Gauge independent effective potential and the Higgs boson mass bound
Guey-Lin Lin and Tzuu-Kang ChyiInstitute of Physics, National Chiao-Tung University, Hsinchu, Taiwan, Republic of China ~Received 2 November 1998; published 26 May 1999!
We introduce the Vilkovisky-DeWitt formalism for deriving the lower bound of the Higgs boson mass. We illustrate the formalism with a simplified version of the standard electroweak model, where all charged boson fields as well as the bottom-quark field are disregarded. The effective potential obtained in this approach is gauge independent. We derive from the effective potential the mass bound of the Higgs boson. The result is compared to its counterpart obtained from the ordinary effective potential.@S0556-2821~99!06211-6# PACS number~s!: 11.15.Ex, 12.15.Ji, 14.80.Bn
I. INTRODUCTION
The effective potentials in quantum field theories are off-shell quantities. Therefore, in gauge field theories, effective potentials are gauge-dependent as pointed out by Jackiw in the early 1970s @1#. This property caused concerns on the physical significance of effective potentials. In a work by Dolan and Jackiw@2#, the effective potential of scalar QED was calculated in a set of Rj gauges. It was concluded that only the limiting unitary gauge gives a sensible result on the symmetry-breaking behavior of the theory. This difficulty was partially resolved by the work of Nielsen @3#. In his paper, Nielsen derived the following identity governing the behavior of effective potential in a gauge field theory:
S
j]j] 1C~f,j! ]]f
D
V~f,j!50, ~1! wherejis the gauge-fixing parameter,fis the order param-eter of the effective potential, and C(f,j) is the Green’s function for certain composite operators containing a ghost field. The above identity implies that, for different j, the local extrema of V are located along the same character-istic curve on the (f,j) plane, which satisfies dj 5df/@C(f,j)/j#. Hence covariant gauges with different j are equally good for computing V. On the other hand, a choice of the multi-parameter gauge Lg f52(1/2j)(]mAm 1sf11rf2)2 @2#, with f1,2 the components of the scalar field, would break the homogeneity of Eq.~1! @3#. Therefore an effective potential calculated in such a gauge does not have a physical significance.Recently, it was pointed out@4# that the Higgs boson mass bound, which one derives from the effective potential, is gauge dependent. The gauge dependence resides in the ex-pression for the one-loop effective potential. Boyanovsky, Loinaz and Willey proposed a resolution@5# to the problem, which is based upon the physical effective potential con-structed as the expectation value of the Hamiltonian in physi-cal states@6#. They computed the physical effective potential of an Abelian Higgs model with an axial vector coupling of the gauge fields to the fermions. A gauge-independent lower bound for the Higgs boson mass is then determined from the effective potential. We note that their approach requires the identification of first-class constraints of the model and a projection to the physical states. Such a procedure is not
manifestly Lorentz covariant. Consequently we expect that it is highly non-trivial to apply their approach to the standard model~SM!. In our work, we shall introduce the Vilkovisky-DeWitt formalism @7,8# for constructing a gauge-independent effective potential, and therefore obtain a gauge-independent lower bound for the Higgs boson mass.
In the Vilkovisky-DeWitt formalism, fields are treated as vectors in the configuration space, and the affine connection of the configuration space is identified to facilitate the con-struction of an invariant effective action. Since this proce-dure is completely Lorentz covariant, the computations for the effective potential and the effective action are straight-forward. We shall perform a calculation with respect to a toy model @9# which disregards all charged boson fields in the SM. It is easy to generalize our calculations to the full SM case. In fact, the applicability of Vilkovisky-DeWitt formal-ism to non-Abelian gauge theories has been extensively demonstrated in the literature@10#.
The outline of this paper is as follows. In Sec. II, we briefly review the Vilkovisky-DeWitt formalism using the scalar QED as an example. We shall illustrate that the effec-tive action of Vilkovisky and DeWitt is equivalent to the ordinary effective action constructed in the Landau-DeWitt gauge@11#. In Sec. III, we calculate the effective potential of the simplified standard model, and the relevant renormaliza-tion constants of the theory using the Landau-DeWitt gauge. The effective potential is then improved by the renormaliza-tion group analysis. In Sec. IV, the Higgs boson mass bound is derived and compared to that given by the ordinary effec-tive potential in the Landau gauge. We conclude in Sec. V, with some technical details discussed in the Appendix.
II. VILKOVISKY-DEWITT EFFECTIVE ACTION OF SCALAR QED
The formulation of Vilkovisky and DeWitt is motivated by the parametrization dependence of the ordinary effective action, which can be written generically as @12#
expi \ G@F#5exp i \
S
W@ j#1Fi dG dFiD
5E
@Df#exp\iS
S@f#2~fi2Fi!dG dFiD
, ~2! where S@f# is the classical action, and Fi denote the back-ground fields. The dependence on the parametrization arisesbecause the quantum fluctuation hi[(fi2Fi) is not a vector in the field configuration space, hence the product hi
•dG/dFi is not a scalar under a reparametrization of
fields. The remedy to this problem is to replace hi with a two-point function si(F,f) @7,8,13# which, at the point F, is tangent to the geodesic connectingF andf. The precise form ofsi(F,f) depends on the connection of the configu-ration space,Gjki . It is easy to show that@12#
si~F,f!5hi21
2Gj k
i hjhk1O~h3!. ~3! For scalar QED described by the Lagrangian:
L521 4FmnF
mn1~D
mf!†~Dmf!2l~f†f2m2!2, ~4!
with Dm5]m1ieAm andf5(f11if2)/
A
2, the connection of the configuration space is given by@7,12#Gj k i 5
H
ij k
J
1Tj ki
, ~5!
where$j ki%is the Christoffel symbol of the field configuration space and Tij k is a quantity induced by generators of the gauge transformation. The Christoffel symbol $jki% can be computed from the following metric tensor of scalar QED:
Gf
a(x)fb(y )5dabd
4~x2y!, GAm(x)An(y )52gmnd4~x2y!,
GAm(x)fa(y )50. ~6! According to Vilkovisky’s prescription@7#, the metric tensor of the field configuration space is obtained by differentiating twice with respect to the fields in the kinetic Lagrangian. For the above metric tensor, we have$j k
i%50 since each
compo-nent of the tensor is field independent. However, the Christ-offel symbol would be non-vanishing if one parametrizes Eq. ~4! with polar variables r andx such that f15rcosx and
f25rsinx. Finally, to determine Tj k
i , let us specify
genera-tors gai of the scalar-QED gauge transformations: gyfa(x)52eabef
b~x!d4~x2y!,
gyAm(x)52]md4~x2y!, ~7! where eab is a skew-symmetric tensor with e1251. The quantity Tij kis related to the generators gai via@7#
Tj ki 52BjaDkgai11 2ga lD lgb iB j aB k b1 j↔k, ~8!
where Bka5Nabgkb with Nab being the inverse of Nab [gakg
b
lG
kl. The expression for Tj k
i can be easily understood
by realizing that i, j, . . . ,l are function-space indices, while a andb are space-time indices. Hence, for example,
Df 1(z)gy Am(x) 5]gy Am(x) ]f1~z!1
H
Am~x! jf1~z!J
gy j , ~9!where the summation over j also implies an integration over the space-time variable in the function j.
The one-loop effective action of scalar QED can be cal-culated from Eq. ~2! with each quantum fluctuation hi re-placed bysi(F,f). The result is written as@12#:
G@F#5S@F#2i2\lndetG1i\
2 lndetD˜i j
21, ~10!
where S@F# is the classical action with F denoting generi-cally the background fields; lndetG arises from the function-space measure @Df#[)xdf(x)
A
detG; and D˜i j21 is the modified inverse propagator:D˜i j215 d 2S
dFidFj2Gi j k@F# dS
dFk. ~11!
To study the symmetry-breaking behavior of the theory, we focus on the effective potential which is obtained fromG@F# by setting each background fieldFi to a constant.
The Vilkovisky-DeWitt effective potential of scalar QED has been calculated in various gauges and different scalar-field parametrizations @11,12,14#. The results all agree with one another. In this work, we calculate the effective potential and other relevant quantities in the Landau-DeWitt gauge @15#, which is characterized by the gauge-fixing term: Lg f 52(1/2j)(]mBm2ieh†F1ieF†h)2, with j→0. In Lg f,
Bm[Am2Aclm, and h[f2F are quantum fluctuations while Aclm andF are background fields. For the scalar fields, we further write F5(rcl1ixcl)/
A
2 and h5(r1ix)/A
2.The advantage of performing calculations in the Landau-DeWitt gauge is that Tjki vanishes@11# in this case. In other words, the Vilkovisky-DeWitt formalism coincides with the conventional formalism in the Landau-DeWitt gauge.
For computing the effective potential, we choose Aclm 5xcl50, i.e., F5rcl/
A
2. In this set of background fields,Lg f becomes Lg f52 1 2j~]mB m] nBn22erclx]mBm1e2rcl 2x2!. ~12! We note that Bm2x mixing in Lg f is j dependent, and
therefore would not cancel out the corresponding mixing term in the classical Lagrangian of Eq. ~4!. This induces mixed-propagators such as
^
0uT„Am(x)x(y )…u0&
or^
0uT„x(x)Am(y )…u0&
. The Faddeev-Popov ghost Lagrangian in this gauge readsLF P5v*~2]22e2rcl 2!
v. ~13!
With each part of the Lagrangian determined, we are ready to compute the effective potential. Since we choose a field-independent flat-metric, the one-loop effective potential is completely determined by the modified inverse-propagators D˜i j21 @16#. From Eqs. ~4!, ~11!, ~12! and ~13!, we arrive at
D˜B mBn 21 5~2k21e2r 0 2!gmn1
S
121 jD
kmkn, D ˜ Bmx 215ikmer 0S
12 1 jD
, D ˜ xx 215S
k22m G 221 je2r0 2D
, D˜rr215~k22mH2!, D˜v*v5~k22e2r0 2!22 , ~14!where we have set rcl5r0, which is a space-time independent constant, and defined mG25l(r0222m2), mH25l(3r0222m2). Using the definition G@r0# 52(2p)4d4(0)Ve f f(r0) along with Eqs.~10! and ~14!, and taking the limit j→0, we obtain Ve f f(r0)5Vtree(r0)
1V1-loo p(r0) with V1-loo p~r0!5 2i\ 2
E
dnk ~2p!nln@~k22e2r0 2!n23 3~k22m H 2!~k22m 1 2!~k22m 2 2!#, ~15! where m12 and m22 are solutions of the quadratic equation (k2)22(2e2r021mG2)k21e4r0450. One notices that the gauge-boson’s degree of freedom in V1-loo phas been contin-ued to n23 in order to preserve the relevant Ward identities. For example, this continuation is crucial to ensure the Ward identity which relates the scalar self-energy to the con-tribution of the tadpole diagram. Our expression forV1-loo p(r0) agrees with previous results obtained in the
unitary gauge @14#. One could also calculate the effect-ive potential in the ghost-free Lorentz gauge with Lg f5 2(1/2j)(]mBm)2. The cancellation of the gauge-parameter (j) dependence in the effective potential has been demon-strated in the case of massless scalar QED where m250 @11,12#. It can be easily extended to the massive case, and the resultant effective potential coincides with Eq. ~15!. In the Appendix, we will also demonstrate the cancellation of gauge-parameter dependence in the calculation of Higgs-boson self-energy. The obtained self-energy will be shown to coincide with its counterpart obtained from the Landau-DeWitt gauge. We do this not only to show that the Vilkovisky-DeWitt formulation coincides with the ordinary formulation in the Landau-DeWitt gauge, but also to illus-trate how it gives rise to identical effective action in spite of beginning with different gauges.
It is instructive to rewrite Eq.~15! as
V1-loo p@r0#5 \ 2
E
dn21kW ~2p!n21„~n23!vB~kW!1vH~kW! 1v1~kW!1v2~kW!…, ~16! where vB(kW)5A
kW21e2r0 2 , vH(kW)5A
kW21mH 2 and v6(kW) 5A
kW21m62. One can see that V1-loopis a sum of the zero-point energies of four excitations with masses mB[er0, mH, m1 and m2. Since there are precisely four physical degrees of freedom in the scalar QED, we see that the Vilkovisky-DeWitt effective potential does exhibit a correct number of physical degrees of freedom. Such a nice property is not shared by the ordinary effective potential calculated in the ghost free Lorentz gauge just mentioned. As will be shown later, the ordinary effective potential in this gauge contains unphysical degrees of freedom.III. VILKOVISKY-DEWITT EFFECTIVE POTENTIAL OF THE SIMPLIFIED STANDARD MODEL
In this section, we compute the effective potential of the simplified standard model where charged boson fields and all fermion fields except the top-quark field are disregarded. The gauge interactions for the top quark and the neutral scalar bosons are prescribed by the following covariant derivatives @9#: DmtL5
S
]m1igLZm2 2 3ieAmD
tL, DmtR5S
]m1igRZm22 3ieAmD
tR, Dmf5„]m1i~gL2gR!Zm…f, ~17! where Zmand Amdenote the Z boson and the photon respec-tively; the coupling constants gL and gR are given by gL 5(2g1/21g2/3) and gR5g2/3 with g15g/cosuW and g2 52etanuWrespectively. The self-interactions of scalar fieldsare described by the same potential term as that in Eq. ~4!. Clearly this toy model exhibits a U(1)A3U(1)Z symmetry where each U(1) symmetry is associated with a neutral gauge boson. The U(1)Zcharges of tL, tR andfare related in such a way that the following Yukawa interactions are invariant under U(1)A3U(1)Z:
LY52y t¯LftR2y t¯Rf*tL. ~18!
Since Vilkosvisky-DeWitt effective action coincides with the ordinary effective action in the Landau-DeWitt gauge, we thus calculate the effective potential in this gauge, which is defined by the following gauge-fixing terms @17#:
Lg f52 1 2a
S
]m˜Z m1ig1 2 h †F2ig1 2 F †hD
2 221b~]mA˜m!2, ~19! with a,b→0. We note that A˜m, Z˜m and h are quantum fluctuations associated with the photon, the Z boson and the scalar boson respectively, i.e., Am5Aclm1A˜m, Zm5Zclm1Z˜m, and f5F1h with Aclm, Zclm and F being the background fields. For computing the effective potential, we takeF as a space-time-independent constant denoted as ro, and set Aclm5Zclm50. Following the method outlined in the previous
sec-tion, we obtain the one-loop effective potential
VVD~r0!5\ 2
E
dn21kW ~2p!n21„~n23!vZ~kW! 1vH~kW!1v1~kW!1v2~kW!24vF~kW!…, ~20! where vi(kW)5A
kW21m i 2 with m Z 25(g 1 2/4)r 0 2, m 6 25m Z 2 11 2(mG 26m GA
mG 214m Z 2) and mF2[mt25y2r02/2. The Gold-stone boson mass mG is defined as before, i.e., mG
25l(r
o 2 22m2) withmbeing the mass parameter of the Lagrangian. One may notice the absence of photon contributions in the above effective potential. This is not surprising since photons do not couple directly to the Higgs boson.
Performing the integration in Eq.~20! and subtracting the infinities with modified minimal subtraction ~MS! scheme prescription, we obtain VVD~r0!5 \ 64p2
S
mH 4 lnmH 2 k21mZ 4 lnmZ 2 k21m1 4 lnm1 2 k2 1m24lnm2 2 k2 24mt 4lnmt 2 k2D
2 \ 128p2 3~3mH 415m Z 413m G 4112m G 2 mZ2212mt4!, ~21! where k is the mass scale introduced in the dimensional regularization. Although VVD(r0) is obtained in the Landau-DeWitt gauge, we should stress that any other gauge with non-vanishing Tijk should lead to the same result. For later comparisons, let us write down the ordinary one-loop effec-tive potential in the Lorentz gauge@removing the scalar part of Eq.~19!# as follows @2#: VL~r0!5 \ 2E
dn21kW ~2p!n21„~n21!vZ~kW!1vH~kW! 1va~kW,a!1vb~kW,a!24vF~kW!…, ~22!where a is the gauge-fixing parameter and va,b(kW,a) 5
A
kW21ma,b2 (a) with ma2(a)51/2•(mG2 1A
mG424amZ2m2G) and mb2(a)51/2•(mG2 2A
mG424amZ2mG2). It is easily seen that there are 6 bosonic degrees of freedom in VL, i.e., two extra degrees of freedom emerge as a result of choosing the Lorentz gauge. In the Landau gauge, which is a special case of the Lorentz gauge witha50, there is still one extra degree of freedom in the effective potential @6#. Since the Landau gauge is adopted most frequently for computing the ordinary effective poten-tial, we shall takea50 in VL hereafter. Performing thein-tegrations in VL and subtracting the infinities, we obtain
VL~r0!5 \ 64p2
S
mH 4 lnmH 2 k213mZ 4 lnmZ 2 k2 1mG 4lnmG 2 k224mt 4lnmt 2 k2D
2128\p2~3mH 415m Z 413m G 4212m t 4!. ~23! We remark that VL differs from VVD except at the point ofextremum wherer0252m2. At this point, one has mG250 and m625mZ2, which lead to VVD(r052m2)5VL(r0
252m2).
That VVD5VL at the point of extremum is a consequence of
the Nielsen identity@3# mentioned earlier.
To derive the Higgs boson mass bound from VVD(r0) or VL(r0), one encounters a breakdown of the perturbation theory at the point of sufficiently large r0 such that, for instance, (l/16p2)ln(lr02/k2).1. To extend the validity of the effective potential to the large-r0 region, the effective potential has to be improved by the renomalization group ~RG! analysis. Let us denote the effective potential as Ve f f
which includes the tree-level contribution and quantum cor-rections. The renormalization-scale independence of Ve f f
im-plies the following equation @18,5#:
S
2m~gm11!]m] 1bgˆ ] ]gˆ2~gr11!t ] ]t14D
3Ve f f~tr0 i ,m,gˆ,k!50, ~24!wherem is the mass parameter of the Lagrangian as shown in Eq.~4!, and bgˆ5k dgˆ dk, gr52k dlnr dk , gm52k dlnm dk , ~25!
with gˆ denoting collectively the coupling constantsl, g1,g2 and y;r0i is an arbitrarily chosen initial value forr0. Solving this differential equation gives
Ve f f~tr0 i ,mi,gˆi,k!5exp
S
E
0 lnt 4 11gr~x!dxD
3Ve f f„r0 i,m~t,m i!,gˆ~t,gˆi!,k…, ~26! with x5ln(r08
/r0 i) for an intermediate scaler0
8
, and tdgˆdt5
bgˆ„gˆ~t!…
m~t,mi!5miexp
S
2E
0lnt11gm~x!
11gr~x!dx
D
. ~28! To fully determine Ve f f at a larger0, we need to calculate the b functions of l, g1, g2 and y, and the anomalous di-mensions gm and gr. It has been demonstrated that the n-loop effective potential is improved by the (n11)-loopb andg functions @19,20#. Since the effectve potential is cal-culated to the one-loop order, a consistent RG analysis re-quires the knowledge ofb andg functions up to a two-loop accuracy. However, as the computations of two-loopb and g functions are quite involved, we will only improve the tree-level effective potential with one-loop b and g func-tions. After all, the main focus of this paper is to show how to obtain a gauge-independent Higgs boson mass bound rather than a detailed calculation of this quantity.To compute one-loopbandgfunctions, we first calculate the renormalization constants Zl,Zg1, Zg2,Zy,Zm2 and Zr,
which are defined by lbare5Z ll, g1 bare5Z g1g1, g2 bare5Z g2g2, ybare5Zyy , ~m2!bare5Zm2m2, rbare5
A
Zrr.~29! In the ordinary formalism of the effective action, all of the above renormalization constants except Zr are in fact gauge-independent at the one-loop order in the MS scheme. For Zr, the result given by the commonly adopted Landau gauge differs from that obtained from the Landau-DeWitt gauge. In Appendix A, we shall reproduce Zrobtained in the Landau-DeWitt gauge with the general Vilkovisky-DeWitt formulation. The calculation of various renormalization con-stants are straightforward. In the MS scheme, we have ~we will set\51 from this point on!
Zl512 1 128p2e
8
S
3g14 l 224g1 2216y 4 l 132y21160lD
, Zg15Zg2512 1 216p2e8
S
27g12 8 12g2 223g 1g2D
, Zy511 1 192p2e8
~9g1 214g 1g2224y2!, Zm2511 1 128p2e8
S
3g14 l 212g1 2216y 4 l 116y2196lD
, Zr511 1 32p2e8
~25g1 214y2!, ~30! where 1/e8
[1/e112 gE2 12 ln(4p) withe5n24. The
one-loopb and g functions resulting from the above renormal-ization constants are
bl5 1 16p2
S
3 8g1 423lg 1 222y414ly2120l2D
, bg15 g1 4p2S
g12 162 g1g2 18 1 g22 27D
, bg25 g2 4p2S
g12 162 g1g2 18 1 g22 27D
, by5 y 8p2S
y 223g1 2 8 1 g1g2 12D
, gm5 1 2p2S
3l 4 1 3g14 128l 2 3g12 32 2 y4 8l 1 y2 8D
, gr5 1 64p2~25g1 214y2!. ~31!Similar to what was mentioned earlier, all of the above quan-tities are gauge-independent in the MS scheme except gr, the anomalous dimension of the scalar field. In the Landau gauge of the ordinary formulation, we have
gr5
1
64p2~23g1
214y2!. ~32!
IV. THE HIGGS BOSON MASS BOUND
The lower bound of the Higgs boson mass can be derived from the vacuum instability condition of the electroweak ef-fective potential@21#. In this derivation, there exists different criteria for determining the instability scale of the elec-troweak vacuum. The first criterion is to identify the insta-bility scale as the critical value of the Higgs-field strength beyond which the renormalization-group- ~RG-! improved tree-level effective potential becomes negative @22–24#. To implement this criterion, the tree-level effective potential is improved by the leading @24# or next-to-leading order @22,23# renormalization group equations, where one-loop or two-loopb andgfunctions are employed. Furthermore, one-loop corrections to parameters of the effective potential are also taken into account @23,24#. However, the effect of one-loop effective potential is not considered.
To improve the above treatment, Casas et al.@25# consid-ered the effect of RG-improved one-loop effective potential. The vacuum-instability scale is then identified as the value of the Higgs-field strength at which the sum of tree-level and one-loop effective potentials vanishes. In our subsequent analysis, we will follow this criterion except that the one-loop effective potential is not RG improved.
To derive the Higgs boson mass bound, one begins with Eq. ~26! which implies
Vtree~tr0i,mi,li!51
4x~t!l~t,li!„~r0
i!222m2~t,m i!…2,
with x(t)5exp„*0lnt@4/11gr(x)#dx…. Since Eq. ~28! implies that m(t,mi) decreases as t increases, we then have Vtree(tr0i,mi,li)'1
4x(t)l(t,li)(r0
i)4 for a sufficiently large t. Similarly, the one-loop effective potential V1-loop(tr0
i
,mi,gˆi,k) is also proportional to
V1-loop„r0 i,m(t,m
i),gˆ(t,gˆi),k… with the same proportional
constantx(t). Because we shall ignore all running effects in
V1-loop, we can take gˆ(t,gˆi)5gˆi and m(t,mi)5(1/t)mi in
V1-loop. For a sufficiently large t, V1-loo p can also be
ap-proximated by its quartic terms. In the Landau-DeWitt gauge with the choicek5r0i, we obtain
VVD'~r0 i!4 64p2
F
9li 2ln~3l i!1 g1i4 16lnS
g1i2 4D
2yi 4 lnS
yi 2 2D
1A1 2 ~g1i,li!lnA1~g1i,li! 1A22~g 1i,li!lnA2~g1i,li! 232S
10li21lig1i 21 5 48g1i 42y i 4D
G
, ~34! where A6(g1,l)5g12/41l/2•(16A
11g12/l). Similarly, the effective potential in the Landau gauge is given byVL' ~r0 i!4 64p2
F
9li 2 ln~3li!1 3g1i4 16 lnS
g1i2 4D
2yi 4 lnS
yi 2 2D
1li 2 ln~li!2 3 2S
10li 21l ig1i 21 5 48g1i 42y i 4D
G
. ~35! Combining the tree-level and the one-loop effective poten-tials, we arrive at Ve f f~tr0 i ,mi,gˆi,k!' 1 4x~t!„l~t,li!1Dl~gˆi!…~r0 i!4, ~36! whereDl represents the one-loop corrections obtained from Eqs. ~34! or ~35!. Let tVI5rVI/r0i
, the condition for the vacuum instability of the effective potential is then @25#
l~tVI,li!1Dl~gˆi!50. ~37!
We note that the couplings gˆi in Dl are evaluated at k 5r0
i
, which can be taken as the electroweak scale. Hence we have g1i[g/cosuW50.67, g2i[2etanuW50.31, and yi 51. The running coupling l(tVI,li) also depends upon
g1,g2and y throughbl, andgr shown in Eq.~31!. To solve Eq. ~37!, we first determine the running behaviors of the coupling constants g1,g2 and y. For g1 and g2, we have
td„gl 2 ~t!… dt 52gl~t! bgl„gˆ~t!… 11gr„gˆ~t!…'bgl 2, ~38!
where l51,2, and the contribution ofgr is neglected in ac-cordance with our leading-logarithmic approximation. Also bg
l
25gl2/2p2•(g12/162g1g2/181g22/27). Although the
dif-ferential equations for g12 and g22 are coupled, they can be easily disentangled by observing that g12/g22 is a RG-invariant. Numerically, we have bg
l
25clgl4 with c152.3
31023 and c
251.131022. Solving the differential equa-tions gives
gl22~t!5gl22~0!2cllnt. ~39!
With g1(t) and g2(t) determined, the running behavior of y can be calculated analytically @4#. Given by2[2yby5c3y4
2c4g1 2
y2with c352.531022and c458.531023, we obtain
y2~t!5
FS
g1 2 ~t! g1i2D
c4/c1S
yi222 c3 c11c4g1i22D
1 c3 c11c4 g122~t!G
21 . ~40!Now the strategy for solving Eq.~37! is to make an initial guess onli, which enters intol(t) and Dl, and repeatedly
adjustliuntill(t) completely cancels Dl. For tVI5102~or
r0'104 GeV) which is the new-physics scale reachable by the CERN Large Hadron Collider ~LHC!, we find li54.83 31022 for the Landau-DeWitt gauge, and l
i54.831022
for the Landau gauge. For a higher instability scale such as the scale of grand unification, we have tVI51013 or r0 '1015 GeV. In this case, we find l
i53.1331021 for both
the Landau-DeWitt and Landau gauges. The numerical simi-larity between the li of each gauge can be attributed to an
identical b function for the running of l(t), and a small difference between the Dl of each gauge. We recall from Eq.~27! that the evolutions of l in the above two gauges will be different if the effects of next-to-leading logarithms are taken into account. In that case, the difference between the grof each gauge gives rise to different evolutions forl. For
a large tVI, one may expect to see a non-negligible
differ-ence between theliof each gauge.
The critical valueli54.8331022corresponds to a lower
bound for the MS mass of the Higgs boson. Since mH 52
A
lm, we have (mH)MS>77 GeV. For li53.1331021,we have (mH)MS>196 GeV. To obtain the lower bound for the physical mass of the Higgs boson, finite radiative correc-tions must be added to the above bounds @4#. We will not pursue these finite corrections any further since we are sim-ply dealing with a toy model. However we would like to point out that such corrections are gauge-independent as en-sured by the Nielsen identity @3#.
V. CONCLUSION
We have computed the one-loop effective potential of an Abelian U(1)3U(1) model in the Landau-DeWitt gauge, which reproduces the result given by the gauge-independent Vilkovisky-DeWitt formulation. One-loopb andg functions
were also computed to facilitate the RG improvement of the effective potential. A gauge-independent lower bound for the Higgs-boson self-coupling or equivalently the MS mass of the Higgs boson was derived. We compared this bound to that obtained using the ordinary Landau-gauge effective po-tential. The numerical values of both bounds are almost iden-tical due to the leading-logarithmic approximation we have taken. A complete next-to-leading-order analysis should bet-ter distinguish the two bounds. This improvement as well as extending the current analysis to the full standard model will be reported in future publications.
Finally we would like to comment on the issue of com-paring our result with that of Ref. @5#. So far, we have not found a practical way of relating the effective potentials cal-culated in both approaches. In Ref.@5#, to achieve a gauge-invariant formulation, the theory is written in terms of a new set of fields which are related to the original fields through non-local transformations. Taking scalar QED as an ex-ample, the new scalar field f
8
(xW) is related to the original field through@6#f
8
~xW!5f~xW!expS
ieE
d3y AW~yW!•¹WyG~yW2xW!D
, ~41! with G( yW2xW) satisfying ¹2G(yW2xW)5d3(yW2xW). To our knowledge, it does not appear obvious how one might incor-porate the above non-local and non-Lorentz-covariant trans-formation into the Vilkovisky-DeWitt formulation. This is an issue deserving further investigations.ACKNOWLEDGMENTS
We thank W.-F. Kao for discussions. This work is sup-ported in part by National Science Council of R.O.C. under grant numbers NSC 87-2112-M-009-038, and NSC 88-2112-M-009-002.
APPENDIX: THE HIGGS-BOSON SELF-ENERGY AND VILKOVISKY-DEWITT EFFECTIVE ACTION
In this Appendix, we calculate the Higgs-boson self-energy of scalar QED from the Vilkovisky-DeWitt effective action. We will focus on the momentum-dependent part of the self-energy, which is not a part of the effective potential calculated in Sec. II. Furthermore only the infinite part of the self-energy will be calculated. We thus perform the calcula-tion in the symmetry phase of the theory.
We begin with the Lagrangian in Eq. ~4! where m2 is negative, i.e.,2m2[u2.0. In this case,f
1andf2 have an identical mass mf252lu2. Let us renamef1asr andf2as
x according to our notation in the symmetry-broken phase. If one follows the background field expansion in Eq. ~2!, one would expand the QED action by writing Am5Aclm1Bm, r 5rcl1h1, andx5xcl1h2, with Aclm, rcl andxcl the
clas-sical background fields, and Bm, h1 andh2 the correspond-ing quantum fluctuations. However, as mentioned earlier, the above quantum fluctuations should be replaced by vectorssi
in the configuration space. Hence the action in Eq.~2! should be expanded covariantly@7# in powers ofsi. To simplify our
notations, we use B˜m and h˜i to denote the new quantum
fluctuations. Since we will only calculate the self-energy of r, we may take Aclm5xcl50 for simplicity. With the
covari-ant expansion, the Lagrangian in Eq. ~4! generates the fol-lowing quadratic terms:
Lquad52 1 4~]mB˜n2]nB˜m! 211 2~]mh˜1!~] mh˜ 1! 11 2~]mh˜2!~] mh˜2!1ercl~]mh˜2!B˜m2eh˜2~]mrcl!B˜m 11 2e 2r cl 2B˜mB˜ m2l
F
1 2rcl 2~3h˜ 1 21h˜ 2 2!1u2~h˜ 1 21h˜ 2 2!G
2GFFmFn l dS dFlf˜mf˜n, ~A1!whereFlandf˜ldenote generically the classical background fields and the quantum fluctuations respectively. We choose the Rj background-field gauge with the gauge-fixing term:
Lg f52 1 2a~]mB˜
m2aer
clh˜2!2. ~A2! The corresponding Faddeev-Popov Lagrangian is then
LF P5v*~2]22ae2rcl2!v. ~A3! Compared to the usual background-field formalism, the qua-dratic quantum flucuations Lquadcontain extra terms
propor-tional to the connectionGij kof the configuration space. These extra terms are crucial for the cancellation of gauge-parameter dependence in the Higgs-boson self-energy. From Eqs. ~5!, ~6!, ~7! and ~8!, we calculate those connections which are relevant to the Higgs-boson self-energy. We find
GAr(z)m(x)An( y )uF52e2rcl~z!~]xmNxz!~]ynNy z! Gx(x)x(y)r(z) uF5e2Nxy„d4~y2z!1d4~x2z!…rcl~z!
2e4r
cl~z!NzxNzyrcl~x!rcl~y! GAr(z)m(x)x(y)uF5e~]mxNzx!d4~z2y!2e3rcl~z!
3~]xmN xz!Nz yr cl~y!, ~A4! where Nx y5
K
xU
1 ]21e2r cl 2 ~X!U
yL
with Xmux
&
5xmux&
; and the notationuF denotes evaluating the connection at the classical background fields. The above connections are to be multiplied by dS/druF[(2]2 22lu22lrcl 2
)rcl with the space-time variable z
inte-grated over. It is interesting to note that the product ofG and dS/dr contain terms which are able to generate the Higgs-boson self-energy. For example, in the expression 2*d4xd4y (G
Am(x)An( y )
r(z) dS/dru
rcl50 in Nxzand Ny z and contract the pair of gauge fields.
This gives rise to, in the momentum space, the following Higgs-boson self-energy: SrAA~p2!52ae 2 8p2 1 e
8
p 2, ~A5! where 1/e8
[1/e11 2 gE2 12ln(4p) with e5n24. For 2*d4xd4y (G x(x)x(y) r(z) dS/dru F)h˜2(x)h˜2(y ), we again set rcl 50 in Nx y and contract the pair of scalar fields. We obtainthe Higgs-boson self-energy Srxx~p2!5 e 2 4p2 1 e
8
p 2. ~A6!Finally, the term 2*d4xd4y (GAr(z)m(x)x(y)dS/
druF)B˜m(x)h˜2(y ) can produce an effective rcl2B˜m2h˜2 vertex, namely, @*d4pd4k/(2p)8#Gm( p,k)rcl(k)B˜m( p)h2 (2p2k), with Gm( p,k)5i(pm/ p2)•(k222lu2). This ver-tex can contribute to the Higgs-boson self-energy by con-tracting with another vertex of the same kind. Similarly, it could contract with an ordinaryrcl2B˜m2h˜2 vertex. It turns out that both contractions produce only finite contributions to the momentum-dependent part of the Higgs-boson self-energy. Therefore Eqs.~A5! and ~A6! are the only divergent contributions ofGj ki to the momentum-dependent part of the
Higgs-boson self-energy. These contributions are to be added to the self-energy obtained by contracting a pair of ordinaryrcl2B˜m2h˜2 vertices. We find
Srordinary~p2!5 e 2 8p2 1 e
8
~31a!p 2. ~A7!From Eqs.~A5!, ~A6! and ~A7!, we arrive at Sr~p2!5SrAA~p2!1Srxx~p2!1Srordinary~p2!5 5e2 8p2 1 e
8
p 2. ~A8! We can see that the gauge-parameter dependence of Srordinary( p2) is cancelled by that ofSrAA( p2). From Eq.~A8!, the wave-function renormalization constant of the Higgs bo-son is found to be Zr5125e 2 8p2 1 e
8
. ~A9!This result can be applied to the model in Sec. III with the replacement e→2g1/2 according to Eq.~17!. Hence Zr51 2(5g1
2
/32p2)(1/e
8
) in that model, which reproduces the rel-evant part of Eq. ~30! calculated in the Landau-DeWitt gauge.@1# R. Jackiw, Phys. Rev. D 9, 1686 ~1974!.
@2# L. Dolan and R. Jackiw, Phys. Rev. D 9, 2904 ~1974!. @3# N.K. Nielsen, Nucl. Phys. B101, 173 ~1975!.
@4# W. Loinaz and R.S. Willey, Phys. Rev. D 56, 7416 ~1997!. @5# D. Boyanovsky, W. Loinaz, and R.S. Willey, Phys. Rev. D 57,
100~1998!.
@6# D. Boyanovsky, D. Brahm, R. Holman, and D.-S. Lee, Phys.
Rev. D 54, 1763~1996!.
@7# G. Vilkovisky, in Quantum Theory of Gravity, edited by S.M.
Christensen ~Hilger, Bristol, 1983!; Nucl. Phys. B234, 215
~1984!.
@8# B.S. DeWitt, in Quantum Field Theory and Quantum Statis-tics: Essays in Honour of the 60th Birthday of E. S. Fradkin,
edited by I.A. Batalin, C.J. Isham, and G.A. Vilkovisky
~Hilger, Bristol, 1987!, p. 191.
@9# G.-L. Lin, H. Steger, and Y.-P. Yao, Phys. Rev. D 44, 2139 ~1991!.
@10# A. Rehban, Nucl. Phys. B288, 832 ~1987!; B298, 726 ~1988!. @11# E.S. Fradkin and A.A. Tseytlin, Nucl. Phys. B234, 509 ~1984!. @12# See, for example, G. Kunstatter, in Super Field Theories,
ed-ited by H.C. Lee, V. Elias, G. Kunstatter, R.B. Mann, and K.S. Viswanathan~Plenum, New York, 1987!, p. 503.
@13# We should point out that the simple replacement of hi with si
(F,f), as suggested in Ref. @7#, is not satisfactory for cal-culations beyond the one-loop order. An effective action ob-tained by such a prescription does not generate one-particle irreducible diagrams at the higher loops. A modified
construc-tion was proposed by DeWitt in Ref.@8#. However, since we are only concerned with one-loop corrections, we shall adhere to the current construction.
@14# I.H. Russell and D.J. Toms, Phys. Rev. D 39, 1735 ~1989!. @15# Although properties of this gauge are discussed in Ref. @11#,
the authors of this paper used another gauge to calculate the effective potential of scalar QED.
@16# In the current gauge, D˜i j215Di j21 sinceGjk i
vanishes. Further-more, for ghost fields, it is the ghost propagator, rather than its inverse, that will appear in the effective action. This is due to the Grassmannian nature of ghost fields.
@17# We have suppressed the fermionic part of the gauge-fixing
term. The justification of this simplified treatment is discussed in the first paper of Ref.@10#.
@18# S. Coleman and E. Weinberg, Phys. Rev. D 7, 1888 ~1973!. @19# B. Kastening, Phys. Lett. B 283, 287 ~1992!.
@20# M. Bando, T. Kugo, N. Maekawa, and H. Nakano, Phys. Lett.
B 301, 83~1993!.
@21# For a review on earlier works, see M. Sher, Phys. Rep. 179,
273~1989!.
@22# C. Ford, D.R.T. Jones, P.W. Stephenson, and M.B. Einhorn,
Nucl. Phys. B395, 17~1993!.
@23# Marc Sher, Phys. Lett. B 317, 59 ~1993!; B331, 448 ~1994!. @24# G. Altarelli and G. Isidori, Phys. Lett. B 337, 141 ~1994!. @25# J.A. Casas, J.R. Espinosa, and M. Quiro´s, Phys. Lett. B 342,