ON THE ZEEMAN EFFECT IN ELECTRONIC TRANSITIONS

OF DIATOMIC MOLECULES

AERT SCHADEE
Sterrewacht “Sonnenborgh”.the Astronomical Institute at Utrecht, The Netherlands

(Received 2 November 1977)

Abstrac&--The Zeeman effect for diatomic molecules is calculated in a representation with Hund’s case (a)
wave functions as a basis set. The resulting expressions are simpler and easier to handle than those
obtained from previous calculations on the basis of case (b) wave functions. The calculation of intensities
and polarization of Zeeman patterns is presented for transitions between doublet states with arbitrary
spin-orbit (if A # 0) or spin-rotation (if A = 0) interaction, and for case (akase (a) transitions of any
multiplicity. These calculations are in good agreement with observations.

1. INTRODUCTION

IN THE late 1920s several laboratory experiments were set up to investigate molecular
electronic spectra in the presence of a magnetic field. In 1934 CRAWFORD(‘) summarized the
theoretical work of Kemble, Mulliken, Crawford, Hill and R. de L. Kronig, that explained the
observed level splittings for *H states with arbitrary spin-rotation interaction, for other doublet
states with arbitrary spin-orbit interaction and for case (a) states of arbitrary multiplicity.
However, the intensity distribution and polarization characteristics were understood only for
case (ajcase (a) transitions. As for other transitions, ALMYand CRAWFORD’*’ noted in 1929: “Due
to their complexity, theoretical intensity calculations as well as a consideration of the polariza-
tion of the patterns are left for later treatment”. This later treatment, for doublet transitions, is
the content of this paper.
There may have been at least two reasons why these calculations were not undertaken some
fifty years ago. First, in those days the usual representation of matrix elements had Hund’s case
(b) wave functions as a basis set, leading to quite tedious expressions for the transformation
matrix (which transforms the basis wave functions to the eigenfunctions) and particularly for
the dipole moment operator. Secondly, even with case (a) wave functions as a basis set, leading
to much simpler matrix elements, the computations are still so involved that they can hardly be
done without a computer, which was not available at the time.

2. THE ZEEMAN OPERATOR

The largest contribution to the Zeeman effect is due to the magnetic moment associated with
the electronic motion. It comprises a contribution pOL from the total electronic orbital angular
momentum and 2& from the total electronic spin, p. being the Bohr magnetron. For a
magnetic field H in the Z-direction of a laboratory-fixed coordinate system (X, Y, Z), the
Zeeman Hamiltonian is

HH = (Lz + =z)P&. (1)

We introduce, in wavenumber units,

Aa,, = &J = 4.66860 x lo-’ H cm-‘. (2)

The matrix elements of HH in the molecular-fixed coordinate system (x, y, z) are found through
the direction cosine matrix with elements a& which transforms the laboratory axes R to the
molecular s [e.g. HouGEN(~)]:

L,+2Sz=;[az+(L_+2s-)+ aZ-(L++2S+)l+az~(d+2S,) (3)

518 A. SCHADEE

where, conventionally, qz = qx 2 iq,. As a basis set, we choose the wavefunctions in strong

(L, S)-coupling, i.e. Hund’s case (a): IASZ; RJM). Here, A = L; is the projection of the
electronic orbital angular momentum on the internuclear axis z, S the electronic spin, C = S, its
projection on the axis; J is the total angular momentum, R = A + Z = Jz and M = Jz are its
projections on the molecular z-axis and laboratory Z-axis (field direction), resDectivelv. I-Z” is
diagonal in M. We then find [see, e.g.. HOIJCEN’~‘]:

(MC; RJMlH,(AS~; RJM) = (A + 2c.) $5 ha,, (44

(ASIX+- l;R& lJM(H,~ASC;RJM)= $+)[(J3R)(J+n+ l)(s+c)(s’c+ l)]“*. (4b)

(ASZ; QJ + I Ml& IASX; fiJM)

=(A+22)Auo (J+R+ l)(J-R+ l)(J+ M+ l)(J- M+ 1) “’

(4c)
J+l C (2J + 1)(2J + 3) I *
(ASZ* 1;R lJ+ lM(H,,(ASz;RJM)

1’
Aa0 (J~~+I)(J~~+~)(J+M+~)(J-M+~)(ST~)(S+~+~) “*
(4d)
=TJ+I [ (2J + 1)(2J + 3)

1’
(A+aC)Au,, (J+Q,(J-n)(J+ M)(J- M) “*
(AS% RJ - 1MIH,IASZ; OJM) = J (4e)
[ (2J- 1)(2J+ 1)

(ASCt I;R-+ lJ- lMIH,IASX:RJM)

(JTR)(JT~- l)(J+ M)(J- M)(STX)(S”X+ 1) I”

(40
(2J - I)(23 + 1) I

For the spin operator, the sign convention used by CONEQN and SHORTLEY’~’is applied
[see SCHADEE”‘, remark 31.

3. TRANSITIONS HUND’S CASE (a)-(a)

3.1. Energy levels
In order to find the energy levels under strong (L, S)-coupling, we apply perturbation theory.
Remembering that, in case (a), the unperturbed levels F*“(J) are widely separated for AR # 0 or
AJf 0, we retain in good approximation only the diagonal elements with AJ = A0 = 0, so that

AFn’(J, M) = Fn’(J, M) - &O(J) = (A + 22) $$ Aa0 (cm-‘). (5)

Often, A + 22 is written as R + z. To this approximation, the perturbed eigenfunctions are the

same as the unperturbed, i.e. the basis functions IASX; RJM).
It follows from eqn (5) that the ‘X, ‘III/~, 3KI~, ‘A,, 4111,~,4A,, 4@3,2 (sub)states show no
Zeeman effect whatsoever.
For case (a)-(a) transitions, the displacement from the no-field wavenumber is readily found
to be
Au(J’M’, JM) = &‘(J’M’) - F*‘(J, M) - Fno(J’) + &O(J)

= (A’ + ZZ)R’M’_ (A + 2C)nM

J’(J’ + 1) J(J + 1)
(6)

with primed quantities for the upper level (C’ = C for allowed transitions).

3.2. Intensities, Zeeman patterns

Intensities are proportional to the square of the matrix elements of the electric dipole
operator P. Elements with ASf 0 vanish for allowed transitions. For case (a)-case (a) tran-
sitions, elements with AX f 0 vanish as well. which makes the case (a) wavefunctions an
 On the Zeeman effect in electronic transitions of diatomic molecules 519

appropriate choice as a basis set. Let the complete wavefunction be IASC; u; RJMp), Ip) for
the parity, which discriminates between A-type doubling components, and Iv) for the vibration,
on which P does not operate. Writing P = R,a, where R, operates on IASZ) and a on IfiJM),
we have (WHITING and NICHOLLS(~))

(A’SI;; v’; R’J’M’p’jPIASZ; u; flJMp> = (u’; v)(A’SCIR,(ASZ)(~~‘J’M’~‘~~(~~JM~). (7)

For non-zero elements, the parities p’ and p are always opposite. The first two factors are
constants for a given vibration band (n’v’, nu): (u’Iu)* = qooum
and IA’SXIR,lASZ)l* = R: are the
Franck-Condon factor and the square of the electronic transition moment, respectively. The
values that the remaining factor qnM = (fVJ’M’p’(alQJMp) can take are collected in Table 1.
The Honl-London factors of the Zeeman components,

can be conveniently written as the product of a factor containing J and M and the Hbnl-
London factor of the no-field transition SsJpp,since this format is independent of AR (Table 2).
The various factors 2 and 7~pin Table 1 are included such that the Honl-London factors obey
the sum rule

F ,c, 2 ,c, SXJMp

= (2 - ~o,K+A)(~~
+ 1)(2J + 1) (9)

in accordance with WHITING and NICHOLLS.(‘) For the II - 2 and C - fl transitions, PX = 2; for all
other transitions Pi = 1.
The Zeeman patterns consist in principle of about 25 equally spaced lines per polarization
mode. As an example, expected patterns for the R3(S), Q3(5), P#)-lines of 3@4-3Aj and
‘A3 - ‘A3 transitions in a field of 3000 Gauss are drawn in Fig. 1 (TiO red and green systems). It

A&m’)
-. 2 0 .2 72 0 .2 -A -.2 0 .2 A

AP

r
+l

-1

0
I. .I
0 .2 72 0 .2 -.4 -.2 0 .2 A

+l ,a

-1 $7

t
I
I
0
II I 1
R3(5) q(5) P3’5’

Fig. 1. Zeeman patterns for case (atcase (a) transitions. Arrows indicate the weighted mean line shift per
polarization mode if different from zero.
520 A. SCHADEE
 On the Zeeman effect in electronic transitions of diatomic molecules 521

Table 2. Hiinl-London factors .S1,~~ for case (akcase (a) transitions and for transition between states for which
the internal molecular interactions (spin-orbit or spin-rotation) are large compared to the magnetic field interaction.

Normalization: n:n ,X,r M’M F Ss,, = (2 - &,,.+,)(2S+ I)(21 t 1).

AM = 0 AM = -+ I

(J+i)’ - M’ (J+I+M.AM) (J+~+M.AM)

SR SR SR
ZJMP (J+l) (ZJ+l) (25+3) zJp 2 (J+l) (2J+l) (25+3) zJp

M* (J-M.AM) (J+~+M.AM)
SQ SQ SQ
ZJMP ~JP 25 (J+I) (‘2Jtl) ~JP
.I (J+I) (ZJ+l)

5’ - Hz (J-M.AM) (J-I-M.AM)
SP SP SP
XJMP 25 (25-l) (2J+l) ~JP
J (25-l) (2J+l) EJp

follows from eqns (5) and (6) that the patterns grow narrower with increasing J. Weighted mean
displacements of the patterns for each AM are defined (arrows in Fig. 1) by weighing the
displacements Au with the HGnl-London factor. The weighted mean positions of r-patterns are
always zero. The separation of the weighted mean positions for the two u-patterns (AM = + 1
and =-l)are
(W)R= (a’+ n + Z)lAfilAu,/(J + l),
(a)o = [(fl’-+ Z)R’+ @‘I+ X,fi]Auo/J(J + I), (10)
(u)p = (nl + R + C)lAQlAu,,D.

In AR = 0 transitions, the (u) of the u-components of P- and R-lines are zero. The weak
Q-lines split into one component for each AM, the separation of the two u-components
decreasing with J2. Clearly, An = 0 transitions, as for the TiO green system, are poor indicators
of magnetic fields.
In An = 2 1 transitions (TiO red system), the stronger Q-branch has a larger splitting of
the u-components that the P- and R-branches only for small .I. The subband with the greatest
fi shows the largest Zeeman effect, R’ + R + P being 8, 5 and 2 for the 3@p4 - ‘A3, 3@3- 3A2and
3@2- ‘Al subbands, respectively.

4. DOUBLET STATES

4.1 Hill’s theory

Already in 1929, HILL”’ showed how the Zeeman effect could be solved for doublet states
with arbitrary (L, S)-coupling. He pointed out that perturbation theory must fail since, for very
loose coupling [IY - 21= 2, Hund’s case (b)], the energies of the unperturbed levels F,’ (N,
J, = N + i) and Ft (N, Jz = N - 4) are equal, so that the correction term HH(Jlr J2)/[Flo(N, J,) -
Fzo (N, J2)] would become infinite. Hill evaded this difficulty as follows (see Fig. 2): with the
Hund’s case (b) wave functions IF&V, J, M)) as a basis set, he calculated the total Hamiltonian
H, i.e. the sum of H,, (the electronic-vibrational part including the spin-orbit interaction YAZ),
H, (the rotational part including the angular momentum operators J and S) and HH.
Hill then split the total Hamiltonian againas follows:

H=H,,+Hr+HH=H,+Hz (11)

such that HI contains only terms with AJ = 0 and HZ the remaining terms with AJ = 2 1. First
neglecting Hz, he diagonalized HI, finding a transformation matrix S and a new basis set,
IE(N, Ji, M)). Next, the H2 is transformed by the same transformation, yielding Hi = S-‘H,S
still containing only elements with AJ = ?l. Out of these, elements with AN = * 1 (Crosses in
Fig. 2) are neglected, as they would be with perturbation theory, since the unperturbed energy
522 A. SCHADEE

<Es (J+l M) 1 H7 HI HP H2 <F,(N=J*JJ+lM)I

<F+(J*l M)j H, H, H, H, <F;(N=J+JJ+IM)I
<E+(J Ml1 H, Hz H\ H, H, H; s O;(N=J+& J M)I
<F$‘J M)I HPH2 H> HI H2 H2 - ~(N=J-h J M)J
@(J-l M)I H, H, HI HI <c(N=J-*J-I M)I
<F&W1 M)I <?(N=J-go-I M)I
Hz H2 HI HI

Fig. 2. The transformation of the Hamiltonian to a representation in which it falls apart into 2 x 2 matrices.
The case (a) basis functions have the subscripts Z rather than ll.

difference (P’(N + AN, J + A./) - F (N, J)( is also large for both ANf 0 and AJf 0. By doing
so, the transformed total Hamiltonian H’ = S-’ KS = S’(H, + H2)S falls apart into a set of
2 X 2 matrices for each N with elements for AN = 0 and AJ = 0, * 1 only. The eigenvalues of
these matrices, F;:‘(N, M), are the desired energies of the levels in a magnetic field. Hill’s
treatment fully explained the observations of the Zeeman effect in MgH by ALMY and
CRAWFORD(‘).
The difficulty of explaining the intensity and polarization of the Zeeman components,
mentioned by Almy and Crawford, is due to Hill’s choice of case (b) basis functions. In case
(a), the matrix elements of the electric dipole operator are simple expressions (see Section 3.2).
In this section, we repeat Hill’s calculation, taking case (a) wave functions as a basis set. In
order to obtain the intensity and polarization of the Zeeman transitions, a second trans-
formation T from the basis set Ifi(N,Ji, M)) to the final set IK(N, M)) is required. T follows
from diagonalizing the H’ = 5’ HS 2 x 2 matrices for each N. The resulting Hiinl-London
factors take the form

r.&;v$sP’ =
IT-‘S-‘qnMSTI? (12)

4.2 States with A # 0

On the basis of case (a) wave functions, [(ASZ; WM) = IFn(J, M)) with fi = A + t] the
elements of If,, i.e. elements with AJ = 0 (Fig. 2), are (HOWEN and Section 2)

(~,)A++,.++ = B{(J + 4+ A)(J + 4- A) + (A2+ +)p ? A( Y - 2 + 3p)/2},

(HI),,:.~\~:=-B(I-~){(J+~+A)(J+~-A))~~, (13)

where B is the rotational constant in cm-‘, Y the dimensionless (L, S)-coupling constant, and

(14)

The eigenvalues of Hi, which become the diagonal elements of S’HS, are

Fi(N,Ji,M)=B{(J;+f+A)(Ji+~-A)+(A2+4)Bi+4(-I)i~(Ji,M)j, (15)

where i = 1 or 2. and

u(J, M) = (P2 + Q2)“2,

P=2(1-p){(J++-A)(J+;+A)}“2,
Q=A(Y-2+3@),

with J, = N +i for F,-levels and J2 = N-t for &levels. For the elements of the
 On the Zeeman effect in electronic transitions of diatomic molecules 523

transformation matrix S, which diagonalizes HI, we define (see Kovlics@‘]

c, = f (ue2+ P) = KU*,
and find
S&i. ,(J,, M) = P(J,)/[2C-(J,, M)li
s*-~.z(Jz~ Ml = - w2, M)/[2C+(J2, M)P,
(16)
S*+$.,(J,, M) = u-(J,. M)/[2C-(J,, M,li
s*+:.*u*.M) = u+(Jz, M)/[2C+(J*, M)li.

The new basis is now

IF;W.Ji,Ml) = ; Sfl.;(Ji,Ml1FdJi, Ml). (17)

The elements of H2 follow from section 2. After transformation by S, the off-diagonal elements
of S’HS. i.e. elements with AN = 0 and AJ = 2 1, are now found, viz.

i&J% Ml = (F,W, J,, M)(S-‘(A, M)H2(J,, M; J2. M)S(J2, M)IF,(N, J2, M)). (18)

After substitution,

M2 f
(N-A)(N+ i+h)
N(N + l)C_(J,, M)C+(J2, M) ‘-(N +t)i>}
x{2[1 -p(J2, M)][N +Al[u-(J,, M)+(l-A)(1 -P(J,, M))(N + 1-A)1
+~[~-P(J,,M)][N+~-AI[U+(JZ,M)+(~+A)(~-P(J~,M))(N+A)I
+ (1 + A)u-(J,, M)u+(J2,WI. (19)

The total energies of the i&man levels are the eigenvalues of the 2 x 2 matrices of S’ HS.
Substracting the no-field energy F, the Zeeman splitting becomes

AIY(N, Ml = t[F,W, J,. M) + F,(N, 52, Ml

+ (- l)‘fo([F~(N,JI, M) - Fz(N, Jz~M)12 + S’(N, M)$l - FiO(N,Ji). m)

The reason for introducing f. = sign [F2’(N, J2) - FIo(N, J,)] will be given below.
The total Hamiltonian S-’ HS can be diagonalized by a transformation T, giving the proper
eigenfunctions

(F,‘(N, M)) = 7 Ki]Fi(N, JiqM)), (21)

T,, = T2* = r/[r2 + S2]i,

TU = - T2, = fo8/[r2 + 6’]f, (22)

where

r = IF,(N, J,, M)- UN, Jz. M)I +{(F, - F’2)2+ a’};.

In the rare case that sign (F, - F2) # sign (FIo- I$“), all factors 6 and r in eqn (22) should be
interchanged. For those case (b) states for which the coupling constant is unknown and
therefore adopted to be zero, we take f. = - 1.

QSRT Vol. 19 No. 5-E

524 A. SCHADEE

If 6(N, M) is not negligible, it follows from eqns (21) and (22) that the final wave functions
are linear combinations of wave functions that contain J, and J2. In other words, no J-value can
be assigned to Zeeman-levels, although M(= M,) remains well-defined all the way. Since
52 = N - !, an exception arises for levels with IMI = N + f, where no mixing of J, - and Jz-wave
functions can occur, since F,(N, J2, IMI = N +$-levels do not exist. For these levels, the
following relation holds:

F,‘[N,M=r(N+~)]=F,[N,J,,M=+(N+~)]=B[(J,+f)*-A*-~u(J,,M)]t”~~(:’:”
I
(23)

and also IF,‘(N, M)) = IF,(N, J,, M)) or, formally, T,, = 1 and T12= T21= r2, = 0.
In Fig. 3, energies of fi’(N, M)-levels in a *II-state as a function of the (L, S)-coupling
constant Y have been plotted, taking N = 2, the rotational constant B = 1 cm-’ and a magnetic
field of 3000 Gauss. We first note that the no-field levels E’(N, Ji) cross at Y = 0 and Y = 4,
However, pairs of levels F’ with equal M obey the non-crossing rule. Hence, the factoif,, is
introduced in eqn (20) so as to make the Ft(N, M)-levels with IMI s N - 4 approach the
J?(N, Ji) - level id the limiting no-field case for any value of Y. Another argument for
introducing f0 is to achieve continuity between the F,‘(N, M)-levels with IM( d N -i and
those with JMI = N + i.
The behaviour of the Zeeman levels at weak (L, S)-coupling is, in fact, a Paschen-Back
effect analogous to the atomic case. In particular, around Y = 0, when the electronic spin is
completely uncoupled from the other angular moments of the system, its magnetic moment
p,=2&=+p o is directed either parallel or antiparallel to the magnetic field. This contribu-
tion to the Zeeman splitting is far larger than that of the orbital motion, which is of the order of
A2Mpo/N(N + 1) [HERZBERG’~‘]. Therefore, (2N + 1) levels, being one half of the total number
of (23, + 1) + (2J2+ 1) levels, crowd together around Au = F’ -p = + Auo. the other half
around Aa = - Aa0. The fact that J loses its meaning for Zeeman levels is clearly illustrated by

0 ‘J 4

Fig. 3. The Zeeman splitting of the ‘II, N = 2 levels at a field of 3000 Gauss as a function of the spin-orbit
interaction (Y). -, F,“- and F,‘-levels; -----, Fzo- and I$‘-levels.
 On the Zeeman effect in electronic transition of diatomic molecules 525

Fig. 3: the F’-levels, which for large Y are grouped around the Fro-level with J = N + i, join
the Fzo-level with J = N -: for 0 c Y G 4.
The (N = A)-levels need special treatment since the first rotational level F,(J = A-f, M) is
irregular (SCHADEE”‘]. Its unperturbed energy is

&(A, A - ;, M) = - f BA( Y - 2) + (A - l)MAao/(A + ;) (24)

and its wave function is the Hund’s case (a) basis function (F_;(A-i, M)), so that no
transformation S is needed. It follows from eqn (24) that the first rotational level of a *II-state is
magnetically insensitive as long as it is not perturbed by the F,(N = A, J, = A + i)-level. The
perturbation element for (N = A)-levels is
I

S(A, M) = Aa,, (2A~l~[;)~~;~,(~~)2’]1{2(A-l)[1-~(J,)]-u_(J,,M)} (25)

with J, = A + i.

4.3 ‘Z-states
ALMY”” published Hill’s calculation of the Zeeman effect for ‘C-states, taking into account
the rotation-spin interaction, characterized by the splitting constant y. The interaction operator
is

y(R.S)=y(J-L-S).S. (26)

On the basis of Hund’s case (a) wave functions, its matrix elements are

(ASC; s1JMIH,(ASX; RJM) = y[2* - S(S + l)],

(ASX+ i;fi+ ~JMIH,IAS~;RJM)=~~[(STZ)(S+~+ i)(J~0)(~+fk+ I+. (27)

Since A = 0, He,. = 0, so that the total Hamiltonian is

H=H,+H,+H”=H,+H*. (28)

As before, HI contains only elements with AJ = 0. Its eigenvalues are

MAa,
E(N,Jj,M)=BN(N+l)-(-1)’ ;y(N+f)+m . (2%
2 I

The elements of the transformation matrix are

_f
s-:, I = s+;. I = - s-j. 2 = s+;, 2 = w (30)

and are independent of Au0 and y. The off-diagonal element in S -’ HS is also independent of y.
viz.
I
S = 2Aao{1 - M2/( N + $)‘}I. (31)

The Zeeman splitting of the levels becomes

AK(N, M) = (- l)i{; y(N +;)+fo[$ y*(N +;)*+ yMAuo+(A~n,)*]f}, (32)

with f. = sign (F2”- FIo) as before. This definition implies that f. = - 1 for y > 0 (most cases).
The second transformation T is the same as for states with A # 0 [eqns (21) and (22)]. If (1) y
and M have opposite signs and (2) f Jyl(N +$) < IMJAuo/(N +$, fi = sign(F2- F,) is the
opposite of f. = sign (F,‘- F,‘). Then S and r must be interchanged in the K]is of eqn (22).
526 A. SCHADEE

Again, Zeeman levels with M = f (N + $ behave differently, the splitting being

AF,‘[N, M =+(N+;)]=+Av,,. (33)

For those *z-states for which y is unknown (i.e. almost all excited states), the best thing to do is
to put y = 0. The above equations then still hold if f0 = - 1 is substituted. The splittings become

AFi’ = - ( - I)‘Ao,, (34)

for all M and the elements of T are

T,, = Tz2 = {(N + ++ M)/(2N + l)}-:,

- T,*= T2,= {(N+ ;- M)/(2N + I)}!. (35)

Again a Paschen-Back effect occurs: the coinciding F,‘(N)- and Fzo(N)-levels are split in two,
each level containing 2N + 1 coinciding Zeeman-levels. Each such Zeeman-level IF’(N, M))
is a linear combination of jF,(N, Jr, M)) and IF2(N, J2, M)) to which no proper J can be
assigned.
The competition between the (R, S)-splitting and the magnetic splitting is illustrated in Figs.
4 and 5. Figure 4 pictures the splitting for a number of rotational levels of the X*Z-ground state

-_ --_
-... -_ --_
--_ -.-._.__ -_ -_
--_ -.__ ---_ -.- ~&N-1/2)
-.2 - _.
_. -_-.-.-_
-_
-_ -.-.
_-__ -.__
-- -1 -_
-=__ -.__
-. F;(N)
---
-.
--_
-_
--._
-.‘I - --_
- _ _.M=N-‘h

Fig. 4. The Zeeman splitting of the X’I: MgH ground state at a field of 3OOOGauss (y = 0.025 cm-‘). Only
levels with [MI = N t f are drawn. -. F,=. F,‘: -----. Fz’. FT.

of MgH for H i= 3000 Gauss. The pure (R, S)-splitting is y(N + $) with y = 0.025 cm-’ and the
pure magnetic splitting is 2Auo = 0.28 cm-‘. Two sets of levels occur, F,’ and F,‘, with
M = N -t,N - 3/2, . ,- (N -4)of which only the outermost levels with M = -+(N - f) are
drawn, the other levels lying in between. Furthermore, there are two F,‘-levels with M =
k (N + $). For small N the magnetic splitting dominates the (R, S)-splitting, causing two groups
of 2N + 1 closely spaced levels, about -yN cm-’ in width, their centres of gravity 2Aaocm-’
apart. With increasing N and, therefore, with increasing (R, S)-interaction, the F,'[N, M =
-(N +:)I-level gradually shifts from the lower to the upper group. At large N, the (R, S)-
interaction dominates the magnetic field interaction. The Paschen-Back effect, at low N, has
vanished. There are now two groups of (2Ji + I)-levels, the centres of gravity are yN cm-’ apart
and each about 2Auocm-’ wide. In Fig. 5, the magnetic splitting for an N = 2-level with
y = 0.025 cm-’ and increasing magnetic field is shown. For small fields, the magnetic interaction
causes weak broadening of each (I?, S)-split level, but a large field causes a Pachen-Back effect:
two groups of 5(= 2N + I)-levels, about y(N + f) = 0.063 cm-’ wide and 2Auo cm-’ apart.
 52-l
On the Zeeman effect in electronic transitions of diatomic molecules

',. ..
-0.1- .. .
'. .
I I I
0 1000
H(Gauss)

Fig. 5. The Zeeman splitting of a ‘I, N = 2 level as a function of the magnetic field. y = 0.025 cm-‘. -.
F,‘: -----, Fz’.

4.4 Approximations
In numerical calculations, large errors in AET(N, M) may be introduced by the subtraction
of the nearly equal quantities Ff and F”. It is recommended to subtract a common term
B[N(N + 1)- A21from all term values F, F and F’.
(a) States with A # 0. (i) The magnetic interaction is characterized by 8, the (L, S)-
interaction by F2’(N, J2) - F,‘(N, J,). If S2(N, M) Q [F2(N, J2, M) - F,(N, J,, M)]‘, the square
root of eqn (20) should be expanded as follows:

S’(N, M)
AFi’(N.M)=fi(N,J,M)-F,O(N,Ji)+ (36)
4[E(N, Jiv M)-fi(N,Jj, M)]’

if i = 1, then j = 2 and vice versa. The factor f. drops out by this expansion and S2 has a
maximum value for M = 0 [eqn (19)]. The elements of T are now
,
TII = T22= I- S2/8(F2- Fl)2, T,2 = - T2, = S/2(F2 - F,). (37)

For the first rotational level,

AFj(A, M) = (A- l)MA~,,/(A+f)+4(~,f F ). (38)

I

(ii) If, moreover, lfil Q 1 and J2 5 A2Y2 [case (b) or intermediate coupling], the square root
in U(Ji) should be expanded in j% Neglecting terms with p2, one finds

MAa, i3A2(Y-2)-4(Ji+~-A)(Ji+~+A) s2
AFi’(N, M)= Ji(Ji+ 1) (39)
2u”(Ji) I +4(F,-e).

This expression also holds for M = 5 (N + i), provided S = 0 is substituted.

(iii) If A2Y* % J* [intermediate coupling approaching case (a)] and A2Y2% /3* (no condition
528 A. SCHADEE

IpI e 1!), the expansion of u(Ji) yields

with X=+f for Fz if Y-++@ and for F, if Y + - cfl. For very large 1YI, eqn (40) approaches
eqn (5), the case (a) expression. The expression (38) for the first level is independent of Y.

(b) *ICstntes
If IS/(F2 - F,)[ G 1, the square root of eqn (20) should again be expanded in 6. In all realistic
cases, sign (Fzo - FIo) = sign (F2 - F,). Under these conditions,

AF,‘(N, M)=(- 1)“‘w+4(F,‘: F,). (41)

2 1,

In the rare case that both IS/(Fz - F, )( B 1 and sign (Fzo - F,‘) # sign (F2 - F, ), one has

AE(N,M)=(-1)’
s*
(42)
4(E-F,)

For the levels with M = +(N +$, eqn (33) holds in all cases.

4.5 Limiting cases in 2C-states

(i) (Aoo)* 6 -y*(N + i)2. The Zeeman splitting becomes

AFi’(N, M)=(,l~;‘{,,,+$} (43)

2

and is independent of sign (y). The elements of T become

T,, = Tz = 1 - S2/8y2(N + t)‘,

7.2, = - TU = 6/2y(N + 4). (44)

(ii) (AcQ)* % r*( N + i)‘.

AFi’= (- l)i{j,(AuO+~ rM)++ r(N +;)} (45)

with f. = sign (F2’ - F,‘) = + 1 for y < 0 and = - 1 for y 3 0. The elements of T may now be
approximated by

‘
T,, = Tz2 =(l+ Y(N +I- M)/4Auo}{(N +:+ M)/(2N + l)}:,

7-n = - T2, =fo{l- y(N +t+ M)/4Aco}{(N +:- M)/(2N + I)}+. (46)

5.INTENSITIESOFZEEMANCOMPONENTLINES
To obtain intensities, one must know the eigenfunctions of the Zeeman levels, (F;(N, M)).
As was shown in the preceding sections, these eigenfunctions are obtained by performing two
transformations, S and T. Since the electric dipole operator is diagonal for case (a) wave
functions and since we have chosen these functions as our basis set, the Hiinl-London factor of
a Zeeman component is in detail [see eqn (12)]:

No general expression for SiNMr, can be given. Neither have we been able to think up a
 On the Zeeman effect in electronic transitions of diatomic molecules 529

convenient short notation for the Zeeman transitions: since each Zeeman level is a mixture of
wave functions with different .I, conventional notations with P, Q or R cannot be used. It seems
that each transition must be written in full: (i’N’M’p’, iNi%@).
Simpler expressions may be derived if the energies of the internal molecular coupling, (R, S)
or (L, S), are much larger than the magnetic field energies in both the upper and lower levels.
Then T,, = Tz2= 1 and Tlz = T2, = 0 and no mixing of wave functions with different J occurs.
Bearing in mind that the Htinl-London factor of a no-field rotational transition is defined as

sirrp E Sig$‘P’ =
&lzi S&(J')q*,(n'J'M', Q/M)Sni(J)
I*
and that qoM can be written as the product qo(WJ’,flJ) (which is independent of M) times
q&‘M’, 04) (which is independent of 0) (see Table l), we have

(49)

as was used to arrive at Table 2. Equation (49) is applicable to transitions of any multiplicity, as
long as an internal molecular interaction that couples the electronic spin to any other angular
momentum [(L, S) or (R, S)] is stronger than the external magnetic field interaction which tends
to decouple the spin.
As was mentioned in Section 3.2, the factors qo and qw are written such that the resulting
Hiinl-London factors for no-field rotational lines obey the summation rule

z3, , &,p = (2- h4~+,)(2s + I)(25 + 1). (50)

6. APPLICATION: THE MgH GREEN SYSTEM

observed the A*II - X*x’

ALMY and CRAWFORD’*’ system of MgH in magnetic fields up to
28,OOOGauss.The splitting patterns were most satisfactorily accounted for by Hill’s theory.

M AF;z(N.M)

‘-9, -0.1401
x2x +

Fig. 6. Energy-level diagram for the MgH A%, iV’ = I to the X2% N = 0,l transition. No-field to the left,
at 3000 Gauss to the right. Energy differences in cm -I. Below each no-field transition. the H8nl-London
factor is given.
530 A. SCHADEE

They described the intensity distribution qualitatively. The intensity calculations in this paper
agree very well with observations. Let us, as an example, take the four no-field transitions
between A%, N’ = 1 levels and the X*Z, N = 0 and = 1 levels (Fig. 6). The resulting patterns at
3000 Gauss are presented in Fig. 7, the numerical details are given in Table 3. The computations
match the observations very well since Almy and Crawford describe both the P,(t) + Q&) and
the Q,(f) patterns as “doublet, approximately of the same intensity” and the Q,(t) + R,,(i)
pattern as a “diffuse doublet”. As to the polarization, each of the two components into which
the P,,(s) + Q,?(i) line is split turns out to comprise contributions from the different polarization
modes AA4= 0, of:1, but not in equal amounts, so that we predict polarization differences
between these components. From the numbers in Table 3, it appears that the (R, S)-interaction
y(N + 4) = 0.038 cm-‘, which is small compared to the Zeeman interaction 2Aao = 0.28 cm-‘,
has a quite noticeable effect on the intensities, the degree of polarization and, to a lesser extent, on
the wavenumbers. Indeed, computations at faster rotatation (increasing N) show that the amount
of (R, S)-coupling is decisive for the shape, intensity distribution and polarization of the Zeeman
patterns. Consequently, accurate values of the splitting constant y are essential for the
computations. In most cases, y is poorly known, particularly for excited 2x-states.
The P,,(s) and Q&) doublet has also been observed in sunspot spectra by W~)HL”“,
HARVEY”*’ and BUURMAN.(‘~) From Table 3, the computed intensity ratio of the violet to the red
component equals 1.13 (provided no self-absorption occurs), whereas we estimate an observed
ratio of roughly 8/7 (- 1.14) from Harvey’s Fig. 2. The influence of y should be noted: for
y = 0, the components would be equally intense. Buurman’s spectra in the two circular

Table 3. Zeeman components of the MgHA% N’ = I to X’Z. n = 0. 1 transitions at a field of 3000 Gauss. First
and second entry: the wavenumber displacement (cm-‘) and the HGnl-London factor. Third and fourth entry: the
same quantities, putting ~(~2) = 0 (y does not affect the 21. N = 0 level). The wavenumber displacements of the
Q&J + R&J and the P,,($ + Q,>(f) transitions are taken with respect to the geometric mean no-field wavenumber
of these close doublets (see Fig. 7).

F; (I$“) Ff (I$“)

x*x+
- .07l - .I59
.496 .I65
+;
- .052 - .I40
( .496 .I67

- .059 - .I18 - .I77 - .I47 - .I47

.I78 .064 .365 .00'3 .303
+$ - .II' - .I69 -. 140 - .I40
- .052
.I57 .063 .343 0 .331

F; (‘,M)
- .I05 - .I64 - .223 - .I35 - .I35
I .I33 .005 .208 .I36 .332
-7 - .I'1 - .I69 - .228 - .I40 - .I40
.I07 .003 .I85 .I66 .333

+ .082 + .033 + ,121

3 .33' .499 .I68
-T + .I11 + .052 + .I40
.331 .499 .I68

+ .236 i .I77 + .I18 + .I47 + .l47

.I68 .0013 .084 .332 .I96
+;
+ .228 + .I70 + .I11 + .I40 + .I40
.169 .0026 .I06 .334 .I68

F: (1.M)
+ .I64 + .I06 + .047 + .I35 + .I35
I .313 .060 .I39 .366 .0015
-7 + .!?O + .'I1 + .05% + .I40 + .I40
.339 .062 .I62 .335 0

- .052 - .III - .I69 - .I40 - .I40

.404 .272 .272 .22l .443

F: (0.M)
+ .I70 + .I11 + .052
I -+ I .I34 .270 .4'0
 On the Zeeman effect in electronic transitions of diatomic molecules 531

AUbm-i’~
-Q2 0 0.2
I 1 I I I, I I

co.4-. I I I 1 I

.2-
O,AM=*l
AM=-1 1
.2- -.4

.4 - - .2
AM=0 o

Fig. 7. Zeeman patterns of the MgH A*Il, N’ = 1to the X*X, N = 0, I transitions at a field of 3000Gauss. The
no-field line positions are indicated.

polarizations confirm the correct sense of the computed polarization, but are not suited for a
quantitative comparison.

Acknowledgements-J thank Dr. G. NIENHUIS of the Fysisch Laboratorium at Utrecht for his advice and Drs. VAN BUEREN
and ZWAANfor their criticism. This work was started while I was visiting fellow at the Department of Astrophysics,
University of Oxford. I thank Prof. BLACKWELLand the other members of the Department for their hospitality and
kindness.

REFERENCES
I. F. H. CRAWFORD, Rev. Mod. Phys. 6, 90 (1934).
2. G. M. ALMY and F. H. CRAWFORD, Phys. Rev. 34. I57 (1929).
3. J. T. HOUGEN.NBS Monog. I I5 (1970).
4. E. U. CONWN and G. H. SHORTLEY, The Theory of Atomic Spectra. The University Press, Cambridge (1951)
5. A. SCHADEE, Astron. Astrophys. 41. 203 (1975).
6. E. E. WHITINGand R. W. NICHOLLS,Astrophys. 1. Suppl. Ser. 27, 1 (1974).
7. E. L. HILL, Phys. Rec. 34. 1507 (1929).
8. I. KovLs, Rotational Srructure in the Spectra of Diafomic Molecules. Adam Hilger Ltd. London (1%9).
9. G. HERZEIERG, Spectra of Diafomic Molecules. 2nd Edn, p. 303. Van Nostrand, New York (1955).
IO. G. M. ALMY, Phys. Rev. 35, 1495 (1930).
II. H. W~)HL,Asfron. Astrophys. 3, 378 (1%9).
12. J. W. Harvey, SolarPhys. 28, 43 (1973).
13. J. BUURMAN.Astron. Astrophys. 29, 329 (1973).