Next, by 3.63 we have W
2
T ≤ Z
R
d
Z
t
2
t
1
s
−γα
Z
t
2
−t
1
p
u
du ∗
e ϕ
T
y
1+ β
dsd y. The Young inequality, substitution u
′
= ut
2
− t
1
, and self-similarity imply W
2
T ≤ Ct
1− γα
2
− t
1− γα
t
2
− t
1 1+
β−dαβ
|| f ||
1+ β
1+ β
||ϕ||
1+ β
1
≤ C
1
t
1− γα
2
− t
1− γα
1 2+
β−dαβ
, 3.67
by 3.41. Combining 3.66, 3.67, 3.61 and using 3.59, we obtain 3.57.
It remains to prove 3.56. Applying the usual substitutions to AT given by 3.58 we obtain
AT =
H
T
T
2−d α−γα
F
2 T
Z
1
Z
1 s
Z
R
3d
|x|
−γ
p
s
x − y e
ϕ
T
yp
u−s
y − z e
ϕ
T
zχsχud x dzd y duds, hence, by 3.60, 3.55 and the Hölder inequality,
AT ≤ C H
T
T
2−d α−γα
F
2 T
Z
t
2
t
1
Z
t
2
s
s
−γα
|| e
ϕ
T
||
1+ β
β
||p
u−s
∗ e
ϕ
T
||
1+ β
duds. We use 3.65, 3.45 and 2.5, obtaining
AT ≤ C
1
t
1− γα
2
− t
1−1 α
1
t
2
− t
1 1−d
αβ1+β
, which implies 3.56 by 3.59. This completes the proof of tightness.
3.4 Proof of Theorem 2.1b
We prove the theorem for γ = d
2 + β
1 + β
α 3.68
see Remark 2.2c. Recall that in this case kT occurring in 2.8 and 2.9 is log T . According to the discussion in Section 3.1 it suffices to prove 3.12, 3.13 and 3.15. By the form
of the limit process see 2.3, 3.12 is equivalent to
lim
T →∞
I
1
T = σS
d−1
α Z
R
d
Z
1
p
s
y Z
1 s
p
u−s
yχudu
1+ β
dsd y Z
R
d
ϕzdz
1+ β
, 3.69
where σS
d−1
is the measure of the unit sphere in
R
d
= 2 if d = 1.
1350
By 3.9, 3.2, 3.6, using similar substitutions as in the previous section, we obtain I
1
T =
1 log T
Z
R
d
Z
1
Z
R
d
p
s
x T
−1α
− y Z
1 s
Z
R
d
p
u−s
y − zχu e
ϕ
T
zdzdu
1+ β
1 1 + |x|
d
d y dsd x, 3.70
where e
ϕ
T
is given by 3.34. We write I
1
T = I
′ 1
T + I
′′
1
T + I
′′′
1
T , 3.71
where I
′ 1
T = 1
log T Z
1 |x|T
1 α
Z
1
Z
R
d
... 3.72
I
′′
1
T = 1
log T Z
|x|≥T
1 α
Z
1
Z
R
d
... 3.73
I
′′′
1
T = 1
log T Z
|x|≤1
Z
1
Z
R
d
.... 3.74
Passing to polar coordinates in the integral with respect to x we have I
′ 1
T = 1
log T Z
T
1 α
1
Z
S
d−1
Z
1
Z
R
d
p
s
w r T
−1α
− yg
s
∗ e
ϕ
T
y
1+ β
r
d−1
1 + r
d
d y ds σd wd r,
where g is defined by 3.35. The crucial step is the substitution r
′
= log r
log T ,
3.75 which gives
I
′ 1
T = Z
1 α
Z
S
d−1
Z
1
Z
R
d
p
s
w T
r−1 α
− yg
s
∗ e
ϕ
T
y
1+ β
T
r d
1 + T
r d
d y ds σd wd r.
It is now clear that if one could pass to the limit under the integrals as T → ∞, then I
′ 1
T would converge to the right hand side of 3.69. This procedure is indeed justified by the fact that for f
defined by 3.22 we have f ∈ L
2+ β
R
d
, 3.76
which follows from 3.68, 3.23 and 3.24. We omit the details, which are similar to the argument in [BGT6] see 3.51 therein.
Next we show that I
′′
1
T and I
′′′
1
T tend to zero. In I
′′
1
T see 3.73 we substitute x
′
= x T
−1α
and we use 3.36, obtaining I
′′
1
T ≤ C
log T Z
|x|1
Z
1
Z
R
d
p
s
x − y f ∗ e
ϕ
T
y
1+ β
T
d α
1 + |x|
d
T
d α
d y dsd x ≤
C
1
log T || f ∗
e ϕ
T
||
1+ β
1+ β
≤ C
1
log T || f ||
1+ β
1+ β
||ϕ||
1+ β
1
→ 0. 1351
I
′′′
1
T see 3.74 is estimated as follows: I
′′′
1
T ≤ C
log T Z
|x|≤1
f ∗ f ∗ e
ϕ
T 1+
β
x T
−1α
d x ≤
C
2
log T || f ∗ f ∗
e ϕ
T 1+
β
||
∞
≤ C
2
log T || f ||
2+ β
|| f ∗ e
ϕ
T 1+
β
||
2+β1+β
≤ C
2
log T || f ||
2+ β
2+ β
||ϕ||
1+ β
1
→ 0, by 3.76. This and 3.71 prove 3.69.
To prove 3.13 we use 3.14 and easily obtain I
2
T ≤ C H
T
T
2−2d α
F
2 T
Z
R
2
f ∗ e
ϕ
T
f ∗ e
ϕ
T
x T
−1α
1 1 + |x|
d
d x. We write the right-hand side as the sum of integrals over {|x| ≤ T
1 α
} and {|x| T
1 α
}. To estimate the integral over {|x| ≤ T
1 α
} we use sup
T 2
1 log T
Z
|x|≤T
1 α
1 1 + |x|
d
d x ∞,
3.77 and in the second integral we apply 1
1 + |x|
d
≤ T
−dα
. For each of the integrals we use ap- propriately the Hölder inequality, properties of the convolution and 3.45, obtaining the estimates
C
1
T
2d α12+β−11+β
and C
2
T
dα12+β−21+β
, respectively the factors involving negative powers of H
T
and log T have been estimated by constants. These bounds tend to zero as T → ∞ by 3.68. We omit details. This proves 3.13.
To prove 3.16 and 3.17 we use 3.18 and 3.19. Again, we consider separately the integrals over {|x| ≤ T
1 α
} and {|x| T
1 α
}, and apply the same tricks as for I
2
T . For J
1
T we obtain the estimate J
1
T ≤ C T
dα1+β2+β−1
→ 0 log T and H
T
appear with negative powers only, whereas J
2
T ≤ C
1
T H
β T
log T
β
+ C
2
T H
β T
log T
1+ β
→ 0 by assumption 2.9. The proof of Theorem 2.1 is complete
3.5 Proof of Proposition 2.3