Analysis and Efficient Computation for Nonlinear Eigenvalue Problems in Quantum Physics and...

Post on 27-Dec-2015

220 views 2 download

Tags:

Transcript of Analysis and Efficient Computation for Nonlinear Eigenvalue Problems in Quantum Physics and...

Analysis and Efficient Computation for Nonlinear Eigenvalue Problems

in Quantum Physics and Chemistry

Weizhu Bao

Department of Mathematics& Center of Computational Science and Engineering

National University of SingaporeEmail: bao@math.nus.edu.sg

URL: http://www.math.nus.edu.sg/~bao

Collaborators: Fong Ying Lim (IHPC, Singapore), Yanzhi Zhang (FSU) Ming-Huang Chai (NUSHS); Yongyong Cai (NUS)

Outline

MotivationSingularly perturbed nonlinear eigenvalue problemsExistence, uniqueness & nonexistenceAsymptotic approximationsNumerical methods & resultsExtension to systemsConclusions

Motivation: NLS

The nonlinear Schrodinger (NLS) equation

– t : time & : spatial coordinate (d=1,2,3)– : complex-valued wave function– : real-valued external potential– : interaction constant

• =0: linear; >0: repulsive interaction • <0: attractive interaction

2 21( , ) ( ) | |

2ti x t V x

( R )dx

( , )x t

( )V x

0

4 ( 1)( . ., )sa Ne g

a

Motivation

In quantum physics & nonlinear optics: – Interaction between particles with quantum effect– Bose-Einstein condensation (BEC): bosons at low temperature

– Superfluids: liquid Helium,

– Propagation of laser beams, …….

In plasma physics; quantum chemistry; particle physics; biology; materials science (DFT, KS theory,…); ….

Conservation laws2 2 22

0 0

2 2 4

02

( ) : ( , ) ( ,0) ( ) : ( ) ( 1),

1( ) : ( , ) ( ) ( , ) ( , ) ( )

2

N x t d x x d x x d x N

E x t V x x t x t d x E

Motivation

Stationary states (ground & excited states)

Nonlinear eigenvalue problems: Find

Time-independent NLS or Gross-Pitaevskii equation (GPE):Eigenfunctions are– Orthogonal in linear case & Superposition is valid for dynamics!!– Not orthogonal in nonlinear case !!!! No superposition for dynamics!!!

2 2

2 2

1( ) ( ) ( ) ( ) | ( ) | ( ), R

2

( ) 0, ; : | (x) | 1

dx x V x x x x x

x x dx

( , ) s.t. ( , ) ( ) i tx t x e

Motivation

The eigenvalue is also called as chemical potential

– With energy

Special solutions– Soliton in 1D with attractive interaction– Vortex states in 2D

4( ) ( ) | (x) |2

E dx

2 2 41( ) [ | ( ) | ( )| ( ) | | ( ) | ]

2 2x V x x x dxE

( ) ( ) immx f r e

Motivation

Ground state: Non-convex minimization problem

– Euler-Lagrange equation Nonlinear eigenvalue problem

Theorem (Lieb, etc, PRA, 02’) – Existence d-dimensions (d=1,2,3):– Positive minimizer is unique in d-dimensions (d=1,2,3)!!– No minimizer in 3D (and 2D) when– Existence in 1D for both repulsive & attractive – Nonuniquness in attractive interaction – quantum phase transition!!!!

| |0 & lim ( )

xV x

( ) min ( ) | 1, | 0, ( )g xS

E E S E

cr0 ( 0)

Symmetry breaking in ground state

Attractive interaction with double-well potential2 2

2 2 2

1( ) ''( ) ( ) ( ) | ( ) | ( ), with | ( ) | 1

2

( ) ( ) & : positive 0 negative

x x V x x x x x dx

V x U x a

Motivation

Excited states:Open question: (Bao & W. Tang, JCP, 03’; Bao, F. Lim & Y. Zhang, Bull Int. Math, 06’)

Continuous normalized gradient flow:

– Mass conservation & energy diminishing

,,, 321

???????)()()(

)()()(

,,,

21

21

21

g

g

g

EEE

2 22

0 0

( (., ))1( , ) ( ) | | , 0,

2 || (., ) ||

( ,0) ( ) with || ( ) || 1.

t

tx t V x t

t

x x x

Singularly Perturbed NEP

For bounded with box potential for

– Singularly perturbed NEP

– Eigenvalue or chemical potential

– Leading asymptotics of the previous NEP

22 21

: , , | ( ) | 1x dx

22 2( ) ( ) | ( ) | ( ), ,

2

( ) 0,

x x x x x

x x

1

4

22 4

1( ) ( ) | (x) | (1)

2

1( ) | | (1), 0 1

2 2

E dx O

E dx O

( ) ( ) ( ) & ( ) ( ) ( ), 1O E E O

Singularly Perturbed NEP

For whole space with harmonic potential for

– Singularly perturbed NEP

– Eigenvalue or chemical potential

– Leading asymptotics of the previous NEP

21/ 2 / 4 1 /( 2) 2, ( ) ( ), , : | ( ) | 1d

d d dx x x x x dx

22 2( ) ( ) ( ) ( ) | ( ) | ( ),

2dx x V x x x x x

1

4

22 2 4

1( ) ( ) | (x) | (1)

2

1( ) ( ) | | | | (1), 0 1

2 2

d

d

E dx O

E V x dx O

1 1 /( 2) 1 /( 2)( ) ( ) ( ) ( ) & ( ) ( ) ( ), 1d d d dO O E E O

General Form of NEP

– Eigenvalue or chemical potential

– Energy

Three typical parameter regimes:– Linear: – Weakly interaction: – Strongly repulsive interaction:

22 2

2 2

( ) ( ) ( ) ( ) | ( ) | ( ), R2

( ) 0, ; : | ( ) | 1

dx x V x x x x x

x x x dx

4( ) ( ) | (x) |2

E dx

22 2 4( ) [ | ( ) | ( )| ( ) | | ( ) | ]

2 2x V x x x dxE

1& 0 1& | | 1

1& 0 1

Box Potential in 1D

The potential:The nonlinear eigenvalue problem

Case I: no interaction, i.e. – A complete set of orthonormal eigenfunctions

0, 0 1,( )

, otherwise.

xV x

22

12

0

( ) ( ) | ( ) | ( ), 0 1,2

(0) (1) 0 with | ( ) | 1

x x x x x

x dx

1& 0

2 21( ) 2 sin( ), , 1, 2,3,

2l lx l x l l

Box Potential in 1D

– Ground state & its energy:

– j-th-excited state & its energy

Case II: weakly interacting regime, i.e.– Ground state & its energy:

– j-th-excited state & its energy

20 0 0( ) ( ) 2 sin( ), : ( ) : ( )

2g g g g g gx x x E E

2 20 0 0( 1)

( ) ( ) 2 sin(( 1) ), : ( ) : ( )2j j j j j j

jx x j x E E

1& | | (1)o

2 20 0 03

( ) ( ) 2 sin( ), : ( ) ( ) , : ( ) ( ) 32 2 2g g g g g g g gx x x E E E

2 20 0

2 20

( 1) 3( ) ( ) 2 sin(( 1) ), : ( ) ( ) ,

2 2

( 1): ( ) ( ) 3

2

j j j j j

j j j

jx x j x E E E

j

Box Potential in 1D

Case III: Strongly interacting regime, i.e.– Thomas-Fermi approximation, i.e. drop the diffusion term

• Boundary condition is NOT satisfied, i.e. • Boundary layer near the boundary

1& 0 1

TF TF TF 2 TF TF TF

1TF 2

0

TF TF TFg g g

( ) | ( ) | ( ), 0 1, ( )

| ( ) | 1

1 (x) ( ) 1, E E , 1,

2

g g g g g g

g

g g g

x x x x x

x dx

x

TF TF(0) (1) 1 0g g

Box Potential in 1D

– Matched asymptotic approximation• Consider near x=0, rescale• We get

• The inner solution

• Matched asymptotic approximation for ground state

, ( ) ( )g

g

x X x x

31( ) ( ) ( ), 0 ; (0) 0, lim ( ) 1

2 XX X X X X

( ) tanh( ), 0 ( ) tanh( ), 0 (1)g

g gX X X x x x o

MA MA MA

MA MA

1MA 2 MA 2 2 TF 2 2

0

( ) ( ) tanh( ) tanh( (1 )) tanh( ) , 0 1

1 | ( ) | 1 2 1 2 2 1 2 , 0 1.

g g g

g g g

g g g g

x x x x x

x dx

Box Potential in 1D

• Approximate energy

• Asymptotic ratios:

• Width of the boundary layer:

MA 2 21 41 2

2 3g gE E

( )O

0

1lim ,

2g

g

E

Box Potential in 1D

• Matched asymptotic approximation for excited states

• Approximate chemical potential & energy

• Boundary layers • Interior layers

MA[( 1) / 2]MA MA

0

MA MA[ / 2]

0

2( ) ( ) [ tanh( ( ))

1

2 1tanh( ( )) tanh( )]

1

jg

j j jl

jg g

jl

lx x x

j

lx C

j

MA 2 2 2 2

MA 2 2 2 2

1 2( 1) 1 ( 1) 2( 1) ,

1 4( 1) 1 ( 1) 2( 1) ,

2 3

j j

j j

j j j

E E j j j

( )O

Harmonic Oscillator Potential in 1D

The potential:The nonlinear eigenvalue problem

Case I: no interaction, i.e. – A complete set of orthonormal eigenfunctions

2

( )2

xV x

22 2( ) ( ) ( ) ( ) | ( ) | ( ), with | ( ) | 1

2x x V x x x x x dx

1& 0

2

2

2

1/ 2 / 21/ 4

20 1 2

1 1( ) (2 !) ( ), , 0,1,2,3,

2

( ) ( 1) : Hermite polynomials with

( ) 1, ( ) 2 , ( ) 4 2,

l xl l l

l xl x

l l

lx l e H x l

d eH x e

dx

H x H x x H x x

Harmonic Oscillator Potential in 1D

– Ground state & its energy:

– j-th-excited state & its energy

Case II: weakly interacting regime, i.e.– Ground state & its energy:

– j-th-excited state & its energy

20 / 2 0 01/ 4

1 1( ) ( ) , : ( ) : ( )

2x

g g g g g gx x e E E

20 1/ 2 / 2 0 00 01/ 4

1 ( 1)( ) ( ) (2 !) ( ), : ( ) : ( )

2j x

j j j j j j j

jx x j e H x E E

1& | | (1)o

20 / 2 0 00 01/ 4

1 1 1( ) ( ) , : ( ) ( ) , : ( ) ( )

2 2 2x

g g g g g g g gx x e E E E C C

0 0

0 0 4j j

-

( 1)( ) ( ), : ( ) ( ) ,

2 2

( 1): ( ) ( ) with C = | ( ) |

2

j j j j j j

j j j j

jx x E E E C

jC x dx

Harmonic Oscillator Potential in 1D

Case III: Strongly interacting regime, i.e.– Thomas-Fermi approximation, i.e. drop the diffusion term

– No boundary and interior layer– It is NOT differentiable at

1& 0 1

TF 2 TFTF TF TF TF 2 TF TF

TF 3/ 2TF 2 TF 2/3

-

/ 2, | | 2( ) ( ) ( ) | ( ) | ( ) ( )

0, otherwise

2(2 ) 1 3 1 | ( ) | ( )

3 2 2

g gg g g g g g

gg g g

x xx V x x x x x

x dx

TF2 gx

Harmonic Oscillator Potential in 1D

– Thomas-Fermi approximation for first excited state

• Jump at x=0!• Interior layer at x=0

TF TF TF TF 2 TF1 1 1 1 1

TF 2 TFTF 1 1

1

TF 3/ 2TF 2 TF 2/31

1 1 1

-

( ) ( ) ( ) | ( ) | ( )

sign( ) / 2, 0 | | 2( )

0, otherwise

2(2 ) 1 3 1 | ( ) | ( )

3 2 2

x V x x x x

x x xx

x dx

Harmonic Oscillator Potential in 1D

– Matched asymptotic approximation

– Width of interior layer:

MA1MA MA

1 1MA MA 2 MA1 1 1

| |tanh( ) 0 | | 2

( ) 2 / 2

0 otherwise

x xx x

x x

( )O

Thomas-Fermi (or semiclassical) limit

In 1D with strongly repulsive interaction– Box potential

– Harmonic potential

In 1D with strongly attractive interaction

0

1 ??? ??? : ( ) ???g g g gE

0 1,11 0 1 1( ) 1

0 0,1 2g g g g

xx W E

x

0 2 00

0 2/3 0 2/3

/ 2, | | 2( ) ( ) [0,0.5)

0, otherwise

3 3 1 3( ) , ( )

10 2 2 2

g gg g

g g g g

x xx x C

E E

1 0 1/2 2

0

0

( ) ( )

( ) ( )g g g gx x x L E

V x V x x

Numerical methods

Runge-Kutta method: (M. Edwards and K. Burnett, Phys. Rev. A, 95’)

Analytical expansion: (R. Dodd, J. Res. Natl. Inst. Stan., 96’)

Explicit imaginary time method: (S. Succi, M.P. Tosi et. al., PRE, 00’)

Minimizing by FEM: (Bao & W. Tang, JCP, 02’)

Normalized gradient flow: (Bao & Q. Du, SIAM Sci. Comput., 03’)

– Backward-Euler + finite difference (BEFD)– Time-splitting spectral method (TSSP)

Gauss-Seidel iteration method: (W.W. Lin et al., JCP, 05’) Continuation method: W. W. Lin, etc., C. S. Chien, etc

( )E

Imaginary time method

Idea: Steepest decent method + Projection

– The first equation can be viewed as choosing in GPE– For linear case: (Bao & Q. Du, SIAM Sci. Comput., 03’)

– For nonlinear case with small time step, CNGF

22 2

1

11

1

0 0

( )1( , ) ( ) | | ,

2 2

( , )( , ) , 0,1,2,

|| ( , ) ||

( ,0) (x) with || ( ) || 1.

t n n

nn

n

Ex t V x t t t

xx t n

x

x x

tt

1( (., ) ) ( (., ) ) ( (., 0) )n nE t E t E

it

0

1

2 1̂

??)()(

)()ˆ(

)()ˆ(

01

11

01

EE

EE

EE

g

Normalized gradient glow

Idea: letting time step go to 0 (Bao & Q. Du, SIAM Sci. Comput., 03’)

– Energy diminishing

– Numerical Discretizations• BEFD: Energy diminishing & monotone (Bao & Q. Du, SIAM Sci. Comput., 03’)

• TSSP: Spectral accurate with splitting error (Bao & Q. Du, SIAM Sci. Comput., 03’)

• BESP: Spectral accuracy in space & stable (Bao, I. Chern & F. Lim, JCP, 06’)

• Uniformly convergent method (Bao&Chai, Comm. Comput. Phys, 07’)

22 2

2

0 0

( (., ))( , ) ( ) | | , 0,

2 || (., ) ||

( ,0) ( ) with || ( ) || 1.

t

tx t V x t

t

x x x

0|| (., ) || || || 1, ( (., )) 0, 0d

t E t td t

Ground states

Numerical results (Bao&W. Tang, JCP, 03’; Bao, F. Lim & Y. Zhang, 06’)

– Box potential• 1D-states 1D-energy 2D-surface 2D-contour

– Harmonic oscillator potential:

• 1D 2D-surface 2D-contour – Optical lattice potential:

• 1D 2D-surface 2D-contour 3D next

otherwise;100)( xxV

2/xV(x) 2

2 2( ) / 2 12sin (4 )V x x x

back

back

back

back

back

back

back

back

back

back

back

Extension to rotating BEC

BEC in rotation frame(Bao, H. Wang&P. Markowich,Comm. Math. Sci., 04’)

Ground state: existence & uniqueness, quantized vortex

– In 2D: In a rotational frame &With a fast rotation & optical lattice

– In 3D: With a fast rotationnext

2 2

2 2

1( ) [ ( ) | | ] ,

2

: | (x) | 1d

dzx V x L x

dx

: ( ) , ,z y x y xL xp yp i x y i L x P P i

: ( ) min ( ), | 1, ( )g gS

E E E S E

back

back

back

back

Extension to two-component

Two-component (Bao, MMS, 04’)

Ground state

– Existence & uniqueness– Quantized vortices & fractional index– Numerical methods & results: Crarter & domain wall

2 2 21 1 11 1 12 2 1

2 2 22 2 21 1 22 2 2

2 22 21 1 2 2

1( ) [ ( ) | | | | ]

21

( ) [ ( ) | | | | ]2

| ( ) | , | ( ) | 1 0 1d d

z

z

x V x L

x V x L

x dx x dx

1 2 1 2: ( ) min ( ), ( , ) | , 1 , ( )g gS

E E E S E

Results

Theorem – Assumptions

• No rotation & Confining potential• Repulsive interaction

– Results• Existence & Positive minimizer is unique

– No minimizer in 3D when

Nonuniquness in attractive interaction in 1D Quantum phase transition in rotating frame

| |lim ( )x

V x

11 220 or 0

211 12 22 11 11 22 12, , 0 or 0 & 0

0

Two-component with an external driving field

Two-component (Bao & Cai, 09’)

Ground state

– Existence & uniqueness (Bao & Cai, 09’)

– Limiting behavior & Numerical methods – Numerical results: Crarter & domain wall

2 2 21 11 1 12 2 1 2

2 2 22 21 1 22 2 2 1

2 2 2 21 2 1 2

1( ) [ ( ) | | | | ]

21

( ) [ ( ) | | | | ]2

| ( ) | | ( ) | 1d

z

z

x V x L

x V x L

x x dx

1 2: ( ) min ( ), ( , ) | 1, ( )g g SE E E S E

Theorem (Bao & Cai, 09’)

– No rotation & confining potential &

– Existence of ground state!! – Uniqueness in the form under

– At least two different ground states under– quantum phase transition

– Limiting behavior

211 11 22 12 12 11 220 & 0 or & 0

12 11 22 0 0 00 & 0 & ( , ) for 0

211 12 22 11 11 22 12, , 0 or 0 & 0

1 2(| |, sign( ) | |)g gg

1 2

1 2

1 2

| | | | & | |

| | 0 & | |

| | & | | 0

g g g

g g g

g g g

Extension to spin-1

Spin-1 BEC (Bao & Wang, SINUM, 07’; Bao & Lim, SISC 08’, PRE 08’)

– Continuous normalized gradient flow (Bao & Wang, SINUM, 07’)

– Normalized gradient flow (Bao & Lim, SISC 08’)• Gradient flow + third projection relation

2 * 21 1 1 0 1 1 1 0

2 *0 0 1 1 0 1 1 0

2 * 21 1 1 0 1 1 1 0

2 2 2 2 2 21 0 1 1 0 1

1( ) [ ( ) ] ( )

21

2 [ ( ) ] ( ) 22

1( ) [ ( ) ] ( )

2

[| ( ) | ( ) | ( ) | ]

n s s

n s s

n s s

V x g g g

V x g g g

V x g g g

x x x

2 2 2 21 1 1 1

1,

[| ( ) | | ( ) | ] ( 1 1)

d

d

dx

x x dx M M

Quantum phase transition

Ferromagnetic gs <0 Antiferromagnetic gs > 0

Dipolar Quantum Gas

Experimental setup – Molecules meet to form dipoles – Cool down dipoles to ultracold – Hold in a magnetic trap – Dipolar condensation – Degenerate dipolar quantum gas

Experimental realization– Chroimum (Cr52)– 2005@Univ. Stuttgart, Germany– PRL, 94 (2005) 160401

Big-wave in theoretical studyA. Griesmaier,et al., PRL, 94 (2005)160401

Mathematical Model

Gross-Pitaevskii equation (re-scaled)

– Trap potential– Interaction constants– Long-range dipole-dipole interaction kernel

References:– L. Santos, et al. PRL 85 (2000), 1791-1797– S. Yi & L. You, PRA 61 (2001), 041604(R); D. H. J. O’Dell, PRL 92 (2004), 250401

2 2ext dip

1( , ) ( ) | | | | ( , )

2i x t V x U x tt

3( , )x t x

2 2 2 2 2 2ext

1( )

2 x y zV z x y z 2

0 dip

20 0

4 (short-range), (long-range)

3s

mNN a

a a

2 2 23

dip 3 3

3 1 3( ) / | | 3 1 3cos ( )( ) , fixed & satisfies | | 1

4 | | 4 | |

n x xU x n n

x x

Mathematical Model

Mass conservation (Normalization condition)

Energy conservation

Long-range interaction kernel:– It is highly singular near the origin !! At singularity near the origin !! – Its Fourier transform reads

• No limit near origin in phase space !! • Bounded & no limit at far field too !!• Physicists simply drop the second singular term in phase space near origin!!• Locking phenomena in computation !!

3 3

2 22( ) : ( , ) ( , ) ( ,0) 1N t t x t d x x d x

3

2 2 4 2 2ext dip 0

1( ( , )) : | | ( ) | | | | ( | | ) | | ( )

2 2 2E t V x U d x E

23

dip 2

3( )( ) 1

| |

nU

3

1

| |O

x

A New Formulation

Using the identity (O’Dell et al., PRL 92 (2004), 250401, Parker et al., PRA 79 (2009), 013617)

Dipole-dipole interaction becomes

Gross-Pitaevskii-Poisson type equations (Bao,Cai & Wang, JCP, 10’)

Energy

2ext

2

| |

1( , ) ( ) ( ) | | 3 ( , )

2

( , ) | ( , ) | , lim ( , ) 0

n n

x

i x t V x x tt

x t x t x t

2 2

dip dip3 2 2

3 3( ) 1 3( )( ) 1 ( ) 3 ( ) 1

4 4 | |n n

n x nU x x U

r r r

2 2 2 2dip

1| | | | 3 & | | | |

4n nUr

3

2 2 4 2ext

1 3( ( , )) : | | ( ) | | | | | |

2 2 2 nE t V x d x

| | & & ( )n n n n nr x n

Ground State Results

Theorem (Existence, uniqueness & nonexistence) (Bao, Cai & Wang, JCP, 10’) – Assumptions

– Results• There exists a ground state if • Positive ground state is uniqueness

• Nonexistence of ground state, i.e. – Case I: – Case II:

3ext ext

| |( ) 0, & lim ( ) (confinement potential)

xV x x V x

g S 0 &2

00| | with i

g ge

lim ( )SE

0 0 & or

2

Conclusions

Analytical study– Leading asymptotics of energy and chemical potential– Existence, uniqueness & quantum phase transition!!– Thomas-Fermi approximation– Matched asymptotic approximation– Boundary & interior layers and their widths

Numerical study– Normalized gradient flow– Numerical results

Extension to rotating, multi-component, spin-1, dipolar cases.