Vibrational Levels Asso

iated with Hydrogen Bonds and

Semi lassi al Hamiltonian Normal Forms
George A. Hagedorn and Alain Joye

Dedi ated to Jean-Mi hel Combes, in elebration of his 65th birthday

Abstra t. We des ribe and extend our re ent proposal to model mathemat-
i ally the vibrational levels asso iated with Hydrogen bonds in symmetri
tri-atomi mole ules. Our approa h is based on modi ation of the usual
Born-Oppenheimer approximation to take into a ount the lighter mass of the
Hydrogen nu leus and the weakness of the Hydrogen bond, using spe ial fea-
tures of the ele tron energy level surfa e asso iated with the Hydrogen bond.
Negle ting bending of the mole ule for simpli ity, we a hieve this by s aling
the mass of the Hydrogen atoms dierently from the heavier atoms, and by
using a modied form for the ele troni energy surfa e.
As a result, anharmoni ee ts play a role at leading order in the limit
where the nu lear masses go to innity. Our analysis is based on lose exam-
ination of the numeri al data available for the ground state energy surfa e of
the F HF ion, and we make a omparison with experimental data for the
vibrational levels of that ion.
The theory we propose is, however, quite general and an a omodate
asymmetri tri-atomi mole ules. Moreover, we provide an extension of our
results to mole ules with nu lei of several dierent spe ies, where we assume
that ea h of the masses s ales dierently. Considering an adapted ground
state energy surfa e, we ompute the leading term of the orresponding vibra-
tional levels in the limit of large nu lear masses by means of a normal form
Hamiltonian.

1. Introdu tion

The standard time{independent Born{Oppenheimer approximation [ ℄ takes ad- 1

vantage of the large masses of the nu lei relative to the mass of an ele tron. It
allows one to ompute the low-lying vibrational states of the nu lear motion from
knowledge of the ground state ele tron energy level surfa e near its minimum un-
der the assumption that the minimum is non-degenerate. To leading order, the
vibrational energy levels are those of a harmoni os illator asso iated with the
non-degenerate minimum.
1991 Mathemati s Subje t Classi ation. 81V55.
Key words and phrases. Hydrogen bonds, vibrational levels, Born-Oppenheimer approxima-
tion, semi lassi al normal forms.
Partially Supported by National S ien e Foundation Grants DMS{0303586 and DMS{
0600944.
1
2 GEORGE A. HAGEDORN AND ALAIN JOYE

More pre isely, onsider a mole ular system whose nu lear onguration spa e
modulo rotations is Rd , and denote the ratio of mean of the nu lear masses to the
ele tron mass be  . For small , there exists a mole ular energy level E () that
4

satises,
 
(1.1) E () = E +  (n + 12 ) ! +


(nd + 21 ) !d + O( );
0
2
1 1
4

where E is the minimum of the ground state ele troni energy level. The numbers
0
nj 2 N , and !j > 0, for j = 1;


; d, hara terize the d-dimensional harmoni
os illator. Under quite general hypotheses, (1.1) represents the rst few terms in a
omplete asymptoti expansion of E () in powers of . See the re ent review [ ℄ for
4
more details and results.
For many mole ules, this approximation is very su essful, but it may fail to
give a urate results when applied to some mole ules that ontain Hydrogen bonds.
The binding energy of Hydrogen bonds is typi ally very small, and the mass of the
Hydrogen nu leus is an order of magnitude smaller than that of other nu lei su h
as Carbon. Moreover, the experimental vibrational spe tra of symmetri tri-atomi
mole ules with Hydrogen bonds display signi ant deviations from the approximate
harmoni spe trum (1.1). See [ ℄ for an a ount of these properties.
2
In [ ℄ we propose an alternative to the Born{Oppenheimer approximation that
3
is spe i ally designed to des ribe mole ules with symmetri al Hydrogen bonds.
There are two main dieren es between our approa h and the standard Born-
Oppenheimer analysis. First, we s ale the masses of the Hydrogen nu lei as  3

while keeping the heavier nu lei s ale as  . Although our analysis applies only
4

for small values of , the mass of a Hydrogen nu leus is 1:015  if  is dened by

3

taking the mass of a Carbon atom to be  . This value of  is roughly  ' 0:082.
4

Se ond, we model the ele tron energy level surfa e in a spe ial way that depends on
. This takes into a ount the smallness of a oeÆ ient in the harmoni potential
asso iated with the Hydrogen bond. That oeÆ ient is weaker than the others,
roughly by a fa tor . Consequently, in our model, anharmoni ee ts ontribute
to the leading order al ulations of vibrational levels, as  ! 0.
Our approa h gives rise to an {independent normal form Hamiltonian HNF
(des ribed below) whi h is not a harmoni os illator. This Hamiltonian des ribes
the vibrational levels: For small , there exists a mole ular level E () that satises
(1.2) E () = E +  E + O(  );
0
2
2
2+

for some  > 0, where E is an element of the spe trum of HNF . Note that the
2
index 2 refers here to the power of  appearing as the leading term. Again, under
appropriate hypotheses, our analysis yields a omplete expansion of E () in powers
of  = .
1 2

Our model leads to a dierent expansion from the usual Born{Oppenheimer

approximation. For Hydrogen nu lei not involved in Hydrogen bonding, the vi-
brational energies are of order  = , while the vibrational energies for the other
3 2

nu lei and the Hydrogen nu lei involved in the symmetri Hydrogen bonding are
of order  . Furthermore, anharmoni ee ts must be taken into a ount for a
2

Hydrogen nu leus involved in Hydrogen bonding at the leading order,  . In the

2

standard Born{Oppenheimer model, all vibrational energies appear at order  and 2

are harmoni at that order. Anharmoni orre tions rst enter at order  . 4
 HYDROGEN BONDS AND SEMICLASSICAL NORMAL FORMS 3

Our analysis is motivated by an examination of symmetri bihalide ions, su h

as F HF or ClHCl , and we work under the simplifying hypothesis that bending
of the mole ule is absent. Although symmetri bihalide ions are quite spe ial, our
approa h is exible enough to des ribe more general phenomena. For example, the
lowest ele tron energy surfa e for F HF has a single minimum with the Hydrogen
nu leus mid{way between the two Fluorines. Our general theory an handle situa-
tions with single or losely spa ed double wells in the oordinates for a Hydrogen
nu leus that parti ipates in Hydrogen bonding.
In this note, we further generalize our approa h to mole ules with nu lei of
several dierent spe ies, where we assume that ea h of the masses s ales with a
dierent inverse power of . We allow the ground state energy surfa e to depend on
 in a similar fashion as in the tri-atomi ase, and we determine a orresponding
normal form HNF . Then, we prove that the leading term of the vibrational levels
E () is des ribed by HNF :
(1.3) E () =  E + O(  ); +

for small , where

E 2 (HNF );  > 0; and  > 0:
Here, the index  refers to the power of  in the leading term.
From a mathemati al point view, our analysis onsists of onstru ting and
studying quasimodes for the full mole ular Hamiltonian in the singular limit  ! 0.
To do this, we use a multis ale analysis te hnique. We identify dierent relevant
length s ales in the system and treat them independently in the rst stage of the
analysis. This leads to a formal algorithm to onstru t an Ansatz for the quasimode
in more variables. We then prove that this yields a bona de quasimode for the
original problem by re-introdu ing the oupling between the length s ales. This is
a traditional, su essful route for getting information on the vibrational spe trum
of the mole ule. See e.g., [ ℄. In the present ir umstan es, when using the same
4
strategy, we fa e a new problem: The spe trum of the normal form Hamiltonian
HNF is not known expli itly, and we have to show that the eigenfun tions of HNF
and their derivatives have suÆ ent de ay properties. For the harmoni os illator of
the standard Born{Oppenheimer theory, everything required is well-known.
2. The Tri{Atomi Model

We onsider a mole ule that onsists of two types of atoms. Two identi al
nu lei are heavier and have masses  , where the mass of an ele tron is 1. The
4

third nu leus is lighter and has mass  . We onsider a simplied model in whi h
3

bending and rotational motions are absent, i.e., all three nu lei lie on a xed line.
After passing to Ja obi oordinates, dis arding kineti energy of the enter of mass,
and res aling by onstant numeri al fa tors (See [ ℄), the mole ular Hamiltonian
3
has the form,
4
  3
Hmol () =
2 x1 2 x2 + h(x ; x );
2 2
1 2

where x denotes the distan e between the two identi al nu lei, and x is the
1 2
dispa ement of the lighter nu leus from the enter of mass of the identi al nu lei.
Further te hni al hypotheses are provided below in Assumption 3.1.
The ele troni Hamiltonian h(x ; x ) depends parametri ally on the positions
1 2
of the nu lei. It is a self-adjoint operator on the ele troni Hilbert spa e Hel . We
4 GEORGE A. HAGEDORN AND ALAIN JOYE

assume h(x ; x ) has a non-degenerate ground state ele troni surfa e E (x ; x )

1 2 1 2
with orresponding normalized eigenve tor (x ; x ). We further assume that the 1 2
mole ule a tually binds, in the sense that that there exists a nu lear onguration
(xeq ; xeq ) that is an isolated minimum of E (x ; x ). We assume xeq 6= 0. (The 1 2
heavy nu lei are not at the same point!) In the symmetri al ase whi h we study
1 2 1

here, the equilibrium position of the lighter nu leus lies at the enter of mass of the
heavy nu lei (xeq = 0). The non-symmetri al ase would have xeq 6= 0.
Thus far, we have only altered the standard Born-Oppenheimer approximation
2 2

by introdu ing two nu lear mass s ales, instead of one. We now make a se ond
modi ation to re e t the stru ture of the ele tron energy level surfa e asso iated
with Hydrogen bonding. (See [ ℄.) We introdu e an arti ial dependen e eqin the
2
ground state ele troni surfa e on . After a shift of the variable x by x , we 1
assume that the ground state has an expansion of the form
1

(2.1) E (; x ; x ) = E + a x + a  x a x x + a x + O(x1 x 2 );

1 2 0 1
2
1 2
2
2 3 1
2
2 4
4
2 1
2
2

where  ;  2 N and  +   3. The Taylor series only ontains even powers

1 2 1 2
of x , due to the symmetry of the mole ule. The fa tor  in the harmoni term
2
a  x expresses the weaker binding of the Hydrogen atom. The smallness of this
2
2

oeÆ ent makes it ne essary to onsider terms of order higher than quadrati in the
2

expansion. We assume that a ; a and a are positive, but a an take any eqvalue.
1 3 4 2
When a is negative, E (; x ; x ) has a losely spa ed double well near (x ; 0),
2 1 2
instead of a single lo al minimum. Note that a = 0 leads to two de oupled one-
1
3
dimensional systems, to leading order. We dene then a s alar normal form by
keeping the leading order terms only in (2.1),
(2.2) 2 x1 2 x1 + E + a x + a  x a x x + a x :
4 3
2 2 2 2 2 4
0 1 1 2 2 3 1 2 4 2

The S hrodinger operator (2.2) plays the r^ole the Harmoni os illator plays in
the standard Born-Oppenheimer approximation. By res aling with (y ; y ) = 1 2
(x =; x = = ), we see that (2.2) minus the onstant term E is unitarily equivalent
1 2
1 2
0
to  times the normal form Hamiltonian HNF ,
2

(2.3) HNF =
1 1
2 y1 2 y2 + ENF (y ; y )
2 2
1 2

with
ENF (y ; y ) = a y + a y 1 a y y +a y:
2 1
2
1 2
2
2 3 1
2
2 4
4
2

We further assume that the oeÆ ents faj g are su h that the normal form potential
ENF is bounded below. More pre isely, we require
2.1
Assumption .

a ; a ; a > 0; 1 3 4

and either
a < 4a a ; 2
3 1 4

or
a = 4 a a and a  0: 2
3 1 4 2
 HYDROGEN BONDS AND SEMICLASSICAL NORMAL FORMS 5

Anti ipating the result of the analysis below, the normal form HNF (2.3) yields
the leading term asymptoti s (1.2) for the vibrational levels.
The spe ial form (2.1) is motivated by numeri al data for the mole ule F HF .
By tting numeri al data for F HF with the value  = 0:0821, we get the following
quantities, where distan es are expressed in Angstroms and energies in Hartrees.
xeq = 2:287; E = 200:215;
1 0

(2.4) a = 0:26; a = 1:22; a = 1:29; a = 1:62 :

1 2 3 4

The experimentally observed values [ ℄ for the symmetri stret h, asymmetri

2
stret h, and ombined vibrational vibrational modes are, respe tively,
583:05 m ; 1331:15 m ; and 1849 m :
1 1 1

Note that the sum of the rst two eigenvalues is signi antly larger than the third
one. This is an expression of anharmoni ity.
By omparison, if we use the data (2.4) in HNF , and numeri ally determine
the vibrational levels using the rst two terms of (1.2) with  = 0:0821, i.e.
E () ' E +  E ; with E 2 (HNF );
0
2
2 2

our model predi ts the following values for these vibrational modes:
600 m ; 1399 m ; and 1942 m :
1 1 1

Besides the relatively good agreement with the the experimental data, we also
point out that the anharmoni ee t mentioned above is present in our leading
term omputation.
Standard Born{Oppenheimer based software yields orresponding harmoni
modes
609 m ; 1118 m ; and 1727 m :
1 1 1

En ouraged by this positive numeri al test, we turn to the mathemati al proof

of assertion (1.2). It is a spe ial ase of the generalization (1.3), whi h we des ribe
in the next se tion.
3. More mass s ales

The tri{atomi model presented in [ ℄ and its mathemati al analysis general-

3
ize naturally. Roughly speaking, we an onsider mole ules with nu lei of several
dierent spe ies, where we assume that ea h of the masses s ales with a distin t
inverse power of . The orresponding ground state energy surfa e depends on  in
a similar way to the tri-atomi ase.
We adopt this more general point of view here and prove the leading order
result (1.3) whi h generalizes (1.2). More pre isely, we work under the following
general assumption.
Assumption 3.1 Let Hmol() be a self-adjoint mole ular Hamiltonian dened
.
on a domain of L (Rn ; Hel ) by
2

n
X j
(3.1) Hmol () = 2 x2j + h(; x); 1  2 


 n > 0:
j =1
6 GEORGE A. HAGEDORN AND ALAIN JOYE

Here, x 2 Rn , and Hel is the ele troni Hilbert spa e. The ele tron Hamiltonian
h(; x) is self-adjoint on an x{independent dense domain D  Hel . We assume
h(; x) is bounded below, uniformly for x 2 Rn . Moreover, h(; x) is C 1 in the
strong resolvent sense and admits a non-degenerate {dependent ground state eigen-
value V (; x) with orresponding normalized eigenstate (x), independent of . We
normalize the ground state energy surfa e so that
(3.2) V (; 0) = 0:
We denote the formal Taylor expansion of V (; x) by
X
V (; x)  v () x ;  = ( ;


; n ) 2 N n ;
1 v () 2 R;
6 2Nn
0=

where the oeÆ ients v () are regular as  ! 0. We impose more restri tions on
these oeÆ ients below.
The rst step in analyzing the spe trum of (3.1) onsists of dening a s alar
normal form Hamiltonian HNF . Let  2 (R )n be a set of positive exponents.
+

We onsider the kineti energy and introdu e res aled variables (y ;


; yn) = 1
(x =1 ;


; xn=n ), whi h we denote by y = x=.
1
n
X j 
2 x2j =
2
y ; for some  > 0:
j =1
This enfor es the relations
(3.3) j 2 j = ; 8 j = 1;


; n:
To determine the normal form potential VNF , we onsider the formal expression
X
(3.4)   V (;  y)  v () ( = )
  jj y ; 1 2 ( +2)

6 2Nn
0=

where we use the notation 
 = Pnj j j and jj = Pnj j . As in the

=1 =1
tri-atomi ase, we want VNF to onsist of the  terms on the right hand side0

of (3.4), and we want therep to be only nitely may su h terms. All other terms
should ontain fa tors of  with p positive. To make this more pre ise, we make
the following assumption:
Assumption3.2 There exist  > 0 and two non-empty nite sets of multi-
.
indi es,   Nn n f0g and F  Nn n  , su h that for small jxj and ,
0 0
 
 
 X  X  
(3.5) V
 (; x) v () 
x    x 
  2 0   2 F
for some > 0. Furthermore,
8
< v () ( = )
  j j
1 2 ( +2)
=  ; 8 2 0
(3.6) :

  (j j + 2) > 0; 8  2 F :
3.3 We address the question of existen e of a suitable  and the
Remark .
orresponding sets  and F for a given general potential at the end of this se tion.
0
 HYDROGEN BONDS AND SEMICLASSICAL NORMAL FORMS 7

Remark 3.4 The ase of the tri-atomi mole ule in R is hara terized by the
.
2

parameters  = 4;  = 3;  = 1;  = 1=2;  = 2; and the sets

1 2 1 2

 = f(2; 0); (0; 2); (1; 2); (0; 4)g;

0

F = f(3; 0); (2; 2); (1; 4); (0; 6)g;

for whi h 
  (jj + 2) = 2 for all  2 F .
Remark 3.5 For any 0 <  < n, the last ondition in (3.6) is always satised
.
for large enough jj, sin e by (3.1) and (3.3),

  (j j + 2)  (n )) j j 2  = 2 (nj j ):
The normal form potential is then dened as
X
(3.7) VNF (y) =  y  ;
 2 0
so that the s alar normal form Hamiltonian be omes
HNF =
1
+ VNF :
2
Taking into a ount our normalization (3.2), we expe t the spe trum of Hmol() to
ontain elements of the form
(3.8) E () =  E + O(  ); +

for small , where

(3.9) E 2 (HNF ) and  > 0:
The index  refers to the power of  in the leading term.
Let us brie y onsider the existen e of a suitable normal form for a given smooth
potential of the form (3.4) and, in parti ular, the denition of the exponent  > 0.
Consider the usual situation in whi h at least one monomial in the Taylor expansion
(3.4) whose oeÆ ient is independent of  appears in the normal form. Su h terms
must dene the exponent , be ause they ontain no extra  dependen e, and after
res aling, their ontribution to the normal form annot diverge. Thus, we introdu e
the set
N = f 2 N n : v () =  6= 0g:
The dimensions orresponding to the minimal value n of the exponents j play
a spe ial r^ole. Let 0  k  n be the integer su h that n = n =


n k <
1
n k . We dene
1

k = f 2 N n :  = (0;


; 0; n k ; n k ;


; n )  (0;


; 0;  00 )g:

+1

Similarly, we write x00 = (0;


; 0; xn k ; xn k ;


; xn ). 00For simpli ity, we as-

+1

sume that there exists a non-zero term of the form x00 in (3.4) with an -
independent oeÆ ient. Note, however, that it is possible to prove results similar
to the following one in ase this hypothesis does not hold.
The next proposition gives onditions on the {dependen e of the oeÆ ients
v () (with  62 N ), ensuring that Assumption 3.2 holds, with an exponent 
dened by means of the indi es of the {independent non-zero oe ients of (3.4).
8 GEORGE A. HAGEDORN AND ALAIN JOYE

Proposition 3.6 Assume N \ k 6= ;. Then Assumption 3.2 holds with

.



 = min
2N 2 + j j
< n ;

0= f  2 Nn : 
 (2 + jj)  0 g; and
 F = f  2 N n n  : j j < F g 0

for some suÆ iently large F > 0, provided

v () ( = )
  jj
1 2
=  ; 8  2  n N :
( +2)
0

Remark 3.7 A tually, F = max 20 jj + 1 will do.

.

Proof. Observe that for jj xed,



 2j+j jnj
2 + j j
with equality if and only if  2 k. Sin e x 7! 2x+nx is stri tly in reasing, we have
j  n > 00 min
j 00 j n  :
 2N \k 2 + j 00 j
Moreover, this shows that the minimum over N is rea hed on the nite set f  2
N n : j j  min00 2k j 00 j g. Then, Remark 3.5 implies that  is nite. Moreover,
0
for any  2 N ; 
 (2 + jj)  0. Hen e, the orresponding terms in the
Taylor series either ontribute to the normal form, or they are of higher order and
an be negle ted. The last ondition ensures that the  dependen e of the oeÆ ents
is strong enough to ompensate for the negative exponents 
 (jj + 2).
4. Mathemati al properties of HNF
The normal form Hamiltonian HNF plays the same r^ole that the harmoni
os illator plays in the standard Born{Oppenheimer approximation. However, in
ontrast with the harmoni os illator, we do not know the spe tral properties of
HNF expli itly. In [ ℄ we prove ertain properties that we require for our analysis.
3
We now des ribe them.
We onsider S hrodinger operators with potentials that satisfy the following
hypothesis.
Assumption 4.1 Let H =
+ V be self-adjoint on a suitable domain
.
1

in L (R ), where x 7! V (x) is a polynomial and V  0.

2
2 n

The rst on ern is the spe trum of the normal form Hamiltonian. We want
purely dis rete spe trum, whi h is not obvious with our polynomial potentials V
that may or may not be onning. Depending on the parameters, the set V (f0g) 1

may be unbounded. In order to ex lude essential spe trum, we ould use Persson's
Theorem, as we did in [ ℄ for the tri-atomi ase. A more general riterion for the
3
absen e of essential spe trum for S hrodinger operators with polynomial potentials
an be found in [ ℄ . Within our framework, it states the following:
7
1

1 We thank B. Heler for pointing this riterion out to us.

 HYDROGEN BONDS AND SEMICLASSICAL NORMAL FORMS 9

For a polynomial V  0, dene X
(4.1) mV (x) = 1 + jD V (x)j;
2Nn
where D = x11 x22


. Note that the sum is nite.
xnn
Proposition 4.2 ([ ℄, Thm (1.3)) Let H satisfy Assumption 4.1 and let mV
7 .
be dened by (4.1). Then, the resolvent of H is ompa t if and only if
lim mV (x) = 1:jxj!1
Remark 4.3 The proof of this result in [ ℄ nds its roots in earlier work on
. 7
hypoellipti operators. See e.g. [ ℄. 5

The fun tion mNF orresponding to the normal form Hamiltonian for tri-atomi
mole ules, i.e., to the potential E (x ;x ) given by NF 1 2

(4.2) E (x ; x ) = a x + a
NF 1 2 a x x + a x;
1
2
1 2 3 1
2
2 4
4
2

is easily seen to tend to innity as jxj tends to innity if Assumption 2.1 holds
Hen e, (HNF ) is dis rete, for all allowed values of the parameters faj g.
For the general normal form potential VNF (3.7), we assume that the orre-
sponding fun tion mNF diverges at innity.
We also need to know that the eigenfun tions de ay exponentially, in a L 2

sense, together with their derivatives. This is the ontent of the next result whi h
is proven in [ ℄.3

Proposition 4.4 Let H satisfy Assumption 4.1, and assume that the spe -
.
trum of H is dis rete. Then, any eigenfun tion x 7! '(x) 2 L (Rn ) is C 1 and, 2

for any a > 0,

' 2 D(eahxi); r' 2 D(eahxi ); and
' 2 D(eahxi );
P
where hxi = (1 + nj xj ) = , and D(eahxi ) denotes the domain of the operator
=1
2 1 2

multipli ation by eahxi .

Proof. The smoothness of the eigenfun tions follows from ellipti regular-
ity arguments. Their exponential de ay properties are onsequen es of lassi al
Combes{Thomas arguments. Sin e the potential is a polynomial, we immedi-
ately dedu e the exponential de ay of
' by using the eigenvalue equation. That
r' 2 D(eahxi) is a onsequen e of the following lemma, with p(x) = eahxi:
Lemma 4.5 Let x 7! p(x) 2 C (Rn ) be positive, and assume that there exists
.
1

a onstant C < 1, su h that Rj(rp(x))=p(x)j  2 C for

all x 2 R .
n

Then, for any f 2 g : Rn jg (x)j + j
g (x)j p(x) dx < 1 2 2

q
(4.3) kp = rf k  C kp = f k + kp = f k kp =
f k + C kp = f k :
1 2 1 2 1 2 1 2 2 1 2 2

The rather te hni al proof of this lemma an be found in [ ℄. 3

Remark 4.6 While Proposition 4.4 is suÆ ient to derive the leading ontri-
.
bution (1.2) of the vibrational spe trum, the derivation of a omplete asymptoti
expansion in powers of  = requires a generalization to arbitrary derivatives of the
1 2

eigenfun tion:
10 GEORGE A. HAGEDORN AND ALAIN JOYE

Proposition 4.7. Let H satisfy Assumption 4.1 and have dis rete spe trum.
Then any eigenfun tion x 7! '(x) 2 L2 (Rn ) is C 1 and, for any a > 0, and any
multi-index 2 N n , D ' 2 D(eahxi ).
This result is proved by indu tion on the order of . The initial step is provided
by Proposition 4.4. The indu tion step involves Paley{Wiener arguments. See [ ℄. 3

Remark 4.8 We also note that while the exponential de ay properties of

.
eigenfun tions of S hrodinger operators have been the obje t of many detailed and
rened investigations (See e.g., [ ℄), we ould not nd any published results that
6
dealt with the de ay properties of their su essive derivatives.
5. Leading order results

In this last se tion, we show how leading order results of the type (1.2), or more
generally, of the type (3.8) an be obtained by means of rather expli it quasimodes.
The following theorem is the main result of this note. It is the generalization of
Theorem 3.6 in [ ℄ to the situation des ribed in Se tion 3.
3

Theorem 5.1 Let H () be dened by (3.1) and satisfy Assumption 3.1.
.
Assume the ground state V (; x) satises Assumption 3.2, and let HNF be the
mol

normal form S hrodinger operator HNF =
+ VNF , with VNF dened by

1
2
(3.7). Further assume that the polynomial VNF is bounded below, and that HNF
has ompa t resolvent.
Let f 2 L2 (Rn ) be a normalized eigenve tor orresponding to an eigenvalue E
of HNF

(5.1) 1
+ V (y)  f (y) = E f (y):
2 y NF 

Then, for small , there exists an element E () of the spe trum of H () that
mol

satises
E () =  E + O(  ): +

for some  > 0.

Remark 5.2 The tri-atomi results are dire t onsequen es of this result, up
.
to a res aling of the energy.
Remark 5.3 Theorem 5.1 applies in parti ular to s alar S hrodinger operators
.
with potential V () satisfying Assumption 3.2.
Remark 5.4 In the ase of the tri-atomi mole ule, it is possible to go beyond
.
leading order results. Theorem 3.7 in [ ℄ states the existen e a omplete asymptoti
3
expansion in powers of  = for E ().
1 2

Proof. We onstru t a quasimode based on the elements above. This quasi-

mode is supported lose to the origin of the oordinates, and we introdu e a uto
fun tion F on Rn , dened as follows:
(5.2) F (z ;


; zn ) = F (z ) F (z )


F (zn )
1 1 2

where F (
) 2 C (R; [0; 1℄) is even, supported on [ 2; 2℄, and equal to 1 on [ 1; 1℄.
1
We introdu e a set of exponents (Æ ;


; Æn) with
1

(5.3) 0 < Æj < j ; 8 j = 1; 2;


; n;

 HYDROGEN BONDS AND SEMICLASSICAL NORMAL FORMS 11

whose values will be determined in the ourse of the proof. We also use the short-
hand notation x=Æ = (x =Æ1 ;


; xn =Æn ).
1
Our quasimode is then dened as
(5.4) Q(; x) = (x) f (x=)F (x=Æ );
where (x) 2 Hel is the normalized ele troni eigenve tor orresponding to V (; x).
Consider its norm,
Z Z
(5.5) k Q()k = n jf (x=)j dx
2 2
n
(1 F (x=Æ )) jf (x= )j dx:
2 2

R R
The rst term is equal to jj by s aling. Using the exponential de ay of f , Propo-
sition 4.4, and the onstraint (5.3) on the exponents, we see that the negative of
the se ond term is bounded above by
Z
j j
Z P P
a nj=1 jyj j e a nj=1 jyj j jf (y )j dy
Æj
j f (x= )j dx
2
  
Æj j
e 2 2 2

jxj j jyj j

P
a nj=1 1=j Æj
Pnj jyj j f k2
 jj e 2
kea =1

(5.6) = O(1 ):
Altogether, this shows that k Q(; x)k = jj= (1 + O(1 )). Thus, to prove the
2

result, we need only show that

(5.7) kHmol () Q ()  E Q ()k = O(jj=   ); 2+ +

for some  > 0. Using (3.1) and the denitions (3.7), (5.1), we see that
Hmol () Q ()  E Q () = (V (; x)  VNF (x= )) Q (; x)
(5.8) R(; x);
where
n
X
R(; x) = j (x) f (x= ) x2j F (x=Æ )=2
j =1
X n
+ j xj ((x) f (x= )) xj F (x=Æ )
j =1
n
X  
(5.9) + j f (x= ) x2j (x)=2 + xj f (x=) xj (x) F (x=Æ ):
j =1
We onsider the rst two terms in R(), where  and its derivatives are uniformly
bounded in a neigborhood of the origin. Be ause of the derivatives of the uto
fun tion, there is always one dire tion in whi h the integration range is away from
the origin. Using this together with the fa t that f and yj f are exponentially
de aying in any dire tion, an argument similar to that used in (5.6) shows that
these ontributions
Pn
are O(1). The last ontribution to R() is simply estimated
by O( j   j j ), where jj= omes from s aling. By (3.3), j j =
j
=1
 j = 
2 2

 + j > , so that R() satises the bound (5.7). To bound the norm of the rst
12 GEORGE A. HAGEDORN AND ALAIN JOYE

term in (5.8), we use (3.5) and (3.7) to see that

Z ! 1 =2
X
k(V ()  VNF (
= )) Q ()k  jx j jf (x= )j dx
2 2

 2F jxj j2Æj

(5.10)  ~ jj= 
Æ ; 2

for some onstant ~. By assumption, for all  2 F , we have


  = (
 (j j + 2)) > 0. Hen e, by ontinuity, hoosing Æj < j lose
1

to j , we get
2

jj= 
Æ = jj=  
Æ   jj=   ;
2 2+
with  > 0;
2+

whi h ends the proof.

Remark 5.5 From the proof, we get that the quasimode for the tri-atomi
.
mole ule is of the form
(5.11) Q(; x ; x ) = (x ; x ) f (x =; x = = ) F (x =Æ1 ; x =Æ2 );
1 2 1 2 1 2
1 2
1 2

for suitable values of Æ and Æ , and where f is an eigenfun tion of the normal form
1 2
(2.3). This shows that the amplitude of the distan e between the two heavy nu lei
is of order , whereas the Hydrogen atom an move a larger distan e away form its
equilibrium position, i.e., of order p.
Referen es

[1℄ M. Born and J. R. Oppenheimer, Zur Quantentheorie der Molekeln. Ann. Phys. (Leipzig) 84
(1927), 457{484.
[2℄ N. Elghobashi, Theory of using few y les IR and UV laser pulses to ontrol the orientation
and sele tive disso iation of Hydrogen-bonded anions, Ph D dissertation, Freie Universitat
Berlin, (2005).
[3℄ G.A. Hagedorn, A. Joye, A Mathemati al Theory for Vibrational Levels Asso iated with Hy-
drogen Bonds I: The Symmetri Case, Preprint (2006).
[4℄ G.A. Hagedorn, A. Joye, Mathemati al Analysis of Born-Oppenheimer Approximations, Pro-
eedings of the Conferen e, \Spe tral Theory and Mathemati al Physi s" in Honor of Barry
Simon, Mar h 27{31 2006, AMS Pro . of Symposia in Pure Math. (to appear).
[5℄ B. Heler, J. Nourrigat, Hypoellipti ite maximale pour des operateurs polyn^omes de hamps
de ve teurs, Birkhaser, Boston, 1985.
[6℄ P. Hislop, Exponential de ay of two-body eigenfun tions: A review in \Mathemati al Physi s
and Quantum Field Theory", Henry War hall (Editor), Ele troni J. Di. Eqns., Conf. 04
(2000), 265{288.
[7℄ A. Mohamed, J. Nourrigat, En adrement du N() pour un operateur de S hrodinger ave un
hamp magnetique et un potentiel ele trique, J. Math. Pures Appl. 70 (1990), 87{99.
Department of Mathemati s and Center for Statisti al Me hani s and Mathemati-
al Physi s, Virginia Polyte hni Institute and State University, Bla ksburg, Virginia
24061{0123, U.S.A.
E-mail address : hagedorn alvin.math.vt.edu
Institut Fourier, Universite de Grenoble I, BP 74, F{38402 Saint Martin d'Heres
Cedex, Fran e, and, Laboratoire de Physique et Modelisation des Milieux Condenses,
Universite de Grenoble I, BP 166, F{38042 Grenoble, Fran e
E-mail address : alain.joyeujf-grenoble.fr