Maintaining Nuclear Weapons
Safe and Reliable Under a CTBT
by
Richard L. Garwin
Senior Fellow for Science and Technology
Council on Foreign Relations, New York
and
IBM Fellow Emeritus
IBM Research Division
P.O. Box 218
Yorktown Heights, NY 10598
Tel: (914) 945-2555
fax: (914) 945-4419
Email: RLG2 at watson.ibm.com
Many papers at http://www.fas.org/rlg
"The Comprehensive Test Ban Treaty and
US National Security Interests"
AAAS Annual Meeting
San Francisco, CA
February 16, 2001
ABSTRACT. I discuss nuclear weapons in general and U.S.
thermonuclear weapons in particular. These have been proven
by nuclear explosive development tests and production
verification tests, and need to be maintained for several or
many decades without further nuclear explosion testing,
under the zero-threshold Comprehensive Test Ban Treaty
(CTBT). The basis for stockpile stewardship has always been
a strict surveillance program by which eleven nuclear
weapons of each type are now removed annually from the
stockpile and inspected; one is destructively dismantled by
the nuclear weapon laboratories. Most of the 4000 or so
parts of a nuclear weapon can be thoroughly tested without
nuclear explosions and may not even be used in an
underground nuclear test.
The weapons can be maintained safe and reliable by
remanufacture of the components-- the primary and secondary
nuclear explosives. The Science-Based Stockpile Stewardship
Program (SBSSP) provides increased understanding of the
necessity for remanufacture, thereby permitting delay in
remanufacture, in some cases, because inspection and
analysis shows that a particular age-related defect is not
significant. The role of advanced computation facilities
(teraflop computers), new flash radiographic systems, and
the National Ignition Facility will be evaluated for the
Stockpile Stewardship Program. The surest way to lose
confidence in the nuclear weapon stockpile is to orient the
Stockpile Stewardship Program toward the introduction of new
designs of the nuclear components. The goal of the SBSSP
should be to maintain the weapons as reliable as they were
during the days of nuclear testing. Historically, weapons
are not most reliable when they first enter stockpile,
because of infant mortality.
Of the other four nuclear weapon states under the NPT,
France and Britain appear to be emulating the United States
in the SBSSP, while Russia and China are more likely to rely
on scheduled remanufacture. Either way they will be able to
maintain a stockpile of proven nuclear weapons safe and
reliable for many decades or centuries. The essential point
is not to make changes, even multiple relatively small
changes, that will eventually lead to lack of confidence.
1046MNWS 021501MNWS Draft 3 Final 02/16/01
INTRODUCTION.
Three important aspects of the analysis of cost and benefits
of a Comprehensive Test Ban Treaty (CTBT) are:
o Can U.S. nuclear weapons be maintained safe and reliable
without nuclear explosion testing?
o Is there a non-proliferation benefit to the United
States in limiting nuclear weapons to those that might
be developed without testing?
o Can the CTBT be adequately verified? What kinds of
nuclear weapons might be developed by experienced and by
inexperienced states below the limit of detectability of
the International Monitoring System and the U.S.
unilateral capabilities?
In this talk, I treat only the first(1) of these topics.
WHAT ARE NUCLEAR WEAPONS?
I have been involved with building and testing U.S. nuclear
weapons from 1950 to the present day, so what I am about to
tell you is limited by what can be said legally and
prudently.
All nuclear weapons are based on a fission chain reaction
carried by neutrons. In a nuclear weapon, a supercritical
mass of fissionable material needs to be assembled, such
that a few fast neutrons introduced will cause fission.
Since 2.5 to 3.5 neutrons are emitted per fission in U-235
or Pu-239, in an infinite medium each neutron disappearing
gives rise to about two new net neutrons, and the neutron
population (and energy density due to fissions that have
occurred) grows exponentially as e&alpha.t. The
"infinite-medium &alpha." corresponds to no loss of neutrons
from the assembly. For a mass that is just (prompt)
critical, the neutron population remains constant, and the
fission energy density increases linearly with time rather
than exponentially.
The &alpha. for an infinite medium increases linearly with
density, since the time between generations depends on the
mean free path of the neutrons, which is inversely as
density. Finally, for any geometrical configuration of
fissionable material (even including a reflector of neutrons
such as beryllium or uranium-238), the amount of material
required to form a critical mass is inversely as the square
of the density.
THE "GUN-TYPE" NUCLEAR WEAPON. The nuclear weapon that
destroyed Hiroshima August 5, 1945, was a gun-assembled
U-235 bomb yielding some 13 kilotons (kt) of high-explosive
equivalent energy-- 13 x 4 TJ. It weighed some 8000 lbs and
is reported to have contained some 60 kg of highly enriched
uranium. Just as important as a supercritical assembly to
make a nuclear weapon is the sub-critical nature of the
material up to the moment a chain reaction is desired.
U-235 has such small spontaneous neutron emission (and the
flux of neutrons from cosmic rays is sufficiently low), that
a normal artillery piece with a projectile velocity of
300 m/s will assemble two sub-critical chunks of U-235 into
a single supercritical mass in a time short enough that the
probability of pre-initiation is acceptably small-- moving
from the just-critical assembly to the maximum criticality
in a time that might be 0.2 m/300 m/s = 7 ms. The Hiroshima
design had a neutron generator to initiate the chain
reaction at maximum criticality. The complete fission of one
kg of material liberates the equivalent of 17 kt.
THE IMPLOSION WEAPON. By placing the U-235 or plutonium at
the center of an assembly of high explosive, in which a
converging spherical detonation wave is launched, even solid
metal can be compressed so that a subscritical mass can
rapidly be made supercritical. With a detonation speed of
0.6 cm/microsecond, a 10 microsecond interval may be
appropriate for this compression. A neutron generator may
be optional for a gun-assembled nuclear weapon, if the
projectile can be stopped so that the supercritical mass
awaits a cosmic-ray neutron, but a neutron generator is
essential for the implosion weapon, which will just bounce
and rapidly become subcritical. But the gun is totally
unsuitable for plutonium, which, with its nominal 6% Pu-240
content and a 6-kg mass in the Nagasaki bomb, emits about
0.36 neutrons per microsecond by virtue of spontaneous
fission of the Pu-240.
Because plutonium has both a larger fission cross-section
and more neutrons emitted per fission, and density
comparable to uranium, smaller amounts of Pu are needed than
U-235. The 6-kg ball of the Nagasaki bomb gave a yield of
almost 20 kt.
EVOLUTION OF FISSION WEAPONS. Implosion is the assembly
method of choice for U-235 as well as Pu. By 1951, with the
aid of nuclear explosion tests, the U.S. had produced and
stockpiled the "half-size" implosion weapon-- the Mk-7,
weighing 1800 lbs and with a yield considerably larger than
the 20 kt of the Nagasaki bomb. South Africa did build six
gun-assembled U-235 weapons, without any explosive test,
just as the United States did not test its gun before using
it on Hiroshima. U.S. nuclear weapons (and perhaps those of
the other four Nuclear Weapon States under the 1970
Non-Proliferation Treaty, NPT) now consist of hollow
plutonium shells surrounded by high explosive. The shell
can be accelerated and then abruptly arrests itself by
symmetry, leading to greater compression than can be
achieved in a solid sphere. Accordingly, the Pu content in
a U.S. nuclear weapon is as little as 4 kg.
BOOSTED FISSION WEAPONS. In 1951, the United States first
tested "boosting." In the modern boosted fission weapon,
deuterium and tritium gas are present in the hollow
plutonium shell at the time of implosion. The d-t gas is
compressed and then suddenly heated by the fission reaction,
which provides thermonuclear reactions between d and t (at
about 100 times the rate of that between d and d). Each d-t
reaction yields a 14-MeV neutron, which has a high
probability of causing fission in the surrounding Pu, which
in turn contributes 150 MeV to the energy of the explosion.
Thus, it is not the energy from thermonuclear boosting, but
the additional flash of neutrons and fission that "boosts"
the fission chain reaction to a higher level and multiplies
the energy release.
In U.S. nuclear weapons, the deuterium and tritium are kept
in high-pressure steel gas "bottles" and provided to the
hollow plutonium shell after the weapon is launched, but
before the high explosive is fired. The plutonium shell is
surrounded by a welded metal enclosure; the whole metal
assembly is a "pit", as it was called in the earliest days
of implosion weaponry.
TWO-STAGE THERMONUCLEAR WEAPONS. In 1951, Edward Teller and
Stanislaw Ulam at the Los Alamos Scientific Laboratory in
New Mexico published a classified paper on "radiation
implosion," (an officially declassified term) in which the
energy of the x-rays in the thermal electromagnetic field is
used to compress and prepare a secondary charge of
thermonuclear fuel and to ignite it by bringing the
compressed fuel to a high temperature. The boosted fission
primary and the secondary are contained in a radiation case,
of material of a high atomic number, such as uranium.(2)
These are called "two-stage" thermonuclear weapons to
distinguish them from other approaches in which substantial
amounts of fusion fuel are in close proximity to the fission
explosive.
Some orders of magnitude may be of interest here. At the
high temperatures resulting from a nuclear explosion,
everything within the bomb is a plasma-- only the heaviest
nuclei have even one electron attached. The thermal kinetic
energy is thus 3/2 kT for each of the particles, and the
particles are by number predominantly electrons. So there
is about one mole of electrons per two grams of material.
If 17 kt of energy were to be confined to 6 kg of Pu, the
temperature (ignoring the black-body radiation field) would
be 170 keV. But this same 70 TJ of energy is far from
enough to heat the vacuum(!) to 170 keV. Indeed, the black
body energy density in the electromagnetic field goes as T4,
and is about 13 GJ/liter at T = 1 keV, or about 5 GJ/0.4 l
at 1 keV. The initial volume of 6 kg of Pu at 15 g/cc is
just 0.4 l. The yield of 70 TJ would heat this same volume
to 11 keV, so that the matter temperature would also be held
to 11 keV.
In fact, the opacity even of Pu is sufficiently low at such
temperatures that much of the black body radiation leaks out
to fill the radiation case at a considerably lower
temperature-- for instance, lower by the fourth root of the
ratio of the volumes. Thus if the radiation case has a
volume of 50 l, the ratio of volumes is some 125, and the
temperature could rise to as much as 3 keV.
Since a given energy density of radiation has almost as much
pressure as the same energy density in matter, the pressure
squeezing on the secondary charge corresponds in this case
to some 70 TJ/50 l or about 1.4 TJ/l. High explosive
provides something like 4 MJ/l, so the radiation implosion
proceeds with a pressure some 0.3 million times that of high
explosive.
RECAPITULATION. The United States has roughly ten types of
nuclear weapons in its stockpile of perhaps 12,000 warheads
and bombs. All are radiation implosion weapons, with
gas-boosted hollow plutonium primaries, and secondaries as
described. The secondary fuel is normally solid lithium
deuteride (LiD).
The nuclear explosion is obtained when electrical firing
pulses are applied with sufficient simultaneity to all of
the electric detonators, which in turn detonate a booster
pellet and then the main charge of high explosive.
Initially, a spherical converging detonation wave was
obtained from the multiple diverging detonation points by
the use of "lenses" of various types-- initially of fast and
slow explosives, later with "ring lenses" to reduce the
enormous mass of the lens region, and still later with "air
lenses" or other means to obtain a detonation wave of the
appropriate shape. The d-t gas from the reservoirs or
bottles enters the pit via an explosively operated valve.
STOCKPILE STEWARDSHIP.
A 1979 memo from the Director of the Los Alamos Scientific
Laboratory to DOE headquarters,(3) states:
"Will the nuclear device work? The reliability of a
nuclear weapon (that has been tested) is determined
primarily by the reliability of its non-nuclear
components and not by its nuclear components. The
non-nuclear components can be demonstrated to be
statistically reliable to a desired level, normally 98%
or better, by doing a sufficient number of non-nuclear
local tests. This procedure has not been altered by
the Threshold Test Ban Treaty and will not be altered
by the Comprehensive Test Ban Treaty. The reliability
of the nuclear performance of a weapon is a different
matter. It is not a statistical quantity. After a few
key nuclear tests are conducted, it is the judgment of
the Laboratories, based upon 30 years of design
experience, that the nuclear performance is guaranteed
if the non-nuclear components function as desired and
the nuclear components are maintained in their as-built
condition. A very important conclusion, then, is that
there has been no reduction in nuclear weapon
reliability as a result of the TTBT and that there will
be none under a CTBT if we utilize current nuclear
systems which have been tested or utilize previously
tested nuclear components as subsystems."
In the United States, nuclear weapons are designed,
developed, manufactured, and cared for by the Department of
Energy. In a sense, they are loaned or transferred to the
military to be ready for delivery as bombs, missile
warheads, or (in the old days) nuclear-armed torpedoes,
anti-aircraft rockets, atomic demolition charges, and the
like. The Los Alamos National Laboratory and the Lawrence
Livermore National Laboratory, both operated for DOE by the
University of California, have responsibility for the
nuclear components-- roughly those within the radiation case
of a thermonuclear weapon. Sandia National Laboratory, with
its main facilities at Albuquerque, NM, and a smaller site
at Livermore, has responsibility for the non-nuclear
components, such as batteries, arming, firing, and fusing,
Prescribed Action Links or other use-control system,
state-of-health monitors, and the like. Of the 4000
components of a typical bomb, all but a few are Sandia's
responsibility.
In analyzing the difference between an era of underground
nuclear tests and the CTBT, it must be recognized that most
of these non-nuclear components are not exercised at all
during an underground nuclear test. For instance, the
environmental sensing system that ensures that a nuclear
weapon will not explode until it has been through the
appropriate accelerations, coast, and the like, is clearly
not applicable to an underground test. Such components for
the most part are testable on the bench. Those that are
destroyed by their activation (such as explosively fired
valves or thermal batteries) are tested in large quantities,
so that those that have not been tested are assured of
performance as random selection from a lot that has been
tested to ensure the desired confidence in reliability.
Explosive detonators (in the first implosion weapons, a
bridge wire of noble metal fired by a substantial pulse of
electrical energy) were initially used in the almost 100
lenses of an early nuclear weapon. Many thousands of bridge
wires had to be fired without fail in order to have
confidence in adequate reliability that not a single one
would miss when fired "for effect."
The Stockpile Stewardship Program has two goals: (I) a safe
and reliable stockpile of nuclear weapons of existing type,
and (II) a viable community to resume weapon development and
testing in case the Comprehensive Test Ban Treaty (CTBT) era
ends. I have been involved in reviewing and assessing the
program for the Department of Energy, primarily as a member
of the JASON group of consultants to the government. Some
of our reports are available as unclassified documents on
the Web:(4)
I believe the relevant question is whether nuclear weapons
can be maintained as safe and reliable as they were before
the CTBT was signed in 1996 (and even before a
congressionally mandated moratorium on nuclear testing took
effect in 1992). Although much of the emphasis, publicity,
and budget go to the advanced facilities and capabilities
involved in stockpile stewardship, such as ASCI (the
Accelerated Strategic Computation Initiative, at the three
weapon laboratories); DAHRT (the Dual-Axis Hydrodynamic
Radiographic facility, at Los Alamos); and NIF (the National
Ignition Facility, at Livermore); the core of the program is
really an enhanced surveillance and in the capability to
remanufacture the weapons in the stockpile, as needed.
WHAT IS REQUIRED TO KEEP NUCLEAR WEAPONS SAFE AND RELIABLE?
Under the Enhanced Surveillance Program, 11 copies of each
of the ten types of nuclear weapons in the enduring
stockpile are removed each year and radiographed and
otherwise inspected. One of each is totally disassembled,
and the pit and secondary cut open for inspection. Samples
of the high explosive are removed and tested, and the
plutonium pit is carefully inspected. Items of concern are
noted and eventually may be categorized as "actionable",
leading to some kind of remedy on the entire stockpile of
such weapons. Problems might be as simple as a loose screw,
excessive friction on a mechanical component such as the
use-control system, or incipient corrosion on some metal
part. Plutonium and uranium both react with hydrogen and
with water.
Plutonium pits were formerly manufactured at Rocky Flats,
CO, now closed. A replacement small-scale pit manufacturing
facility is being qualified at TA-55 (Los Alamos) which
might produce on the order of 30-50 pits per year. The
United States has experience with pits on the order of 30
years old, which show no signs of deterioration, and the
general feeling from recent inspections and from aging
experiments done with higher Pu-238 content (78 years
half-life vs. 24,000 for Pu-239) is that current pits are
not expected to deteriorate for 60-90 years or more. Note
that the lack of a pit production facility is not a CTBT
issue; it would be the same problem whether or not one could
use nuclear explosion testing.
Without being excessively tedious, I note the conclusion of
the JASON studies (and laboratory assessments) that a
secondary charge, if driven with the design radiation energy
in the radiation case, will provide the overall energy
output.
Given the ability to test-fire detonators and to inspect
them carefully in the disassembly process, and also to fire
samples of the high explosive-- and even one per year of the
explosive in an actual weapon-- the major uncertainty in the
performance of a nuclear weapon comes in the details of the
boosting process. This is affected by how much plutonium is
mixed with the d-t gas by the implosion itself, and that is,
in turn, affected by the surface finish and corrosion (if
any) of the plutonium surrounding the d-t gas.
To address this question, so-called subcritical experiments
are being performed at the Nevada Test Site, specifically to
monitor the ejecta produced when a plutonium plate or
partial shell is subject to high explosive shock on the
other side.
What is actually being done in the Science-Based Stockpile
Stewardship Program (SBSSP)? In addition to the normal
"custodianship" sketched above, the SSP adopted by the
Department of Energy and its laboratories is "science
based." In part, this adds assurance that the impact of
defects or artifacts discovered during the surveillance
process will be fully understood. The experimental and
analytical capabilities, in principle, could be used to
assess defects of substantial magnitude, and might show that
remedial action was not needed until they had grown further.
By delaying replacement or refurbishment, there is in
principle here the possibility of saving money. But,
especially with small numbers of weapons in the stockpile,
the acquisition of these capabilities might be more costly
than early or routine replacement of the warheads.
Specifically, DAHRT extends the conventional pulsed x-ray
facility used for dynamic imaging of actual implosion
systems (with the fissionable material replaced by a
simulant-- for instance, depleted uranium). The first axis
is operating in Los Alamos, and the second will be available
in about a year-- imaging at right angles, and with the
possibility of having four pulses spread over a two
microsecond interval. These flash x-ray images were
initially captured on film, but are now caught on an array
of scintillation crystals-- electronic imaging. DAHRT is an
excellent tool of exploring new weapon designs that have
been developed by analysis and computation, as well as the
assessment of the influence of flaws that may be discovered
in the surveillance program.
No nuclear weapon in the stockpile was designed with
computational capabilities exceeding the 500 MHz processor
now common in your desktop PC. Yet the ASCI program has
already resulted in systems at each laboratory operating at
the three tera-operation per second (TOPS) level. Such
facilities are used in the SBSSP to try to understand the
fundamentals of detonation of high explosives, and the
details of hydrodynamics, neutron propagation, and the like.
The need for such enormous computing capabilities is often
stated to derive from the three-dimensional (3-D) nature of
a flawed or aged nuclear weapon, in contrast with the
strictly 2-D modeling that suffices for a two-stage
radiation implosion-- which, after all, has an axis of
symmetry. In 2-D, the dynamics can be analyzed by assigning
every volume element a variable radius and axial position
(R,Z), and thus modeling the system as elementary rings of
material. Of course, at every R,Z there are material
properties-- identity, density, temperature(s), velocities,
etc. Designs were done on such a 2-D basis, even though
turbulence, or instabilities at interfaces have no reason to
respect the 2-D assumption.
Furthermore, in a flawed system, there is no reason to
believe that a flaw will automatically have azimuthal
symmetry-- a nick or gap in the R-Z plane, for instance,
must be modeled (at least locally) in 3-D.
Straightforwardly, if a 1-D spherical implosion would be
modeled now with 1000 radial points, and a 2-D system with
1000 x 1000, then a 3-D model would need an extra factor
1000-- a billion points.
The National Ignition Facility at Livermore (NIF) has been
in the news recently, with technical problems and budget and
schedule difficulties. It consists of 192 laser beam lines,
intended to focus more than 1 MJ of short-pulse laser light
into a mm-size gold cylindrical radiation case (or Hohlraum)
where they will equilibrate to produce soft x-rays with a
black body radiation temperature on the order of 300 eV.
Considerably lower in temperature than the black body
radiation available from a nuclear primary, and hence much
lower in energy density (remember T4!), NIF will nonetheless
allow validation on the sub-mm scale of radiation flow and
implosion calculations relevant to secondaries.
At Sandia Albuquerque, the "Z-pinch" has produced more than
1 MJ of thermal x-rays, filling a larger Hohlraum than NIF
with a black body temperature of more than 200 eV. Such
results are obtained by imploding a cylindrical shell of
hundreds of fine wires by the self-magnetic field of
20 megamp currents. Interesting results have been obtained
not only on radiation flow but also by the use of magnetic
pressure to drive flyer plates to produce shocks in
materials of relevance to nuclear weapons. In other
experiments on the Z machine, the magnetic pressure is used
to provide isentropic compression of materials, such as
deuterium, to explore their equation of state.
THE CTBT AND THE STEWARDSHIP PROGRAM. My judgment is evident
from the comments above-- the enhanced surveillance program
and a remanufacturing capability, and people to carry it
out, will serve to maintain the primary and secondary of
nuclear weapons in good shape for many decades or even
centuries. Atoms do not age, and a weapon rebuilt in the
year 2100 (using perhaps rather archaic processes) will be
just as good as when it was manufactured in 1985. Recall the
1979 letter from the director of Los Alamos.
The facilities that get all the attention, in my view, serve
primarily to attract, challenge, and validate the people who
will be involved in the surveillance and remanufacturing
efforts. But it would be totally incorrect to equate the
ability to obtain "ignition" of d-t gas or ice in NIF with
confidence in the U.S. stockpile of nuclear weapons. And it
would be equally wrong to assess a failure to achieve
ignition (for whatever reason) as impugning in any way the
capability of our stockpiled weapons.
In fact, new-design weapons entering the stockpile have
sometimes had an infant mortality problem, so that existing
and remanufactured weapons are likely to be more reliable,
and we should have more confidence in them than if we were
replacing them by weapons of new design.
The important point is that nuclear explosive tests are of
little use in maintaining a stockpile of weapons of existing
type.
CONCLUSION.
The inability to test with nuclear explosion yield does not
inhibit the United States from maintaining its stockpile of
existing nuclear weapons safe and reliable. And the same is
true for the other four nuclear states with fully developed
and tested nuclear weapons in their stockpiles-- Russia,
Britain, France, and China.
----------------
1 In a paper for the American Geophysical Union meeting
May 31, 2000, (at http://www.fas.org/rlg), I discuss the
other two topics as well.
2 I contributed to the design of the first radiation
implosion, the MIKE test of November 1, 1952, which
yielded almost 11 megatons (MT) from a secondary charge
of liquid deuterium surrounded by uranium.
3 Harold M. Agnew to J.K. Bratton, February 13, 1979.
4 At http://www.fas.org/rlg (search for "JASON" or "JSR").