diff f6bf760..6fbd3ac weighted_dilation_totient_fibers_v2.txt
diff --git a/weighted_dilation_totient_fibers_v2.txt b/weighted_dilation_totient_fibers_v2.txt
index 16ee20d..c3b94f7 100644
--- a/weighted_dilation_totient_fibers_v2.txt
+++ b/weighted_dilation_totient_fibers_v2.txt
@@ -431,7 +431,7 @@ The constants are absolute: no factorization of the coefficients enters.
 PROOF.  Integration gives the diagonal H sum_n |c_n|^2.  For n != m the
 exponential integral is at most 2/|log(n/m)| << Y/|n-m|.  Use
 2|c_n c_m| <= |c_n|^2 + |c_m|^2 and sum 1/|n-m|.  For the discrete form
-use the one-dimensional Sobolev inequality on unit intervals: |P(t)|^2 is
+use the one-dimensional Sobolev ineuality on unit intervals: |P(t)|^2 is
 bounded by a constant times the integral of |P|^2 + |P'|^2 over that unit
 interval, and unit intervals overlap boundedly.  The derivative has
 coefficients i (log n) c_n, so apply the first bound twice.  [End]
@@ -455,7 +455,7 @@ gives a normalized bound of the shape
 
 where B_(k+1) controls the differenced sum, and the differenced phase has
 second derivative of size lam_k H_0^(k-2) on an interval which is again an
-interval.  (The inequality follows by averaging H_0 translates and
+interval.  (The ineuality follows by averaging H_0 translates and
 applying Cauchy-Schwarz; its zero-shift term is precisely the displayed
 H_0^(-1).)
 
@@ -587,7 +587,7 @@ For a block put
    e_j = E E_j <= ( sum_{p in B_j} g(p) )^(h_j+1) / (h_j+1)!
               <= C^(h_j+1) / (h_j+1)!,                               (4.23)
 
-using positivity of the monomial coefficients for the first inequality,
+using positivity of the monomial coefficients for the first ineuality,
 and (4.14) for the second: a block is contained in (w, w^2] for w equal to
 its left endpoint, so the sum of g over the block is at most C.  Also
 
@@ -1846,7 +1846,7 @@ positive mass satisfies
    sup |K_Lambda| <= Q^2/xi,
    sup_X E_D |K_Lambda(D,X)| <= Q^2.                                   (7.12)
 
-The last inequality holds because, for fixed retained X, the mass nu_(Lambda
+The last ineuality holds because, for fixed retained X, the mass nu_(Lambda
 j) of its cell cancels the denominator; the kernel vanishes for discarded X
 and is supported only where |log D - log X| < h.
 
@@ -2571,7 +2571,7 @@ of j_0 primes; Cauchy-Schwarz on each fiber proves this even with
 repeated primes.
 
 For a polynomial Q(t) = sum over n <= Y of c_n n^(it) and unit-separated
-points in [-O(Y), O(Y)], the unit-interval Sobolev inequality and
+points in [-O(Y), O(Y)], the unit-interval Sobolev ineuality and
 bounded overlap give
 
    sum over j of |Q(t_j)|^2                                            (8.42)
@@ -2676,7 +2676,7 @@ Otherwise put eps_0 = 1/10.  The overlapping ranges
 cover cN < |u| <= x^2 using 2 <= k <= K = ceil(4/tau) + 3.  The k-th
 derivative of the phase has constant sign and magnitude comparable to
 |u| N^(-k).  For k = 2 the second-derivative test directly saves a
-positive power.  For k > 2 the van der Corput differencing inequality
+positive power.  For k > 2 the van der Corput differencing ineuality
 is applied k-2 times with positive integer shifts at most
 H_0 = floor(N^(eps_0/(2k))).  After shifts h_1,...,h_(k-2), the second
 derivative is an iterated integral of the k-th derivative, hence of
@@ -2691,7 +2691,7 @@ On an interval of length m <= CN the second-derivative bound gives
 
 To justify the estimate uniformly for short subintervals, extend each
 sequence by zero in a containing interval of length O(N).  The
-differencing inequality for its normalized sum has the form
+differencing ineuality for its normalized sum has the form
 B_j^2 << H_0^(-1) + B_(j+1), where B_(j+1) bounds the average of the
 absolute normalized correlations.  Their supports are intersections of
 translates of the original interval, hence again intervals.  All shifts
@@ -2699,7 +2699,7 @@ are o(N), preserving the derivative bounds.  Iterating from the final
 exponent eps_0/4 proves a bound O(N^(-delta_k)) for the original
 normalized sum, with delta_k = eps_0/(k 2^(k-1)).  Taking
 delta = eps_0/(K 2^(K-1)) proves (8.51) in every case.  The differencing
-inequality itself follows by averaging H_0 translates and applying
+ineuality itself follows by averaging H_0 translates and applying
 Cauchy-Schwarz; its zero-shift term is the stated O(H_0^(-1)).
 
 Partial summation of 1/[d(q'y+a)] costs O(1/P') times the uniform
@@ -2775,7 +2775,7 @@ to retain the discrepancy while converting multiplicative factorizations
 into additive correlations.
 
 We split u = eh, with e slightly smaller than the scale of m, using a
-smooth allocation of the band primes.  Cauchy's inequality in (e, n)
+smooth allocation of the band primes.  Cauchy's ineuality in (e, n)
 then has a small enough outside factor to control its diagonal.  Off the
 diagonal the two factorizations give cross-products whose difference is
 a small nonzero integer.  These are the shifted endpoints.  All group
@@ -2905,7 +2905,7 @@ into factorizations u = eh with
    E L^(-C_s) <= e <= E,                                               (9.13)
 
 where C_s depends only on the band constants.  The reason for placing e
-below U is that Cauchy's inequality will contribute an outside factor
+below U is that Cauchy's ineuality will contribute an outside factor
 EV = x L^(-K_0).  We will use this logarithmic saving to control the
 diagonal, so the splitting costs must be independent of K_0.
 
@@ -2987,7 +2987,7 @@ Everything in this paragraph is independent of K_0.
 Separate Psi(eh/x) by Fourier inversion in log(eh/x) as well.  Its
 Fourier L^1 norm is fixed, and the tail outside frequency L has
 arbitrary power saving; its two factors are powers of e and h.  All
-these errors can be summed before Cauchy's inequality.  Indeed, the
+these errors can be summed before Cauchy's ineuality.  Indeed, the
 absolute sum over the original factorizations is bounded by
 
    L^O(1) sum over u = x of tau(u) tau(2u+1) << x L^O(1)                (9.20)
@@ -3210,7 +3210,7 @@ REMARK 9.2 (order of precisions).  The band constants, j_**, b_*, the
 coefficient bound C, and the target D_* are fixed first.  The transition
 functions, their total variation, and the coefficient and diagonal
 exponents are independent of K_0; only then is K_0 chosen.  The smooth
-separation after Cauchy's inequality and the accuracy required from
+separation after Cauchy's ineuality and the accuracy required from
 Theorem 8.1 are fixed next, and the required discrepancy exponent B is
 fixed last.  Accordingly, when alpha is the difference between a test
 and its rough-integer proxy, the precision of the cells defining that
@@ -3956,7 +3956,7 @@ primes is at least
    x L^(-C_A-1) exp(-O(sqrt L)) = x^(1-o(1)).                            (11.2)
 
 Their range is 2x < p <= 4x + 1 <= 5x.  Moreover, P^+(p-1) = P^+(2u) =
-max(2, P^+(u)) <= x^delta for large x.  The strict inequality c_2 s' <
+max(2, P^+(u)) <= x^delta for large x.  The strict ineuality c_2 s' <
 delta in (10.1) leaves room for every subpower P-prime, and the factor 2
 is eventually below x^delta.  This proves Theorem B.
 
@@ -3977,7 +3977,7 @@ yields a set Pp of primes with
 
    X/3 < p <= X,   p-1 is X^delta-smooth,   |Pp| >= X^(1-delta).      (12.1)
 
-The last inequality uses X^(1-o(1)) >= X^(1-delta) for large X.
+The last ineuality uses X^(1-o(1)) >= X^(1-delta) for large X.
 
 Many distinct products.  Put M = |Pp| and k = floor(X^(2 delta)).
 For large X, k <= M because 3 delta < 1.  The elementary bound
@@ -4044,7 +4044,7 @@ orientation:
    Baker, R. C. and Harman, G.  Shifted primes without large prime
    factors.  Acta Arith. 83 (1998), 331-361.
 
-   Montgomery, H. L. and Vaughan, R. C.  Hilbert's inequality.  J.
+   Montgomery, H. L. and Vaughan, R. C.  Hilbert's ineuality.  J.
    London Math. Soc. (2) 8 (1974), 73-82.
 
    Ford, K. and Halberstam, H.  The Brun-Hooley sieve.  J. Number