HAL Id: jpa-00247288
https://hal.archives-ouvertes.fr/jpa-00247288
Submitted on 1 Jan 1996
HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.
P. Chaikin
To cite this version:
P. Chaikin. Field Induced Spin Density Waves. Journal de Physique I, EDP Sciences, 1996, 6 (12),
pp.1875-7898. �10.1051/jp1:1996169�. �jpa-00247288�
J.
Phys.
l France 6(1996)
1875-1898 DECEMBER 1996, PAGE 1875Field Induced Spin Density Waves
P-M-
Chaikin (*)
Department
ofPhysics,
PrincetonUniversity, Princeton,
NJ 08544, USA andExxon Research and
Engineering Company,
Route 22 East,Annandale,
NJ o8801, USA(Received
20 June 1996, revised 20August
1996, accepted 26August 1996)
PACS.75.30.Fv
Spin-density
wavesPACS.72.15.Gd
Galvanomagnetic
and other magnetotransport elfects PACS.74.70.KnOrganic superconductors
Abstract. Trie Field Induced
Spin Density
Waves(FISDWS)
found inorganic
conductorsrepresent a unique serres of transitions which meld trie one-dimensional physics of trie Peierls
instability
with trie two-dimensionalphysics
of theQuantum
Hall Elfect. This paper presents apedagogical
introduction tu trie FISDW'S in theBechgaard softs, along
with recentexperimental
results on related
high magnetic
fieldphenomena.
1, Introduction
The
Bechgaard salts, (TMTSF)2X (where
X=
PF6, Cl04, Re04 etc.),
areprobably
trie mostinterestiug
electronic materials ever discovered.Dependiug
oucomposition, temperature,
pressure and
magnetic
fieldthey
exhibit most of trieground
states audpheuomena
associated withinteracting
electrons invastly
differentsystems.
There arecompetitions
between metallic andinsulating, magnetic
andsuperconductiug, semicouducting
and semimetallicphases.
TrieBechgaard
softs exhibit ail of the electronic transport mechanismsyet discovered,
metalhcconductivity, sliding density
waveconductivity, superconductivity
and thequantum
Hall effect.What is even more remarkable is that ail of the above
properties
cau be observed in ouesiugle crystal
of one of trieBechgaard
softs(TMTSF)2PF6
astemperature,
pressure, andmagnetic
field are variediii.
Trie basis for
understanding
trie widevariety
of behaviors is to be found in triestrongly anisotropic
bandstructureresultiug
from triequasi-one-dimensional
chainlikecrystal
structure.Platelike TMTSF molecules stock face to face lu a zigzag chaiu. Trie wavefunction
overlap
from one molecule to trie next is
responsible'for
trielarge
bandwidth(1 eV)
in trie chair a direction.Neighboring
chairs in trie b direction are alsosulficiently
close for Se orbitaloverlap
and
yield
a bandwidth of 0.1 eV. Triecoupled
chairs form two-dimensionalplanes
in which trie electrons are delocalized. Trieplanes
areseparated
in trie third directionby
a sheet of anious.Trie
overlaps
are small aud trie baudwidth is dowuby
an additional factor ofr-
30 to 0.003 eV. There is a fuit
charge
trausfer of one electron per unit cell to trieanion, leaving
trie two TMTSF molecules with half a noie ou each. Were it uot for aslight
dimerization of triezigzag
(* e-mail: chaikin@
pupgg.princeton.eau
©
LesÉditions
dePhysique
1996lE+04
Ge lE+02
~Si
-
E+00
~_jgçj c~(Ncsj
~
~
)
lE~2
À~
g
lE~4YBCO
à
lE-06
i~~~
COPPer
iE-io
i io ioo iooo
Tenperature (K)
Fig.
l.Log-log plot
of trie resistivities of some common conductors. Circles indicate the super-conducting
transition temperatures. Note that(TMTSF)2Cl04
looks most similar to Cu menai. Bonnare very clean, with mean free
patins
of many microns ai low temperature. Trie main dilference is trie factor ofr- lo00 carrier
density
from thelarger
unit cell in theorganic.
chaiu,
we wouldexpect
a quarter filled TMTSF baud. But trie dimerizationsplits
trie TMTSF baud into two bauds and leaves us with a one half filled Upper baud asrequired by
triecharge
transfer aud
stoichiometry.
Trie
highly anisotropic bandwidths, 4ta 4tb 4tc
re eV 0.1eV 0.003 eVimply conductivity anisotropies
in trie ratio (t~/tj)~.
Trie measured conductivities are: aa ab i a~ m10~
10~10~~(Q cm)~~
[2]. Thetemperature dependent
resistivities of severalpreseutly interesting
materials is shown inFigure
1. The transition temperatures of thesupercouductors
are iudicated
by
thelarge
circlesterminatiug
trie low eud of trieresistivity
curves. It is clear that trie(TMTSF)2Cl04
sait is agood
metal with resistaucedecreasiug
about three orders ofmagnitude
oucooliug
from roomtemperature
to il K. It is very similar toCopper,
trie main differencebeing
that trie carrierdensity
is much less. TrieBechgaard
salts bave about carrier per 1000 cubicAngstrom
unitcell,
while Cu bas one carrier per 1 cubicAngstrom
Cu atom.This
essentially
accounts for trie thousandfold difference inconductivity.
Trie low temperature
phase diagram
for trie wunder material(TMTSF)2PF6
is shown inFigure
2[3,4].
Above 12 K triesystem
is metallic.Upon coohng
at ambient pressure there is a ver»weakly
first order transition to aSpin Density
Wave(SDW) insulating phase
as evidencedby antiferromagnetic
resonance, NMR and muon spm rotation [5]. Trie SDW transition tem-perature con be
suppressed by application
of moderate pressure until ai r- 6 kbar trie metallicstate is reestablished. Once trie SDW is
suppressed
triecrystal
becomessuperconducting
at1.2 K
[6j.
This material was trie first organicsuperconductor
discovered. Further increasing trie pressuregradually
reduces triesuperconducting
transitiontemperature. Applying
a smallmagnetic
fieldalong
trie leastconducting
direction(to
mduce screemng currents in triehighly conducting plane)
kilts triesuperconductivity
above a critical field ofr- 500 Gauss [7].
N°12 FIELD INDUCED SPIN DENSITY WAVES 1877
'~~,
, ,
' ',
' ',
' ',
,~ i' ',,~
'~~Î~'~ ~~~éÎ
' ',
'
i
,
%$Î~
'
$~
~
~ ~
j~ ~
1~ ,
/~
fl
',
' '
~
, ,
s
,,
'
»j ~WÙ
Fig.
2. The low temperaturephase diagram
of(TMTSF)2PF6 (3,4].
While trieproximity
of trie SDWor
antiferromagnetic
insulatorphase
to triesuperconduct,ing phase
is common to manysy~tems, trie field induced SDW cascade seen above 6 kbar and
r- 5 testa is su f~r
unique
tu triequasi-one-dimensional
organic conductors [8]. Each separatephase
in trie cascade withincreasing
fieldcorresponds
to a different quantum Hall state.All of trie
phases
discussed up to thispoint
bave been observed in other materials. Triejuxtaposition
ofsuperconducting
and SDW oraiitiferromagnetic phases
is now often seen mhighly
correlated systems such as triehigh
temperaturesuperconductors.. However,
what isnew and so far unique to these materais [8] is trie
phase
which results from furthermcreasing
trie
magnetic
field while in trie lowtemperature
metallic state[9]. Imagine starting
at 10kbar,
at 0.5 K and
applymg
a field(along
trie cdirection).
500 Gauss kilts triesuperconductivity
and leaves us with a "normal" metal
(more
about thatlater).
Atslightly
more than 5 tes-las we encounter another
phase boundary
where trieresistivity sharply
mcreases as does trieHall coefficient. Further
increasing
field this transition is followedby
a cascade of at least 9 consecutive first order transitions to different semimetal states until atr- 20 testas we enter trie "final"
insulating phase.
Most remarkable of ail. each transition takes us to a state witha well defined and
sequenced (..1/4,1/3,1/2,1) quantized
Hall resistance[10j.
It isparticularly striking
since trieQuantum
Hall Effect(QHE)
isintrinsically
trieproperty
of a two-dimensionalelectron system and this materai was trie first to exhibit trie
QHE
in a bulk three-dimensionalcrystal.
This cascade ofphase
transitions is to Field InducedSpin Density
Wave(FISDW)
states and is trie main
topic
of this paper.2. One Dimensionalization in a
Magnetic
FieldThe Fermi surface which we associate with this bandstructure is cartooned in
Figure
3. It consists of twonon-intersecting slightly warped
sheets. Trie actual Fermi surface and Brillouink~
2x/b
Fig.
3. Cartoon of the idealized Fermi surface of theBechgaard
salis.(Trie
realcrystal
structure is triclinic nonorthorhombic).
Trie Fermi surface consists ofnonintersecting
sheetsalong
triehighly
conducting
a directionwarped by 4tbleF
and4tcleF
in the b and c directions.Zone are more
complex il ii.
Triecrystal
structure istriclinic,
non orthorhombic as in trie car- toon and there isoverlap
between one TMTSF molecule on a chair and several molecules on aneighboring
chair. If trie softs weretruly
one-dimensionalelectronically
then trie Fermi surface would consist of twoparallel
sheets at+2kF.
Trie warping is due to trie finite bandwidths m trie b and c direction(trie
bwarping
is about trie correct size, but trie c axis isexaggerated by
about a factor of30.)
Agood
deal of triephenomena
observed in theBechgaard
softs aredirectly
related to this Fermi surface which at firstglance
seems innocuous. There are no closed orbits and hence no chance for Landauquantization.
Triewarping
seems toolarge
to allow theone-dimensional instabilities associated with trie Peierls transition. For H (
c
(perpendicular
to trie
highly conducting plane)
we expect smallsaturating magnetoresistance along
trie a axis and sizablenonsaturating magnetoresistance along
b and c[12].
Let us
forget
about trie leastconducting
c direction for trie moment and concentrate on the Fermi surface and bandstructure in trie a-bplane.
Trie essence of theproblem
is that oflarge amsotropy
andonly
open orbits of trie Fermi surface. Adispersion
relation with theseproperties
con be wntten as£(kx, ky
=-2ta
cosk~a 2tb
coskg
b re1~ ~) 2tb
caskyb (2.1)
where we
temporarily
choose a free electron form for thedispersion along
x to avoid com-plications
due tonesting
andcommensurability [13,14].
We want to see the effect of a fieldperpendicular
to theplane.
A Landau-Peierls substitution k ~ iveA/hc,
with the choice of a Landau gauge A=
(0, Hz)
leads to:£ $il
2tb Cos(ib) ~)~x)il
= fil(2.2)
m x y c
Since y
only
appears in triepartial denvative,
trie wavefunction con be factored asil(~, vi
=
e~~YY
<(-ri
anddefining
themagnetic wavelength
as2~/À
=
eHb/hc
we bave:~ÎÎÎ2~~~~ ~~~~°~~~~~ Î~~~~~~ ~~~~~
~~'~~N°12
FJjLD
INDUCED SPIN DENSITY WAVES 1879eHv~b
a~=-
4ick space
ÎÎÎ~ ~Î~
~~+~
r~al SPaC~
~- ~~eHb
Fig.
4. Real space and momentum spacequasi-classical
trajectories of electrorts in the presertce ofa
magnetic
fieldalong
c. In k space the motionalong
the open orbitsperiodically
crosses the Brillouinzone with
frequency
uJb. In real space the motionis Iocalized to r~
4tô/~~b
oc1/H
chainsalong
b.The motion is extended along the chains but there is a
spatial
modulation ai themagnetic Iength
=
hcleHb
= ~o
/Hb (where
~o is the fluxquanta).
Since
ky
appearsonly
in theargument
of thecosine,
itonly
serves to shift theorigin
of x(hence
the center of mass of thewavefunction,
x'~ x
lbky/2~). Equation (3)
is then a one- dimensionalSchrodinger's equation
for aperiodic potential (in
fact it is Mathieu'sequation) [15].
In amagnetic
field thedispersion depends
onkz, ejkz
rather than trie zero fielddispersion e(k~, ky)
Thus themagnetic
field makes triesystem electromcally
one-dimensional.We can understand this one dimensionalization tram the
quasi-dassical
motion of the elec-trons on trie Fermi surface. Trie two-dimensional Fermi surface is
schematically
shown inFig-
ure 4. Trie electron
velocity
vk=
s7ke
isperpendicular
to trie Fermi surface. In trie presence of amagnetic
field electrons are constrained to move on constant energy surfaces. Electrons follow trieequation
of motionltôk/ôt
= evk xB/c
~ trie real spacevelocity
vk ~ôr/ôt
and k space motion
ôk/ôt
aresimply
rotatedby
90°. There is a characteristicfrequency
uJb +
(ôkb/ôt)/(2~/b)
=
eufHb/hc
with which the electron crosses trie Brillouin zone in trie b(y)
direction.(u~
isapproxiniately
constant at tif on the Fermi surface. Thedispersion along
xis often taken in linearized form as
ltufk~.)
The real space electron motion is shown inFigure
4 bottom. It is limitedalong
y and extendedalong
x, 1-e- one-dimensional. Trie width of trie orbit(which
isactually
the extent of thequantum
wavefunctionalong
y(16]
is(4tb/ltuJb16
and there is a newperiodicity,
G = 2~Il
=
eHb/ltc, given by
triemagnetic length
1along
xil 7j.
(The length
is such that#o
" lbH with çio "
hcle
a fluxquanta.
Note that for anisotropic
two-dimensional electron system the
magnetic length lH
is such that po "1(H.
For 10 testasQa Q°
§Qt.K
.4
(Q"l~~~
Fig.
5.Susceptibility
of trie electromc system to a distortion of wavevectorQ. x(Q)
[29]. The FISDW wavevector is determinedby
theQ
ai the maximum of x,m ibis
case
Q". (SmaII changes
m bandstructure can shift trie maximum to either of the Qo's
Ieading
tochanges
in thesign
of the quantum Hallresistance.)
trie wavefunction is
spread
over about 60 airains in trie b direction and trieperiodicity
inducedon trie airains in trie a direction is about 200 unit
cells.) [18]
The effective one-dimensionalization of trie electrons
by
triemagnetic
field can and does baveinteresting
consequences. One-dimensional metals are unstableagainst
trie formation ofcharge
andspin density
waves(Peierls transitions) (19].
Trie effects of electron interactionsbecome more
important
in onedimension,
Fermiliqmds
give way toLuttinger liquids
and other manybody ground
states[20, 21].
Disorderplays
a drastic rote andalways
leads to locahzation andinsulating
behavior at least fornon-interacting
electrons. There is nolong
range order and hence no finite
temperature
transitions in one dimension. Trie absence of any closed currentloops
wouldsuggest
that triesuperconducting
cntical field should tend tomfinity
for one-dimensionalsystems (22].
Of this nch set ofproposed
consequences of field induced onedimensionahty,
trieonly
one of which wepresently
bave conclusive evidence is trie Field InducedSpin Density
Wave(FISDW)
transitions[9,17, 23j.
3. Trie FISDW Phase
Diagram
The
instability
of trie metallic state can be obtainedby calculating
triesusceptibility
to aperi-
odic spm
density
modulation at trie wavevector q. Thissusceptibility
is shown mFigure
5. Trie one-dimensional character of trie electronicdispersion
in amagnetic
field assuresdivergences
in trie
susceptibility
as temperature is lowered toward zero.Using
a Stoner critena we thenexpect
a transition into an SDW statecorresponding
to trie wavevector at which triesuscepti- bility
is maximum. The wavevectors of trie maxima shift withmagnetic
field(17, 24, 25]. They
alwavs occur at
2kF ~112~/À along ka
but trie kb value varies with bandstructure and field in such a way as toyield quantized
k 9pace areas.N°12 FIELD INDUCED SPIN DENSITY WAVES 1881
À
qualitative explanation
of the FISDW'S aise follows fromconsidering
thepossible ground
states of the
system (rather
than theinstability
oncooling
from the metallicstate).
À Due-dimensional
system
is unstableagamst
the formation of adensity
wave because a distortion ofwavevector
2kF exactly
nests trie two sides of trie Fermi surface. Thisproduces
a gap at theFermi energy. The total electronic energy is lowered
by A~
In eF/~l
while trie distortion costs anenergy of order
KA~
where K is an elastic constant or anexchange
interaction(19).
Because of thelog
term the electronic energyalways
wins(for
smallAl
and thedensity
wave is stable. Fora
quasi-one-dimensional dispersion,
thenesting
of the Fermi surface is netperfect
as illustratedschematically
inFigure
6a. À distortion of2kF Produces
a gap atonly
a fewpoints
on the Fermi surface. These gaps close off electron and halepockets
and thedispersion
is that of a semi-metal. Trie electronic energylowenng
is of order ti~ In eFIA (with
m >3),
netenough
to overcome the cost of
making
trie distortion. Trie metallic state of triequasi-one-dimensional
system
is stable. This is the situation for(TMTSF)2PF6
under pressure without amagnetic
field. It remains metallic to T= 0.
Dimension ail states at sf
,
£
coupled by q=2kf
,~ £~ in gap
everywhere
c
distortion stable
£~ 2 W
2n/b
(6aÎ)
-kf ka~ kf
h e
Imper§ect Nesting
leaves pockets distodion~
~
~b
~f ~°~~~~~~(6a2)
k
~
a)
Fig.
6.ai
Distortion with H= 0.
ii
Forperfect nesting (as
m one
dimension)
asurgie
wavevector maps one side of the Fermi surface to the other opening acomplete
gap ai the Fermi energy,stabilizirtg
the distortion and the
msulating phase. 2) Imperfect
nesting Ieaves electron andfor
holepockets
ona
partially gapped
Fermi surface. There is nolonger
a gap ai EF and ibis state is non lower energy thon trie menai without dïstortion.b)
No distortion, finite H 1) For closed orbits area, ertergy artd Hall resistance arequantized. 2)
For the open orbitBechgaard
salis there are nointeresting
elfects ai EF.c)
Distortion and field. The areas and energies m the electron and holepockets
arequantized,
the distortion wavevector adjusts so that EF lies between Landau Ievels. Since EF iscompletely
in a gap trie distortion is stabilized andconcurrently
we baveouly
filled Landau Ievels and trie quantum Halleifect.
Closcd
°'bit landau
Lcvcls
QHE
g
(6b1) En
=(n
+la)
Jfoc
Magnetic Field
Quasi-ID
~Opencfbit
No Landau
£j QllwtiZation
(6b2)
bj ka~
Distonion
+Field
£f in Gap distonion
q ~f "~~
+
Landau Quantization QHE
~j (6ci
Fig.
6. Contmued.Now lets consider the effect of a
magnetic
field on an undistorted metal~Figure
6b. If we bave closed orbits then we bave Landauquantization,
discrete energy levels and trieQuantum
Hall effect(in
a two-dimensionalsystem).
However. for open orbits there is noquantization
condition and trie energy
dispersion
remains continuous with a dassicaljoften zero)
Hall effect(26].
Trie quasi-one dimensional metal seemsumnteresting
bath in itsstability
and lack ofmagnetic
field effects.N°12 FIELD INDUCED SPIN DENSITY WAVES 1883
If we bave bath a distortion and a
magnetic
field weregain something interesting, Figure
6c
(27].
Trie distortion leads to dosed orbits in trie semi-metalpockets.
These must be Landauquantized
withspacing huJ,
uJ =eH/m*c.
If trie Fermi energy lies between Landaulevels,
thenwe bave a
completely gapped system,
trie electronic energygain again
beats eut trie distortion energy and triesystem
lowers its energyby distorting, forming
an SDW. How con eFalways
lie in thegap? Suppose
we were tochange
themagnetic
field. Thespacing
of the Landaulevels shifts and we
might expect
eF to enter a Landau level. Without a gap the SDW wouldcollapse. However,
eF remains in the gap aslong
as we haveonly
filled Landaulevels,
1.e. thearea in the k space
pockets
must be anintegral multiple
ofeH/hc (the
area which canexactly
accommodate trie number of states in a Landaulevel).
Trie system canadjust
trie area of triepockets by changing
the wavevector of the SDW distortion.Equivalently,
the distortionwavevector dictates the
top
and bottom of the electron and halepockets.
As the fieldchanges
the qsDw
changes
SO as tokeep
eF in the gap between Landau levels. Àt some field it may be that trie energy is loweredby
qsDwjumping
to a different value SO that eF sits betweenanother set of Landau levels.
Trie situation where there are
only completely
filled landau levels isprecisely
trie condition whichgives
rise to theQuantum
Hall effect[28].
In a conventional two-dimensional electron gas eF sits between Landau levelsonly
m the presence of disorder induced localized states. In the present case eF sits in the gap for intrinsic energy reasons. In fact we could net have a stable FISDW without theQHE (and conversely).
As field ischanged
we should have first order transitions between SD~V states with different quantum numbers. Trieordering
of triequantum occupations
as field is mcreaseddepends
on details of trie bandstructure and trienesting [29].
In ail cases triehigh
field state should bave a widesplitting
between the lowest energy electron Landau level and triehighest
energy hale Landau level. Thishigh
field limit is an insulator. In triesimplest picture
as trie field is lowered there is asingle pocket
and triequantum
numbers follow n=
0.1, 2, 3, 4,...
as trie field is lowered.Trie cartoon
description
given above is culled from agreat
deal of theoretical workby
anumber of groups who bave
converged
on a "Standard model" or"quantized nesting
model"for trie FISDW'S
[17, 24,25].
Triepicture
which emerges from these detailed calculations is illustrated inFigure
7.Starting
fromhigh
field we should bave trie FISDW insulator state with no filled Landaulevels,
N=
0,
andessentially
trie same transitiontemperature
as for triezero field SDW with
perfect nesting.
Onlowering
trie field we enter trie N=
phase
with semi- met allicinteger
quantum Hall behavior. Furtherlowering
trie fieldyields
trie cascade of FISDW transitions to differentphases, separated by
first order transitions and each associated with a different Landau levelfilling
andquantum
Hallplateau.
Within eachphase
the wavevector for trie SDWchanges continuously
with field(qfisDw
=
(2kF
~1~2~Il,
~lb
+e)
to maintain triecomplete filling
of trie Landau levels. Betweenphases
trie wavevectorchanges discontinuously
with the acomponent having
a different value of n. In thesimplest
case i~= N and the cascade results m the
stepwise
decrease of n assuggested
in thefigure. However, depending
on thebandstructure,
bath the sequence of wavevectors i~ and their association with N(trie
number of filled Landaulevels)
maychange [29].
Àt pressures somewhat above the critical pressure needed to suppress the ambient field
SDW,
bath trie
PF6
Salt and trieÀsF6
Salt show trie transitions aspredicted by
thesimplest
form of triequantized nesting
model. Trie Hall resistance(at
T= 0.5
K)
and transitiontemperature
are shown for a
PF6 sample
at llkbar inFigure
8, and Hall andlongitudinal
resistance are shown inFigure
9 for anÀsF6 sample
at T = 0.5 K and 10 kbar[30].
We see trie
stepwise
decrease m trie Hall resistancefollowing
p~~ = h/2Ne~
asexpected [31].
The absolute value of the Hall resistance m trie
plateaus
isroughly
at triequantized
value given aboveii.
e. 13 k QID
perplane)
for"good" samples (where good
meansagreement
andF-
1 n=0
1
Àlxy
f
2e~o
H
Fig.
7.According
to trie standard or quantizednesting
mortel thereis a cascade of FISDW tran- sitions
as field is increased. Associated with each transition the Iow temperature Hall resistance should show a quantized
plateau.
In triesimplest
case trie plateaus, p~~=
h/2Ne~,
mcrease as
N
= ..5, 4, 3,2. 1,0, with
increasing
field.io
o
-
50 C
jzs
ce
o
0 5 10 15 z0 z5
H(T)
Fig.
8. Data for(TMTSF)2PF6
ai 11 kbar pressureshowing
trie phasediagram
and the Hall resistance ai 0.5 K[3,4].
N°12 FIELD INDUCED SPIN DENSITY WAVES 1885
~~z
(TMTSF)2AsF6 to kbar oSK
6
1
~-z
DÎ~
ioo
6
50 1
DÎ
o
0 à 10 15 z0 25
H(T)
Fig.
9.(TMTSF)2AsF6
shows similar behavior to PF6. Here thelongitudinal
resistance is seen to decrease withincreasing
field in theregiou
of the Hallplateaus
and to exhibitpeaks
in the region of the transitions betweenFISDW-QHE
states. Above 18 testa thesample
enfers the N= 0 state which
is a SDW insulator
[3,4].
presumably
that the currentpaths
areuniformly
distributed between theconducting planes
that make up triecrystal).
Trie differences from
conventionalinteger quantum
Hall Effect insingle layer
two-dimensional electronsystems je-g- GaÀs)
are the factor of 2 in the denominator(from
thespin degeneracy resulting
from thespin pairing
in the SDWstate) [31],
the absence of any linearregion
between theplateaus (from
the first order nature of the transitions between the different N states, orequivalently,
the fact that thesystem
canenergetically
never find itself with eF located in a Landaulevel),
and the tact that trieplateaus
do net sit on a hne p~y =Hli~~e (with
n~ the fixed electrondensity)
whichextrapolates
to zero at zero field(from
trie fact that the effective carrierconcentration is
changing
with field as trienesting changes
and triepockets
shrink toward zero athigh field.)
In trie conventionaltwo-dimensional electron
system (2DES) QHE
thelongitudinal
resistance p~~ goes toward zero in the Hallplateaus
and has local maxima between trieplateaus.
À similarbehavior is seen m
Figure
10. In fact p~~ decreases withdecreasmg temperature
in trieplateaus (approximately exponentially)
and increasesexponentially
in the N= 0
phase [32].
There are several otherinteresting
differences between thissystem
and the 2DES. The dosed orbits whichquantize
into the Landau levels are createdby
the SDW distortion and itsresulting
SDW gap. The gap is
relatively
small and themagnetic
fields present are ofcomparable magnitude.
We cari therefore bavemagnetic
breakdownthrough
these gaps and the result istunneling
between Landau levels. Trie Landau levels become Landau bauds.Depending
on trieparticular
bandstructure and qsDw we bave an intricate set of bauds and gaps. Anexample
isshown in
Figure
11 which shows trie calculated gaps about eF as trie field is vaned and different states m trie cascade are present[33].
io~
10°
§
~$
n"0
n=1
n-Z n~3 75
_
)
50
ç/
~~ 25
nm3
o 0
(K)
Fig.
10.Temperature dependence
of thelongitudinal
and Hall resistance of(TMTSF)2AsF6.
TheHall resistance
monotonically
increases to ils quantized value as temperalure is Ioweredthrough
theFISDW transition. For the N = 0 SDW
insulating
state the resistance mcreases due to theopening
of the SDW gap and the absence of any filled Landau levels. For N
#
0 states the resistance risesjust
below TsDw from the Ioss of carriers but then decreases as thedissipationless
transport associated with theQHE
takesover ai Iower temperatures.
4. Tilted
Magnetic
Fields and Lebed ResonancesSO far trie FISDW bas been descnbed as trie
mstability
of aquasi-one-dimensional
metal m a two-dimensional space.(Expenmentally
triephase diagram basically
scalesOrly
with the fieldperpendicular
to the a-bplane
when trie field istilted.)
Trie real electromcsystem
is of course three-dimensional and here we consider some of trie effects of triethree-dimensionality.
If trie bandwidth in trie third direction islarger
than trie SDW gap, or triespacing
of trie Landau levels(these
numbers arecomparable)
then trie bands Willoverlap
and trie FISDW and trie associatedQHE
willdisappear.
Trie best measurements and calculations agree that4t,
m 0.003 eur- 30
K,
easily enough
to kilt trie two-dimensional effect. The clever solution that nature bas found andapphed
in thesesofts,
is to choose a wavevector of~/c along
c, 1-e- to alternate trie SDW inneighbonng planes.
Just as the i~>avevector(2kF, ~/b)
leads toperfect nesting
andcomplete
gapping of the Fermi surface for triedispersion
relatione(k)
=
ltvfkz 2tb coskyb, (2kF, ~/b, ~/c)
does trie same fore(k)
=
ltvfkz 2tb
Coskyb 2tb
Coskzc.
If trie xdispersion
is
quadratic,
trienesting
isimperfect
and what remains of triedispersion
is tbe~ ~t)lef
and tc~~ ~wt)lef
~w
10~~ev
~w o-1 K. For a
strictly
two-dimensionalsystem
any finitemagnetic
fieldN°12 FIELD INDUCED SPIN DENSITY WAVES 1887
uJ
o
' 6 à lu 20 3o
H 11)
Fig.
ii. The gap structure m the FISDW states are shown as a functiou ofmaguetic
field [33].The dark
regions
are the gaps which separate the white Landau bands. Note that thelargest
gap isceutered around eF.
would
produce
an FISDW if we cooled to lowenough temperature [17].
À limite tc~~ mandatesa threshold field below which we carnet attain an FISDW.
Experimentally
this bas been round to be whenhuJcr,t
+~ tc~~
r-
o-1 K as
expected
aise from consideration of trieQHE
and Landauquantization [34].
Lebed
[35] suggested
that there was a way around trie threedimensionality
and trie zerotemperature
threshold field.Up
to thispoint
we bave treated trie case when triemagnetic
field isaligned along
trie c axis. If trie field is tilted in trie b-cplane,
trie electronic orbits sweepacross trie Fermi surface as illustrated in
Figure
12. Trie k spaceequation
of motion issimply
hdk/dt
= eV x B m evf~ x B so that there are now two raturai
frequencies,
Due forcrossing
trie Brillouin zone in trie
direction, nô
"
Îô/(2~/b)
=
evfbB cos9/lt
and Due for trie(irection,
Q~ =Îc/(2~/c)
=
evfcB
sin9/h.
Àt certainangles,
tan 9=
pb/qc,
with p and qintegers,
thesefrequencies
arerationally related, Qc /Qb
"
P/Q.
For theseparticular angles,
trie electron orbitsretrace their motion and the
trajectory
is ahue,
for otherangles
trie electron orbits do trotretrace and end up covenng trie entire Fermi surface. Lebed
argued
that the commensuratemotion at trie
"magic" angles couples
triedispersion
in trie two open orbit directions and reduces the three-dimensionaldispersion
to one-dimensional. Àt theseangles
trie threshold2z/b
« »
a) arbibary angle j.space) c) p/q= In magic angle
/Î=1 /q=iQ
. . .
/~
. .
b) P'q=' ma#c anfle d) mai space iatùce Î
Fig.
12.a)
Motion of an electron across the Fermi surface in the presence of amagnetic
fieldapplied
in the b-cplane (perpendicular
toa)
ai anangle
à from the c axis.Umklapp scattering
at the Brillouin zone boundariesproduce
atrajectory
which covers the entire Fermi surface.b), c)
Fora
specific angle
à=
tan~~(pb/qc)
the k spacetrajectories
retrace on the Fermi surface.d)These "magic"
or Lebed
angles
correspond to themagnetic
fieldpointing
between molecular centers(or along
real space translationvectors).
field would be reduced to zero
(at
zerotemperature).
Trieanalogous effect,
trie increase of trie transitiontemperature
of trie FISDW at trie Lebedangles
bas been seen[36].
À few years after Lebed's
prediction
of triecommensurability
effect on trie thresholdfield,
came trie idea that trie
angular
resonance on trie Fenni surface should aise bave drastic effects on trietransport propeities [37]. Expenmentally,
these were first observed in trieCl04
Salt a axis resistance[38].
Laterthey
were seenalong
bath a and c directions in bath trieCl04
andPFô
salts when
they
are in trie metallic state at lowtemperature [39].
Ii isfairly
easy to understandwhy transport properties
arechanged
at trie Lebedmagic angles:
trierepeating
orbits cari avoidhigh scattering
regions of trie Fermi surfaceand/or provide
different averages over trie velocities.Trie one or two
dimensionality
of triedispersion
bas drastic effects on trie nature of bathimpufity
and electron-electronscattefing.
To date there arer- 10 different
modelslexplanations
for trie magic
angle
resonances, none of which bassatisfactorily explained
ail of trie data[38,40,41].
This ispartly
because trie effects differ from one sali to trie orner. There areespecially Sharp
and unusual structures found in triePF6
Salt under pressure:Figure
13.Àlthough
the basic idea for trie Lebed resonances cames from k space, Fermi surface argu- mentsFigure
12, trie faon that k space vectors areperpendicular
to real space vectors~ and that trievelocity
isperpendicular
to triefield,
translates to trie fact thon trie"magie" angles
bavea real space
interpretation. They
aresimply
trieangles
at which trie field is oriented in trie direction between trie actual molecules in different chains(Fig. 12d).
This observation led toN°12 FIELD INDUCED SPIN DENSITY WAVES 1889
7
~ (Hcoso)~/~
6 xx
5 Q
6~4
x
2
1
/
~~
~
~
zz
0.3 Ô~
0.
'~-ioo
d
Fig.
13. Angulardependent magnetoresist~nce along
trie a(top)
and c(bottom)
axes for(TMTSF)2PF6
ai 11 kbar. 4 testa and 0.5K. trie Lebeddips
aip/q
= 0, +1 are evident. The "back-
ground"
magnetoresistance bas an unusual form which seems todepend
ou the componeut of the fieldperpendicular
to the mostconducting plane
to a power.one of the most
intrigmng
ideas ou theongin
of themagic angle
effects, at least for thePFU
Salt. The resistivitiesalong
the most and leastconducting
directions are shown for this sali inFigure
13. Triesuggestion
was that trie Salt ismarginally
a Fermihquid [41].
It ispresumed
that theplanes
would be m a non-Fermiliquid (NFL)
state ifthey
wereuncoupled.
The elec- troniccoupling along
the c axis, tc, is(presumably) just barely enough
to lead to coherent transportalong
c andconsequently
trie destruction of trie 2 dimensional NFL state m favor ofa conventional 3 dimensional Fermi
liquid.
In this scenario a small fieldalong
b reducestaon
below trie critical value and leaves a 2D NFL. In such a 2Dsystem
ailmagnetotransport
mustdepend only
on the fieldperpendicular
to the 2Dplane
andmight
be m trie form of powero-o
~
© ~W~ -ST
tlC
,oQ
1 ~
~~
~
~l~
~ m
~llÎ -l.5
O
log[Hcos8]
Fig.
14. Trie power Îawdependence
on fieldperpendicular
to a-b plane(H cos(à))
is illustrated byIog-Iog
plot of c axismagnetoresistance
forsample
rotations taken ai diifereut fields. Trie same curveresults when field sweeps are taken at fixed
angles.
Trie deviations occur ai trie Lebedangles.
Same parameters as inFigure
13.laws, (H
Cos6)~.
At the magieangles tc~~
is net as reduced and trie resistancedraps.
Measure- ments show that aside from triemagie angles
themagnetoresistances
vary p~~ cç(H
Cos6)°.~
and pzz cç
(Hcos6)~
~ ~° ~.~ when either H is varied at constant 6 or 6 is vaned at constant H(42, 43], Figure
14. Fiom this point of view trie "normal" metal state of(TMTSF)2PF6
is anon Fermi
liqmd
in trie presence of a small fieldalong
b.Trie
deceptively simple
Fermisurface,
which ai firstglance yields
no hint ofinteresting
behav-ior in a
magnetic field,
alsoproduces large angular dependent magnetoresistance
for rotations in trie a-cplane [44]
and mostrecently
in trie a-bplane [45j.
SO for these resonances bave beenreadily explainable,
at least in trieCl04
sali in terms of conventional Boltzniann transport, with a coherent Fermi surface[46]. They
bave been able toprovide
useful information about trie bandstructure parameters much as bas been donepreviously
with closed orbit menais.5. Trie Anomalous behavior of
(TMTSF)zCl04
Àlthough
thePF6
sait of TMTSF seems an idealexample
of triesimplest
form of trie standard model for trieFISDW'S,
trieCl04
sait presents a much morecomplex
behavior. To date trieCl04
sait is more studied than triePF6
for both FISDW'S and forsuperconductivity
even
though
bothphenomena
were discovered first in thePF6. Cl04
is an anibient pressuresupercondiictor
and FISDWsystem
whilePF6 requires
a pressure ofgreater
thon 6 kbar.This makes it much barder to meurt,
align,
and rotate triePF6 samples
and makes mostthermodynamic
and elasticproperties
almostimpossible
to measure. Ail of these measurements bave been done onCl04 (4T48j.
Cl04
has an anionordenng
transition at 24K which leads to adoubling
of trie unit oeil intrie b direction
[49j.
It was well known that trie lowtemperature properties
of thissait,
its resistance, itssuperconducting
transitiontemperature
etc.depended strongly
on howrapidly
triesample
was cooledthrough
this anionordenng
transition.However~
thé extremesensitivity
of triequantum
Hall steps,especially
triechange
in sign or "Ribault" anomali[50j,
was un-expected.
For many years trieonly
Hall data was fromCl04
and masked trieacceptance
ofN°12 FIELD INDUCED SPIN DENSITY WAVES 1891
~.~
b'
(TVTSF)~PF,
~'~~
fi
a 5
~'~~
2 3 ' Sompie #1
0.50 P * il kb°~
i o 0.25
~~'~°Ù
~'~~ ~~~
ç~
0 5p = 8.3 kbor -
O'~
(~)So~np,e il o o
~~~
~ ~~ ~~~
o.z5 - -o 5
o oo ~~~
~~~~j~ ~~
_~ ~~ p « BS kbar
~'-~
Tm 380 mK.? ~'/~ oiz
p~= h/2
e'p~~ /Po
0.080.6 (d) -
0.04
$
ù_~ somp,e ;2 o.ooli
t~ ~ -
~
je) O-O
5 lO 15 20
MagneLic
field(Lesla)
Fig.
15.(TMTSF)2
PF6 exhibits"uegative"
Hall steps ai pressures below about 9 kbar [51], similar to what, bas been observed eallier in(TMTSF)2CI04s by
Ribault et ai. [50].They
have been attributed to aspecific
warpmg of the Fermi surface. po =H/2e~.
trie
QHE
in trieBechgaard
salts until several groups(10]
looked atPF6
andgot
trie data hke that shown inFigure
8. More recent data on bathPF6 (3, 51j, Figure
15 andCl04
(4]land
onRe04) Î52j
at different pressures show that triesign changes
con be found in both materials and are mostprobably
associated withchanges
in thenesting
vectorresulting
from subtletiesin the bandstructure
[29].
The lack of a
simple
staircase for the Hall resistance isOrly
Due of many anomalous behaviors of trie FISDW inCl04. Although expenments
have been carried out to veryhigh
fields(50
T at
~w
il
trie N= 0
insulating
state basyet
to be found. At lowtemperature
there is avery wide field
region
from 7.5 to 27 Tesla where trie Hall resistance is flot and trielongitudinal
resistance is