
Use R=Pi+A+C+X, h=d+z, 2(Pi-h)=sum f, and Omega=-2e_M,w=2M.
The baseline is Pi>=25 and Pi>=23+h. Actual center star mass is1.
There is one inactive center q, with f(q)>=5 and no f6. Helper-free
active centers have f>=9. A nonzero class center has f>=8; f8 forces
its sole class to1/2 and all marginals zero. A sole nonzero class at
f9 is1, with a glue-free F4 line and a dense F5 together line.
There is no f10 for a sole unit class. A nonzero class with a nonzero
separating marginal requires f>=13. These local inputs permit balanced
collisions. The literal block equation is star sum+A sum=beta.

## 1. Unit-class stars at cost at most11 remain on together lines

Let a center have exactly one nonzero class, of value1 and source U,
with arbitrary additional balanced classes, and f<=11. It has exactly
two occupied N-lines, both together lines for U; all its glues and
actual stars lie on those two lines.

An actual star is required, so N<=4. Two main N lie on distinct lines.
Three N must all belong to the nonzero class and occupy three distinct
lines, each of cost at least4, exceeding11. Four N in that class occupy
four distinct lines and are also too costly. The other four-N option
is a minimal main pair plus one balanced pair. Each N-line costs at
least4, including a two-N collision whose aggregate is zero. Thus at
most two such lines fit, and the balanced pair is confined to the two
main lines. Any additional N-free nonempty line costs at least5, beyond
the two main lines' combined minimum8. This proves the assertion.

In particular the actual stars of such a center have source pair equal
to its containing half minus that center, hence containing its U-mate.
If V is disjoint from U, none can have source U or V. This conclusion
uses incidence, not positivity of actual stars at cost11.

## 2. Z=0: reduce the source graph to two disjoint D

An M-edge containing q cannot be a D source, so its array remains
nonzero and A>=1. Nonzero coordinate flow would require C>=6, exceeding
Pi+A+C>=32. Thus flow is zero, with idle loops still counted.

Repeated or intersecting D sources give Pi>=28 by the degree
charge. R29 then forces Pi28,A1,C=X0. But A1 requires cancellation of
two M-arrays and no outside nonzero array. With Z=0, every occupied
block has beta=2D!=0, so the D are exactly the two unit columns in
those M-blocks. Their sources are disjoint. This contradicts repeated
incidence and also excludes d>=3 on five active vertices.

With one D, E23 gives endpoint cost at least23. The other three active
centers cost at least27, and q costs5; hence Pi>=29 and A>=2 exceeds
R29. Therefore exactly two D remain, on disjoint sources U,V, with
coefficients a,b. Let w be the remaining helper-free active center.
If t of a,b differ from1/2, then Pi>=25+t. The possible Pi are25--28.

### Pi25

Both coefficients are halves; their four endpoints have f8 and B=0,
while w has f9 and q has f5. With A<=4, both D cannot be outside M,
since that would require five arrays. At a source U whose block is in
M, H0_U=0. Its own helper contributes1 and the other contributes either
0 or1/2. Zero-flow F instead requires H0_U>=1/2. Contradiction.

### Pi26

If both coefficients are halves, A<=3 forces both D in M. Each endpoint
has cost at most10 and all its marginals vanish. F at either source
requires H0=1/2 instead of0.

Otherwise t=1 and equality in Pi>=25+t forces f9 at the nonhalf D's
endpoints, f8 at the half D's endpoints, f9 at w and f5 at q. The
nonhalf coefficient is therefore1. A<=3 forces its D block into M:
if the unit D were outside, no M-array could cancel, giving A>=4.
At its source U, the f9 endpoint rows give H0_U=-1. The other half D
contributes either0 or1/2 there, so F requires H0_U=0 or-1/2, impossible.
This permits an outside half D and any mixed slack that fits the count.

### Pi27

A<=2 forces both D in M and at least one, say a, to equal1. The total
line cost is50. Each of its endpoints has cost at most11, since the
other costs at least9, the other two D endpoints at least16 together,
w at least9 and q at least5. Section1 puts their total marginal1/2
on together lines, giving H0_U=-1. In F the other D contributes b,
so H0_U=-b and b=1.

Equality now forces all four D endpoints to have f9, w f9 and q f5.
There is just one nonzero M-array. With A<=2, an outside zero-total
array cannot carry support, because it would need two columns in
addition to the nonzero array. All A are in group M, which has no
block-alternating columns in (M,U) or(M,V). Both blocks therefore
need actual star mass2. Section1 excludes the four D endpoints from
those blocks. The remaining centers w,q have nonnegative actual stars
of total mass1 each, by the B9 and inactive-F5 laws. Their
combined contribution is at most2, whereas the required sum is4.
Mixed columns vanish in these block equations; idle loops cannot fit
because C<=1 and Z=0 forbids C1.

### Pi28: the additional direction is covered

The exact budget is A1,C=X0. Array cancellation again forces a=b=1
in two M-blocks. The minimum doubled line cost is50; the actual total
is52. Unit endpoints have no f10, and q has no f6. Thus precisely one
of the following occurs:

1. One D endpoint has f11; the other D endpoints and w have f9, q f5.
2. The inactive q has f7; all five active centers have f9.
3. The helper-free active w has f11; all D endpoints have f9, q f5.

In case1, section1 still excludes all D-endpoint stars from the two
D blocks. B9 and inactive F5 again supply at most2 instead of4.

In case2, all other centers' nonzero glues have magnitude1/2, so every
glue at q has that magnitude. Its N-free f7 line has total marginal
-1/2 and is its only nonempty line. The F7 form is(r,t,-t), up to
permutation. If every glue has magnitude1/2 and r=-1/2, the star
vector has just one nonzero entry, making the line F5 rather than F7.
This is impossible. No nonempty zero-marginal line fits in cost7.

In case3, w has at least two nonzero glues. Here is the full incidence
argument. Choose an endpoint u of U and its dense f9 half, which
contains its U-mate. Its third vertex cannot lie in V: the opposite
V-endpoint would receive a glue on a separating line, impossible at
its f9 cost. Thus the third vertex is w or q, and both V-endpoints
lie in the complementary half. Both receive a nonzero glue, forcing
their dense lines into this same split. Those two dense lines in turn
force the other U-endpoint's dense line into it. Consequently all
four D endpoints are dense in one split, with U and V on opposite
sides. The vertex w receives distinct glues from the two D endpoints
opposite it. The exact checker covers every matching, choice of D
edges and choice of w, with both possible splits.

Write f(w)=g+2s+2N=11. Its glue count is odd, so g>=2 implies g>=3.
Its balanced N count is positive and even, and s>=1. Therefore N=2
and s<=2. Both D blocks still require mass2, all four D endpoints
contribute zero, and q has nonnegative stars of total mass1. The w
contribution to each D block must consequently be at least1. At most
two w-stars are available, so there must be one positive star in each
block and none elsewhere. Their total is1. Together with q's total1
they cannot supply the required4. No positivity premise was imposed
on w before the two block equations forced this conclusion.

All Z=0 cases are excluded.

## 3. Z present: cancellation costs exclude every budget

A sole Z needs C>=6 and is impossible since Pi>=25. Thus z>=2,C>=2,
Pi<=27 and h<=4.

Suppose A0. All three M-arrays cancel and h>=3. The M-edge containing
q has just one active endpoint, whose class must equal2 and whose
cost is at least11. Other active costs are at least8 and q costs5,
so Pi>=24+h. For h4 this already exceeds27. For h3 exactly one helper
must cancel each M-array, with no other helper. Every D is1 and every
Z is2. If there are one D and two Z, their supports have four distinct
active endpoints, costing at least18+22, while the fifth costs9 and
q costs5; hence Pi>=30. Three Z give Pi>=31. Thus A>=1, Pi<=26,h<=3.

For h2 there are exactly two Z. If attached at the same center, their
cost13, the four other active costs36 and q's5 give Pi>=29. Otherwise
the two Z centers cost at least8 each and the other three active
centers cost9 each, giving Pi>=26. R29 forces Pi26,A1,C2,X0. Two
M-arrays would have to cancel, so the Z are in different M-blocks and
both equal2. Their cost rises from8 to at least11 each, contradicting
Pi26.

For h3, three Z can reach at most three active vertices; at least two
active vertices are helper-free. Their minimal total gives

    Pi >= ceil((3*8+2*9+5)/2)+3 = 27.

The only other possibility is one D and two Z. They reach at most four
active vertices. Pi26 gives doubled line budget46, attained only with
four distinct helper-bearing f8 vertices, one helper-free f9 vertex,
and q f5. Every nonzero class is1/2, so every block has beta at most1.
No M-array cancels, contradicting the necessary A1. More collisions
or helper-free vertices increase the cost. This excludes every Z case.
