Assume matching corrections, R=28, nonzero normalized helpers, and m in {5,6}.

## 1. Exact stated inputs

Use the notation and identities of the preceding lower-bound-31 component:

    R=Pi+A+C+X,  Pi=stars+glues+N+D+Z,  C=L3+E,
    h=d+z,  m=|K|,  2Pi=sum_x f(x)+2h,
    Pi>=15+2m,  Pi>=15+ceil(3m/2)+h.

Matching rigidity O29 applies since R<=29: Omega=-2e_M, w=2M. The
array totals are beta-2 in the three M-blocks and beta elsewhere, with
beta=2D+Z_u+Z_v. Every nonzero array total needs its own alternating
column; a zero-total array with nonzero support needs at least two.

Actual star mass is one at every center. Inactive f>=5. Active f>=8,
or f>=9 when Gamma=sum(zeta)-1/2 is nonzero. Active f8 is glue-free
and has a single nonzero class of value 1/2. The proved f9 classification
puts all glue in one of two class lines. At a single nonzero class with
nonzero Gamma and f9, its value is one and those two lines are together
lines for its source. Arbitrary extra balanced classes were allowed in
the proof of that classification.

Part H section 3.2 supplies the unconditional sharing lemma: **two
distinct nonzero helper classes at a center imply f>=13**, including
classes with the same source and arbitrary balanced collisions.
Each nonzero class requires at least two N. The four columns of one
class lie on four distinct private lines; there are two columns on each
h side. At a zero class, each h-side support is zero or two, so its
support is even. At Z=0, a vertex of D-degree r>=3 has f>=4r+2 from
the 2r N and at least one actual star alone.

At Z=0, flow is either zero or costs C>=6; C=1 is impossible directly
from S. If any Z is present, normalized C>=2: a C=1 configuration would
admit the proved support-decreasing kernel move, contradicting the
already proved matching R>=28. A single Z globally costs C>=6 by
Part G, Lemma L1+. Each Z block needs a loop.

The general mixed-free equations (F), (B) and source balance (T) are
used with their stated scope. In particular, when flow is zero and
all nonzero classes have one source U, (T) says NT_U=w_U-1; balanced
classes at other centers remain allowed. The block-alternating supports
of (M,U), U in M, are disjoint from all twelve group-M alternating
columns (the proved literal incidence check). Block-star supports form
a partition; this prevents double counting in section 5.

## 2. A zero-helper f9 center has nonnegative actual stars

**Lemma B9.** Suppose every zeta class at an active center is zero
and its total cost is f=9. Then it has exactly two nonzero N and every
actual star coefficient at that center is nonnegative. More precisely,
its nonzero actual stars are either two coefficients 1/2 or one
coefficient one. All its nonzero glues have magnitude 1/2.

Here Gamma=-1/2 requires a glue, and actual star mass one requires a
star. Thus f>=1+2+2N. Balanced parity and activity imply N=2. They
belong to one balanced class on one h side, with opposite coefficients
on two distinct lines. Each line costs at least four; no third nonempty
line fits. The costs are therefore F4 and F5. The F4 line has no glue
and one star equal to the negative of its N coefficient.

Let n be the N coefficient on the F5 line, whose marginal is r=-1/2,
and let nu be its single-position N vector. The other N is -n, so the
F4 star is n. The F5 star vector is l-r*1-nu. Its cost is
g+2s+2=5, giving exactly two possibilities.

- If g=1,s=1, the glue r is at one position. The N must be at a
  different position and cancel one of the two entries -r. Hence
  n=-r=1/2, and the surviving F5 star is -r=1/2. Together with the
  F4 star this gives two positive halves.
- If g=3,s=0, summing l=r*1+nu gives r=3r+n, so n=-2r=1. The F4
  star is one. The three glues are 1/2,-1/2,-1/2.

This proves the claim. No general positivity premise is imposed on
other active vertices, on larger line costs, or on effective stars.

## 3. The cost of intersecting D sources when Z=0

For Z=0, let r_x be the number of incident nonzero D columns and assign
pi(x)=f(x)/2+r_x/2. This counts each D exactly once in sum_x pi(x)=Pi.
At an active center:

    r_x=0: pi(x)>=9/2,
    r_x=1: pi(x)>=9/2,
    r_x=2: pi(x)>=13/2+1=15/2,
    r_x>=3: pi(x)>=1+5r_x/2>=17/2.

The second line uses f>=8; equality has its unique D coefficient 1/2.
The third line uses the proved two-class f13 lemma. An inactive center
has pi>=5/2. Therefore

    Pi >= 15+2m,
    any center incident to two or more D forces Pi >= 18+2m.    (DC)

The latter statement is only asserted here for Z=0. It counts the extra
three units at that center once, and places every other active center
at its valid 9/2 baseline. Parallel D sources in different blocks are
included, because they are distinct nonzero classes.

In particular, at m=5 and Pi<=27, all D source pairs are disjoint, so
d<=2. If t of their coefficients differ from 1/2, each such D makes
both endpoint Gamma values nonzero, giving the sharper bound

    Pi>=25+t.                                                (D5)

Equality specifies f8 at the half-valued D endpoints, f9 at the other
D endpoints and at every helper-free active center, and f5 at the sole
inactive center. No assumption about the existence of these equality
patterns is made.

## 4. Six attachment vertices

Here Pi>=27. Z is impossible because C>=2 would give R>=29. Thus Z=0.

If Pi=27, (DC) excludes repeated incidence, and every D has coefficient
1/2 at its f8 endpoints. Such a D cannot cancel an M-array: its beta
is one. All three M-arrays remain nonzero, so A>=3 and R>=30.
If there were no D, the same A>=3 conclusion holds.

If Pi=28, the budget forces A=C=X=0. All three M-arrays must then
cancel. At Z=0 each needs its D=1 in that M-block. Hence d>=3 and
all six endpoints have a value-one class. If there is any repeated
incidence, (DC) gives Pi>=30 already. Otherwise the three D are the
three matching edges, each endpoint has Gamma=1/2 and f>=9, and
Pi>=6*9/2+3=30. Contradiction. Larger Pi cannot fit R28.

Thus every m6 case is excluded, including its zero-helper case.

## 5. Five attachment vertices, Z=0 and a nonzero D

One matching edge contains the inactive center. No D can have that
source, so its array remains -2 and A>=1. Hence Pi<=27. Equation (DC)
excludes overlapping or repeated D sources. There are one or two D on
disjoint source pairs. Since Pi>=25 and A>=1, C<=2. Z-free nonzero
flow requires C>=6, so the flow is zero. The following support budgets
also exclude idle loops whenever necessary.

### 5.1 Pi=25

All D coefficients are 1/2 by (D5). Consequently A>=3, so the exact
R28 budget is A3,C=X0. Any outside-M D would require another array,
so all D lie in M. At the source U of any D, both endpoints have f8
and are glue-free, hence H0_U=0. In (F), its own beta is one; the only
possible other D is another M edge and contributes 1/2 on U. Thus

    H0_U = 2-1-(d-1)/2 >= 1/2,

contradiction. This proof uses the actual helper-modified (F) equation;
it does not assert H0=2M.

### 5.2 Pi=26

Now A<=2. The unchanged M-array gives A>=1, so C<=1; the S rows
exclude C1, and C0 follows. At least one M-array must cancel, hence
one D is the coefficient-one column in (M,U). If there is a second D,
it must also lie in M: otherwise two unchanged M-arrays plus its
outside array require A>=3. Bound (D5) permits at most one D whose
coefficient differs from 1/2.

With just one D=1, equality forces both its endpoints to have f9.
All their glues are on together class lines, so NT_U=0. Source balance
(T) requires NT_U=1. The other three active centers can have balanced
N; they do not occur in this source balance.

With two D, their coefficients are 1 and 1/2 on two disjoint M-edges.
Equality gives f9 at the first pair and f8 at the second. At either
endpoint of the D=1 edge U, all marginal Gamma=1/2 is on together
lines, so B toward its mate is -1. Hence H0_U=-1. Equation (F), with
own beta two and the other beta one, instead requires

    H0_U=2-2-1/2=-1/2.

Contradiction. Mixed X would not alter these equations.

### 5.3 Pi=27: the new star-mass obstruction

Only A1,C=X0 fits. Canceling two M-arrays forces exactly two D=1 on
two disjoint matching edges U,V. There are no other D, by disjointness
and the inactive endpoint. The four D endpoints have f9, the fifth
active center has f9 with all helpers zero, and the inactive center has
f5: these are forced by equality in Pi>=25+t with t=2.

The only nonzero alternating array is the third M-array. Since A=1,
its unique column belongs to group M. The proved block/group incidence
therefore puts no block-alternating column in either D block (M,U) or
(M,V). The literal B equation requires actual star sum two in each of
those blocks, for combined sum four.

At a D endpoint, f9 confines its actual stars to its two together class
lines. Their star source pairs are the containing half minus the center:
they contain its D mate and exclude the center. Such a source is neither
U nor V. Thus none of the four D endpoints can contribute a star to
either D block. This excludes their coefficients by incidence, without
assuming their stars have a sign.

The remaining active center has nonnegative actual stars by Lemma B9, with total mass one. The inactive f5 center also has nonnegative
actual stars and total mass one: its marginal is -1/2 and its F5 form
is a single -1/2 glue or three glues 1/2,-1/2,-1/2. Their combined
contribution to any subset of distinct star blocks is at most two.
It cannot supply the required four. Contradiction.

The bounds Pi>=25 and A>=1 exclude any additional Pi budgets. This
completes every Z=0, nonzero-helper m5 case.

## 6. Five attachment vertices with Z present

Pi>=25 throughout. Exactly one Z is excluded by L1+, since C>=6
would give R>=31. Hence z>=2 and C>=2. The helper count satisfies

    Pi>=23+h,  h=d+z>=2.

If h>=4, Pi+C>=27+2>28. Thus h=2 or h=3.

For h=2, there are exactly two Z and no D. They give nonzero helper
values at at most two centers. At least three of the five active
centers therefore have Gamma=-1/2 and f>=9. The others have f>=8,
and the inactive center has f>=5. Consequently

    Pi >= ceil((3*9+2*8+5)/2)+2 = 26.

With C>=2, the budget leaves A=0. But two helpers can affect at most
two of the three M-arrays, leaving a nonzero array. Contradiction.

For h=3, Pi>=26 and C>=2 again force A=0. Canceling all three
M-arrays uses all three helpers, precisely one per M-block, with no
other helper block. Each single Z must consequently equal two; a
single D equals one. Since z>=2, at least two distinct centers have a
sole nonzero class of value two, and each costs at least eleven. The
other three active centers cost at least eight, and the inactive one
at least five. Therefore

    Pi >= ceil((2*11+3*8+5)/2)+3 = 29,

contradiction. This exhausts Z placements, including the possibility
A=0; no positive lower bound on A was assumed for the general Z case.
