Newform orbit 10.6.a.c
-expansion
Coefficient field
The coefficients generate . The eigenvalue of is a root of
AtkinβLehner signs
| 2 | |
| 5 |
For every form in the orbit has . The Fricke sign is , so the sign of the functional equation is : each L-function in the orbit has even analytic rank.
Traces of Hecke eigenvalues
| 2 | 3 | 5 | 7 | 11 | 13 | 17 | 19 | 23 | 29 | 31 | 37 | 41 | 43 | 47 | 53 | 59 | 61 | 67 | 71 | 73 | 79 | 83 | 89 | 97 | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 4 | 6 | β25 | β118 | 192 | 1106 | 762 | β2740 | 1566 | 5910 | β6868 | β5518 | β378 | β2434 | 13122 | β9174 | β34980 | β9838 | 33722 | 70212 | 21986 | 4520 | β109074 | 38490 | β1918 |
for every is in the JSON.
SatoβTate
The normalized for the primes not dividing 10, against the semicircle :
Provenance
Stored in the atlas: computed by Sagebrush's modular symbols engine in 0.105 s (one thread) and proven: the dimensions are certified, and the characteristic polynomial of over is reconstructed by CRT past a rigorous coefficient bound.
The engine's checks
dim M^+ at levels [1, 2, 5, 10]: [1, 2, 3, 9] (certified at ell = 2147483629)T = [(3, 1), (7, 1)]; h of degree 3 = 1 x 3 (phi(ord eps) x dim S_k^new); 1 primes, 0 rejected
Recompute it yourself
In Sage mode at sagebrush.space (or npx sagebrush --sage):
newform_orbits(10, 6)[2] # 10.6.a.c
With the Rust engine (source), in engine/:
cargo run --release -p sagebrush-web --example atlas -- 10:6 1000