You are on page 1of 10

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/234987744

A re-examination of the Buneman–Hartree condition in a cylindrical smooth-


bore relativistic magnetron

Article in Physics of Plasmas · March 2010


DOI: 10.1063/1.3328804

CITATIONS READS

45 4,531

7 authors, including:

Yue-Ying Lau John Luginsland


University of Michigan Confluent Sciences, LLC
772 PUBLICATIONS 10,612 CITATIONS 245 PUBLICATIONS 2,769 CITATIONS

SEE PROFILE SEE PROFILE

Keith Cartwright David Henry Simon


Sandia National Laboratories University of Michigan
124 PUBLICATIONS 802 CITATIONS 46 PUBLICATIONS 276 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by R.M. Gilgenbach on 20 May 2014.

The user has requested enhancement of the downloaded file.


PHYSICS OF PLASMAS 17, 033102 共2010兲

A re-examination of the Buneman–Hartree condition in a cylindrical


smooth-bore relativistic magnetron
Y. Y. Lau,1,a兲 J. W. Luginsland,2 K. L. Cartwright,3 D. H. Simon,1 W. Tang,1,3 B. W. Hoff,1,3
and R. M. Gilgenbach1
1
Department of Nuclear Engineering and Radiological Sciences, University of Michigan, Ann Arbor,
Michigan 48109-2104, USA
2
Air Force Office of Scientific Research, Arlington, Virginia 22203, USA
3
Air Force Research Laboratory, Kirtland AFB, New Mexico 87117, USA
共Received 26 August 2009; accepted 1 February 2010; published online 8 March 2010兲
The Buneman–Hartree condition is re-examined in a cylindrical, smooth-bore, relativistic
magnetron using both the conventional, single particle model, and the Brillouin flow model. These
two models yield the same result for the Buneman–Hartree condition only in the limit of a planar
magnetron. When b / a = 1.3, where a is the cathode radius and b 共⬎a兲 is the anode radius, the
difference in the two models becomes significant. When b / a = 4 the difference is acute, the
Buneman–Hartree magnetic field at a given voltage in the Brillouin flow model exceeds four times
that in the single particle model. Such a difference is always present, whether the voltage is
relativistic or not. These results are quantified for b / a Ⰷ 1 using Davidson’s model, conveniently
cast in terms of the normalized gap voltage and normalized magnetic flux imposed on the cylindrical
magnetron. A comparison with the University of Michigan/L-3 relativistic magnetron experiment is
given. © 2010 American Institute of Physics. 关doi:10.1063/1.3328804兴

I. INTRODUCTION For the BH condition, Eq. 共3兲, Davidson et al.11 showed


that it is identical for the single particle model and for the
The Buneman–Hartree 共BH兲 condition, together with the Brillouin flow model in the planar geometry whether the gap
Hull cutoff condition, govern the operation of crossed-field voltage is relativistic or nonrelativistic. Surprisingly, much
devices.1–9 The BH condition is a synchronous condition re- less is known for the circular geometry. This paper focuses
quiring that an electron is trapped in the moving frame of the on this issue. Equally surprising is that a most useful form of
wave when it grazes the anode.1 The Hull cutoff condition the solution on Brillouin flow in a cylindrical, relativistic
gives the minimum magnetic field that prevents an electron smooth-bore magnetron was published only in a conference
from reaching the anode. While the BH condition and the proceeding by Davidson et al.11
Hull cutoff condition, given respectively by Eqs. 共3兲 and 共4兲 In this paper, we use Davidson’s solution to compare the
below, may be derived from single particle orbit in crossed
BH condition in cylindrical geometry between the single par-
electric and magnetic fields in vacuum, a close examination
ticle and Brillouin flow description. We found that the devia-
of their derivation1,8 shows that Eqs. 共3兲 and 共4兲 remain valid
tion between the two is already quite significant, whether the
if this orbital theory includes an arbitrary but axisymmetric
voltage is relativistic or not, when b / a ⬎ 1.3. This difference
distribution of space charge within the crossed-field gap.10
is acute when b / a = 4, and we have quantified analytically
The Hull cutoff condition and the BH condition have
also been studied using the Brillouin flow model.5,6,11–13 The this difference for b / a Ⰷ 1. These findings are quite unex-
Brillouin flow is a laminar shear flow solely in the E ⫻ B pected, considering oven magnetrons and conventional high
direction. It extends from the cathode to a distance h from power magnetrons often have an aspect ratio b / a ⬃ 2, and
the cathode and this region is known as the Brillouin hub. some relativistic magnetrons have b / a ⬃ 5 共see Ref. 7兲.
Unlike the single particle model, the effect of space charge in Our motivations to re-examine the old topic of BH con-
Brillouin flow is extremely important. The Hull cutoff con- dition are the following. First, the operating voltage in rela-
dition then determines the minimum magnetic field to pre- tivistic magnetrons is time varying, leading to mode hopping
vent the Brillouin hub from reaching the anode at a given during the finite pulse length.2–4,7 This mode hopping is com-
voltage. The BH condition is taken as the synchronous con- pounded by beam loading, caused by the intense space
dition between the wave’s phase velocity and the Brillouin charge14–16 and by the plasmas17–20 that are produced by
flow velocity, both evaluated at the outer edge of the Bril- electron bombardment on the anode or ion bombardment on
louin hub.5,6,11,13 It is known that the Hull cutoff condition, the cathode after being accelerated across the crossed-field
Eq. 共4兲, is identical in the single particle model and in the gap. Mode identification demands a knowledge of the oper-
Brillouin flow model, whether the geometry is circular or ating condition according to BH. When we surveyed the lit-
planar, and whether the gap voltage is relativistic or erature, we were surprised that this issue is not addressed at
nonrelativistic.5 all from the view point of cylindrical Brillouin flow which is
well known to be the dominant state long before spoke
a兲
Electronic mail: yylau@umich.edu. formation.1,21–26 While Davidson et al.11 provided a complete

1070-664X/2010/17共3兲/033102/9/$30.00 17, 033102-1 © 2010 American Institute of Physics


033102-2 Lau et al. Phys. Plasmas 17, 033102 共2010兲

solution to the cylindrical relativistic Brillouin flow, they


only considered the planar limit for the BH condition. Mak-
ing use of Davidson’s original formulation, we discover and
quantify the remarkable difference in the BH condition ac-
cording to the single particle cycloidal orbit and the Brillouin
flow picture, as mentioned in the preceding paragraph. Such
differences were noted in various particle simulations in the
past, performed by us and by others.27 Here we provide a
comprehensive, general, analytic formulation.
The tremendous differences in the BH conditions are
reminiscent of the old controversy on the adequacy of the
multistream model 共where electrons execute cycloidal orbits兲
versus single stream model 共where electrons form a laminar
shear flow兲 in the description of the magnetron
operation.6,21,23,28 This controversy apparently has been long
settled in favor of the Brillouin flow model.22,23,25,26 It ap-
pears that the issue of single particle model versus Brillouin
flow model resurfaces with the present re-examination of the
BH condition for cylindrical magnetrons. The resolution of
this fundamental issue will have implications on the various
priming or locking techniques on magnetrons,7 including
magnetic priming,29,30 cathode priming,31–33 anode
priming,34 rf priming,1–3,7,35 and peer-to-peer locking.36,37
FIG. 1. A cylindrical, smooth-bore magnetron with cathode radius a, anode
The adequacy of the cycloidal orbit model38 versus the Bril- radius b 共⬎a兲, and Brillouin hub between r = a and r = rb.
louin flow model12 has also been raised in the literature on
intense ion beam diodes.
This paper is organized as follows. Section II introduces
all linear dimensions by b, where m is the electron rest mass,
the natural, nondimensional parameters that are applicable
e is the magnitude of electron charge, and Dⴱ = 共b2 − a2兲 / 2b is
for arbitrary gap voltage, magnetic field, and the geometry.
the equivalent gap width. These normalized quantities are
In Sec. III, we summarize Davidson’s important results on
denoted with a bar
cylindrical relativistic Brillouin flow. In Sec. IV, we compare
the BH conditions obtained from single particle and from the V̄ = eV/mc2, B̄ = eBDⴱ/mc, a = a/b, rb = rb/b. 共2兲
Brillouin flow and show the significant difference between
the two when b / a ⬎ 1.3. Such a difference is quantified ana- In the planar limit, b → ⬁, a → ⬁, and a / b → 1 with D = b − a
lytically for b / a Ⰷ 1, borrowing Davidson’s formulation. A being finite. Then Dⴱ reduces to the planar gap width D.
comparison with the experimental data from the University Moreover, in this planar limit, n → ⬁ in such a way that both
of Michigan/L-3 relativistic magnetron will also be given in the wavenumber k = n / b and vph = ␻ / k remain finite 关see Eq.
this section. Section V contains some concluding remarks. 共1兲兴. There is also the implicit assumption, almost universally
adopted in the literature on magnetrons1–4,6–9,22,24 that the
slow wave structure at the anode has no effect on the particle
II. BUNEMAN–HARTREE CONDITION ACCORDING orbit and on the Brillouin flow, and the inner radius of the
TO SINGLE PARTICLE ORBIT
vane tip is taken to be the anode radius 共b兲 of the equivalent
Consider a cylindrical smooth-bore magnetron of cath- smooth bore magnetron 共Fig. 1兲. It is these slow wave struc-
ode radius a and anode radius b 共⬎a兲, as shown in Fig. 1. If tures that allow ␤ ⬍ 1 in Eq. 共1兲.
there is an angular wave with dependence ej␻t-jn␪ in this A few remarks on the voltage scale and magnetic field
smooth-bore magnetron, the angular velocity of the wave scale are in order. Just from the velocity scale c and the
pattern is ␻ / n 共in rad/s兲 where ␻ is the frequency of the length scale b, all other physical scales may be constructed
wave and n is the azimuthal mode number representing the from these two. For example, the energy scale is mc2, and
number of electron spokes. The linear velocity of this wave the voltage scale is Vs = mc2 / e. Thus V̄ in Eq. 共2兲 is simply
in the azimuthal direction is vph = 共␻ / n兲b at the anode r = b. the gap voltage V divided by Vs. Moreover, the frequency
We normalize this phase velocity at the anode with respect to scale is c/b, leading to a time scale Ts = 2␲ / 共c / b兲. Since the
c, the light speed in vacuo unit of magnetic flux is volt-second, the magnetic flux scale
is naturally ⌽s = VsTs = 2␲bmc/ e. One may then easily verify
vph ␻b
␤⬅ = , 共1兲 that Eq. 共2兲 is simply B̄ = ⌽ / ⌽s, the total magnetic flux ⌽
c nc
= B共␲b2 − ␲a2兲 within the gap divided by ⌽s. Since the gap
so that the constant ␤ is assumed known once the mode voltage and the magnetic flux are externally imposed quan-
under consideration is specified. We find it convenient to tities, independent of the space charge or space current
normalize the dc voltage V by mc2 / e, vacuum dc magnetic within the gap, the scaled parameters in Eqs. 共1兲 and 共2兲 are
field B by mc/ eDⴱ in Systeme International 共SI兲 units, and the natural parameters for both the single particle description
033102-3 A re-examination of the Buneman-Hartree condition… Phys. Plasmas 17, 033102 共2010兲

electron from reaching the anode for a given gap voltage V.


In the dimensionless variables, Eq. 共2兲, the Hull cutoff con-
dition reads

B̄ = 2V̄ + V̄2 共Hull cutoff兲. 共4兲
Equation 共4兲 is relativistically correct and is applicable to
both cylindrical and planar diode.5 Once more Eq. 共4兲 is
valid even if there is an arbitrary space charge distribution
that is axisymmetric within the gap.10
It is straightforward to show that in the V̄ − B̄ plane the
BH straight line 关Eq. 共3兲兴 is always tangent to the Hull cutoff
curve 关Eq. 共4兲兴, as shown in Fig. 2. The point of tangency is
readily shown to be 共B̄ , V̄兲 = 共␤␥ , ␥ − 1兲, where ␥ = 共1
− ␤2兲−1/2 is the relativistic mass factor constructed from the
normalized phase velocity of the mode ␤ = vph / c 关Eq. 共1兲兴. In
Appendix A, we provide a simple derivation of this point of
FIG. 2. The universal normalized BH and Hull cutoff condition according to tangency, using only conservation of energy and conserva-
single particle orbit theory. These curves are valid for both relativistic and tion of canonical angular momentum.
nonrelativistic voltages, for both circular and planar geometries and for ar- The planar limits of the BH and Hull cutoff conditions,
bitrary normalized phase velocity ␤ = vph / c at the anode.
Eqs. 共3兲 and 共4兲, are obtained by simply setting Dⴱ = D, the
planar gap separation, as discussed in the sentences follow-
ing Eq. 共2兲. We shall demonstrate toward the end of Sec. IV
and the Brillouin flow description, whether the gap voltage is the established fact11 that Eqs. 共3兲 and 共4兲 are also valid for
relativistic or not. the Brillouin flow model in this planar limit, whether the
The BH condition was most conveniently derived from diode voltage is relativistic or not.
orbital consideration of a single electron exposed to the dc The nonrelativistic limits of the BH and Hull cutoff con-
voltage V and dc vacuum magnetic field B. This electron is ditions, Eqs. 共3兲 and 共4兲, may be obtained by taking the limit
emitted at the cathode with zero initial velocity. An angular c → ⬁, in which case ␤ → 0 and V̄ → 0 from Eqs. 共1兲 and 共2兲.
wave with linear phase velocity vph at the anode is then im- Using the binomial expansion, 共1 − ␤2兲1/2 ⬃ 1 − ␤2 / 2, in Eq.
posed. This angular wave is assumed to have an infinitesi- 共3兲, we obtain the nonrelativistic BH condition in a cylindri-
mally small amplitude so that it will create a torque that cal smooth-bore magnetron
brings the electron to the anode without altering the electron
energy; but the torque alters the angular momentum of the V̄ = B̄␤ − ␤2/2 共BH, nonrelativistic兲. 共5兲
electron so that this electron reaches the anode despite the
Upon ignoring the V̄2 term in Eq. 共4兲 we obtain the nonrel-
fact that the gap originally satisfies the Hull cutoff condition,
ativistic Hull cutoff condition in a cylindrical smooth-bore
i.e., that the gap is originally magnetically insulated. The BH
magnetron
condition further assumes that this electron reaches the an-
ode with negligible radial velocity, and that its linear velocity 冑
B̄ = 2V̄ 共Hull cutoff, nonrelativistic兲. 共6兲
in the azimuthal direction matches the wave’s phase velocity
vph at the anode. The latter assumption is essential to the BH In dimensional forms, Eqs. 共5兲 and 共6兲 become
condition, as it reflects the physical argument that this elec-
tron must be captured by the wave 共in the moving frame of eV = eBvphDⴱ − mvph
2
/2 共BH, nonrelativistic兲, 共5⬘兲
the wave兲 if it is to strongly interact with the wave.1,21 This
condition reads,2,6–8,22,24 in the normalized quantities defined
in Eqs. 共1兲 and 共2兲
冉 冊 冑
eB ⴱ
m
D =
2eV
m
共Hull cutoff, nonrelativistic兲,

共6⬘兲
V̄ = B̄␤ − 共1 − 冑1 − ␤2兲 共BH兲. 共3兲 which are independent of c. Equation 共6⬘兲 shows that V is
proportional to B2 and is thus represented by a parabola in
Equation 共3兲 is represented by a straight line of slope ␤ in the the B-V plane. The BH line 关Eq. 共5⬘兲兴 is also tangent to this
V̄ − B̄ plane 共Fig. 2兲. It is relativistically correct8 and is appli- parabola 关Eq. 共6⬘兲兴 in the nonrelativistic limit. Toward the
cable to a cylindrical or planar magnetron.2,6–8,22,24 It is also end of Appendix A, we present a simple derivation of the BH
valid even if the crossed-field gap contains an arbitrary axi- condition, Eq. 共5⬘兲, in the nonrelativistic, planar limit. There,
symmetric space charge distribution.10 we illustrate that, except at the point of tangency, canonical
The BH condition is usually plotted along with the Hull angular momentum is not conserved on the BH curve ac-
cutoff condition 共Fig. 2兲. From single particle orbit in a cording to the single particle theory. The derivation there
vacuum crossed-field diode, the Hull cutoff condition gives makes it clear that Eq. 共5⬘兲 is valid even if there is a stratified
the minimum magnetic field that is required to prevent an space charge distribution in the gap.10
033102-4 Lau et al. Phys. Plasmas 17, 033102 共2010兲

III. CYLINDRICAL BRILLOUIN FLOW IN RELATIVISTIC Once r̄b, ¯␬, and ¯␹ have been determined, the complete
SMOOTH-BORE MAGNETRONS Brillouin flow profile may be deduced. Specifically, the nor-
malized Brillouin flow velocity profile ␤b共r兲 = vb共r兲 / c may be
Under the same title of this section, Davidson et al.11
obtained from
provided a comprehensive study of the relativistic cylindrical
Brillouin flow. Here, we will quote their main results and ␤b共r兲 = tanh ␹共r兲, 共10兲
cast them in the dimensionless parameters, Eq. 共2兲. For the where ␹共r兲 can be numerically integrated from r = a to r = ra
geometry shown in Fig. 1, Davidson et al. formulated the 关see Eq. 共9兲兴
Brilluoin flow by assuming that the dc voltage V and the
共initial兲 vacuum magnetic field B are known. The initial d␹ 冑 2
r = ¯␬ + sinh2 ␹, ␹共a兲 = 0. 共11兲
magnetic flux ⌽ = B␲共b2 − a2兲 is assumed to be preserved dr
even after the Brillouin flow is set up. That is, even if the
Brillouin flow produces significant self-magnetic field, the From ␤b共r兲, one may construct the relativistic mass factor,
total magnetic flux within the smooth-bore magnetron is still ␥b共r兲 = 关1 − ␤2b共r兲兴−1/2 = cosh ␹共r兲. The electrostatic potential
given by this initial ⌽ through a redistribution of magnetic profile ␾共r兲 within the Brillouin hub is then given by
flux. The Brilluoin flow is a cold, laminar shear flow only in − e␾共r兲 + 关␥b共r兲 − 1兴mc2 = 0. 共12兲
the ␪-direction 共Fig. 1兲; its flow velocity is 0 at r = a, at which
the electric field also equals 0. It extends from r = a to r = rb, a From ␾共r兲, the electron density profile within the Brillouin
region known as the Brillouin hub. The electron density pro- hub follows from the Poisson equation.
file nb共r兲 and the flow velocity vb共r兲 within the Brillouin hub Given a dc voltage V, we may next inquire what is the
need to be determined. The magnetic field B is assumed to be value of B so that rb = b, i.e., the Brillouin hub just fills the
sufficiently large so that the Brillouin hub radius rb is less entire gap. It turns out that this critical value of B is identical
than b 共Fig. 1兲. The radial electric field and the axial mag- to the Hull cutoff condition, Eq. 共4兲, that is derived from
netic field of the Brillouin flow are solved self-consistently single particle orbit. To verify this well-known result, con-
with the flow’s space charge density −enb共r兲 and with the sider this critical case r̄b = 1. Equation 共7兲 gives cosh ¯␹ = 1
flow’s azimuthal current density −enb共r兲vb共r兲. For every elec- + V̄. Thus, sinh ¯␹ = 共cosh2 ¯␹ − 1兲1/2 = 共2V̄ + V̄2兲1/2. Substitution
tron, there is a radial force balance between the electric of this expression into Eq. 共8兲 immediately yields Eq. 共4兲 in
force, the Lorentz force, and the relativistic centrifugal force. the limit r̄b = 1.
共This centrifugal force vanishes in the planar limit, in which
case the flow velocity is simply the E ⫻ B drift that accounts
IV. SYNCHRONISM BETWEEN THE PHASE
for all self electric and magnetic field effects.兲 Also at each
VELOCITY AND ELECTRON FLOW VELOCITY
radius within the Brillouin hub, the electron’s total energy is AT THE OUTER EDGE OF THE BRILLOUIN HUB
equal to zero 关see Eq. 共12兲 below兴.
Once a, b, V, and B are specified, the dimensionless pa- Davidson’s complete solution of the Brillouin flow11 en-
rameters ā, V̄, and B̄ are known by Eq. 共2兲. The complete ables us to answer the following questions. Given the phase
Brillouin flow solution may be obtained by first solving for velocity vph at the anode 关see Eq. 共1兲兴, what should be the
the three parameters r̄b, ¯␬, and ¯␹ from the following three relation between V and B so that the phase velocity at the
equations, originally given as Eqs. 共20兲, 共21兲, and 共17兲 in outer edge of the Brillouin hub vphrb / b is equal to the Bril-
Davidson et al.,11 respectively louin flow velocity there? How is this relation between V̄ and
B̄, in nondimensional form, compared with the BH condition,
V̄ = cosh ¯␹ − 1 + 共sinh ¯␹兲冑¯␬2 + sinh2 ¯␹ ⫻ ln共1/r̄b兲, 共7兲 Eq. 共3兲, that is obtained from single particle orbit consider-
ation?
In terms of the normalized parameters, Eqs. 共1兲 and 共2兲,
关共1 + r̄2b兲sinh ¯␹ + 共1 − r̄2b兲共cosh ¯␹兲冑¯␬2 + sinh2 ¯␹兴,
1
B̄ = the normalized phase velocity of the wave at r = rb is
2r̄b
共vphrb / b兲 / c = ␤rb. We next recognize that the value of ␹ at
共8兲 r = rb is simply ¯␹, by comparing Eq. 共11兲 with its solution,
Eq. 共9兲. Thus, from Eq. 共10兲, we have ␤b共rb兲 = tanh ¯␹ which
ln 冉 冊 冕冑
r̄b

=
¯␹

0
d␹
␬ + sinh2 ␹
¯2
. 共9兲
is the normalized Brillouin flow velocity at r = rb. The condi-
tion for synchronism between the phase velocity and electron
flow velocity at r = rb then reads
Equation 共7兲 关Eq. 共8兲兴 gives the total voltage 共magnetic flux兲 ␤r̄b = tanh ¯␹ . 共13兲
in the gap as the sum of voltage 共magnetic flux兲 within the
Brillouin hub and outside the Brillouin hub. Equation 共9兲 is This synchronism condition, at fixed values of b / a and ␤, is
the solution to the Poisson equation within the Brillouin hub, represented in the V̄ − B̄ plane in Figs. 3–6, and is designated
i.e., it is the explicit solution to Eq. 共19兲 of Davidson11 where as the BH condition according to the Brillouin fluid flow
¯␬ = ␬a ⬅ ␻ pa / c and ␻ p is the nonrelativistic electron plasma model 共the dash-dotted curves in Figs. 3–6兲. It is solved us-
frequency of the Brillouin flow evaluated at r = a. 关An algo- ing the algorithm given in Appendix B.
rithm yielding simultaneous solutions to Eqs. 共7兲–共9兲 and In Figs. 3–6, we progressively increase the geometrical
共13兲 below, is given in Appendix B.兴 aspect ratio, 1 / ā = b / a = 1.1, 1.3, 2, and 4. In each of these
033102-5 A re-examination of the Buneman-Hartree condition… Phys. Plasmas 17, 033102 共2010兲

FIG. 4. 共Color online兲 共Top兲 The normalized Hull cutoff condition 共black
FIG. 3. 共Color online兲 共Top兲 The normalized Hull cutoff condition 共black curve兲, the normalized BH condition according to single particle orbit theory
curve兲, the normalized BH condition according to single particle orbit theory 共color, solid curves兲, and the normalized BH condition according to the
共color, solid curves兲, and the normalized BH condition according to the Brillouin flow model 共color, dash-dotted curves兲 for b / a = 1.3. 共Bottom兲 En-
Brillouin flow model 共color, dash-dotted curves兲, for b / a = 1.1. 共Bottom兲 larged plot near the point of tangency.
Enlarged plot near the point of tangency.

figures, we display the relation between B̄ and V̄ for three


values of ␤ : ␤ = 0.1, 0.2, and 0.3. The range of voltage V increases markedly. Figure 6 shows that for b / a = 4, this dif-
ranges from 0 共V̄ = 0兲 to about 1 MV 共V̄ = 2兲. Thus, a fairly ference is huge. The BH magnetic field, at a given voltage, in
inclusive range of operating voltage and geometry is covered the Brillouin flow model exceeds four times that in the single
in Figs. 3–6. For comparison, the BH condition, Eq. 共3兲, and particle model. A close-up view of Fig. 6共b兲 shows that the
the Hull cutoff condition, Eq. 共4兲, are also displayed in BH curve according to the fluid model remains tangent to the
Figs. 3–6. Hull cutoff curve at the point of tangency, even though it
For b / a = 1.1 共Fig. 3兲, the circular magnetron is almost does not appear to be so in Fig. 6共b兲. Below, we consider the
planar. The BH condition according to the fluid model 共dash- decidedly cylindrical limit b / a Ⰷ 1 and the planar limit b/a
dotted line兲 deviates from the BH condition according to approaching 1.
orbit theory 共solid line兲 by about 10%. That is, given a value To quantify the large deviations for 1 / ā = b / a Ⰷ 1, let us
of V̄, the value of B̄ according to the fluid BH model is larger examine the limit B̄ Ⰷ 1 so that the Brillouin hub is expected
than the single particle BH model by about 10%. This is true to be very close to the cathode, i.e., r̄b → ā, 共ā Ⰶ 1兲. In this
for relativistic voltage 关V̄ = 2, Fig. 3共a兲兴 and for nonrelativis- limit, Eqs. 共13兲, 共9兲, 共7兲, and 共8兲 simplify to read, respec-
tively
tic voltage 关V̄ Ⰶ 1, Fig. 3共b兲兴. This is also true regardless of
the value of ␤ chosen, i.e., regardless of the mode. As b/a ¯␹ ⬵ ␤r̄b ⬵ ␤ā Ⰶ 1, 共14兲
progressively increases from Fig. 4–6, the departure from
planar geometry increases accordingly. The difference in the
BH condition according to fluid theory and single particle ln共r̄b/ā兲 ⬵ ¯␹/¯␬ ⬵ ␤ā/¯␬ Ⰶ 1, 共15兲
033102-6 Lau et al. Phys. Plasmas 17, 033102 共2010兲

FIG. 6. 共Color online兲 共Top兲 The normalized Hull cutoff condition 共black
curve兲, the normalized BH condition according to single particle orbit theory
FIG. 5. 共Color online兲 共Top兲 The normalized Hull cutoff condition 共black
共color, solid curves兲, and the normalized BH condition according to the
curve兲, the normalized BH condition according to single particle orbit theory
Brillouin flow model 共color, dash-dotted curves兲, for b / a = 4. 共Bottom兲 En-
共color, solid curves兲, and the normalized BH condition according to the
larged plot near the point of tangency.
Brillouin flow model 共color, dash-dotted curves兲, for b / a = 2. 共Bottom兲 En-
larged plot near the point of tangency.

¯␹2
V̄ ⬵ + ¯␹¯␬ln共1/ā兲 ⬵ ¯␹¯␬ln共1/ā兲 ⬵ ␤ā¯␬ln共1/ā兲, 共16兲
2

1 ¯␬ ¯␬
B̄ ⬵ 共¯␹ + ¯␬兲 ⬵ ⬵ . 共17兲
2r̄b 2r̄b 2ā
Division between Eqs. 共16兲 and 共17兲 yields

V̄ ⬵ B̄ ⫻ ␤ ⫻ 关2ā2ln共1/ā兲兴, ā Ⰶ 1. 共18兲
Equation 共18兲 is shown by the solid curves in Fig. 7 for the
case b / a = 4. It is in qualitative agreement with the BH con-
dition according to the fluid theory, shown by the dash-dotted
curves in Fig. 7 关and also in Fig. 6共a兲兴. In comparison, at
large values of B̄, the BH condition according to single par-
ticle orbit reads

V̄ ⬵ B̄ ⫻ ␤ , 共19兲
FIG. 7. 共Color online兲 The normalized BH condition according to the Bril-
from Eq. 共3兲. Equation 共18兲 quantifies the very significant louin flow model 共dash-dotted curves兲 and to the asymptotic formula Eq.
deviation in the BH condition using the fluid model, as com- 共18兲 共solid curves兲. Here, b / a = 4.
033102-7 A re-examination of the Buneman-Hartree condition… Phys. Plasmas 17, 033102 共2010兲

experimental data show that the single particle orbit BH con-


dition cannot adequately explain the operation of the Univer-
sity of Michigan relativistic magnetron, while the Brillouin
flow BH condition is consistent with the experiments.

V. CONCLUDING REMARKS
This paper provides a critical re-examination of the Hull
cutoff condition and BH condition according to the single
particle orbit model and Brillouin flow model. It employs the
natural dimensionless parameters V̄ and B̄ that are con-
structed for a smooth-bore magnetron which may either be
relativistic or nonrelativistic, planar or circular. For the Hull
cutoff voltage, we confirm that it is the same for the single
particle and Brillouin flow model. For the BH condition, we
also confirm that it is the same for the single particle and
Brillouin flow model, but only in the planar magnetron limit.
If the magnetron is cylindrical, the BH condition differs pro-
FIG. 8. 共Color online兲 The normalized Hull cutoff condition 共black curve兲, gressively as b/a increases, between the single particle and
the normalized BH condition according to single particle orbit theory 共color, Brillouin flow model. This difference is huge for b / a = 4,
solid curves兲, and the normalized BH condition according to the Brillouin regardless of whether the voltage is relativistic or not. This
flow model 共color, dash-dotted curves兲 for the ␲-mode and 2␲ / 3-mode of
the University of Michigan/L-3 relativistic magnetron. The data show the
difference is quantified by Eqs. 共18兲 and 共19兲. If magnetron
maximum and average values for normalized voltage and magnetic field operation lies between the BH curve and Hull cutoff curve in
strength recorded during magnetron operation. the V̄ − B̄ plot, then the Brillouin flow model allows a signifi-
cantly wider range of operating conditions. A comparison
with the University of Michigan/L-3 relativistic magnetron,
as given toward the end of Sec. IV, shows that this is indeed
pared to the conventional single particle model 关Eq. 共19兲兴,
the case, enabling an interpretation of a large amount of ex-
when the cathode radius is small and the anode radius is
perimental data.
large 共ā Ⰶ 1兲, regardless of the diode voltage and the phase
Since an overwhelming majority of magnetrons are cy-
velocity of the wave.
lindrical, a large fraction7 with aspect ratio b / a ⬃ 2, and
For easy reference, we shall recapitulate Davidson’s
some with b / a ⬃ 5, it is interesting that the Brillouin flow
proof11 that the BH condition is indeed the same for the fluid
description and the single particle description give such
model and for the single particle model in the planar limit. In
vastly different BH conditions. Arguably, magnetrons must
the planar limit, a → ⬁, b → ⬁, r̄b = 1 − 共b − rb兲 / b → 1, and ¯␬
start with emission of single particles that execute cycloidal
= ␻ pa / c → ⬁, while the quantities 共b − a兲, 共b − rb兲, ␤, and ¯␹
orbits at the very early stage and presumably that was the
remain finite. The leading terms of Eqs. 共7兲 and 共8兲 give
physical basis in the original derivation of synchronism by
Buneman and Hartree. However, simulations and theoretical
V̄ = cosh ¯␹ − 1 + 共sinh ¯␹兲¯␬共b − rb兲/b, 共20兲
analysis invariably show that the Brillouin flow is the pre-
ferred state,21–23,25,26 and that this state is fully developed
B̄ = sinh ¯␹ + 共cosh ¯␹兲¯␬共b − rb兲/b. 共21兲 long before the spokes are formed. Indeed, Slater repeatedly
emphasized that the electron cycloidal orbits that now bear
We next multiply Eq. 共20兲 by cosh ¯␹, Eq. 共21兲 by sinh ¯␹, and his name are not accessible, and that the Brillouin flow
subtract to obtain a relation between B̄ and V̄. This relation is would be the preferred state 共see pp. 341, 342, 350, 355, and
readily shown to be Eq. 共3兲 upon using tanh ¯␹ = ␤ which is 356 of Ref. 1兲. This would seem to settle in favor of the
Eq. 共13兲 in the planar limit. Brillouin flow model. However, one needs to be very bold to
Finally, we present in Fig. 8 the experimental data on the abandon the venerable BH condition Eq. 共3兲 which was
University of Michigan/L-3 relativistic magnetron which has based on single particle orbit consideration whenever b / a
been described elsewhere.7 This magnetron has a b/a ratio of ⬎ O共1兲.
5, ␤ = 0.2261 for the ␲-mode, and ␤ = 0.3258 for the From a description of the guiding center orbits,
2␲ / 3-mode. The maximum values for the voltage and mag- Riyopoulos39 suggested that the correct synchronism condi-
netic field values are extracted from each pulse. In addition, tion is between the phase velocity of the wave and the guid-
the values of the voltage and magnetic field are averaged ing center velocity of the electron, rather than the electron
over the pulse. Both of these values are shown in Fig. 8 velocity itself. Since Riyopoulos’ treatments employed a
along with the theoretical models for both the single particle Cartesian geometry, they cannot explain the differences in
and Brillouin flow BH condition. The data show that all the synchronism condition due to the cylindrical geometry
of the 2␲ / 3 and many of the ␲ mode pulses are beneath the according to the single particle model and the fluid model.
single particle theoretical BH curve. Since oscillations Once more, in the derivation of BH condition, Eq. 共3兲,
only occur between the Hull cutoff and the BH curves, the we only compare the end states of a single electron’s cycloi-
033102-8 Lau et al. Phys. Plasmas 17, 033102 共2010兲

dal orbit 共i.e., on the cathode and on the anode surfaces兲. It P␪ = ␥mrv␪ − erA␪ = constant, 共A1兲
demands conservation of energy, but not conservation of ca-
nonical angular momentum, as illustrated in Appendix A. On where v␪ is the electron’s azimuthal velocity and A␪ is the
the other hand, in the Brillouin flow solution, both energy azimuthal component of the vector potential A, which is re-
and canonical angular momentum are conserved everywhere lated to the magnetic field B by ⵜ ⫻ A = B. Thus, the total
within the Brillouin hub. In the Brillouin flow model, an magnetic flux, ⌽, within the magnetron is
electron next to the cathode surface is not the same electron
at the top of the Brillouin hub, while the single particle orbit
theory follows the motion of the same electron.
⌽= 冕冕 magnetron
B · ds = 冕冕magnetron
ⵜ ⫻ A · ds

Particle simulations may further help resolve the dis-


crepancy between the BH condition in the fluid and single
particle model. The new revelation given in this paper invites
= 冕 A · dᐉ = 2␲bA␪共b兲 − 2␲aA␪共a兲. 共A2兲

a re-examination of the experimental evidence on the appli- Next recall that v␪ = 0 at r = a and that v␪ = vph = ␤c at r = b, for
cability of Eq. 共3兲, especially for the typical cases with b / a the electron that satisfies both the Hull cutoff and BH condi-
⬃ 2. tion. For this electron, we evaluate Eq. 共A1兲 at r = b, and then
Finally, we should remark that in the relativistic, cylin- at r = a, and subtract. This gives, upon using Eq. 共A2兲
drical Brillouin flow solution, the BH condition 共3兲 is indeed
⌽ ⌽
satisfied at the top of the Brillouin hub, r = rb, if V̄, B̄, and ␤ = ⬅ B̄ = ␤␥ , 共A3兲
in Eq. 共3兲 are taken to be, respectively, the normalized elec- ⌽s 2␲bmc/e
trostatic potential, the normalized magnetic flux, and the nor- where we have used the magnetic flux scale ⌽s = 2␲bmc/ e
malized Brillouin flow velocity, all evaluated at r = rb. How- that is introduced in the paragraph after Eq. 共2兲 in the main
ever, the latter quantities are not measurable and are of little text. Equation 共A3兲 proves the abscissa at the point of tan-
use to experiments. Other interesting but hitherto little- gency shown in Fig. 2.
explored properties of the relativistic, cylindrical Brillouin Finally, we provide a simple derivation of the BH con-
flow solution are beyond the scope of the present paper and dition, Eq. 共5⬘兲, in the planar, nonrelativistic limit. This
will be given elsewhere. serves to illustrate some subtlety from the single particle pic-
ture. Consider energy conservation in the frame moving at
ACKNOWLEDGMENTS the phase velocity of the wave vph. This wave is assumed to
have an infinitesimally small amplitude so that it can create a
We wish to thank the referee, whose multiple reviews
共small兲 torque that brings the electron to the anode, but with-
helped improve our manuscript. This work was supported by
out altering the total energy of the electron.1 In this moving
AFOSR Grant No. FA9550-09-1-0086, Air Force Research
frame, the initial energy of electron, at the cathode, is
Laboratory, Office of Naval Research, L-3 Communications
mvph2
/ 2, and its potential energy is zero. Thus, the total initial
Electron Device Division, and Northrop Grumman Corpora-
energy of the electron in the moving frame is
tion.
Einitial = 21 mvph
2
. 共A4兲

APPENDIX A: POINT OF TANGENCY When this electron just grazes the anode, in the moving
ON THE HULL CUTOFF AND BH CURVES frame, its kinetic energy is zero because this electron is
trapped by the wave at the anode.1 This electron suffers a
The physically appealing forms of the ordinate V̄ = ␥ − 1 potential energy drop −eV+ e共vph ⫻ B兲D in the moving
and of the abscissa B̄ = ␤␥ at the point of tangency shown in frame, when it reaches the anode. Here, the term ⫺eV ac-
Fig. 2 prompted their derivation from conservation of energy counts for the anode-cathode voltage drop, whereas the last
and conservation of canonical angular momentum, respec- term accounts for the constant, additional Lorentz force
tively. We use the single particle orbit model. On this point e共vph ⫻ B兲, that the electron experiences in the moving frame
of tangency, both the Hull cutoff condition and the BH con- throughout its journey to the anode. Thus, in the moving
dition are simultaneously satisfied. Some subtlety on the BH frame, the total energy of the electron at the anode is
condition1 is revisited in the last paragraph of this Appendix,
Efinal = − eV + evphBD. 共A5兲
using a simple case for illustration.
An electron emitted from the cathode with 0 initial ve- Equating Eq. 共A4兲 with Eq. 共A5兲, we obtain the nonrelativ-
locity obeys the energy conservation law, Eq. 共12兲, where istic BH condition, Eq. 共5⬘兲, in the planar limit. Thus, on the
␥b共r兲 now stands for the relativistic mass factor of the elec- BH curve 共Fig. 2兲, the electron energy is conserved, but the
tron when it reaches a radius r. Evaluation of Eq. 共12兲 at r electron canonical angular momentum is not 共except at the
= b immediately yields V̄ = ␥ − 1, remembering that this elec- point of tangency兲. This subtle point, together with the math-
tron has only a tangential velocity equal to the phase velocity ematical derivation of the BH condition, may be found in
vph at the anode, at the point of tangency in Fig. 2. Slater.1 Note also that Eqs. 共A4兲 and 共A5兲 remain valid if
This electron also satisfies conservation of canonical an- there is a space charge distribution that depends only on x,
gular momentum P␪. In the limit of zero rf electric fields, this where x is the coordinates orthogonal to the cathode
gives surface.10
033102-9 A re-examination of the Buneman-Hartree condition… Phys. Plasmas 17, 033102 共2010兲

APPENDIX B. PROCEDURES TO SOLVE tion of the BH condition toward the end of Appendix A illustrates this fact
EQUATIONS „7…–„9… and „13… AND THE BH CONDITION which was mentioned in Refs. 1 and 8.
11
ACCORDING TO THE BRILLUOIN FLOW MODEL R. C. Davidson, G. L. Johnston, K. T. Tsang, and A. T. Drobot, Proc. SPIE
1061, 186 共1989兲.
We assume that ā and ␤ are specified. The four equa-
12
T. M. Antonsen and E. Ott, Phys. Fluids 19, 52 共1976兲.
13
E. Ott and R. V. Lovelace, Appl. Phys. Lett. 27, 378 共1975兲.
tions, 共7兲–共9兲 and 共13兲, may be solved simultaneously to 14
R. Kowalczyk, Y. Y. Lau, T. M. Antonsen, J. W. Luginsland, D. P. Cher-
yield the BH condition according to the Brillouin flow nin, C. B. Wilsen, W. Tang, and R. M. Gilgenbach, IEEE Trans. Electron
model, as follows: Devices 52, 2087 共2005兲.
15
Y. Y. Lau and D. P. Chernin, Phys. Fluids B4, 3473 共1992兲.
A. Assume a value of r̄b 共ā ⬍ r̄b ⬍ 1兲. 16
B. E. Carlsten, R. J. Faehl, M. V. Fazio, W. B. Haynes, R. D. Ryne, and R.
B. Calculate ¯␹ = tanh−1共␤r̄b兲 according to Eq. 共13兲. M. Stringfield, IEEE Trans. Plasma Sci. 22, 730 共1994兲.
Find ¯␬ so that Eq. 共9兲 is satisfied.
17
C. D. Price, J. S. Levine, and J. N. Benford, IEEE Trans. Plasma Sci. 26, 256
共1998兲.
D. Use the values of r̄b, ¯␹, and ¯␬ in Eq. 共7兲 to obtain V̄ and 18
M. D. Haworth, K. L. Cartwright, J. W. Luginsland, D. A. Shiffler, and R.
in Eq. 共8兲 to obtain B̄. J. Umstattd, IEEE Trans. Plasma Sci. 30, 992 共2002兲.
19
J. Benford, F. J. Agee, D. M. Goebel, F. Hegeler, K. J. Hendricks, R. M.
E. The value of 共V̄ , B̄兲 in Step D corresponds to the value Gilgenbach, C. Grabowski, H. Jory, and J. P. Verboncoeur, in High-Power
of r̄b assigned in Step A. Changing r̄b in Step A and Microwave Sources and Technologies, edited by R. J. Barker and E.
repeating Steps B–D then traces a curve in the V̄ − B̄ Schamiloglu 共IEEE, New York, 2001兲 Chap. 4; K. J. Hendricks, J. Ben-
plane. This curve is labeled in Figs. 3–6 as the BH ford, J. Eastwood, M. Friedman, M. Haworth, R. W. Lemke, J. Lugin-
sland, R. B. Miller, and J. Pasour, High-Power Microwave Sources and
synchronous condition according to the fluid model, for Technologies, 共IEEE, New York, 2001兲 Chap. 3; Also, F. J. Agee, IEEE
the values of ā and ␤ specified at the beginning of this Trans. Plasma Sci. 26, 235 共1998兲.
20
Appendix. Y. Y. Lau, J. W. Luginsland, K. L. Cartwright, and M. D. Haworth, Phys.
Rev. Lett. 98, 015002 共2007兲.
In Step C, there is only one real, positive value of ¯␬ that 21
O. Buneman, in Crossed-Field Microwave Devices, edited by E. Okress
satisfies Eq. 共9兲. This may be proved by differentiating the 22
共Academic, New York, 1961兲, p. 218.
RHS of Eq. 共9兲 with respect to ¯␬ under the integral sign. The A. Palevsky, Ph.D. dissertation, Massachusetts Institute of Technology,
Cambridge, MA, 1980.
result is always negative. Therefore, the RHS of Eq. 共9兲 is a 23
T. J. Orzechowski and G. Bekefi, Phys. Fluids 22, 978 共1979兲.
monotonically decreasing function of ¯␬ and there is only one 24
A. Palevsky and G. Bekefi, Phys. Fluids 22, 986 共1979兲.
25
root of ¯␬ to Eq. 共9兲. Finally, Step E gives 共B̄ , V̄兲 = 共␤␥ , ␥ P. J. Christenson, D. Chernin, A. Gardner, and Y. Y. Lau, Phys. Plasmas 3,
4455 共1996兲.
− 1兲 for r̄b = 1, the point of tangency in Fig. 2. 26
P. J. Christenson and Y. Y. Lau, Phys. Plasmas 1, 3725 共1994兲; Phys. Rev.
1
Lett. 76, 3324 共1996兲.
J. C. Slater, Microwave Electronics 共Van Nostrand, New York, 1951兲, p. 27
P. Mardahl and J. Watrous, personal communication.
302. 28
G. D. Sims, in Crossed-Field Microwave Devices, edited by E. Okress
2
J. Benford, J. A. Swegle, and E. Schamiloglu, High Power Microwaves, 共Academic, New York, 1961兲, p. 180.
2nd ed. 共Taylor & Francis, New York, 2007兲. 29
V. B. Neculaes, R. M. Gilgenbach, and Y. Y. Lau, Appl. Phys. Lett. 83,
3
R. J. Barker, N. C. Luhmann, J. H. Booske, and G. S. Nusinovich, Modern 1938 共2003兲; U.S. Patent No. 6,872,929 共29 March 2005兲; U.S. Patent No.
Microwave and Millimeter Wave Power Electronics 共IEEE, Piscataway, 6,921,890 共26 July 2005兲.
NJ, 2004兲. 30
4 V. B. Neculaes, P. Pengvanich, Y. Hidaka, Y. Y. Lau, R. M. Gilgenbach,
V. L. Granatstein and I. Alexeff, High-Power Microwave Sources 共Artech
W. M. White, M. C. Jones, H. L. Bosman, and J. W. Luginsland, IEEE
House, Norwood, MA, 1987兲.
5 Trans. Plasma Sci. 33, 654 共2005兲.
R. C. Davidson, Physics of Nonneutral Plasmas 共Addison-Wesley, Red- 31
M. C. Jones, V. B. Neculaes, Y. Y. Lau, R. M. Gilgenbach, and W. White,
wood City, CA, 1990兲.
6 Appl. Phys. Lett. 85, 6332 共2004兲.
Y. Y. Lau, in High-Power Microwave Sources, edited by V. L. Granatstein 32
and I. Alexeff 共Artech House, Norwood, MA, 1987兲, p. 309. M. C. Jones, V. B. Neculaes, Y. Y. Lau, R. M. Gilgenbach, W. M. White,
7
R. M. Gilgenbach, Y. Y. Lau, H. McDowell, K. L. Cartwright, and T. A. B. W. Hoff, and N. M. Jordan, Appl. Phys. Lett. 87, 081501 共2005兲.
33
Spencer, in Modern Microwave and Millimeter Wave Power Electronics, M. Fuks and E. Schamiloglu, Phys. Rev. Lett. 95, 205101 共2005兲.
34
edited by R. J. Barker, N. C. Luhmann, J. H. Booske, and G. S. Nusinov- J. I. Kim, J. H. Won, G. S. Park, H. J. Ha, and J. C. Shon, Appl. Phys. Lett.
ich, 共IEEE, Piscataway, NJ, 2004兲 Chap. 6. 88, 221501 共2006兲.
35
8
R. L. Walker, in Microwave Magnetrons, edited by G. B. Colloins P. Pengvanich, V. B. Neculaes, Y. Y. Lau, R. M. Gilgenbach, M. C. Jones,
共McGraw-Hill, New York, 1948兲, p. 227. W. M. White, and R. D. Kowalczyk, J. Appl. Phys. 98, 114903 共2005兲.
9 36
J. R. M. Vaughan, IEEE Trans. Electron Devices 20, 818 共1973兲. P. Pengvanich, Y. Y. Lau, E. Cruz, R. M. Gilgenbach, B. Hoff, and J. W.
10
It is often forgotten that the orbital theory results, Eqs. 共3兲 and 共4兲, are also Luginsland, Phys. Plasmas 15, 103104 共2008兲.
37
valid even when there is an axisymmetric space charge distribution in a E. J. Cruz, B. W. Hoff, P. Pengvanich, Y. Y. Lau, R. M. Gilgenbach, and J.
cylindrical crossed-field gap. Since these two equations compare the en- W. Luginsland, Appl. Phys. Lett. 95, 191503 共2009兲.
38
ergy and canonical angular momentum of an electron at the anode and at K. D. Bergeron, Appl. Phys. Lett. 28, 306 共1976兲; Phys. Fluids 20, 688
the cathode, an axisymmetric space charge distribution would not alter 共1977兲.
39
these quantities at the end points 共anode and cathode兲. The simple deriva- S. Riyopoulos, Phys. Rev. E 47, 2839 共1993兲.

View publication stats

You might also like