarajapura, S., Zentner, M., Denny, N. and Klimeck, G., 2013. nanoHUB. org: cloud-based services for nanoscale modeling, simulation, and education. Nanotechnology Reviews, 2(1), pp.107-117. 15. Klimeck, G., McLennan, M., Brophy, S.P., Adams III, G.B. and Lundstrom, M.S., 2008. nanohub. org: Advancing education and research in nanotechnology. Computing in Science & Engineering, 10(5), pp.17-23. 16. Dragica Vasileska, Matteo Mannino, Michael McLennan, Xufeng Wang, Gerhard Klimeck, Saumitra Raj Mehrotra, Benjamin P Haley (2014), "PN Junction Lab," https://nanohub.org/resources/pntoy. (DOI: 10.4231/D3GH9B95N). 17. Ning Yang, Tong Wu, Jing Guo (2021), "2DFET," https://nanohub.org/resources/2dfets. (DOI: 10.21981/MCT5-1694).  ...