ABSTRACT 
 
 
 
 
 

Tittle of thesis:  
 
 
 
 
 
 
 
 
 
Thesis directed by: 
 
 

RAPID DESTABILIZATION OF DEEP, SUPERHYDROUS 
MAGMA PRIOR TO THE LARGEST KNOWN PLINIAN 
ERUPTION OF CERRO MACHIN VOLCANO, 
COLOMBIA 
 
 
Silvia Camila Castilla Montagut  
Master of Science, 2022 
 
 
Professor Megan Newcombe 
Department of Geology   
 

 

 
 

A detailed stratigraphy of the pyroclastic fall deposits associated with Cerro Machin 

volcano (CMV) is presented following a previously defined categorization of the different 

eruptive units: El Espartillal, P0, P1, El Guaico, P2 and El Anillo. For the largest Plinian 

eruption in the sequence (P1), two lithofacies were distinguished on the basis of 

sedimentary features, grain size and componentry analysis. Early stages of the eruption 

could have been associated with vulcanian-type phases characterized by conduit/plug 

clearing explosions, producing a monolithologic lithic-rich laminated basal layer. The 

climactic event is represented by a white to grey, clast-supported, reverse to normally 

graded pumice-rich lapilli layer formed by a sustained eruptive column that gradually 

waned towards the end of the eruption. Associated deposits were identified up to 40 km 

from the vent. Pumice clasts from the most explosive phase were sampled along the 



 

deposit layer in order to characterize storage conditions and ascent rates for the magma 

erupted.  

Pumice samples were classified as medium-K, calc-alkaline dacites (63-67 wt.% SiO2). The 

mineral assemblage includes plagioclase+amphibole+biotite+quartz and olivine and 

orthopyroxene (Fo 89-92) as accessory phases with amphibole overgrowths. 

Geothermobarometry of unzoned amphiboles, cores of reversely zoned crystals and 

rims of normally zoned crystals indicate a temperature range from 825±17°C and 

913±45°C, a pressure range from 270±75 MPa to 1000±320 MPa indicating crystallization 

depths of 8-29 km. Thermobarometry of minor populations of unzoned amphibole 

crystals, cores of normally zoned crystals and rims of reversely zoned amphiboles show 

the same crystallization pressures as the dominant amphibole populations, but higher 

temperatures between 850±17°C and 978±29°C. The presence of these small populations 

of high-temperature amphiboles suggests a minor recharge event that did not drastically 

change the average crystallization conditions of the main dacitic reservoir but that could 

have been the trigger mechanism for the explosive eruption.   

CMV dacites display several geochemical signatures of adakites (low Y < 14 ppm, high 

Sr >700ppm, high Al2O3 > 16 wt.%, low MgO < 3 wt.%) which suggest that CMV magmas 

are produced by fractional crystallization of primitive hydrous magmas at the Moho 

boundary. The primitive magmas could have been the result of the interaction between 

Ba-enriched fluids dehydrated from the subducted slab with mantle peridotite in the 

mantle wedge. Three lines of evidence support the presence of superhydrous magma 

(containing >~8 wt% H2O) beneath Cerro Machin: 1) water concentrations of 2–11 wt% 

measured in plagioclase- and quartz-hosted melt inclusions; 2) the presence of Fo89-92 

olivine rimmed by high Mg# amphibole; and 3) measurements of ~100–167 ppm H2O in 



 

plagioclase phenocrysts. Assuming a partition coefficient of 0.002, measured plagioclase 

crystals crystallized in a melt with 5-8 wt% H2O. 

Water-diffusion modelling in plagioclase crystals indicates a minimum time of 1 day for 

the magma to ascend from shallow depth (<250 MPa, ~5 km) before the eruption. Rapid 

ascent times are also suggested by the absence of breakdown rims in amphibole crystals 

which indicate a magma ascent timescale of <4 days from 8 km to the surface 

(Rutherford & Hill, 1993). Results from this study indicate that the 3600 yr BP eruption 

was preceded by rapid mobilization and ascent of superhydrous dacitic magma from 

mid- to deep-crustal storage regions beneath Cerro Machin. 

This thesis comprises eight appendices (A-H), Appendix A contains fieldwork 

information including: map of locations, a description of sample labels and schematic 

stratigraphic columns. In Appendix B, descriptions of componentry analysis are 

presented as well as pictures of the different grains identified under binocular 

observations. Appendix C is conformed by ten petrographic forms in which 

petrographic observations are detailed. This is followed by Appendix D in which whole 

rock chemistry of five samples is listed. In Appendix E, water measurements in melt 

inclusions are presented. In Appendix F, measurements of water in feldspar are detailed 

while in Appendix G, glass, melt inclusions, mineral chemistry and pictures of each 

crystal are presented, it also contains plagioclase-melt and amphibole-melt equilibrium 

tests and geothermobarometry calculations. Finally, Appendix H contains the link with 

the repository of the Matlab code used for volatile diffusion modelling in feldspar.  

 
 
 



 

RAPID DESTABILIZATION OF DEEP, SUPERHYDROUS MAGMA PRIOR TO THE LARGEST 
KNOWN PLINIAN ERUPTION OF CERRO MACHIN VOLCANO, COLOMBIA 

 
 
 
 

by  
 
 

Silvia Camila Castilla Montagut 
 
 
 
 
 

Thesis submitted to the Faculty of the Graduate School  
of the University of Maryland, College Park, in partial fulfillment  

of the requirements for the degree of  
Master of Science 

2022 
 
 
 
 
 
 
 
 
 
 

Advisory committee: 
Professor Megan Newcombe, Chair 
Professor. Philip Piccoli  
Professor Sarah Penniston-Dorland  

 
 

 
 
 
 

 



 

 
 
 
 
 
 
 
 
 
 
 
 
 
 

© Copyright by 

Silvia Camila Castilla Montagut 

2022 

 



ii 
 

Acknowledgments 

My journey would not have been the same without all the people I met along the way 

and for whom I will always feel grateful and blessed. With mixed feelings of happiness 

and nostalgia, I would like to thank all of you who, in one way or another, contribute to 

my professional and personal growth in these two years and made this experience one 

of the best decisions of my life.  

I would like to start expressing my deepest gratitude to my amazing and wonderful 

supervisor Megan, in who I always found guidance, expertise and interesting discussions. 

Your continuous support and encouragement, your enthusiasm for science and the 

compassion for your students have inspired in me the kind of scientist I want to be. 

Thank you for always be a confident place to express my ideas and opinions, a friendly 

face every week during two years and of course the most contagious laugh ever. You are 

by far the best definition of a mentor, and I will be eternally grateful for the amazing 

experience we had working together. I also would like to say a huge thank you to the 

members of my thesis committee, Professor Sarah Penniston-Dorland and Professor 

Philip Piccoli. I would like to express my gratitude to Professor Phil, whose most 

intriguing and interesting questions vastly influenced my critical thinking and my 

passion for science, for his assistance with the electron microprobe analysis and for 

always being a helping hand without hesitation. To my lab mates: Cosmo, Kyle, Liam, 

MaKayla, Ayomide, Sumedah and Kathy, whose their great sense of humor always 

brought me joy on our group meetings. I am grateful to the Geology Department 

community of the University of Maryland for providing a warm and fertile working 

environment. 



iii 
 

I am deeply indebted to the Fulbright program for financing this study, for giving me 

the opportunity to represent my country Colombia and for the incredible friendships I 

made through the Fulbright community. Major thanks to the Colombian Geological 

Survey, for the support in the initial stage of my research providing samples of Galeras 

volcano which unfortunately could not been used for my studies. Special thanks to my 

constant support and guidance, profe Natalia Pardo, for always being part of my 

academic growth and for your unconditional help in every stage of my professional life. 

To Ricardo Mendez, Diego Gomez, Gloria Patricia Cortes, Marta Calvache, Ana Maria 

Correa and Carlos Zuluaga for their support and guidance at the any stage of this 

project.  

My experience was indelibly enriched by the wonderful friendships I gained from this 

journey. I would like to mention my friends Jacek, Sergio, Cesar, my friends from the 

Fulbright community: Ariko, Juliane, Nina, Khursheed, Juan, Vera and Andrea, my 

friends from the Latin crew: Jairo, Julio, Aldana, Mari, Share, Nico, Sofi, Max, Manuel, 

Tania, William, and Sergio – thank you for the parties, barbecues, laughs, beers, arepas 

and the uncountable memories I will keep forever. I would like to thank the members of 

the soccer team Hakuna Matata, present and past: Kyle, Karla, Liam, Jiayang, Elaine, 

Sergio, Diego, Sam, Sarah, Melody, Maya, Jacek, Garyson, Tania, Julio, Manuel, Cesar, 

Jacob and Megan. it was an honor being your captain and a total pleasure to play with 

you every weekend.  I also thank my roommates Yousra, Veronica and Johanna, for 

listening to me after long and stressful days, for sharing with me your life stories, for 

supporting me during my personal challenges, for becoming a family to me in this new 

country, for the laughs and secrets we shared and the bond we have built. I want to 

specially thank my roommate and friend Johanna in who I found a sister, and who always 



iv 
 

made me feel close to Colombia. Thank you for spoling me, for motivating me every time 

I needed, for always being willing to listen and share, for crying with me and of course 

for the delicious muesli, arepas, pan de bonos, pan de yucas, chocolate caliente and 

everything you always did with love for me. Thank you from the bottom of my heart to 

my family in Colombia, to my dad, my sister and specially to my beautiful mom who 

every day sent me her love, support and funny pictures of Matata. To my friends all 

around the world Ana, Mari, Ale, Laura, Natis and Sergio that were always supporting 

me and have been fans #1 of my dreams.    

Last, but not least, I would like to dedicate this thesis to my lovely husband, Juan David, 

who has witnessed my process since the first day and has helped me to get through all 

the existential crisis that graduate school can potentially arise. Thank you for always 

believing in me, for supporting my dreams, for your help in the fieldwork we did for this 

research and for using it as the perfect excuse to propose and made me the happiest 

woman ever. Thank you for helping me during long summer days in the lab polishing 

melt inclusions, for practicing with me every presentation I did during my master’s, and 

for being source of strength, support, patience, and motivation during this entire 

experience. Thank you for being the best life partner. ¡Te amo! 

  

 

 

 

 

 

 

 



v 
 

Table of Contents 

 
Acknowledgments ........................................................................................................................... ii 

Table of Contents ............................................................................................................................ v 

List of figures ................................................................................................................................. vii 

List of tables .................................................................................................................................... xi 

Chapter 1: Introduction ................................................................................................................. 1 

1.1 Geological background ................................................................................................... 4 

1.2 Literature review .............................................................................................................. 7 

1.2.1 Plinian Eruptions ...................................................................................................... 8 

1.2.2 The role of water in melting processes ............................................................. 17 

1.2.3 Magmatic water at storage conditions .............................................................. 19 

1.2.4 Water behavior during magma-ascent decompression .................................. 23 

Chapter 2: Motivation .................................................................................................................. 29 

2.1 Colombian perspective ................................................................................................ 32 

2.2 Statement of purpose ................................................................................................... 34 

2.2.1 Specific objectives ................................................................................................. 34 

Chapter 3: Methods ..................................................................................................................... 35 

3.1 Field work ....................................................................................................................... 35 

3.2 Sample preparation ...................................................................................................... 36 

3.3 Componentry and petrographic analysis ................................................................. 37 

3.4 Whole rock geochemistry ............................................................................................ 38 

3.5 Fourier Transform Infrared Spectrometer (FTIR): volatile concentrations ........ 38 

3.6 Electron Probe Microanalyzer (EPMA): glass, melt inclusions and mineral 

chemistry ................................................................................................................................... 42 

Chapter 4: Results ........................................................................................................................ 45 

4.1 Stratigraphy: pyroclastic fall deposits ...................................................................... 45 

4.2 Componentry and petrography .................................................................................. 51 

4.3 Volatiles .......................................................................................................................... 55 

4.4 Water in feldspar ........................................................................................................... 61 

4.5 Plagioclase and amphibole geothermobarometry .................................................. 65 

4.6 Whole rock chemistry .................................................................................................. 71 



vi 
 

Chapter 5: Discussion ................................................................................................................. 75 

5.1 Dynamics of main Plinian eruptions in CMV ........................................................... 75 

5.2 Geothermobarometry ................................................................................................... 76 

5.3 Evidence of a superhydrous magma ......................................................................... 78 

5.4 Origin and evolution of CMV dacitic melts .............................................................. 80 

5.5 Triggering mechanism for the largest Plinian eruption of CMV .......................... 86 

5.6 Magma ascent history for the largest Plinian eruption of CMV ........................... 87 

Chapter 6: Conclusions ............................................................................................................... 91 

Appendixes .................................................................................................................................... 94 

Appendix A:  Locations map and stratigraphic columns ................................................. 95 

Appendix B: Componentry analysis.................................................................................... 100 

Appendix C: Petrography analysis ...................................................................................... 103 

Appendix D:  Whole rock geochemistry ............................................................................ 148 

Appendix E: Volatile content in melt inclusions .............................................................. 149 

Appendix F: Water content in feldspar .............................................................................. 156 

Appendix G: Glass, melt inclusions and mineral chemistry of amphibole, plagioclase, 

olivine and orthopyroxene ................................................................................................... 176 

Appendix H:   Matlab code for water diffusion in plagioclase ...................................... 226 

References ................................................................................................................................... 227 

 



vii 
 

List of figures  

Figure 1. Location of Cerro Machin volcano, Colombia. A. Neotectonic map of Colombia 

with the main fault systems. B. Geological setting of Cerro Machin volcano delimited by 

Machin and Cajamarca faults. Geomorphological features are also shown, symbols taken 

from (Castilla et al., 2021). ............................................................................................................ 5 

Figure 2. Digital elevation model of the outcrop locations (blue dots). Samples from 

location CMV-008 (represented by a yellow star in the map) were used for lab analysis. 

Isopachs are shown in black - thickness in cm – towns, streams and roads in grey. ....... 7 

Figure 3. Spatial distribution of Plinian eruptions over the last 50 kyr. Color shows 

compositions while size of symbol represents VEI. Symbols with more than one color 

are representing the different compositions of multiple eruptions of the same VEI. 

Database taken from LaMEVE (Large Magnitude Explosive Volcanic Eruptions) of 

VOGRIPA project. ......................................................................................................................... 10 

Figure 4. Different regions of a typical Plinian-sub Plinian eruptive column (modified 

from Cioni et al., 2015). .............................................................................................................. 12 

Figure 5. Frenkel defects in the NaCl structure. It occurs when an atom leaves its normal 

site vacant and instead occupies an interstitial site (Aschcroft & Mermin, 1976). ......... 23 

Figure 6. Compilation of published water content in feldspar crystals by (Mosenfelder et 

al., 2020). ....................................................................................................................................... 31 

Figure 7. Geological context of active Colombian volcanoes. A. Map of Colombia with the 

different active volcanoes. B. Most tragic eruptions in Colombia, each bar represents one 

eruption from a specific volcano and the associated number of deaths. ........................ 33 

Figure 8. Sketch of methodology for measuring water concentration in a plagioclase 

crystal by FTIR. A single plagioclase crystal is cut in half and the two halves are doubly 

polished. One half is polished to create a slab parallel to the cut face of the crystal and 

the other half is polished to create a slab perpendicular to the cut face. Green lines in 

each half show the different profiles taken by the FTIR on each half of the crystal. Red 

lines show the polarization direction at 0° and 90° of the IR beam. Note that this method 

allows us to make polarized FTIR measurements in three perpendicular directions 

through the crystal, which enables us to quantify the total concentration of dissolved 

water. X, Y and Z denote crystallographic axes, but it is important to note that water 

concentration can be quantified using light polarized in any three perpendicular 

directions (i.e., it is not necessary to perfectly align the slabs with the crystallographic 

axes). ............................................................................................................................................... 41 

Figure 9. Proximal outcrops of the study eruption units - 3 km from the vent. A. 

Interbedding between clast-supported pyroclastic fall deposits and matrix-supported 

pyroclastic flows overlined by grey to brownish fine-ash cross-laminated pyroclastic 

flow deposits. Scale 20 cm. B. Occasional charcoal fragments in the matrix-supported 

pyroclastic flow deposits at the bottom of the sequence. C. Photograph of different 

eruptive units identified in medial areas. D. Close up of the selected unit P1 for following 

lab analysis. ................................................................................................................................... 46 



viii 
 

Figure 10. Stratigraphic correlation of the marker horizon identified from proximal to 

distal areas. ................................................................................................................................... 47 

Figure 11. Schematic stratigraphic column of location CMV-008. Medial areas outcrop - 

13 km from the vent. ................................................................................................................... 50 

Figure 12. Photographs of glassy particles under binocular microscope. A. white, fresh, 

highly vesiculated, coarse-grain, glassy particles with phenocryst of plagioclase, 

amphibole, biotite and quartz. B. gray, fresh, highly vesiculated, fine-grain glassy 

particles with crystals of plagioclase, quartz, amphibole and biotite. C. Banded particles 

combining white and gray textures. D. Fresh, dense, hypohyaline, porphyritic lithic 

fragments. E. Altered dense lithic fragments. F. Irregular fragments of the metamorphic 

basement in CMV. ........................................................................................................................ 52 

Figure 13. Photomicrographs of the different regions identified in P1 pumice products. 

A. dense, holocrystalline, porphyritic lithic fragments with plagioclase-amphibole 

crystals in a crystalline matrix. B. Irregular metamorphic fragments from the basement 

of CMV. C. Moderate to highly vesiculated, medium-grained region. D. Highly vesiculated, 

fine-grained region. E. and F. Banded pumice between same regions. .............................. 53 

Figure 14. Photomicrographs of the different textures identified in plagioclase (Pl), 

amphibole (Amp), quartz (Qz), biotite (Bt), olivine (Ol) crystals. The top left region of 

each panel is imaged in plane-polarized light, whereas the lower right region of each 

panel is imaged in cross-polarized light. A. Zoned and subhedral phenocryst of 

plagioclase and subhedral biotite microcrysts in a hypohyaline and vesiculated 

groundmass. B. Microcrystal of plagioclase with a reabsorbed rim and anhedral shape. 

C. Euhedral crystal of amphibole, fractured. D. Zoned and euhedral microcryst of 

amphibole with at least one rim. E. Anhedral microphenocryst of quartz with melt 

inclusions. Euhedral crystal of biotite also observed. F. Anhedral microphenocryst of 

olivine surrounded by a rim of amphibole. ............................................................................ 55 

Figure 15. Volatiles in melt inclusions hosted in plagioclase and quartz crystals. A. 

Photomicrograph of conical-shaped melt inclusion hosted in quartz. B. H2O vs CO2 plot 

showing 36 melt inclusions hosted in plagioclase and quartz crystals. Also shown are 

calculated vapor saturation isobars (contours of constant pressure) and vapor isopleths 

(contours of constant vapor composition) for 10, 20, 40, 60 and 80% CO2, all calculated 

using VolatileCalc (Newman and Lowenstern, 2002). Closed system degassing paths with 

0, 2, 5 and 10% of exsolved vapor also calculated using VolatileCalc are shown. 

Uncertainty for each melt inclusion H2O and CO2 value was calculated using a Monte 

Carlo simulation (Newcombe, pers. communication, 2022): each parameter (absorbance, 

wafer thickness, density, absorption coefficient) was sampled 1000 times from a normal 

distribution (with a mean and standard deviation representative of the measured values) 

to produce H2O and CO2 concentrations. A representative standard deviation of the 

resultant H2O and CO2 distributions is plotted in the lower right corner of the plot. ... 57 

Figure 16. Plots of volatile content in melt inclusions and matrix glasses. A. Comparison 

between water results from EPMA and FTIR analysis. B. Plot of volatile content versus 

SiO2 content in melt inclusions measured by EPMA, matrix glasses and melt inclusions 

measured by FTIR. ....................................................................................................................... 59 



ix 
 

Figure 17. Plots of volatile content (F, S, Cl) versus water (A, C, E) and SiO2 (B, D, F).. ... 61 

Figure 18. Water content in feldspar crystals. A. Polarized spectra from three 

perpendicular directions in a feldspar crystal. Similar shapes for each direction were 

obtained for oligoclase crystals (An10-30) by Johnson and Rossman (2013). B. Total 

water concentrations measured in three different feldspar crystals. Note that different 

crystals have different water contents, but the water contents of each crystal are 

approximately homogeneous from core to rim (See Appendix F for results from 17 

feldspar crystals). ......................................................................................................................... 62 

Figure 19. Diffusion model of feldspar crystal with no zoning in water (see Matlab results 

in Appendix H). Red lines indicate different magma ascent times to illustrate how 

profiles should be for each scenario. ....................................................................................... 65 

Figure 20. Thermometry plots applying different methodologies. A. Anorthite content vs 

temperature plot calculated using the plagioclase-melt thermometer of Putirka et al. 

(2008). B. Magnesium number vs temperature calculated using the amphibole-melt 

thermometer of Molina et al. (2015). ....................................................................................... 66 

Figure 21. Thermobarometry plots applying different methodologies (see results in 

Appendix G, Table J). Grey regions represent seismicity from 2008-2015 (Londoño, 2016) 

as well as the depths recorded by a CO2 isotope study (Inguaggiato et al., 2017). A. 

Amphibole thermobarometer of Ridolfi (2021).  B. Amphibole thermobarometer of 

Higgins et al. (2022). C. Amphibole-melt thermobarometer of Putirka (2016). D. Water 

content vs saturation pressure plot for 36 melt inclusions. Solubility curves were 

calculated with VolatileCalc (Newman and Lowenstern, 2006) for the minimum and 

maximum temperatures recorded by amphibole crystals, both curves are shown to 

illustrate best-fit to high water melt inclusions. ................................................................... 67 

Figure 22. Whole-rock geochemical plots of the analyzed samples (whole rock chemistry 

from this study: orange triangles, melt inclusions and glass from this study: light orange 

circles and orange circles, respectively). For comparison CMV data available in the 

literature are also plotted (whole rock chemistry: gray triangles, glass: gray circles) 

(Laeger et al., 2013; Régnier et al., 2015). A.Total Alkalis vs. Silica (TAS) diagram (after 

Le Bas et al., 1986); (B-J) variation diagrams of the main major oxides. Arrows indicate 

samples tendencies...................................................................................................................... 72 

Figure 23. Whole-rock geochemical plots of the analyzed samples (whole rock chemistry: 

orange triangles). For comparison CMV data available in the literature are also plotted 

(whole rock chemistry: gray circles). (Laeger et al., 2013; Régnier et al., 2015). (A-L) 

variation diagrams of the trace elements with respect to SiO2. Arrows indicate tendencies 

for samples collected in this sutdy. ......................................................................................... 73 

Figure 24. Multielement diagram normalized to primitive mantle (Sun & McDonough, 

1989). Whole-rock geochemical CMV samples are orange triangles. For comparison CMV 

data available in the literature are also plotted (gray circles). (Laeger et al., 2013; Régnier 

et al., 2015) .................................................................................................................................... 74 

Figure 25. Discrimination diagrams for adakites. Adakite fields are from Defant and 

Drummond (1990) High-silica adakite and Low-silica adakite fields are from Martin et al. 



x 
 

(2005) and Ribeiro et al. (2016). A. Sr/Y versus Y. B. Sr versus SiO2. C. Cr/Ni versus TiO2 

which discriminates high-silica adakites (HSA) from low-silica adakites (LSA). .............. 74 

Figure 26. Experimental constraints for deep pressures in Cerro Machin. A. Phase 

diagram taken from Krawczynski et al. (2012) to show the T-P conditions for amphibole 

and olivine coexistence. B. Microphotograph of high-Fo olivine microphenocryst with a 

overgrowth rim of amphibole (Mg# 80) and microphenocryst of amphibole in CMV 

dacite. ............................................................................................................................................. 79 

Figure 27. Comparison of total H2O in melt from melt inclusions data and plagioclase 

hygrometer using a partition coefficient of 0.002 (Caseres et al., 2018) and a partition 

coefficient of 0.004. Both values agree with melt inclusions water contents. ................. 80 

Figure 28. Composition of olivine microphenocrysts and whole rock chemistry in CMV 

dacites. A. Ni/(Mg/Fe)/1000 versus 100xMn/Fe in olivine crystals. The fields for 

pyroxenite and peridotite are from Sobolev et al. (2007). B. Ni versus Fo content in CMV 

olivines. Blue fields correspond to olivines from 1963-65 Irazu eruption in Straub et al., 

(2011) and in Ruprecht and Plank (2013). Green and red fields are olivines in equilibrium 

with peridotite and pyroxenite source compositions (taken from Ruprecht and Plank, 

2013). C. Whole-rock and olivine FeO/MgO for CMV dacites. Dashed lines are olivine and 

whole-rock compositions predicted to be at equilibrium based on Fe–Mg exchange 

coefficients (Kd’s) of 0.30 (Kd = [FeO*/MgOolivine]/[FeO/MgOmelt]) from Roeder & Emslie 

(1970). D. Rhode’s diagram to illustrate processes that result in deviations from 

equilibrium, as shown by arrows. ............................................................................................. 83 

Figure 29. Samples display the distinctive geochemical features of adakite A. Ba/Nb vs. 

Nb diagram (Bourdon et al., 2002) support the role of slab fluids in adakite genesis. B. 

Th/La versus Th/Nb (Chiaradia, 2009) supports an Altered oceanic crust (AOC) fluid 

input from the slab. Compositions from N-MORB, E-MORB and OIB after Sun and 

McDonough (1989). C. Rb/Y versus Nb/Y (Wang et al., 2008) supports an adakite 

signature from fluid related enrichment. D. Ni versus Cr diagram compare subduction-

related adakite with lower crust-derived adakites (fields shown of Guan et al., 2012) . 84 

Figure 30. Histograms showing abundance of different textures of A. amphibole and B. 

plagioclase. .................................................................................................................................... 87 

Figure 31. A. Summary of the different pressures and depth recorded by amphibole 

crystals, melt inclusions, partition coefficient of water in plagioclase. ............................ 90 

 



xi 
 

List of tables  

Table 1. Plagioclase parameters published in the literature. The absorbance coefficient 

(I) determines how far light of a particular wavelength can penetrate into a material 

before it is absorbed. Diffusion parameter refers to Do is the maximal diffusion 

coefficient at infinite temperature in m2/s; Ea is the activation energy for diffusion in 

kJ/mol and D is the diffusion coefficient value which refers to the amount of a particular 

substance that diffuses across a unit area in 1 s under the influence of a gradient of one 

unit. ................................................................................................................................................. 22 

Table 2. Magma ascent range covered by each tool used in the literature. Advantages and 

disadvantages of each method are also listed as well as the conduit process evidenced 

by each approach. ........................................................................................................................ 26 

Table 3. Summary of melt inclusions data. Saturation pressures calculated by different 

methodologies are also presented. ........................................................................................... 58 

Table 4. Water in feldspars summary. Only average from each profile is shown. ......... 63 

Table 5. Summary of T, P conditions calculated using Higgins et al. (2022), Ridolfi (2021) 

(temperature ± 22°C, pressure ±12% MPa ) and Putirka (2016) (temperature ± 30°C, 

pressure ±200MPa) geothermobarometers. ............................................................................ 70 





1 
 

Chapter 1: Introduction 

Volcanic eruptions are one of the most impactful geologic events on our planet, in fact, 

during the last century, hundreds of minor eruptions occurred across the globe (National 

Academies of Sciences Engineering and Medicine, 2017). Twelve were large enough to 

cause economic and social consequences. Some of these eruptions were explosive, 

violent and their impact was propagated to further distances from the volcano (e.g. 

Hunga Tonga volcano, January 2022). In contrast, others were effusive, gentler but still 

destroyed and affected thousands of people (e.g. La Palma eruption, 2021). 

Despite the variability of eruptive styles, all eruptions are governed by a set of physical 

and chemical processes that respond to the magmatic conditions at which these 

processes occur (National Academies of Sciences Engineering and Medicine, 2017).  

Understanding what controls volcanoes’ behavior requires investigation of the melt 

generation process in the subduction zone, the storage conditions of magma reservoirs, 

the magma compositional evolution in the crust and the magma ascent history from 

source to surface. In each of these stages, water plays a significant role influencing the 

magmatic evolution of the system (Gill, 1981; Plank et al., 2013; Sisson and Grove, 1993).  

Arc magmas are the consequence of mantle melting produced by H2O-rich fluids derived 

from the slab at subduction zones. The presence of H2O lowers the solidus temperature 

of the mantle promoting the ascent of hydrous melts by reactive porous flow in the core 

of the mantle wedge to the shallow mantle lithosphere where they stall and fractionate 

into high SiO2 melts (Annen et al., 2006; Bryant et al., 2011; Collins et al., 2020; Grove et 

al. 2003; Krawczynski et al., 2012; Prouteau & Scaillet, 2003). The amount of water 

involved in these processes is still unknown but estimates up to 10 wt.% have been 

reported on the basis of comparison studies between mineral stability with experimental 



2 
 

products and thermodynamic calculations (Carmichael, 2002, 2004; Grove et al., 2003; 

McCanta et al., 2007).  

At the same time, water has a significant effect on fractional crystallization and 

differentiation paths by controlling the mineral phase stability at any storage condition 

(P-T). Thus, crystals and melt inclusions – small drops of melt trapped inside 

crystallizing mineral phases – constitute a valuable source of evidence for the active 

processes happening in the storage zone and the initial conditions for eruption (e.g. 

Crabtree & Lange, 2011; Blatter & Carmichael, 1998, 2001; Pichavant et al., 2002; Moore 

et al., 1998; Rutherford et al., 1985; Rutherford & Devine, 1988; Scaillet & Evans, 1999;). 

Likewise, the role of water has been reported as the driving force for magma to ascend 

and erupt (Cashman, 2004; Edmonds, 2008; Edmonds & Wallace, 2017; Sparks, 2003; 

Wallace, 2005; Wallace & Anderson, 2000). However, eruptive style seems to be less 

influenced by the initial dissolved volatile content and more controlled by the efficiency 

of volatile exsolution and degassing style, both of which are controlled by magma 

decompression and ascent rate (Barth et al., 2019; Cassidy et al., 2018). In basaltic 

systems, magma decompression rate seems to correlate with eruption magnitude 

(Moussallam et al., 2019) likely due to the influence of magma decompression rate on 

melt-vapor segregation in the conduit. However, few studies have been developed in 

dacitic-rhyolitic systems in which no such correlation is observed (Moussallam et al., 

2019).  

In this study, a petrological investigation of the largest known eruption of Cerro Machin 

volcano in Colombia is presented. Interpretations of melt generation in the subduction 

zone, storage conditions prior to the eruption, triggering mechanisms and degassing 

and ascent history are discussed based on major elements and trace element chemistry, 



3 
 

mineral and mineral-liquid geothermobarometry, volatile content in melt inclusions and 

nominally anhydrous minerals and volatile diffusion modelling in feldspar crystals.  

Cerro Machin volcano is continuously monitored by the Colombian Geological Survey 

(Servicio Geológico Colombiano SGC in Spanish) who reported from 2007 to 2015 an 

increase in the seismic activity, recording a maximum peak of 9000 volcano-tectonic (VT) 

events in 2010 (Londoño et al., 2011), 6500 VT in 2011 and 2000 VT in 2012. Two VT 

earthquakes on the 7th of October 2012 reached magnitudes of 4.7 and 4.1 (Londoño et 

al., 2011). Combined with seismic studies, petrological constraints are a powerful tool 

for understanding magma dynamics, eruption triggering processes and magma ascent 

histories. Geophysics and petrology studies provide meaningful information for the 

correct interpretation of changes detected by monitoring the magmatic system and 

possible onset of magma ascent (Bräuer et al., 2005; Londoño, 2016). These parameters 

comprise invaluable information for hazard assessment studies and represent the input 

data for modeling work.  

Studies have demonstrated that explosive eruptions (Volcanic Explosivity Index or VEI > 

2) are a latent hazard for communities living close to volcanoes with an explosive history 

(Saucedo et al., 2005). In case of eruption, Cerro Machin volcano could affect several 

main cities of Colombia such as Armenia (295,208 inhabitants), Ibague (529,635 

inhabitants) and Cajamarca (17,309 inhabitants) as well as impact important transport 

connections between the capital of Colombia – Bogota, and the southern part of the 

country (Méndez et al., 2002a, 2002b; Murcia et al., 2008). Therefore, understanding the 

dynamics of Cerro Machin’s magmatic system is essential for developing effective 

strategies that could save thousands of people and let the government prepare when an 

eruption occurs.  



4 
 

1.1 Geological background  

Cerro Machin volcano (CMV) is located in the northern segment of the Central Cordillera 

of Colombia (lat 4⁰29’N; long 75⁰23’W; 2750 m.a.s.l.), as the result of the subduction of 

Nazca Plate beneath the South American Plate in an east-west direction (Figure 1A) with 

a velocity of 46 mm/yr and a slab dip of 38 (Hayes et al., 2018, Ha et al., 2020). In the 

Colombia-Ecuador subduction zone, the age reported at the subducted plate trench is 

16 Myr, the thickness of the overriding plate is 40 km (Hayes et al., 2018) and the locus 

of the slab surface below the arc is reported at a depth of 110 km (Wada & Wang, 2009) 

Along with eight other volcanoes – including Nevado del Ruiz volcano, well known for 

the Armero tragedy - Cerro Machin belongs to the Cerro Bravo-Cerro Machín volcanic 

complex. Its edifice is emplaced at the intersection between two major faults, the 

Cajamarca fault which represent a strike-slip fault with vertical displacement and the 

Machín fault interpreted as a normal fault with dextral-lateral displacement (Mosquera 

et al., 1982; Murcia et al., 2008; Rueda, 2005) (Figure 1B).   

Cerro Machin basement comprises the Cajamarca metamorphic complex (Maya & 

González, 1995; Núñez et al., 1980) primarily named as the Cajamarca Group by Nelson 

(1959). It includes para- and orthogneisses, phyllites, quartzites, greenschists, graphitic 

schists and locally marbles (Vargas et al., 2005; Villagómez & Spikings, 2013) affected 

by a low to moderate regional metamorphism – greenschist to amphibolite facies (Cediel 

& Shaw, 2003). The Cajamarca Complex has been described as a Triassic sequence 

intruded by hypabyssal igneous bodies (Ibagué Batholith) with a K/Ar hornblende and 

biotite ages of 150-140 Ma (Brook, 1984; Vesga & Barrero, 1978; Villagómez et al., 2011).  



5 
 

Figure 1. Location of Cerro Machin volcano, Colombia. A. Neotectonic map of Colombia with the 

main fault systems. B. Geological setting of Cerro Machin volcano delimited by Machin and 

Cajamarca faults. Geomorphological features are also shown, symbols taken from (Castilla et 

al., 2021). 

The volcanic edifice is 2750 m of altitude above sea level with a diameter of 2.4 km and 

it has been classified as a pyroclastic ring complex with a dacitic intra-crater lava-dome 

complex (Arango Palacio, 2012; Cepeda et al., 1995).  The volcanic sequence comprises 

at least six major events with ages of >5000, 4500, 3600, 2500, 1200 and 900 yr BP, as 

determined by carbon dating (Cepeda et al., 1995; Méndez et al., 2002a; Rueda, 2005). 

The volcaniclastic deposits from the 5000 yr BP event were named “El Espartillal” by 

Cepeda et al. (1996) and constitute concentrated and diluted pyroclastic density currents 

(PDC) up to 40 m thick. This event was dated in burnt material found in a PDC 10 km 

far from the vent (Méndez et al., 2002b). The second period of activity occurred at 4800-

4300 yr BP, named P0 by Rueda (2005). This section of the deposit is interpreted to have 

been produced by a Plinian eruption with a calculated plume height of 29 km (Rueda, 

2005). The third and most explosive event occurred at 3600 yr BP (called P1 by Rueda, 

2005), and involved a beginning phase of dome disruption followed by an open-conduit 



6 
 

event. The pyroclastic fall deposits related to this event reached up to 60 km from the 

vent (Méndez et al., 2002b; Rueda, 2005). Consequently, several lahars took place along 

the Coello River.  

The following eruptive episodes (2500 and 1200 yr BP - also named Guaico and P2, 

respectively by Rueda (2005) comprised multiple dome collapse events resulting in block 

and ash deposits that traveled up to 35 km from the vent. The last event occurred at 

900 yr BP, also called El Anillo, and led to the emplacement of the lava-dome complex 

occupying the center of the volcanic edifice (Méndez et al., 2002b; Rueda, 2005). Each 

episode was associated with considerable lahar events that traveled more than 100 km 

from the vent.  

This contribution focuses on the most spread deposits of CMV described up to 40 km 

away from the vent (Figure 2). According to previous studies (Méndez et al., 2002a; 

Rueda, 2005), such deposits correspond to the 3600 yr BP eruption or P1 event. Although 

no historical eruptions have been reported for CMV, archeological studies have 

suggested that indigenous communities were strongly affected by the 5100 and 3600 yr 

BP eruptions (Méndez, 1999). Seismic activity in the Cerro Bravo-Cerro Machín Volcanic 

Complex was reported by the Colombian Geological Survey from 2007 to 2015 (Londoño, 

2016) consistent with an active magmatic system with a high probability of new unrest 

or an increase in volcanic activity. Since 2021, CMV activity level remains in III (yellow 

color) which means ‘changes in the volcanic activity in this case related to seismic events 

and fumarolic activity’ (Colombian Geological Service webpage, 2022). 



7 
 

 

Figure 2. Digital elevation model of the outcrop locations (blue dots). Samples from location 
CMV-008 (represented by a yellow star in the map) were used for lab analysis. Isopachs are 

shown in black - thickness in cm – towns, streams and roads in grey. 

1.2 Literature review 

The scientific eagerness to communicate the processes occurring within volcanoes has 

promoted the invention of different terminologies, classifications and definitions that 

seeks to describe accurately volcanic behavior observations (Lockwood & Hazlett, 2010). 

Multiple approaches made in the past century have resulted in different classifications 

based on eruptive style, volume and composition of the erupted material (Gèze, 1964; 

Self & Sparks, 1978; Wood, 1980). Now, eruptions are most frequently described by the 

Walker (1973) classification system – Hawaiian, Strombolian, Vulcanian, Surtseyan, 

SubPlinian, Plinian - as well as by its Volcanic Explosivity Index (or VEI). The latter 



8 
 

proposed by Newhall & Self (1982) correlates the volume of material ejected, eruption 

column height and eruption duration scaling from 0 or non-explosive to 10 or very 

explosive. This section comprises the literature background for the main topics 

discussed in this study. A detailed description of the most common eruption style 

reported in Cerro Machin volcano, along with a description of the typical associated 

deposits is presented. Moreover, a comprehensive summary of the physical and chemical 

processes that occur from the mantle source in subduction zones to the surface are 

discussed for a better understanding of the current document. 

1.2.1 Plinian Eruptions 

Plinian eruptions, named after Pliny the Younger for his accounting of the 79AD 

Vesuvius eruption, represent the most explosive end-member of volcanic eruptions (VEI: 

4-8). Commonly, they are associated with volatile-rich, highly viscous magmas that 

experience high speed discharge forming vertical columns of several tens of kilometers 

of height (Sigurdsson, 2000; Sigurdsson et al., 2015; Sparks, 1986; Sparks & Wilson, 1976; 

Walker, 1973, 1981; Wilson, 1976, 1980; Wilson et al., 1978). Some of the most 

documented andesitic-dacitic Plinian eruptions are the 1980 Mt. St. Helens event in the 

United States (e.g. Carey et al., 1995; Cashman & Blundy, 2000; Cashman & Hoblitt, 2004) 

and the 1991 eruption of Mt. Pinatubo (e.g. Hammer et al., 1999; Koyaguchi & Tokuno, 

1993; Polacci et al., 2001). Examples from the 20th century that include basaltic Plinian 

eruptions are: The 1902 eruption of Santa Maria volcano, Guatemala (Andrews, 2014; 

Rose, 1972; Williams & Self, 1983), the 1932 eruption, Quizapu volcano in Chile 

(Ruprecht & Bachmann, 2010), the 1783 eruption of Asama volcano, Central Japan (Yasui 

& Koyaguchi, 2004), the 1875 eruption of Askja volcano, Iceland (Carey et al., 2010; 

Thorarinsson & Sigvaldason, 1962), the 2008 eruption of Chaiten volcano in Chile 

(Alfano et al., 2011, 2012) and the 1912 eruption of Novarupta volcano (Fierstein & 



9 
 

Hildreth, 1992; Hildreth & Fierstein, 2012) in Alaska. Other well characterized Plinian 

eruptions include: the 7700 yr BP eruption, Mt. Mazama, United States (Klug et al., 2002; 

Suzuki‐Kamata et al., 1993; Young, 1990), the 800 yr BP eruption of Quilotoa volcano in 

Ecuador (Mothes & Hall, 2008; Rosi et al., 2004), the 10.5 ka BP eruption of Nevado de 

Toluca volcano in Mexico (Arce et al., 2003), the ~14.1 ka BP Tutti Frutti (Siebe et al., 

2017; Sosa-Ceballos et al., 2014) and the ~5 ka BP “Ochre Pumice” (Arana-Salinas et al., 

2010) eruptions of Popocatépetl volcano in Mexico, the 1.8 ka BP eruption of Taupo 

volcano (Talbot et al., 1994; Wilson, 1985) (Figure 3).



10 
 

 

 

Figure 3. Spatial distribution of Plinian eruptions over the last 50 kyr. Color shows compositions while size of symbol represents 
VEI. Symbols with more than one color are representing the different compositions of multiple eruptions of the same VEI. Database 

taken from LaMEVE (Large Magnitude Explosive Volcanic Eruptions) of VOGRIPA project.



11 
 

Dynamics controlling Plinian eruptions include magma composition, rheology, volatile 

content and conduit geometry (Fisher & Schmincke, 1984; Papale et al., 1998; Sigurdsson, 

2000). Although they have been defined as a sustained phase, Plinian eruptions typically 

consist in a complex time-evolution of eruption dynamics consisting of a succession of 

phases, sometimes including different eruptive styles (Cioni et al., 2015). In fact, many 

Plinian eruptions begin with vulcanian style phases, also called open-phases or vent-

clearing phases, where the older material is ejected preceding the larger magmatic 

eruption (Lockwood & Hazlett, 2010). Eruptive columns associated with the climax of 

the eruption can be described in three different regions. A lower part is called the gas-

thrust region in which the material is carried upwards by its momentum (Sparks & 

Wilson, 1976). The eruptive jet occurs as a result of incorporated air that expands due 

to heating from the hot pyroclasts and eventually gains a positive buoyancy with respect 

to the ambient atmosphere (Woods, 1988). Above the gas-thrust region, the column 

behaves as a buoyancy-driven convective plume where the thermal energy causes the 

column to continue rising until it reaches the neutral buoyancy height. The ejected 

material is finally propagated laterally from the umbrella region by winds (Figure 4) 

(Cioni et al., 2015). Columns formed by magmas with high volatile content >5 wt.% water 

are likely to show convective motion, whereas those with low volatile content <1 wt.% 

water tend to form collapsing columns (Cas & Wright, 1987). For magmas with 

intermediate content of volatiles, collapse occurs when the vent radius exceeds a certain 

value defined by Wilson (1976).  



12 
 

 

Figure 4. Different regions of a typical Plinian-sub Plinian eruptive column (modified from Cioni 
et al., 2015). 

During Plinian eruptions large volumes of material are ejected forming widespread wind-

controlled fall deposits (Cas & Wright, 1987) that can be used as marker horizons for 

tephrochronology (Lowe et al., 2008; Thorarinsson, 1981; Walker, 1981). In addition, 

tephra deposits can reflect the complex variability of the eruptive regimes and the 

multiple pulses that occurred during the eruption. This enables the reconstruction of 

the plume dynamics (Cioni et al., 2003) and the assessment of eruptive parameters such 

as volume, mass discharge rate and column height (Carey & Sparks, 1986; Pyle, 1989). 

1.2.1.1 Associated deposits  

Plinian deposits cover a dispersal area greater than 1000 km2 inside the 0.01 Tmax isopach, 

where Tmax is the maximum thickness of the deposit (Walker, 1973). The study of these 

has demonstrated to be a valuable source for calculations of volume, mass discharge 

rate, column height, wind velocities, eruption intensity and magnitude (Fierstein & 



13 
 

Nathenson, 1992; Pyle, 1989). This is how several models of grain size distribution and 

dispersal patterns have been developed in the literature deposits (Carey & Sparks, 1986; 

Settle, 1978; Walker, 1981; Wilson et al., 1978; Wilson & Walker, 1987). 

The systematic variations with distance from the vent of parameters such as thickness 

and grain size follow a simple exponential decay law -or power law- and are controlled 

by plume dynamics and wind field at the time of the eruption (Bonadonna & Costa, 2012; 

Bonadonna & Houghton, 2005; Pyle, 1989). For instance, the deposition of large clasts 

that follow ballistic trajectories in proximal areas yield important information about exit 

velocities and explosion pressures in the gas-thrust region (Wilson, 1976; Wilson et al., 

1978). On the other hand, finer fragments can provide information about the column 

height and eruption intensity since these fragments are transported laterally in the 

umbrella region and are subsequently accumulated in distal areas based on their 

terminal fall velocity (Carey & Sparks, 1986; Wilson, 1980; Wilson et al., 1978).   

Proximal deposits 

In proximal areas (5-10 km from the vent), deposits result from the rapid accumulation 

from the gas-thrust regions and from the margins of the convective column. They can 

present certain complexities such as 1) recycling of clasts falling from the umbrella 

region into the rising column, 2) erosion or overthickening of proximal fall deposits by  

pyroclastic density currents or PDCs, 3) simultaneous deposition of material from 

different regions of the eruption column, 4) poor preservation of the deposits due to the 

steepness of the volcano edifice or collapse of the volcanic edifice, 6) interbedding 

between PDCs separate fall units (Cas & Wright, 1987; Cioni et al., 2015; Self et al., 1984). 



14 
 

Medial deposits 

Depending on the size of the eruption, medial deposits can cover up to 40-50 km. They 

are well preserved and represent the typical pyroclastic fall sequence of lapilli layers. 

Based on their internal configurations, they provide information about changes in 

magma discharge, column height and eruptive column steadiness (Adams et al., 2006; 

Alfano et al., 2011; Carey et al., 1995; Cole et al., 1995; Jurado-Chichay & Walker, 2001; 

Taddeucci & Wohletz, 2001).  

Massive, normal or reversely graded deposits are the result of steady eruptive columns 

(Houghton & Carey, 2015; Williams & Self, 1983) with changes in mass discharge rate 

(MDR). They are classified as: Simple Plinian deposits (Rosi, 1996) in which multiple-

graded layers are the result of the gradual variation of the column height. On the other 

hand, under column instability, it is common to observe internal bedded layers within 

the sequence these deposits are categorized as Simple-stratified Plinian deposits by Rosi 

(1996). Occasional partial column collapses can result in transitions between convective 

and collapsing phases (i.e. interlaying between ash fallout deposits and PDCs) (Self et al., 

1984). PDCs can be absent within the depositional sequence but represented by fine-

grain size layers which represent the co-PDC ash dominated by a turbulence regime 

(Brown & Andrews, 2015). It is important to notice that compositional vertical changes 

do not vary with distance and help to correlate among different outcrops (Cioni et al., 

2015).   

Distal deposits 

Distal deposits cover from 50 km up to several hundreds of km from the vent. They are 

represented by very fine-grain size deposits and are very poorly preserved in the 



15 
 

geological record. This means that only recent eruptions contain a complete description 

for these deposits (Sigurdsson, 2000). 

SubPlinian deposits 

SubPlinian eruptions are characterized by lower intensity and unsteadiness in magma 

discharge which result in strongly oscillatory eruptive columns (Bursik, 1993; Self, 1976; 

Sieh & Bursik, 1986). Therefore, the associated pyroclastic fall deposits show a strong 

stratification (Cioni et al., 2003). In contrast to Plinian eruption deposits, they tend to be 

thinner and to show greater variability in the juvenile component. 

Componentry 

The most common attribute of Plinian eruption deposits is the presence of juvenile 

highly vesiculated material (i.e. pumice or scoria) (Fisher & Schmincke, 1984).  However, 

accessory and accidental lithics derived from the conduit wall can also be present (Cas 

& Wright, 1987; Németh & Martin, 2007). A certain number of eruptions show evidence 

of compositional zoning or juvenile material represented by streaky mixed pumice clasts 

suggesting mechanical mixing before the eruption (Martí et al., 2010; Sparks et al., 1981). 

Componentry vertical variations are evidence of changes in the vent such as widening 

of the conduit (i.e. promoting erosion rate and therefore increasing the proportion of 

accessory lithics), changes in the position of the fragmentation level (i.e. changes in 

accidental lithic composition) or sampling different parts of the magma chamber. 

Magma fragmentation mechanisms  

As magma rises from depth, the decrease in pressure causes volatiles to exsolve by 

nucleation of new bubbles and to grow due to diffusion into the existing bubbles (Cas & 

Wright, 1987; Gonnermann, 2015; Gonnermann & Manga, 2007). The process in which 



16 
 

the continuous molten phase with disperse bubbles breaks into a gas phase with 

dispersed magma fragments is called fragmentation (Cashman & Scheu, 2015; 

Gonnermann, 2015). Fragmentation of magma can occur by two different mechanisms 

that can be interpreted based on the abundance and shapes of the juvenile ash particles 

in the deposits (Heiken, 1972, Wohletz & Krinsley, 1982, Dellino et al., 2001).  

Magmatic fragmentation mechanism is driven by the exsolution and expansion of gases 

originally dissolved in the molten phase (Blake, 1984; Jellinek & DePaolo, 2003; McLeod 

& Tait, 1999). In this case, fragmentation can occur by rapid acceleration of magma to 

the surface due to overpressure within the magma chamber. This overpressure can be 

the result of new and hot magma input or because of volatile enrichment induced by 

cooling and crystallization processes (Cas & Wright, 1987; Wallace, 2001). Rapid 

decompression caused by collapse of the edifice or lava domes can also represent a 

potential fragmentation trigger (Gonnermann & Manga, 2007; Lipman & Mullineaux, 

1981). Deposits from magmatic eruptions tend to be rich in juvenile material (Cas & 

Wright, 1987).  

Phreatomagmatic mechanism is generated by the physical interaction between the rising 

magma and external water (Wohletz, 1983). The contact between these two phases 

creates a vapor-film that expands and collapse promoting the melt fragmentation 

(Lorenz, 1985, 1987; Sheridan & Wohletz, 1981, 1983; Wohletz & Sheridan, 1983). The 

continuous energy and heating transfer facilitate the formation of a new vapor-film 

which generates an effective chain-reaction (Németh & Martin, 2007; Wohletz et al., 2013; 

Zimanowski et al., 2015). When magma heats surrounding fluids without any direct 

contact, it is called phreatic or hydrothermal fragmentation (Cas & Wright, 1987; 

Sigurdsson et al., 2015). Dense lithic fragments tend to be dominant in the associated 

deposits (Cas & Wright, 1987).  



17 
 

1.2.2 The role of water in melting processes  

The generally accepted model for melt generation in subduction zones involves partial 

melting in the mantle wedge by the addition of hydrous fluids released from the 

subducted slab (e.g. Annen et al., 2006; Davies & Stevenson, 1992; Forneris & Holloway, 

2003; Grove et al., 2002; Schmidt & Poli, 1998; Ulmer, 2001). Experimental studies and 

numerical modeling suggest that primary melt compositions range from basalt to high 

magnesium andesite (Conrey et al., 1997; Grove et al., 2002) depending on the amount 

of magmatic water involved in the melting process (Grove et al., 2012). On the other 

hand, silica-rich counterparts have been explained by: a) partial melting of harzburgite 

in the mantle by water released from the subducted slab (Carmichael, 2004; Parman & 

Grove, 2004), b) crystallization of mantle-derived basaltic or basaltic andesitic melts at 

shallow crustal pressures (Sisson & Grove, 1993), c) crystallization of mantle-derived 

basaltic or basaltic andesitic melts at high pressures close to the Moho boundary 

(Prouteau & Scaillet, 2003), d) partial melting of metabasalts in the lower-middle crust 

promoted by intrusions of hot mantle derived magmas (Rushmer 1991; Sen & Dunn 

1994; Rapp & Watson 1995; Petford & Atherton, 1996) and e) mixing between mafic and 

silicic magmas – including assimilation of silicic crustal rocks (Grove & Kinzler 1986; 

Hildreth & Moorbath, 1988).  

The adakite term was introduced in the literature by Defant & Drummond (1990) to 

describe melts produced by melting young subducted oceanic slab (< 25Myr) – hot 

oceanic crust (Kay, 1978). The rock is characterized by high Al2O3 (>15 wt %), high Sr/Y 

(>40), Sr (>400 ppm), low-Y (<18 ppm), Yb <1.9 ppm, La/Yb > 20, and a lack of Eu 

anomalies (Castillo, 2006, 2012; Defant & Drummond, 1990). However, the source of 

adakite signature in magmas has been highly debated as they can result from multiple 

models besides the initially proposed slab melting environment (Castillo et al., 2006 and 



18 
 

references therein). Some occurrences of adakites in cold subduction settings have been 

explained by melting of thickened mafic lower crust or crystal fractionation of wet 

basaltic arc magmas (Atherton & Petford, 1993; Castillo, 2006; Macpherson et al., 2006; 

Prouteau & Scaillet, 2003; Ribeiro et al., 2016; Rodriguez et al., 2007; Sajona et al., 1993; 

Yogodzinski & Kelemen, 1998). This topic is still controversial and remains in 

discussion. 

The presence of high amounts of water (> 10 wt.%) in the mantle wedge has been 

suggested as the controlling factor to generate SiO2-rich primitive arc magmas (57-60 

wt.% SiO2) (Grove et al., 2003; Grove et al., 2012). Water is capable of lowering the mantle 

liquidus temperature, which increases the mineral-melt partition coefficient D for MgO 

as the temperature of the silicate liquid drops. In contrast, D for SiO2 remains invariant, 

favoring the SiO2 enrichment of water-rich mantle peridotite melts. These water-rich 

melts are in equilibrium with the mantle residue before ascending into the overlying 

crust where they undergo fractional crystallization at water-saturated conditions. 

During cooling and crystallization, magmas produce an aqueous phase which infiltrates 

the surrounding crust and promotes crustal melting in the base of the arc crust (Collins 

et al., 2020).  

Although the presence of primitive superhydrous melts (>8 wt. H2O) beneath some arcs 

has been inferred on the basis of volatile measurements in lower-crustal cumulates 

(Urann et al., 2022) and mafic enclaves (Goltz et al., 2020), experimental phase equilibria 

studies (Camichael et al., 2002; Conrad et al., 1988; Grove et al., 2005; Krawczynski et 

al., 2012; Lu et al., 2015; McCanta et al., 2007; Prouteau & Scaillet, 2003), mineral 

barometry coupled with water saturation pressures, and microprobe measurements in 

melt inclusions (Grove et al., 2003), the maximum amount of dissolved water remains 

unknown. Volatile concentration measurements in melt inclusions persist as the main 



19 
 

tool to explore water abundance in primary arc magmas, but another promising 

technique come from the analysis of water in nominally anhydrous minerals (Urann et 

al., 2020). 

1.2.3 Magmatic water at storage conditions  

1.2.3.1 Volatile concentration in melt inclusions  

Melt inclusions are defined as tiny samples of undegassed melt trapped inside of 

phenocrysts (Wallace, 2005). They record significant information about the geochemical 

cycles of the major volatiles (H, C, S and Cl) before eruptions as they are physically 

isolated from external processes. Over the last few decades, a great set of studies have 

determined H2O and CO2 contents in melt inclusions of subduction-related magmas 

(Lowenstern, 1995; Lowenstern, 2003; Rose-Koga et al., 2021, and references therein). 

Water and CO2 concentrations in melt inclusions can reflect magma degassing history 

through the crust recording the entrapment pressures and tracking the volatile 

saturation evolution of magmas (Atlas et al., 2006; Bali et al., 2018; Hartley et al., 2018; 

Johnson et al., 2008; Rust et al., 2004; Spilliaert et al., 2006). It is important to notice 

that post-entrapment processes such as crystallization or diffusive exchange can modify 

the composition of melt inclusions after they are trapped in the crystal host (Bucholz et 

al., 2013; Danyushevsky et al., 2000; Gaetani & Watson, 2002; Wallace, 2005). In addition, 

thermal contraction can result in the formation of vapor bubbles that deplete the melt 

in CO2 due to its low solubility (Anderson & Brown, 1993; Cervantes et al., 2002). 

1.2.3.2 Water in Nominally Anhydrous Minerals  

Nominally Anhydrous Minerals or NAM refer to those minerals whose chemical formula 

can be written without any hydrogen, but they contain trace to minor amounts of 



20 
 

hydrous components (Bell & Rossman, 1992; Keppler & Smyth, 2006; Solomon & 

Rossman, 1988). Hydrogen is incorporated as defects in NAMs and it can be studied by 

diffraction methods and spectroscopic methods (IR and Raman spectroscopy). The latter 

relies on the idea of molecular bond vibration in which O-H vibrations and their 

absorption bands are easily assigned in the high energy part of the spectrum between 

3000-4000 cm-1 (Libowitzky & Beran, 2006). 

NAMs have been considered as a promising approach to estimate H2O contents of arc 

magmas (Urann et al., 2022). Recent studies have suggested the presence of 

superhydrous magmas at depth (> 8 wt.% H2O) (Goltz et al., 2020; Krawczynski et al., 

2012; Laumonier et al., 2017) which have not been recorded by melt inclusion 

measurements due to re-equilibration processes (Plank et al., 2013) or quenchability 

limitations (Gavrilenko et al., 2019). Using experimental water partition coefficients, 

NAMs can be used as potential hygrometers to estimate water contents in the host 

magmas (Urann et al., 2022) when water diffusion has not taken place. 

Similarly, NAMs have been proposed as potential chronometers from processes 

happening in a relative short-time window before the eruption. As magmas degas and 

ascend to the surface, the solubility of water decreases resulting in water loss through 

the crystals by diffusion. Assuming a degassed magma, NAMs are expected to contain 

water concentration gradients from core to rim that can be used to calculate syn-eruptive 

magma decompression and ascent rates (Demouchy et al., 2006; Newcombe et al., 2020; 

Peslier et al., 2015; Peslier & Luhr, 2006). This approach has been successfully applied 

in olivine crystals, a common phase present in basaltic and intermediate systems. 

However, this mineral phase is exceptionally rare in dacitic-rhyolitic systems which 

makes it valuable to test other mineral phases such as plagioclase and quartz as 

potential chronometers for silica-rich systems.    



21 
 

H in feldspars 

Volcanic feldspars contain OH as the structurally incorporated hydrous species (Johnson 

& Rossman, 2004). Pioneering studies of water concentration in feldspar using FTIR 

(Johnson & Rossman, 2003) indicate that for quantitative H concentration 

determinations, it is necessary to measure the total integrated absorbance in three 

mutually perpendicular polarized IR spectra. In order to calculate H concentrations, the 

Beer-Lambert law is applied in which the absorbance coefficient has been reported and 

updated by different authors (Table 1). Reported concentrations of water in feldspar are 

up to hundreds of wt. ppm H2O (Caseres et al., 2018; Hamada et al., 2011; Johnson & 

Rossman, 2004; Seaman et al., 2006) and have been used to constrain water content in 

their host melts (Caseres et al., 2017, 2018; Hamada et al., 2013). Available partition 

coefficient values between plagioclase and melts comes from experimental studies 

0.0025 – for melts with > 3 wt% - to 0.005 - for melts with < 3 wt% (Hamada et al., 2013; 

Mosenfelder et al., 2015). In dacitic-rhyolitic systems such as Mt. Saint Helens (Johnson, 

2005), Mt. Hood (Caseres et al., 2018), Mt. Mazama (Mosenfelder et al., 2018) and 

Huckleberry Ridge Tuff, Yellowstone (Rappoccio et al., 2020) a partition coefficient of 

0.002 was estimated. 

The solubility of water in volcanic feldspar at different temperatures and oxygen 

fugacity has been explored in a wide range of compositions (Mosenfelder et al., 2020; 

Yang, 2012). Reported results from experimental studies show how solubility of water 

has a strong dependence on fO2 for a variety of compositions. Solubility experiments at 

800°C and NNO condition – similar conditions to dacitic and rhyolitic magmas - show no 

statistically meaningful effect of plagioclase composition on solubility.  



22 
 

Sample type must be considered when interpreting water concentration data. As 

demonstrated in Hamada et al. (2011), different types of samples - scoria, lava flow 

samples and bombs - can show water concentration variations which can be explained 

by different cooling rates where rapidly quenched samples such as scoria samples retain 

the highest recorded water content (Mosenfelder et al., 2020). 

Table 1. Plagioclase parameters published in the literature. The absorbance coefficient (I) 
determines how far light of a particular wavelength can penetrate into a material before it is 

absorbed. Diffusion parameter refers to Do is the maximal diffusion coefficient at infinite 
temperature in m2/s; Ea is the activation energy for diffusion in kJ/mol and D is the diffusion 
coefficient value which refers to the amount of a particular substance that diffuses across a 

unit area in 1 s under the influence of a gradient of one unit.  

Absorbance coefficient Assessment method Author 

I = 15.3 ± 0.7 ppm–1·cm–2 

I = 107000 ± 5000 L·mol–1 

H2O cm–2 

The slope of a best-fit line 

through the concentration 

of H for feldspars with 

different H concentrations. 

Johnson & Rossman (2003) 

I = 202600 ± 20260 L·mol–1 

H2O cm–2 

Comparison between FTIR 

and SIMS from several 

samples 

Mosenfelder et al. (2015) 

Diffusion parameter Assessment method Author 

log D0=-1.62 ± 0.31 (m/s2) 

and Ea= 266 ± 77 kJ/mol 

log D0=-0.97 ± 0.35 (m/s2) 

and Ea= 278 ± 90 kJ/mol 

D = 3.7465E-15 m2/s at 

812°C 

 

Dehydration experiments 

assuming a spherical 

plagioclase of 1mm at 

different temperatures 

(800,900, 1000C) 

Johnson & Rossman (2013) 

 

Water diffusion experiments performed in plagioclase at different temperatures (800-

1000°C) show that hydrogen diffusion in feldspar is isotropic. The diffusion rates of 

hydrogen and sodium in plagioclase are very similar suggesting that H+ diffusion occurs 

as H+-Na+ interdiffusion on large cation site vacancies, via Frenkel defects (Figure 5). 



23 
 

Based on hydrogen diffusion coefficient, a 1 mm spherical plagioclase during ascent and 

eruption can retain 50% of its water over periods of hours to days (Johnson & Rossman, 

2013). The diffusion parameters published in the literature are listed in Table 1. 

 

Figure 5. Frenkel defects in the NaCl structure. It occurs when an atom leaves its normal site 
vacant and instead occupies an interstitial site (Aschcroft & Mermin, 1976). 

Water in volcanic feldspars has been explored in few volcanic systems despite its 

potential to be used as a chronometer (i.e., useful for constraining magma 

decompression rates based on volatile diffusion) or as a hygrometer (i.e., useful for 

constraining pre-eruptive magmatic water concentrations). In this study, we explored 

water behavior in plagioclase crystals from the largest known eruption of Cerro Machin 

volcano as a potential constraint for magma storage depths and magma migration 

through the crust. 

1.2.4 Water behavior during magma-ascent decompression 

Magmas are defined as silicate melts that contain solids – crystals and gases – volatiles 

(Fisher & Schmincke, 1984). According to measurements of volcanic gasses emitted at 

volcanoes, these volatiles are dominantly H2O, followed by CO2 and SO2. Minor amounts 

of H2O, CO, HCl and HF have also been detected (Burgisser & Degruyter, 2015). Although 

the amount of volatiles is only a few weight percent, they play an important role in 

triggering eruptions (Edmonds, 2008; Sparks, 2003; Woods & Huppert, 2003). The 

solubility of these volatile species is pressure-dependent which means that once the 



24 
 

pressure drops below the saturation pressure they can exsolve into a gas volatile phase 

(Dixon et al., 1995; Klug & Cashman, 1996; Moore et al., 1998; Newman & Lowenstern, 

2002). During magma-ascent decompression, magmas become volatile-saturated which 

results in the formation of bubbles (first boiling) that grow and expand for volatile 

diffusion (Edmonds, 2008; Edmonds & Wallace, 2017; Wallace et al., 2015). Volatile 

exsolution can also occur as a result of isobaric cooling and crystallization (second 

boiling) which increases the reservoir pressure (Blundy & Cashman, 2001; Cashman & 

Blundy, 2000; Christopher et al., 2015).  

Bubbles can form by homogeneous nucleation -spontaneous bubble formation- or by 

heterogeneous nucleation – crystals acting as nucleation sites (Gardner & Denis, 2004; 

Mangan & Sisson, 2000; Toramaru, 1989). Once bubbles are formed, the solubility and 

diffusivity processes govern the degassing process during ascent-driven decompression 

(Gonnermann & Manga, 2007). Degassing of different volatile species vary with melt 

composition, temperature, pressure and oxygen fugacity (Dixon & Stolper, 1995; 

Hamilton et al., 1964; Newman & Lowenstern, 2002).  

Magma ascent and eruptive style are strongly related to magma degassing behavior 

(Cassidy et al., 2018). During fast ascent and rapid decompression rates, disequilibrium 

degassing is promoted since volatiles remain coupled to the melt increasing buoyancy 

and overpressure - closed-system degassing (Cassidy et al., 2018; Johnson et al., 2008; 

Métrich & Wallace, 2008; Sparks, 2003; Stasiuk et al., 1996). In contrast, during slow 

ascent and low decompression rates volatiles are decoupled from the melt allowing 

bubbles to migrate from deeper to shallower levels. This is referred as open-system 

degassing or equilibrium degassing in which permeable paths formed by the coalescence 

of bubbles, enable outgassing which reduces buoyancy and overpressure (Burgisser & 



25 
 

Degruyter, 2015; Cassidy et al., 2018; Gonnermann & Manga, 2007; Hammer et al., 1999; 

Yoshimura & Nakamura, 2008).  

Estimation on magma decompression rates of different types of eruption (effusive vs 

explosive) is a powerful tool that can help to understand the physical processes that 

influence degassing regime and eruptive style of volcanoes. In basaltic systems magma 

decompression rate correlates with eruption magnitude (Moussallam et al., 2019), 

however, the lack of this relation in more silicic and higher magnitude eruptions need 

to be addressed from the study of magma ascent rates of more silicic systems.  

1.2.4.1 Tools for estimation of magma ascent rates  

Different techniques have been performed to constrain the rate at which magma 

decompresses: 1) experimental studies of amphibole breakdown rims (McCanta et al., 

2007; Rutherford & Devine, 2003; Rutherford & Hill, 1993), 2) microlite number density 

(Andrews, 2014; Cassidy et al., 2015; Castro & Gardner, 2008; Couch et al., 2003; 

Hammer & Rutherford, 2002; Martel, 2012; Miwa et al., 2009; Noguchi et al., 2008; Sano 

et al., 2015; Scott et al., 2012; Toramaru et al., 2008; Wright et al., 2012), 3) bubble 

number density (Carey et al., 2010; Castro & Gardner, 2008; Giachetti et al., 2010; Myers 

et al., 2021; Nguyen et al., 2014; Shea et al., 2010; Toramaru, 2006), 4) feldspar 

composition (Scarlato et al., 2017) however, this relation is not always met (Brugger & 

Hammer, 2010) 5) diffusion of lithium in plagioclase and quartz (Charlier et al., 2012; 

Neukampf et al., 2021), 6) diffusion of volatiles in apatite (Li et al., 2020), 6) diffusion of 

volatiles in melt embayments (Anderson, 1991; Humphreys et al., 2008; Liu et al., 2007; 

Lloyd et al., 2014; Myers et al., 2016, 2018, 2021) 7) diffusion modelling of water loss 

from melt inclusions into their olivine-host crystal (Barth et al., 2019) and 8) diffusion 

of water loss out of nominally anhydrous minerals (Newcombe et al., 2020). In Table 2 



26 
 

the range of ascent rates covered by each technique and the processes reflected by each 

one are shown. 

Table 2. Magma ascent range covered by each tool used in the literature. Advantages and 
disadvantages of each method are also listed as well as the conduit process evidenced by each 

approach. 

Approach 

Ascent 
rate 

range 
(m/s) 

Advantage Disadvantage 
Conduit 
process 

Amphibole 
breakdown rims 

0.0001- 1 

Thin sections are 
easy to prepare. 

Rim width can be 
measure in Back 
Scattered Eletron 

images (BSE) 

Experimental 
data for rim 

development is 
require for 

specific 
magma 

Reflect 
decompression 
from storage 
conditions to 
the surface 

Microlite/Crystal 
number density 

(CND) 

0.001– 
100 

Thin sections are 
easy to prepare. 
Software have 

been designed to 
determine CND 

on back scattered 
electron images 

Data of 
storage depth, 
temperature 
and water 
content is 

required. CND 
cannot resolve 

ascent rate 
without 

independent 
knowledge of 
crystal growth 

rate 

Reflect 
decompression-

induced 
crystallization  

Bubble number 
density 

0.001 - 10 

Shea et al. (2010) 
provide FOAMS 

software to 
calculate magma 
decompression 

rate based on BSE 
images. 

Big difference 
if 

heterogeneous 
or 

homogeneous 
nucleation is 

assumed 

Reflect 
nucleation 

peaks 
commonly 

occurring in the 
upper part of 
the conduit  

Feldspar 
composition 

0.01 – 0.1 

Feldspar is a 
common mineral 
phase in volcanic 

products 

Crystals 
formed at 

equilibrium 
conditions are 
required. The 

relation 
between 

decompression 

Reflect 
decompression 
from storage 
conditions to 
the surface 



27 
 

Approach 

Ascent 
rate 

range 
(m/s) 

Advantage Disadvantage 
Conduit 
process 

rate and 
composition is 
not always met 

Diffusion of 
lithium in 

plagioclase and 
quartz 

1.5 - 5 

Capture syn-
eruptive 

processes on very 
short timescales 

Relation 
between Li 

diffusivity and 
pressure 

changes is not 
totally 

understood 

Reflect late-
stage ascent 
processes 

Diffusion of 
volatiles in 

apatite 

0.01 – 
0.001 

An online tool for 
apatite diffusion 
modelling and 

estimating 
magma ascent 

times is available 

A weakness of 
the model is 

the 
assumption of 

a fixed 
boundary 
condition 

(composition 
of crystal rim) 
which can vary 
according to 

Cl-F-OH 
changes of the 
melt. Also, a 
decrease in 
temperature 
during the 

ascent could 
lead to slower 
diffusion rates.  

Zoned 
groundmass 

apatite reflects 
magma ascent 
times of days 

before eruption 

Diffusion of 
volatiles in melt 

embayments 
0.01 – 10 

Diffusion 
coefficients in 
different melt 

composition are 
well constrained 

Melt 
embayments 

tend to be 
rare. Difficult 

sample 
preparation  

Deep conduit. 
It can reflect 

changes in the 
conduit 

geometry: 
widening 
and/or 

lengthening. 

Diffusion 
modelling of 

water loss from 

0.01 - 1  
The method is in 
good agreement 

with the melt 

Useful for 
basaltic- 
andesitic 

Reflects 
syneruptive 

magma 



28 
 

Approach 

Ascent 
rate 

range 
(m/s) 

Advantage Disadvantage 
Conduit 
process 

melt inclusions 
into their 

olivine-host 
crystal 

embayment 
approach. 

volcanic 
products 

containing 
olivine 

decompression 
rates 

Diffusion of 
water loss in 

nominally 
anhydrous 
minerals 

0.01 - 1  

Olivine is a 
mineral phase 

easy to identify. 
The method is in 
good agreement 

with the melt 
embayment and 

MI water loss 
approaches. 

Water 
diffusion in 

olivine is 
anisotropic 

Reflects syn-
eruptive 
magma 

decompression 
rates 

 

  



29 
 

Chapter 2: Motivation 

A major goal of volcanology is to be able to anticipate volcanic eruptions in order to 

prevent loss of life and to mitigate economic impacts. This requires that volcanologists 

develop models of volcanic systems that integrate data and methodologies from 

multiple disciplines (National Academies of Sciences Engineering and Medicine, 2017). 

Several studies have successfully explained how and why volcanoes erupt; nevertheless, 

some challenges remain related to the physical-chemical processes that control volcanic 

explosivity.  

Water represents the essential component in every stage of arc magmatism. It has a 

major effect in melt generation (Grove et al., 2012), it controls magma crystallization 

phases and differentiation (Grove et al., 2003) and it provides the fuel for explosive 

eruptions (Cashman, 2004).  Over the last few decades, studies of water in melt 

inclusions have provided significant information about the amount of water dissolved 

in magma reservoirs before eruptions, and along with CO2 measurements, have given 

insights of the storage pressures of magma chambers (Wallace, 2005).  However, the 

pressure solubility dependency of water as well as the multiple port-entrapment 

processes occurring at surface constitute a challenge to quantify the maximum amount 

of water dissolved in subduction-related magmas when they last equilibrate with the 

mantle (Grove et al., 2012). These challenges are still a very active research area (Rose-

Koga et al., 2021).   

Similarly, the role of volatiles on the bubble exsolution rate and degassing style has been 

highlighted as an important controlling factor for the eruptive style (Barth et al., 2019; 

Cassidy et al., 2018). Both processes are governed by the magma decompression rate 

which seems to be poorly constrained in the literature (Barth et al., 2019), especially for 



30 
 

dacitic magmas. According to Moussallam et al. (2019), magma decompression rates of 

felsic eruptions show an unorganized pattern compared to basaltic systems. The lack of 

a consistent trend underlines the importance of studying magma decompression from 

more dacite eruptions to identify what controls eruptive style in these systems. Cerro 

Machin volcano in Colombia represents a good example to explore with at least six 

dacitic Plinian eruptions occurring in the last 10000 years BP, and no volatile data 

published in the literature. Our approach for estimating magma decompression rates in 

dacitic eruptions involves the application of a volatile diffusion model based on water 

concentration gradients in plagioclase, a nominally anhydrous mineral (NAM) and a 

common mineral phase from any volcanic product. 

As confirmed by Newcombe et al. (2020), NAMs represent an important insight to study 

water behavior during magmatic processes, nevertheless, there are hardly any 

measurements of water in volcanic feldspar phenocrysts (Figure 6) (Hamada et al., 2013; 

Johnson & Rossman, 2003, 2004; Mosenfelder et al., 2020; Shepherd & Johnson, 2016). 

One complexity of feldspar crystals is that concentration of water must be estimated by 

the integrated absorbance of three mutually perpendicular directions. Authors from 

previous studies have polished four orthogonal faces of one plagioclase block, losing 

information from the crystal edges of the polished faces. In this study, a different sample 

preparation method is explored in which the crystals are cut in half and subsequently 

only two faces from each half are polished. With this method we will be able to measure 

water concentration from the core to the crystal edge which is in contact with the host 

magma.  



31 
 

 

Figure 6. Compilation of published water content in feldspar crystals by (Mosenfelder et al., 
2020). 

Loss heating experiments in plagioclase have been conducted in the past by Johnson & 

Rossman (2013) who calculated the hydrogen diffusivity in natural OH-bearing 

plagioclase feldspar (Ab66An31Or3, with 510 ppm H2O). However, natural volcanic feldspar 

are commonly zoned presenting chemical variations in very high-resolution bands. A 

recent study by Behrens (2021) found further complexities associated with the diffusion 

of water in feldspar: the diffusivity of water in both plagioclase and alkali feldspar was 

observed to slow down over the course of their experiments. Clearly there is a need for 

further exploration of water diffusion in feldspar and such experiments represent an 

important avenue of future investigation. In particular, it will be important to conduct 

dehydration experiments in natural volcanic feldspars in order to assess the effect of 

the plagioclase major element zoning on the diffusivity of H through the crystal. This 

would help us to understand if water gradients in feldspar respond to the chemical 

zoning of the crystals or in contrast they are intimately related to degassing and magma 

ascent processes.   



32 
 

2.1 Colombian perspective  

Colombia has more than 20 active volcanoes (Figure 7A), which are permanently 

monitored by the Colombian Geological Survey (Servicio Geológico Colombiano, SGC) 

(Gomez et al., 2021). However, Colombia not always had the monitoring network they 

have today. The disasters of Doña Juana (1899, 1906), Puracé (1949), Galeras (1993) and 

Nevado del Ruiz (1545, 1845, 1985) volcanoes are clear examples of the importance of 

studying volcanoes to prevent future tragedies (Figure 7B). According to the geological 

record, volcanoes such as Animas, Azufral and Cerro Machin have had several Plinian 

eruptions. The severe implications of Plinian events make it very important to 

understand the processes that trigger this type of explosive activity and how it can be 

potentially related to the seismological and geochemical signature recorded by the 

volcanic monitoring. In particular, Cerro Machin is considered one of the most explosive 

volcanoes in the country with a very explosive geological record that include at least six 

Plinian eruptions (Cortes et al., 2006). Moreover, in case of eruption Cerro Machin would 

affect one of the most populated areas of the country as well as the only land transport 

connection between the capital of Colombia – Bogota- and the south of the country 

(Gomez et al., 2021).  



33 
 

 

Figure 7. Geological context of active Colombian volcanoes. A. Map of Colombia with the 
different active volcanoes. B. Most tragic eruptions in Colombia, each bar represents one 

eruption from a specific volcano and the associated number of deaths. 

In this study, a petrological investigation of the largest known eruption of Cerro Machin 

volcano is presented. Bulk rock composition of the juvenile material together with major 

element, trace element and volatile content in melt inclusions hosted in plagioclase and 

quartz crystals, provide the key criteria for the interpretation of melt generation 

processes happening beneath CMV. In addition, we reconstruct the storage conditions, 

triggering mechanism and degassing history of this eruption based on mineral and 

mineral-liquid geothermobarometry, volatile content in melt inclusions and NAMs and 

water diffusion modeling in feldspar. This new contribution will be one of the inputs 

that can be incorporated into the studies of volcanic hazards of Cerro Machin volcano, 

which are very useful to the responsible government entities for the vulnerability and 

risk studies. 

Therefore, the politicians and decision makers can carry out proposals for integral 

planning of the Colombian territory and also create adaptation strategies and resilience 



34 
 

mechanisms for the communities who live in the areas of influence of this volcano (Vega 

& Diaz, 2013). 

2.2 Statement of purpose 

This project aims to constrain the pre-eruptive conditions that influenced the largest 

known Plinian eruption of Cerro Machin volcano, Colombia in order to understand the 

plumbing system of this volcano and evaluate new approaches to estimate magma 

ascent rates. 

2.2.1 Specific objectives 

1. To identify and describe the pyroclastic fall deposits associated with the largest 

eruption of Cerro Machin volcano. 

2. To describe the vertical and distance-related variations of the deposits in order 

to characterize the nature and evolution of the eruption.  

3. To characterize the mineralogy and componentry of representative samples in 

order to recognize the juvenile material.  

4. To determine the storage conditions of the magma (temperature, pressure) 

related to this eruption using the chemistry of the juvenile material. 

5. To assess magmatic water content and chemical composition prior to the 

eruption based on phenocryst-hosted melt inclusions. 

6. To estimate magma ascent rates based on modeling of diffusive water loss in 

plagioclase phenocrysts.  

  



35 
 

Chapter 3: Methods 

3.1 Field work 

The pyroclastic fall deposits of Cerro Machin volcano have been identified as the 

products of six major eruptions: a 5000 yr BP eruption named El Espartillal; a 4800 yr 

BP eruption named P0; a 3600 yr BP eruption named P1; a 2600 yr BP eruption named 

Guaico; a 1200 yr BP eruption named P2; and a 900 yr BP eruption named Anillo (Rueda, 

2005). The best exposures of the largest known explosive eruption P1 (3600 yr BP) are 

found to the west of the volcano. Fieldwork outcome was a detailed description of 28 

stratigraphic profiles (Appendix A). Each deposit was described in terms of the following 

characteristics: bed thickness, bed geometry, type of contacts, color, sedimentary 

structures, sorting, framework, roundness, shapes, and macroscopic components were 

recorded at a field scale to correlate the different deposits from proximal to distal areas. 

The different parameters mentioned were used to interpret the beginning, evolution and 

final stage of each event. Pyroclastic fall deposits from different eruptions were 

recognized based on the identification of sharp contacts representing erosion times and 

fine-grain size layers representing lower energy events or different fragmentation 

mechanisms. Pauses in the eruptive activity were recognized by the presence of 

paleosols (Fisher & Schmincke, 1984).  

The sampling process required the characterization of different lithofacies within each 

unit. For P1, clasts of pumice from the bottom, middle and top sections of the layer were 

collected in proximal to medial areas. The size of these clasts ranges from medium lapilli 

to coarse lapilli with a diameter of 10-64 mm as they represent rapidly quenched 

material that did not experience water loss due to post-eruptive cooling (Lloyd et al., 

2013). When the average grain size of the layer was less than 2 mm, samples from the 

whole layer were collected without any distinction. Laminated lithofacies were sampled 



36 
 

according to fabric and grain size association. Matrix-supported fine-grain size layers 

were sampled separately from clast-supported coarse-grain size layers. In distal areas, 

bulk samples of each lithofacies were collected. 

3.2 Sample preparation 

Ten samples of pumice clasts from the bottom, middle and top regions of the P1 layer—

the target eruption—were ranked by decreasing size from 10 for the largest clast to 1 

for the smallest clast. Ten of them were cut in half and one half was made into a thin 

section by Spectrum Petrographics, Inc. The other half from each clast was lightly 

crushed and sieved, and plagioclase and quartz phenocrysts were separated by hand 

using a Leica S9i Stereomicroscope. The phenocrysts were immersed in isopropanol in 

order to screen for cracks, melt inclusions and melt re-entrants. Phenocrysts were 

mounted in Crystalbond and polished on two parallel sides until the melt inclusions 

were doubly exposed. We chose melt inclusions that were wholly enclosed (i.e., no cracks 

or re-entrant features) regardless of their position within the crystal - core or closer to 

the rim. Sometimes more than one melt inclusion was analyzed in the same crystal. All 

melt inclusions contained multiple small vapor bubbles, especially those hosted in 

quartz crystals. These vapor bubbles made it difficult to measure water concentration 

profiles in melt re-entrants. 

For analysis of water concentration gradients, crystals of plagioclase were cut in half 

using a Wire Saw (WS 25) with a 20-micron wire and each half was polished to intersect 

different crystallographic axes (see Figure 8). Following FTIR analysis, a subset of 36 

melt inclusions and 16 crystals cut in half—a total of 32 wafers—were placed in dental 

resin for EPMA analysis. Once hardened, the dental resin mount was polished with 0.3-

μm aluminum paste to remove residual material from the dental resin preparation and 



37 
 

expose the melt inclusions and crystals. Finally, the ten thin sections of pumice clasts 

from the bottom, middle and top of the layer were analyzed by petrographic microscope 

and then carbon coated for Electron Probe Microanalyzer (EPMA) at the Advanced 

Imaging and Microscopy Laboratory (AIMLab) in the Maryland Nanocenter at the 

University of Maryland. 

3.3 Componentry and petrographic analysis  

Component analysis was conducted in the 0φ, 1φ and 2φ (1, 0.3 and 0.2 mm) size 

fractions of individual lithofacies from P1. A total of ˜300 particles per size fraction 

were described following the methodology of Cas and Wright (1987) in which particles 

are classified as juvenile, accessory and accidental material (Appendix B, Table A). 

Component variations in the different phases within the eruptive units were key to 

interpret potential fragmentation mechanisms and changes in eruptive style. In addition, 

different textures of pumices were described under a binocular microscope to support 

interpretations of the following analysis.  

The petrographic analyses of ten thin sections were conducted with a Leica DM750P 

Petrographic Microscope (Appendix C). The different mineral phases were classified by 

size, and detailed descriptions of growth textures (e.g. oscillatory zoning, patchy zoning, 

presence of inclusion rims) and disequilibrium textures (e.g. rounded shapes, sieve 

textures, breakdown rims) were provided. During petrographic sessions, detailed maps 

of each thin section were made in order to identify mineral phases for analysis by 

Electron Probe Microanalyzer (EPMA). 



38 
 

3.4 Whole rock geochemistry  

Whole rock chemistry by X-ray fluorescence of five pumice clasts from P1 was conducted 

in the Department of Earth & Environment in Franklin and Marshall College by Professor 

Stanley Mertzman (Appendix D). According to the lab methodology, the total volatile 

concentration (%LOI) is determined by weighing out 1 gram of sample, placing in a muffle 

furnace at 950C for 1.5 hours, removing and cooling it to room temperature in a 

desiccator and re-weighing in order to see the weight change. A portion of this 

anhydrous powder is mixed with lithium tetraborate, placed in a platinum crucible and 

heated until molten. After mixing it is transferred to a platinum casting dish and 

quenched. The resulting glass disk is then analyzed for oxides including SiO2, TiO2, Al2O3, 

Fe2O3 Total, MnO, MgO, CaO, Na2O, K2O, and P2O5. 

For trace element analysis a high purity copolywax powder is added to the whole rock 

powder. It is mixed and pressed into a briquette at 50,000 psi. Data for Rb, Sr, Y, Zr, V, 

Ni, Cr, Nb, Ga, Cu, Zn, Co, Ba, U, Th, La, Ce, Sc, and Pb are reposted in parts per million 

(ppm).   

3.5 Fourier Transform Infrared Spectrometer (FTIR): volatile 

concentrations 

Transmission infrared spectra of the melt inclusions were obtained using a Thermo 

Nicolet iN10 MX Fourier transform infrared spectrometer (FTIR) with a liquid nitrogen-

cooled MCT detector at the University of Maryland. The sample and surroundings were 

continuously purged with dry, CO2-free air to eliminate any contribution from 

atmospheric CO2 to the spectra. The aperture of the infrared beam was set to 10, 15, 20, 

25, 30, 35, 40 and 45 microns based on melt inclusion size. Spectra from at least three 

different apertures at 4 cm-1 resolution and an average of 256 scans were measured per 



39 
 

melt inclusion in order to calculate the uncertainty in the absorbance peak heights 

(Appendix E, Table B). Water concentration was calculated from peak heights using the 

Beer-Lambert Law (Eq. 1) 

 

Eq (1)                                                           𝐶 =
(𝑚𝑤)∗(𝐴)

𝜌∗𝑑∗𝜀
     

 

where C is concentration of water in weight fraction of the absorbing species, mw is the 

molecular weight (g/mol) of the volatile species (18.02 for total H2O and 44.01 for CO2), 

A is the absorbance peak height measured with the OMNIC software using a straight 

baseline in all cases, ρ is the density of the glass (g/L), d is the inclusion thickness (cm) 

, and ε is the molar absorption coefficient (L/mol·cm). In rhyolitic compositions, glass 

densities strongly depend on total H2O concentration, therefore the use of an iterative 

process is required to converge on appropriate values [Eq. 2 (Skirius et al., 1990)]: 

 

Eq (2)                                                         ρ = 2350 – 12.6 CH2O  

 

Water dissolved in silicate glass absorbs infrared light at 5,200 cm−1 and ∼1,630 cm−1 

(attributed to molecular water, H2O) at 4,500 cm−1 (attributed to hydroxyl, OH), and at 

3,550 cm−1 (the fundamental OH stretching vibration, with contributions from both 

molecular water and hydroxyl). In case of melt inclusions with the 3550 cm-1 peak 

saturated (i.e., close to 100% of the light was absorbed at this frequency), the 5200 cm−1 

and 4500 cm−1 were used for water calculations. When observed, CO2 is present as 

molecular CO2 at 2,350 cm−1. The molar absorption coefficient (ε) for the peaks were 

calculated based on the calibration equations of Newman et al. (1986), Zhang et al. 



40 
 

(1997), Ohlhorst et al. (2001) and Nishimura et al. (2005) for 5,200 cm−1 and 4,500 cm−1 

peaks (Appendix E, Table A). The absorption coefficient (e) for molecular CO2 (2350 cm–

1) in rhyolitic glass is 1214 L/mol·cm (Behrens et al., 2004). 

Inclusions and crystal thickness were measured using a Mitutoyo micrometer. For thin 

crystals (< 40 microns), we use a calibrated stereozoom lens with a 1 mm objective. 

Measurements were done with the measuring tool box of the LEICA S9 software on the 

Leica S9i Stereomicroscope. Several measurements across the crystal edge were taken 

and then averaged for following calculations. Thicknesses of eight singly exposed melt 

inclusions hosted in quartz (see Table 3) were calculated using the host quartz silicate 

overtones [Eq. 3 (Tollan et al., 2019)]: 

 

Eq (3)                                     Thickness (MI) = 
𝐴𝑄𝑢𝑎𝑟𝑡𝑧 − 𝐴𝑄𝑢𝑎𝑟𝑡𝑧+𝑀𝐼

𝐴𝑄𝑢𝑎𝑟𝑡𝑧
× measured thickness, 

 

Where AQuartz is the integrated absorbance of the 2136 cm−1 overtone (molecular vibration 

in the near-infrared spectra) of pure quartz adjacent to the melt inclusion, AQuartz+MI is the 

integrate