Review of Smelt Water Explosions

advertisement
i^zt/yTj'S/ r, m- Ace
h
,-
C-?-,,
f'r
nrY-
t" AP 19 1982
INFORMATiON SERVICES
TECHNICAL FILES
Smelt
Review of
Water Explosions
A SPECIAL REPORT
by
Philip E. Shick
and
Thomas M. Grace
""~S~
R~~~~I~
THE INSTITUTE OF PAPER CHEMISTRY
Office Box 1039
Appleton, Wisconsin 54912
Phone: 414/734-9251
~ :~,~ ~Post
~l
March 26, 1982
TO:
Members of The Institute of Paper Chemistry
REVIEW OF SMELT-WATER EXPLOSIONS
A SPECIAL REPORT
Attached is a Special Report entitled "Review of Smelt-Water Explosions."
The report was prepared for the American Paper Institute's Recovery
Boiler Research and Development Subcommittee by Dr. P. E. Shick (OwensIllinois, Inc.) and Dr. T. M. Grace (The Institute of Paper Chemistry).
Because the subject is one of considerable interest to the paper industry,
and relates directly to current research at the IPC, the Institute has
undertaken to publish the review and distribute it to its membership.
The report covers the current state of understanding of vapor explosions.
New theories are developed to help explain smelt-water interactions.
Implications for recovery boiler safety are discussed. I am sure you will
find this an interesting report.
If you have any questions or comments, please contact the co-authors of
this review.
Sincerely yours,
Earl W. Malcolm
Director
Division of Chemical Sciences
EWM/hmf
Enclosure
1043
East
South
River
Street
THE INSTITUTE OF PAPER CHEMISTRY
Appleton, Wisconsin
REVIEW OF SMELT-WATER EXPLOSIONS
Project 3473-2
Philip E. Shick, Senior Research Scientist
Owens-Illinois Inc., Toledo, OH 43666
Thomas M. Grace, Senior Research Associate
The Institute of Paper Chemistry, Appleton, WI 54912
January, 1982
FOREWORD
This review was jointly prepared by Dr. P. E. Shick, Senior Research Scientist,
Owens-Illinois Inc., Toledo, Ohio, and Dr. T. M. Grace, Senior Research Associate,
The Institute of Paper Chemistry, Appleton, Wisconsin, acting as a two-man task
group on emerging theories of smelt-water explosions for the American Paper
Institute Recovery Boiler Research and Development Subcommittee.
It is being
published by The Institute of Paper Chemistry and will. be distributed to Members of
The Institute of Paper Chemistry and API Recovery Boiler Committee as well as to
other interested parties.
TABLE OF CONTENTS
ABSTRACT
v
INTRODUCTION
1
VAPOR EXPLOSIONS
3
Stable and Metastable States
Superheat Limit
.
3
3
.
Film Boiling
4
Mixtures
4
5
Physical Properties
-
Temperature Distributions
6
Nucleating Agents
8
Chemical Reactions
8
Conditions of Contact
9
Size of the System
10
Ambient Pressure
11
Energy Release
11
EXPLOSION MECHANISMS
14
Background
14
Fragmentation
15
20
Fragmentation Summary
21
Superheat
24
Superheat Summary
Detonations
24
28
Detonation Summary
Present Status
Initial Configuration
28
.29
Triggering
30
Escalation
30
-iv-
Propagation
SMELT-WATER EXPLOSIONS
31
32
Dissolving Tank Studies
32
Furnace Explosion Studies
34
Smelt-water Research Group
34
Battelle Review
39
Arthur D. Little Review
40
Battelle Study
44
Other Work
47
Revised Mechanism
49
Continuing Experiments
50
Swedish Work
50
Current IPC Studies
52
Safe Firing Concentration
53
CONCLUSIONS
54
REFERENCES
55
REVIEW OF SMELT-WATER EXPLOSIONS
Philip E. Shick, Senior Research Scientist
Owens-Illinois Inc., Toledo, OH 43666
Thomas M. Grace, Senior Research Associate
The Institute of Paper Chemistry, Appleton, WI 54912
ABSTRACT
The current state of understanding of vapor explosions, especially those involving
contact of two liquids is reviewed..
Using this as background, all of the previous
work on the sme.lt-water system. is reviewed. Theories developed for other liquidliquid systems..appear applicable to smelt-water. Explanations are given for previously unexplainable smelt and water composition effects. The effects of the scale
of the system on explosion intensity and the implications on recovery boiler safety
are discussed. Directions of. ongoing work are indicated.
INTRODUCTION
The API Recovery Boiler R&D Subcommittee has a task group on emerging theories of
smelt-water explosions. This task gr6up carried out an extensive review' of the
literature relevant to smelt-water explosions, which is presented in this report.
The major objective of the review is to summarize the current state of knowledge of
smelt-water explosions.
A related objective is to aid the further development of
explosion theory by assembling the important information and observations on smeltwater systems and similar systems. Answers are sought to the following questions:
1.
What are the underlying phenomena. involved in smelt-water explosions?
2.
What elements are necessary for an explosion?
3.
What factors govern the extent of violence?
4.
Can the existing body of experimental observations on smelt-water
explosions be rationalized?
There has been a tremendous amount of work done on similar explosions in other
industries since the last major study of smelt-water explosions, and a very extensive literature exists.
Reid (1,2) has published two very complete literature surveys. Other reviews or papers giving extensive bibliographies include references
(3-14).
When a volatile liquid expands into vapor at or above its boiling point, thermal
energy is converted into mechanical energy. When such an expansion is inertially
limited and is propagated at sonic or supersonic velocities, we have a so-called
vapor explosion.
Such explosions can occur in many ways.
Thus, Reid (1) closes his
"When you spit
1977 lecture on superheated liquids with an old blacksmith's adage:
on a piece of red hot iron, it dances around in a little ball. If you hit the ball
with your hammer, the explosion will throw the hammer over the barn and blow a hole
in the iron."
Another case is a so-called boiling liquid expanding vapor explosion (BLEVE), such
as occurred at 12:10 am on November 11, 1979, at Mississauga, Ontario (15).
This
explosion blew up a propane tank car, showering the surrounding area with chunks of
metal.
It knocked fire fighters to the ground and..broke windows half a mile away.
BLEVE's in two more propane tank cars hurled the- cars as far as 700 yards.
According to Reid (16), such explosions may be.anticipated when the ambient pressure
over a heated liquid falls to a pressure corresponding to the superheat limit of the
liquid. The violence of high pressure explosions and some volcanic action might
also be attributed at least in part to the boiling liquid expanding vapor explosion
process.
Still another form of vapor or superheated liquid explosion is represented by the
so-called liquid-liquid explosions, which occur under some circumstances when a hot
liquid is brought into contact with a cold volatile liquid. Such explosions have
been experienced when molten steel or aluminum have been spilled into water (17) and
have also been produced under laboratory conditions when certain liquefied petroleum
mixtures are spilled onto water (18).
It is considered probable that a liquidliquid, molten fuel-coolant interaction occurred during a transient test of the
experimental Spert 1-D nuclear reactor (19), in which a destructive pressure pulse
greater than 20.7 MPa (3000 psi) occurred 15 milliseconds after initiation of a
-2-
power excursion. Explosive interactions with water of molten smelt produced in the
chemical recovery of pulping chemicals represent a special case of such a liquidliquid explosion. These various forms of vapor explosions have many features in
common, but also important differences.
-3VAPOR EXPLOSIONS
Vapor explosions occur only when the vaporization process is extremely .rapid. They
are a highly nonequilibrium process and can occur in either of. two ways:
1.
2.
Heat transfer occurs before the vaporization process, creating an
unstable, superheated liquid which explosively flashes'into vapor-.
'
Heat transfer and vaporization occur simultaneously at very rapid .
rates due to intense mixing or radically enhanced heat transfer
mechanisms.
Any given system might involve either or both of these processes.
Stable and Metastable States
To interpret and perhaps to understand liquid-liquid vapor explosions, it is
necessary to consider the fundamental properties of the liquids involved.
Reid (1) discusses the physical states of a liquid. A plot of equilibrium vapor
pressure vs. temperature defines conditions under which a liquid may exist in
equilibrium with its saturated vapor up to the critical temperature, at which point
the vapor and liquid become indistinguishable and only one phase can exist. When-'
ever a liquid is heated above its boiling point, or the ambient pressure is lowered
below the vapor pressure at a given temperature, the liquid is said to be
superheated. This is a not uncommon situation, leading to such phenomena as bumping
when heating fluids in clean glassware.
Superheat Limit
There is a limit to the temperature to which a fluid may be heated above its boiling
point, whereupon spontaneous nucleation will occur, (1) and the energy stored as
sensible heat above the boiling point will be converted to latent heat and/or to
mechanical energy. 'This limit is pressure dependent and is equal to the critical
temperature at the critical pressure.
For lower pressures, it can be estimated from
a thermodynamic equation of state for the liquid as the locus of points where the
partial differential of pressure with regard to volume at constant temperature is
equal to zero. This is equivalent to the point of maximum stress in the stress
strain diagram of a tensile test. The superheat-limit temperature can also be estimated from the statistical probability of forming a sufficient concentration and
size of bubble nuclei in a given liquid. This probability is an extremely strong
function of temperature, increasing by several orders of magnitude for a few degrees
temperature change. Homogeneous nucleation is a negligible factor until a liquid
comes within a degree or so of the superheat limit temperature (SLT).
At the SLT,
however, an enormous number of nuclei are formed and the liquid can vaporize explosively.
The superheat-limit temperature may also be approached by various experimental
methods, one of which is to heat a small droplet of fluid suspended in another
fluid, which wets it and thus provides no nucleating sites.
For most fluids, the
theoretically estimated and measured values of superheat-limit temperatures at
atmospheric pressure correspond to approximately 0.9 times the critical temperatures
expressed on the absolute scale. The handbooks give the critical temperature of
-4-
water as 374°C (705 0 F) at 22.1 MPa (3206 psi).
The superheat-limit temperature for
water at atmospheric pressure has been estimated by the statistical approach as
approximately 310°C (590°F), whereas a temperature of 279.5°C (535°F) has been
achieved experimentally by Apfel (20), using the heated drop method.
Theoretically, the tensile strength of;-a liquid, i.e.., the cohesiveness of a liquid
which prevents a column from breaking or forming a bubble under tension or negative
pressure, is simply an extension of the superheat-limit temperature curve. By subjecting droplets of ether suspended in glycerine to acoustic stress over a range of
temperatures, Apfel (21) demonstrated such a relationship from a superheat-limit
temperature of approximately 147°C at one atmospheric pressure down to 127°C and a
negative pressure of approximately 16 atmospheres.
Film Boiling
Another important phenomenon in the mechanics of vapor explosions is film boiling.
In the conventional boiling process nucleate boiling.occurs.with many small bubbles
forming at the surface between the liquid and the heat source, and rising in the
fluid. This process continues as the. surface temperature is increased until a critical heat flux is reached at which nucleate boiling becomes unstable. As the temperature is increased, an.unstable, transition region is.encountered, and the average
heat flux decreases. Finally, a point is reached (called the Leidenfrost point) at
which heat transfer becomes stable again at a minimum heat flux. In this region
(temperatures at or above the Leidenfrost point) a stable vapor film separates the
volatile liquid from the heat source, and the heat transfer takes place across the
vapor film.
The Leidenfrost point depends upon the degree that the liquid wets the solid surface.
If wetting is complete, it should approach the SLT of the liquid. There is a
relationship between the SLT and the onset of film boiling (13,22).
If a fluid is
being heated from a surface, temperature gradients will exist in the fluid and the
hottest fluid will be in the thin boundary layer next to the hot surface. When the
surface is above the SLT, spontaneous nucleation will occur in the boundary layer,
resulting in the formation of a vapor film.
Mixtures
Reid (1) notes from the theoretical calculations of Hirshfelder (23), Beegle et al.
(24), and the experimental results of Porteous and Blander (25) that the superheatlimit temperature of a mixture of similar liquids approximates a mole fraction
average of the superheat-limit temperatures of the pure components.
We all know that a heated liquid tends to release dissolved gases. Lienhard (26)
has illustrated by calculations the effect of such a gas in reducing the equilibrium
vapor pressure and thus the superheat required for nucleation in boiling.
Using the heated drop method of estimating superheat-limit temperature and working
with solutions of carbon dioxide in Freon 22, propane, and isobutane, Mori et al.
(27) found that the presence of the dissolved gas shifted the superheat-limit curve
to lower temperatures, with the new limits for a given concentration of dissolved
gas essentially parallel to the original superheat-limit curve and actually crossing
the pure liquid saturation curve of pressure plotted against temperature. Forest
and Ward (28) obtained similar results with nitrogen dissolved in ethyl ether. They
also noted that as the gas content increased, the pressure pulse when nucleation
-5-·
occurred became much less noticeable.' Thus the presence or- release of a fixed gas
not only lowers the superheat-limit temperature at a given pressure,' but also
decreases the violence when the superheat 'is,relieved.
'
By analogy, it might be expected that the presence of a nonvolatile solute would
raise the superheat-limit temperature of a liquid. We have so far found no data on
the SLT of such solutions. However,'the following might be indicative.
Yayanos (29) has shown that data'on the compressibility of both water and sodium
chloride solutions can be plotted against each other as straight lines. Extrapolation of these lines to the volume corresponding to maximum stress gave a predicted
negative pressure for the tensile strength of pure water at 25°F of -2273
atmospheres. For a 25% sodium chloride solution at 25°F, the negative pressure
would be approximately 68% greater than that for pure water. Using this point, if
the superheat-limit temperature curve of a 25% salt solution were shifted to higher
temperatures, essentially parallel to the superheat-limit temperature curve of pure
water, this would correspond to an increase of approximately 213°C (384°F) in the
superheat-limit temperature to approximately 523°C (974°F) at atmospheric pressure.
With regard to salt solutions, when Long (17) quenched molten aluminum at 750°C
(1380°F) in water, there were no explosions, provided the water temperature'was
above 60°C (140°F), but explosions were obtained using a 15% sodium chloride solution for quenching at a temperature of 78°C (172°F). On the other hand, it might be
noted that Buxton et al. (30) did not find a difference between salt water or
borated water and deionized water in flooding tests of small arc-melted samples of
iron oxide and cobalt oxide, which required short-duration pressure transients to
initiate explosions.
'
In further consideration of mixtures of materials which may be involved in a liquidliquid or other type of vapor explosion, it should be noted that the melting point
or freezing point of mixtures, particularly eutectic mixtures, is frequently significantly lower in temperature than any of the pure components. It should also be
noted, as most organic chemists would attest, that it is more difficult to
crystallize an impure material than a pure material. Thus, we might anticipate a
lower melting point and greater supercooling when the hot liquid is a mixture of
compounds.
Physical Properties
Practically all other physical characteristics must be considered as potentially
important to the interaction of two liquid materials to cause a vapor explosion.
These involve both the properties important to the heat balance and the energy
available, as well as those which influence the ease and rapidity of mixing. A list
of such properties follows:
-Density,
Surface tension, interfacial surface tension and wetting,
Solubility or miscibility,
Viscosity,
Specific heat and thermal conductivity,
Heat of vaporization of the cooler liquid,
Heat of fusion of the hotter liquid (if freezing can occur),
Heat of solution (when a solution can be formed),
-6-
Speed of sound and. relative speeds of sound .(as related to
shock wave propagation and shearing effects-),
Tensile strength and thermal expansion of solids (as
related to possible encapsulation and thermal shock).
As noted above, the superheat-limit temperature is a function of the surface tension
of a liquid. Long (17-) found that wetting agents tend to suppress explosions in the
quenching of molten aluminum and water. Thus 0.01% of wetting agent lowered the
threshold temperature of the quench solution from 60°C (140°F) to 46°C (64°F),
whereas with a 0.5% solution no explosions were.obtained at any temperature down to
the freezing point of water.
Skripov et al. (31) found no discontinuities in the measurements of density, sound
velocity, viscosity, and heat conductivity in passing from the normal state to a
superheated state.
The importance of physical properties, particularly as they may affect mixing, is
illustrated by an example given by Reid (2) that.even in a spill of a sensitive
liquid.ethane-butane mixture on water, substitution of a very dilute water-agar gel
with essentially the same thermal.-.properties as liquid water prevented a vapor
explosion. Also, Flory et al. (32) found that a five-fold increase in the viscosity
of water by the addition of carboxyme.thyl cellulose greatly reduced or totally prevented fragmentation of molten lead, tin, and bismuth.
Temperature Distributions
The preponderance of evidence indicates that in a liquid-liquid system, the temThus, if the interface temperature is signiperature at the interface is critical.
ficantly below the freezing point of the hot liquid, it may be expected to freeze,
providing a solid surface which could prevent developing a significant superheat in
the cooler liquid and might also interfere with subsequent mixing of the two
liquids. However, if this transformation took place after a film of the cooler
liquid had developed significant superheat, it could be the trigger for a liquidliquid explosion. This might.be the case for injection of molten tin or lead-tin
alloys at 260°C (500°F) into 24°C water, which fragments violently according to
Witte et al. (5), even though the interface must be below the 310°C (590°F)
superheat limit of water.
Interface temperatures between the boiling point and either the superheat-limit temperature or the Leidenf.rost temperature of the cooler liquid would permit development of a superheated layer in the. cooler fluid, since nucleation would tend to be
suppressed at a liquid-liquid interface.
An illustration of the possible. importance of these temperatures is cited by Reid
(2). When liquid propane with asuperheat-limit temperature of 53°C and initially
at -42°C is poured onto water below 53°C, boiling.results with the formation of surFor initial water temperatures..between 53°C and 71° C there is a high
face ice.
probability of a vapor explosion. However., with water above 71°C quiet boiling
occurs with only an occasional "pop."
A common method for calculating the interface temperature is that given in Carslaw
and Jaeger (33) for the time-independent contact temperature between two infinite
slabs of material, each initially at a uniform temperature.
--7TH + YTc
Ti'-
1 +ty
kHPHCH 1/2
¥ = kI
,
I
where k, p, and C are the thermal conductivity, density and specific heat of the hot
and cold fluid.
Fauske (34) has pointed out that although the initial interface temperatures may be
estimated by this equation, both the bulk temperature and the interface temperature
of a small droplet of one fluid'suspended in a large mass of the other will with
time approach the temperature of the surrounding fluid.
A few other examples of the importance of temperatures might be cited.
Williams
(35) notes, "If the hot liquid freezing.point is below the temperature for the onset
of transition boiling, film collapse and propagating interaction are likely to occur
spontaneously. If the hot liquid solidifies at much higher temperatures, high film
stability may require substantial triggering pressure pulses to force liquid-liquid
contact before freezing occurs, and. to produce interactions generating sufficient
pressures to continue propagation."
Water poured on the surface of molten aluminum boils without exploding; however, if
water is impacted upon. the surface of the molten aluminum, violent explosions can be
produced. In such shock tube experiments, Wright and Humberstone (36) impacted
water on the surfaces of solid and molten aluminum in a 1-cm-thick pool at temperatures from room temperature to 950°C (1742°F). A water hammer impact pressure
of about 14 MPa (200 psia) was observed on the solid surface. The peak pressure
increased essentially linearly with temperature above the melting point of 650 ° C
(1202°F) to 200 MPa (2900 psia), suggesting that only the sensible heat of the
molten aluminum was available to produce high-pressure steam and that neither latent
heat nor the sensible heat of the frozen aluminum contributed appreciably in this
system.
Just as Long (17) found an upper water temperature limit for explosions when molten
aluminum was poured into water, Reynolds et al. (37) found explosions limited to a
definite temperature range for both water and tin when molten tin was dropped into
the water. Henry and McUmber (38) found a similar relationship for the initial temperatures.of Freon 22 and mineral oil to give a definite range of interface temperatures below which or above which explosive interaction did not occur when the
subcooled liquid Freon was poured into mineral oil. On the other hand, Buxton and
Benedick (39) did not note a significant difference in the explosivity or violence
of explosions when a thermite reaction product consisting of molten iron and molten
alumina at a temperature of approximately 2727°C (4940°F) was poured into water at
the boiling point as compared with water at room temperature.
The lower temperature limit for the hot fluid has been ascribed to interface temperatures being below the SLT or to the hot fluid freezing at the interface. The
upper temperature limits for both fluids are usually ascribed to increased likelihood of film boiling.
Henry and Fauske (40) have proposed that a necessary condition for vapor explosions
is for the interface temperature to be above the SLT. This seems to fit the data on
-8-
liquid hydrocarbon vapor explosions quite.well (41). However, the criterion does
not appear to fit.with the results of tests on the U02 - Na system, as pointed out
by Anderson and Armstrong (42). The interface temperature between molten U02 at
300 0 °C and 400°C-molten Na is only 1150°C, whereas the homogeneous nucleation temperature is 2050°C. Yet explosive interactions..have been observed in small-scale
tests on this system.
Nucleating Agents
It is often suggested that the addition of nucleating agents to either of the fluids
might prevent superheated liquid explosion.
Reid (16) cites work by Buivid and Sussman (43) which suggested that the superheatlimit temperature for liquids was not greatly affected.even when the liquids were
loaded with suspended hydrophobic or hydrophilic particles; possibly because of the
relatively small number. of-:nucleation sites provided in comparison with those.
needed. He..suggests for further consideration essentially colloidal dispersion of
particles less than 1 micrometer'in size, as proposed by Shanes (44). Anderson and
Armstrong (42) also mention that adding. a powder to water failed to prevent explosions in LNG-water tests.
Chemical Reactions
It has been clearly demonstrated that chemical reactions are not necessary for
liquid-liquid or other:types of vapor explosions. However, in some cases, chemical
reactions may add thermal..energy.and. also gaseous products, which could increase the
Evidences for chemical reactions in connection with
violence of the reaction.
liquid-liquid explosions have been reported. in numerous cases. Thus, Harrison. and
Ivins (45) reported ignition with..violent production of aluminum oxide and hydrogen
when reactor fuel plates containing aluminum were heated to temperatures on the
order of 1400 to 1500°C (2552 to 2732°F) in contact with water at atmospheric
pressure.
Hess and Brondyke (46) in experiments with molten aluminum and water measured temperatures at the container bottom. These were more than 214°C (385°F) higher than
the initial temperature of the aluminum, which was 768°C (1415°F), when explosions
took place in an uncoated steel container. They also obtained particularly violent
explosions, as well.as production of large quantities of aluminum oxide particles
when the container was coated with calcium hydroxide. They suspected the possibility of a thermite-type reaction between the aluminum and an oxide or hydroxide
followed by secondary reaction between aluminum and water as a trigger for their
explosions.
Reaction products.may also have different physical properties, thus altering the
limits of the system. Thus Flory et al. (32) suggested that an oxide layer on
slightly oxidized copper might have acted as an insulation between water and molten
copper, which melts at 1087°C (1989°F).
According to the circumstances, gases given off by chemical reaction might be
expected to maintain film boiling and prevent an explosive interaction or to provide
the nucleation to initiate a release of superheat and. an explosion. Buxton and
Benedick (39) noted the evolution of hydrogen from the reaction of molten iron
during the 1 to 3 seconds' nonexplosive initial contact of a thermite-melt with
water. Corradini (14) presents calculations ..
to show how the presence of such a gas
would prevent film collapse, and thus could-explain the difficulty in producing
explosions with a high iron content melt as. opposed to an iron oxide melt in the
presence of water.
.
*
It might also be noted that Segev (10) demonstrated that the introduction of a non-,
condensible gas between two fluids in a shock tube system decreased the rate of
pressure buildup and reduced maximum pressures obtained.
Conditions of Contact
The manner or conditions in which a hot fluid and a cold volatile fluid are brought
into contact may have a decisive effect on whether or not explosions are produced.
The relative and absolute masses of two materials which are brought into contact or
mixed may be expected to determine final equilibrium temperatures and vapor volume.
Thus, if a small quantity of the hotter 'liquid is dispersed in a very large volume
of the cooler liquid, the total heat of the hotter material could be absorbed as
sensible heat in the cooler liquid, without any generation of vapor.
On the other
hand, if a small quantity of the cooler liquid is dispersed in a large quantity of
the hotter liquid, it would be totally converted into superheated vapor. However,
the total violence will be limited by the amount of liquid being vaporized.
The mode and energy with which two phases are initially brought together may have a
profound effect on their apparent explosivity. Thus, Anderson and Bova (7) claimed
explosions when they injected a few grams. of.. water under. the surface of 80 grams of
molten sodium chloride; no-explosions were produced using the reverse geometry with
a small-diameter jet of the hot liquid penetrating into a bulk of cold liquid.
Asher et al. (47) did not obtain explosive interactions of water injected into
molten tin above 450°C, whereas'Reynolds et al. (37) obtained explosive interactions
at tin temperature up to 100.0° C, in the dropping mode. As noted by Reid (2),
pouring of liquid propane at the right temperature on the surface of water leads to
vapor explosion; however, pouring of water on a surface of molten aluminum does not.
Nelson and Buxton (8) did not obtain explosions when melts of stainless steel, aluminum and silver, as well as melts of "metallic" and "oxidic" reactor core simulants
were quickly but gently flooded with water, but did obtain explosions with the
"oxidic" materials when the system was subjected to a shock wave from an exploding
bridge wire. On the other hand, Buxton and Benedick (39) obtained spontaneous
explosions in most of the cases in which a molten thermite product was poured or
dropped into water. They speculated that the trigger mechanism for those explosions
may have been the momentum of the metal slug, which would cause the vapor film to
collapse in front of the slug, or in some cases contact with the walls or bottom of
the containers. Anderson and Bova (7), in studying water-molten salt systems, had
difficulty in producing large.'explosions when the water was injected from above the
surface into the molten salt, presumably because of entrained air. This was overcome by reducing the velocity of the jets as well as by introducing the water jet
below the surface of the smelt.
Even here it was found necessary to insulate the
tip of the injection needle to prevent premature nucleation and boiling. Their
experiments with glass spheres of'water, which were broken either by internal
pressure or by contact with the bottom of the crucible, failed to give explosive
interactions (possibly because the glass surface would promote nucleation).
In
other cases, as described by Porteous (25) for liquid ethane and by Wright and
-10-
Humberstone (36) for aluminum and water-, explosions can be brought about by
impacting one fluid upon the other. Segev (10) reports an extensive study of such
shock tube experiments in which the pressure. impulses obtained are compared with
those to be expected. from simple water hammer. effects. For the more reactive
systems, the yield is greater than that to be expected from simple water hammer and
tends to increase with each bounce as the one fluid tends to bounce on the other,
with a progressive mixing at the interface.
After the initial conditions of contact, the systems may also be subjected to
further shocks, either spontaneous or induced, such as from the use of exploding
Maeshima, as
bridge wires or explosive detonators by Nelson and Buxton (8).
reported by Bankoff et al. (48), describes the collapse of film boiling by a sudden
pressure rise for pentane drops on hot oil.
Size of the System
It appears that the gross scale of the system is important to both the mechanism and
relative violence of the explosions. Williams (35), considering the acoustic wave
propagation from a small interaction in a mixed two-fluid system as a function of
the system size, estimated both.the acoustic relief times and the explosion
pressures. He points out that "for many reasonable values of the parameters, the
model predicts only modest interaction will be observed in small-scale experiments,
yet efficient explosions can occur in large-scale accidents."
Segev (10) quotes from a.paper by Henry and Cho (49) which predicts only small-scale
explosions when the temperature. at. the liquid-liquid interface is below the spontaneous nucleation (Leidenfrost) temperature, with. nucleation of individual drops of
cold liquid entrained. in the hot liquid (incoherent nucleation); but which predicts
the potential for large-scale explosions if the interface temperature is above the
spontaneous nucleation temperature to give. stable film boiling (whether the cold
liquid is dispersed in the hot liquid or the. hot in the cold), so that a shock wave
can collapse the stable film boiling to give extensive (coherent) nucleation. Henry
and Fauske (12) thus rationalize the observed explosions in the U02 -Na system as
incoherent, small-scale explosions.
Buxton et al. (30,39).describe experiments in which from 1- to 27- kg quantities of
iron/alumina melts were poured into an open tank of water. They found "that the
most efficient explosions occurred for full tanks of water, which.is believed to be
a tamping effect."
With such considerations in mind, Board and Hall (9) quote a private communication
from Jakeman "that materials which fragment readily on a small scale (e.g.,
tin/water, sodium/U02) appear to be just those which do not interact violently on a
this seems plausible in view of the stability requirements for the
large scale:
pouring mode of contact;" however, they go on to caution, "stability is known to
depend strongly on the conditions - so that a wide range of conditions must be
investigated if such general arguments are to be substantiated--."
Bergman and..Laufke.(50) have suggested .a.s.imilar .effect.of size in work with smeltwater systems.
-11-
Ambient Pressure
Peckover et al. (51) consider the effects of pressure upon underwater vulcanism.
They note that at depths of greater. than about 2 km, where the pressures correspond
to the critical pressure of water (216 bar), there is only one water phase, and a
quiet effusion of lava would be expected. However, based upon unpublished work by
Buchanan which considers a bubble collapse mechanism for mixing the hot and cold
liquids,'they estimate that this. threshold depth may be much less, i.e., on the
order of 130 m, if homogeneous nucleation occurs at the water-lava interface, and
approximately 700 m for heterogeneous nucleation.
In discussing a boiling droplet fragmentation mechanism, which would be sensitive to
the ambient pressure, Henry and Fauske (12) refer to the work of Henry and McUmber
(52) who found that when Freon was dropped on mineral oil, explosive interactions
were obtained at one atmosphere ambient pressure over an oil temperature range of
approximately 135 to 170°C with peak interaction pressures of about 22 atmospheres;
no explosive interactions were obtained over a much wider temperature range from
100°C to about 240°C when the same experiments were carried out either at 2.2
atmospheres or 8 atmospheres ambient pressure. The Henry and Fauske model also predicted that explosive interactions with water should cease at about 0.9 MPa, which
seems to be confirmed by work reported by Nelson and Buxton (53), who obtained
explosive events for system pressures less than or equal to 0.5 MPa but no explosions at a pressure of 0.75 MPa, working with simulated oxidized reactor core
material, arc-melted, flooded with water and triggered by a submerged exploding
wire.
Segev (10) in his shock tube experiments found that thermal interaction of either
water or butanol with Woods metal could be prevented by an increase in the ambient
pressure. With water and molten salt at 600°C no explosive thermal interaction
occurred, but rather slow vaporization, under an ambient pressure of 1.35 MPa (196
psi).
Henry et al. (54) reported on a test series in which two kilograms of molten salt
was poured into a pool of water using a factorial type experiment with molten salt
temperatures from 850 to 1200°C, water temperatures from 20 to 90°C, and with
ambient pressures of 0.1, 0.5, 1, 2, 4, and 5 MPa, and using dropping, submersion
and injection modes of contact. No vapor explosions were experienced at ambient
pressures of 0.5 MPa (72 psia) or higher.
However, Corradini (14) states, "If the magnitude of the external trigger is sufficiently increased, a coherent vapor explosion can once again be induced. This has
been experimentally observed by Kottowski and Sharon in shock tube geometries for
simulant fuel-coolant systems."
They (55) obtained energy conversion ratios of 1 to
2% in a steel/water system with an ambient pressure of 2.6 MPa when the impacting
water column pressure was increased from 4 to 20 MPa.
Energy Release
A liquid-liquid explosion involves the conversion of the thermal energy in the hot
fluid into the pV energy in the blast wave. The relevant questions on energetics
are:
-12-
1.
How much of the thermal energy present ends up in pV work?
2.
What parameters are important in determining.the explosive energy?
Anderson and Armstrong (42) analyzed-the damage.patterns from two incidents: a'
foundry accident in which about 10.0 lb of steel was dumped into a drainage gutter
containing 625 lb of water, and a destructive transient at the SL-1 experimental
nuclear reactor. They estimated that about 13% of the energy in the initially
spilled steel (about 40% of the maximum thermal. energy theoretically convertible to
work) was delivered to the shock wave. In the SL-l incident, 10-15% of the total'
available thermal energy was converted into destructive work. This indicates that
large-scale vapor explosions can be fairly efficient.
A common method of calculating the maximum potential work from vapor explosions is
the Hicks-Menzies approach (56). A unit mass .of.the hot liquid is mixed with a specific mass of cold liquid and an equilibrium temperature is reached in a constant
volume process. The system is..then expanded, maintaining equilibrium until the
final temperature is reached. An alternative procedure is to assume the expansion
proceeds adiabatically (no further heat transfer between the two fluids). This
gives a smaller estimate of the maximum possible work.
Anderson and Bova (7) estimated the pV work done in small-scale experiments when
water was injected under various conditions into molten sodium chloride. They
obtained as little as 0.0035% of the theoretical maximum work (0;0088% of the
theoretical maximum adiabatic work.) and up to as high as 14% of the theoretical
maximum work.(36% of the theoretical maximum adiabatic work).
The above analyses neglect the freezing of the hot fluid. There is some evidence
that only the sensible heat in the hot liquid..is available for vaporization. In the
shock tube experiments of Wright and.Humberstone. (36).discussed previously, the peak
pressure increased essentially linearly with temperature above the melting point of
the aluminum, suggesting that only the sensible heat of the molten aluminum was
available to produce high. pressure. steam. and that.neither latent heat nor the sensible heat of the frozen-aluminum contributed appreciably.
Buxton and Benedict (39) estimated explosion energies from the crushing of a
honeycomb block support under a tank of water into which from 1 to 27 kg of thermite
was poured. In most tests less than 0.5% of the thermal energy in the melt was converted to work. The highest value obtained was 1.34%. This compares with a
Hicks-Menzies efficiency of about 30%.
All of the laboratory tests tend to,give less of the maximum energy into work than
the two industrial accidents discussed above. This may be an effect of the scale of
the system. As mentioned previously, energy efficiencies are expected to increase
as the size of the system increases.
Lienhard (26) made an analysis of the potential work obtainable from the flashing of
a superheated liquid, which indicates that the available energy will be proportional
to the square of the superheat. This would be expected to be applicable if the
explosions are superheated liquid explosions and if there is negligible heat
transfer after vaporization is triggered.
-13-
Another approach to the energy release in a vapor explosion is to'relate it to the
'
explosion of a given amount of TNT. Lipsett.(3) made such a calculation for the
foundry accident and concluded-that the explosive evaporation of 1 lb of water was
equivalent to the explosion of 0.4 lb of TNT. Similar.calculations have been done
by others, giving slightly different numbers. depending-on the assumed expansion 'path
and the initial and final states of the water.
-14-,
EXPLOSION MECHANISMS
.
There are two fundamental problems in developing a satisfactory mechanistic.explanation of liquid-liquid explosions. The first is to'explain the extremely rapid
vaporization rates and the associated heat transfer required to supply the heat of
vaporization. The second is to explain the coherence of the explosive event.
There is no doubt that inertial confinement of a vaporizing liquid can lead to an
energetic shock wave provided that the vaporization proceeds rapidly enough. Two
general ways that rapid vaporization can occur can be identified:
1.
The heat transfer takes place by conventional processes (at relatively
slow rates) before the explosion takes place. Part or all of the explosive energy is stored in a metastable state as superheat in the liquid.
2.
The heat transfer takes place at very rapid rates simultaneously with
the explosive event. This requires extremely high heat transfer rates
which could involve either new ultra-high rate heat transfer processes
or the fragmentation and intermixing of the two fluids to give a very
high interfacial contact area.
It should be noted that these two paths are not mutually exclusive, but may occur in
sequence.
Coherence is a fundamental problem which becomes more important as the scale of the
For a large-scale energetic explosion to occur there must be
interaction increases.
a coordinated vaporization of a large mass of volatile liquid on a time-scale comIn this way the expansion energy from the
parable to the pressure-rise time.
vaporizing liquid serves to reinforce the blast wave rather than simply to disrupt
the liquid-liquid system. The need to explain coherence requires the introduction
of some form of coupling between the explosive shock wave and the vaporization process.
Numerous mechanisms have been proposed to explain liquid-liquid explosions, probably
indicating that various systems have different controlling features. The particular
experimental system used undoubtedly had a great influence on mechanisms suggested
It is quite possible that no single mechanism can be used
by various investigators.
In attempting to sort out all
to describe all possible liquid-liquid interactions.
to
of the different explanations it is helpful to keep in mind the basic goal:
understand large-scale explosive interactions capable of doing significant damage.
Background
To properly interpret the developments in the understanding of the mechanism of
vapor explosions, it is necessary to put them in a historical context. It is also
necessary to consider the different industries from which concepts evolved: since
the systems are quite different, and the controlling phenomena are not necessarily
the same.
Vapor explosions have long been a recognized problem in the metal processing
industry and the pulp and paper industry. Experimental studies of explosions were
Although attention was paid to the
initiated by these industries during the 1950's.
studies was to try to find pracof
these
focus
the
major
mechanism of explosions,
of
the effort at this time was
A
part
tical methods for preventing explosions.
-15-
simply to demonstrate that physical explosions could occur. Explanations of the
explosion phenomena were mainly mechanical. One of the earliest proposed mechanisms
was the encapsulation theory, which had its origin with Epsteins (57) suggestion
that explosions might be initiated by the. rupture of a frozen shell of the hot
substance surrounding the cold fluid.
The explosive rupture of the shell fragments
and intermixes the liquids, generating a much. larger explosion. Other fragmentation
mechanisms were also considered in.this period.. The problems of coherence and of
the scale of the event were not given much attention.
In 1962 a small experimental, nuclear reactor was destroyed when the main control rod
was rapidly removed during, shutdown maintenance (42). A nuclear power excursion
melted -the fuel element which penetrated the cladding and contacted the cooling
water, generating. a large vapor, explosion. As a result, the nuclear power industry
became very active in studies of liquid-liquid explosions.
The prime interest of
the nuclear industry was to understand. the conditions necessary for an explosion to
occur (to predict the likelihood of an occurrence) and to predict the intensity or
energetics of the explosion (to size the containment vessel).
This need to develop
a predictive capability and the availability of considerable government support led
to elaborate studies of mechanism by a considerable number of investigators. Much
of the current understanding of liquid-liquid explosions comes from the nuclear
Investigators extended. the fragmentation concepts coming from the metals
industry.
industry, adapted spontaneous nucleation concepts when they became available,
recognized that there may be.pronounced effects of system scale and 'were very
conscious of coherence and the need for coupling in the explosion mechanism.
In 1968, the U.S. Coast Guard unexpectedly produced a vapor explosion during a test
spill of liquefied natural gas (LNG) on water (42). Thus, a fourth industry became
involved with liquid-liquid explosions.
Subsequent studies of LNG-water interactions showed that a critical factor was the suppression of normal nucleate boiling
at the liquid-liquid interface and the buildup of a considerable superheat in the
cold liquid.
This was the source of the concept that spontaneous nucleation and the
superheat limit is a critical factor in vapor explosions.
It is now generally
recognized that LNG-water interactions are not highly energetic events, and this has
led to some doubt as to the applicability of simple. superheat limit concepts to
systems which do undergo' highly energetic explosive events.
To organize the discussion of explosion mechanisms, the material is grouped into
three general categories:
fragmentation, superheat, and detonation. These categories are not absolute. As will become apparent, some. explosion theories involve
elements from each of these categories.
Fragmentation
Much of the work on vapor explosions has sought explanations for the rapid increase
in heat transfer contact area. Mechanisms have been proposed for the fragmentation
of the two liquids and their interdispersion. This section of the report is concerned only with fragmentation mechanisms that do not involve the passage of the
explosive shock wave itself. The latter are considered in the detonation section.
Long (17) carried out experiments in which molten aluminum, magnesium or a sodiumpotassium chloride eutectic were poured into water in an iron container. He
observed that the explosions with aluminum could be prevented by greasing or
painting the bottom of the container. As a result he suggested that the critical
-16-
factor was the entrapment of a thin .layer.of,..water below the incoming metal.
Explosiv.e.vaporization of..the...entrapped-..water-.would. fragment the. metal and disperse
it in the water. .Hess and. Brond-yke.. (46). later:suggested that the critical factor
might be a .thermite-type chemical reaction between aluminum and an iron. oxide
coating on the container suface.. This was.. based-on their own experiments with
molten aluminum and. water in.which.they. found-that the. temperature at the container
bottom was nearly 400°F hotter than the initial aluminum temperature.
Based on Long.'s work, Epstein (57) suggested that -explosions might be initiated by
the rupture of a solidified shell of the met.al. containing encapsulated water.
Brauer (58) used. high-speed photographic techniques to study the behavior when aluHe observed the blowing up and both the
minum was poured into cold (4°C) water.
He also observed the violent ejection of
bursting and collapse of aluminum bubbles.
Experiments with Woods
spongy or mossy lead under similar experimental conditions.
metal poured into cold ethylene glycol-., so. that no boiling occurred, gave hollow
He proposed a liquid entrainment theory in which
spherical droplets of Woods metal.
liquid water gets inside a molten mass of metal and is rapidly vaporized, causing a
pressure pulse which fragments the metal drop.
Flory et al. (32) carried out similar pouring experiments using a variety of metals
They considered four possible mechanisms for fragmentaas well as polyethylene.
tion:
1) the violent boiling theory, 2) the shrinking metal shell theory, 3) the
Their experichemical reaction theory, and 4) Brauer's liquid entrainment theory.
ments produced clear evidence that liquid water could be. incorporated into a mass of.
They also found evidence that bursting of aluminum
a hotter, molten material.
They took high-speed motion pictures
bubbles can take place after solidification.
that showed concentric ripples due to Helmholtz instabilities with lead, tin, and
mercury.
Growth of these instabilities could provide a mechanism for enclosing
Fragmentation also depended on the physical properliquid water .in the metal drop.
Addiction of one .atomic percent bismuth to aluminum (which
ties of both liquids.
Increasing water
lowers surface tension 50%) gave much greater fragmentation.
viscosity 5-fold by addition of carboxyme.th.yi cellulose greatly reduced fragmenThey concluded that their work strongly supported the. liquid entrainment
tation.
theory and was incompatible with the other theories.- With respect to violent
boiling they noted..that films of lead and. tin .fragmentation showed no evidence of
With aluminumdrops they noted that the entire drop was
any boiling action.
surrounded by a film of water vapor during its. descent, and that "a front of .
nucleate (violent.). boiling passed over the surface of the solidified drop only after
a considerable time. on the bottom of the container."
Witte et al. (4) in their review noted that heat .fluxes in boiling processes rarely
exceed 3 x 107 W/m 2 (107 Btu/hr ft2 ). Using this upper limit for the heat flux, and
applying it to a documented explosion where 45.4 kg (100 lb) of molten steel at
1560°C (2840°F) was spilled into water to produce about 16 kg (35 lb) of steam in 3
milliseconds, they concluded that the steel would have to be fragmented into parIn support of this required
ticles with an average diameter of about 200. microns.
fragmentation level,, they cited an-experiment by Higgins (59) in which explosions of
In
molten metal were triggered by a blasting cap 13 cm below the water surface.
those cases where intensive explosions occurred, the recovered particles were less
When the particle size was greater than 1400
than 200 micrometers. in diameter.
micrometers, little explosive reaction was recorded.
-17-
They considered four theories to account for fragmentation of molten material:
1.
Entrapment of volatile liquid between the molten metal and the
surface of the container
2.
Violent boiling, at the transition from film boiling to nucleate
boiling
3.
Entrapment of. liquid within the molten metal globule and its
sudden vaporization
4.
Weber number effects in which inertial..forces acting on the molten
globule overcome the surface tension forces
No conclusions .on the applicability of these mechanisms were drawn.
any discarded.
Neither were
Stevens and.Witte (60). stud.ied.the action of al1-inch-diameter silver sphere when
passed through a Plexig.las tank of water at 5 ft/sec with an initial sphere temperature varying between 500 and 1000°F. They observed two mechanisms for the
destruction of a stable vapor shell. The first was a "transplosion" which was
described as follows: "With no visible indication of an impending change in the
mode of boiling, the thin vapor shell (estimated to be no greater than 0.010 inch
thick) surrounding the sphere suddenly appears to explode. The effect of this
miniature explosion is to cause the spherical shell to become unstable. The time
duration for a single transplosion was observed to be less than 0.25 ms.
Immediately following a transplosion, the vapor film goes into a condition of pulsation
boiling - (which) takes the form of a rhythmic expansion and contraction of the
vapor shell normal to the surface of the sphere.
"The second method of stable vapor film destruction can be described as a progressive vapor film instability. This instability is induced in an otherwise stable
vapor shell when small irregularities which have the appearance of hemispherical
bubbles on the liquid-vapor interface, begin to grow and collapse while traveling
irregular paths over the surface of the vapor shell. These small perturbations
trigger oscillations in the vapor shell, resulting .in the: pulsation boiling phenomena described previously. In comparison to the. ..
transplos.ion, the progressive vapor
film instability is a gradual phenomenon., although its duration time was consistently less than 100 ms."
In a subsequent study, Witte et al. (5) made a high-speed photographic study of the
quenching of mercury, lead,. zinc, bismuth, tin,. and.aluminum in both water and
liquid nitrogen. It was repeatedly observed that.bismuth dispersed immediately as
it penetrated the water, apparently from the Weber number effect. However, as the
dispersed drop penetrated farther into the water, the droplets fragmented massively.
"In some cases the metal entered the water as several droplets rather than a single
drop. In some of these cases, fragmentation of a single droplet was followed immediately by fragmentation of the other drops - there is little doubt that the first
event initiated the other."
An evaluation of proposed fragmentation theories was made in the light of their
experimental results. Three theories were examined:
-18-
1.
Violent boiling associated with the transition from film boiling to
nucleate boiling
2.
Entrainment of cold liquid within the hot liquid
3.
The Weber number effe.ct (inertial forces overcome surface tension
forces)
The violent boiling hypothesis was rejected because the time scale for vapor
collapse and reformation-was -about the same. as.. for- the -entire fragmentation process.
Therefore if transition from film boiling, has.any effect on fragmentation, it is
only in the initial collapse of the vapor .filmt. -Liq.uid.entrainment was rejected
In addition, the obserbecause there. was no photographic. evidence,for entrainment.
vation that fragmentation of one droplet "set-off" the fragmentation of other
droplets indicates that. the triggering.. is caused by forces external to the droplet.
The Weber number effect does cause dispersal of the.-hot liquid, but it is not the
major cause of fragmentation. This was apparent in photographic sequences showing
hot liquid dispersal on entering cold fluid, with. violent fragmentation subsequently
at greater depths.
Having rejected previously proposed fragmentation hypotheses, Witte et al. discussed
what might be involved. They noted two experimental facts which they felt must be
considered. in any explanation of fragmentation:
1.
Fragmentation is a response to an external stimulus.
2.
Fragmentation occurs only when the sample is still molten.
They also noted that fragmentation seemed to be associated with the initial collapse
of the vapor blanket separating the hot and cold fluids. They suggested three
possible ways that collapse of a vapor film could cause fragmentation:
1.
The impact pressure generated when the cold liquid decelerates
as it strikes the hot liquid (or vice versa)
2.
An abrupt.reduction in surface tension as the cold liquid
touches the hot liquid surface
3.
Flashing of a superheated liquid
They made an attempt to measure the surface area increase from fragmentation in
their experiments. The maximum increase found was about a factor of ten, although
factors of 2 to 4 were more. common. The..procedure used -to estimate surface areas
was conservative and tends to underestimate the area increase. Despite this the
area increases seem. quite low, especially in the light of Witte's earlier calculation (4) that average particle diameters of about 200 microns (a thousand-fold
expansion) were required.
Buchanan and Dullforce (61) proposed the following vapor bubble collapse mechanism
for a hot liquid entering a subcooled volatile liquid:
1.
the liquids come into intimate contact and a vapor bubble is formed.
-19-
2.
The bubble expands and then collapses as a result of condensation in
the subcooled liquid.
3.
The bubble.collapse is asymmetric and a high velocity jet of liquid
coolant is formed., directed toward the.hot liquid, penetrates it and
is disintegrated.
4.
Heat is transferred from the surrounding hot liquid to the disintegrating jet.
5.
This suddenly vaporizes the liquid, in the jet to create a high-pressure vapor bubble.
6.
Rapid expansion of this bubble disperses the surrounding fuel into
the coolant as a spray of small drops.
The process is then repeated.
Ochiai and Bankoff. (62) proposed, a "splash" theory for local propagation of vapor
explosions. When a random liquid-liquid contact. is made, explosive growth and
collapse of vapor bubbles occurs as soon as the surrounding pressure is relieved,
resulting in a high local. pressure. at the liquid-liquid contact area. This is
treated as an impact pressure -applied to the free surface, and potential flow theory
is then used to obtain the velocity distribution. It may be noted that the vapor
bubble collapse portion of this mechanism would seemingly require that the cold
liquid be at a.temperature lower than the saturation temperature.
Anderson and Bova (7) carried out a high-speed photographic study of the injection
of water in molten sodium chloride.
Subsurface movies showed that explosions
resulted from a two-step sequence:
an initial bulk-mixing phase in which the two
liquids intermix on a .large scale, and a second step, immediately following breakdown of the separating vapor layer, in which the two liquids fragment locally,
intermix and pressurize very rapidly. Fragmentation occurred in times as short as
77 microseconds, and particles with a mean.diameter of 40 im were produced. As a
result they discarded the frozen shell, cold liquid entrainment, transition boiling
and a Weber-type jet instability mechanisms for fragmentation.
They suggested that
dynamic forces produced by the release of some degree of superheat at the contact
surface were responsible for fragmentation.
In recent work, Froehlich.et a.l. (63) again studied entrapment. They poured molten
lead, copper, aluminum, iron, and. glass into a water-filled conical mold with an
average diameter of. 1" and a depth of. 2". Melt. temperatures varied from 30°C above
the melting point.to 1400°C. The water temperature.was held at 15°C. Experiments
were carried out with and without external triggering... Only lead and copper produced explosions without external triggering. When these did not cause an explosion, the melt filled out the cavity completely, which meant that the water escaped
completely before the freezing. of. the me.l.t.
Aluminum did not fill out the cavity
completely unless it was subjected.to vertical shaking. Neither the iron nor the
glass completely filled out the contour nor volume of the mold. In such cases, a
vertical channel. remained.. When explosions were triggered with the molten iron, a
high percentage of hollow spherical fragments was obtained. No explosions were
obtained. with the glass melt under the same experimental conditions. Only weak
-20-
explosions occurred when the aluminum melts were triggered.
this work to fragmentation mechanisms is not obvious.
The significance of
A final possibility that needs to be considered is thermal..fragmentation of solidified materials due to stresses set up.by rapid cooling and high temperature gradients. Nelson and Buxton (8) observed thermal.fragmentation of oxidic melts.
Colgate and. Si.gurgeirsson (6) called attention to-the sudden chilling, solidification, and finally cracking or spalling of the fractured surface of underwater lava.
They also note that such spalling..occurs when molten glass is plunged into cold
water. Cronenberg.et al. (64).proposed a thermal stress mechanism for uranium oxide
in sodium, and Epstein (65) also proposed a gas release. thermal fragmentation mechanism. Corradini (14) dismissed thermal-stress fragmentation because it involves a
solid material, and the fragmentation process is not..fast enough for explosive heat
transfer.
Fragmentation Summary
A distinction must be made between dispersion, which.results in a relatively mild
increase in surface area .(say less than an order .of.magnitude), and explosive
fragmentation, which occurs on a time scale less than one millisecond and produces
particles on the order of 100 microns in diameter...The latter is the critical
fragmentation step important to vapor explosions.
A number of processes have been found adequate for producing relatively vigorous
dispersion, but inadequate for explaining intense fragmentation. These include:
1.
Rupture of solidified shells of hot fluid from internal pressurization
2.
Entrainment of cold-liquid within a globule of hot liquid and
its rapid vaporization
3.
Violent boiling..associated with the transition from film boiling
to nucleate boiling
4.
Low velocity Weber-type instabilities associated with competing
inertial and surface tension forces
5.
Entrapment. of cold liquid between hot liquid and a container
surface
6.
Spalling of solidified material due to thermal stresses
7..
Oscillatory vapo.r bubble growth and collapse
The key fragmentation. event occurs very rapidly (in one case less than 77 microseconds).
Most available evidence suggests that fragmentation is initiated when a
vapor film separating the hot and cold. liquids collapses and the two liquids come
into direct contact. Three possible explanations. are the impact pressure developed
when the two liquids come together, release of superheat at the contact interface,
and hydrodynamic instabilities associated with an explosion shock wave.
-21-
Superheat
.
..
An alternative explanation of vapor explosions is the sudden release of superheat in
the cold liquid and resulting,.extremely rapid vaporization. Normally this
"mechanism" considers the suppression of nucleate boiling at a liquid-liquid.interface so that the cold liquid can be superheated significantly above its boiling
point. When the homogeneous nucleation temperature is reached at the interface the
If a large fraction of the:
liquid spontaneously vaporizes at an explosive.rate.
volatile cold liquid can be. brought to its spontaneous nucleation .temperature at the
In liquid-liquid systems,with
same time; an energetic explosion would be obtained.
very high-temperature gradients,. only a small fraction (a thin boundary layer) of
the cold liquid can be superheated.before the spontaneous nucleation temperature is
reached and vaporization begins. -Thus, the spontaneous nucleation process is
usually considered as an initiator leading to. mixing of the fluids and an escalating
interaction.
Initial work on .spontaneous nucleation came. from .the cryogenics field. -Enger and
Hartman (66) spilled mixtures of hydrocarbons on water and. controlled the mole fractions of various hydrocarbon. components in the mixture. and: the prespill temperature
Distinct temperature. thresholds between. explosive and nonexplosive
of both liquids.
events were found which could be rationalized. .They explained their experimental
results by proposing that .liquid-liquid contact .suppressed vigorous nucleate boiling
and extended the nucleate.boiling region in the boiling curve to superheat as high
as the homogeneous nucleation temperature, at which point the cold liquid would
Three conditions prevented explosions: hot liquid temperatures '
explosively boil.
less than the homogeneous nucleation temperature; ice layer formation on the water,
which provided nucleation- sites; and temperature differences high enough to sustain
film boiling.
Nakanishi and Reid (67) discussed numerous experiments involving cryogens.
They
proposed that when the cryogen first contacts the warmer liquid, film boiling
results, cooling (but not freezing). the lower liquid. At a critical temperature
difference, the cryogen suddenly wets the surface. No nucleation sites exist at the
interface and the cryogen superheats rapidly. Superheating continues until homogeneous nucleation occurs, followed by flashing and an "explosion."
Subsequently Porteous and. Reid (41) analyzed the: data on cryogenic explosions from
other experiments as well as. their own.. They showed that essentially all of the
explosions involving cryogens spilled on water occurred at values of Tw/TsL between
1.0 and 1.1, where Tw is the initial water temperature and TSL is the superheatlimit temperature of the cryogen...
The lower temperature limit (for the water) was
quite sensitive. The upper temperature limit was less definite and could be overcome by impacting. the cryogen and. water together.. Systems which were nonexplosive
in the pouring mode of contact..(such as methane on water) because Tw/TSL was too
high, could be exploded by. imp.acting the fluids- together. The lower temperature
limit is of course associated. with. superheating the. cryogen to the superheat-limit
temperature. The upper limit is associated with the stability of film boiling.
Thus, explosions of cryogens. and water appear to take place whenever film boiling
is suppressed,. giving.direct liquid-liquid contact and the cryogen can reach ;the
superheat-limit temperature.. This is strong evidence for the superheat mechanism in
cryogenic systems.
In a later review article Reid .(2) states, "It 'is now generally
agreed .that light hydrocarbon vapor explosions on water result from superheating a
thin layer of hydrocarbon on the water surface. Very rapid nucleation then leads to
further fragmentation and mixing that culminates in a large-scale vapor explosion."
-22-
Although the superheat mechanism-seems to apply very. wel.l to the data on cryogenic
systems, there -are difficulties in applying it to..molten metal-water, molten saltwater and nuclear systems.
In these cases, the hot. liquid temperature is usually
well above the spontaneous nucleation temperature of the-cold liquid and often above
its critical temperature.. The-cryogenics results would simply predict stable film
boiling and a nonexplosive system under these conditions.
Fauske (68) extended the superheat theory by pointing out that the interface temperature between the hot and cold. liquids was the significant temperature.
He
postulated that a..necessary condition for a vapor explosion is that the interface
temperature be. greater. than the. spontaneous nucleation temperature.
Fauske (34)
applied this concept. to the uranium dioxide-sodium-system (an important one in
safety analysis of the Liquid Metal Fast Breeder Reactor).
A significant feature of
the U02 -Na system is that the initial. interfacia.l contact temperature is considerably below the spontaneous nucleation temperature' for sodium.
Thus a vapor
explosion cannot occur immediately. on contact.
Fauske indicated that homogeneous
nucleation could occur in this system- if the liquid sodium is completely entrained
in a sufficiently large bulk .of U02 and. if the. minimum..temperature needed to sustain
film boiling is:.above the freezing-po.int. of U02 [it. is according to Henry's theory
He was thus able to rationalize Armstrong.'s. laboratory explosions with
(69)].
UO2 and Na.
He. also concluded that a. large-scale coherent vapor explosion between
U02 -Na is not possible in a reactor environment.
This distinction of-scale effects
was an important contribution.
Anderson and Armstrong (42) obtained explosions by injecting water into molten NaCl
and injecting molten sodium into molten U02 . They pointed out that the simple
homogeneous nucleation theory based on water-hydrocarbon systems would predict
nonexplosive film boiling with no surface contact for the NaCl-water system, since
the contact temperature would be well above the homogeneous nucleation temperature.
The contact temperature is below Henry's minimum film boiling temperature, so the
water-NaCl system would be predicted to be in the transition boiling region.
In
high-speed movies of NaC.l-water explosions it was evident that "every explosive case
was initiated by an external force which tended. to drive the two liquids across the
insulating vapor film into near or actual contact with each other."
The explosions obtained by Anderson and Armstrong with UO2-Na gave considerable
problems to the. spontaneous. nuc.leation theory,. and led to extensive analyses of
explosion mechanisms.
U02 -Na form an unusual- pair in that the high-conductivity
metal is the cold..liquid, while the hot liquid is a high-temperature ceramic.
The
contact temperature. between 400°C Na and 3000-°.C. U02 is only 1150°C, whereas the
homogeneous nucleation temperature. of sodium .is 205.0°C.
Thus, even if the sodium is
entrained in excess U02 so that it can reach the. spontaneous nucleation temperature,
the sodium is very likely to be in contact with.. frozen U02 before reaching 2050°C.
Fauske (34) proposed that freezing of the hot liquid does not necessarily provide
nucleation sites if the hot liquid is wetted. .by the cold liquid before freezing
occurs.
In a later paper Henry and Fauske (40) refined. their concepts.
They listed three
conditions as necessary but not sufficient for vapor explosions:
1) two liquids
must be present, 2) the liquids must come into intimate contact, and 3) the interface temperature must be .greater than the spontaneous nucleation temperature for the
system.in question.
They consider that spontaneous nucleation may be either heterogeneous (where random vapor embryos are generated at the liquid-liquid interface
because of reduced interfacial tension) or homogeneous (where embryos are generated
within the bulk of the cold liquid), depending on the transient wetting characteristic of the given system.
.
'
'They considered the conditions needed for random nuclei to grow in a nonisothermal
liquid and the nature of the.heat transfer process under explosive boiling conditions. This involved an analysis of conditions after direct contact on time
scales between 10- 9 and .10-5 sec. Under these conditions nucleation cannot proceed
until a sufficiently thick thermal boundary layer has been developed to support the
growth of the vapor embryos to. the-limit of their stability. From this analysis
they developed a criterion for the stability of a given size cold-liquid drop in
terms of wetting and capture by the hot liquid or sustained film boiling, as a function of interface temperature.. For a given interface temperature, droplets larger
than a critical size remain in film boiling despite initial direct liquid-liquid
contact, while droplets .smaller than the critical. size will wet and be captured by
the hot liquid.
For very high .interface temperatures, essentially all drops remain
in film boiling.
When a drop is captured by the surface, vapor bubbles, attempt to grow but cannot
approach coalescence prior to the onset of condensation.
In this case, the liquid
has established a very efficient means of transporting energy away from the interface, which can far exceed conduction. This process will tend to reduce the interface temperature.
However the interface temperature cannot decrease too much
without eliminating the responsible mechanism, spontaneous nucleation. Thus the
interface temperature would tend to remain somewhere above the spontaneous
nucleation temperature and control of the heat transfer rates will be in the hot
liquid.
This model was extended in a subsequent paper by Henry and Fauske (12) to' include
predictions that explosions are eliminated by an elevated system pressure or a
supercritical contact interface temperature. Data for Freon-12 on mineral oil are
presented which show very good agreement with the droplet capture model. Data are
also provided which show large-scale explosive interactions fit the contact interface temperature criterion. Henry and McUmber (38) give data showing a dramatic
change in system behavior as the interface temperature changes by a few degrees near
the spontaneous nucleation temperature.
These data were for Freon-22.
Segev (10) in his review summarizes the relation between the spontaneous nucleation
temperature and small-scale and large-scale explosions.
If the interface temperature is below the spontaneous nucleation temperature, there is no stable film
boiling, incoherent nucleation only, and small-scale explosions which are only
obtained when the cold liquid, is entrained in the hot liquid.
If the interface temperature is above the spontaneous nucleation temperature, there is stable film
boiling and.coherent nucleation, and large-scale explosions can be obtained for
either cold liquid into hot or vice versa. He also comments on the "capture theory"
of Henry and Fauske and its prediction that the explosions are not possible if the
interface temperature is greater than the critical temperature. Excellent agreement
with this upper temperature threshold exists with Freon-oil and most LNG experiments.
However, liquid metal-water systems have given explosions even when the contact temperature exceeds the thermodynamic critical temperature. One possible '
explanation is that strong variations, in physical properties occur close to the critical point.
-24-
Superheat Summary
There is impressive evidence that superheating. volatile liquids to the spontaneous
nucleation temperature is a major contributing factor to vapor explosions. This is
especially the case for cryogenic fluids on water, and Freon-water systems. -In
these cases the spontaneous nucleation temperature serves as an effective lower
bound for explosive interactions and the critical temperature as an upper bound.
Other systems are more complex.
Important contributions in applying spontaneous nucleation concepts to noncryogenic
systems have been mad.e by Eauske and.Henry. These include: the identification of the
interface (contact) temperature as the important temperature, a distinction between
small-scale explosions and. large.-sc.al.e explosions, a treatment of heat transfer in
the boundary layer and. resulting delayed nucleation, and an explanation for-direct
liquid-liquid contact: at interface temperatures. above.the spontaneous nucleation
temperature. They sugge.st.a mechanism for very. rapid.heat transfer during direct
liquid-liquid contact and. provide a criterion for when this will occur.
The analyses of Fauske and. He.nr.y eliminate .most objections to the spontaneous
Most of the. work on fragmentation showed that the
nucleation theory of explosions.
initiation of intense fragme.ntation involved. collapse. of a vapor film and direct
liquid-liquid contact. This would allows -the processes treated by Fauske and Henry
to occur.
Detonations
One aspect that the fragmentation and spontaneous nucleation approaches do not
completely address .is the coherence problem. How do all of the events happen in a
Some
coordinated manner so that the expansion energy is directed into a blast wave?
form'of coupling between the fragmentation and/or spontaneous nucleation processes
and the developing shock wave must be present.
Board and Ha.ll (70) first applied the concept of a classical detonation to the vapor
explosion problem. They modeled vapor explosions as a one-dimensional plane explosion front propagating through a coarsely mixed region of hot and cold fluid. They
were able to predict detonation velocities and pressures without a detailed
knowledge of the fragmentation and energy processes. They indicated that fine
fragmentation within a few centime.ters of the front sets a minimum scale for effiOne result of their analysis was
cient thermal explosions in unconstrained systems.
in a fast reactor whole core
occur
the prediction that efficient explosions .might.
prediction.
the
Henry-Fauske
.to
contradiction
accident. This was in direct
Board, Hall, and.Ha.ll. (71) presented an analysis of the detonation process. They
assumed that. close to the shock front the flow. velocities are such as to cause fine
fragmentation. and rapid. heat transfer.. Expansion of .this material as it equilibrates drives the front forward. Thus, the propagation of a strong shock through a
coarsely mixed region of hot and.,cold liquids.by itself is sufficient to lead to an
escalation of the- energy content of the shock and an explosion. For the tin-water
system they calculated the propagation velocity at 300 m/sec and estimated that
fragmentation by Taylor instabilities could occur in a region about 10 cm behind the
front.
-25-
In a later paper, Board and Hall (9.)refined their concepts and indicated there were
1) a quasi-stable
three stages to a large-scale, high-efficiency vapor explosion:
initial configuration, 2) triggering, and 3) propagation.
The first step requires that the hot and-cold fluids be intermixed on a scale small
with respect to the exploding region and close enough together to give coupling for
coherence. Board and Hall (9) state that stable film boiling may be important in
establishing the initial configuration and that spontaneous nucleation may serve as
a not wholly reliable criterion for stable film boiling.
The requirements for the triggering process depend on the stability of the initial
configuration. A sufficient disturbance might. arise spontaneously by several
possible mechanisms or an external trigger may be required.
Board and
The propagation step requires- a coupling mechanism to. ensure coherence.
the
explosion
front comHall (9) state that energy transfer. and expansion close to
pared to overall, dimensions is.sufficient to sustain the motion of the front. The
mechanisms suggested .for giv.ing. rapid .energy transfer -are vapor collapse and its
associated fragmentation. and: mixing.,..or high velocity differentials due to differing
accelerations of hot and cold liquid and vapor. as the.. shock passes.
They make a
distinction between fully developed. detonating. explosions (in which displacement
velocities are sufficient -to -ensure fine. fragment-at'ion.)-.and slower propagation or
the escalation from a low-energy: trigger to the detonation state (where other processes may be:.needed for fragmentation).
Board and Hall (9) argue that the fact that high-yield.large-scale explosions occur
implies either that the interaction is triggered everywhere simultaneously or that
it spreads very rapidly from a local trigger.
The latter is most likely.
Two
pressure waves producing vapor blanket
mechanisms are suggested for propagation:
collapse, and explosive displacements producing rapid mixing. They discuss experimental evidence for propagation, including films by Dul.lforce which show that propagation between tin drops in a water-filled tube was clearly associated with vapor
blanket collapse.
They also mention films of Laber and Lemmon and of Briggs taken
of large-scale aluminum-water explosions which clearly show a rapidly moving explosion front.
Board and Hall took movies with simultaneous pressure measurement of
Freon 22-water which. show the explosion spreads by means of a shock front.
Nelson and Buxton (8) carried out. experiments in which 10-35 grams of arc-melted
material were flooded with 1.5 liters of water. Materials used were Type 304
stainless steel, "metallic" Corium-E., "oxidic" Corium-E (Corium-E is a simulant of
what would be present during a reactor core meltdown consisting of mixtures of iron,
uranium, and zirconium.at various levels of oxidation) and, in a few experiments,
aluminum and silver. Most of the tests were with. stainless steel and oxidic
Corium-E.
Very few. experiments were made with metallic Corium-E because of phase
separation in the melt. Only one flooding, experiment with metallic Corium-E and an
applied pressure pulse was performed.. No explosions, were obtained with metallic
materials under any conditions,. although. it.should. be noted that these were melted
in a water-cooled copper cavity, and thus would-.cool very rapidly during the tests.
With the oxidic Corium-E, explosions could be triggered by an exploding bridge wire.
No explosions occurred when .triggering pulses were not used.
The interactions were
most vigorous when the pressure pulse. was supplied soonest after flooding.
The
-26-
maximum pressurization also increased as the delay time .between the application of
the trigger and..the. measurement of. the. interaction increased. High-speed. photographs. showed that. these interactions.-occurred in two stages. The exploding
bridge wire produced a shock wave which in turn produced..coarse fragmentation of the
"Each droplet then seems.-to. drift through the water encased in a
molten material.
bag of .wate.r...vapor. as film b.oiling..ar.ound the. droplet .continues. Five to 20 frames
later (1.5 to 6 ms) a violent .second interaction occurs which blanks out viewing
for the remainder of the .film."
Buxton,.. Nelson, and. Bened.ick.(30.) did...exper.iment.s.. with simulated fuel melts flooded
with wat.e.r.. Natural triggering of, .vapor..exp-osions. was....not observed. Short duration pressure pulse.s...could...trigge.r..vigorous..muLt.i-.s.tage steam explosions. They
values for the...oxygen: content of-the. melt (below which explodetermined. .threshold..
. Thresholds..for..water..subcooling and chamber pressure
triggered).
sions could. not be
a.narrow. time. window:.during which a triggering
was.
..
There.
were also observed..
.an
explosion.. Large.r..applied pulses usually did not
ini.tiate
could
pulse
pressure
nonexplosive conditions.
otherwise
trigger an explos.ion..for
Buxton. and...Be.nedick .(39) carried out a series. of intermediate-scale experiments by
pouring 1- to 27-kg..quantitie.s .of an.iron/alumina. mixtur.e..into an open tank of water.
Spontaneous explosions occurred...in. most experiments.,- with several exhibiting
multiple explosions.. When. detonators were used;,. they. triggered explosions under
almost every condition (the one exception.was when. the:.detonator was fired quite
The detonators did not seem to alter. the efficiency of the explosions.
late).
Cronenberg (13) reviewed.proposed explosion mechanisms and concluded, "All model
concepts are consistent in that.an initial .per.iod.of stable film boiling, separating
molten fuel from coolant, is considered necessary (at least for large-scale interactions and efficient intermixing)., with subsequent breakdown of film boiling due to
pressure and/or thermal effects., followed by intimate fuel-coolant contact and a
rapid vaporization .process which is sufficient-to cause shock pressurizationn"
Although differences exist regarding the:conditions. and.energetics of film boiling
destabilization and fragmentation and intermixing, the principal area of difference
seems to be what constitutes the requisite conditions for.rapid vapor production and
shock pressurization. . Cronenberg. (13) concluded that there was.strong experimental
evidence for a vapor film collapse/fragmentat:ion mechanism for explosive vaporization, and that therefore. fine-scale fragmentation and..int.ermixing are necessary conditions for large-scal.e vap.or..explosions.,. while -attainment. of spontaneous nucleation
is not necessary.
Bankof.f..et al. (72) in a 1977 report on mechanisms.. stated that two principal concepts dominate the .fuel-coolant .int.eract.ion.fiel.d:: .. Fauske's spontaneous nucleation
theory .and .the Board-Hal.l.. theory. of thermal..detonation waves. They summarized the
Fauske hypothesis .as...follows; .for a large-scale .vapor explosion to occur the
following are necessary:
1.
The two liquids initially must be in.-a film-boiling mode to allow
coarse premixing on a global scale.
2.
Liquid-liquid contact.must be. established (triggering event).
-27-
3.
Spontaneous nucleation must occur upon contact.
4.
Inertial and/or structural constraints must be sufficient to allow
the development of high pressures
.
The third requirement-is the most controversial.
Bankoff et al. state that the
third requirement-can be replaced by.the necessity for rapid local pressurization
(10- 3 to 10-4 sec time constant) in order to sustain a pressure shock; Spontaneous
nucleation is then a sufficient but not necessary condition for rapid pressurization. They also review the Board-Hal.l approach which hypothesizes a fully developed
steady-state detonation wave supported by fragmentation of fuel droplets fragmented
by Taylor instabilities.due to the large relative velocities behind the shock. They
state that although Board-Hall predictions have been seriously questioned, the
viewpoint of a propagating detonation wave has been useful.
Williams (35) adopted a '.'vapor film model" in. which film boiling initially' inhibits
liquid-liquid contact. He assumes the system is coarsely premixed and that vapor
collapse and resulting liquid-liquid contact yields at.least a moderately vigorous
local interaction. His model predicts that pressures.rise quickly enough to
collapse most vapor films so that the interaction propagates at the shock wave propagation speed and the explosion intensity increases with system size. For many
reasonable values of the parameters, the model.predicts only modest interactions for
small-scale experiments, yet efficient explosions can occur in large-scale accidents.
The vapor film model also suggests..that explosions may have to be triggered.
Rabie and Fowles (73) consider a thermodynamic system having a phase transformation
with change in volume as an elementary explosive. They note the possible differences compared to the detonation waves predicted for conventional explosives but
confirm that one could expect detonation, if the material is in a well-developed
metastable state and has a sufficient mass so that the detonation will propagate.
In another paper Fowles (74) proposed that a superheated liquid can act as an elementary explosive and sustain a detonation analogous to those that occur in chemical
explosives. However, his approach fails to address the question of how a large
quantity of liquid can acquire significant superheat in the first place.
Corradini (14) proposed. that .vapor.explosions can be thought of as consisting of 4
phases:
1.
Coarse Intermixture:
Hot and cold liquids become intermixed and
the mode of heat transfer is quiescent (e.g., film boiling).
Present knowledge is sparse on the extent of mixing necessary
(characteristic dimensions and mass ratios).
2.
Trigger: The hot and cold liquids are brought into direct
liquid-liquid contact and rapid heat transfer begins. This
could be initiated by an external pressure pulse.
3.
Propagation:
The rapid heat-transfer process escalates as the
whole system becomes involved by.fragmentation and generates
high-pressure vapor.
-28-
4.
The high pressure vapor expands against inertial
Expansion:
constraint and may cause destructive work.
Corradini analyzed the experimental data of Nelson and coworkers at Sandia and
concluded (with the aid of a mathematical model) that noncondensible gases generated.
at a melt-coolant interface would retard. film.collapse and tend to suppress vapor
explosions. He also concluded that the dominant fragmentation mechanism for-explosion propagation with light-water-reactor materials is Taylor instabilities due to
local pressurization.
Detonation Summary
Certain aspects of.the..detonation theory are very appealing. It provides an
excellent conceptual framework. for explaining coherence and also leads logically to
the effects o.f system size (explosions become more -efficient as the size of the
Another benefit is.that it permits viewing the
interacting system increases-).
explosion sequence in terms..of individual steps. -These include the quasi-stable.
initial configuration, the trigger concept and a-rapid..vaporization process associated with the passage of a shock wave (this defines time and distance scales for
the interaction).
The original detonation theory as developed by Board and Hall considered only the
conditions needed to sustain a shock wave propagating through a mixture of two
liquids. It did not consider at all.how.the. detonation wave was established in the
first place. This is now considered to be a key part of the explosion phenomenon how the disturbance escalates from a small.triggering pulse to a full-fledged
detonation wave.
Another problem with the original detonation theory was that it did not consider the
details of the rapid heat transfer and vaporization process behind the shock front.
It merely assumed that 'they occurred and then defined the time and distance scales
in which the interaction had to occur. Subsequent-work has refined the analyses of
the interaction processes considerably, but. there is still an element of "detonations occur because they are assumed to occur" in the theory.
Initially, the detonation theory, and the spontaneous nucleation theory were considered as. distinctly different (and-.in some. cases,. contradictory) explanations of
vapor explosions.. This controversy focused mainly on the U0 2-Na system and
questions of.. safety. with liquid-metal...fast .breeder reactors. Recently it is
becoming.more apparent..that. thes.e. theo.ries are. not mutually exclusive and may
describe different:.aspects of the.whole. explosion phenomenon. The latest approaches
are achieving. a.synthesis of these viewpoints.
Present Status
It appears. .that a.consensus. is developing that. large-scale, high-efficiency vapor
explosions involve 4 stages:
1.
Establishment of a quasi-stable-initial configuration in which
the hot and.co.ld. fluids are coarsely premixed
2.
A. triggering. event
.region.and.permits
a pressure impulse
superheat in those
that collapses the vapor film in a localized
direct liquid-liquid contact (alternatively
may trigger explosive release of stored
systems without a vapor film)
1
-29-
3.
Escalation of the interaction by rapid heat transfer, vaporization
and pressurization
4.
Propagation of a fully developed detonation wave through the
coarsely premixed system
Initial Configuration
A necessary condition.for. a large-scale explosion is to have the individual hot and
cold regions intermixed on a scale small compared.with the size of the total
exploding region and close, enough to give coupling for coherence. The configuration
before the explosion is initiated.must be quasi-stable, in that the temperature difference between.hot and. cold. liquids must be retained for as long as is necessary
for the coarse intermixing to develop.. Film.boiling. is normally considered to be
the.means by which the system is stabilized as a coarse intermixture. Film boiling
prevents direct liquid-liquid contact because of the intervening vapor film.
Moderate fragmentation due to forces developed from the film boiling process itself
or to hydrodynamic forces, surface tension forces or other factors can be operating
during this period to achieve a finer degree of mixing. This fragmentation and
dispersion would be. quite slow compared with -that occurring during escalation and
propagation.
The superheat limit or spontaneous nucleation temperature can serve as a criterion
for establishing -film boiling. According to the Henry-Fauske (40) theory, stable
film boiling will occur for larger droplets at temperatures very close to the spontaneous nucleation temperature,.while sufficiently small droplets can be captured
and retained at the hot liquid surface at temperatures gre.ater than the spontaneous
nucleation temperature. Other workers also support the spontaneous nucleation temperature as a criterion for stable film boiling.- If this is correct, then one
necessary criterion for a large-scale, efficient vapor explosion is that the contact
interface temperature for the ho.t and cold liquids exceed the spontaneous nucleation
temperature, since this is needed..to establish stable film boiling. This might be
tested by impacting a..large..quant.i.ty of molten sodium into molten uranium dioxide in
a large shock-tube experiment.
The preexplosive configuration may. be-the most important factor governing explosive
yield (9). The greater the degree of subdivision before initiation of the interaction and the greater the amount. of the two liquids involved in the intermixed
region, the higher the yield.. -The. smaller. the individual regions of hot and cold
fluid, the easier the achievement of coupling .and coherence. The motions required
for rapid fine scale mixing..and heat transfer. are, smaller, and a greater fraction of
the vaporization can take place.on a time scale comparable to the propagating
disturbance wave.
Many mechanisms exist for achieving the initial dispersion: hydrodynamic instabilities, Weber instabilities, growth and collapse of vapor bubbles, thermal fragmentation, shock waves from external events, encapsulation and even the film boiling
process itself. Although these may be important in any given experiment, particularly small-scale experiments, the mechanism of dispersal is generally of
lesser importance than the stability.
A noncondensible gas being emitted by either of the two liquids, or formed by chemical reaction on contact, would tend to: collect in the interfacial region and act as
-30-
a separating. gas barrier. It would .tend to stabilize the dispersion in the same way
as film.boiling, thus .tending to make the dispersion more stable. If fragmenting
mechanisms...are present, this increased stability could lead to a more finely
dispersed system before an interaction is initiated..
Triggering
Most explosions appear to be initiated by the collapse of the separating vapor film
in a localized region, permi.t.ting-direct liquid-liquid contact. This may arise
spontaneously, or it could.be. triggered by a. local pressurization arising. from an
external source. A number of. potential triggering .mechanisms have been identified,
including:
1.
Spontaneous vapor collapse as film boiling becomes unstable
2.
Entrapment of cold fluid by hot fluid at a solid boundary
3.
Thermal stress fragmentation
4.
Superheating to the spontaneous nucleation temperature and
sudden onset of boiling
5.
An external mechanical disturbance (e.g.,
a detonator)
Trigger requirements are directly linked to the stability of the initial configuration. If the coarsely intermixed system is quite stable, a strong triggering
pulse is needed to produce liquid-liquid contact.. If the initial configuration is
only marginally stable, an interaction can be triggered. quite easily. To the extent
that the stability of the coarsely intermixed stage.-depends on film boiling, the
conditions which govern.film boiling. stability determine triggering requirements.
The importance of the triggering stage. in any particular system depends on whether
or not the system.is easily triggered. If very.energe.tic triggers are needed, this
could reduce the explosion hazard considerably, and knowledge of the triggering process could be very important in devising preventive measures. If low-energy
triggers are effective,.or if the system self-triggers, then it is not important to
understand the details of the triggeringprocess.
The triggering.. step may have an undue. influence on the interpretation. of small-scale
laboratory experiments. Board. and.Hall (9.)quote Colgate as follows: "Experiments
which show an extreme.sensitivity of.explosion occurrence to small details of
experimental test set up do so because. triggering,.is the sole phenomenon being
investigated.". On the other.hand,..there are also, systems that have .only exploded
when a de.tonator.was used (8,39.,6.3.),, suggesting that. laboratory-scale tests to
determine the "explosivity" of' various systems are bes.t. carried 'out using an external detonator.
Escalation
The localized disturbance.which brings the two liquids into direct contact must
escalate to a full-fledged detonation wave for a large-scale explosion to occur.
Localized, direct contact of the two liquids creates an interaction which generates
fine fragmentation and rapid heat transfer which then feeds mechanical expansion
-31-
energy into the disturbance. The processes .governing-heat transfer and fragmentation during escalation. may be. different from those occurring..during a full-fledged
detonation. This fact may help.reconcile the differences between the. Fauske spontaneous nucleation approach and the Board-Hall detonation.'concept.
There is very good experimental evidence. that the' key step in initiating an interaction is bringing the hot- and. cold liquids into direct .contact by collapsing the
separating vapor .film.- There is also good evidence that such direct contact results
in a very rap.id. (on the. order of O.L millisecondn) fine fragmentation of the two
liquids and generation of .a.pressure pulse. There. is considerable disagreement on
the mechanism.of the fragmentation process at-this stage. Candidate mechanisms
include. flow. in. response to the impact pressure generated on vapor film collapse,
spontaneous nucleation and explosive boiling., and. cyclic bubble growth and collapse.
It is quite possible that the Henry-Fauske analysis that predicts conditions under
which direct liquid-liquid contact can be maintained, at contact temperatures greater
than the. spontaneous nucleation temperature. (droplet capture) and which also
suggests a mechanism. for .greatly enhanced. heat transfer is.applicable to the escalation step.
If..this view is correct,. it is the release of superheat due to spontaneous nucleation that permits. disturbances in liquid-liquid systems to escalate to
the point where a fully developed detonation wave exists.
Propagation
The final stage in a large-scale vapor explosion is the propagation of a detonation
wave through the quasi-s.table, coarsely intermixed region..
The passage of .the shock
wave causes fine fragmentation and.rapid heat transfer and vaporization, and .the
resultant expansion energy..accumulates in the. shock wave :and intensifies it. When
an interaction reaches. this.. stage the expl.osion..pr.oce.s.s-is self-sustained, and the
final blast energy is.determined- by the. extent -of. .the coarsely mixed region and the
proportions and degree of subdivision of the two liquids.
Board, Hall, and.Hall. (.71) were the first to apply detonation theory to liquidliquid explosions, treating the. shock as a.plane, steady, one-dimensional detonation
wave passing through a coarsely mixed hot liquid, cold liquid, vapor system. The
shock collapses the vapor blankets and induces a large relative velocity between the
hot and cold liquids which can.cause fragmentation.
Boiling mechanisms can also be
involved in the fragmentation process behind the shock front;
Jump mass, momentum, and. energy balances can be made across the shock front which,
together with assumptions about the. nature of multiphase flow and a parametric
treatment of the fragmentation, provide a means for quantitative analysis of the
detonation process. Some analyses., with simplifying assumptions, have been performed which suggest that purely hydrodynamic mechanisms (such as Taylor instabilities or boundary layer stripping), are. .adequate for achieving the fragmentation
needed to sustain the detonation.. Spontaneous nucleation is then not a critical
factor at this stage of the explosion process.
Much of the current theoretical work on vapor explosions is concerned with an analysis of conditions behind the shock .front.. One important question is whether or not
a detonation wave will propagate if the. pressure behind. the front is supercritical.
Another is the definition of conditions during,.the-escalation of the pressure shock
wave from a small to a large value.
-32-
SMELT-WATER EXPLOSIONS
Smelt-water explosions. have.been associated with kraft. chemical recovery operations
from the inception of the kraft process. Roberts.(75). commented on the early kraft
rotary furnace -with.a refractory smelter...as follows.,' "Air pipes were. placed into
the smelter for the. admission of air.from'.-a positive blower. Later these air
nozzles were water-cooled and because...this was before the days of water treatment,
the nozzles oft.en..plugged. due to scale.formation with resultant explosions. The
small. smelt ..
streams..due to. the low capacity of the .unit.s developed freeze-ups and
explosions."
Roberts further quotes from-.:.a.1925 addres.s by H. K. Moore: "As the
black liquor proceeds. down..the incinerator, it is liab-le to form a ring around the
inside damming...up..:the. black liquor on one side of the ring or dam, whereas we may
have.molten alkali on the other side of.the ring. Then in the course of the operation, the ring..br.e.aks. and. the molten alka.l.i is submerged:in a flood of black liquor.
Steam is generated..suddenly. A violent.explosion takes place, and men are carried
to the hospital to have their burns treated. Where an industry is built up gradually with a large .excess of stolid men to.draw upon,..we find that in the end a
crew can be obtained .who: have become.so.experienced that they take things as a
matter of course,.-and thus become so proficient that accidents of this sort are
reduced to a minimum."
Dissolving Tank Studies
The first published study of the.smelt-water system was an investigation of
dissolving. tank; explosions by.Sallack (76).. He. carried.out laboratory experiments
by pour.ing.smel.t from a.crucib-le into a pan...of..water .or green liquor. He was. unable
to obtain explosions.with pure -Na2 CO3 sme.l.ts, but found.. that the addition of 5% NaCl
or 10% NaOH.made. smelt..explosive.. A.correlation.was. found between dissolving tank
violence and the NaCl content of the smelt.. .Green liquor reacted more violently
than water, and increased green liquor temperature reduced explosion violence. Smelt
temperature did not seem to have any effect. He. concluded. that combustible gases
were not involved in the explosions.
Nelson and Kennedy (7.7) reported on similar experiments.with over 50 synthetic
smelts -as well as 5 commercial. soda and 38 kraft smelt samples. They concluded that
the action was primarily a physical phenomenon-and-.also observed the following:
As long. as the smelt was fluid enough to pour, temperature in
the range from 1500 to 19.00°F had little effect on quenching
violence.
Cold water produced the greatest violence..
With a nonviolent mixture, quenching produced only a wisp of
steam. The chilled mix crystallized into leaflets just below
the surface with no audible noise.
Still other compositions of smelt formed small glowing spheres
beneath the surface and exploded only after a delay of four
to seven seconds.
-33-
Molten sodium carbonate did not produce an explosion under any
circumstances.
Pure sodium chloride and pure sodium hydroxide both quenched
violently. These were also sensitizers as constituents in kraft
smelts, i.e., they' shortened the delay time and increased
the violence of the smelt-water explosions.
Kraft smelts were about 10.0 times more sensitive to sodium
hydroxide and about. 10. times more sensitive to sodium'chloride
Also, high reduction favored
additions than soda smelt.
explosions, while sodium sulfate seemed to inhibit them.
A reduction of the smelt melting point and viscosity by the
presence of sodium sulfide and-sodium hydroxide was noted.
Small additions of caustic to the quench solution had no effect.
In more than 500 quench experiments, including some carried out
at night, no flames or gaseous combustion were observed.
Smelts, to which oxidizing.compounds had been added
any elemental sodium present, were still explosive.
to destroy
Addition of either finely ground graphite or ferrous sulfide
These also reduced
to smelt inhibited explosive quenching.
the tensile strength of the cooled smelt.
Quenching violence increased with
increasing particle size.
They concluded that dissolving. tank explosions were caused by unbroken streams of
smelt during, rushes, which..coul.d..be prevented .if.the streams were adequately
They recommended use of a
dispersed by either.green liquor or steam shatter. jets.
vertical steam shatter.jet, with a fishtail.orifice about 1/32-inch wide, using 100
psig steam.
Nelson and Kennedy mentioned. work on an automatic device to regulate the pressure of
A patent issued
the shatter steam by the rate of flow through the smelt spout.
later to Gettle (78) and. assigned to Combustion Engineering, Inc., described a
method of control based. on sensing the.temperature rise of the smelt spout cooling
water.
Of related interest, Honda, Tamou, and Morimoto (79) describe a method of
preventing.explosions in the dispersion of. molten metals and slag in water, in which
the rate of addition of the molten material and the water was controlled by the
pressure pulse or noise level in the receiving vessel.
The widespread adoption of smelt shatter jets to break up the smelt stream and the
recognition that undissolved smelt should not be allowed to accumulate in the bottom
of the dissolving tank resulted in a significant decline in the frequency and
The more. serious problem has been exploseverity of dissolving tank explosions.
sions within the recovery boiler itself. After. the work.of Nelson and Kennedy (77)
on dissolving tank explosions., attention focused on furnace explosions.
Furnace Explosion Studies
Rogers., Markant, and Dluehosh (80) reported the..results. of a series, of experiments
in..which- one gallon of..water or green liquor. was....injected under the- surface of 20
pounds of synthetic smelts. prepared. from..sod.iumn.-carbonate, sodium s.ul.fate, sodium
They useda..manomet.er-.as.-an approximate measure of.
sulfide, and. mixtures..of. these..
the pressure pulse produced-by the. steam and...gas.es evolved in the 9-inch I.D. furThey ran x-ray analyses on. samples. of the. sme.lt.. before and after the explonace.
They also. sampled...and...analyzed evolved-gas.e.s.for carbon dioxide, oxygen,
sion.
carbon monoxide,. ill.uminants, .hydrogen, hydrocarb.ons--and nitrogen (by difference).
They found that the.pressure- puls.e obtained with sodium.carbonate and water was only
The pressure pulse
about one-third that obtained with sodium sulfide and :water.
°
increased significantly. wi.th-increase.-in smelt temperature in the range from 1600
Injection of. green liquor into a-20% sulfid.ity smelt gave about twice
to 200 0°F. ..
The
the .pressure pulse obtained when water was injected into. pure sodium sulfide.
They found
presence of.20 pounds of coke in the smelt increased the-pressure pulse.
hydrogen was released and sodium .sulfate was, formed during these violent interacThe quantities of .H2 and.Na 2 SO 4 increased with increasing smelt sulfidity
tions.
They explained-their results in terms of the production of hydroand temperature.
gen by the reaction
Na
2
S + 4H20 + 4H 2 + Na
2 SO 4
Thus they considered smelt-water explosions as being.
and its subsequent combustion.
combustible gas explosions, with a smelt-water reaction. responsible for generating
the combustible gas.
In retrospect, it is doubtful that this. endcothermic, hydrogen-forming reaction.
contributed significantly to the violence .of the. explosive interaction of smelt and
water; however., it is interesting...to. note that. they did find a good correlation
between the size of.the pressure pulse and. thepercentage of hydrogen in their gas
samples, -except for the experiment with green liquor.
Smelt-water Research Group
In 1963, a group of 58 pulp.manufacturers joined- with the Fourdrinier Kraft Board
Institute,.Inc. to form the Smelt-water.. Research Group to sponsor research on
smelt-water reactions. . The experimental. work-was carried out by the Babcock and
Wilcox Company and Combustion Engineering, Inc. under contract, while The Institute
of Paper Chemistry was contracted to provide project management, coordination, and
Howard S. Gardner was named as coordinator.
guidance, and to make progress reports.
Malcolm L.. Taylor served as chairman of the FKI Smelt-water Technical Committee.
The references to
Seven reports (81)., including the summary report, were issued.
these reports in this review are concerned only with the mechanism of the smeltAny emphasis is that of these reviewers.
water interaction.
The first progress report outlined the scope of the program, which included consideration of explosions of molten metals as,.well as other molten salts with water.
It was noted that explosions do not always occur when water and smelt come together
An
and also that. there may be. significant delay between contact and an explosion.
encapsulation mechanism,. in which water. trapped .ins.ide a solidified shell of the
molten. material. would. burst .out violently, .as .propo.sed..for molten metal-water reactions by Epstein- (57),. was. .cited as one possible. trigger, for a physical explosion.
-35-
Thermodynamic data were collected for.consideration of. possible,.chemical reactions
of smelt with water. In laboratory experiments in which a small quantity of water
was injected.under the surface of molten smelt, the violence of the explosions
seemed to be.proportional to the quantity of.water added. Water added to the surSolutions containing.
face of the smelt.did. not explode, but evaporated rapidly.
or prevented exploretarded
above
5%
bicarbonate
or
ammonium
agents
surface-active
sions. According.. to. a.recent. communication from -one of the experimenters (82),
however, violent explosions were later obtained when 10% solutions of either sodium
or potassium bicarbonate were injected into active kraft smelts at about 1520°F.
In the second progress report, the encapsulation theory was further developed with
the additional consideration that a solid shell of smelt might also form around
molten smelt immersed in a quench liquid. It was observed experimentally that
solidified samples of smelt containing mixtures of either sodium sulfide or sodium
chloride with sodium carbonate were much stronger than were test specimens of pure
In further tests in which small
sodium carbonate, which did not give explosions.
quantities of water or solution were injected into a crucible of molten smelt, the
following observations were made:
Explosions rarely occurred on the first injection.
on the fifth to eighth injection.
Most occurred
The smelt was cooled by the injected water.
Explosions with 24-27% sulfidity smelt and water took place only,
when the smelt temperature was between 1498 and 1670°F.
The results with a sodium. carbonate quench solution were similar
to those with water, but boosted the upper temperature limit.
Both the..violence of the explosions, and the upper temperature
limit were increased when sodium chloride solutions were substituted for water.
Surface-active agents tended to prevent, delay, or reduce the
violence of explosions and also reduce the temperature at which
they could occur.
Five and.20% solutions of ammonium bicarbonate foamed without
violence, whereas a more stable 10% urea solution, which decomposes above 270°F, exploded more violently than water.
Explosions were obtained when.10, 25, and 30% black liquor were
injected, but not when 50% liquor was used; however, it was not
possible to inject the 50% liquor beneath the smelt surface with
the apparatus employed.
When the smelt consisted of 15 or 20% sodium chloride in sodium
carbonate, water injections gave violent explosions with smelt
at smelt temperatures of 1442 and 1338°F.
,
-36-'
In the. third progress report, results of both experimental and theoretical studies
of possible chemical reactions of..smelt' with water..were presented, as well as a
further development of the phys.ical. explanation. of explosions, with emphasis on
encapsulation.
Twenty-one possible reactions of water with smelt constituents were examined,
looking. in.particula.r for the possibilities of producingg: sodium.vapor, hydrogen,
carbon monoxide or other combustible gases.
Thermodynamic consideration did show
the possibility of producing significant. concentrations of hydrogen, carbon monoxide
and hydrogen sulfide.
However, in..the. presence. of. water, sodium hydroxide was.produced instead of sodium.
Hydrogen., carbon monoxide and carbon dioxide were produced
'
experimentally by exposure of molten smelt to water-vapor .
In these experiments it
was also noted that. the first water vapor was absorbed by the molten smelt, and that
the production of the gaseous products. took place after an induction period. 'About
one-half gram of water was reversibly absorbed by-100 grams of smelt at 1600°F with
only 30 mm mercury water vapor pressure.
Although the 'reversible hydrolysis reaction of-sodium sulfide with water to produce sodium hydrosulfide and sodium
hydroxide was mentioned, it was not considered. further, possibly because it would
not produce any gaseous products.
Today, the ability of kraft smelt to absorb
water, whether by hydration or chemical reactions, must be considered as an important observation.
In the further study of the explosion'mechanism, smelt-water explosions were created
by forceful injections of water or solutions into a.crucible of molten smelt under
an inert atmosphere in a closed vessel.
Explosions were also obtained by flowing
water over the surface of molten smelt.
One violent ..
explosion was obtained when an
otherwise nonexplosive 1.0% solution of ammonium bicarbonate was allowed to run on
the solidifed top surface of a crucible .of partially cooled smelt.
This explosion
dislodged one-half-inch-thick chunks of. solidified smelt, revealing a core of molten
smelt surrounded on all sides in the crucible by solidified smelt.
At the time,
this explosion was. attributed -to the possible. presence of fissures in the solidified
smelt.
It is nevertheless important as. perhaps. the first reported explosion of a
solution in contact with solidified smelt.
In a further. search for materials which might be used to cool a smelt bed, solutions
of methyl and isopropyl alcohol., ammonium-hydroxide and, ammonium-sulfate were found
nonexplosive under the test. conditions.
The performance of' ammonium bicarbonate was
improved .by the addition of a small quantity of a. surfactant.
The effectiveness of
concentrated. solutions of ammonium sulfate, which did not give explosions under any
of the test conditions, was attributed to its reaction with alkali in the smelt to
form sodium sulfate and to liberate gaseous ammonia'.
A mathematical model was developed to simulate the heat transfer taking place
between encapsulated water and molten smelt.
The model predicted that the contents
of the capsule at the time of rupture would be hot water rather than steam, and that
pressure increase before rupture would be due to the volumetric expansion of the
water.
This necessitated a revision of the encapsulation theory to the consideration of capsule rupture as a triggering event rather than the main explosion.
The compression strength of pure sodium'carbonate'was measured at about 997 psi,
while that of a mixture of about 78% sodium carbonate. with about 22% sodium chloride
was 6060 psi.
It was also noted that contact with.a solution or molten salt might,
by eliminating imperfections, greatly. increase the strength of a solid shell.
*
-37-
With. this concept .in mind-, additions. of.,'a large number of. materials which',might
In all., .118 tests were run.
decrease 'the tensile strength of the smelt.. were tried.
Most. were. not effective in reducing the violence of explosions. Halogen salts
increased. the violence... Sodium aluminate. (which als.o greatly;increased the smelt
viscosity) reduced.the.violence and frequency of explosions but did not eliminate
them.
.
The quenching of smelt was compared to the. quenching of metals. The effects of
water temperature and of dissolved salts.-on the persistence of film boiling and. rate
of cooling.was quoted from French (83). Thus, high water temperatures prolonged
film boil.ing.and.reduced the. rate of cooling; low water temperature or the presence
of sodium hydroxide in solution decreased-the period of film boiling, thus favoring
contact of the cooling water with the hot metal surface.
In the fourth progress report, the possibility of superheat in the water phase
followed by "bumping"."-was presciently suggested as one means by which-water might
become mixed with smelt.
In a further. study of smelt components, sodium oxide and sodium hydroxide were not
found more .reactive than sodium chloride. High levels of sodium sulfate increased
the violence of the reaction of high sulfidity smelts with injected water., In a
further study. of possible quenching solutions, injection of 10% solutions of ammonium chloride or bromide into a 30% sulfidity smelt gave explosions, whereas a 5%
sulfuric acid solution was nonexplosive under .the same experimental conditions.
Pressure rise .times for smelt-water interactions of from 100 to 250 milliseconds
were measured.in one laboratory.. in which water was injected with a hypodermic
needle under the surface. of.molten smelt,. whereas only 0.1 to 3 milliseconds was
recorded in another. laboratory under different conditions in which the water was
injected forcibly from .a.tube po.si.tioned. about. 2 inches above the smelt surface..
Similar pressure transducers and recorders were employed by both laboratories.
In the fifth progress report, it.was reported that. sodium hydroxide was formed and
carbon dioxide lost upon repeated injections' of water into molten sodium carbonate.
As noted earlier, the explosivity of kraft smelts is increased by additions of
Thus., both the. drop in.-temperature .to a critical range and this
sodium hydroxide.
increase in sodium hydroxide.:content might help to account for those explosions
occurring only after multiple injections. It was also noted that exposure of a
smelt to a carbon dioxide atmosphere tended to decrease its explosivity. However,
no sodium hydroxide was detected in the smelt residue after an explosive interaction
of water with an eutectic mixture. of sodium chloride and potassium chloride initially at 1350°F. In one experiment, 'a pressure surge was experienced while heating
up a 35% sulfidity smelt plus about 5% NaOH and 3% water.
With regard to potential quench solutions, a violent explosion was reported in one
test in which a 20% solution of ammonium carbonate was injected into a 30% sulfidity
smelt at about 1600°F. Very violent explosions were also obtained when a 40% ammonium sulfate solution was poured onto molten kraft smelt containing 10% sodium
chloride. Violent explosions were also observed when 51% liquor was injected into
kraft smelt and when about 50.0 grams of kraft smelt was poured into a gallon pail
containing several inches of 50% kraft liquor. On the other hand, no explosions
were experienced when molten smelt was poured into a 60% solution of kraft liquor as
-38-
well as. a 20% and. a. 50% polyethylene glycoL.40.0. solution.
Also no explosion took
place when smelt was poured into a similar pail of water with a greased bottom.
The low elemental. sodium content of. kraft smelt found, after exposure to carbon at
elevated temperatures (less than 0.05% att.1700°E and a-maximum of about 0.15% at
200 0 °F) essentially eliminated.a reaction of. sodium with water as a possible contributor to smelt-water explosions.
With regard. to smelt additives, 5% calcium carbonate and 1% sodium aluminate were
found to. decrease both the explos.ivity and the .tensile strength of kraft smelts.
(from more. than .100.0. psi to about 350 and. 600. psi, respectively)..
Ferric oxide also
moderated.or prevented explosions, while smelt.s with added sodium or aluminum borate
exploded violently.
The sixth progress report in this series was concerned with the probable differences
in the pressure signatures and the. structural damage to be expected from combustible
gas explosions and smelt-water explosions.
The seventh progress report largely summarized the work previously reported.
However, a further suggested mechanism was proposed, in which smelt might dissolve
in the water,. become superheated, and release this superheat suddenly, as has been
observed when heating a concentrated solution of soluble salts on a hot plate.
Probably the main contribution of this work was the conclusive demonstration that
smelt-water explosions were noncombustible.. The key experiment involved obtaining
explosions when water was injected into molten smelt under an inert atmosphere in a
closed container.
A large amount of da.ta..was .obtained from small-scale experiments in which water was
injected into crucibles containing molten smelts.
These showed the same effects of
smelt composition as the dissolving tank work.
The. explosiveness was dependent on
numerous minor variables-. and. was. not. completely reproducible.
(This calls to mind
Colgate's comment that.. "experiments which show extreme sensitivity of explosion
occurrence .to small details of experimental test setup- do so because triggering is
the sole phenomenon being investigated.)
The possibility of using .gas-evolving. quench agents was .tested with mixed results.
Subsequently patents were obtained..for. the use. of solutions of ammonium sulfate or
polymeric glycols. (84) and. solid sodium or ammonium bicarbonate or ammonium carbonate (85) to quench char beds in recovery furnaces.
An empirical study of the effects of agents added to the. smelt was carried out.
Over 118 tests. were run.
Most agents were not effective in reducing the violence of
explosions (halogen .salts increased the violence).
A few agents reduced violence
and sometimes eliminated explosions.
However, no pattern was recognized in the
data, and the attempt to find smelt additives was not successful.
A summary of. the. results of this .work with recommendations of operating safeguards
to prevent sme.lt.-water explosions was released in a special technical report (86).
It was noted.that the exact mechanism involved in smelt explosions had not been
identified..
With regard .to the safe firing solids for black liquor it was also
noted that "the exact value for a particular furnace will probably depend on the
chemical composition of it.s own black liquor and smelt."
-39-
Battelle Review
.
.
.
:.
,
Following the completion of the original smelt-water. study, the FKI Smelt-Water
Technical Committee sponsored project reviews aimed at the development of proposals
for further work at both Battelle Memorial Institute and Arthur D. Little Inc.
The Battelle review (87) concluded on the basis of a theoretical analysis that the.
primary explosion mechanism or triggering mechanism could not be the.previously proposed encapsulation of water in smelt, because the shell of frozen smelt would not.
be strong enough.to contain any significant pressure.
As an alternative, inertial
constraint would. also be inadequate as long, as the encapsulated water was heated
only by infrared radiation and thermal conduction through the smelt and steam.
A limited study to determine the explosiveness at different dilutions of black
liquors from several mill.s was suggested, as was the determination of the effec-.
tiveness of various concentrations of known sensitizers, NaOH and NaCl, as constituents of black liquor.
Consideration was given to the possibility of forming hydrates of the major constituents of smelt, .with the observation that sodium sulfide monohydrate (Na2 S * H2 0)
was stable at very high- temperatures., a temperature of. about 1292°F being required
to obtain an anhydrous salt.
Consideration was aLso given to the relative heats of
fusion and of solution as well as to the sensible heat of the smelt. Thus, sodium
hydroxide and sodium sulfide both have much higher exothermic heats of solution than
does sodium carbonate.. On the other hand, the solution of sodium chloride is
slightly endothermic.
In this regard, it was estimated that about 400 grams of
steam would be generated. if 347 grams of. sodium hydroxide at 1650°F were poured into
500 grams of. water initially at 86°F; physical data..were presented which would
show that about 24% of this heat would come from the heat of solution, about 3.5%
from the heat of fusion and the balance from the sensible heat of the molten salt.
The work of Keevil (88) on the vapor pressure of aqueous solutions at high temperatures was cited. He noted that the vapor pressure of saturated aqueous solutions of sparingly soluble salts with inverse solubility characteristics, such as
sodium carbonate .and..sodium sulfate, approached the vapor pressure of pure water as
the temperature was increased. These salts appeared to become insoluble as the critical temperature of water was approached, so that the solution acted like water.
On the other hand, the vapor pressure of more soluble salts, such as sodium
chloride, remained well below that of.'water.. Further, their solubilities increased
with temperature, so that their saturated solutions could be heated well above the
critical temperature and equilibrium pressure of. water without reaching a critical
point. After reaching a maximum, the.equilibrium vapor pressure decreased to essentially zero as. the temp.eratur.e. and solubility were increased up to the melting point
of the salt.
Thus the. equi.l.ibr.ium vapor pressure of a saturated sodium chloride
solution would reach. a maximumn..of. about 16,330. psig..at a temperature of about 1112°F
and 69% solids, but would.decrease to zero pressure at the melting point of salt
(1474°F).
The Battelle review noted that for such soluble salts, "...the internal molecular
forces in solution appear to be strong. enough to. prevent critical conditions (i.e.,
vaporization) at temperatures as high as 800°C (1480°F)." Here lies one possible
explanation for the effective sensitizing action not only of the alkali halides, but
-40- -
also of two other very. solub.l.e and therma-Ly...s.table smelt constituents, sodium
sulf-ide. and sodium hydroxide.. However, the...relationship of critical temperature to
superheat-limit temperature and, in turn, to superheated liquid explosions was not
made at that time.
The experimental work of...Nelson on additives .to. smelt was reviewed and a summary
table. prepared. This table. is included herein. as Table I.. At the time,.no pattern
was recognized that would indicate why one compound should be more effective than
another. .Now it is apparent that all of the.materials in the nonexplosive column
and many in the mild column are .materials that will liberate CO2 when present in
molten smelt.
Arthur D. Little Review
Arthur D. Little Inc. also prepared a comprehensive review of-past work and proposals for further study (89). Only highlights can be covered here.
A thorough statistical analysis was made of the available recovery boiler explosion
data as.. collected by the Black. Liquor Recovery 3oiler Advisory Committee (BLRBAC, an
organization of insurance carriers, boiler manufacturers and users) with additional
information from the boiler manufacturers.. This covered such items as auxiliary
The..sources of the water as by various pressure
fuel versus smelt-water explosions.
part failures, cooling water leaks, fuel.,. liquor,, or waste stream were listed.
Differences in furnace design and firing-.method were also considered.
The prior research on smelt-water explosions was analyzed in the light of prior and
ongoing studies..of. other molten materiaL.-water reactions,- including work sponsored
by the Atomic Energy Commission on -interaction of nuclear fuel cladding materials
They noted the following -characteristics.of such explosions:
with water.
The. conspicuous absence of explosions between molten glass and
water as. well as coal slags and. water. and, at that time, between
pure molten sodium carbonate and water
The reported violence and explosivelike detonation with shock
waves of sensitive systems
The poor reproducibility of most experiments; introducing the
,element of probability (as. first used by Sallack) as well as
sensitivity and violence
The dependence of explosivity on the specific experimental
method employed
Pressure rise times on the order of 1 to 2 milliseconds, with
maximum pressures up to 5,000 to 23,000 psi close to the
explosion center and 60 to 360 psi at distances of 3 to 4 inches
Some of these features are readily explainable now with our present understanding of
However, they were very mysterious when the
the mechanism. of.vapor explosions.
review was conducted.
-41- :
'ITABLE I
-
Effect of Additives on Explosiveness of Smelt
Violent
NaC I
NaBr
NaOH
Na 2 O
Na2 SO3
Na2 MoO 4
10% Na2 Sn0 3
Na2 S2 03
Na2 ZnO 2
Na2 PbO 2
Na2 SiC3Ì€
Na 2 W04
Na3 PO4
Na3 V04
Mild
Na4 P2 07
Na2 B4 07
Nonexplosive
0.5-5% NaAlO 2
Na2 CrO 4
5% Na2 SnO 3
KA 102
K 2 C03
K2 S304
LiAlO 2
LiOH
Li 2 C03
Li2 B4 0 7
Li2 SO4
GaO
CaSiO 3
GaGN 2
Ca(Al0 2 )2
Ca C2
CaF 2
5, 7 .5, and 20% CaGO 3
Ca S04
GaS
CGa 3 (P04 ) 2
Ba(OH)2
Ba I2
10% BaCO 3
MgO
MgG03
10% SrCO 3
SrO
Mg 304
MgSiO 3
WdS
ZnO
Zn S
Zn3 (P04 )2
Pb3 04
PbS
P
10 and 15% GaCO 3
-42-
TABLE I (Continued)
Effect of Additives on Explosiveness of Smelt
Violent
Mild
Nonexplosive
CuO
H3B0 3
AlPO 4
A1BO 3
A1 2S 3
A1 2 (Si0 3) 3
A1(OH)3
A1 2 (CrO 4) 3
A12 (SO4 )3
FeS
5% Fe2 0 3
Fe3 04
Fe4 (P 2 0 7 )3
Cr 2 O3
Cr 2 (SO0
4 )3
10% NiS
10% NiTiO 3
Ni(OOCCH 3 )2
5% NiS
5% NiTiO 3
NiO
10% Fe203
NiSO 4
CoSO 4
CoS
C0 2 03
MnS
MnSO4
MnO
Mn(OOCCH 3 )2 '
MnC1 2
C
SiC
SiO 2
SnS
CeO 2
TiO 2
Sb 2 S3
ZrO 2
B4 C
With regard to smelt-water explosions, they noted the following:
Reported delays between initial exposure and the explosion of
from a fraction of a second to 1-1/2 hours in laboratory and
mill experience
-43- -
A probable lower melting point and viscosity for.:smelt containing
sodium hydroxide as well as sodium.chl.oride and some of the other
known sensitizers
A decreasing probability of explosions with an increase in the
temperature .of either the smelt or the water .'
.-
The potential for water to dissolve smelt and form a saturated
green liquor at its boiling point., which would be somewhat higher
than that for water
The apparent hydration of smelt by water and the'greater explosivity
of hydrated smelt
The decreased explosivity of some smelts with added calcium car*bonate, sodium aluminate, carbon and iron sulfide
An estimate that about 4 pounds of water in a smelt-water explosion
should be equivalent to 1 pound of TNT'and that 5 to 25 pounds of
water was probably involved in typical destructive furnace explosions
The possibility (in accordance with previous recommendations) that
solutions of gas-producing salts, high-viscosity water-soluble
polymers or 60% black liquor could be used as quench liquids
The fact that the delay times in.the injection experiments in the
previous smelt-water study corresponded to the time for the entering
jet to reach. the bottom of the crucible
It was speculated that there might be. no absolutely safe aqueous quench solution,
i.e., that the proposed solutions.had only a very low probability of producing
explosions. Concern. was .also. expressed about the effects of dilution of such solutions by water from a boiler leak. With regard to black liquors as well as other
quench solutions, none should be classified as "nonexplosive" unless tested in at
least two ways, i.e., smelt into black liquor and black liquor into smelt.
It was further noted that a gradual transition might be expected between a high
probability of explosions with a dilute liquor to very low probability at high
solids levels.
In this sense, the only completely safe liquor would be a dry one.
The frequency was noted of weak black liquor explosions under low bed, start-up or
black out conditions, with direct firing onto the bed., which would favor getting
incompletely dried liquor into the bed.
It was anticipated that the physical properties and behavior of the black liquor would be.determined by such individual mill
factors as wood species, liquor to wood ratio, yield, additives, and tall oil recovery. It was suggested that the net heating value of a liquor might be a more
meaningful variable than the percent solids.
The prior explosion mechanism models (encapsulations, combustible gas, and smelt
hydration) were rejected and a new. "turbulence" model was proposed. The turbulence
model would involve the following processes:
-44-
A metastable condition characterized -by a steam :film between the
two phases (film boiling)
A turbulent mixing. process initiated by a rupture of the steam
film (start of nucleate boiling) and.-involving breakup of both the
aqueous and smelt phase with rocketlike..,acceleration of droplets of
both phases - causing .an..exponential. rate of increase of surface
area. of- both molten smelt and. liquid water .in a rapidly. expanding
volume of steam
A final stage of rapidly increasing.heat transfer from the dispersed
smelt phase- to the dispersed .water.phase by a superheated steam:
atmosphere (superheated boiling)
Calculations- showed that such a mechanism.would be consistent with the results both
Physical properties such as the
of laboratory experiments and furnace explosions.
andl
the aqueous phase "under dynamic
smelt
the
viscosity and surface tension of both
to
influence this turbulent
be.expected
could
conditions of mutual solubility"
explosivity.
mixing and thus to control the
The first two parts of the ."turbulence" model are not that different from some
aspects of. present thinking on. mechanism... The proposed rapid heat transfer mechanism (from.a-superheated s.team...atmosphere.) is not at all realistic.
A comprehensive. program .for further, work was proposed along .three potentially independent lines:
1.
The development of nonexplosive smelt compositions (in analogy
with glass...and .coal slags) by.use of-modifying additives,
especially rod-shaped.insolub.le materials, to change the
theological properties of the smelt
2.
A complete mill systems process control study, which might
ultimately. become-a computer control .system, to prevent those
upsets which produce low solids black liquor
3.
Development. of rapid. smelt cooLing or blanketing techniques,
using suitable liq-uids. or -sol.ids-,. to. be proven effective under
simulated tube leak conditions
Instead, the FKI Smelt-Water Technical
The A. D. Little proposal was not accepted..
Committee selected Batelle-Columbus. to carry out further studies.
Battelle Study
The Battelle study was completed in 1972 and the summary report issued, in January,
1973. This report contains a wealth of information on explosivity vs. the chemical composition of smelts, on the physical. properties of smelt and smelt components,
and on explosion mechanisms.. The Battelle explosivity tests were performed by
injecting. 30. to 1000 milligrams .of..wat.er into a graphite crucible containing about
70 grams of. molten smelt. or by .introduction of.. 80-300. milligrams of water as a drop
-45-
on the end of a ceramic tube, wh:ich:was dropped..into..the crucible of smelt. .It is
probable that these experimental techniques had. some influence both on the experimental results and on. the. conception of an explosion mechanism.
Pertinent observations. and. conclusions are listed below:
Of the reagent grade chemicals evaluated, pure sodium carbonate
could not be exploded.
NaC. and..NaOH were strong explosion
sensitizers, with Na2 S nearly as strong, and with K2 C0 3, Na 2S0 4,
and Na2 SO 3 mild sensitizers in descending order.
CaCO 3 or NaAlO 2 reduced the. incidence of explosions, particularly
at higher temperatures. It was suggested that this might be
related to an increased evolution of CO.2.; It was also noted that
those compounds have limited solubility in smelt and would be
present as solids.
MgO and dolomite were without effect, while additions of)BaCO 3
or Li2 CO3 increased the incidence of explosions. The latter
effect may be due. to melting point changes.
Data are given on the (1) viscosity, (2) surface tension, (3)
density, and. (4) sound velocity for molten smelts. In general
the differences between explosive and. nonexplosive compositions
were relatively small.
Although. in the evaluation of 12 mill smelts it was stated. that
"the injections were made at temperatures from 30 to 100°F above
the melting point of the smelts, which.range from 1250 to 1350°F,"
and it was further noted that "explosions can also occur down
to lower temperatures when NaCl is present, because the freezing
point of the mixture is lower when.the.NaCl is present," no further
consideration was given to the possible importance of smelt freezing point and eutectic temperatures to the explosion mechanism.
The incidence of explosions increased with increasing quantity
of injected water..
The.incidence of explosions was decreased, usually dropping to
zero, as the temperature of the water was increased from room
temperature to the boiling point.
The incidence of explosions
with room temperature water was also decreased, to zero for the
less-sensitive smelts, as smelt temperatures were increased
above the melting point to about 1800°F.
For one smelt, the.substitution of a typical green liquor at
concentrations of 5 to 20% total solids for water as the injection fluid increased the incidence of explosions from 50 to
100%.
The incidence of explosions also increased with the injection
pressure. Thus, two smelts which did not give explosions at 50
-46-
to 75 psi injection pressure. at any temperature could be made
to explode at 1560-1580°F by injections at 100 psi and 1700 to
1730°F at 150 psi.
Explosions did not occur when the injected stream of water was
broken up into a spray before impact with the smelt.
Explosions occurred if the water
thin film of molten smelt on the
crucible; however, they were not
was impinged at a 90 ° angle on a
stream impinged obliquely on a
side wall of the graphite
experienced when the water stream
similar surface film of smelt.
When a pressure transducer was attached to an aluminum rod inserted
into the smelt and the response observed on an oscilloscope, the
explosions occurred within 1 to 2 milliseconds after the water
injection.
Typically 6 to 10 pulses were detected over a
further period of about 1 to 2 milliseconds; however, this
periodicity might. have been related to the natural frequency of
the measuring system. Pressure peaks were typically about 20 psi,,
with some up to 40 psi.
The rates. of carbon dioxide evolution from molten smelt stirred
at high. speeds to produce. cavitation or under vacuum were determined.
The more. explosive smelts containing Na2 S or a combination of
Na 2 S and. NaCi. gave. off significantly less CO2 than Na2 CO 3 alone
(about 58% and 71%, respectively of. the amount corresponding to
the mole fraction of Na 2 CO3 ). On the other hand, smelts containing
the explosion inhibitors. CaCO 3 and NaA10 2 gave off 4 to 8% more
CO2 than..would be predicted from the mole fraction of NaCO 3 . Thus,
it was suggested. that. sensit.izati-on..might be related to the inhibi-:
:
........
tion of C0 2 release
Smelts containing either NaA102 or NaOH foamed upon heating to
higher temperatures. This was attributed to the possible
release of CO2 . Thus, when 5% NaOH. replaced 5% NaCl in a highly
explosive mixture, the mixture began. to foam and ceased to be explosive above 1600°F, whereas the smelt with the NaCl was explosive up
Although it was suggested
to 1800°.F, the maximum temperature tried.
that. the gas evolved was CO2 , it is.more likely that in this case it
was water vapor from the water which would. have been introduced with
the reagent.grade NaOH.
The Battelle report discussed the other known types of physical explosions which
occur when a hot and. cold. liquid are brought together, noting their similarities to
the smelt-water system.
"Investigators generally agree.that volatilization of the
colder liquid, must occur at a very high rate.for a physical explosion to-occur.
For
inertial restraint to expansion., an extremely high rate of heat transfer would be
required, three and four. orders of..magnitude greater than the peak heat flux
involved in the transition .from nucleate to unstable violent boiling.
-47-
"Attaining. a.high. rate. of heat.. transfer is.. generally. thought to involve a high
degree of fragmentation of one liquid and-it.s ..rapid ,dis.per.sion into the other. What
produces the rap.id. -fragmen.t.ation and. dispers.ian of..one liquid into the.other, is a
more speculative matter."
Entrainment as a result of Helmholtz type.instab.ili.ty during injection or by strong
turbulent and-convective flows.in violent boiling 'at.-the- interface was considered,
as well as. Taylor..type instability involved-in .the collapse of a vapor globule.
Also considered was the possibility of the sudden-release of superheat energy of a
film of the volatile liquid heated at. the interface to its superheat limit...
However, this superheat limit concept was .di.scarded..because of "the probable presence of nucleating sources" and a question. as to how the assumed small energy
release could. cause "an appreciable extent of fragmentation and dispersion."
To overcome these self-imposed.restrictions,
a.mathematical model was formulated in
which a large. (i.e., one. pound) globule-of cool water.would be initially surrounded
by a very large reservoir of molten smelt. A partial mixing of this water and
smelt, to give an initial heat transfer rate. ab.out. 1000. times greater than would
occur by simple..contact, is-trigge.red..by. an external event such as the collapse of
one or more small. (one mL) vapor-filled cavities which in turn might have resulted
from the release- of superheat at the interface
.With this very high rate of heat
input, the water. start.s. to boil, generating .a pressure. shock wave in the surrounding
molten smelt. As .the vapor bubble.expands,. it. is alternately compressed and
expanded...by. the. shock wave, meanwhile increasing..the degree of mixing and heat
transfer until all. of the water.. is vaporized.. Whether this is the actual mechanism
in a smelt-water explosion, the model. does. exhibi.t.-.some. of the characteristics of a
smelt-water system, including the time frame- and. peak pressures up to 132
atmospheres and .can account for the. conversion of about 20% of the thermal energy
into kinetic energy and mechanical damage.
According. to the model.,.. generation. or release of. noncondensible gas, such as CO2 at
normal rates. of..heat- -exchange, would.-att.enuate -the pressure cycles associated with
vapor film collapse in the smaller triggering globules. This would explain the
lower incidence of explosions for those. sme.lt.:compositions giving off CO2 .
Other Work
In a report presented at the 1972 Black Liquor Recovery Boiler Advisory Committee
meeting., Nelson (91) pointed out the very great -similarities of the smelt-water
explosion system to the purely physical explosions reported between many molten
metals and.water as well as liquefiednatural gas and water.
These were recognized
at that. time as examples of liquid-liquid (superheated liquid or vapor) explosions.
The characteristics were given as "(1.).a large.. temperature difference between two
touching liquids, (2) direct liquid-to-liquid contact., and (3) the development of a
substantial superheat in the interfacial liquid shell layer of the vaporizing
liquid."
In this paper, the. concept of. encapsulation of.the aqueous phase in a shell of
solidified smelt was abandoned- .In its.place., we now have, a large globule of molten
smelt immersed.in the aqueous. phase and. initially. enveloped in a steam blanket,
under conditions- of film. boiling. As-the system-cools, the vapor film collapses and
the surrounding water or smelt solution is brought into: immediate contact with the
-48-.
In an instant the.water film-in contact with the still molten or
molten smelt.
supercooled smelt, is heat.ed. to. its. Limit.-of.-superheat.,. at which point it explodes.
This superheat explosion creates a shock wave,:-which fragments the smelt and mixes
it with the balance..o.f. the. water. The steam.released ,in this sudden quenching of
the smelt creates the.damaging smelt-water blast.
In the.discussion, it was noted that peak. pressures.. can.be reached in a matter of
microseconds in such.liquid-liquid explosions., as opposed to milliseconds for chemical .reaction explosion types. It was. also. noted.that hot liquid mercury, which
cannot freeze in water, explodes violently when injected into room temperature
water.
It was also observed, without explanation, that explosions had not been
experienced.between molten sodium carbonate and.water. or molten glass and water.
This new theory helped to- explain the previous..observations that hot water or water
containing volatile materials was less likely to give explosions than cold water
without .such additions. These would tend to prolong .the film boiling phase and to
permit. the .surface .of the. submerged smelt to.solidify. Dissolved salts, which have
been found to increase explosivity, would raise. the .boiling point of the aqueous
phase-.and. would .restrain vapor blanketing or. .film.boiling. Small globules of smelt
would. be chilled below their freezing point-. and would, fail to develop the necessary
film of superheated. water, whereas. large .bodies-of...smelt would remain molten for
longer periods.. To prevent smelt-water. explosions, it:was suggested that nucleating
agents be. added.ei.ther to the smelt. or the. water-to prevent the formation of the
superheated films.
(As was- discussed earlier., prevention of vapor explosions by
addition of.nucleating..agents is very difficult.)
Although .this "superheat theory" did incorporate .some. of -the spontaneous nucleation
concepts into the explanation of smelt-water..explosions,. it is not complete since it
does not address coherence.
In a 1974 status report for BLRBAC on the frequency, causes, and prevention of
recovery boiler explosions, Taylor and Gardner (92) summarized the results of 10
years of research with the following statement:
"Research has identified the smeltwater reaction. as an extremely. rapid,.. extremely intense physical explosion consisting. of the explosively .rapid generation of steam. However, the research has not
been encouraging..with respect to the possibility of rendering molten kraft smelts
nonexplosive with water solutions, by means compatible with the chemistry and economics of the kraft process. Thus, as a.practical matter, our principal hope for
reducing the incidence of smelt-water explosions, in smelting recovery furnaces lies
in keeping the molten smelt and water phases apart."
This ended group-sponsored research on smelt-water reactions in North America.
Attent.ion--then. shifted to the very practical matter of preventing water from
entering the recovery furnace.
At this point there was.a lapse in the study of the
smelt-water system, with only scattered references and-very few publications. Payne
(93) suggested that .rapid.combustion of :furnace char dispersed by the smelt-water
reaction might account for much of the upper furnace .damage and for the multiple
explosions reported for some recovery boiler explosions..
In response, Nelson (94)
noted. that. multiple. smelt .exp.losions. had. been.. obtained. in the laboratory and also
reported in.the case of -furnace,explosions. without. a.char bed. The furnace damage
typically- included highly .localized. damage .in the... immediate vicinity of the smeltwater .reaction, but. als.o a more symmetrical .patt.ern. in the upper furnace as the
pressure decreased with distance from the source.
-49-
Revised Mechanism
Meanwhile, having' studied the superheated. l.iquid.. review, article by Reid (2) and the
first smelt-water experiments by Sallack (76), in a series of letters to Howard
Gardner, Shick (95) had begun to question the simple application of the superheated
liquid trigger mechanism .to. smelt-water explosions, under nonimpact conditions. The
problem was to reconcile the 600-700°F temperature difference between the superheatlimit.temperature of. water and the melting. point of the smelt, particularly when the
explosion is delayed.
Could there be a mechanism which would be compatible with a
frozen shell of smelt at the interface.? An experience was recalled in which a
freshly solidified rivulet or sliver of smelt was observed to disintegrate with a
"kerchunk" when placed in a jar of .water. The possibility was suggested that sudden
cooling of the smelt may have prevented crystallization or had introduced strains
into the solidified smelt, similar to those produced.. by the sudden chilling of glass
in the formation of "Prince Rupert's.Drops."
However., further evidence suggests
that such strains are not a necessary condition for an explosive reaction between an
aqueous solution and.a .solidified sme.lt.
Thus., we. have recently had reports of a
case in which small-explosions were observed when .drops of sweat fell on a hot solidified smelt. surface. In addition, Coulter (96) has. recounted another example of
reaction with solid smelt.
Recovery operators on the graveyard shift in one mill
would. take. a.sample of. smelt in a conical sampler,. allow the sample to solidify,
knock it out on the. floor. and throw it..- still hot- - into a nearby lake. The smelt
would explode with. sufficient violence to kill or stun fish.
How can nucleation .be avo.ided- so as to permit development of superheat at such a
solid smelt .surface? Shick sugge.sted..that. the-.wat.er.-woul4 not be in direct contact
with the solid surface, but. instead would.be..in..contact with a very hot saturated
solution of smelt. constituents.
In. support. of this concept, the boiling point of a
saturated high sulfidity smelt. solution had been estimated as about 360°F, whereas
the International Critical Tables gave .the. following .values for the boiling points
of concentrated aqueous solutions of some typical smelt constituents:
95%
67%
28%
30%
NaOH
K2 CO3
NaCl
Na2 CO3
528°F
275°F
228 F
219 0°F
Certainly this table corresponded to the reported relative sensitivities of NaOH,
K2 CO3 , and Na2 CO3 , but would not explain the. very great sensitivity of NaCl. Two
further possibilities remained:
the extent to..which.such concentrated solutions
could themselves be superheated and the effects of such smelt constituents on the
freezing point and/or eutectic of the smelt mixture. These thoughts were also conveyed to the API Recovery Boiler Research Subcommittee (97).
The full potential for increase in the superheat-limit temperature of a salt solution was not realized by these reviewers until. they had studied the articles by
Apfel (21) and Yayanos. (29)., in..the course- of..prepar.ing.a draft interim report on
Emerging Theorie.s. on Explosion Mechanisms. to. the APL.Rec.overy Boiler Committee (98).
Apfel had experimentally demonstrated the theoretically predictable continuity
between the. tensile strength of a..liquid at low temperatures and its superheat-limit
temperature at normal pressures. Yayanos predicted from. compressibility data that a
-50-
25% solution of NaC1 should have a.tensile.strength. .68%. greater than that of pure
water. By parallel extrapolation this..would correspond-to a superheat-limit temperature of about 935°F for the 25% salt solution compared with about 589°F estimated for water.
.
This concept was further. elaborated by Shick (99) in his concentration gradient.
trigger mechanism. He. proposed that.whenever there is contact between smelt and an
aqueous liquid, the properties .of b.oth. are changed at the interface in accordance
with the. solubility of each in the other. .Thus, the freezing point of the interfacial film of smelt is probably depressed by absorption of water,: while the
superheat-limit temperature of the interfacial aqueous phase will be raised by formation of a concentrated (essentially saturated) solution of such soluble smelt
constituents as Na2 S, NaOH, and NaCl.
In this way, nucleation can be delayed even
with a very high interface temperature., permitting a -very significant degree of
superheat to develop in the bulk of the aqueous phase.. Similarly compositions which
will release gases or reduce. the surface tension of the. aqueous phase will reduce
the superheat-limit temperature.,- reduce the possible superheat and reduce the incidence of. explosions. The hypothesized. mechanism can be used to explain most all of
the reported effects of composition on smelt-water exp.losivity. All of the known
smelt sensitizers. (NaOH, Na2 S.,.and .NaCl.) are. species whose solubilities increase
with temperature. This. is unlikely to be just a.coincidence.
Continuing Experiments
Swedish Work
A major study of smelt-water explosions was. recently carried out under sponsorship
of the Swedish Recovery Boi.ler. Committee (50.).
Thes.e. experiments were undertaken to
permit comparison.of. smelt-water blast e.ffects.. with tho.se. of conventional explosives
and to determine the potential for minimizing..furnace damage and danger. to. personnel
by provision of suitable pressure relief areas in the furnace construction.
The experiments.. involved the addition.of.10-100 g.of water into 10-30 kg of molten
smelt. Three addition techniques were used:
1.
Subsurface injection with 145 psi on the piston
2.
Bursting a ceramic capsule containing the water under the
smelt surface by an "electric fuse" (possibly by a spark
or the fusion of a fine wire in the water)
3.
Bursting a ceramic capsule of water under the smelt
surface with a.stronger, .high VA ."fuse." (detonator?)
The effectiveness of the mixing and the violence, of. the. smelt-water explosions were
judged to. be approximately the same for the first two methods of injection, whereas
the third.was. more effective. This was. demons.trated.-.by. high-speed photographs of
simulation tests, in which colored water .was.injected into clear water, as well as
by the violence of the. sme.lt-water explosions.
Thus., the third method gave an
explosive. dispersion of the colored water. in the. clear.water and. also gave about 65%
higher impulse in the smelt-water tests than did -the other methods.
-51-
of
Na2 C03 , Na2 S.,Na2 SO4 , and. NaC were used, including pure
Various combinations.
sodium carbonate as well as smelts with from 15 to 36.% Na2 S and from 0.8 to, 5% NaCl.
Explosions took place for all compositions when the water was forcibly ejected below
the smelt surface. However, the impulse in the pure sodium carbonate smelts was
Beyond that there did not
only about one-half that with the other compositions.
appear to be a systematic trend with smelt composition. In a few experiments in
which water was ejected onto the surface of smelt, no explosions were obtained, the
water simply vaporizing.
It is significant that explosions were obtained with pure Na2 CO3 , since'the earlier
studies at Combustion Engineering and Battelle never gave explosions with Na2 CO3 and
water. This showed that carbonate can react explosively with water when suitably
In this connection it is of interest to note a damaging dissolving tank
triggered.
explosion reported by Ludwig (100) with a nonsulfur system.
In some injection experiments,- a pressure sensor was mounted on the pneumatic
piston. In two examples with 25 grams of water injected., the pressure during injection fell rapidly at .firs.t from the initial. 145 .psi injection pressure to about zero
during the.8--.to .12-millisecond. injection period..
This was followed by a 136- to
238-millisecond. induction period at essentially zero.pressure, which was in turn
followed .by a.very sharp smelt-water explosion pressure spike. The peak in one case
was about 145 psi and. in the other about 360 psi., the higher peak corresponding to
the longer induction period... The ris.e. was.. almost instantaneous followed by a rapid
return to zero, with the entire event taking. place in from 1.25 to 2.5 milliseconds.
Pressures.were also measured. by .sensors.mounted. in the walls of a simulated or model
furnace enclosing the space above the smelt. pot. With injections of 100 grams of
water, the induction period would be followed,, for example, by a rise to a maximum
pressure of about 16 to 17 psi in 7.5 to 15 milliseconds, with a return to zero
pressure in about 33 to 36 milliseconds.
It was found that essentially the same.pressure-time.curves could be satisfactorily
produced with conventional explosives, such as TNT.and especially black powder.
Depending.upon the efficiency of the smelt-water explosion, as determined at least
in part by..the injection procedure, one pound of water was found to be equivalent to
from 0.03 to 0.2 pounds of TNT, or 0.3 to 2 pounds of black powder. With this
information further experiments were conducted with black powder to determine the
probable effectiveness of pressure relief sections in a recovery furnace.
With regard to the relative explosiveness. of water injected into sodium carbonate in
these experiments, the authors. hypothesized that. this would be a function of the
quantity of water injected, rising from- zero with about 50 milligrams, as reported
in the Bat.telle .(90) studies, to 50% at 50 gr.ams in these tests and projected to
reach 100% at about 50. kgThe authors neglected here..the difference in injection
techniques between the two. studies and also the. work. of Rogers et al. (80) in which
the impulse for sodium carbonate when one gallon (3.8 kg) of water was injected
under 20 pounds of sme.lt. was also only about one-third to one-half the pressure
pulse obtained with a kr.af.t-typ.e. sme.l.t.
The. difference in explosiveness might
better be related to differences in the magnitude of triggering forces.
The authors note that. the. maximum..blas.t effect they obtained under ideal conditions
corresponded to about. .0.2.pounds of TNT per pound. of water, vs. a theoretical
value in the literature of about 0.5 pounds per pound.
In actual furnace
-52-
practice, much lower yields would..normally be.expected, because the conditions would
not be ideal.... They sugge.sted that the frequency and violence of smelt-water explosions might therefore be. expected to fall on some sort of probability curve, which
would be useful in specifying the required relief.. ar.eas..to limit the potential for
more serious furnace damage and personnel injury.
There-are serious questions concerning the effectiveness of pressure relief areas in
smelt-water explosions. Large-scale, high-efficiency vapor explosions involve a
detonation wave or shock wave. Damage..occurs wherever the wave impacts on the
furnace structure.
It is only when the wave degenerates to a subsonic pressure wave
that relief surface could have any benefit.
Current IPC Studies
Small-scale studies of.smelt-water interactions are currently in progress at The
Institute of Paper Chemistry... These studies originated when one of us (Grace)
noticed that some of. the substances be.ing.considered. as autocausticizing agents were
the same materials. that had. been partially successful. in reducing explosion intensity.when added to smelt. Autocausticiz.ing...agents. areamphoteric metal oxides or
amphoteric salts which will react with molten Na2 CO 3 to,.l.iberate CO2 and produce a
residual comp.lex.which. wiLl form. caustic directly when dissolved in water. These
reactions can be represented by
Na 2 C0
and
3
+ MxOy + Na20
* MxOy + C02
Na2 0 * MxOy + H2 0 + 2NaOH + MxOy
A review of the earlier work on smelt additives showed a definite pattern in that
most of the materials that reduced.or eliminated.explosions in small-scale tests
were materials that would release CO2 in molten.smelt. The weak relationship that
Battelle (90.). found between. C02 -releasing tendency of.a. smelt and "explosiveness"
was also noted. Accordingly, an experimental program.-to investigate CO2 release
agents as smelt desensitizers was initiated.
The experimental approach was to.use a dropped-tube containing a suspended drop
similar to that used in the Battelle study. Initial.tests using small amounts of
additives (5 to 0% by weight on the smelt) showed a..high degree of temperature
sensitivity and gave very mixed results.. It was. recognized that any beneficial
effects of reduced explosiveness would only be obtained while gases were actively
being liberated, and that the kinetics of gas.liberation would be highly temperature
dependent. Thus the study. shifted.. to a study of the kinetics of gas liberation from
smelt and different additives; this. is currently in progress.
It may be noted.. that..if the hypothes.is..of gas. liberation from smelt acting.as an
explosion inhibitor is correct, then smelt in.which..active reduction is occurring
should be nonexplosive, since CO2 and.. CO. are given off. during reduction. Reduction
is known to be highly temperature sensitive, with reaction rates becoming very slow
in the temperature range. of 14.00-15.00°..... Thus.,. the ideal additive would be a
material that.would generate C02- in..the..low, temper.ature range (between the smelt
freezing..point and...s.ay 1500.°F) while not .reacting.. so fast at higher temperatures that
it was used. up before the smelt.cooled to the critical, temperature range. Thus, it
should have a relatively.low.activation energy (temperature sensitivity).
Unfortunately most of the additives examined-so far- have.-high activation energies.
-53-;
Safe Firing Concentration
One of the sources of smelt-water explosions in recovery boilers has been so-called
weak liquor explosions. These involve an interaction between black liquor and
Since black liquormust be.fired into a recovery boiler as an aqueous
smelt.
liquid, there has been considerable interest in determining the black liquor concentration at which it can be safely fired.
Based on the limited experimental evidence at the time, the Advisory Technical
Committee of the Smelt-WaterResearch Group (86) regarded 55% concentration as the
minimum safe firing concentration. for kraft black liquor. This was-later revised on
the basis of field experience by the BLRBAC Emergency Shutdown Procedure Subcommittee, as noted by Nelson (101), who defined. a "safe" firing black liquor as one which
has a solids concentration of 58% or above. Meanwhile, in accordance with the
original recommendations of the Smelt Water Research Group, Owens-Illinois, Inc.
contracted for measurement of the relative explosivity as well as other physical
properties of the liquors from its kraft and nonsul.fur mills. When these liquors
were compared by multi.ple.injection into a kraft smelt, one of the kraft liquors was
explosive at 45% solids and one at 50%, but not at. higher solids under the conditions of these tests. On the other hand-neither..nonsulfur liquor was explosive at
Here there was a difference
any concentration down to 35% solids, the lowest tried.
between liquors of at least 10 to 15% in the critical solids concentration for
explosivity under identical testing conditions. Clearly percent solids, as such,
cannot be the determining factor on explosivity of liquors in contact with molten
smelt. A summary of these test results was transmitted. to the API Recovery Boiler
Research Subcommittee (102) with the recommendations that further work be undertaken
along the same lines.
-54-
CONCLUSIONS
Smelt-water explosions involve the same basic mechanism as other systems which
experience liquid-liquid explosions. 'The four stages in the development of largescale explosions are:
a quasi-stable intermixed. initial configuration, a triggering
step, escalation to a fully developed. detonation wave.,.-and propagation of a detonation wave through the quasi-stable mixture.. Smelt-water systems generally seem to
be easily triggered (and may be self-triggered). Escalation and propagation steps
involve fragmentation, s.o the viscosities and relative densities of the two liquids
would be important. .Spont.aneous
nucleation processes and sudden release of
superheat are also likely to be important.
According to this mechanism., large..explosions can be prevented by avoiding the
quasi-stable initial configuration, e.liminating-.the triggering pulse, or by inhibiting the escalation and propagation step.. The most. promising avenue seems to be
avoidance. of. the .quas.i-s.tab.le, well-intermixed. initial- configuration. This could be
done in either of. two ways .- by having the system so reactive that small-scale
interactions occur immediately on contact and prevent a large coherent explosion or
by increasing the stability of the film boiling so..that. the water simply evaporates
before an interaction is triggered.
Smelt-water systems differ. from most other liquid-liquid systems in that the components.are solub.le in .each.other.
Although this does not likely change the basic
mechanism, .it does lead to major composition-effects, and possibly to a more sensitive system. The known effects of smelt and solution composition can be explained
by appropriate combinations of the concentration gradient mechanism, the release of
CO2 from Na2 CO 3 and additives, and the viscosities, .densities, and melting points of
the components.
There is increasing evidence, that the size of the system is important in smelt-water
explosions.. Explosions tend to become more.efficient in larger scale systems. This
means that. caution must be used in extrapolating results from laboratory-scale
tests. It also suggests that the most critical parameter in actual smelt-water
emergencies is the amount of smelt. and water which can.come in contact. Reducing
molten smelt inventory in the furnace would then. be.one means for reducing the
potential explosive violence.
L~~~
-55-,
REFERENCES
*
....
:
1.
Reid, R. C. Superheated liquids - a laboratory curiosity and, possibly, an
industrial curse. Chem. Eng. Education Spring:60,;Summer:108, Fall:194 (1978).
2.
Reid, R. C.
3.
Lipsett, G. S. Explosions from mo.lten. materials and water.
118-26(May, 1966).
4.
Witte, L. C., Cox, J. E.,
22(2):39(Feb., 1972).
5.
Witte, L. C., Vyas, T. J., and Gelabert, A. A.· Heat transfer and fragmentation
during molten-metal/water interactions. Trans. ASME..J. Heat Transfer 95(4):521
(Nov., 1973).
6.
Colgate, S.. A. and Sigurgeirsson, T.
244:552(Aug. 31, 1973).-
7.
Anderson, R. P., and Bova, L. Final report on the small-scale vapor explosion
experiments using a molten NaCl-H 2 0 system.. Argonne National Laboratory,
ANL-76-57, April, 1976. 45 p.
8.
Nelson, L. D. and Buxton., L. D. Steam explosion triggering phenomena:
stainless steel and Corium-E simulants studied. with a floodable arc melting
apparatus.. Sandia Laboratories, SAND 77-0998, April, 1977.
155 p.
9.
Board, S. J. and Hall, R. W. Recent advance in understanding large scale
vapour explosions. Central Electricity Generating Board, CEBG-RD/B/N-3666,
April, 1977. 32 p.
Superheated liquids.. Am.
Scientist 64:146-56(March-April, 1966).
and Bouvier, J. E.
Fire Tech.
The vapor explosion.
2(2):
J. Metals
Dynamic.mixing.-of water and lava.
Nature
10.
Segev, A. Liquid-liquid contact in vapor explosion.
Laboratory, ANL-8-92, August, 1978.
188 p.
11.
Henry, R. E. Test plan:
large scale molten salt - water vapor explosion
studies to be conducted at Ispra, Italy. Argonne National Laboratory,
ANL-79-20, Feb., 1978.
33 p.
12.
Henry R. E. and Fauske, H. K. Nucleation processes in large scale vapor explosions. Trans. ASME, J. Heat Transfer 101(2):280(May, 1979).
13.
Cronenberg, A. W. On the similarities and differences in vapor explosion criteria. Idaho National Engineering Lab, CONF-781105-19, 1978. 5 p.
14.
Corradini., M. L. Phenomenological modelling of. the small-scale vapor explosion
experiments.
Sandia Laboratories, SAND 7.9-2003, April, 1979. 82 p.
15.
Anon. Mississauga:
a disaster that became a miracle.
58(47):23(Nov. 24, 1.980) .
16.
Reid, R. C. Possible.mechanism for pressurized-liquid tank explosions or
Science 203(23):1263(March, 1979).
BLEVE's.
Argonne National
Chem. Eng. News
-56-
17.
Long, G..
Explosions of molten aluminum-.in-water .-
cause and prevention..:
-Metal
Progr. 71(5):107(May, 1957).
18.
Porteous, W. Superheating and cryogenic ,vapor'explosions. Ph.D. Thesis,
Massachusetts Institute of Technology, Cambridge, Massachusetts, 1975.
19.
Miller,.. R. W.,...Sparo, A. H., Dugone., J., Wieland, D. D., and Houghfailing, J.
E. Experimental results and damage effects'of. destructive test.' Trans.'Am.
Nuclear Society 6(1):138(1963).
20.
Water superheated to 279.5°C at atmospheric pressure.
Apfel,. R. E.
Physical Science 238:63(July 24, 1972).
21.
Apfel, R.. E. A novel technique for measuring the strength of liquids.
.
Acoust. Soc. Am. 49(1):Part 2, 145(1971).
22.
Gunnerson, F...S.., Cronenb.erg,. A.. W. Thermodynamic prediction of the temperature for..film boiling destabilization and its relation to vapor explosion
15 p.
phenomena....Idaho National Engineering..Lab., CONF-780622-36, 1978.
23.
Hirschfelder.,. J. O. Kinetics of-hhomogeneous nucleation in many component
systems. Theoretical Chemistry Institute, The University of Wisconsin,
WIS-TCI-510, April 5, 1972(?).
24.
Beegle, B.. L.,. Mo.de.ll, M., and Reid.,..R...C.. Thermodynamic stability criterion
for pure substances and. .mixtures. AIChE..J. 20(6):1200(Nov., 1974).
25.
Porteous, W. and. Blander,. M. Limits of superheat and explosive boiling of
light hydrocarbons, halo.carbons and hydrocarbon mixtures. AIChE J. 21:560
(1975).
26.
Lienhard,
systems.
27.
Mori, Y., Hijikata, K., and Nagatani, T. Effect of dissolved gas on bubble
nucleation. Int. J. Heat Mass Transfer 19:1153(1976).
28.
Forest, T. W. and Ward, C. A.. Effect of a dissolved gas on the homogeneous
nucleation pressure of a liquid. J. Chem.. Phys. 66(6):2322(March 15, 1977).
29.
Yayanos, A. A. Equation o.f'state. for P-V .isotherms of water and NaCI solutions. J. Appl. Physics 41(5):2259(April, 1970).
30.
Buxton, L. D.., Nelson, L. S., and:Benedick, W. B. Steam explosion triggering
3 p.
and efficiency studies. SAND-78-0261C, 1979.
31.
Skripov, V.., Ermakov, G., Sinitsin, E., Pavlov, P., Baidakov, V.', Bulanov, N.,
thermophysical properties,
Danilov, .N.., and Nikitin, E. Superheated. liquids.:
homogeneous nucleation and explosive boiling:.up... Presented at 'AIChE-ASME Heat
Transfer. Conference, Salt. Lake City, August 15-17, 1977.
32.
Flory, K., Paoli,. R., and Me.sler R.
Prog. 65(12):50(Dec., 1969).
Nature
J.
J. H. Some generalizations of..the stability of liquid-gas-vapor
Int. J. Heat and Mass Transfer 1:813(1964).
Molten.metal-water explosions.
Chem. Eng.
-57-
33.
Carslaw, H. S; and Jaeger, J.'C. Conduction o'f heat in solids.
tion. Clarendon Press, Oxford, 1959.
'
.
..-
Second edi:.
34.
Fauske, H. K. On theniechanism of'uranium dioxide - sodium explosion interactions. Nuc. Sci. Eng. 51:95(1973')..
35.
Williams; D. C. The importance of scale and triggering effects in vapor explosions. Sandia Labs., SAND-77-1101C, 1977. 6 p.
36.
'Wright, R. W. and Humberstone, G.'H.
Dispersal and pressure generation-by
water impact upon molten:aluminum. Trans. Am. Nucl.'Soc. 9:305(1966).
37.
Reynolds, J. A., Dullforce, T. A., Peckover, R. J., and Vaughn, G. J. Fuelcoolant interactions-some basic'studies. at the UKAEA Culham Laboratory. Proc.
Third Specialist Meeting on SFI in Fast Reactors, .PNC N251 76-12,' 97:(1976).
1p.
38.
Henry, R. E; and McUmber, L. M. Vapor-explosion experiments with subcooled
freon. Argonne National Lab.', ANL-77-43, June, 1977. '32'p.
39.
Buxton, L. D. and Benedick, W. B. Steam explosion efficiency studies.
Labs., SAND 79'-1399, Nov., 1979. 60 p:
40.
Henry, R.. E. and Fauske, H.K. .Energetics of vapor explosions. Paper
75-HT-66, presented at AIChE-ASME Heat Transfer Conf., San Francisco, CA, Aug.
11-13, 1975.
41.
Porteous, W. M. and Reid, R. C;- Light hydrocarbon vapor explosions. 'Chem.
Eng.. Prog. 72(5'):83-9(May, '1976).
42.
Anderson, R. P. and Armstrong., D. R. Comparison between vapor explosion models
and recent experimental results. AIChE Symposium Series 138(70):31-47(1973).
43.
Buivid, M. G. and Sussman, M. V. Superheated liquids containing suspended particles. Nature 275:203(Sept. 21, 1978).
44.
Shanes, L. M., Ph.D. Thesis, Massachusetts Ins:titute of Technology, 1977.
45.
Harrison,' L. J. and Ivins. A photographic study of-the transient meltdown
behaviour of aluminum/uranium fuel plates in water. Trans. Am. Nucl. Soc.
9:305(1966).
46.
Hess, P. D.. and Brondyke, K. J. Causes of molten aluminum-water explosions and
their prevention. Metal Progress 95(4):93(April, 1969).
47.
Asher, R. C., Bullen, D.., and Davis, D. 'Vapor explosion (fuel-coolant
interactions) resulting from the subsurface projection of water into molten
metals: ' preliminary results;' U.K. Atomic Energy Authority, Harwell,
'
14 p.
N77-19840/6SL, March, 1976.
Sandia
'
48." Bankoff, S'. G., Sharon, A.,' Maeshima, M., and Inoue, A. Mechanism of vapor
explosions. Northwestern University;' C00-2512-10, 1977.
16 'p.
I
-5849.
Henry, R. E. and.-Cho, D.. H. An evaluation of. the.potential for energetic fuelcoolant interactions in hypothetical. LMFR-.accidents.. Presented at ASME Winter
Annual Meeting, Atlanta, Georgia, 1977.
50.
Bergman,. S. G...
A. and..Laufke., H. RecoYery boiler explosions. -The
Steam Users Association SPC1-Report No. 35.
51.
Peckover, R. .S., Buchanan, D.. J., and-Ashby, D.,E. T. F. Fuel-coolant interactions in submarine vulcanism... Nature 245:307(0ct.. 12, 1.973).
52.
Henry., R. E. and, McUmber, L. M. Vapor explosion potentials under LWR hypothetical accident conditions. Proc. Light Wat.er.Reactor Safety Meeting., Sun
Valley, ID, Conf.-770708, Vol. 3:414-25(Aug., 1977).
5.3.
Nelson, L. S. and Buxton, L. D., Trans. ANS, Vol. 28:448(1978),
See also
effects of pressure on steam explosion triggering inCorium-E simulants, Trans.
ASME 101:448(May, 1979).
54.
Henry, R. E...,Hohmann, H., and.Kottowski.. The effect of pressure on NaCl-H 2 0
explosions. Fourth CSNI Specialis.t Meeting. on Fuel-Coolant Interaction.in
Nuclear Reactor Safety,. Bournemouth, U.K., FC14/P14, April 2-5, 1979.
55.
Kottowski, et al. Importance of the..coolant. impact on the violence of the
vapor explosion. Fourth CSNI Specialist -Me.eting...on Fuel-Coolant Interactions
in Nuclear Reactor Safety, Bournemouth, England, April, 1979.
56.
Hicks, E. P. and Menzies., D.. C. Theoretical .studies on the fast reactor maximum..accident... Proceedings. of-the.Conference on Safety, Fuels and Core Design
in LargeFast Powe.r...React.ors,.Argonne-.Natio.na.l.-Laboratory,
Illinois, October
11-14, 1965; Argonne. Laboratories Report ANL-7120,. 1965:654-70.
57.
Epstein, L. E. Recent.developments.in the study. of.metal-water reactions.
Nucl. Energy, Part .B 3:,273-95(1962).
58.
Brauer,.F. E.. Metal/water explosions.
59.
Higgins, H. M. The.reaction.of molten uranium..and zirconium alloys with water.
Aerojet General Report No. Agc-AE-7, April 30, 1965.
60.
Stevens, J. W., and Witte, L. C. Transient vapor film behavior during
quenching. Trans. Am. Nucl. Soc. 15:418(1972).
61.
Buchanan, D. J. and Dullforce, T. A.
245:32(Sept. 7, 1973).
62.
Ochiai, M. and...Bankoff, S.. G.... Local. propagation.. theory for vapor explosions.
Northwestern. Univ. Evanston, Illinois, ERDA,. NU-2512-5, Feb. 3, 1976. 23 p.
63.
Froehlick, G., Mueller, G., and Unger, H.. Analysis of shapes of solidified
melts from entrapment experiments. Trans. Am.. Nucl. Soc. 28:449(1978)..
64.
Cronenberg., .A. W., Chawla,. T... C.., .and..Fauske.?.H. K.. A..thermal stress mechanism
for the fragmentati.on..of molten UO2 upon contact with sodium coolant. Nucl.
Eng. Design 30:434(1974).
Swedish
J.
Nuclear Science & Eng. 31:551(1968).
Mechanism for vapor explosions.
..
Nature
gas. release phenomenon.
Nucl. Sci. Eng.
65.
Epstein, M. Thermal fragmentation -a
55:462(1974).
66.
Enger, T. and Hartman, D. Rapid phase transformation during LNG spillage on
water. Proc. Third Intl.. Conf. on LNG, Washington, D.C., 1972.
67.
Nakanishi, E. and Reid, R. C. Liquid natural--gas-water reactions.
Eng.. Prog..67(12):36-41(Dec., 1971).
68.
Fauske, H. K. The 'role of nucleation in vapor explosions.
Soc. 15:813(1972).
69.
Henry, R. E. A correlation for the minimum'film boiling temperature.
Natl. Heat Transfer Conf., Atlanta, Georgia, 1973.
70;
Board, S. :J. and Hall, R. W. Propagation of thermal explosions Part -2: A
theoretical model. CEGB&Report RD/B/N3-249 Berkeley.Nuclear Laboratories, Dec.,
19-74.
..
Chem.
Trans. Am. Nucl.
Fourteen
:
.
Detonation of fuel coolant explo-
71.
Board, S. J., Hall, R. W., and Hall, R. S.
Nature 254-:319-21(March 27, 1975).
sions.
72.
Bankoff, S.. G., Sharon, A., Maeshima, M., and Inoue, A. Mechanisms of vapor
explosions, Report No. COO-2512-11, Northwestern Univ., 1977.' -
73.
Rabie, R. L. and. Fowles, G. R.
22(3):422(March, 1979).
74.
Fowles, G. R. Vapor phase explosions:
204:168-9(April 13, 1979).
75.
Roberts, F. W. The evolution of the modern recovery unit for' soda and sulfate
pulp mills. Paper Trade J. 131(22):191-4, 196-200, 202(Nov.'30, 1950).-
76.
Sallack, J. A. An investigation of explosions in the soda smelt dissolving
operation. Pulp Paper Mag. Can. 56(10):114-18(Sept., 1955).
77.
Nelson, W. and Kennedy, E.. H; What causes kraft dissolving tank explosions. I. Laboratory experiments.
II. Mill investigations. .Paper Trade J. 140(29):
50-6; (30):30-2(July 16, 23, 1956).
78.
Gettle, T. J. Apparatus and method of operating a chemical recovery furnace.
U.S. pat. 3,122,421(Feb. 25, 1964); Can. pat. 641,338(May 15, 1962).
79.
Honda, M., Tannov, T., and Morimoto, H. Method. for preventing vapor explosions
in a liquid comminuting process. U.S. pat. 4,218,412(Aug. 19, 1980).
80.
Rogers, C. E., Markant, H. P., and Dluehosh, H. N.
water. Tappi 44(2): 146-51(Feb .,.1961). '
81.
The Institute of Paper Chemistry. Research on smelt water reactions. Project
2419-. Reports One through Seven to Members of Smelt Water Research Group
(Reports One,'Aug....21, ..1964.;. Two, Nov. 20, 1964; Three, Sept'. 27, 1965; Four,
Dec. 14, 1965; Five and Six, June 30;, 196-5;...and Seven Aug. '15, 19'66).
The polymorphic detonation.
Phys. Fluids
Elementary detonations?
Science
Interreactions of smelt and
-60-
82.
Nelson, W.
Personal communication, Feb. 3, 1981...
83.
French, H. J. The quenching of steels.. -The American Society for Steel
Treating, Cleveland, Ohio, 1930.
84.
Nelson, H. W. and Norton, C. L. Method of preventing smelt-water explosions.
U.S. pat. 3,447,895(June 3, 1969).
85.
Nelson, H. W. Preventing physical explosions due. to the interaction of liquid
water and molten chemical compounds.. U...S. pat. 3,615,175(0ct. 26, 1971).
86.
Smelt-Water .Research Group... Explosions in kraft recovery furnaces.
Southern
Pulp and Paper Manufacturer, 29(.11):40-1, 44, 46, 48(Nov. 10, 1966); Tappi
49(12):51-55A(Dec., 1966).
87.
Lougher, .E...H.,. Blue., G., Goddard, S.., Gurey, H. S.,. Miller, J. F., Putnam, A.
A., and Simon,..R.. Summary..report on feasibility study on smelt-water-explosions to Fourdrinier Kraft Board .Inst.itute,...Inc. (Battelle Memorial Institute)
May 3, 1968.
88.
Keevil, N.. B. Vapor pressure of aqueous solutions at high temperatures..
Am. Chem....Soc. 64:841-50(April, 1942).
89.
Arthur D. Little, Inc.
Feasibility of further research on smelt-water explosions. Report to Fourdrinier Kraft Board-Institute, Inc., C-69480, July, 1968.
90.
Krause.,. H. H.., Simon, R., and Levy, A. Final report on smelt-water explosions
to Fourdrinier Kraft Board Institute, Inc. Battelle Columbus Laboratories, Jan.
31, 1973.
91.
Nelson, W. A new theory to explain physical explosions. Presented at Black
Liquor Recovery Boiler Advisory Committee Meeting, Atlanta, Georgia (Oct. 11,
1972),. Tappi. 56(3):121-5(March, 1973).
92.
Taylor, M. L. and Gardner, H. S.
57(11):76-8(Nov., 1974).
93.
Payne, G. Suggested new mechanism for recovery furnace explosions.
Can. 76(5):130(May, 1975).
94.
Nelson, W. Sme-lt-wat.er explosions
Paper Can. 77(5):100(May, 19.76).
95.
Shick,. P....E.... Personal communications to Howard.. S...Gardner, May 6,
July 27, and Aug. 2, 1976).
96.
Coulter, J.. H.
97.
Shick, P. E. Personal communication to Peter E. Wrist with copy to Loraine
Atter, Aug. 5, 1977.
98.
Grace, T. M. and Shick,* P. E.. Emerging theories on explosion mechanisms.
Draft. report .to the API Recovery Boiler Research & Development Subcommittee,
prepared for the meeting in Chicago, May 6, 1980.
J.
Causes of recovery boiler explosions.. Tappi
Pulp Paper
in kraft chemical recovery furnaces.
Pulp
June 4,
14,
Personal communication, March 5, 1981.
-61-
99.
Shick, P. E. Concentration-gradient trigger mechanism for smelt-water explosions. Presented at The American Paper Institute Annual Recovery Boiler
Committee Meeting in Chicago, Oct. 30-1, 1980.
100.
Ludwig, R. M. Report to the API Recovery Boiler Research Subcommittee Meeting
in Chicago, May 6, 1980.
101.
Nelson, W. Relation of firing method on a C-E kraft chemical recovery boiler
to explosion safety (part of a talk delivered at BLRBAC in Atlanta on Oct. 5,
1977).
102.
Shick, P. E. Personal communication to Peter E. Wrist, copy to Loraine Atter
(Oct. 4, 1977). Distributed to API/ADL. Study Committee Meeting in Chicago,
Oct. 21-28, 1977.
I
PST HASELTON LIBRARY
5 0602 01064624
Download