Sub-signature operators and the Dabrowski-Sitarz-Zalecki type theorems

for manifolds with boundary

Hongfeng Li, Yong Wang∗

School of Mathematics and Statistics, Northeast Normal University, Changchun, 130024, China
arXiv:2311.03589v1 [math.DG] 31 Oct 2023

Abstract
In this paper, we define the spectral Einstein functional associated with the sub-signature operator for
manifolds with boundary. Motivated by the spectral Einstein functional and the sub-signature operator,
we relate them to the noncommutative residue for manifolds with boundary. And we give the proof of the
Dabrowski-Sitarz-Zalecki type theorems for the spectral Einstein functional associated with the sub-signature
operator on 4-dimensional manifolds with boundary.
Keywords: Sub-signature operators; spectral Einstein functional; the Dabrowski-Sitarz-Zalecki type
theorems.

1. Introduction

An eminent spectral scheme that generates geometric objects on manifolds such as residue, scalar cur-
vature, and other scalar combinations of curvature tensors is the small-time asymptotic expansion of the
(localised) trace of heat kernel[1, 2]. The theory has very rich structures both in physics and mathematics.
Recently, Dabrowski etc. [3] defined bilinear functionals of vector fields and differential forms, the densities
of which yield the metric and spectral Einstein functionals on even-dimensional Riemannian manifolds, and
they obtained certain values or residues of the (localised) zeta function of the Laplacian arising from the
Mellin transform and the coefficients of this expansion.
Let E be a finite-dimensional complex vector bundle over a closed compact manifold M of dimension n,
the noncommutative residue of a pseudo-differential operator P ∈ ΨDO(E) can be defined by
Z
−n P
res(P ) := (2π) Tr(σ−n (x, ξ))dxdξ, (1.1)
S∗ M

where S ∗ M ⊂ T ∗ M denotes the co-sphere bundle on M and σ−n

P
is the component of order −n of the
complete symbol X
σ P := σiP (1.2)
i

of P , cf. [4–7], and the linear functional res : ΨDO(E) → C is in fact the unique trace (up to multiplication
by constants) on the algebra of pseudo-differential operators ΨDO(E). In [8], Connes used the noncom-
mutative residue to derive a conformal 4-dimensional Polyakov action analogy. Connes proved that the
noncommutative residue on a compact manifold M coincided with Dixmier’s trace on pseudo-differential
operators of order -dimM in [9]. And Connes claimed that the noncommutative residue of the square
of the inverse of the Dirac operator was proportioned to the Einstein-Hilbert action. Kastler[10] gave a
brute-force proof of this theorem, and Kalau and Walze[11] proved this theorem in the normal coordinates

∗ Corresponding author.
Email addresses: [email protected] (Hongfeng Li), [email protected] (Yong Wang)

Preprint submitted to Elsevier November 8, 2023

system simultaneously, which is called the Kastler-Kalau-Walze theorem now. Based on the theory of the
noncommutative reside introduced by Wodzicki, Fedosov etc.[12] constructed a noncommutative residue on
the algebra of classical elements in Boutet de Monvel’s calculus on a compact manifold with boundary of
dimension n > 2. With elliptic pseudo-differential operators and noncommutative residue, it’s a natural
way to study the Kastler-Kalau-Walze type theorem and operator-theoretic explanation of the gravitational
action for manifolds with boundary.
On the other hand, Wang generalized the Connes’ results to the case of manifolds with boundary in
[13, 14], and proved the Kastler-Kalau-Walze type theorem for the Dirac operator and the signature opera-
] + D−1 ◦ π + D−1 ]
tor on lower-dimensional manifolds with boundary [15]. In [15, 16], Wang computed Wres[π
] + D−2 ◦ π + D−2 ], where the two operators are symmetric, in these cases the boundary term van-
and Wres[π
] + D−1 ◦ π + D−3 ], Wang got a nonvanishing boundary term [17], and gave a theoretical
ished. But for Wres[π
explanation for gravitational action on boundary. In other words, Wang provides a kind of method to study
the Kastler-Kalau-Walze type theorem for manifolds with boundary. In [18], the authors got the spectral
Einstein functional associated with Dirac operators with torsion on compact manifolds with boundary. For
lower dimensional compact Riemannian manifolds with boundary, they computed the lower dimensional
residue of ∇ e Y D−4 and get the Dabrowski-Sitarz-Zalecki theorems. Motivated by [3, 18], we define the
e X∇
T
spectral Einstein functional associated with the sub-signature operator for manifolds with boundary, and
] + (∇A
the motivation of this paper is to compute the noncommutative residue Wres[π A −2 +
X ∇Y DA ) ◦ π (DA )]
−2

] + (∇A
and Wres[π A −1 + −3
X ∇Y DA ) ◦ π (DA )] on 4-dimensional compact manifolds, where DA is the sub-signature
operator.
The paper is organized in the following way. In Sec.2 and 3, we define the spectral Einstein functional
associated with the sub-signature operator and get the noncommutative residue for manifolds without bound-
ary. In Sec.4 and 5, we prove the Dabrowski-Sitarz-Zalecki type theorems for the spectral Einstein functional
associated with the sub-signature operator on 4-dimensional manifolds with boundary.

2. The spectral functional for the sub-signature operator

Firstly, we introduce some notations about the sub-signature operator. Let M be an n-dimensional
(n ≥ 3) oriented compact Riemannian manifold with a Riemannian metric g T M . And let F be a subbundle
of T M , F ⊥ be the subbundle of T M orthogonal to F . Then we have the following orthogonal decomposition:
M
TM = F F ⊥;
M ⊥
gT M = gF gF , (2.1)
⊥
where g F and g F are the induced metric on F and F ⊥ .
Let ∇L denote the Levi-Civita connection about g T M . In the local coordinates xi ; 1 ≤ i ≤ n and the
fixed orthonormal frame {e1 , · · · , en }, the connection matrix (ωs,t ) is defined by

∇L (e1 , · · ·, en ) = (e1 , · · ·, en )(ωs,t ). (2.2)

Let ǫ(e∗j ), ι(e∗j ) be the exterior and interior multiplications respectively, where e∗j = g T M (ej , ·). Write

c(ej ) = ǫ(e∗j ) + ι(e∗j ); c(ej ) = ǫ(e∗j ) − ι(e∗j ),

b (2.3)

which satisfies

b c(ej ) + b
c(ei )b c(ei ) = 2g T M (ei , ej );
c(ej )b
c(ei )c(ej ) + c(ej )c(ei ) = −2g T M (ei , ej );
c(ej ) + b
c(ei )b c(ej )c(ei ) = 0. (2.4)

2
By [19], we have
n
X  
e =d+δ = 1X
D c(ei ) ei + ωs,t (ei )[b
c(es )b
c(et ) − c(es )c(et )] . (2.5)
i=1
4 s,t
⊥
Let π F (resp. π F ) be the orthogonal projection from T M to F (resp. F ⊥ ). Set
∇F = π F ∇L π F ;
⊥ ⊥ ⊥
∇F = π F ∇L π F , (2.6)
⊥
then ∇F (resp. ∇F ) is a Euclidean connection on F (resp. F ⊥ ), let S be the tensor defined by
⊥
∇L = ∇F + ∇F + S. (2.7)

Let e1 , · · ·, en be the orthonormal basis of T M and f1 , · · ·, fk be the orthonormal basis of F ⊥ . The

sub-signature operator DA acting on ∧∗ T ∗ M ⊗ C is defined by
n X
X k
DA = d + δ + c(ei )b
c(S(ei )fα )b
c(fα )
i=1 α=1
n
X   X n X k
1X
= c(ei ) ei + ωs,t (ei )[b
c(es )b
c(et ) − c(es )c(et )] + c(ei )b
c(S(ei )fα )b
c(fα ), (2.8)
i=1
4 s,t i=1 α=1

where c(ei ) denotes the Clifford action.

The following lemma of Dabrowski etc.’s Einstein functional plays a key role in our proof of the Einstein
functional. Let V , W be a pair of vector fields on a compact Riemannian manifold M of dimension n = 2m.
Using the Laplace operator ∆−1 −2 m
T = DT = ∆ + E acting on sections of a vector bundle E0 of rank 2 , the
spectral functional over vector fields defined by
Lemma 2.1. [3] The Einstein functional equals to
Z Z Z

υn−1 m υn−1 1
e e
W res ∇V ∇W ∆T−m
= 2 G(V, W )volg + F (V, W )volg + (trE)g(V, W )volg , (2.9)
6 M 2 M 2 M
where G(V, W ) denotes the Einstein tensor evaluated on the two vector fields, F (V, W ) = tr(V a Wb Fab ) and
2π m
Fab is the curvature tensor of the connection T , trE denotes the trace of E and υn−1 = Γ(m) .
The aim of this section is to prove the following.
2
Theorem 2.2. For the Laplace (type) operator ∆A = DA , the Einstein functional equals to
Z
A −m

2m+1 π m 1

W res ∇A X ∇Y ∆A = Ric(V, W ) − sg(V, W ) volg
6Γ(m) M 2
Z
− 22m−3 sg(V, W )volg , (2.10)
M
Pn Pk ∗ ∗
where s is the scalar curvature, A = i=1 α=1 c(ei )b c(fα ) and ∇A
c(S(ei )fα )b ∧ T
X = ∇X
M
− 21 [c(X)A +
Ac(X)].
Proof. By (2.12) in [18], for any ψ ∈ Γ(M, ∧∗ T ∗ M ⊗ C), we let
∧ ∗ 1
T ∗M
∇A
X ψ =∇X ψ − [c(X)A + Ac(X)]ψ
2
1
=Xψ + σ(X)ψ + a(X)ψ − [c(X)A + Ac(X)]ψ
2
=Xψ + B(X)ψ, (2.11)

3
where
1X 1X
σ(X) = − ωs,t (X)c(es )c(et ), a(X) = ωs,t (X)b
c(es )b
c(et ). (2.12)
4 s,t 4 s,t
Pn Pn
Let V = a=1 V a ea , W = b=1 W b eb , in view of that
n
X ∗
T ∗M
F (V, W ) = tr(Va Wb Fab ) = V a W b tr∧ (Fea ,eb ), (2.13)
a,b=1

we obtain
Fea ,eb =(ea + B(ea ))(eb + B(eb )) − (eb + B(eb ))(ea + B(ea )) − ([ea , eb ] + B([ea , eb ]))
=ea ◦ B(eb ) + B(ea ) ◦ eb + B(ea )B(eb ) − eb ◦ B(ea ) − B(eb ) ◦ ea
− B(eb )B(ea ) − B([ea , eb ])
=B(eb ) ◦ ea + ea (B(eb )) + B(ea ) ◦ eb + B(ea )B(eb ) − B(ea ) ◦ eb − eb (B(ea ))
− B(eb ) ◦ ea − B(eb )B(ea ) − B([ea , eb ])
=ea (B(eb )) − eb (B(ea )) + B(ea )B(eb ) − B(eb )B(ea ) − B([ea , eb ]). (2.14)
Also, straightforward computations yield
∗ ∗
∗ ∗
h  1X 1X
tr∧ T M ea (B(eb )) =tr∧ T M ea − ωs,t (eb )c(es )c(et ) + ωs,t (eb )b
c(es )b
c(et )
4 s,t 4 s,t
1 i
− [c(eb )A + Ac(eb )]
2
h 1X 1X i
∧∗ T ∗ M
=tr − ea (ωs,t (eb ))c(es )c(et ) + ea (ωs,t (eb ))b
c(es )bc(et )
4 s,t 4 s,t
=0, (2.15)
where
ωs,t (eb ) = 0 (s = t); tr[c(es )c(et )] = tr[b
c(es )b
c(et )] = 0 (s 6= t); tr[c(eb )A] = 0, (2.16)
where we take the normal coordinate about x0 , it follows that
∗ ∗

tr∧ T M B(ea )B(eb ) − B(eb )B(ea ) (x0 )
∗ ∗
h 1 X 1X 1 
=tr∧ T M − ωs,t (ea )c(es )c(et ) + ωs,t (ea )b c(et ) − [c(ea )A + Ac(ea )]
c(es )b
4 s,t 4 s,t 2
 1X 1 X 1 i
× − ωs,t (eb )c(es )c(et ) + ωs,t (eb )b c(et ) − [c(eb )A + Ac(eb )] (x0 )
c(es )b
4 s,t 4 s,t 2
∗ ∗
h 1 X 1 X 1 
− tr∧ T M − ωs,t (eb )c(es )c(et ) + ωs,t (eb )b c(et ) − [c(eb )A + Ac(eb )]
c(es )b
4 s,t 4 s,t 2
 1X 1X 1 i
× − ωs,t (ea )c(es )c(et ) + ωs,t (ea )b c(et ) − [c(ea )A + Ac(ea )] (x0 )
c(es )b
4 s,t 4 s,t 2
=0, (2.17)
and
∗
T ∗M


tr∧ B([ea , eb ]) (x0 )
∗
 1 
T ∗M
=tr∧ σ([ea , eb ]) + a([ea , eb ]) − [c([ea , eb ])A + Ac([ea , eb ]) (x0 )
2
=0. (2.18)

4
Let ∆A = ∆ + E. By (2.17) in [20], we have
1X 1 1X
E= Rijkl b c(ej )c(ek )c(el ) − s − A2 −
c(ei )b [c(ej )A + Ac(ej )]2
8 4 4 j
ijkl
1 ∗ ∗ ∗ ∗
+ [(∇e∧j T M A)c(ej ) − c(ej )(∇e∧j T M A)], (2.19)
2
and
∗
T ∗M 1
tr∧ (E) = − str[id] = −22m−2 s. (2.20)
4
Summing up (2.15)-(2.20) leads to the desired equality (2.10), and the proof of the Theorem is complete.

3. The noncommutative residue for manifolds with boundary

In this section, to define the noncommutative residue for the sub-signature operator, some basic facts
and formulae about Boutet de Monvel’s calculus can be found in Sec.2 in [13]. Let M be an n-dimensional
compact oriented manifold with boundary ∂M . Some basic facts and formulae about Boutet de Monvel’s
calculus are recalled as follows.
Let Z
F : L2 (Rt ) → L2 (Rv ); F (u)(v) = e−ivt u(t)dt

denote the Fourier transformation and ϕ(R+ ) = r+ ϕ(R) (similarly define ϕ(R− )), where ϕ(R) denotes the
Schwartz space and
r+ : C ∞ (R) → C ∞ (R+ ); f → f |R+ ; R+ = {x ≥ 0; x ∈ R}. (3.1)
We define H + = F (ϕ(R+ )); H0− = F (ϕ(R− )) which are orthogonal to each other. We have the following
property: h ∈ H + (H0− ) iff h ∈ C ∞ (R) which has an analytic extension to the lower (upper) complex half-
plane {Imξ < 0} ({Imξ > 0}) such that for all nonnegative integer l,
∞
X d l ck
dl h
l
(ξ) ∼ ( ) (3.2)
dξ dξ l ξ k
k=1

as |ξ| → +∞, Imξ ≤ 0 (Imξ ≥ 0). L L

Let H ′ be the space of all polynomials and H − = H0− H ′ ; H = H + H − . Denote by π + (π − )
respectively the projection on H + (H − ). For calculations, we take H = He = {rational functions having no
e is a dense set in the topology of H). Then on H,
poles on the real axis} (H e
Z
1 h(ξ)
π + h(ξ0 ) = lim dξ, (3.3)
2πi u→0− Γ+ ξ0 + iu − ξ
where Γ+ is a Jordan close curve included Imξ > 0 surrounding all the singularities of h in the upper
′
half-plane and ξ0 ∈ R. Similarly, define π on H, e
Z
1
π′ h = h(ξ)dξ. (3.4)
2π Γ+
T 1
R T 1
So, π ′ (H − ) = 0. For h ∈ H L1 (R), π ′ h = 2π R h(v)dv and for h ∈ H
+
L (R), π ′ h = 0. Denote by
B Boutet de Monvel’s algebra (for details, see Section 2 of [13]).
An operator of order m ∈ Z and type d is a matrix
  C ∞ (X, C ∞ (X,
π+ P + G K L E1 ) L E2 )
e=
A : −→ .
T S
C ∞ (∂X, F1 ) C ∞ (∂X, F2 )
5
where X is a manifold with boundary ∂X and E1 , E2 (F1 , F2 ) are vector bundles over X (∂X). Here, P :
C0∞ (Ω, E1 ) → C ∞ (Ω, E2 ) is a classical pseudo-differential operator of order m on Ω, where Ω is an open
neighborhood of X and Ei |X = Ei (i = 1, 2). P has an extension: E ′ (Ω, E1 ) → D′ (Ω, E2 ), where
E ′ (Ω, E1 ) (D′ (Ω, E2 )) is the dual space of C ∞ (Ω, E1 ) (C0∞ (Ω, E2 )). Let e+ : C ∞ (X, E1 ) → E ′ (Ω, E1 ) denote
extension by zero from X to Ω and r+ : D′ (Ω, E2 ) → D′ (Ω, E2 ) denote the restriction from Ω to X, then
define
π + P = r+ P e+ : C ∞ (X, E1 ) → D′ (Ω, E2 ).
In addition, P is supposed to have the transmission property; this means that, for all j, k, α, the homo-
geneous component pj of order j in the asymptotic expansion of the symbol p of P in local coordinates near
the boundary satisfies:
∂xkn ∂ξα′ pj (x′ , 0, 0, +1) = (−1)j−|α| ∂xkn ∂ξα′ pj (x′ , 0, 0, −1),
then π + P : C ∞ (X, E1 ) → C ∞ (X, E2 ) by Section 2.1 of [13].
Let M be a compact manifold with boundary ∂M . We assume that the metric g M on M has the following
form near the boundary
1
gM = g ∂M + dx2n , (3.5)
h(xn )
where g ∂M is the metric on ∂M . Let U ⊂ M be a collar neighborhood of ∂M which is diffeomorphic
∂M × [0, 1). By the definition of h(xn ) ∈ C ∞ ([0, 1)) and h(xn ) > 0, there exists e
h ∈ C ∞ ((−ε, 1)) such
e e c
that S small ε > 0. Then there exists a metric gb on M =
S h|[0,1) = h and h > 0 for some sufficiently
M ∂M ∂M × (−ε, 0] which has the form on U ∂M ∂M × (−ε, 0]
1
gb = g ∂M + dx2n , (3.6)
e
h(xn )

such that gb|M = g. We fix a metric b c such that gb|M = g. Now we recall the main theorem in [12].
g on the M
Theorem
 + 3.1. (Fedosov-Golse-Leichtnam-Schrohe)
 Let X and ∂X be connected, dimX = n ≥ 3,
e= π P + G K
A ∈ B , and denote by p, b and s the local symbols of P, G and S respectively. Define:
T S
Z Z
] e
Wres(A) = trE [p−n (x, ξ)] σ(ξ)dx
X S
Z Z
+ 2π {trE [(trb−n )(x′ , ξ ′ )] + trF [s1−n (x′ , ξ ′ )]} σ(ξ ′ )dx′ . (3.7)
∂X S′

Then ] A,
a) Wres([ e B]) = 0, for any A,
e B ∈ B; b) It is a unique continuous trace on B/B −∞.
e the l-order symbol of an operator
Let p1 , p2 be nonnegative integers and p1 + p2 ≤ n, denote by σl (A)
e an application of (3.5) and (3.6) in [13] shows that
A,
Definition 3.2. The spectral Einstein functional of compact manifolds with boundary is defined by

Ein{p
n
1 ,p2 } ^
M := W res[π + (∇A A 2 −p1
X ∇Y (DA )
−2 p2
) ◦ π + (DA ) ], (3.8)
−2 p2
where π + (∇A A 2 −p1
X ∇Y (DA ) ), π + (DA ) are elements in Boutet de Monvel’s algebra[15].
For the sub-signature operator ∇A A −2 −2 e
X ∇Y DA and DA , denote by σl (A) the l-order symbol of an operator
e
A. An application of (2.1.4) in [13] shows that

^
W res[π + (∇A A −2 p1 + 2 −p2
X ∇Y (DA ) ) ◦ π (DA ) ]
Z Z Z
A A 2 −p1 2 −p2
= tr∧ T M⊗C [σ−n (∇X ∇Y (DA )
∗ ∗ ◦ (DA ) ]σ(ξ)dx + Φ, (3.9)
M |ξ|=1 ∂M

6
where
Z Z +∞ ∞ X
X (−i)|α|+j+k+1
Φ= tr∧∗ T ∗ M⊗C [∂xj n ∂ξα′ ∂ξkn σr+ (∇A A 2 −p1
X ∇Y (DA ) )(x′ , 0, ξ ′ , ξn )
|ξ ′ |=1 −∞ j,k=0 α!(j + k + 1)!

× ∂xαn ∂ξj+1
n
∂xkn σl ((DA ) )(x′ , 0, ξ ′ , ξn )]dξn σ(ξ ′ )dx′ ,
2 −p2
(3.10)

and the sum is taken over r − k − |α| + ℓ − j − 1 = −n, r ≤ −p1 , ℓ ≤ −p2 .

A −1 −3
For the sub-signature operator ∇A
X ∇Y DA and DA , similarly we have

^
W res[π + (∇A A
X ∇Y (DA )
−1 p1 3 −p2
◦ π + (DA ) ]
Z Z Z
= tr∧∗ T ∗ M⊗C [σ−n (∇A ∇
X Y
A
(D A )−p1
◦ (D 3 −p2
A ) )]σ(ξ)dx + e
Φ, (3.11)
M |ξ|=1 ∂M

where
Z Z +∞ ∞ X
X
e= (−i)|α|+j+k+1
Φ tr∧∗ T ∗ M⊗C [∂xj n ∂ξα′ ∂ξkn σr+ (∇A A
X ∇Y (DA )
−p1
)(x′ , 0, ξ ′ , ξn )
|ξ ′ |=1 −∞ j,k=0 α!(j + k + 1)!

× ∂xαn ∂ξj+1
n
∂xkn σl ((DA ) )(x′ , 0, ξ ′ , ξn )]dξn σ(ξ ′ )dx′ ,
3 −p2
(3.12)

and the sum is taken over r − k − |α| + ℓ − j − 1 = −n, r ≤ −p1 , ℓ ≤ −p2 .

4. The residue for the sub-signature operator ∇A A

X ∇Y DA and DA
−2 −2

In this section, we compute the spectral Einstein functional for 4-dimension compact manifolds with
2
boundary and get a Dabrowski-Sitarz-Zalecki type theorem in this case. We will consider DA . Since
A A −2 −2 A A −2 −2
[σ−4 (∇X ∇Y DA ◦ DA ]|M has the same expression as [σ−4 (∇X ∇Y DA ◦ DA ]|M in the case of manifolds
without boundary, so locally we can use Theorem 2.2 to compute the first term.
Theorem 4.1. Let M be a 4-dimensional compact manifold without boundary and ∇A be an orthogonal
A −2 −2
connection. Then we get the spectral Einstein functional associated to ∇A
X ∇Y DA and DA on compact
manifolds without boundary
A −2 −2
W res[σ−4 (∇A
X ∇Y DA ◦ DA )]
Z   Z
4π 2 1
= Ric(X, Y ) − sg(X, Y ) volg − 2 sg(X, Y )volg , (4.1)
3 M 2 M

where s is the scalar curvature.

R
So we only need to compute ∂M
Φ. By (2.8), we have
√
σ1 (DA ) = −1c(ξ); (4.2)
1X 1X
σ0 (DA ) = ωs,t (ei )c(ei )b
c(es )b
c(et ) − ωs,t (ei )c(ei )c(es )c(et ) + A, (4.3)
4 s,t,i 4 s,t,i

Pn Pk
where A = i=1 α=1 c(ei )b
c(S(ei )fα )b
c(fα ).
∗ ∗
1
P 1
P
We define ∇∧ X
T M
:= X + 4
L
ij h∇X ei , ej ic(ei )c(ej ) − 4
L
ij h∇X ei , ej ib
c(ei )b
c(ej ), which is a connection
on ∧∗ T ∗ M . Set
1X L 1X L
B(X) = h∇X ei , ej ic(ei )c(ej ) − h∇X ei , ej ib
c(ei )b
c(ej ). (4.4)
4 ij 4 ij

7
 1 1
Let ∇A A
X = X + B(X) − 2 [c(X)A + Ac(X)] and ∇Y = Y + B(Y ) − 2 [c(Y )A + Ac(Y )], we obtain
  
A A 1 1
∇X ∇Y = X + B(X) − [c(X)A + Ac(X)] Y + B(Y ) − [c(Y )A + Ac(Y )]
2 2
1 1
= XY + X[B(Y )] + B(Y )X − X[c(Y )A + Ac(Y )] − [c(Y )A + Ac(Y )]X
2 2
1 1
+ B(X)Y + B(X)B(Y ) − B(X)[c(Y )A + Ac(Y )] − [c(X)A + Ac(X)]Y
2 2
1 1
− [c(X)A + Ac(X)]B(Y ) + [c(X)A + Ac(X)][c(Y )A + Ac(Y )], (4.5)
2 4
Pn Pn
where X = j=1 Xj ∂xj , Y = l=1 Yl ∂xl .
P P
Let g ij = g(dxi , dxj ), ξ = j ξj dxj and ∇L
∂i ∂j = k Γkij ∂k , we denote that

1X 1X
σi = − ωs,t (ei )c(es )c(et ); ai = ωs,t (ei )b
c(es )b
c(et );
4 s,t 4 s,t
ξ j = g ij ξi ; Γk = g ij Γkij ; σ j = g ij σi ; aj = g ij ai .
(4.6)

Then we have the following lemmas.

Lemma 4.2. The following identities hold:
1 1
σ0 (∇A A
X ∇Y ) =X[B(Y )] + B(X)B(Y ) − [c(X)A + Ac(X)]B(Y ) − X[c(Y )A + Ac(Y )]
2 2
1 1
− B(X)[c(Y )A + Ac(Y )] + [c(X)A + Ac(X)][c(Y )A + Ac(Y )]; (4.7)
2 4
Xn X X
∂Y √ √ √
σ1 (∇A A
X ∇Y ) = Xj l −1ξl + B(Y )Xj −1ξj + B(X)Yl −1ξl
∂Xj j
j,l=1 l
1X √ 1X √
− [c(Y )A + Ac(Y )]Xj −1ξj − [c(X)A + Ac(X)]Yl −1ξl ; (4.8)
2 j 2
l
n
X
σ2 (∇A A
X ∇Y ) = − Xj Yl ξj ξl . (4.9)
j,l=1

Hence by Lemma 2.1 in [15], we have

Lemma 4.3. The symbol of the sub-signature operator
√
−1 −1c(ξ)
σ−1 (DA )= ; (4.10)
|ξ|2
c(ξ)σ0 (DA )c(ξ) c(ξ) X h i
−1 2 2
σ−2 (DA )= + c(dxj ) ∂xj (c(ξ))|ξ| − c(ξ)∂ xj (|ξ| ) . (4.11)
|ξ|4 |ξ|6 j

By (8a) and (11a) in [10], we get

Lemma 4.4. The following identities hold:
−2
σ−2 (DA ) =|ξ|−2 ; (4.12)
√  
−2
σ−3 (DA ) = − −1|ξ|−4 ξk Γk − 2σ k − 2ak − c(∂i )A − Ac(∂i )
√
− −1|ξ|−6 2ξ j ξα ξβ ∂j g αβ . (4.13)
8
 P 1 α α
By Lemma 4.2, Lemma 4.4 and σ(p1 ◦ p2 ) = α α! ∂ξ [σ(p1 )]Dx [σ(p2 )], we obtain
Lemma 4.5. The following identities hold:
n
X
A −2
σ0 (∇A
X ∇Y DA ) = − Xj Yl ξj ξl |ξ|−2 ; (4.14)
j,l=1
A −2 −2 −2
σ−1 (∇A A A A A
X ∇Y DA ) =σ2 (∇X ∇Y )σ−3 (DA ) + σ1 (∇X ∇Y )σ−2 (DA )
n
X    −2

+ ∂ξj σ2 (∇A A
X ∇Y ) Dxj σ−2 (DA ) . (4.15)
j=1

Since Φ is a global form on ∂M , so for any fixed point x0 ∈ ∂M , we choose the normal coordinates U of x0
in ∂M (not in M ) and compute Φ(x0 ) in the coordinates U e = U ×[0, 1) ⊂ M and the metric 1 g ∂M +dx2n .
h(xn )
The dual metric of g M on U e is h(xn )g ∂M + dx2n . Write g M = g M ( ∂ , ∂ ); g ij = g M (dxi , dxj ), then
ij ∂xi ∂xj M
 1 ∂M
  ij

M
[gij ] = h(xn ) [gij ] 0
; ij
[gM ] =
h(xn )[g∂M ] 0
, (4.16)
0 1 0 1
and
∂M M
∂xs gij (x0 ) = 0, 1 ≤ i, j ≤ n − 1; gij (x0 ) = δij . (4.17)
∂M
Let {ee1 , · · ·, ef
n } be an
porthonormal frame field
p in U about g ∂ which is parallel along geodesics and
∂
eei = ∂xi (x0 ), then {e1 = h(xn )ee1 , · · ·, en−1 = h(xn )] e
en−1 , en = ∂xn } is the orthonormal frame field in U
about g M . Locally ∧∗ T ∗ M |U e∼ =U e × ∧∗ (n). Let {f1 , · · ·, fn } be the orthonormal basis of ∧∗ (n). Take a spin
C C
frame field σ : U e → Spin(M ) such that πσ = {e1 , · · ·, en } where π : Spin(M ) → O(M ) is a double covering,
then {[σ, fi ], 1 ≤ i ≤ 4} is an orthonormal frame of ∧∗ T ∗ M |Ue . In the following, since the global form Φ is
independent of the choice of the local frame, so we can compute tr∧∗ T ∗ M in the frame {[σ, fi ], 1 ≤ i ≤ 4}.
en } be the canonical basis of Rn and c(ebi ) ∈ Hom(∧∗C (n), ∧∗C (n) be the Clifford action. By [21],
Let {eb1 , · · ·, c
then
∂ ∂
c(ei ) = [σ, c(ebi )]; c(ei )[σ, fi ] = [σ, c(ebi )fi ]; = [σ, ], (4.18)
∂xi ∂xi
∂
then we have ∂xi c(ei ) = 0 in the above frame.

Lemma 4.6. [15] With the metric g T M on M near the boundary

(
2
0, if j < n,
∂xj (|ξ|gM )(x0 ) = (4.19)
h′ (0)|ξ ′ |2g∂M , if j = n,


0, if j < n,
∂xj [c(ξ)](x0 ) = (4.20)
∂xn (c(ξ ′ ))(x0 ), if j = n,

where ξ = ξ ′ + ξn dxn .
R
Now we need to compute ∂M Φ. When n = 4, then tr∧∗ T ∗ M⊗C [id] = 2n = 16, the sum is taken over
r + l − k − j − |α| = −3, r ≤ 0, l ≤ −2, then we have the following five cases:

Case (a) (I) r = 0, l = −2, k = j = 0, |α| = 1.

9
By (3.10), we get
Z Z +∞ X
Φ1 = − tr[∂ξα′ πξ+n σ0 (∇A A −2 α −2 ′ ′
X ∇Y DA ) × ∂x′ ∂ξn σ−2 (DA )](x0 )dξn σ(ξ )dx . (4.21)
|ξ ′ |=1 −∞ |α|=1

By Lemma 2.2 in [15], for i < n,

−2 ∂xi (|ξ|2 )(x0 )

∂xi σ−2 (DA )(x0 ) = ∂xi (|ξ|−2 )(x0 ) = − = 0, (4.22)
|ξ|4

so Φ1 = 0.

Case (a) (II) r = 0, l = −2, k = |α| = 0, j = 1.

By (3.10), we get
Z Z +∞
1
Φ2 = − tr[∂xn πξ+n σ0 (∇A A −2 2 −2 ′ ′
X ∇Y DA ) × ∂ξn σ−2 (DA )](x0 )dξn σ(ξ )dx . (4.23)
2 |ξ ′ |=1 −∞

By Lemma 4.4, we have

−2 6ξn2 − 2
∂ξ2n σ−2 (DA )(x0 )||ξ′ |=1 = ∂ξ2n (|ξ|−2 )(x0 )||ξ′ |=1 = . (4.24)
(1 + ξn2 )3

It follows that
n
X
A −2
∂xn σ0 (∇A
X ∇Y DA )(x0 )||ξ ′ |=1 =∂xn (− Xj Yl ξj ξl |ξ|−2 )(x0 )||ξ′ |=1
j,l=1
n
X
1
= Xj Yl ξj ξl h′ (0). (4.25)
(1 + ξn2 )2
j,l=1

By integrating formula, we obtain

πξ+n ∂xn σ0 (∇A A −2 + A A −2

X ∇Y DA )(x0 )||ξ ′ |=1 =∂xn πξn σ0 (∇X ∇Y DA )(x0 )||ξ ′ |=1
n−1
X
iξn 2 − iξn
=− 2
Xj Yl ξj ξl h′ (0) + Xn Yn h′ (0)
4(ξn − i) 4(ξn − i)2
j,l=1
n−1
X n−1
X
i i
− Xj Yn ξj − Xn Yl ξl . (4.26)
4(ξn − i)2 j=1
4(ξn − i)2
l=1

From (4.24) and (4.26), we obtain

tr[∂xn πξ+n σ0 (∇A A −2 2 −2

X ∇Y DA ) × ∂ξn σ−2 (DA )](x0 )||ξ ′ |=1
n−1
X
−3iξn3 + iξn −3iξn3 + 6ξn2 + iξn − 2
=8 5 3
Xj Yl ξj ξl h′ (0) + 8 Xn Yn h′ (0)
(ξn − i) (ξn + i) (ξn − i)5 (ξn + i)3
j,l=1
n−1
X n−1
X
(1 − 3ξn2 )i (1 − 3ξn2 )i
+8 5 3
X Y ξ
j n j + 8 5 3
Xn Yl ξl . (4.27)
(ξn − i) (ξn + i) j=1 (ξn − i) (ξn + i)
l=1

10
 R
We note that i < n, |ξ′ |=1 {ξi1 ξi2 · · · ξi2q+1 }σ(ξ ′ ) = 0, so we omit some items that have no contribution for
computing Φ2 . Therefore, we get
Z Z +∞  X n−1
1 −3iξn3 + iξn
Φ2 = − 8 Xj Yl ξj ξl h′ (0)
2 |ξ ′ |=1 −∞ (ξn − i)5 (ξn + i)3
j,l=1

−3iξn3 + 6ξn2 + iξn − 2
+8 Xn Yn h′ (0) dξn σ(ξ ′ )dx′
(ξn − i)5 (ξn + i)3
n−1
X Z Z
−3iξn3 + iξn
=−4 Xj Yl h′ (0) 5
ξ ξ dξ σ(ξ ′ )dx′
3 j l n
|ξ ′ |=1 Γ+ (ξn − i) (ξn + i)
j,l=1
Z
−3iξn3 + 6ξn2 + iξn − 2
− 4Xn Yn h′ (0)Ω3 5 (ξ + i)3
dξn dx′
Γ+ (ξn − i) n
n−1
X  (4)
4π 2πi −3iξn3 + iξn
=−4 Xj Yl h′ (0) dx′
3 4! (ξn + i)3 ξn =i
j,l=1
 (4)
2πi −3iξn3 + 6ξn2 + iξn − 2
− 4Xn Yn h′ (0)Ω3 dx′
4! (ξn + i)3 ξn =i
2 n−1
X
π 15
=− Xj Yj h′ (0)dx′ − Xn Yn h′ (0)πΩ3 dx′ , (4.28)
2 j=1
8

where Ω3 is the canonical volume of S 2 .

Case (a) (III) r = 0, l = −2, j = |α| = 0, k = 1.

By (3.10), we get
Z Z +∞
1
Φ3 = − tr[∂ξn πξ+n σ0 (∇A A −2 −2 ′
X ∇Y DA ) × ∂ξn ∂xn σ−2 (DA )](x0 )dξn σ(ξ )dx
′
2 |ξ′ |=1 −∞
Z Z +∞
1
= tr[∂ξ2n πξ+n σ0 (∇A A −2 −2 ′ ′
X ∇Y DA ) × ∂xn σ−2 (DA )](x0 )dξn σ(ξ )dx . (4.29)
2 |ξ′ |=1 −∞

By Lemma 4.4, we have

−2 h′ (0)
∂xn σ−2 (DA )(x0 )||ξ′ |=1 = − . (4.30)
(1 + ξn2 )2

An easy calculation gives

n−1
X
i 1
πξ+n σ0 (∇A A −2
X ∇Y DA )(x0 )||ξ ′ |=1 = Xj Yl ξj ξl − Xn Yn
2(ξn − i) 2(ξn − i)
j,l=1
n−1
X n−1
X
1 1
− Xj Yn ξj − Xn Yl ξl . (4.31)
2(ξn − i) j=1 2(ξn − i)
l=1

Also, straightforward computations yield

n−1
X
i 1
∂ξ2n πξ+n σ0 (∇A A −2
X ∇Y DA )(x0 )||ξ ′ |=1 = Xj Yl ξj ξl − Xn Yn . (4.32)
(ξn − i)3 (ξn − i)3
j,l=1

11
From (4.30) and (4.32), we obtain

tr[∂ξ2n πξ+n σ0 (∇A A −2 −2

X ∇Y DA ) × ∂xn σ−2 (DA )](x0 )||ξ ′ |=1
n−1
X
−16h′ (0)i 16h′ (0)
= X j Yl ξj ξl + Xn Yn . (4.33)
(ξn − i)5 (ξn + i)2 (ξn − i)5 (ξn + i)2
j,l=1

Therefore, we get
Z Z +∞  n−1
X 
1 −16h′ (0)i 16h′ (0)
Φ3 = Xj Yl ξj ξl + Xn Yn dξn σ(ξ ′ )dx′
2 |ξ ′ |=1 −∞ (ξn − i)5 (ξn + i)2 (ξn − i)5 (ξn + i)2
j,l=1
n−1
X Z Z
i
=−8 Xj Yl h′ (0) ξj ξl dξn σ(ξ ′ )dx′
|ξ ′ |=1 Γ+ (ξn − i)5 (ξn + i)2
j,l=1
Z
′ 1
+ 8Xn Yn h (0)Ω3 dξn dx′
(ξn − i)5 (ξn + i)2
Γ+
n−1
X  (4)  (4)
4π 2πi i 2πi 1
=−8 Xj Yl h′ (0) dx′
+ 8X Y
n n h ′
(0)Ω 3 dx′
3 4! (ξn + i)2 ξn =i 4! (ξn + i)2 ξn =i
j,l=1
2 n−1
X
5π 5i
=− Xj Yj h′ (0)dx′ − Xn Yn h′ (0)πΩ3 dx′ . (4.34)
3 j=1
4

Case (b) r = 0, l = −3, k = j = |α| = 0.

By (3.10), we get
Z Z +∞
Φ4 = −i tr[πξ+n σ0 (∇A A −2 −2 ′
X ∇Y DA ) × ∂ξn σ−3 (DA )](x0 )dξn σ(ξ )dx
′
|ξ ′ |=1 −∞
Z Z +∞
=i tr[∂ξn πξ+n σ0 (∇A A −2 −2 ′ ′
X ∇Y DA ) × σ−3 (DA )](x0 )dξn σ(ξ )dx . (4.35)
|ξ ′ |=1 −∞

By Lemma 4.4, we have

 X
−2 i 1 ′
σ−3 (DA )(x0 )||ξ′ |=1 = − h (0) ξk [c(ek )c(en ) − b
c(ek )b
c(en )]
(1 + ξn2 )22
k<n
X 
3 2ih′ (0)ξn
− ξk [c(∂k )A + Ac(∂k )] + h′ (0)ξn − . (4.36)
2 (1 + ξn2 )3
k<n

An easy calculation gives

n−1
X
i 1
∂ξn πξ+n σ0 (∇A A −2
X ∇Y DA )(x0 )||ξ ′ |=1 = − 2
Xj Yl ξj ξl − Xn Yn
2(ξn − i) 2(ξn − i)2
j,l=1
n−1
X n−1
X
1 1
+ 2
X j Yn ξj + 2
Xn Yl ξl . (4.37)
2(ξn − i) j=1 2(ξn − i)
l=1
R
We note that i < n, |ξ ′ |=1
{ξi1 ξi2 · · · ξi2q+1 }σ(ξ ′ ) = 0, so we omit some items that have no contribution for

12
computing Φ4 . Then, we have

tr[∂ξn πξ+n σ0 (∇A A −2 −2

X ∇Y DA ) × σ−3 (DA )](x0 )||ξ ′ |=1
  n−1
X n−1
X 
−ih′ (0)
=tr Xj Yn ξk ξj [c(ek )c(en ) − b
c(ek )b
c(en )] + Xn Yl ξk ξl [c(ek )c(en ) − b
c(ek )b
c(en )]
4(ξn − i)4 (ξn + i)2
k,j=1 k,l=1
 n−1
X n−1
X 
i
+ Xj Yn ξk ξj [c(∂k )A + Ac(∂k )] + Xn Yl ξk ξl [c(∂k )A + Ac(∂k )]
2(ξn − i)4 (ξn + i)2
k,j=1 k,l=1
n−1 
h′ (0)(3ξn3 + 7ξn ) X ih′ (0)(3ξn3 + 7ξn )
− Xj Yl ξj ξl + Xn Yn . (4.38)
4(ξn − i)5 (ξn + i)3 4(ξn − i)5 (ξn + i)3
j,l=1

By the relation of the Clifford action and tr(AB) = tr(BA), then we have the equalities:

tr[c(ek )c(en )] = 0(k < n); tr[b

c(ek )b
c(en )] = 0(k < n);
tr[c(∂k )A] = 0; tr[Ac(∂k )] = 0. (4.39)

Therefore, we get
Z Z  n−1 
+∞
4h′ (0)(3ξn3 + 7ξn ) X 4ih′ (0)(3ξn3 + 7ξn )
Φ4 =i − Xj Yl ξj ξl + Xn Yn dξn σ(ξ ′ )dx′
|ξ ′ |=1 −∞ (ξn − i)5 (ξn + i)3 (ξn − i)5 (ξn + i)3
j,l=1
n−1
X Z Z
3ξn3 + 7ξn
= − 4i Xj Yl h′ (0) ξj ξl dξn σ(ξ ′ )dx′
|ξ ′ |=1 Γ+ (ξn − i)5 (ξn + i)2
j,l=1
Z
3ξn3 + 7ξn
− 4Xn Yn h′ (0)Ω3 5 3
dξn dx′
Γ+ (ξn − i) (ξn + i)
n−1
X  (4)  (4)
′ 4π 2πi 3ξn3 + 7ξn ′ ′ 2πi 3ξn3 + 7ξn
= − 4i Xj Yl h (0) dx − 4Xn Yn h (0)Ω3 dx′
3 4! (ξn + i)3 ξn =i 4! (ξn + i)3 ξn =i
j,l=1
2 n−1
X
17π 51i
= Xj Yj h′ (0)dx′ − Xn Yn h′ (0)πΩ3 dx′ . (4.40)
4 j=1
16

Case (c) r = −1, ℓ = −2, k = j = |α| = 0.

By (3.10), we get
Z Z +∞
Φ5 = −i tr[πξ+n σ−1 (∇A A −2 −2 ′ ′
X ∇Y DA ) × ∂ξn σ−2 (DA )](x0 )dξn σ(ξ )dx . (4.41)
|ξ ′ |=1 −∞

By Lemma 4.4, we have

−2 2ξn
∂ξn σ−2 (DA )(x0 )||ξ′ |=1 = − . (4.42)
(1 + ξn2 )2

Since
A −2 −2 −2
σ−1 (∇A A A A A
X ∇Y DA )(x0 )||ξ ′ |=1 =σ2 (∇X ∇Y )σ−3 (DA ) + σ1 (∇X ∇Y )σ−2 (DA )
n
X    −2

+ ∂ξj σ2 (∇A A
X ∇Y ) Dxj σ−2 (DA ) . (4.43)
j=1

13
 (1) Explicit representation the first item of (4.43),
−2
σ2 (∇A A
X ∇Y )σ−3 (DA )(x0 )||ξ ′ |=1
Xn  √ √ 
=− Xj Yl ξj ξl × − −1|ξ|−4 ξk (Γk − 2σ k − 2ak − c(∂i )A − Ac(∂i )) − −1|ξ|−6 2ξ j ξα ξβ ∂j g αβ
j,l=1
Xn  h1 X X
i
=− Xj Yl ξj ξl × −
2 2
h′ (0) ξk [c(ek )c(en ) − b
c(ek )b
c(en )] − ξk [c(∂k )A + Ac(∂k )]
(1 + ξn ) 2
j,l=1 k<n k<n

3 i 2ih′ (0)ξ 
n
+ h′ (0)ξn − . (4.44)
2 (1 + ξn2 )3
(2) Explicit representation the second item of (4.43),
−2
σ1 (∇A A
X ∇Y )σ−2 (DA )(x0 )||ξ ′ |=1
X n
∂Y √ X √ X √
= Xj l −1ξl + B(Y )Xj −1ξj + B(X)Yl −1ξl
∂Xj j
j,l=1 l
1X √ 1X √ 
− [c(Y )A + Ac(Y )]Xj −1ξj − [c(X)A + Ac(X)]Yl −1ξl × |ξ|−2 . (4.45)
2 j 2
l

(3) Explicit representation the third item of (4.43),

Xn X
1 α  α −2

∂ξ σ2 (∇A A
X ∇Y ) Dx σ−2 (DA ) (x0 )||ξ ′ |=1
j=1 α
α!
n
X   √ −2

= ∂ξj σ2 (∇A A
X ∇Y ) (− −1)∂xj [σ−2 (DA )
j=1
n
X n
X
  √  
= ∂ξj − Xj Yl ξj ξl (− −1)∂xj |ξ|−2
j=1 j,l=1
n X
X n
√
= −1(Xj Yl + Xl Yj )ξl ∂xj (|ξ|−2 ). (4.46)
j=1 l=1
R
We note that i < n, |ξ′ |=1 {ξi1 ξi2 · · · ξi2q+1 }σ(ξ ′ ) = 0, so we omit some items that have no contribution for
computing Φ5 . An easy calculation gives
tr[πξ+n σ2 (∇A A −2 −2
X ∇Y )σ−3 (DA ) × ∂ξn σ−2 (DA )](x0 )||ξ ′ |=1
n−1
X n−1
12h′ (0)ξn 2ih′ (0)(2iξn2 + 6ξn ) X
=− X j Yl ξj ξl + Xj Yl ξj ξl
(ξn − i)4 (ξn + i)2 (ξn − i)5 (ξn + i)2
j,l=1 j,l=1
12h′ (0)ξn 16ih′ (0)ξn
− 4 2
Xn Yn − Xn Yn . (4.47)
(ξn − i) (ξn + i) (ξn − i)5 (ξn + i)2
By the relation of the Clifford action and tr(AB) = tr(BA), then we have the equalities:
tr[B(X)] = 0; tr[B(Y )] = 0; tr[c(X)A] = 0;
tr[Ac(X)] = 0; tr[c(Y )A] = 0; tr[Ac(Y )] = 0. (4.48)
Then
tr[πξ+n σ1 (∇A A −2 −2
X ∇Y )σ−2 (DA ) × ∂ξn σ−2 (DA )](x0 )||ξ ′ |=1
∂Yn −16iξn
=Xn . (4.49)
∂xn (ξn − i)3 (ξn + i)2

14
Also, straightforward computations yield
 X 
n X
1 α  α 
tr πξ+n ∂ξ σ2 (∇A ∇
X Y
A
) D x σ−2 (D −2
A ) × ∂ξn σ−2 (D −2
A ) (x0 )||ξ′ |=1
j=1 α
α!
−64ih′ (0)ξn2
= Xn Yn . (4.50)
(ξn − i)2 (ξn + i)2

Substituting (4.47)-(4.50) into (4.41) yields

n−1
11π 2 X
Φ5 = − Xj Yj h′ (0)dx′ − 32Xn Yn h′ (0)πΩ3 dx′
3 j=1
∂Yn π
− Xn Ω3 dx′ . (4.51)
∂xn 2
Pn−1
Let X = X T + Xn ∂n , Y = Y T + Yn ∂n , then we have j=1 Xj Yj (x0 ) = g(X T , Y T )(x0 ). Now Φ is the sum
of the Φ(1,2,···,5) . Therefore, we get

5
X 542 + 71i 19π 2 ∂Yn π
Φ= Φi = − Xn Yn h′ (0)πΩ3 dx′ − g(X T , Y T )h′ (0)dx′ − Xn Ω3 dx′ . (4.52)
i=1
16 12 ∂xn 2

Then, we obtain the following theorem.

Theorem 4.7. Let M be a 4-dimensional compact manifold with boundary and ∇A be an orthogonal connec-
A −2 −2
tion. Then we get the spectral Einstein functional associated to ∇A
X ∇Y DA and DA on compact manifolds
with boundary

] + (∇A
Wres[π A −2 +
X ∇Y DA ) ◦ π (DA )]
−2
Z   Z
4π 2 1
= Ric(V, W ) − sg(V, W ) volg − 2 sg(V, W )volg
3 M 2 M
Z  542 + 71i 19π 2 ∂Yn π 
+ − Xn Yn h′ (0)πΩ3 − g(X T , Y T )h′ (0) − Xn Ω3 vol∂M , (4.53)
∂M 16 12 ∂xn 2

where s is the scalar curvature.

5. The residue for the sub-signature operator ∇A A

X ∇Y DA and DA
−1 −3

In this section, we compute the 4-dimension spectral Einstein functional for the sub-signature operator
A −1 −3 A A −1 −3 A A −1
∇AX ∇Y DA and DA . Since [σ−4 (∇X ∇Y DA ◦ DA )]|M has the same expression as [σ−4 (∇X ∇Y DA ◦
−3
DA )]|M in the case of manifolds without boundary, so locally we can use Theorem 2.2 to compute the first
term.
Theorem 5.1. Let M be a 4-dimensional compact manifold without boundary and ∇A be an orthogonal
A −1 −3
connection. Then we get the spectral Einstein functional associated to ∇A
X ∇Y DA and DA on compact
manifolds without boundary
A −1 −3
W res[σ−4 (∇A
X ∇Y DA ◦ DA )]
Z   Z
4π 2 1
= Ric(X, Y ) − sg(X, Y ) volg − 2 sg(X, Y )volg , (5.1)
3 M 2 M

where s is the scalar curvature.

15
 From Lemma 4.2 and Lemma 4.3, we have
Lemma 5.2. The following identities hold:
A −1 −1 −1
σ0 (∇A A A A A
X ∇Y DA ) =σ2 (∇X ∇Y )σ−2 (DA ) + σ1 (∇X ∇Y )σ−1 (DA )
n
X    −1

+ ∂ξj σ2 (∇A A
X ∇Y ) Dxj σ−1 (DA ) ; (5.2)
j=1
n
X
A −1
√
σ1 (∇A
X ∇Y DA ) = − −1 Xj Yl ξj ξl |ξ|−2 . (5.3)
j,l=1

Write
−3 P∞
Dxα = (−i)|α| ∂xα ; σ(DA
3
) = p3 + p2 + p1 + p0 ; σ(DA )= j=3 q−j . (5.4)

By the composition formula of pseudo-differential operators, we have

X 1
3 −3 3 −3
1 = σ(DA ◦ DA )= ∂ α [σ(DA )]Dxα [σ(DA )]
α
α! ξ
= (p3 + p2 + p1 + p0 )(q−3 + q−4 + q−5 + · · · )
X
+ (∂ξj p3 + ∂ξj p2 + +∂ξj p1 + ∂ξj p0 )(Dxj q−3 + Dxj q−4 + Dxj q−5 + · · · )
j
X
= p3 q−3 + (p3 q−4 + p2 q−3 + ∂ξj p3 Dxj q−3 ) + · · · , (5.5)
j

so X
q−3 = p−1 −1 −1
3 ; q−4 = −p3 [p2 p3 + ∂ξj p3 Dxj (p− 3−1 )]. (5.6)
j

Then, it is easy to check that

Lemma 5.3. The following identities hold:
−2
σ−2 (DA ) = |ξ|−2 ; (5.7)
√  
−2
σ−3 (DA ) = − −1|ξ|−4 ξk Γk − 2σ k − 2ak − c(∂i )A − Ac(∂i )
√
− −1|ξ|−6 2ξ j ξα ξβ ∂j g αβ ; (5.8)
−3
√
σ−3 (DA ) = −1c(ξ)|ξ|−4 ; (5.9)
√
−3 c(ξ)σ2 (DA 3
)c(ξ) −1c(ξ)  4
σ−4 (DA ) = + |ξ| c(dxn )∂xn [c(ξ ′ )]
|ξ|8 |ξ|8

− 2h′ (0)c(dxn )c(ξ) + 2ξn c(ξ)∂xn [c(ξ ′ )] + 4ξn h′ (0) , (5.10)

where
X
3
σ2 (DA )= c(dxl )∂l (g i,j )ξi ξj + c(ξ)(4σ k + 4ak − 2Γk )ξk − 2[c(ξ)Ac(ξ) − |ξ|2 A]
i,j,l
1 X
+ |ξ|2 ωs,t (el )[c(el )b c(et ) − c(el )c(es )c(et )] + |ξ|2 A.
c(es )b (5.11)
4
s,t,l
R
Now we need to compute ∂M Φ. e When n = 4, then tr∧∗ T ∗ M⊗C [id] = 2n = 16, the sum is taken over
r + l − k − j − |α| = −3, r ≤ 0, l ≤ −2, then we have the following five cases:

16
Case (a) (I) r = 1, l = −2, k = j = 0, |α| = 1.

By (3.12), we get
Z Z +∞ X
e1 = −
Φ tr[∂ξα′ πξ+n σ1 (∇A A −1 α −3 ′ ′
X ∇Y DA ) × ∂x′ ∂ξn σ−3 (DA )](x0 )dξn σ(ξ )dx . (5.12)
|ξ ′ |=1 −∞ |α|=1

Similarly, for i < n,

−3
√
∂xi σ−3 (DA )(x0 ) =∂xi ( −1c(ξ)|ξ|−4 )(x0 )
√ ∂x c(ξ) √ c(ξ)∂xi (|ξ|4 )
= −1 i 4 (x0 ) + −1 (x0 )
|ξ| |ξ|8
=0, (5.13)

e 1 = 0.
so Φ

Case (a) (II) r = 1, l = −3, k = |α| = 0, j = 1.

By (3.12), we get
Z Z +∞
e2 = − 1
Φ tr[∂xn πξ+n σ1 (∇A A −1 2 −3 ′ ′
X ∇Y DA ) × ∂ξn σ−3 (DA )](x0 )dξn σ(ξ )dx . (5.14)
2 |ξ ′ |=1 −∞

By Lemma 5.3, we have

−3
∂ξ2n σ−3 (DA )(x0 )||ξ′ |=1 = ∂ξ2n (c(ξ)|ξ|−4 )(x0 )||ξ′ |=1
4i(5ξn2 − 1) ′ 12i(ξn3 − ξn )
= 2 4
c(ξ ) + c(dxn ). (5.15)
(1 + ξn ) (1 + ξn2 )4

By Lemma 5.2, we obtain

n
X
A −1
√
∂xn σ1 (∇A ∇
X Y D A )(x )|
0 |ξ |=1
′ = ∂ xn (− −1 Xj Yl ξj ξl c(ξ)|ξ|−2 )(x0 )||ξ′ |=1
j,l=1
n
X  
∂xn [c(ξ ′ )] h′ (0)c(ξ)
= Xj Yl ξj ξl + . (5.16)
1 + ξn2 (1 + ξn2 )2
j,l=1

17
Then, we have

πξ+n ∂xn σ1 (∇A A −1 + A A −1

X ∇Y DA )(x0 )||ξ ′ |=1 =∂xn πξn σ1 (∇X ∇Y DA )(x0 )||ξ ′ |=1
n−1
X  
′ ic(ξ ′ ) c(ξ ′ ) + ic(dxn )
=i Xj Yl ξj ξl h (0) +
4(ξn − i) 4(ξn − i)2
j,l=1
n−1
X ∂xn [c(ξ ′ )]
− Xj Yl ξj ξl
2(ξn − i)
j,l=1
 
∂xn [c(ξ ′ )] 2ic(ξ ′ ) − 3c(dxn )
− iXn Yn + h′ (0) −
2(ξn − i) 4(ξn − i)

[c(ξ ′ ) + ic(dxn )](iξn + 2)
+
4(ξn − i)2
n−1
X 
i∂xn [c(ξ ′ )] ih′ (0)[c(ξ ′ ) + 2ic(dxn )]
− Xj Yn ξj −
j=1
2(ξn − i) 4(ξn − i)
′

[ic(ξ ) − c(dxn )](iξn + 2)
−
(ξn − i)2
n−1
X 
i∂xn [c(ξ ′ )] ih′ (0)[c(ξ ′ ) + 2ic(dxn )]
− Xn Yl ξl −
2(ξn − i) 4(ξn − i)
l=1
′

[ic(ξ ) − c(dxn )](iξn + 2)
− . (5.17)
(ξn − i)2

Then, there is the following formula

tr[∂xn πξ+n σ1 (∇A A −1 2 −3

X ∇Y DA ) × ∂ξn σ−3 (DA )](x0 )||ξ ′ |=1
n−1
X  
′ 5ξn2 − 1 (5ξn2 − 1) + 3i(ξn3 − ξn )
=16 Xj Yl ξj ξl h (0) 2i +
(ξn − i)5 (ξn + i)4 (ξn − i)6 (ξn + i)4
j,l=1
 
′ (i − 1)(ξn2 − 1) + 12(ξn3 − ξn ) (5ξn2 − 1) + 3i(ξn3 − ξn )
+ 16Xn Yn h (0) −
(ξn − i)5 (ξn + i)4 (ξn − i)6 (ξn + i)4
n−1
X  
[6 − 3ih′ (0)](ξn3 − ξn ) − 2i(5ξn2 − 1) 2(5ξn2 − 1) + 6i(ξn3 − ξn )
+8 Xj Yn ξj +
j=1
(ξn − i)5 (ξn + i)4 (ξn − i)6 (ξn + i)4
n−1
X  
[6 − 3ih′ (0)](ξn3 − ξn ) − 2i(5ξn2 − 1) 2(5ξn2 − 1) + 6i(ξn3 − ξn )
+8 Xn Yl ξl + . (5.18)
(ξn − i)5 (ξn + i)4 (ξn − i)6 (ξn + i)4
l=1
R
We note that i < n, |ξ ′ |=1 {ξi1 ξi2 · · · ξi2q+1 }σ(ξ ′ ) = 0, so we omit some items that have no contribution for

18
 e 2 . Therefore, we get
computing Φ
Z Z +∞  n−1 X  
e 1 ′ 5ξn2 − 1 (5ξn2 − 1) + 3i(ξn3 − ξn )
Φ2 = 16 Xj Yl ξj ξl h (0) 2i +
2 |ξ′ |=1 −∞ (ξn − i)5 (ξn + i)4 (ξn − i)6 (ξn + i)4
j,l=1
 
(i − 1)(ξn2 − 1) + 12(ξn3 − ξn ) (5ξn2 − 1) + 3i(ξn3 − ξn )
+ 16Xn Yn h′ (0) − dξn σ(ξ ′ )dx′
(ξn − i)5 (ξn + i)4 (ξn − i)6 (ξn + i)4
n−1
X Z  
′ 4π 5ξn2 − 1 (5ξn2 − 1) + 3i(ξn3 − ξn )
=8 Xj Yl h (0) 2i + dξn dx′
3 Γ+ (ξn − i)5 (ξn + i)4 (ξn − i)6 (ξn + i)4
j,l=1
Z  
(i − 1)(ξn2 − 1) + 12(ξn3 − ξn ) (5ξn2 − 1) + 3i(ξn3 − ξn )
+ 8Xn Yn h′ (0)Ω3 − dξn dx′
Γ+ (ξn − i)5 (ξn + i)4 (ξn − i)6 (ξn + i)4
n−1
2368π 2 X
=− Xj Yj h′ (0)dx′ − (461 + 23i)Xn Yn h′ (0)πΩ3 dx′ , (5.19)
3 j=1

where Ω3 is the canonical volume of S 2 .

Case (a) (III) r = 1, l = −3, j = |α| = 0, k = 1.

By (3.12), we get
Z Z +∞
e3 = − 1
Φ tr[∂ξn πξ+n σ1 (∇A A −1 −3 ′
X ∇Y DA ) × ∂ξn ∂xn σ−3 (DA )](x0 )dξn σ(ξ )dx
′
2 |ξ′ |=1 −∞
Z Z +∞
1
= tr[∂ξ2n πξ+n σ1 (∇A A −1 −3 ′ ′
X ∇Y DA ) × ∂xn σ−3 (DA )](x0 )dξn σ(ξ )dx . (5.20)
2 |ξ′ |=1 −∞
By Lemma 5.3, we have

−3 i∂xn [c(ξ ′ )] 2ih′ (0)c(ξ)

∂xn σ−3 (DA )(x0 )||ξ′ |=1 = − . (5.21)
(1 + ξn2 )4 (1 + ξn2 )6
By integrating formula, we obtain
n−1
c(ξ ′ ) + ic(dxn ) X c(ξ ′ ) + ic(dxn )
πξ+n σ1 (∇A A −1
X ∇Y DA )(x0 )||ξ ′ |=1 = − Xj Yl ξj ξl − Xn Yn
2(ξn − i) 2(ξn − i)
j,l=1
n−1 n−1
ic(ξ ′ ) − c(dxn ) X ic(ξ ′ ) − c(dxn ) X
− Xj Yn ξj − Xn Yl ξl . (5.22)
2(ξn − i) j=1
2(ξn − i)
l=1

Then, we have
n−1
c(ξ ′ ) + ic(dxn ) X c(ξ ′ ) + ic(dxn )
∂ξ2n πξ+n σ1 (∇A A −1
X ∇Y DA )(x0 )||ξ ′ |=1 = − X j Yl ξj ξl − Xn Yn . (5.23)
(ξn − i)3 (ξn − i)3
j,l=1
R
We note that i < n, |ξ ′ |=1
{ξi1 ξi2 · · · ξi2q+1 }σ(ξ ′ ) = 0, so we omit some items that have no contribution for
e 3 , then
computing Φ
tr[∂ξn πξ+n σ1 (∇A A −1 −3
X ∇Y DA ) × ∂ξn ∂xn σ−3 (DA )](x0 )||ξ ′ |=1
n−1
X
−8h′ (0) −8h′ (0)
= X j Yl ξj ξl + Xn Yn .
(ξn − i)5 (ξn + i)2 (ξn − i)5 (ξn + i)2
j,l=1

(5.24)

19
Therefore, we get
Z Z +∞  n−1
X 
e3 =1 −8h′ (0) −8h′ (0)
Φ Xj Yl ξj ξl + Xn Yn dξn σ(ξ ′ )dx′
2 |ξ ′ |=1 −∞ (ξn − i)5 (ξn + i)2 (ξn − i)5 (ξn + i)2
j,l=1
n−1
X Z Z
1
=−4 Xj Yl h′ (0) ξj ξl dξn σ(ξ ′ )dx′
|ξ ′ |=1 Γ+ (ξn − i)5 (ξn + i)2
j,l=1
Z
′ 1
− 4Xn Yn h (0)Ω3 dξn dx′
(ξn − i)5 (ξn + i)2
Γ+
n−1
X  (4)  (4)
4π 2πi 1 2πi 1
=−4 Xj Yl h′ (0) dx′
− 4X Y
n n h ′
(0)Ω 3 dx′
3 4! (ξn + i)2 ξn =i 4! (ξn + i)2 ξn =i
j,l=1
2 n−1
X
10iπ 5i
= Xj Yj h′ (0)dx′ + Xn Yn h′ (0)πΩ3 dx′ . (5.25)
3 j=1
2

Case (b) r = 0, l = −3, k = j = |α| = 0.

By (3.12), we get
Z Z +∞
e 4 = −i
Φ tr[πξ+n σ0 (∇A A −1 −3 ′ ′
X ∇Y DA ) × ∂ξn σ−3 (DA )](x0 )dξn σ(ξ )dx . (5.26)
|ξ ′ |=1 −∞

By Lemma 5.3, we obtain

−3 ic(dxn ) 4iξn c(ξ)

∂ξn σ−3 (DA )(x0 )||ξ′ |=1 = − . (5.27)
(1 + ξn2 )2 (1 + ξn2 )3

By Lemma 5.2, we have

A −1 −1 −1
σ0 (∇A A A A A
X ∇Y DA ) =σ2 (∇X ∇Y )σ−2 (DA ) + σ1 (∇X ∇Y )σ−1 (DA )
n
X    −1

+ ∂ξj σ2 (∇A A
X ∇Y ) Dxj σ−1 (DA ) . (5.28)
j=1

(1) Explicit representation the first item of (5.28),

−1
σ2 (∇A A
X ∇Y )σ−2 (DA )(x0 )||ξ ′ |=1
Xn  h i
c(ξ)σ0 (DA )c(ξ) c(ξ) X 2 2
=− Xj Yl ξj ξl + 6 c(dxj ) ∂xj [c(ξ)]|ξ| − c(ξ)∂xj (|ξ| ) , (5.29)
|ξ|4 |ξ| j
j,l=1

1 P 1 P
where σ0 (DA ) = 4 s,t,i ωs,t (ei )c(ei )b
c(es )b
c(et ) − 4 s,t,i ωs,t (ei )c(ei )c(es )c(et ) + A.

We denote
1X
P1 (x0 ) = ωs,t (ei )(x0 )c(ei )b
c(es )b
c(et );
4 s,t,i
1X
P2 (x0 ) = − ωs,t (ei )(x0 )c(ei )c(es )c(et ). (5.30)
4 s,t,i

20
Then
h c(ξ)P (x )c(ξ) i h i
1 0 + c(ξ)A(x0 )c(ξ)
πξ+n σ−2 (DA −1 )(x0 )||ξ′ |=1 =πξ+n + π ξn
(1 + ξn2 )2 (1 + ξn2 )2
h c(ξ)P (x )c(ξ) + c(ξ)c(dx )∂ [c(ξ ′ )](x ) c(ξ)c(dxn )c(ξ) i
2 0 n xn 0
+ πξ+n − h ′
(0) .
(1 + ξn2 )2 (1 + ξn2 )3
(5.31)

By computations, we have
h c(ξ)P (x )c(ξ) i h ′ ′ i h ′ i
1 0 + c(ξ )P1 (x0 )c(ξ ) + ξn c(ξ )P1 (x0 )c(dxn )
πξ+n =π ξn + πξn
(1 + ξn2 )2 (1 + ξn2 )2 (1 + ξn2 )2
h ξ c(dx )P (x )c(ξ ′ ) i h 2 i
n n 1 0 + ξn c(dxn )P1 (x0 )c(dxn )
+ πξ+n + πξn
(1 + ξn2 )2 (1 + ξn2 )2
′ ′ ′
−c(ξ )P1 (x0 )c(ξ )(2 + iξn ) −ic(ξ )P1 (x0 )c(dxn )
= +
4(ξn − i)2 4(ξn − i)2
′
−ic(dxn )P1 (x0 )c(ξ ) −iξn c(dxn )P1 (x0 )c(dxn )
+ + , (5.32)
4(ξn − i)2 4(ξn − i)2

and
h c(ξ)P (x )c(ξ) i h 2 ′ ′ i h 3 ′ i
1 0 + ξn c(ξ )P1 (x0 )c(ξ ) + ξn c(ξ )P1 (x0 )c(dxn )
πξ+n ξn2 =π ξn + π ξn
(1 + ξn2 )2 (1 + ξn2 )2 (1 + ξn2 )2
h ξ 3 c(dx )P (x )c(ξ ′ ) i h ξ 4 c(dx )P (x )c(dx ) i
n 1 0 n 1 0 n
+ πξ+n n + πξ+n n
(1 + ξn2 )2 (1 + ξn2 )2
−c(ξ ′ )P1 (x0 )c(ξ ′ )iξn c(ξ ′ )P1 (x0 )c(dxn )(2ξn − i)
= +
4(ξn − i)2 4(ξn − i)2
′
c(dxn )P1 (x0 )c(ξ )(2ξn − i) c(dxn )P1 (x0 )c(dxn )(3iξn + 2)
+ + . (5.33)
4(ξn − i)2 4(ξn − i)2

Since
n−1
X
1
c(dxn )P1 (x0 ) = − h′ (0) c(ei )b
c(ei )c(en )b
c(en ), (5.34)
4 i=1

then by the relation of the Clifford action and tr(AB) = tr(BA), we have the equalities:

tr[c(ei )b c(en )] = 0 (i < n); tr[c(ξ ′ )c(dxn )] = 0;

c(ei )c(en )b
tr[P1 (x0 )c(dxn )] = 0; tr[P2 (x0 )c(dxn )] = 12h′ (0);
tr[∂xn [c(ξ ′ )]c(dxn )] = 0; tr[∂xn [c(ξ ′ )]c(ξ ′ )](x0 )||ξ′ |=1 = −8h′ (0). (5.35)

21
By (5.27), (5.32) and (5.33), we have
 h X 
c(ξ)P1 (x0 )c(ξ) i
n
tr πξ+n − Xj Yl ξj ξl × ∂ σ
ξn −3 (D −3
A ) (x0 )||ξ′ |=1
(1 + ξn2 )2
j,l=1
 h 
c(ξ)P1 (x0 )c(ξ) i
n−1
X
=− Xj Yl ξj ξl tr πξ+n × ∂ σ
ξn −3 (D −3
A ) (x0 )||ξ′ |=1
(1 + ξn2 )2
j,l=1
 h i 
+ 2 c(ξ)P1 (x0 )c(ξ) −3
− Xn Yn tr πξn ξn × ∂ξn σ−3 (DA ) (x0 )||ξ′ |=1
(1 + ξn2 )2
n−1
X  
−1 ′ −2iξn + 2ξn ′
=− Xj Yl ξj ξl tr[c(ξ )P (x
1 0 )] + tr[c(ξ )P (x
1 0 )]
2(ξn − i)4 (ξn + i)2 (ξn − i)5 (ξn + i)3
j,l=1
 
−2iξn − 1 ′ (−2 + 4i)ξn + (2 + 2i)ξn ′
− Xn Yn tr[c(ξ )P1 (x0 )] + tr[c(ξ )P1 (x0 )] . (5.36)
2(ξn − i)4 (ξn + i)2 (ξn − i)5 (ξn + i)3
R
We note that i < n, |ξ′ |=1 {ξi1 ξi2 · · · ξi2q+1 }σ(ξ ′ ) = 0, so tr[c(ξ ′ )P1 (x0 )] has no contribution for computing
e 4.
Φ
Similar to (5.36), we have
 h X 
c(ξ)A(x0 )c(ξ) i
n
tr πξ+n − Xj Yl ξj ξl × ∂ξ σ−3 (D −3
A ) (x0 )||ξ′ |=1
(1 + ξn2 )2 n
j,l=1
 h 
c(ξ)A(x0 )c(ξ) i
n−1
X
=− Xj Yl ξj ξl tr πξ+n × ∂ σ
ξn −3 (D −3
A ) (x0 )||ξ′ |=1
(1 + ξn2 )2
j,l=1
 h 
c(ξ)A(x0 )c(ξ) i
− Xn Yn tr πξ+n ξn2 × ∂ σ
ξn −3 (D −3
A ) (x0 )||ξ′ |=1
(1 + ξn2 )2
=0. (5.37)

By computations, we have
h c(ξ)P (x )c(ξ) + c(ξ)c(dx )∂ [c(ξ ′ )](x ) i h i
2 0 n xn 0 + c(ξ)c(dxn )c(ξ)
πξ+n − h ′
(0)πξn := C1 − C2 , (5.38)
(1 + ξn2 )2 (1 + ξn2 )3

where
−1 
C1 = 2
(2 + iξn )c(ξ ′ )P2 (x0 )c(ξ ′ ) + iξn c(dxn )P2 (x0 )c(dxn )
4(ξn − i)

+ (2 + iξn )c(ξ ′ )c(dxn )∂xn [c(ξ ′ )](x0 ) + ic(dxn )P2 (x0 )c(ξ ′ ) + ic(ξ ′ )P2 (x0 )c(dxn ) − i∂xn [c(ξ ′ )](x0 ) ,
(5.39)

and
 
h′ (0) c(dxn ) c(dxn ) − ic(ξ ′ ) 3ξn − 7i ′
C2 = + + [ic(ξ ) − c(dxn )] , (5.40)
2 4i(ξn − i) 8(ξn − i)2 8(ξn − i)3

where P2 (x0 ) = c0 c(dxn ) and c0 = − 43 h′ (0).

22
By (5.27), (5.39) and (5.40), we have
 n
X  
−3
tr − Xj Yl ξj ξl (C1 − C2 ) × ∂ξn σ−3 (DA ) (x0 )||ξ′ |=1
j,l=1
n−1
X  −3

=− Xj Yl ξj ξl tr (C1 − C2 ) × ∂ξn σ−3 (DA ) (x0 )||ξ′ |=1
j,l=1
 
c(ξ)P2 (x0 )c(ξ) + c(ξ)c(dxn )∂xn [c(ξ ′ )](x0 )
− Xn Yn tr πξ+n ξn2
(1 + ξn2 )2
 
′ 2 c(ξ)c(dxn )c(ξ) −3
− h (0)ξn × ∂ξn σ−3 (DA ) (x0 )||ξ′ |=1
(1 + ξn2 )3
n−1
X  
ξ 2 − 6iξn − 7 12ξn3 − 36iξn2 − 23ξn
=− Xj Yl ξj ξl 2h′ (0) n + 2ih ′
(0)
(ξn − i)5 (ξn + i)2 (ξn − i)6 (ξn + i)3
j,l=1
 
′ 2ξn − i ′ ξn4 + iξn3 + 2iξn
− Xn Yn 6h (0) + 16h (0) . (5.41)
(ξn − i)4 (ξn + i)2 (ξn − i)6 (ξn + i)3

Substituting (5.41) into (5.26) yields

Z Z +∞    
+ −1 −3
−i tr πξn σ2 (∇X ∇Y )σ−2 (DA ) × ∂ξn σ−3 (DA ) (x0 )dξn σ(ξ ′ )dx′
A A
|ξ ′ |=1 −∞
Z Z +∞  n−1
X  
ξn2 − 6iξn − 7 12ξn3 − 36iξn2 − 23ξn
=−i − Xj Yl ξj ξl h′ (0) 2 + 2i
|ξ ′ |=1 −∞ (ξn − i)5 (ξn + i)2 (ξn − i)6 (ξn + i)3
j,l=1
 
2ξn − i ξn4 + iξn3 + 2iξn
− Xn Yn h′ (0) 6 + 16 (x0 )dξn σ(ξ ′ )dx′
(ξn − i)4 (ξn + i)2 (ξn − i)6 (ξn + i)3
n−1
X Z  
4π ξ 2 − 6iξn − 7 12ξn3 − 36iξn2 − 23ξn
= Xj Yl h′ (0) 2i n 5 − 2 dξn dx′
3 Γ+ (ξn − i) (ξn + i)2 (ξn − i)6 (ξn + i)3
j,l=1
Z  
2ξn − i ξn4 + iξn3 + 2iξn
+ Xn Yn h′ (0)Ω3 6i 4 (ξ + i)2
+ 16i 6 (ξ + i)3
dξn dx′
Γ + (ξn − i) n (ξn − i) n
n−1
X   2 (4)  (5) 
4π 2πi ξn − 6iξn − 7 2πi 12ξn3 − 36iξn2 − 23ξn
= Xj Yl h′ (0) 2i − 2 dx′
3 4! (ξn + i)2 ξn =i 5! (ξ n + i) 3
ξn =i
j,l=1
  (3)  (5) 
2πi 2ξn − i 2πi ξn4 + iξn3 + 2iξn
+ Xn Yn h′ (0)Ω3 6i + 16i dx′
3! (ξn + i)2 ξn =i 5! (ξn + i)3 ξn =i
2 n−1
X
55π 3
= Xj Yl h′ (0)dx′ − Xn Yn h′ (0)πΩ3 dx′ . (5.42)
3 8
j,l=1

(2) Explicit representation the second item of (5.28),

−1
σ1 (∇A A
X ∇Y )σ−1 (DA )(x0 )||ξ ′ |=1
X n
∂Y √ X √ X √
= Xj l −1ξl + B(Y )Xj −1ξj + B(X)Yl −1ξl
∂Xj j
j,l=1 l

1X √ 1X √  √−1c(ξ)
− [c(Y )A + Ac(Y )]Xj −1ξj − [c(X)A + Ac(X)]Yl −1ξl . (5.43)
2 j
2 |ξ|2
l

23
By integrating formula, we get

1X √ 1X √  √−1c(ξ) 
+
πξn − [c(Y )A + Ac(Y )]Xj −1ξj − [c(X)A + Ac(X)]Yl −1ξl
2 j 2 |ξ|2
l

1X
n−1
√ 1X
n−1
√  √−1c(ξ) 
+
=πξn − [c(Y )A + Ac(Y )]Xj −1ξj − [c(X)A + Ac(X)]Yl −1ξl
2 j=1 2 |ξ|2
l=1

1 √ 1 √  √−1c(ξ) 
+ πξ+n − [c(Y )A + Ac(Y )]Xn −1ξn − [c(X)A + Ac(X)]Yn −1ξn
2 2 |ξ|2
1 X  −ic(ξ ′ ) + c(dx )
n−1 n−1
X n
= [c(Y )A + Ac(Y )]Xj ξj + [c(X)A + Ac(X)]Yl ξl
2 j=1 2(ξn − i)
l=1
1  c(ξ ′ ) + ic(dx )
n
+ [c(Y )A + Ac(Y )]Xn + [c(X)A + Ac(X)]Yn . (5.44)
2 2(ξn − i)
R
We note that i < n, |ξ′ |=1 {ξi1 ξi2 · · · ξi2q+1 }σ(ξ ′ ) = 0, and by the relation of the Clifford action and
tr(AB) = tr(BA), then we have the equalities:

tr[c(X)Ac(ξ ′ )c(dxn )] = 0; tr[Ac(X)c(ξ ′ )c(dxn )] = 0;

tr[c(X)Ac(dxn )c(ξ ′ )] = 0; tr[Ac(X)c(dxn )c(ξ ′ )] = 0, (5.45)

so
h   i
tr πξ+n σ1 (∇A A −1 −1
X ∇Y )σ−1 (DA ) × ∂ξn σ−3 (DA ) (x0 )||ξ ′ |=1 = 0. (5.46)

(3) Explicit representation the third item of (5.28),

Xn X Xn
1 α A A
 α −1
   √  −1

∂ξ σ2 (∇X ∇Y ) Dx σ−1 (DA ) (x0 )||ξ′ |=1 = ∂ξj σ2 (∇A A
X ∇Y ) (− −1)∂xj σ−1 (DA )
j=1 α
α! j=1
Xn h X n i √  √−1c(ξ) 
= ∂ξj − Xj Yl ξj ξl (− −1)∂xj
j=1
|ξ|2
j,l=1
Xn X n
√  √−1c(ξ) 
= −1(Xj Yl + Xl Yj )ξl ∂xj . (5.47)
j=1
|ξ|2
l=1

By integrating formula, we obtain

X
n X 
1 α  α 
πξ+n ∂ξ σ2 (∇A ∇
X Y
A
) D x σ−1 (D −1
A )
j=1 α
α!
 n−1
X√  √−1c(ξ)  
√  √−1c(ξ) 
=πξ+n −1(Xn Yl + Xl Yn )ξl ∂xn + π +
ξn −1(X Y
n n + X Y )ξ ∂
n n n xn
|ξ|2 |ξ|2
l=1
n−1
X h i∂ [c(ξ ′ )] (2 + iξn )c(ξ ′ ) ic(dxn ) i
xn
= (Xn Yl + Xl Yn )ξl − h′ (0) − h ′
(0)
2(ξn − i) 4(ξn − i)2 4(ξn − i)2
l=1
h −∂ [c(ξ ′ )] ic(ξ ′ ) iξn c(dxn ) i
xn
+ Xn Yn − h′ (0) + h ′
(0) . (5.48)
(ξn − i) 2(ξn − i)2 2(ξn − i)2

24
Substituting (5.48) into (5.26) yields
Z Z h X
+∞ n X
1 α  α 
−i tr πξ+n ∂ξ σ2 (∇A A
X ∇Y ) Dx σ−1 (DA )
−1
|ξ ′ |=1 −∞ j=1 α
α!
i
−3
× ∂ξn σ−3 (DA ) (x0 )dξn σ(ξ ′ )dx′
Z Z +∞
−3ξn3 + 4iξn2 − 3ξn + 4
=−i −8Xn Yn h′ (0) dξn σ(ξ ′ )dx′
|ξ ′ |=1 −∞ (ξn − i)5 (ξn + i)3
Z
′ −3ξn3 + 4iξn2 − 3ξn + 4
=8iXn Yn h (0)Ω3 dξn dx′
Γ+ (ξn − i)5 (ξn + i)3
 (4)
2πi −3ξn3 + 4iξn2 − 3ξn + 4
=8iXn Yn h′ (0)Ω3 dx′
4! (ξn + i)3 ξn =i
 7 15i 
′ ′
= − Xn Yn h (0)πΩ3 dx . (5.49)
2 2
Summing up (1), (2) and (3) leads to the desired equality
n−1
55π 2 X  25 15i 
e
Φ4 = Xj Yj h′ (0)dx′ + − Xn Yn h′ (0)πΩ3 dx′ . (5.50)
3 j=1 8 2

Case (c) r = 1, ℓ = −4, k = j = |α| = 0.

By (3.12), we get
Z Z +∞
e5 = −
Φ tr[πξ+n σ1 (∇A A −1 −3 ′ ′
X ∇Y DA ) × ∂ξn σ−4 (DA )](x0 )dξn σ(ξ )dx
|ξ ′ |=1 −∞
Z Z +∞
= tr[∂ξn πξ+n σ1 (∇A A −1 −3 ′ ′
X ∇Y DA ) × σ−4 (DA )](x0 )dξn σ(ξ )dx . (5.51)
|ξ ′ |=1 −∞

An easy calculation gives

n−1
c(ξ ′ ) + ic(dxn ) X c(ξ ′ ) + ic(dxn )
∂ξn πξ+n σ1 (∇A A −1
X ∇Y DA )(x0 )||ξ ′ |=1 = 2
Xj Yl ξj ξl − Xn Yn
2(ξn − i) 2(ξn − i)2
j,l=1
n n
ic(ξ ) − c(dxn ) X
′
ic(ξ ′ ) − c(dxn ) X
+ Xj Yn ξj + Xn Yl ξl
2(ξn − i)2 j=1 2(ξn − i)2
l=1
n−1
X n−1
X
1 i
= Xj Yl ξj ξl c(ξ ′ ) + Xj Yl ξj ξl c(dxn )
2(ξn − i)2 2(ξn − i)2
j,l=1 j,l=1
2iξn − 1 −2ξn − i
+ Xn Yn 2
c(ξ ′ ) + Xn Yn c(dxn )
2(ξn − i) 2(ξn − i)2
n−1
X i
+ (Xj Yn + Xn Yj )ξj 2
c(ξ ′ )
j=1
(ξn − i)
n−1
X −1
+ (Xj Yn + Xn Yj )ξj c(dxn ). (5.52)
j=1
(ξn − i)2

25
By (4.62) in [20], we have

c(ξ)σ2 (DA 3 )(x0 )||ξ′ |=1 c(ξ) c(ξ) X
 √−1c(ξ) 

σ−4 (DA −3 )(x0 )||ξ′ |=1 = − ∂ ξj c(ξ)|ξ| 2
D xj
|ξ|8 |ξ|4 |ξ|4
j
1 1 X 1 X
= 8 c(ξ) h′ (0)c(ξ) ξk c(ek )c(en ) − h′ (0)c(ξ) ξk b
c(ek )bc(en )
|ξ| 2 2
k<n k<n
5 1 
− h′ (0)ξn c(ξ) − h′ (0)|ξ|2 c(dxn ) − 2c(ξ)Ac(ξ) + 3|ξ|2 A] c(ξ)
2 4
ic(ξ)  4 ′ ′ ′ ′

+ |ξ| c(dx n )∂ x [c(ξ )] − 2h (0)c(dx n )c(ξ) + 2ξn c(ξ)∂ x [c(ξ )] + 4ξn h (0) .
|ξ|8 n n

(5.53)
R
We note that i < n, |ξ ′ |=1
{ξi1 ξi2 · · · ξi2q+1 }σ(ξ ′ ) = 0, so we omit some items that have no contribution for
e 5 . Here
computing Φ

tr[c(ei )b c(en )] = 0 (i < n); tr[c(ξ ′ )c(dxn )] = 0;

c(ei )c(en )b
tr[Ac(ξ ′ )] = 0; tr[Ac(dxn )] = 0;
tr[∂xn [c(ξ ′ )]c(dxn )] = 0; tr[∂xn [c(ξ ′ )]c(ξ ′ )](x0 )||ξ′ |=1 = −8h′ (0). (5.54)

Also, straightforward computations yield

tr[∂ξn πξ+n σ1 (∇A A −1 −3

X ∇Y DA ) × σ−4 (DA )](x0 )||ξ ′ |=1
n−1
X 
1
= Xj Yl ξj ξl h′ (0) [4iξn3 + (4 − 22i)ξn2 − (24 − 12i + 4ξi2 )ξn + 2iξi2 + 4]
(ξn − i)2 (1 + ξn2 )3
j,l=1

1 2 2
+ [48ξn − 64iξn − 16ξi ]
(ξn − i)2 (1 + ξn2 )4

1
+ Xn Yn [−8ξn4 + (44 + 4i)ξn3 − (28 + 26i + 8iξi2 )ξn2 + (16 − 4i)ξn − 6iξi2 − 4]
(ξn − i)2 (1 + ξn2 )3

1 3 2 2 2
+ [96iξn + 80ξ n + (64i − 32iξi n)ξ + 16ξi .] (5.55)
(ξn − i)2 (1 + ξn2 )4

Substituting (5.55) into (5.51), we get

  n−1
X  5 103i 
Φe 5 = 323 − 31i π 2 Xj Yj h′ (0)dx′ + − + Xn Yn h′ (0)πΩ3 dx′
60 60 j=1
4 32
 7 49i 
+ − + Xn Yn h′ (0)π 2 dx′ . (5.56)
3 24
Pn−1
Let X = X T + Xn ∂n , Y = Y T + Yn ∂n , then we have j=1 Xj Yj (x0 ) = g(X T , Y T )(x0 ). Now Φe is the sum
e (1,2,···,5) . Combining with the five cases, this yields
of the Φ

5
X h3673 793i   7 49i 
e=
Φ ei =
Φ − − Xn Yn Ω3 + − + Xn Yn π
i=1
8 32 3 24
 45937 169i  i
+ − + πg(X T , Y T ) πh′ (0)dx′ . (5.57)
60 60
So, we are reduced to prove the following.

26
Theorem 5.4. Let M be a 4-dimensional compact manifold with boundary and ∇A be an orthogonal connec-
A −1 −3
tion. Then we get the spectral Einstein functional associated to ∇A
X ∇Y DA and DA on compact manifolds
with boundary

] + (∇A
Wres[π A −1 + −3
X ∇Y DA ) ◦ π (DA )]
Z   Z
4π 2 1
= Ric(V, W ) − sg(V, W ) volg − 2 sg(V, W )volg
3 M 2 M
Z h 3673 793i   7 49i 
+ − − Xn Yn Ω3 + − + Xn Yn π
∂M 8 32 3 24
 45937 169i  i
+ − + πg(X T , Y T ) πh′ (0)vol∂M , (5.58)
60 60
where s is the scalar curvature.

Acknowledgements

This work was supported by NSFC. 11771070. The authors thank the referee for his (or her) careful
reading and helpful comments.

References

References
[1] P. B. Gilkey, Invariance theory, the heat equation, and the Atiyah-Singer index theorem. Inc. USA, (1984).
[2] H. Figueroa, J. Gracia-Bondía, and J. Várilly, Elements of noncommutative geometry. Birkhäuser Boston, (2001).
[3] L. Dabrowski, A. Sitarz, P. Zalecki, Spectral metric and Einstein functionals. Adv. Math. 2023, 427: 109128.
[4] T. Ackermann, A note on the Wodzicki residue. J. Geom. Phys. 20, 404-406, (1996).
[5] M. Wodzicki, Local invariants of spectral asymmetry. Invent. Math. 75(1), 143-178, (1984).
[6] M. Wodzicki, Non-commutative residue I, Lecture Notes in Math. Springer, New York, Vol. 1289, 320-399, (1987).
[7] V. W. Guillemin, A new proof of Weyl’s formula on the asymptotic distribution of eigenvalues. Adv. Math. 55, no. 2,
131-160, (1985).
[8] A. Connes, Quantized calculus and applications. 11th International Congress of Mathematical Physics(Paris,1994), Inter-
nat Press, Cambridge, MA, 15-36, (1995).
[9] A. Connes, The action functinal in noncommutative geometry. Comm. Math. Phys. 117, 673-683, (1988).
[10] D. Kastler, The Dirac operator and gravitation. Comm. Math. Phys. 166, 633-643, (1995).
[11] W. Kalau, M. Walze, Gravity, noncommutative geometry and the Wodzicki residue. J. Geom. Physics. 16, 327-344, (1995).
[12] B. V. Fedosov, F. Golse, E. Leichtnam, E. Schrohe, The noncommutative residue for manifolds with boundary. J. Funct.
Anal. 142, 1-31, (1996).
[13] Y. Wang, Differential forms and the Wodzicki residue for manifolds with boundary. J. Geom. Physics. 56, 731-753, (2006).
[14] Y. Wang, Differential forms and the noncommutative residue for manifolds with boundary in the non-product case. Lett.
Math. Phys. 77, 41-51, (2006).
[15] Y. Wang, Gravity and the noncommutative residue for manifolds with boundary. Lett. Math. Phys. 80, 37-56, (2007).
[16] Y. Wang, Lower-dimensional volumes and Kastler-kalau-Walze type theorem for manifolds with boundary. Comm. Theor.
Phys. Vol 54, 38-42, (2010).
[17] J. Wang, Y. Wang, The Kastler-Kalau-Walze type theorem for six-dimensional manifolds with boundary. J. Math. Phys.
56, 052501 (2015).
[18] J. Wang, Y. Wang, T. Wu, Dirac operators with torsion, spectral Einstein functionals and the noncommutative residue.
J. Math. Phys. 64, 102505(2023).
[19] H. Li, T. W, Sub-signature operators and the Kastler-Kalau-Walze type theorem for five dimensional manifolds with
boundary. J. Nonlinear Math. Phys. 30, 1032-1068(2023).
[20] T. Wu, S. Wei, Y. Wang, Sub-signature operators and the Kastler-Kalau-Walze type theorem for manifolds with boundary.
J. Geom. Phys. 174(2022)104455.
[21] Y. Yu, The Index Theorem and The Heat Equation Method, Nankai Tracts in Mathematics-Vol.2, World Scientific
Publishing, (2001).

27