Theoretical Models of Magma Ascent Through A Geothermal Borehole
Theoretical Models of Magma Ascent Through A Geothermal Borehole
Theoretical Models of Magma Ascent Through A Geothermal Borehole
BOREHOLE
David Dempsey1, Craig McConnochie1
1Department of Civil and Natural Resources Engineering, University of Canterbury, New Zealand
david.dempsey@canterbury.ac.nz
ABSTRACT
As drilling technology improves and shallower resources become tapped, interest is growing in deeper, hotter geothermal resources.
In volcanic settings, the possibility of encountering a magmatic intrusion increases with each metre of borehole. Such intrusions may
not be resolved by surface geophysical surveys, particularly if they are small, deep or sit beneath other bodies that shadow their
presence. Further, if the intrusion contains a mobile fraction of partial melt, this could enter the borehole and begin to ascend.
Magma has been encountered at least three times during geothermal drilling: in Iceland, Kenya and Hawai’i. In the latter instance,
dacitic magma at a depth of 2.5 km entered the borehole and ascended 5.5 m over several minutes before stopping. Several
international projects have proposed to drill shallow magma bodies or their thermal aureoles for scientific inquiry, energy generation
or hazard mitigation. However, the possibility of uncontrolled magma ascent within a wellbore has not been quantitatively addressed.
Here, we present an analysis of the competing time scales controlling the thermal survival of a magma column ascending the borehole.
Rise time is most rapid for low viscosity magmas (like basalts) under large overpressure that are encountered at shallow depths in
large-diameter boreholes. This magma must reach the surface before it is stalled by crystallization – either through radial cooling or
growth of a crystal plug that exerts a drag on the borehole wall. Magma rise is also rapidly accelerated in the presence of a low-
viscosity outer annulus of degassed volatiles, although it is not clear if this configuration could persist stably.
Investigating a typical parameter range, we show that rhyolite magmas are too viscous to feasibly ascend wellbores, which is
consistent with the experience at Hawai’i and Iceland. In exceptional circumstances, low-viscosity basaltic magmas could reach the
surface before freezing, such as occurred in Námafjall geothermal field in 1977 during the Krafla Fires.
1. INTRODUCTION
As conventional geothermal systems are increasingly developed, the search for new high-temperature resources has turned to greater
depths. For instance, the Iceland Deep Drilling Project (IDDP) has explored the use of deep geothermal wells to access supercritical
temperatures of 450-600°C at 4-5 km depth in conventional geothermal fields (Friðleifsson and Elders, 2015). The concept has also
been proposed in unconventional geothermal settings, such as at Avachinsky Volcano in Kamchatka (Fedotov et al., 2007), and at
Newberry Volcano in Oregon (Cladouhos et al. 2018). Drilling to magma would offer unique insights to the scientific community
(e.g., the proposed Krafla Magma Testbed; Hólmgeirsson et al., 2018) and has even been suggested as a mechanism to prevent future
volcanic eruptions (Wilcox et al., 2017).
There have been several notable instances in which geothermal drilling has encountered magma at shallow depths. In 2005, drilling
of KS-13 at Puna geothermal field, Hawai’i, encountered 1050°C dacitic magma at a depth of 2.5 km (Teplow et al., 2009). Magma
flowed into the borehole (25.88 cm diameter) and ascended 5.5 m in several minutes before coming to a halt. Teplow et al. (2009)
concluded that radial cooling would not have been rapid enough to crystallise and arrest the magma. Instead, they suggested a
crystallised plug of magma must have formed in contact with the drilling fluid.
In 2009, drilling of IDDP-1 at Krafla geothermal field, Iceland, encountered >900°C rhyolite magma (Friðleifsson et al., 2015).
Drilling logs show that, on the second side track, the drill string became stuck at a depth of 2.1 km. It was freed and pulled back 9.5
m up the hole and then, three minutes later, re-lowered a depth of 2 m where it became stuck again (Pálsson et al., 2014). One
explanation is that an encounter with magma caused the first stoppage and, when the string was pulled up, it flowed 7.5 m up the
12 1/4” borehole where the second encounter occurred. The well was eventually completed and flowed steam at 450°C. It was lost
two years later due to casing failure after a valve failed and the well had to be killed with cold water.
This was not the first instance of magma flow in boreholes in the Krafla region. In 1977, during the Krafla Fires (a period of rifting
and an eruption between 1975 and 1984), basaltic magma flowed horizontally into a 1138 m borehole in the Námafjall geothermal
field (Larsen et al., 1979). In a 20-minute period, 1.2 m3 of magma was erupted at an estimated rate of 25 kg s-1. From an estimated
column height of 20 to 40 m, the magma was inferred to have travelled at 20 to 30 m s -1 and ascended the (average) 6” diameter
borehole in 0.5 to 1 min.
Finally, a study of drill cuttings from wells MW-04 and MW-06 at Menengai geothermal system, Kenya, revealed fresh glassy,
quenched cuttings from depths of 2 km (Mbia et al., 2015), indicating the drilling had encountered magma.
A further field study relevant to basaltic magma transport in boreholes is the 1981 drilling campaign of the Kilauea Iki lava lake (Helz
and Wright, 1983), originally erupted in 1959. Here, drilling of the crystallized carapace to the crust-melt interface at depths up to
100 m produced multiple observations of “ooze-back”, a phenomenon where the hole was refilled with melt. Ascent rates shortly
For simplicity, we have not considered the possibility of explosive interactions with borehole fluids, interactions between magma and
the borehole casing or cement, or degassing of the magma as it ascends and depressurizes.
Figure 1: Schematic illustration of borehole transport and freezing processes relevant to magma ascent. Symbols and
equations are referred to in the text.
2.1. Viscous flow through a cylindrical pipe
The incompressible Navier-Stokes equation describes unsteady flows of incompressible, viscous fluids:
𝜕𝒗
𝜌 ( + (𝒗 ⋅ ∇)𝒗) = −∇𝑃 + 𝜇∇2 𝒗, (1)
𝜕𝑡
where 𝒗 is the fluid velocity vector, 𝑃 = 𝑝 − 𝜌𝑔𝑧, is the hydrodynamic pressure, 𝑝 is fluid pressure, 𝑔 is gravity, and 𝜌 and 𝜇 and
fluid density and viscosity, respectively. Conservation of mass is expressed by the continuity equation: ∇ ⋅ 𝒗 = 0.
For a well-developed flow along a vertical wellbore of radius, 𝑅, an axisymmetric geometry can be adopted with radial and vertical
coordinates 𝑟 and 𝑧, and the velocity vector written, 𝒗 = (0, 𝑣𝑧 ). From the continuity equation it follows that 𝜕𝑣𝑧 /𝜕𝑧 = 0. Assuming
inertial effects are small, (1) simplifies to:
∂𝑃 𝜇 𝜕 𝜕𝑣𝑧
= (𝑟 ). (2)
∂𝑧 𝑟 𝜕𝑟 𝜕𝑟
Eq. (2) is solved under standard fluid mechanical assumptions: no slip at the borehole wall (𝑣𝑧 (𝑟 = 𝑅) = 0), 𝑃 is a function of 𝑧
only, and the flow velocity is finite at 𝑟 = 0. We define 𝜕𝑃/𝜕𝑧 as the pressure drop Δ𝑃 from the bottom of the borehole to the
ascending magma front at height, ℎ(𝑡).
Integrating Eq. (2) twice with respect to 𝑟, we obtain the quadratic solution for pipe flow, with the column height, ℎ, given implicitly
as the integral of 𝑣𝑧 :
where the magma front is constrained by balancing the volume of the magma column with the area and time integrated flow producing
the column. By making the substitution for ℎ in (3), integrating over 𝑟, and then differentiating with respect to 𝑡 to eliminate the outer
integral, we obtain a differential equation for ℎ that can be solved with initial conditions ℎ(𝑡 = 0) = 0. Setting ℎ to magma depth,
𝑧𝑚 , gives the time to reach the surface, 𝑡𝑎𝑠𝑐𝑒𝑛𝑡 :
𝑅2 − 𝑟 2 Δ𝑃 Δ𝑃𝑡 2
4𝜇𝑧𝑚
𝑣𝑧 = √ , ℎ(𝑡) = 𝑅√ , 𝑡𝑎𝑠𝑐𝑒𝑛𝑡 = 2
. (4)
𝑅 4𝜇𝑡 4𝜇 𝑅 Δ𝑃
Thus, magma ascent is most rapid for large diameter boreholes that access shallow, low viscosity magma under large overpressure.
The maximum overpressure due to buoyancy will occur as magma first enters the borehole and is given by the difference between
the lithostatic load and the mud weight (approximately hydrostatic) in the wellbore at depth, 𝑧𝑚 , i.e., Δ𝑃𝑚𝑎𝑥 = (𝜌𝑂𝐵 − 𝜌𝑚𝑢𝑑 )𝑔𝑧𝑚 ,
where 𝜌𝑂𝐵 is the overburden density. For typical depths (2-2.5 km), rock density (𝜌𝑂𝐵 = 2600 kg m-3) and mud weight (𝜌𝑚𝑢𝑑 = 1050
kg m-3), Δ𝑃 ranges between 30 and 34 MPa. However, as magma ascends the well, Δ𝑃 will be reduced under the weight of the magma
column. The minimum overpressure due to buoyancy, when magma reaches the surface, is Δ𝑃𝑚𝑖𝑛 = (𝜌𝑂𝐵 − 𝜌𝑚𝑎𝑔𝑚𝑎 )𝑔𝑧𝑚 , where
density difference between magma and country rock is now explicit. For 𝜌𝑂𝐵 − 𝜌𝑚𝑎𝑔𝑚𝑎 ∼ 100 kg m-3, Δ𝑃𝑚𝑖𝑛 is between 2 and 2.5
MPa. This calculation does not account for magma overpressured above lithostatic, which would decrease 𝑡𝑎𝑠𝑐𝑒𝑛𝑡 , nor local depletion
of the magma pressure due to mass flow.
The calculation above assumes that flow in the pipe is laminar. Onset of turbulent flow in a pipe occurs when the Reynolds number,
𝑅𝑒 = 2𝑅 𝑣̅𝑧 𝜌/𝜇, exceeds ∼ 2300. For typical parameter values used here, the flow will transition from initially turbulent to laminar
in, at most, about one second.
For a standard 9-5/8” diameter geothermal borehole intersecting a rhyolite (𝜇 = 107 Pa s) at 2 km depth with average overpressure 40
MPa during transport, 𝑡𝑎𝑠𝑐𝑒𝑛𝑡 is about eight years. For the same well intersecting a basalt (𝜇 = 102 Pa s) at 2.5 km depth with average
overpressure 4 MPa, 𝑡𝑎𝑠𝑐𝑒𝑛𝑡 is about half a day.
𝑟2 𝑟2
𝑣𝑧 = 𝐹 ( 2 + 𝛼ℛ2 − 𝛼 − ℛ2 ), 𝑣𝑎 = 𝐹𝛼 ( 2 − 1). (5)
𝑅 𝑅
−Δ𝑃𝑅2
𝐹= , (6)
4𝜇ℎ
𝛼 is the ratio of the viscosity of the magma to that of the fluid in the annulus, and ℛ is the fraction of the borehole radius occupied by
rising magma:
(𝑅 − 𝛿)
ℛ= , (7)
𝑅
with 𝛿 being the thickness of the annulus.
The velocity given in Eq. (5) can be compared to that from Eq. (3) in each of the two described situations to determine the importance
of the annulus formation. For the first situation, we use an annulus viscosity of 10 -5 Pa s based on superheated steam at temperatures
similar to the magma temperatures at abovementioned geothermal fields (i.e. 900-1000oC). For the situation of a crystallizing annulus,
we consider the annulus viscosity to be much smaller than that of the magma and assume that 𝛼 = 0. Fig. 2 shows the multiplicative
factor that results from accounting for an annulus of either low (A) or high (B) viscosity fluid. This model suggests a thin annulus of
gas can allow the magma to rise at much higher velocities than if the magma directly contacted the borehole. In contrast, a crystallizing
annulus will reduce the rise velocity in a manner that is equivalent to the borehole reducing in size.
where Δ𝑇 = 𝑇𝑚𝑎𝑔𝑚𝑎 − 𝑇𝑟𝑜𝑐𝑘 , 𝜅 = 𝐾/𝜌𝑐 is the thermal diffusivity of the rock, 𝐾, 𝜌 and 𝑐 are the thermal conductivity, density and
heat capacity of the rock, and 𝐽0 and 𝑌0 are zeroth order Bessel functions of the first and second kind (Fig. 3A). Note, we have assumed
that rock and magma density are sufficiently similar to not warrant distinction here.
Figure 3: Temperature and heat flux evolution around a cylindrical borehole at fixed temperature. (A) Radial temperature
profiles at different times, as given by Eq. (8). (B) Radial heat flux at the wellbore (black; Eq. 9) and total heat loss
(red; Eq. 10) with time. The heat loss and time required to freeze a plug of rhyolite within a 9-5/8” wellbore is indicated
by red dashed lines (Eq. 11).
4Δ𝑇𝐾 ∞ −𝜅𝑢2𝑡 𝑑𝑢
𝑞= ∫ 𝑒 . (9)
2
𝑅𝜋 0 𝑢[𝐽0 (𝑢𝑅) + 𝑌02 (𝑢𝑅)]
2
If the magma is instantaneously emplaced in the wellbore at 𝑡 = 0, then at time 𝑡, the heat lost is given by the integral of Eq. (9):
𝑡
8Δ𝑇𝐾 ∞ 2 𝑑𝑢
𝑄 = 2𝜋 ∫ 𝑞(𝑡 ′ ) 𝑑𝑡 ′ = ∫ (1 − 𝑒 −𝜅𝑢 𝑡 ) 3 2 (10)
0 𝑅𝜋𝜅 0 𝑢 [𝐽0 (𝑢𝑅) + 𝑌02 (𝑢𝑅)]
Equating this to the latent heat of crystallization for the cylindrical magma volume in the borehole, 𝑄𝑓𝑟𝑒𝑒𝑧𝑒 = 𝐿𝑚 𝜌𝜋𝑅2 , gives an
expression for the time for borehole magma to completely freeze over:
2
∞
(1 − 𝑒 −𝜅𝑢 𝑡𝑓𝑟𝑒𝑒𝑧𝑒 ) 𝐿𝑚 𝜋 2 𝑅 3
∫ 3 2 2 𝑑𝑢 = . (11)
0 𝑢 [𝐽0 (𝑢𝑅) + 𝑌0 (𝑢𝑅)] 8𝑐Δ𝑇
Eq. (11) can be solved for 𝑡𝑓𝑟𝑒𝑒𝑧𝑒 either by numerical quadrature or simulation of the discretized energy conservation equation. For
magma at 800°C with latent heat of crystallization, 𝐿𝑚 = 2.9×105 J kg-1 (Caricchi and Blundy, 2015) in a 9-5/8” diameter borehole
in contact with rock cooled to 20°C with specific heat 10 3 J kg-1 K-1, 𝑡𝑓𝑟𝑒𝑒𝑧𝑒 is about 10 days (Fig. 3B).
The heat extraction, 𝑄, to freeze a cylindrical plug of height, Δℎ, from magma at its eutectic is:
If there is ample mixing of the borehole fluid, then one can assume that 1D linear heat flow through the plug, 𝑞, is controlled to first
order by an equilibrated conductive temperature gradient. This yields an integrodifferential equation that can be solved for plug
growth, Δℎ(𝑡):
𝑡
Δ𝑇 2𝐾Δ𝑇𝑡
𝑞(𝑡) = 𝐾 , 𝑄(𝑡) = 𝜋𝑅2 ∫ 𝑞(𝑡 ′ )𝑑𝑡 ′ , Δℎ(0) = 0, Δℎ = √ , (13)
Δℎ(𝑡) 0 𝜌𝐿𝑚
where Δ𝑇 = 𝑇𝑚𝑎𝑔𝑚𝑎 − 𝑇𝑚𝑢𝑑 . The frictional strength, 𝐹𝑓𝑟𝑖𝑐𝑡𝑖𝑜𝑛 , due to contact of the plug with the borehole wall will increase with
the growing contact area:
𝐹𝑓𝑟𝑖𝑐𝑡𝑖𝑜𝑛 = 2𝜋𝑅Δℎ
⏟ × 𝑓
⏟ ×⏟
𝜌𝑚𝑢𝑑 𝑔 𝑧𝑚 ,
contact friction normal (14)
area coef. stress
where 𝑓 is the friction coefficient and we have assumed that normal stress in the plug is in equilibrium with the overlying mud weight
at the time of freezing, i.e., the isotropic magma pressure is “frozen” in place. This model does not account for density and volume
changes due to the phase transition, i.e., as magma freezes, it contracts and will exert a smaller normal stress on the borehole. In
addition, vertical advection of the plug would reduce the axial stress applied by the overlying mud column, further reducing the radial
normal stress.
The frictional strength is compared against the maximum borehole overpressure the plug must resist if it is to arrest flow, which is
given by the difference between magma overpressure below and mud weight above:
Equating (14) and (15) and substituting (13) for Δℎ yields an expression for the time to freeze a large enough borehole plug that can
arrest magma ascent:
2
𝑅(𝜌𝑂𝐵 − 𝜌𝑚𝑢𝑑 ) 𝜌𝐿𝑚
𝑡𝑝𝑙𝑢𝑔 = ( ) . (16)
𝑓𝜌𝑚𝑢𝑑 8𝐾Δ𝑇
For a magma at 1050°C with 𝐿𝑚 = 2.9×105 J kg-1 and 𝐾 = 2.5 Wm-1K-1, in contact with 20°C drilling mud with density 1.05 g cm-3
in a 9-5/8” diameter borehole with friction coefficient of 0.6, 𝑡𝑝𝑙𝑢𝑔 is about 1 hour. This is about an order of magnitude too large to
account for magma arrest within “several minutes” as observed by Teplow et al. (2009) for magma ascent in KS-13.
The time scale analysis indicates that geothermal wells from slim hole to big bore (Garg and Combs, 1997; Bush and Siega, 2010)
are not large enough for viscous rhyolite magmas to ascend (decades to millennia) before freezing (hours to months) or plug formation
(minutes to days) occurs. Instead, a borehole would need to have a radius of 7.5 m or more for 𝑡𝑝𝑙𝑢𝑔 to reliably exceed 𝑡𝑎𝑠𝑐𝑒𝑛𝑡 and,
even then, it would take weeks to reach the surface. In fact, a natural analog to these ‘boreholes’ is the conduits that feed rhyolite
dome extrusions. For instance, during the 1995-97 Soufrière Hills eruption in Montserrat, a lava dome was extruded from an estimated
30 m diameter conduit (Melnik and Sparks, 1999) at an average rate of 2.1 m3/s (Sparks et al., 1998), which corresponds to 𝑡𝑎𝑠𝑐𝑒𝑛𝑡
of about 20 days from the 5 km deep magma chamber.
The implications for encountering basaltic magma while drilling are quite different. As its viscosity would be many orders of
magnitude lower than rhyolite, borehole ascent would be a lot more rapid. In particular, for large diameter wellbores, magma ascent
could occur on time scales of hours, which is comparable to the time required to freeze a plug.
To date, basaltic magmas have not been encountered during drilling and their occurrence at shallow (drillable) depths is probably
rare. In any event, were a basaltic magma to be inadvertently drilled, borehole transport may not be as simple as the description given
above. First, magma need not completely solidify for flow to become arrested, as its viscosity will tend to increase with crystallinity,
greatly slowing the rate of ascent. On the other hand, as the plug solidifies, it will tend to contract, reducing borehole friction. Other
processes, such as lubrication by along the borehole wall by superheated drilling fluid or volatiles exsolved from the magma, could
also reduce the frictional resistance of the plug. Addressing these processes would likely require sophisticated numerical simulators.
Figure 4: Comparison of 𝒕𝒂𝒔𝒄𝒆𝒏𝒕 for rhyolitic (blue) and basaltic (red) magmas across a plausible parameter range, along with
𝒕𝒇𝒓𝒆𝒆𝒛𝒆 (green) assuming a radial heat loss model, and 𝒕𝒑𝒍𝒖𝒈 (black) assuming frictional plug formation for the same
magmas. Both ascent and freezing times are strong functions of wellbore radius, and a typical range of geothermal
wellbores are shown (left, vertical dashed lines). Eruption is assumed where both green and black curves plot above
the red or blue curves. The two righthand plots show the wellbore radii required to “erupt” basalts and rhyolites.
Inferred magma ascent during the Námafjall eruption (0.5 to 1 min) is much more than the 30 mins to 1 day expected for basalts in
6 in boreholes. However, the inferred ascent rate was based on observations of the eruption column and this only constrains magma
velocity in the upper section of the borehole. If significant acceleration occurred during ascent, e.g., due to fragmentation or degassing
processes, the ascent rate would be overestimated. Fault movement and earthquake activity in the Námafjall area was noted up to one
hour before the eruption (Larsen et al., 1979) and it is possible that magma entry to the borehole occurred at this time. Alternatively,
faster than expected rise could occur if a lubricating volatile annulus formed (Fig. 2).
4. CONCLUSION
We have developed models to describe the physics of magma ascent and arrest in a borehole, and to characterize time scales of the
system. Magma ascent is conceptualized as a balance between viscous drag and driving overpressure, which yields the Hagen
Poisueille equation for pipe flow. Magma viscosity is shown to be the primary constraint on magma ascent, with rise times of years
to centuries for high viscosity rhyolites and as low as minutes to hours for low viscosity basalts. However, these could be dramatically
reduced in the event that a low-viscosity annulus of degassed volatiles forms.
Erupting magma from a borehole is a thermal survival problem, with both radial cooling and plug formation through drill fluid contact
tending to freeze and arrest the rising column. Plug formation was formulated as a steady heat flow and crystallization problem and,
for feasible parameter ranges, is nearly always more rapid that the ascent time scale. Only for the very particular case that drilling
encounters a shallow, pressurized basalt, does the prospect of an eruption become feasible.
ACKNOWLEDGEMENTS
The authors thank Susan Ellis for reviewing an early draft of this manuscript.
REFERENCES
Bush, J., and Siega, C.: Big Bore Well Drilling in New Zealand – A Case Study, Proceedings, World Geothermal Congress, Bali,
Indonesia (2010).
Caricchi, L., and Blundy, J.: Experimental petrology of monotonous intermediate magmas. Geological Society Special Publications
422, (2015), 105-130.
Carslaw, H. S., and Jaeger, J. C.: Conduction of heat in solids. Oxford: Clarendon Press, (1959), 2nd ed.
Cladouhos, T. T., Petty, S., Bonneville, A., Schultz, A., and Sorlie, C. F.: Super Hot EGS and the Newberry Deep Drilling Project,
Proceedings, 43rd Workshop on Geothermal Reservoir Engineering, Stanford, California (2018).
Elders, W. A. et al.: Origin of a rhyolite that intruded a geothermal well while drilling at the Krafla volcano, Iceland. Geology, 39,
(2011), 231-234.
Fedotov, S. A., Sugrobov, V. M., Utkin, I. S., and Utkina, L. I.: On the Possibility of Using Heat Stored in the Magma Chamber of
the Avachinsky Volcano and the Surrounding Rock for Heat and Power Supply. Journal of Volcanology and Seismology, 1,
(2007), 28-41.
Friðleifsson, G. O., and Elders, W. A.: The Iceland Deep Drilling Project: a search for deep unconventional geothermal resources.
Geothermics, 34, (2005), 269-285.
Friðleifsson, G. O., Pálsson, B., Albertsson, A. L., Stefánsson, B., Gunnlaugsson, E., Ketilsson, J., and Gíslason, Þ.: IDDP-1
Drilled Into Magma – World’s First Magma-EGS System Created, Proceedings, World Geothermal Congress, Melbourne,
Australia (2015).
Garg, S. K., and Combs, J.: Use of slim holes with liquid feedzones for geothermal reservoir assessment. Geothermics, 26, (1997),
153-178.
Helz, R. T., and Wright, T.: Drilling report and core logs for the 1981 drilling of Kilauea Iki lava lake, Kilauea volcano,
Hawai’i,with comparative notes on earlier (1967-1979) drilling experiences. USGS Open-file Report 83-326, (1983).