VOLUME 42, NUMBER 5 PHYSICAL REVIEW LETTERS 29 JANUARY 1979

Reemergent Order of Chaotic Circular Couette Flow

R. W. Walden ' and R. J. Donnelly

Department of Physics, University of Oregon, Eugene, Dregon 97403
(Received 4 August 1978)
Measurements of the fluctuations in local velocity in fluid motion between a rotating
inner cylinder and a fixed outer cylinder are reported for rotation rates up to 67R, '

where P, is the critical Reynolds number for Taylor instability. The Reynolds number
R& for which the transition from periodic or quasiperiodic to aperiodic flow occurs de-
pends strongly on aspect ratio (F = fluid height/gap}: R, ~ 22R, for F = 20; R, = 26R, for
I'=80. For F &25 a sharp periodic component reemerges in the power spectrum for
28R SR: 868 .
In couette flow between concentric cylinders gap, except for a 6.4&&10 '-cm gap between the
with the outer cylinder fixed a series of instabil- inner cylinder and the ring. The upper boundary
ities occurs with slowly increasing rotation rate is a free surface. Visual observations have been
of the inner cylinder. The fluid motion, which is made through a glass outer cylinder with fine
laminar and parallel to the motion of the inner aluminum flakes (aluminum paint pigment) sus-
cylinder for small rotation rates, is transformed pended in the working fluid (carbon tetrachloride).
to a steady pattern of toroidal rings or "Taylor Electrical measurements are made using a
cells" following the first instability. With in- pair of gold-plated brass cylinders. The outer
creasing stress these rings acquire azimuthal cylinder has an array of 1500 ion collectors from
waves and evolve through a series of "states" de- which any four may be connected for computer
fined by various combinations of axial and azi- sampling at a given time. The 1-mm-diam ion
muthal wavelength. ' These waves generate a dis- collectors produce an output current of the order
tinct spectral line with an appropriate detector. of 0. 1 picoampere and provide a local probe (-5
At some Reynolds number denoted R, all line x 10 ' cm') of the radial component of fluid veloc-
spectral features disappear; the flow seems to ity in a narrow boundary layer at the outer cylin-
have lost its azimuthal periodicity and in that '
der. The entire apparatus is enclosed in a wood-
sense to be disordered, ' We call the change at en box which maintains temperature constant to
"
R, an "aperiodic transition, but caution the read- + 0. 1' C, and inner-cylinder rotation rate is
er to recall that axial periodicity persists to the locked to a programmable synthesizer which
highest Reynolds numbers yet observed. Analo- maintains constant speed to better than + 0. 01%%uo
gous behavior has been reported for other sys- or permits uniform acceleration at any conven-
tems' as, for example, in Rayleigh-Benard con- ient rate under computer control. '
vection. The Reynolds number R = 2it QR,d/v (where 27tQ
We now find, however, that for sufficiently = inner cylinder rotation rate, v =kinematic vis-
large aspect ratio F=L/d (L =fluid height; d=R, cosity, and d =R, -R, ) will serve as a dimension-
-R, = gap between cylinders) a sharply defined less measure of rotation rate in the following dis-
azimuthal wave reemerges for inner cylinder ro- cussion. R, = 118.3 is the critical Reynolds num-
tation rate well beyond the "aperiodic transition" ber for the onset of Taylor cells.
at R, . Further, the aspect ratio greatly influenc- For the experiments reported here aspect ratio
es the rotation rate for which the aperiodic tran- I' has ranged from 18 to 80, and data have been
sition occurs and the set of possible states ac- taken for Reynolds numbers up to R = 67R, . The
cessible to the system immediately below the observable set of "states" accessible to the sys-
transition. tem as R is increased through R, has been de-
The apparatus for our experiments consists of scribed in Refs. 1 and 2. One example is shown
two pairs of cylinders of nearly identical physi- in Fig. 1. The spectral peaks denoted co, and +,
cal dimensions one pair for visual observations scale roughly with cylinder rotation frequency ~;
and the other for electrical measurements. In the ratio to, /&o, varies slowly as a function of R
each case the inner cylinder radius is Ry= 2 2225 as reported by Fenstermacher et a/. ' As the as-
+ 0.0003 cm, and the outer cylinder radius is R, pect ratio I is increased, R, increases and the
= 2. 5383+ 0.0005 cm. The lower fluid boundary ranges over which spectral features such as coy
is a stationary brass ring which fills the fluid and co, are present change. For example, Fig. 1

1979 The American Physical Society 301

VOLUME 42, +UMBER 5 PHYSICAL REVIEW LETTERS 29 JANUARY 1979

IQ IO
R 23 Rc
= (a)
Q = 3.44 Hz IO . R=2~Rc
Q= 4.00 Hz
I" =so
I IQ I

IQ 2Q 0 5.0 IO.O
24) g I
02
(b)
R 26Rc
Q= 3.85 Hz
'
0 5.0 IO.O
IOz
(c)
—IO'
CL
R=25Rc
Q= 3.&0 Hz
iQ
IO
Q l2 0 5.0 I 0.0
f(Hz) IOz

FEG. 1. Power spectrum for which co~ and cu3 are IOo. R= 22 Rc
present as well as many harmonics and difference fre- Q= 3.25 Hz

quencies. Spectral peaks at the cylinder rotation speed I6'-

0 5.0 I 0.0
and its harmonics are artifacts of minute periodic IO 2
Ed( 24)) (e)
changes in capacitance between the rotating inner cyl- i

inder and the ion collector mounted in the outer cylin-

der. Unlike the other peaks they are present without
10--
0

lO
' '

Jj„g~l
1 3~,

50
I
R= l9Rc
Q= 2.85 Hz

the fluid. 0 I 0.0

f(Hz)
FEG. 2. Power spectral density (on a logarithmic
illustrates data taken for I" = 80 and R =23R„a scale) as a function of frequency. The abscissa scale
Reynolds number which is just beyond the aperi- has been normalized to the rotation speed for each
odic transition observed with I" =20. plot. co&=0.68Q appears with many harmonics at R
= 19',; as R is increased the higher harmonics and
Figure 2 shows data taken for the same aspect
finally the fundamental disappear from the power spec-
ratio F=80 and near the same Reynolds number
trum. j." = 80.
as that displayed in Fig. 1, but it represents a
different "state" of the system. When a bifurca-
tion in the flow leads to the appearance of ~, as olds numbers as great as 36R, . This new peak
R slowly increases, we find that the fundamental has a frequency which scales very closely (within
frequency first emerges followed by higher har- experimental resolution) with cylinder rotation
monics. Conversely, as the system approaches speed 0 [v„=(1.435+ 0.005)&I. The scale factor
R„ the higher harmonics decline in amplitude un- (1.435) is constant for all data taken with I =40
til at last the fundamental fades into the back- and I' =80 but appears to increase slightly (by
ground spectrum as shown in Fig. 2. Once the less than 1%) for I =28. (By contrast, the scale
system enters a particular state, that state seems factor varies slowly with ~ for peaks below the
to persist for an indefinite period of time as the aperiodic transition.
Reynolds number is varied slowly over a moder- Figure 3 presents a summary of the informa-
ate range. However, no mell-defined reliable tion above. The upper boundary of region I marks
procedure for initiating one or another of the the critical Reynolds number R, for the Taylor
"accessible states" of the system has been de- transition, ' the boundary above it represents the
termined. transition to time-dependent "wavy" modes ob-
Beyond the Reynolds number R, for which the served by Cole. ' Note that the Taylor transition
last sharp spectral peak fades from view is a re- is not much affected by aspect ratio unless there
gion of apparently disordered fluid motion, al- are only a very few cells; but the onset of azi-
though the faint outline of Taylor cells persists muthal (wavy) modes occurs at considerably high-
in visual experiments to extremely great R, and er R for short aspect ratio. End effects appear
the broad spectral feature denoted "B"in Figs. 2 to play a similar role in determining the upper
and 4 likewise persists for the largest Reynolds cutoff R, of the sharp azimuthal modes. As in-
numbers (R = 6VR, ) for which data have been tak- dicated previously, there are often several pos-
en. However, for sufficiently large aspect ratio sible alternative states accessible to the system
(I'~ 25), a new sharp spectral peak emerges for for a given R ~ 27', . The Reynolds number B,
Reynolds number R„»28K, and persists to Reyn- for which the sharp spectral peaks vanish as R
302
VOLUME 42, NUMBER 5 PHYSICAL REVIEW LETTERS 29 JANUARY 1979

40 10
Col lector
+ 39.
IO0
8
-I
IO
50-
IO

IO
Col lec tor
CL 0 +27
20- IO
CL
j
10
I
I
I
I I
IO I
I I

I I
"j':'('",
I".'(!@&&&&II(. I I

3.5"
'++++++4(Wjii$$
10 I
I
PI Col lector I
I
I

;;n o +24 I
I

IO I I

I ~ 398 '
I
i$&p I

"I
~ 278
I

0 Z&i(@42$mp~p&
IO
I

~ 24 8',
I

0 20 40 60 80 I
I
I 218 I

I I

IO
I

FIG. 8. Domains of various flow states for circular IO

couette flow: (I) two-dimensional laminar flow; (II) Collector
o +21
I =2e
three-dimensional time-independent Taylor flow; 10 - 8 38 48 R = 29Rc

(III) time-periodic and quasiperiodic azimuthal (wavy} A=43Hz

IO
modes; (IV) aperiodic flow; (V) reemergent peak, cu
= 1.440. Not illustrated is the domain of the broad 10 I

spectral peaks which appear at Reynolds numbers 0 5.0 10.0

(sometimes) as 11R and persist well beyond the top of f(Hz)

the diagram for all aspect ratios studied. FIG. 4. Power spectral densities recorded simul-
taneously at the four collectors illustrated on the right.
For this aspect ratio, I'=28, there is a clear depen-
increases depends on the initial state of the sys- dence of the amplitude of su& on the collector position.
tem. The dashed line denoting the upper limit of (Amplitude of co~ has also been observed. to decline for
the wavy-mode region (III) thus indicates the collector approaching the upper fluid boundary. )
maximum R for which any sharp spectral modes
have been observed. Similarly the dashed bound-
ary of the region (V) for which a sharp spectral upper liquid surface. Evidently, then, end ef-
peak reemerges denotes the limits within which fects play a significant role in short-aspect-ra-
that peak ~ay appear. Several experimental tio experiments.
runs with aspect ratios of 20, 28, 40, 56, and 80 The experimental results reported here leave
probed this reemergent region both by slowly 'in- many more questions to be answered. The re-
creasing the Reynolds number (R '&R/&t - 0.02%/ emergence of a sharp peak in the power spectrum
sec) between data sets and by starting at a very- after a range of Reynolds numbers of apparent
high Reynolds number (R -60R,) and slowly re- disorder has not been predicted and suggests that
ducing R. In some instances the reemergent region (IV) of Fig. 3 is really a region of "latent
peak was present throughout the entire range of "
order. If so, is the order still observable in
R indicated, but at other times (with ostensibly some fashion in the absence of sharp spectral
identical test protocol) it persisted through only peaks? Current models for the transition from
part of the range. laminar flow to turbulence' do not offer many
Figure 4, which displays power spectra for da- clues.
ta taken simultaneously at four different points, Comparison of our results with the observa-
provides a clue as to why the reemergent peak tions of Ahlers and Behringer' for Rayleigh-Be-
would not have been observed in an earlier ex- nard convection presents another enigma. Al. -
periment for which I"= —20.' Near the center of though a series of flow transitions similar to that
the fluid depth the peak is quite distinct, but clos- observed for circular couette flow is reported
er to the end of the annulus the amplitude de- for the Rayleigh-Benard experiment, changing
clines until it is no longer discernible above the aspect ratio has opposite effects for the two ex-
background spectrum. The same is true near the periments in determining the range over which

303
VOLUME 42, NUMBER 5 PHYSICAL REVIEW LETTERS 29 J&NU/RY 1979

time-independent flow and time-periodic flow 1978); J. P. Gollub and H. L. Swinney, Phys. Rev. Lett.
occur. In the limiting case of very large aspect 35, 927 (1975).
ratio for the convection experiment, only time- See, e.g. , J. B. Mc Laughlin and P. C. Martin, Phys.
dependent aperiodic flow is observed when con- Bev. A 12, 186 (1975); G. Ahlers and B. P. Behringer,
Phys. Rev. Lett. 40, 712 (1978); C. Normand, Y. Po-
vection is initiated, while for circular couette
meau, and M. G. Velarde, Bev. Mod. Phys. 49, 581
flow the greatest range of periodic states occurs (1977); J. P. Gollub, T. O. Brunner, and B. G. Danly,
for large aspect ratio. Science 200, 48 (1978).
This research has been supported by the Na- The ion technique utilizes naturally occurring impur-
tional Science Foundation Grant No. ENG-76- ity ions in the fluid; the ion probes and both cylinders
07354. J.
are held at ground potential. See B. Donnelly and
J.
D. Tanner, Proc. Boy. Soc. London, Ser. A 288,
:"0 (1965).
For a detailed description of the apparatus, see R. W.
Walden, Ph. D. thesis, University of Oregon, 1978 {un-
~ ~Current
address: Bell Laboratories, Murray Hill, published).
N. J. 07974. Fernstermacher et al. , Ref. 2.
~See, e.g. , D. Coles,J. Fluid Mech. 21, 885 {1965); J. A. Cole,J. Fluid. Mech. 75, 1 (1976).
H. A. Snyder, let. J. Non-Linear Mech. 5, 659 (1970). p. C. Martin, in Proceedings of the International
P. B. Fenstermacher, H. L. Swinney, S. V. Benson, Conference on Statistical Physics, Budapest, 1975, ed-
and J.P. Gollub, in Bifurcation Theory and Application ited by L. Pal and P. Szepfalusy (North-Holland, Am-
in Scientific LVisciplines, edited by O. Gurel and O. E. sterdam, 1975).
Bossier (New York Academy of Sciences, New York, Ahlers and Behringer, Bef. 8.

Doppler-Broadening Measurements of X-Ray Lines for Determination of

the Ion Temperature in Tokamak Plasmas
M. Bitter, S. von Goeler, R. Horton, M. Goldman, K. W. Hill, N. R. Sauthoff, and W. Stodiek
Plasma Physics Laboratory, Princeton University, Princeton, ¹zv Jersey 08540
(Received 3 November 1978)
Ion-temperature results are deduced from Doppler-broadening measurements of the K~
(1s-2P) resonance line emitted from heliumlike iron impurity ions in the hot central core of
PLT {Princeton Large Torus) tokamak discharges. The measurements were performed using
a high-resolution Bragg-crystal spectrometer with a multiwire proportional counter.
The larger tokamaks of the future require new difficult to interpret if the ion-velocity distribu-
techniques for the measurement of the ion tem- tion is not Maxwellian. This may occur, for ex-
perature (T, ) in the hot central core ot the plas- ample, during plasma heating by injection of in-
ma since the standard methods become more dif- tense neutral deuterium beams. In this paper
ficult with increasing plasma diameters and elec- we demonstrate that the inner-core ion tempera-
tron temperatures (T,). Neutral-charge-exchange ture of a tokamak plasma can be determined from
diagnostics' cannot be used for the determination Doppler-broadening measurements of suitable x-
of the central ion temperature when the mean ray lines emitted from the plasma.
free path for neutral charge exchange is much As T, increases, a larger fraction of the elec-
smaller than the plasma diameter. The line radi- tromagnetic radiation is emitted in the x-ray re-
ation in the vacuum ultraviolet (vuv) region com- gion. This radiation consists of a continuous
monly used for Doppler-broadening measure- bremsstrahlung spectrum and characteristic line
ments' of T; is emitted primarily from the edge spectra from high-atomic-number impurity ions,
of the plasma, at higher central electron tempera- which occur in different degrees of ionization or
tures. An exception is the recently observed' for- charge states. In a previous experiment' we
bidden line of Fe XX although this line will also used a germanium Bragg-crystal spectrometer
originate from outer regions of the plasma col- to investigate the Kn-line spectra of the charge
umn when the electron temperature exceeds 2 states of iron in Princeton Large Torus (PLT)
keV. An alternate method, the determination of discharges as a function of T, . The measured
T, from a measurement of the neutron yeild, 4 is line intensities and wavelengths were in agree-

304 1979 The American Physical Society