JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 89, NO.

B10, PAGES 8253-8268, SEPTEMBER 30, 1984

Laboratory Studiesof Volcanic Jets

SUSAN WERNER KIEFFER

U.S. GeologicalSurvey,Flagstaff,Arizona

BRADFORD STURTEVANT

California Institute of Technology,Pasadena

The studyof the fluid dynamicsof violentvolcaniceruptionsby laboratoryexperimentis described,

and the important fluid-dynamicprocessesthat can be examinedin laboratorymodelsare discussedin
detail.In prelim•inary
experiments,
puregasesare eruptedfromsmallreservoirs.
The gasesusedare
Freon 12 and Freon 22, two gasesof high molecularweightand high densitythat are good analogsof
heavyand particulate-laden volcanicgases;nitrogen,a moderatemolecularweight,moderatedensitygas
for which the thermodynamicpropertiesare well known; and helium, a low molecularweight, low-
densitygasthat is usedas a basisfor comparisonwith the behaviorof the heaviergasesand as an analog
of steam,the gasthat dominatesmanyvolcaniceruptions.Transientjets eruptfrom the reservoirinto the
laboratoryuponruptureof a thin diaphragmat the exit of a convergent nozzle.The gasaccelerates from
rest in the reservoirto high velocityin thejet. Reservoirpressures and geometriesare suchthat the fluid
velocityin the jets is initially supersonic
and later decaysto subsonic. The measuredreservoirpressure
decreases as the fluid expandsthroughrepetitivelyreflectingrarefactionwaves,but for the conditionsof
theseexperiments,a simple steady-discharge model is sufficientto explain the pressuredecay and to
predictthe durationof the flow. Densityvariationsin the flow field havebeenvisualizedwith schlieren
and shadowgraphphotography.The observedstructureof the jet is correlatedwith the measured
pressurehistory.The startingvortexgeneratedwhenthe diaphragmrupturesbecomes the headof thejet.
Thoughthe exit velocityis sonic,the flow headin the heliumjet decelerates to aboutone-thirdof sonic
velocityin the first few nozzlediameters,the nitrogenhead decelerates to about three-fourthsof sonic
velocity,while Freon maintainsnearly sonicvelocity.The impulsiveaccelerationof reservoirfluid into
the surroundingatmosphereproducesa compression wave.The strengthof this wavedependsprimarily
on the soundspeedof the fluid in the reservoirbut also,secondarily with oppositeeffect,on the density:
helium producesa relativelystrongatmosphericshockwhile the Freonsdo not produceany optically
observablewave front. Well-formed N waves are detected with a microphone far from the reservoir.
Barrel shocks,Mach disks, and other familiar featuresof steady underexpandedsupersonicjets form
insidethe jet almostimmediatelyafter passageof the flow head.Thesefeaturesare maintaineduntil the
pressurein the reservoirdecaysto sonicconditions.At low pressures thejets are relativelystructureless.
Gas-particlejets from volcaniceruptionsmay behaveas pseudogases if particleconcentrations and mass
and momentumexchangebetweenthe componentsare sufficientlysmall.The soundspeedof volcanic
pseudogases
canbe aslargeas 1000m s- • or assmallasa fewtensof metersper second
depending
on
the mass loading and initial temperature.Fluids of high sound speedproducestrongeratmospheric
shock waves than do those of low sound speed.Therefore eruption of a hot gas lightly laden with
particulatesshould producea strongershock than eruption of a cooler or heavily laden fluid. An
empiricalexpression suggeststhat the initial velocityof the head of supersonic
volcanicjets is controlled
by the soundspeedand the ratio of the densityof the eruptingfluid to that of the atmosphere.The
duration of gas or pseudogas eruptionsis controlledby the soundspeedof the fluid and the ratio of
reservoir volume to vent area.

1. INTRODUCTION studiedexperimentallyso that key parameterscan be varied in

During violent volcaniceruptions,magma transformsfrom
a liquid containing dissolvedgasesinto a fragment-ladengas
(magmaticeruptions);liquid water within pores and cracks
transformsinto steam laden with crushedhost rock (phreatic
eruptions);and/or magma and external water may interact
(phreatomagmaticeruptions).The thermodynamicand fluid-
dynamic behavior of complex fluids expanding through re-
gions of complex geometry within the volcano and into the
surrounding atmosphereis not well understood and is not
amenableto direct observation.Transformation of magma or
liquid water plus surroundingrock into a particulate-laden
gas, accelerationand motion of this gas, and its interaction
with the surroundingatmospherein a jetlike flow involve the
transport of mass, momentum, and energy by complex and
nonlinear mechanisms.It is important that theseprocessesbe
This paper is not subjectto U.S. copyright.Publishedin 1984 by
the AmericanGeophysicalUnion.
Paper number4B0362.
8253
a controlled fashion, because reliable models for volcanic
eruptions and their hazards can be formulated only if the
fundamental dynamic processesthat govern the behavior of
volcanicjets are understood.Therefore in the laboratory we
abstractthe governingphysicsfrom the field situation insofar
as is possiblewithout altering the essentialfeaturesof the flow.
Becauseour laboratory experimentsare of small scale com-
pared with volcanicsystems(by a factor of from 10-5 to
10-7), it is not possibleto modelsimultaneously
the effectsof
fluid inertia and gravity on the flowing fluid. On the other
hand, the effectsof complex geometry, both within the vol-
canoesand in the surroundingtopography,can be studiedat
reduced scale.
Several workers [Shoemaker et al., 1962; Bennett, 1971,
1974; McGetchin and Ulrich, 1973] have suggestedthat fluid
flow during large gas-rich volcanic eruptions resemblesthe
flow from laboratory shock tubes, one-dimensionaldevicesin
which a high-pressuregas at one end is separated from a
low-pressurechamber at the other by a diaphragm that rup-
8254 KIEFFER AND STURTEVANT' LABORATORYSTUDIESOF VOLCANIC JETS

2. What are the flow characteristics of the first fluid out of

(a) INITIAL CONDITIONS (b) TRIGGERING: Undisturbed
See DEVELOPMENT----"•. air
the vent, that is, of the unsteady flow before steady flow is
Inset established ?
3. How does the steady flow regime evolve from the un-
/• Oldvolcanic
/Jrocks
•/ / •empted•Compressed
•r• •• /Head of• • •
steady flow? At a fixed point in space,what is the relation
betweenthe conditionsof unsteadyand steadyflow, i.e., what
is the thermodynamichistory at a point ?
4. What influence does entrainment of ambient atmo-
sphericair at the edgesof the jet have on the flow of the fluid
within the jet, and how doesit dependon the propertiesof the
two fluids ?
5. What is the relation between the observableproperties
/...' of the external flow and the internal history of the reservoiras
•Gassy
ß 5•'"• Ma.ma
still the dischargedevelops?
'•imagm• //'"/ at rest
/•.'.•/
• Magma
with
lY/.•issolved volatiles ••.'..
• 2. VOLCANIC JETS AND PLUMES

(c) QUASI-STEADY FLOW: (d) DECLINE To illustrate the evolution of volcanicjets, it is useful to
EVOLUTION
OFSYSTEM .••;:. considera simplified large-scale,gas-rich eruption that pro-
GEOMETRY
• '::•,••
Intercaldera
•':' duces a caldera. Figure 1 shows how an eruption sequence
might be divided into four stages:(1) initial conditions(the
statesof the atmosphere,magma, and host rock immediately
prior to an eruption); (2) triggering(the event that exposesa

Erosion
fissure
int•
s•••7•
'•••ldera
'• •
o/• •l••fissure:
Widening
of magma
chamber
reservoir at high pressureto ambient pressure)and devel-
opment of flow; (3) quasi-steadyflow (during which constant
flow conditionsnear the sourcepersistfor measurabletimes);
and (4) decline(whenflow velocitiesand massflux are small).
Relative durationsof the stagesdependon circumstances of

crater•
• ••!•ng
of , •emporal
•
L/ magma
changesin

properties
triggering, reservoir size and geometry, and fluid properties.
For example,in some sequences,caldera collapsemight not
occur, as in small Hawaiian eruptions. On the other hand,
fluid flow processesmay overlap severalof thesestages,e.g.,
Fig. 1. Schematicdiagram indicatingstagesof a gas-richeruption calderacollapsemay actually beginwhen an eruptionis trig-
(suchas an eruption from a calderaor a stratovolcano)in relation to geredand continueuntil after the eruption terminates.Never-
the fluid flow processes.(a) Initial conditions:intrusion of magma theless,thesefour stagesprovide a framework for comparing
into the near-surfaceenvironmentand propagationof a conduit(frac- laboratory experimentsto volcanicevents.
ture) to the surfacejust prior to magmaeruption.(b) Triggeringand
Jets and plumes are fluid flows in which mass,momentum,
developmentof flow: exposureof the magma to the ambient atmo-
sphere,propagationof a compressionwave into the atmospherein and energy are convectedfrom an isolated sourceinto a sur-
responseto the eruption of the fluid, rise of material in the fracture,
continued developmentof the fracture vertically and laterally. (c)
Quasi-steadyflow: developmentof the flow behind the unsteady
front, emptying of the bulk of the magma reservoir,evolution of
systemgeometry and magma properties;episodesof unsteadyflow
may occur. (d) Decline of flow: formation of caldera. Processesmay
overlap stages.

tures to initiate the flow. In our laboratory work we have

examined three-dimensional flow from nozzles attached to res-
ervoirs of various geometries,a generalizationof the shock
tube concept. Our research focuseson the behavior of the
erupting material rather than, for example, on the fracture
mechanicsof host rock, the geochemistryof the fluids, or on
magma-water interactions [Wohletz and McQueen, 1983]. In
particular,a detailedstudyis made of thejet that expandsinto
the external atmosphereduring a laboratory-scaleeruption.
The shapeand propertiesof the external flow fields of erup-
tions,expandinginto an infinite half-space,are studiedin rela-
tion to changing reservoir conditions. We examine the influ-
ence of flow unsteadiness,compressibility,and turbulent
mixing. Relationsbetweenpropagatingflow fronts and atmo-
sphericshock and compressionwaves are described.Some of
the questionsthat we addressare:
1. How does the atmospherein the vicinity of an eruption
respond to the eruption of fluids of differing thermodynamic Fig. 2a. Crosssectionof the reservoirusedfor the experiments.
The
properties? nozzlediameteris 1.5 cm (0.6 in.).
 KIEFFER AND STURTEVANT: LABORATORY STUDIES OF VOLCANIC JETS 8255

rounding atmosphere.In terms of end-memberbehavior, pure TABLE 1. Propertiesof GasesUsed in Experiments

plumesare entirely buoyancydriven, and purejets are entirely
Critical Critical Boiling
momentumdriven. As commonlydescribed,plumesare driven Mol. T, P, Point, c,
primarily by buoyancy,whilejets can be momentumdriven or Weight K MPa K •, m/s
buoyancydriven [Turner, 1973; Chen and Rodi, 1980]. Thus
Helium 4 5.2 0.228 4.22 1.67 1002
buoyantjets may equally well be referredto as forcedplumes.
He
It is well known that the effectsof gravity are frequently
Nitrogen 28 126.2 3.4 77.4 1.40 353
importantin the developmentof volcanicjets and plumes,but N:
it has only recentlybeen fully appreciatedthat in many vol- Freon 12 121 385 4.12 243.4 1.137 153
canicflows, particularly in those describedas "explosive,"the CCI:F•
Freon 22 86.5 369.2 4.98 232.4 1.184 185
momentumimparted to the fluid by the sourceis substantially
CHCIF•
greater than that acquired from (or lost to) gravity, so the
effectsof gravity may be neglected[Wilson, 1980; Kieffer, T, temperature.Sound speedc calculatedfor room temperature
1982a]. It is to theseflows that we have directed our attention 27øC.

in this work. Accordingly,we refer to the flows as "jets" in

order to emphasizeboth that the rapidly developingvolcanic
flows we treat are principally momentum driven and that the nearly unity, so the flow is essentiallyisothermal.However,
laboratoryexperimentsdo not modelgravitationaleffects. very little is known about the behavior of transientjets such
as those that occur in explosivevolcanism,even for such rela-
2.1. Multiphase Flows tively simple materials as gasesor pseudogases.Steady free
Upon exit from the conduit, most volcanicjets are a dis- jets have beenwell documentedby laboratory studies[cf. e.g.,
persedmultiphasemixture of liquid dropletsand ash particles Ricou and Spaulding,1961], but little work has been done on
in vapors and gases.Considerablemasstransfertakes place the unsteadyconditionsthat occur at the start-up of these
betweenthe particulate and vapor phasesby exsolutionof flows. Therefore before examining laboratory jets of nonideal
dissolvedgasesand by evaporation and condensationof the fluids for applicationto analysesof volcanicflows,it is neces-
volatile components.Analytical treatment of the flow is sim- sary to developa data baseof the behavior of transientjets of
plifiedwhen (1) the massexchangeis small,(2) the momentum pure gases.Even so, thesepreliminary studiesserveto model
and energy exchange between the phases occurs rapidly those explosive volcanic eruptions that may be reasonably
enoughthat their velocitiesand temperaturesare nearly equal, analyzedby the theory of single-phase gas dynamics.All cur-
and (3) the concentration of ash is so small that interactions rent models of gaseousvolcanic eruption dynamicsassume
(including collisions) between particles are unimportant. that an eruptinggas-pyroclastmixture behavesas a pure gas
Under these conditions,the mixture approximatesa perfect or pseudogasI-McGetchin,1968; McGetchin and Ulrich, 1973;
gas whose density is equal to that of the mixture and whose Wilson, 1976, 1980; Self et al., 1979; Wilson et al., 1980;
pressureis equal to that of the gaseousphase. The fluid Wilsonand Head, 1981; Eichelbergerand Hayes, 1982; Kieffer,
propertiesare describedby a gasconstantR and an isentropic 1982a, b; Sparksand Wilson, 1982]. Hence these laboratory
exponent• that are not simply the gas constantand the ratio experimentsare directly relevant to interpretationof volcanic
behavior based on theories to date.
of specific
heatscp/cv,respectively,
of the gas,but are mass
weightedwith the correspondingvalues for the solid. In this If spatial or temporal variations of the flow are too rapid
casethe fluid is called a pseudogas,and the well-known theory for the velocity and temperatureof liquid and solid particles
of single-phasegas dynamicscan be used to describethe be- to follow, then "relaxation phenomena"occur during the ac-
havior of the jet. Typically, the sound speedof the pseudogas celerative and heating stages,and the pseudogasapproxi-
mixture is so low that the flow velocity is supersonic.Fur- mation is inapplicable.Sucha flow can only be describedby a
thermore,the ratio of specificheats of the pseudogasis very two- or three-fluid theory, no satisfactoryversion of which
existsat this time. Additional (and usually intractable) com-
plexity results
• whereinterphasemasstransferoccurs.If the
II
particulatematerial is sufficientlyconcentratedthat interparti-
cle interactionsoccur, it is not even known what physical
effectsto expect! Studiesof laboratory and industrial fluidized
beds ['see,for example, Rowe, 1971; EI-Kaissey and Homsy,
1976] in which gas velocitiesare generallysmallerthan those
of interesthere show that even thoseflows are inherently un-
stable,so no steadyuniform flow state can exist. Laboratory
experiments designed to resolve some of these issues are
needed.

2.2. Nonsteady Flows

Explosivevolcaniceruptionsinitially generateflowsthat are
inherentlynonsteady;that is, no transformationinto a coordi-
nate systemmoving at constantvelocityrelative to the earth's
surface(whichwe refer to as the laboratory frame)can render
the flow steady. In particular, in the laboratory frame the
Fig. 2b. Layout of the experimentalfacility: sg, spark gap light
source;11 and 12,collimatingmirrors; M1 and M2, plane mirrors,k, "head" of the jet, that is, the first fluid to exit from the source
schlierenknife edgeor dot; e, cameralens,f, film plane; pg, pressure and to push continuously out into the surrounding atmo-
gaugeand high-pressuregassupply.Not to scale. sphere,is inevitably time varying. However, if conditionsin
8256 KIEFFERAND STURTEVANT' LABORATORYSTUDIESOF VOLCANIC JETS

is termed "incompressible."Conservationof energy or, more

directly, momentum is expressedby the Bernoulli equation,
which for the low-speedsteadyflow of an inviscidnon-heat-
conductingfluid is
50 1

(cm)
tI
Po= P+ • Pu2 Po,
P= const (1)
4O
wherePo is the stagnationpressure,P is the staticpressure,u
is the flow velocity, and p is the density.The pressureP re-
spondssolelyto changesof velocityand is not a thermody-
namic variable. On the other hand, as the Mach number M
(= u/c, where c is sound speed)increases,the effectsof com-
pressibilitybecomemore important,and Bernoulli'sequation
// / PP M2
MI
in the aboveform doesnot apply but insteadbecomes

ß 4 •,,.R2•'/ 5 •R4/,• etc•.

0 0.2 0,4 0,6 0.8 I. 0 I 2
time (mllllsec)
Fig. 3. An x-t diagram showing wave and fluid motion in the
axial direction for nitrogenjet; initial pressure7.25 bars, 13-cm-long
reservoir. Above the reservoir the flow is three dimensional, so the
graph representsthe on-axisbehavior.S, atmosphericshockor com-
pression
wave;D, flow head;R•,2,3.... rarefaction
waves;M•,2, Mach
disks;PP, path of fluid particle.The velocitiesof the wavesshown on
the graph and flow propertiesin the differentfieldsare givenin Table
2. The atmosphericshockS is shownwith a velocityinitially greater
than acoustic,but decayingquickly to acousticvelocitywith distance
becauseof radial spreading.The flow head D is shown leaving the
orifice at sonic velocity but with velocity decreasingrapidly to the
measuredvalue of 250 m/s. Note the accelerationsand decelerations
of the particle PP as it passesconsecutivelythrough the first rarefac-
tion wave, the area change, the rarefaction waves in the jet (not
shown)and the Mach disks.
Po_
P - l1
+7-
2 1M2)
3'/(3'-
1) (2)
1.4 Now the pressureplaysboth a dynamicand a thermodynamic
role. For supersonicvelocities(M > 1), the nature of the den-
sityvariationschanges,and for example,the density(and pres-
sure) can undergo large discontinuouschangesin shock
waves.(For a discussionof compressibleflow, see Thompson
[1972].) The most important property of supersonicfree jets
for volcanicapplicationsis that the pressureand densitycan
vary by largefactorsalongthe axisof thejet, while in subson-
ic jets the pressureis nearly atmosphericthroughoutand the
densityis constant.An initially subsonicjet can changeden-
sity in responseto decreasingambient pressure,say, with in-
•P4' Nozzle

the reservoirremain relatively constantfor times longer than

severalacoustictransmissiontimes acrossthe jet, the jet flow
near the sourceis quasi-steady;the flow only becomescom-
pletely steady when reservoir conditions are truly constant
and thejet head has propagatedfar downstream.In the steady
or quasi-steadyportion of a supersonicjet, acousticwavesand
shockwavesmay occur,but they do not move relative to the I I I
lab frame.
Time (ms)
During an eruption,conditionswithin the reservoirchange o
with time (e.g.,becauseof the finite size of the reservoir,be-
causethe magma propertieschange,or becausethe boundary
I I I
conditionsmay change).The changesare communicatedto
the externalflow field by wavesthat propagatefrom the reser- • P4 Baas.
Pi(3
t.
voir through the jet fluid; these waves move downstreamat .e_ , %
the local speedof sound relative to the jet fluid. The smallest o

time for significantchangesin the reservoir is the acoustic

o 4. B
transit time over the largestdimensionof the sourcechamber,
but if a substantial constriction ("throat") occurs at the exit,
the decay time can be much longer; many acoustic transit ß-• 2. •

times may be required before reservoir conditions substan-

tially change.Quasi-steadyflow occursin many volcanicerup-
tions where a narrow conduit or vent overliesa large reser- B lB 2B

voir. Probable exceptionsare in brief eruptions such as Vul- Time (ms }

canian or Strombolian bursts. One purpose of the present
studyis to determinethe dependenceof quasi-steadyflow on
initial conditionsand reservoirgeometry.
2.3. Compressibility
In subsonicflow, changesof fluid density due to acceler-
ation and decelerationof the flow are very small, and the flow
b
Fig. 4. (a) Pressurehistory recordedby the pressuretransducer
locatedin the nozzlejust outsideof the diaphragmfor nitrogenemp-
tying from a reservoir13 cm long, initially at 7.25 bars. P4,*, Ps*
indicate quasi-steadyconditionsin the nozzle exit. (b) Same for the
transducerlocatedat the bottom of the reservoir.P,• and P5 indicate
initial and quasi-steadyconditions,respectively.Dashed line shows
resultfrom pressuredecaymodel (10).
 KIEFFER
ANDSTURTEVANT:
LABORATORY
STUDIES
OFVOLCANIC
JETS 8257

o b

½ d
Fig.5. Schlierenphotosofa nitrogen
jetdevelopment; initial
pressure7.25bars,
reservoir
length13cm,jetexit
diameter1.5cm.(a)292#s,reservoir
atstate
(4'),P,•,= 6.9bars.
Notetheatmosphericshock,
thevortexlike
flowhead,and
thefaintwavesconverging
towardtheaxisandashort Mach diskjustbelowtheflowhead.
(b)At702#s,reservoir
stillat
state
(4').Notethesimilarity
ofthenear-field
structure tothatinFigure5a,themultiple
rarefaction
andshock sequences,
andtheenlargement oftheflowhead.
(c)At2.11 ms,reservoiratstate
(7),5.8bars.
TheMach disks
areslightly
closer
to
theorifice
andaresmaller
indiameter
thanin Figure
5b.(d)At9.92ms,reservoir
pressure
is2.6bars.
Subsonic
jet.Note
theabsenceofshocks
andrarefactions.
Anestimate
ofexperimental
reproducibility
canbeobtained
bycomparing
Figure
5ctoFigures
6cand6d.Allwere taken
atthesamenominal
timesindifferent
runs.
Notealso
thatFigures
9cand9dand
Figure
7bcanbefitted
intothesequence
here,sothatphotos
ofseven
different
stages
oftheeruption
ofnitrogen
from
a
reservoir
initiallyat 7.25barsareshownin thispaper.

creasing
altitudein theambientatmosphere,onlyby becom- tingsarecommonly
100-500
m s-1onearthand,forexample,
ing"compressible,"e.g.,supersonic.
Anothermechanism for more than 1 km s-1 on Io. Compressibility
becomesimpor-
changingthe densityof jet fluid,namely,mixingand en- tant in theseflows,not generally
because of abnormallyhigh
trainment
(seesection2.4),doesnot dependto firstorderon velocities,
butbecause thesoundspeed of theflowingmaterial
externalpressuregradients. can be low. For example,
Velocitiescan be highsubsonic
or evensupersonic
in geo- 1. The soundspeedof boilingwateris in the rangeof a
logicflows,particularly
in geothermal andvolcanic settings fewtensto 100m s- • overa largerangeof steamquality,but
in theseset- canbe aslow as 1 m s-1 [cf.Kieffer,1977].Thereforeflowsin
[Wilsonet al., 1980;Kieffer,1982b].Velocities
8258 KIEFFERAND STURTEVANT:LABORATORYSTUDIESOF VOLCANICJETS

cl b

½ cl
Fig. 6. Schlierenphotographscomparingimagesobtainedwith two differentspark durations: 1 #s (left) and 10 #s
(right).All photographsare of nitrogeneruptingfrom a 13-cm-longreservoirinitially at 7.25 bars;jet exit diameter1.5cm.
(a and b) taken 500 #s after diaphragmrupture when the reservoirwas at state(4'). (c and d) Taken at 2.25 msecwhenthe
reservoirpressurewas about 5.75 bars. The illumination of Figures6b and 6d providesthe best visualizationof the
rarefactionwavesthat emanatefrom the orifice,the interceptingshocksthat form when the rarefactionsreflectfrom the
flow boundary,the Mach disk shockthat standsacrossthe flow on the axis,and the obliqueshockwavesthat emanate
from the Mach disk.(The smallbright line extendingout from the orificein the centerof the flow field is the wake of the
knife blades.)

geyser or volcanic eruptions, or in producing geothermal as 50 m s- • [Marble, 1963, 1970;Rudinger,1965].Therefore

wells,involving the flow of boiling water at speedsfrom a few particulate-ladenflows from volcanoescan be supersonic.
meters to a few tens of meters per secondcan have high sub- In most of these cases,the flow velocities are subsonic rela-
sonicor even supersonicMach numbers. tive to the surrounding atmosphere, e.g., much less than the
2. Wet steamcan have a sound speedof only a few to 100 soundspeedin air (330m s-•). It is not knownwhateffect,if
m s-•. Many volcanicand geothermaljets consistprincipally any, such a disparity between external and internal Mach
of wet steam and can have fluid velocities as large as the numbers can have on the behavior of the flow.
sound speed. A volcanicjet can be supersonicwhen the reservoirpressure
3. Gas flows containing a large massof entrained particu- is more than about twice the atmosphericpressure,as is fre-
late material can have very low sound speeds,perhapsas low quently the case.When the constrictionat the exit is modest
 KIEFFER AND STURTEVANT' LABORATORY STUDIES OF VOLCANIC JETS 8259

or severe(e.g., when a volcanic vent or conduit is small com- TABLE 2. Summary of Initial States and Calculated Early Flow
pared with the feeding reservoir),the jet is "underexpanded" Conditions for Experiments
becausethe exit pressure,though lessthan in the reservoir,is
He N2 Fr 22 Fr 12
much greater than atmospheric.Typically, underexpandedjets
grow rapidly in diameter as they expandinto the atmosphere. P4, bar 7.25 7.25 7.25 7.25
c4, m/s 1002 353 185 153
2.4. Density Effects P4,kg/m3 1.13 7.96 25.4 35.8
Fluid densitiesin volcanicjets vary over several orders of T4, K 300 300 300 300
P4,, bar 6.85 6.92 6.97 6.96
magnitude; pure steam jets can be less dense than the sur- c4,,m/s 990.9 350.6 184.4 152.6
rounding air, whereas particulate-ladenjets are much denser P4,,kg/m3 1.09 7.70 24.5 34.6
than the atmosphere.The high density introducesimportant T4,,K 298.3 299.0 299.5 299.5
gravitational effects and affects the flow dynamics. Even v4,,deg 20.3 25 34.6 33
P4,*, bar 3.55 3.84 4.13 4.20
though steadysubsonicand supersonicjets have been studied
c4,*, m/s 868 32.2 177 150
extensivelyin the laboratory, there is little information avail- P4,*,kg/m3 0.73 5.01 21.5 22.2
able about the effectsof greatly disparatejet and atmospheric T4,*, K 225 250 274 281
densities. Dense subsonic jets flowing into a less dense m4,*,kg m- 2 s- • 300 1240 730 630
medium (suchas a water jet into air) grow downstreammuch Ps, bar 6.53 6.60 6.68 6.69
P6, bar 6.13 6.30 6.41 6.43
less rapidly than jets of density similar to the surrounding
atmosphere(such as an air jet into air) [Tombach, 1969], but For notation seetext and appendix.The initial pressureP4 is esti-
the mechanismfor this is not clear. Presumably,the inertia of mated to be 7.25 _+0.3 bars; the initial temperatureT4 is taken to be
thejet makesit lesssusceptible to the developmentof contort- room temperature,27 +_1 øC. The initial densitiesP4, were obtained
from tables of thermodynamicproperties; c4 was calculated from
ions on its surfaceby instability.Distortion of jet boundaries c4= (7RT)•/2
is the processby which entrainment,turbulent mixing, and
thereforejet growth occur.The principal instabilitiesin dense
jets are the Kelvin-Helmholtz instability of shearingflow and
Froude numbers,respectively.Becausethe experimentsdo not
the Rayleigh-Taylorinstability of an acceleratingnonuniform
fluid.
simulategravity, we discussonly viscouseffects.The Reynolds
number is
As mentionedabove,the diameterof underexpandedsuper-
sonicjets tends to grow rapidly in the immediate vicinity of pUL
the orifice. Since a dense pseudogashas a smaller ratio of Re - (3)
specificheats •, than a less-denseone does, its initial growth
tends to be greater, becausethe angle by which the flow is where L is a characteristiclength, usually the diameter of the
deflected(the Prandtl-Meyer angle) in the expansion wave jet, U is a similar characteristicvelocity,p is the fluid density,
that reducesthe exit pressureto ambient, increaseswith de- and/• is the viscosityof the gaseouscomponentof thejet.
creasing•,. Thus the mechanismsby which jet diameter in- In gassyvolcanic eruptions, viscousforcesplay essentially
creasesdownstreamare different for subsonicand supersonic no role. That is, the Reynoldsnumberis very large. For exam-
jets, and the effect on each of increasingthe densityof the jet ple,for manyvolcanicjets L ,-• 1 km, U ,-• 100m s-x, p ,-• 20
fluid is opposite. kg m-3, and/• ,-• 10-5 kg m-a s-a. The corresponding Reyn-
oldsnumberis greaterthan 10TM.On the otherhand,typical
2.5. Scaling Reynoldsnumbersfor the experimentsreported in this work
Our experimentsare at reducedscale,with different fluids, are about 5 x 10'L
and at lower temperatureand pressurethan in volcanicerup- Becauseof fluid-dynamicinstability,large Reynoldsnumber
tions.In orderto simulatevolcaniceruptions,it is necessary to flows are turbulent. Experiments [e.g., Reynolds,1962] show
scalegeometric,dynamic,and thermodynamicquantities. that freejets are fully turbulent for Reynoldsnumbersgreater
2.5.1. Geometric. Spatial and temporal scalesof a flow than a few hundred.Turbulent mixing and thereforethe devel-
processare interrelated; their ratio is the characteristicflow opment of a turbulent jet is independentof viscosity.Thus in
velocity.The bestchoiceof sucha velocityin a sonicor super- both the lab experimentsand in volcanicjets, fluid viscosity
sonicjet is the sound speed of the jet fluid. Becauseof the has no effect on the structure of the jet. The only issuescon-
universalityof the behaviorof multiphasefluids,the velocities cerningReynoldsnumbermodelingin the presentstudyare
in the experimentaljets are of the same order as in volcanic 1. The turbulence level of the reservoir fluid as it enters
jets (tensto hundredsof metersper second).Becausethe spa- the jet should approximate that of the full-scaleflow, which
tial dimensionsof the experimentalapparatus are at least 5 can be achievedby proper experimentaldesign.
orders of magnitude smaller than volcanoes,the laboratory 2. The jet Reynoldsnumber shouldbe large enoughthat a
timescalesare about 10-5 timesthoseof volcaniceruptions. range of turbulent eddy sizes occurs in the flow sufficient to
That is, 1 s in the life of a volcanicjet correspondsto about 10 approximate the turbulence dynamics [Townsend, 1976] of
/•s in the life of a laboratoryjet. actualvolcanicjets. For the Reynoldsnumbersof theseexper-
2.5.2. Dynamic. The important dynamic quantities are iments the ratio of maximum to minimum eddy sizes is of
the variousforcesthat can alter the momentum(and energy) order 100. The maximum eddy size is of the order of the jet
of the fluid. The validity of the simulationcan be judgedby diameter,and the minimum eddy size is the Kolmogorov mi-
comparingthe numerical valuesof nondimensionalparame- croscale,at which viscousdissipation occurs [Tennekes and
ters that express the ratios of forces. In momentum-driven Lureely, 1972].
flows the inertia of the fluid is the reference force. Other im- 2.5.3. Thermodynamic. If during laboratory experiments
portant forcesto considerare viscousand gravitational,so the the thermodynamic trajectory of the model fluid, from initial
relevant nondimensionalparameters are the Reynolds and conditions to downstream infinity, approximatesthat of the
8260 KIEFFER AND STURTEVANT: LABORATORYSTUDIESOF VOLCANIC JETS

o b c
Fig. 7. The structureof the flow headson jets of similar size:(a) heliumjet at 410 #s, (b) nitrogenjet at 516 #s, and (c)
Freon 22 jet at 620 #s. All eruptionsare from reservoirsinitially at 7.25 bars;jet exit diameter 1.5 cm. Note the relatively
large head on the heliumjet and the well-forrncdring vortex structureson the nitrogenand Freon jets. The timesat which
thesephotoswere taken provide an indication of the differentvelocitiesof thejets.

gasesin an actual volcanicjet, it can be said that the simula-
tion is accurate.This principle is expressedformally by the
Law of CorrespondingStates [Hirschfelderet al., 1954], a
similarity law which states that if thermodynamic variables
are properly normalized by their values at the critical point,
all pure substancesbehave similarly. In particular, when nor-
malized properly, the temperature-entropydiagrams are the
same (see section 5.1). Consequently,insofar as the Law of
CorrespondingStatesapplies,it is possibleto model the fluid-
dynamic and thermodynamic behavior of one fluid by an-
other, more convenient,one. The test fluidsusedin the present
experimentsare chosen to simulate the condensiblevapors
and pseudogases of volcanicjets as dosely as possible.
3. THE EXPERIMENTS
3.1. Apparatus
The experimental apparatus is shown schematically in
Figure 2. The test fluid is pressurizedin a reservoir that is
sealedfrom the atmosphereby a thin aluminum diaphragm
(Figure 2a). The diaphragm ruptures at a prescribedvalue of
the reservoir pressure,allowing the fluid to expand into the
atmosphere.Use of cruciform knife blades fixed just down-
streamof the diaphragminsuresthat diaphragmrupture reli-
ably occurswithin lessthan 0.01 MPa of the designvalue and
that the rupture geometryis repeatable.Diaphragm rupture
takes about 1 ms, but the instant of rupture is not controlled
in the presentexperiments.
The reservoirsare designedas convergentnozzlesto ensure
sonic flow at the throat and supersonicflow in the external
flow. Thus the experimental configuration resemblesa vol-
canic system in which a magma reservoir is tapped by a
straight-walledor upward convergingfissure,not one in which
the fissurefeedsa divergingsurfacecrater.If sucha geometry
were incorporatedinto the laboratory apparatus,many flow
featuresnear the throat in the region of supersonicflow would
occur below the field of view of the optical system.
Pressurehistories in the reservoir are measuredwith pi-
ezoelectric transducers at two different locations and are re-
corded on a digital oscilloscope.One pressuretransducerat
the bottom of the reservoirrecordsthe high initial and subse-
quently decreasingpressure.The other transducer,at the top
of the apparatusjust outside the diaphragm, initially records
atmospheric pressure,then follows a sudden increase to the
exit pressureand finally a subsequentdecrease,paralleling the
decline of the interior reservoir pressure.To date, the maxi-
mum ratio of initial reservoirpressureto ambient atmospheric
pressuretestedhas been 7:1.
The development of the external flow field is recorded by
high-speedshadow and schlierenphotography (Figure 2b; cf.
Liepmannand Roshko[1957]). The light sourceis a spark gap,
triggered at a preset time after the flow begins.Data are ac-
quired by a single Polaroid photograph for each run, and a
photographic sequenceof the flow evolution is obtained by
repeatedruns of different time delay. This method allows ade-
quate spatial resolution (the equivalent of 30 line pairs per
millimeter on a 16-mm image), which is not possible with
high-speedmotion picture photography.The information re-
corded on a photographdependson the duration of the spark
gap. A short spark duration (1 #s) highlights the small-scale
turbulencein the shear layers at the edge of the jet; however,
it obscuresthe internal jet structure.Longer sparks(•> 10 #s)
average out the unsteady motions of fine-scale turbulence,
thereby providing more resolution of steadyinternal features.
Due to their motion, fluid entities such as the head are also
slightlysmearedout by the longer spark duration.
3.2. Test Fluids
The test gasesstudiedto date include helium (He), nitrogen
(N2), Freon 12 (dichlorodifluoromethane, CCI2F:), and Freon
22 (chlorodifluoromethane, CHC1F:). Their physicaland ther-
modynamic propertiesare compiled in Table 1. Helium and
nitrogenare usedas typical light, high soundspeedvapors to
provide a basis for comparison with the heavy, low sound
speedvapors and multiphasefluids to be used as analogsto
magmatic flows. Freon vapor has a relatively low ratio of
 KIEFFER AND STURTEVANT: LABORATORY STUDIES OF VOLCANIC JETS
..
....
"'":%....
;. -,. ..
ß
f
..,
...:;}-:;!i..
!71:
...
:..
Fig. 8. A vortexlikestructureat the head of a smalleruption of Sakurajima,Japan.
specificheats(e.g.,y - 1.14),like the volcanicpseudogaspro-
posed by Kieffer [1982a] as the eruptive fluid in the lateral
blast of May 18, 1980,at Mount Saint Helens(y - 1.04);how-
ever, its density is a factor of 40 lessthan that assumedfor the
volcanicpseudogas.At somewhathigher pressure,the Freons
can be liquified in the laboratory and provide a good analog
for the two-phaseflow of water because,in addition to their
high density and low ratios of specificheats, near room tem-
perature they are superheatedabove their boiling points by
about the same amount, relative to their critical temperature,
as is water at conditionsthought to be representativeof many
shallowcrustal environments,i.e., low-pressure(P < 10 kbar),
low-temperature(T < 1200 K) sourceregions.
3.3. Flow History
The evolution of jets erupting from the laboratory reser-
voirscan be describedwith referenceto a space-time(or "x-t")
diagram. A quantitativeexampletaken from a typical experi-
ment in the presentstudyis shownin Figure 3.
When the diaphragmruptures,an atmosphericcompression
or shockwave (S) travelsfrom the diaphragminto the labora-
tory. Simultaneously,a rarefactionwave R• propagatesdown-
ward, signalingthe diaphragm rupture to the fluid in the res-
ervoir and acceleratingand expandingit in the upward direc-
tion. The head of the rarefaction moves downward
8261
speed of sound of the undisturbed reservoir fluid (Table 1).
The velocityof the tail of the wave through fluid that has been
acceleratedupward and has been cooled is smaller than the
velocity of the head. However, in the experimentsto date the
rarefaction is rather weak, because the major acceleration
occurs later when the fluid encounters the substantial area
contraction (17.4:1) at the exit. Therefore the difference in
velocity betweenthe head and tail of the wave is too small to
be resolvedin Figure 3. A typical particlepath of fluid initially
in the reservoiris indicated by the dotted path PP. The mild
acceleration
by the rarefaction(to only 10 m s- • in the case
illustrated)and the strong accelerationin the contraction (to
sonicvelocity,322m s-•)are indicatedby thegreatlydiffering
slopesof particlepath in thoseregions.
When the rarefaction wave reaches the bottom of the reser-
voir, it reflectsfrom the rigid wall as another rarefaction,R 2,
and traversesagain the full length of the tube. At any position
x in the reservoirthere is an interval betweenpassageof the
first and secondwaves when all quantites are constant (state
4'). During this and subsequentsimilar intervals(5, 6, etc.) the
measuredpressurehistories(Figure 4) show plateaus between
the passageof rarefactions.The values of pressureon suc-
cessiveplateaushave beendesignatedP4,, Ps, etc.,correspond-
ing to the relevant flow state. It is through this series of
with the rarefactions that the pressure drops, stepwise, to ambient.
8262 KIEFFER
ANDSTURTEVANT'
LABORATORY
STUDIES
OFVOLCANIC
JETS

½ d
Fig.9. Atmosphericwave frontscausedbytheemergence of(a)helium (160#s),(b)Freon22(340#s),(c)nitrogen
(216
#s),and(d)nitrogen
(322#s).Allflowsarefromreservoirsinitially
at 7.25bars;
jetexitdiameter1.5cm.Notethestrong
shockassociatedwiththeheliumeruption (Figure9a),thelackof optically detectableatmospheric
wavewithFreon
(Figure
9b),andthesteepeningofaninitially
dispersed
wave asthenitrogen flowevolves(fromFigure
9cto9d).

Thustheunsteady flowprocess of emptying thereservoirof its reservoirlargeenoughto producequasi-steady flow nearthe

high-pressure gasoccursby a sequence of steadyflow con- nozzleexit. Figure 6 illustrateshow the duration of the illumi-
ditionsat theexit.Equations for flowconditions in regions 4', nationenhances differentflowfeatures. The features depicted
5, and 6 are givenin the appendix. in Figures5 and 6 are typicalof the jets generatedin all
In summary, interpretationof thedevelopment ofjetsema- experiments to date. In this section we discuss the devel-
natingfromreservoirs of finitevolumeis meaningful onlyin opmentof theflowwith reference to the fourstages definedin
reference to changing conditions in the reservoir.
Extrapola- Figure 1.
tionto larger,morecomplicated volcanicsystemsmaybe ac-
complished throughthe scalingof experimentally measured 4.1. Initial Conditions
histories.
The resultspresentedhere were obtained in runs at an ini-
4. RESULTS
tial reservoir
pressure
of 7.25barsandtemperature of 27øC.A
summaryof conditionsin the reservoirduring the first few
Dark-backgroundschlierenphotographs (Figure5) show pressure plateaus,
calculatedfromthe initialconditions (state
the development
of a nitrogenjet flowinginto air from a 4) andthereservoirgeometry, is givenin Table2. Notepartic-
 KIEFFERAND STURTEVANT:LABORATORYSTUDIESOF VOLCANIC JETS 8263

ularly that the initial densitiesof the gasesvary by a factor of

35.
Aberdeen,.
4.2. Developmentof the Flow
4.2.1. Unsteadyflow at the head of the jet. The initial,
rapid accelerationof fluid from the reservoirgeneratesa start-
ing vortex structure,which sometimeshas the appearanceof a
vortex ring. This structurebecomesthe head of the jet (Fig-
ures 6a and 7). Similar vortical structurescommonly develop
in volcanic eruptions(Figure 8) and buoyant plumes.From
flow visualization studies in water it is known that the ap- Fig. 10b. (N wave recorded by a barograph at Aberdeen, Scot-
pearanceof the headsof jets can vary greatly [Zukoski, 1982], land, from the eruption of Krakatoa in 1883 [Strachey, 1888]). Nei-
but the causesfor the variations have not yet been isolated ther the time scalenor the amplitude of the trace is given in the cited
reference:we estimatethat the duration of the N wave may be as long
and documented.Varying the density of the jet fluid has a as several minutes.
visibleeffect (Figure 7); interaction of a light jet with a heavy
fluid (i.e., a helium jet moving into air) yields a much more
voluminous head than the inverse. The interaction chey, 1888] and is frequentlydemonstratedby the sonic boom
has a simi-
larly pronounced effect on the velocity of the head (approxi- from aircraft, after the wave propagatessome distancefrom
mate velocitiesare obtainedby comparingphotographstaken the source in spherical symmetry, the pressurebehind the
at different stagesof flow)' a few orificediametersdownstream front falls below atmospheric.A trailing compressionis then
of the exit, the helium head has slowed to one-third (300 m required to bring the pressureback to ambient. When fluid-
s-x) of the heliumexit velocity(868m s-x); the nitrogenhead dynamicnonlinearityand shocksteepeningare important, the
hasslowedto three-quarters (250m s-x) of the corresponding pressurehistory in the acousticfar field acquiresthe shapeof
exitvelocity(322m s-x); but to withinexperimental accuracy, the letter N, the so-calledN wave. Figure 10a showsthe re-
the Freonheadsmoveat exit velocity(about175m s-x). This sponseof a microphonesituateda large distancefrom a tran-
trend may reflectthe fact that entrainmentof a heavyfluid has sient jet in the present experiments.Several well-formed N-
a much greater effect on the momentumof a light fluid than shapedatmosphericwaveswere observedfar from the event at
vice versa. Krakatoa (Figure 10b; $trachey[ 1888]). The extent of nonlin-
4.2.2. Atmosphericshockwave. Figure 9 showsthe head earity and absence of dispersion of these acoustic gravity
of the jet and the atmosphericdisturbanceshortly after dia- wavessuggestedby the waveformsis remarkable.
phragm rupture for three differenttest gases.The strengthof
4.3. Steady State Jet Structure
the atmospheric wave dependson the velocity of the head,
which in turn, scaleswith the exit velocity of the jet or, for Our experimentsshow that, at any distancefrom the reser-
supersonicjets, with the jet fluid soundspeed.If the velocityof voir exit, steady flow features develop almost immediately
the head is large enough, a discontinuous(shock) wave may after passageof the flow head (cf. especiallyFigure 5a). Fur-
form in the atmospherewhen the compressioncausedby the thermore, even for the shortest reservoir (2.5 cm), which ex-
jet steepensby fluid-dynamic nonlinearity. Helium, with its haustsmost rapidly, flows develop that are qualitatively very
soundspeedof 1002m s-x, producesa strongatmospheric similar to steady flows at equivalent conditions. Except for
shock near the orifice (Figure 9a). The Freons, with sound brief eruptionsin which most of the massis ejectedin the flow
speedsof 150 and 185 m s-x, do not producean optically head itself, a steady flow model based on contemporaneous
detectable compressionwave (Figure 9b). Nitrogen, with a reservoirconditionsgivesa valid picture of conditionsin the
soundspeedof 353 m s-x, initially producesonly a strong jet after passageof the starting vortex. A qualitative illustra-
compressionwave. It steepensinto a shock as it travels tion of the structure of a steady supersonicjet is given in
through the atmosphere(Figures9c and 9d). Figure 11 for comparisonwith the laboratory photographs.
Due to the subsequentdecelerationof the flow head and In general,details of the jet structure,including divergence
also to the radial spreadingof the wave front, the pressure angle, curvature of the external boundary, distance to the
decreasesrapidly behind the initial compression.As was al- Maeh disk shock, and diameter of the Maeh disk, depend on
ready known at the time of the explosionof Krakatoa [Stra- the thermodynamic properties of the fluid, the ratio of reser-
voir to ambient pressures,and the ratio of exit to throat area
(1.07, ½f. Figure 2a, which shows how the blockage by the
4000 I I I I
knife blades causes the effective throat area to be less than the
exit area). The lateral divergenceof the jet at the orifice is
calculatedas the differencebetweenthe Prandtl-Meyer angle
2000
for flow from P4, to Pa and the Prandtl-Meyer angle for the
exit Mach number M e. For all gaseswe estimate the exit
Mach number as 1.3. The calculatedlateral divergenceis then
14ø for helium, 18ø for nitrogen, and 26ø and 28ø for Freon 12
and 22, respectively.These are comparableto, or larger than,
the observed angles (see Figures 6 and 7, which show the
CHANNEL
1 I I I I divergenceof nitrogenjets while the reservoiris in state4').
_
2 .0 .2 .4 .õ ;)
The rarefactionsthat deflect the flow laterally at the orifice
Time (ms)
reflect from the boundary as compressionsand coalesceto
Fig. 10a. Pressuretrace of the N wave in air measuredby a micro- form so-called "intercepting" or "barrel" shocks. These fea-
phone located38 nozzlediametersfrom the orifice. tures are prominent as bright oblique lines in the schlieren
 8264 KIEFFER
ANDSTURTEVANT.'
LABORATORY
STUDIES
OFVOLCANIC
JETS

Rarefac t ion reservoiryields

wave

dp rh A
dt
..... V p'c* -V (5)
Discontinuity
Surface
where quantitieswithout superscripts
refer to reservoircon-
Axia
Machdisk Flow
'• ditions and V is the reservoir volume. Now, from the isen-
Boundary] .' shock • tropicflow relations[e.g.,LiepmannandRoshko,1957],
layer ß

c* = c p* (6)
7+1 =P 7+1
and,fromtheequationof statefor the adiabaticexpansion
of
Streamline
,
a perfectgas,
ß

C__(p•
Co (•'-
1)/2 (7)
Weak (intercepting) M>>I
,• Rarefaction where subscript0 refersto initial conditionsin the reservoir.
fan Then
Compression
waves

d___p_ •A(
dt- -PøCø 2)(•+1)/2(7-1)(p•(7+1)/2
7+1
If 7 • 1, thisintegratesto

--
Conduit or
X' X
L\Po/ 7- 1 7+ 1 coA
crater wall
(9)
for to = O. For an adiabaticexpansion,
this equationalter-
Overpressured nativelycanbe writtenin termsof pressure:
flow

Fig. 11. Schematic illustrationof the featuresin an underexpanded

jet [JointArmy,Navy,NASA, Air Force,1975]. t=LLPo/ - 1 y- 1 y+ 1 coA
(10)
photos(e.g.,Figures6b and 6d).The barrelshocksare connec-
tedacrosstheflowat theirdownstream endby the Machdisk,
1500
and their reflections radiate outward from the Mach disk 1200
I i I I i I .:.:.:. :.:.:. :.1.:.1.:.:.:.:
triple point as obliqueshocks.The reflectedshocks,in turn,
intersectthe outerboundaryof the flow at its planeof mini-
mumdiameter.The entirepatternrepeatsagainandagainin lOOO
- KIMBEllLITE / 1300

thedownstream direction(Figure6d).
4.4. Declineof the Flow
800
As time during the eruption increases,and the reservoir
pressuredecreases,
the Prandtl-Meyerangledecreases, and the
diameter and stand-off distance of the Mach disk decrease.In •o:':':::::::::::'•v
600 • :::::.:.:.:.:.'.o
- 900 •
effect,thejet "collapses."
Thesechanges
are mostnoticeable
whenthe reservoirpressuredecreases
to a valuefor whichthe
flow is everywheresubsonic.
700•
In thissectionwe derivean approximate expressionfor the I-- 400

decayof reservoirpressure with timeandcompareit with the

experimentalresults.We use the theory of one-dimensional
compressible flowandassume quasi-steady flow.In particular, 200

nonuniformities in the reservoirare neglected.

Thusthe calcu-
lation only appliesto reservoirgeometries for whichthe con-
striction at the exit is rather severe. We further assume that 0
0 2 4 6 8 -•0
the reservoirpressureis so high that the exit flow is choked -1
Entropy, Joule g K
and thejet flow is supersonic.In principle,a furthercalcula-
tion shouldbe madeof the flow decayafter the nozzleun- Fig. 12. Temperature-entropy
diagramfor H20. The shaded
chokesandthe flowbecomes
subsonic,
but we applythe re- regionon the right showsthe rangeof initial conditionsfrom which
sultsonly to large pressureratio cases,for which the time in isentropic
decompression(vertical
paths)of pureH20 wouldproduce
vaporalone.Otherinitialconditions
resultin a two-phase
mixture
supersonic flow is muchgreaterthan in subsonic.
eitherby boiling(if theinitialentropyis lessthanthecriticalpoint
The massflux• throughtheexit planeis entropy)or by condensation, as shownby the samplepath labeled
m = 0 from an initial conditionappropriateto a kimberlitereservoir.
rh= p*c*A (4)
The pathsslantedto higherentropyrepresent the thermodynamic
where p* and c* are the sonicconditions,assumedconstant pathof theH:O component in an adiabatically
tureof H:O andparticulate
decompressing
matter.The ratio of particulate
mix-
massto
across the orifice area A. The conservation of mass in the H:O is denotedby m.
 KIEFFER
ANDSTURTEVANT.'
LABORATORY
STUDIES
OFVOLCANIC
JETS 8265

The pressurehistorycalculatedfrom (10) for the experi-

mentalconfiguration,
whereV - 384cm3 andA - 1.81cm2,
is shownin Figure4b. The agreementwith measurements is
satisfactory.
The differencebetweenthe predictedand mea-
suredcurvesat late time may be dueto dischargeof about 160
cm3 of vaporin recharge tubingconnected to andnot isolated
from the reservoir.
For thegases
Studied
thetransition
froma supersonic
to a
subsonicjet shouldoccur when the reservoirpressurede-
creasesto about 2 times ambient. The structure of weakly
supersonic steadyjets was studiedby Love and Grigsby
[1955]. In the presentexperiments, because the shocksand
rarefactionsin thejet are so weakand the obscuration of the
internalstructureby shearlayer turbulenceis so strong,the
wavesseemto disappearbeforethe transitionto subsonic
conditionswhilethe pressure is still abouttwicethe transition
pressure.Whenthe reservoir pressure is in the rangeof 2-4 Fig. 13. A geothermalbore discharging in the Imperial Valley.
The flow is chokedat the pipeexit and producesa supersonic
jet with
timesatmospheric, the effectsof compressibility are lessim- the characteristicflared shape.
portantand, whenthe jet becomes subsonic, are negligible
(Figure 5d).
There are no volcanic eruptions for which observations
5. DISCUSSION
permit a well-constrained
flow model, even usingthe sim-
Thoughthe primarypurposeof the presentstudyis to pro- plifiedpseudogasassumption.The mostimportantof the ill-
videa basisfor comparisonwith experiments usingmultiphase constrainedvariablesare (1) the abundanceof volatiles,(2) the
fluids,enoughvolcaniceruptionsresembletheseexperiments geometryof the volcanicsystem(expressed
throughV, A, and
to justifybriefdiscussion
of the analogies.
The fluidsin such reservoir-nozzlegeometry),and (3) initial pressureand temper-
eruptionsare eithergasesor behavelike pseudogases. Erup- ature in the reservoir(expressedthrough an averagePo and
tionsthat producejets similarto our laboratoryjets include To).Thusapplicationto specificvolcanoes is limited.The pur-
phreaticto phreatomagmatic eventswherepulverizedcountry poseof this sectionis to point out severalphenomenathat
rockand fragmented magmaare drivenlargelyby magma- may be observedand their relationto reservoirconditions.
heatedexpanding meteoricwatervaporand magmaticpyro-
5.1. Initial Conditionsand ThermodynamicPaths
clasticjets drivenlargelyby exsolvedjuvenilevolatiles.As far
asthejet structureis concerned,the originof the gasis imma- The geometryof volcanicsystemsat depthis rarely known.
terial. Examplesmight include(1) the Mount Saint Helens Field observationssuggestthat eruptions commonly begin
phreatic eruptions OfMarchandApril 1980;(2)theVulcanian with the propagationof narrow fissuresfrom magma reser-
eruptions of Ngauruhoe,February1975;(3) thelateralblastat voirs to the surface,a configurationthat resemblesa conver-
Mount Saint Helens on May 18, 1980 (assumingeither that gentnozzleabovea reservoir(Figure la). Repeatederuptions
therewassubstantial hydrothermal contribution to the erup- cantake placefrom the samefissureor from a roughlycylin-
tion or that any magmaticvolatilesinvolvedhad exsolved drical conduitthat developswithin the fissure.
prior to eruption);and (4) gas-rich,reticulite-forming phases Plausible initial conditionsand thermodynamicpaths of
of someHawaiianeruptions.More speculatively, if the magma fluids that might satisfy the requirement that no phase
involvedin large calderaeruptionsexsolvesits gas prior to changes occurin thejet flow are shownon the TS diagramof
eruption,then the largecaldera-forming eruptionsthat result H20 in Figure 12. Use of the TS diagramto analyzethe
in massiveashflows(e.g.,the emplacementof the BishopTuff behavior
ofthegaseous
component
assumes
equilibrium
ther-
700,000yearsago in Long Valley,California)may form vol- modynamicconditionsthroughoutthe flow. At initial reser-
canicjets that act as pseudogases. voir temperaturesof 600-1100øCand at pressuresup to 1
kbar the fluid remains on the vapor side of the coexistence
region,so long as its ascentis eitherisentropicor the entropy
TABLE 3. Propertiesof VariousVolcanicGasesand Pseudogases of the vaporincreases (by thermodynamic irreversibilityor by
heattransfer,e.g.,from the wallsof the conduitor from partic-
• c, m s-1 ulate material entrainedin the gas).Volcanicfluids satisfying
theserequirements rangefrompuresteamto particulate-laden
at at R, at at
m 600øC 1100øC J/kg K 600øC 1100øC P*/Po c*/Co
gasesof mixedcomposition(H20, CO2, SO2, H2S, CO, C1,
etc.).For simplicity,we assumethat the dominantvapor in
0 1.26 1.22 463 714 895 0.55 0.94 particulate-ladenjets is H20; if the flow is evenmoderately
1 1.16 1.14 231 484 607 0.57 0.96 massloaded,the compositionof the gasis unimportantunless
2 1.12 1.10 154 388 486 0.58 0.97
5 !.06 1.06 77 267 335 0.59 0.98
it participates
in phasechanges, whichdisqualifiesthemixture
10 !.03 1.03 42 194 243 0.59 0.99 from a pseudogas treatmentanyway.The propertiesof pure
25 1.02 1.01 18 127 159 0.60 0.99 steam and of steam laden with a weight ratio m of solidsat
50 1.01 1.01 9.1 90 113 0.60 0.99 600 and 1100øC are shown in Table 3. Note that the ratio of
100 1.004 i.004 4.6 63 79 0.60 0.99 specific
heats• of the mixturedecreases rapidlyfrom the value
200 1.002 1.002 2.3 45 56 0.60 0.99
for pure steam toward unity, as the massloading increases
Here m is the ratio of the massof solidsto the mass of vapor per from zero to valuesthat are only at the low end of the range
unit volume. encounteredin volcanic flows (m ~ 10) and that • is quite
8266 KIEFFER AND $TURTEVANT: LABORATORY STUDIES OF VOLCANIC JETS

lOOO
l , ] ]][[]l i i i ] i [

'øI:
I
I
I
2 . .5. e,
I I I IIIIII
10 -•o 10-9 10-8
I
10-7
I
10-6
I

10-5
I
10-':
I
10 3
-
l
10
-2
['•.3-'%%
10
-1
'-
I

P/Po

Fig. 14. Pressuredecayof the supersonic

part of gasor pseudogas
eruptionsversusnormalizedtime.

independentof temperature.In addition, decreaseof the gas may be stronglyaffectedby head deceleration.If the eruption
constant R with mass loading dramatically affects the reser- beginswith the dischargeof relatively gas-richmaterial, e.g.,a
voir soundspeed. steam cap, it will produce a stronger, sharper compressive
wave than an explosiveeventthat beginswith the dischargeof
5.2. Triggering and Developmentof the Flow more ash or frothy magma. Becausethe soundspeedof a fluid
For explosiveeruptions in which the triggering is rapid increaseswith increasingtemperature,hot gaseswill generate
comparedwith the duration of the eruption, the flow history strongercompressionwavesthan cooler ones.
should be similar to that shown in Figure 3, providing the Where atmosphericshockshave been observedupon in-
axesare properly scaledby the soundspeedof the fluid and itiation of a volcaniceruption,the settinghas often beensuch
the dimensions of the volcanic system. As the fluid in the that the occurrenceof a volatile-rich"cap" overlyinga magma
reservoir acceleratesfrom rest toward the vent, its pressure is plausible: at the Vulcanian-styleeruptions of Ngauruhoe
decreases.The pressuredrop between reservoir and vent is (New Zealand), Sakurajima (Japan), and at Mount Pe16.In
0.55-0.60 for the entire range of pseudogasesconsidered contrast, due to the manner by which the lateral blast at
(Table 3). Fluids at initial pressureshigher than about 1.8 Mount Saint Helens on May 18, 1980, was triggered,a dense
times atmosphericare overpressured(underexpanded)when mixture was suddenlyexposedto the atmosphere,and an at-
they leave the conduit.The sonicvelocitiesfor thesefluids are mosphericshock wave was absent. The smooth compression
within a few percentof the reservoirsoundspeed(0.94-0.99 c). wave generatedby the blast steepenedby atmosphericrefrac-
Dependingon temperatureand massloading,chokedconduit tion into a shock only after it had propagated about 200 km
velocities may rangefrom nearly900 m s-• for pure steam from the source [Reed, 1980]. Although no audible shock
down to only a few tens or a hundred metersper secondfor wave was reported near the volcano, an N wavelike record
particulate-ladenjets. was obtained at Toledo, 54 km northwest.
5.2.1. Unsteadyflow at the head. The presentexperiments Though not apparent in the photographspresentedin this
demonstrate that velocities in the choked conduit are the paper, but seen in other photos from these experiments,es-
highestfor low-densitygases(e.g., helium) and are lower for pecially of heliumjets at small times, the shapeof the atmo-
densegases(e.g., Freon). However, the flow head velocities, sphericshock is nonspherical,protruding outward along the
when normalized by sound speed,show the oppositedepen- jet axis.The portion of the shockthat is pushedalong the axis
denceon density' The helium jet is deceleratedmost rapidly by the (decelerating)flow head retains a finite strengthlonger
(presumablyby entrainment of relatively heavy ambient air), than the portions that diffract around to the sidesof the jet.
whereasthe Freon jet showslittle deceleration.An empirical Thus the laboratory resultsare in accord with the field obser-
expressionderived from the data presentedin section 4.2.1 vations of Nairn [1976] concerningwave shape.
that relatesthe head velocity una few diametersfrom the exit
to the sound speed c and molecular weight m of the three 5.3. Steady State Jet Structure
gasesstudiedto date is approximately The internal structure of volcanic jets, particularly the
steady shock and rarefactionwavesthat arise when the jet is

un• c 1/3+ m/m a

m/ma (11)
underexpanded(as describedin section4.3), is difficultto doc-
ument, both due to the paucity of field observationsand be-
where unsubscriptedquantities refer to reservoir fluid and cause of the opacity of volcanic fluids. Reported internal
( )a refers to the atmospheric fluid. Equation (11) expresses waves (at Tolbatchik, USSR [Livshits and Bolkhovitinov,
the fact that for light gasesthe head is deceleratedmuch more 1977]), were not the stationary waves described here, but
rapidly than for heavy. The equation does not include any rather were traveling surgewavesattributed to oscillationsof
variation of un in the streamwise(x) direction, a dependence exit flow velocities above and below sonic. Actually, con-
that has not yet been examined with any precision in the clusiveevidencefor supersonicflow and internal waveswould
present experiments.Further study of flow head velocity is be the simple observationof substantialjet pluming at the
required;field observations shouldincludedocumentation of conduit exit, due to Prandtl-Meyer expansionand subsequent
observedvent and flow head velocitieswheneverpossible. inward curvature of the jet boundaries'(sections2.3 and 2.4).
5.2.2. Atmosphericshockwave. In view of the dependence Unfortunately, the part of volcanic jets where such effects
of head velocity on flow parameters,volcanicfluids with low occur is often deep within a volcanic crater, or is otherwise
mass loading should initially produce the strongest atmo- obscuredfrom observationby volcanic topographyor debris.
sphericcompressionwaves,but the subsequentwave strength Such boundariesare often prominent on dischargesfrom geo-
 KIEFFER AND STURTEVANT: LABORATORY STUDIES OF VOLCANIC JETS 8267

thermal bores (Figure 13), which are sometimesmodeled as
pseudogasflows.
The high internal pressurethat existsin underexpandedjets
can erode the conduit (Figure lb) into a crater (Figure lc)
[Kieffer, 1982b]; deposition of ejecta may also modify
conduit-cratergeometry[Wilson et al., 1980]. The structureof
jets emergingfrom craterswill be differentfrom thoseemerg-
ing from simplestraight conduits.
5.4. Duration of Eruptions
The results of the calculation of reservoir discharge (10)
exhibit the variables that influence reservoir pressuredecay
under conditions for which all the simplifyingassumptionsof
the model apply; that is, for high reservoirpressuresand for
reservoirgeometriesin which the exit constrictionis severe,so
the flow is quasi-steadyand supersonic.The variablescontrol-
ling the dischargeare the .gas propertiesthrough 7 and the
equation of state, the initial reservoirconditionsthrough the
soundspeedc, and the volumeto area ratio VIA. The pressure
decayscaleslinearly with a characteristictime, z = V/(Aco), as
shown by the dimensionsof (5). The duration of an eruption
can be approximated by the time that it takes the reservoirto
discharge
to thepressure
Ps that yieldsambientpressure
Poin
the sonic throat. Eruption durations definedin this way scale
with z. The z is also the characteristictime for the dischargeof
a cylindrical reservoirwith no exit constriction,which accord-
ing to one-dimensionalshock tube theory, is dominated by
nonsteadywave processesand for which VIA = L, the tube The pressurein terms of the soundspeedis
length.Thus reservoirsmodeledas very long cylinders("cylin-
der models") or reservoirsmodeled as large volumes with
small exits ("pinhole models") both have very large VIA and
meroles)or that the dischargeis from a finite reservoirbut the
fluids do not behave as gasesor pseudogases for the duration
of the eruption (e.g., the discharge of fluids initially liquid
when the eruption begins is controlled by the saturation
curve).Sincemany terrestrial eruptionssimilarly involve satu-
rated liquids, future experimentsshould examine such cases.
The results of the present study cannot be extrapolated to
eruptionsin whichchangesof phasecontrolthe flow field.
APPENDIX: PRESSURES IN THE RESERVOIR
We presenthere the equationsfor the first few stepsof the
sequenceof steadyflow conditionsthat exist in the reservoir
during its discharge.We use the geometry and notation of
Figure 3. The expansion of the reservoir fluid behind the
shockcan be approximatedas the following two-stepprocess:
(1) an unsteady,constant area expansionthrough the rarefac-
tion waves,and (2) a steadyexpansionthrough the area con-
traction to sonic conditions.
Since the flow is assumed isentropic, all thermodynamic
variables can be expressedin terms of the sound speed.The
sound speed is obtained from the appropriate Riemann in-
variant; for example,for the first rarefactionwave propagating
downwardand expandingthe flow from state(4) to state(4'),
7-1
c,• = c,•,+ #,•, (A1)
2
-P4,(½4,•
P,• \c,•,/
27/(7-1' (A2)

long durations.
Calculated results from (10) for a range of volcanic con- Thus the pressure ratio in terms of the Mach number in
ditions and for low-7 fluids are shownin Figure 14. For a few region(4') is
eruptionswhere adequatedata are available, the model seems
to give reasonable relations between t, A/V, and P. As an
example, a cylindrical terrestrial steam reservoir (VIA = L) p4_(l+Y_l
P4, )2•,/(y-
2 M4' •)• 1+7M. (A3)
500 m long initially at 180 bar, 600øCwould dischargesuper-
where the approximation is valid for small M4,. The density
sonicallyuntil the pressuredroppedto 1.8 bar, i.e., to Psi and temperaturemay be obtained from (A3) and the equation
P0 = 0.010. The dischargetime would be 5 s. If it were a
of state,
pseudogaswith rn = 100 and y = 1.004 (Table 3), subsonic
conditionswould be reachedwhen P• = 1.7 bars or when
P•/Po = 0.0094.The discharge
timewouldbe 60 s. Durations
P4,_ P4,•/•
of 5-60 s are typical of the durations of phreatic or phreato- (A4)
magmatic bursts when the dischargeapproximatesthat of a
pseudogasfrom a reservoir;the Ngauruhoe eruptionsof 1974
T½,(P4,•
•-•/•
and 1975 consisted of bursts of tens of seconds duration
[Nairn, 1976], and the duration of the lateral blast at Mount M•, is as yet unspecified.It is determinedby the hct that, for
Saint Helens on May 18, 1980, was similar [Kieffer, 1982a]. sonic conditions at the throat, there is a unique upstream
For discharge
to ambient
conditions
on Io (P•/Po• 10-xø), Mach number M, for a given area ratio A•/Ae from the reser-
the duration would be about 50 s for the pure steam example voir to the exit. When the flow is sonic at the throat (that is,
and about 5 min for the pseudogasexample. when M** = 1.0 and Ae • A*), it is termed "choked."Thus
Terrestrial reservoirs that are simple cylindrical conduits M, is given implicitly by the area Mach number relation,
ranging up to a few kilometersin length will, if they contain
gaslike fluids, be emptied in a few hundred seconds.Devi-
ations from simple cylindrical geometry that give very large
volume to area ratios (V/A) could significantlyincreasethis
A,- M4,
A* 1(7+21+Y-1M4•
7+1 2)(•+
x)/2(•-
x) (AS)
time. A gas-rich eruption would be about 1 order of mag- It is convenientto construct perfect gas tables for the gasesof
nitude longer if it occured on Io, where the ambient atmo- interest and then to obtain values of M4, and the sonic con-
sphericpressureis 10-7-10-x2 that of earth.Gassyvolcanic ditions(.) directly from the tables.
eruptions that last more than a few hours (for example, vir- The rarefaction R, travels down the reservoir until it hits
tually all large pyroclasticevents on the earth and the long and reflectsfrom the bottom wall as R2. In state 5 after the
eruptionson Io) indicateeither that the erupting reservoirsare reflectionthe fluid is at rest; us = 0. Using now the Riemann
open to recharge(e.g.,resemblecontinuouslydischargingfu- invariant for the oppositefamily of waves than was the case
8268 KIEFFER AND STURTEVANT.' LABORATORY STUDIES OF VOLCANIC JETS
for (A1), we derivethe analogof (A3)'
P5- (1--y--
P4, 21M4,
)2•/(•-
1) (A6)
The rarefactionwave R 2 now reflectsfrom the area change
as R 3. This processis exactlyanalogousto the 4-4' transition,
so
Ps-(1
P6 +•'-1
2 M6 (A7)Ricou,
axisymmetricalturbulentjets, J. Fluid Mech., 11, 21, 1961.
whereM 6 = M•,. Then, P6/P,• may be calculatedfrom
P6 P6 P5
(A8)
P5 P5 Pc, Pc
and (A3), (A6), and (A7). The valuesof flow variablesin states
1, 4, 4', 4'*, 5, and 6 are given in Table 2 for He, N2, Freon 12,
and Freon 22 initially at P,• = 7.25 bars, T,• = 300 K.
Acknowledgment.This work was supportedjointly by the Nation-
al ScienceFoundation under grant EAR-8218445 and the U.S. Geo-
logical Survey.Approved by Director, U.S. Geological Survey, No-
vember 10, 1983.
REFERENCES
Bennett, F. D., Vaporization waves in explosivevolcanism,Nature,
234, 538, 1971.
Bennett, F. D., On volcanic ash formation, Am. J. Sci., 274(6), 648-
661, 1974.
Chen, C. J., and W. Rodi, Vertical TurbulentBuoyantJets,Pergamon,
New York, 1980.
Eichelberger,J. C., and D. B. Hayes, Magmatic model for the Mount
St. Helens blast of May 18, 1980, J. Geophys.Res.,87, 7727-7738,
1982.
E1-Kaissey,M. M., and G. M. Homsy, Instability waves and the
origin of bubblesin fluidized beds,lnt. J. MultiphaseFlow, 2, 379-
395, 1976.
Hirschfelder,J. O., C. F. Curtiss,and R. B. Bird, Molecular Theory of
Gasesand Liquids,John Wiley, New York, 1954.
Joint Army, Navy, NASA, Air Force (JANNAF), Handbook of
rocket exhaust plume technology, Publ. 263, chap. 2, 237 pp.,
Chemical Propulsion Information Agency, Washington, D.C.,
1975.
Kieffer, S. W., Sound speed in liquid-gas mixtures: Water-air and
water steam,J. Geophys.Res.,82, 2895-2904, 1977.
Kieffer, S. W., Fluid dynamics of the May 18 blast at Mount St.
Helens,U.S. Geol.Surv.Prof Pap., 1250, 379-400, 1982a.
Kieffer, S. W., Fluid dynamics and thermodynamicsof Ionian vol-
canism,chap. 18, in The Satellitesof Jupiter, edited by D. Mor-
rison,pp. 647-723, Universityof Arizona Press,Tucson,1982b.
Liepmann,H. W., and A. Roshko, The Elementsof Gasdynamics, John
Wiley, New York, 1957.
Livshits, L. D., and L. G. Bolkhovitinov, Weak shock waves in the
eruption column, Nature, 267, 420-421, 1977.
Love, E. S., and C. E. Grigsby, Some studiesof axisymmetricfreejets
exhaustingfrom sonicand supersonicnozzlesinto still air and into
supersonicstreams,NACA RM L54L31, 178 pp., Natl. Advisory
Comm. on Aeronaut.,Washington,D.C., 1955.
Marble, F. E., Dynamics of a gas containingsmall solid particles,in
Proceedingsof the Frith AGARD Combustionand PropulsionCol-
loquium,pp. 175-213, Pergamon,New York, 1963.
Marble, F. E., Dynamics of dusty gases,Ann. Rev. Fluid Mech., 2,
397-446, 1970.
McGetchin, T. R., The Moses Rockdike: Geology, petrology and
mode of emplacementof a kimberlite-bearingbreccia-dike,San
Juan County, Utah, Ph.D. thesis,Calif. Inst. of Technol.,Pasadena,
1968.
McGetchin, T. R., and G. W. Ulrich, Xenoliths in maars and dia-
tremeswith inferencesfor the Moon, Mars, and Venus,J. Geophys.
Res., 78, 1833-1853, 1973.
Nairn, I. A., Atmosphericshockwavesand condensationcloudsfrom
Ngauruhoe explosiveeruptions,Nature, 259, 190-192, 1976.
Reed, J. W., Air pressurewaves from Mount St. Helens eruptions
(abstract),Eos Trans. AGU, 61(46), 1136, 1980.
Reynolds,A. J., Observationsof a liquid-into-liquidjet, J. Fluid
Mech., 14, 552-556, 1962.
F. P., and D. B. Spaulding,Measurementsof entrainmentby
Rowe, P. N., Experimental properties of bubbles,in Fluidization,
edited by J. F. Davidson and D. Harrison, Academic,New York,
1971.
Rudinger, G., Some effectsof finite particle volume on the dynamics
of gas-particlemixtures,AIAA J., 3, 1217-1222, 1965.
Self, S., L. Wilson, and I. A. Nairn, Vulcanian eruption mechanisms,
Nature, 277, 440443, 1979.
Shoemaker,T. R., C. H. Roach, and F. M. Byers,Jr., Diatremes and
uraniumdepositsin the Hopi Buttes,Arizona,in PetrologicStudies,
a Volumeto Honor A. F. Buddington,pp. 327-355, Geological So-
ciety of America,Boulder,Colo., 1962.
Sparks,R. S. J., and L. Wilson, Explosivevolcaniceruptions,V, Ob-
servationsof plume dynamicsduring the 1979 Soufriereeruption,
St. Vincent, Geophys.J. R. Astron. Soc.,69, 551-570, 1982.
Strachey,R., On the air wavesand soundscausedby the eruption of
Krakatoa in August 1883, in The Eruption of Krakatoa and Subse-
quentPhenomena'Report of the Krakatoa Committeeof the Royal
Society,editedby G. J. Symons,Trubner, London, 1888.
Tennekes,H., and J. L. Lumely, A First Coursein Turbulence,MIT
Press,Cambridge,Mass., 1972.
Thompson, P. A., Compressible-Fluid Dynamics,McGraw-Hill, New
York, 1972.
Tombach, I. H., Velocity measurementswith a new probe in inhomo-
geneousturbulentjets, Ph.D. thesis,Calif. Inst. of Technol.,Pasa-
dena, 1969.
Townsend,A. A., The Structureof Turbulent Shear Flow, Cambridge
University Press,New York, 1976.
Turner, J. F., BuoyancyEffectsin Fluids,CambridgeUniversityPress,
New York, 1973.
Wilson, L., Explosive volcanic eruptions, III, Plinian eruption col-
umns,Geophys.J. R. Astron.Soc.,45, 543-556, 1976.
Wilson, L., Relationships between pressure, volatile content and
ejecta velocity in three types of volcanicexplosion,J. Volcanol.
Geotherm.Res., 8, 297-313, 1980.
Wilson, L., and J. W. Head III, Ascent and eruption of basaltic
magma on the earth and moon, J. Geophys.Res.,86, 2971-3001,
1981.
Wilson, L., R. S. J. Sparks,and G. P. L. Walker, Explosivevolcanic
eruptions,IV, The control of magma propertiesand conduitgeom-
etry on eruption column behavior, Geophys.J. R. Astron.Soc.,63,
117-148, 1980.
Wohletz, K. H., and R. G. McQueen, Experimental studiesof hy-
dromagmaticvolcanism,in ExplosiveVolcanism:Inception,Evolu-
tion, and Hazards,edited by F. R. Boyd, pp. 158-169, Geophysics
Study Committee, National Academy Press, Washington,D.C.,
1983.
S. W. Kieffer, U.S. GeologicalSurvey,Branch of Igneousand Geo-
thermal Processes, 2255 North Gemini Drive, Flagstaff,AZ 86001.
B. Sturtevant, Graduate Aeronautical Laboratories, California In-
stituteof Technology,Pasadena,CA 91125.
(ReceivedNovember 29, 1983;
revisedFebruary3, 1984;
acceptedMarch 1, 1984.)