Real zeros of the Scorer function Gi
edit · history · discussion · files · long url · special functions zero
Numbers
$n$ 
$g_n$
1:
-0.7376403323201060346755275107651336209591008174033227485941609784857018146900092291880113378167299374
2:
-3.390831509448388299019904733863988520987801637018427573124781728535808443440077344631669094455639486
3:
-4.751600790643217841330497605986484761252535519200056582721524247558522417864025779066335226288427246
4:
-6.227070834562610414311286356278210460210287640065052224598626394165065851544295344058950449629716548
5:
-7.330170703258512449120389230804608887478109696622147239847546957904095644603685084119273255680849252
6:
-8.530647818615232317127904186507838278478821941173430080386111849263495575734016986970790246907544074
7:
-9.504435473068411469490236915270933444297419747185653608121410242329910376541418002735621698064946481
8:
-10.55956788510568338704659090510256622816101014198936473068834720014298138659693645176986910281030936
9:
-11.45018302721606039218172323311071509623366551394474819174980820221941925092074590552459902236651785
10:
-12.41065271985533325958333605937484214724193958120615716156012620958760825580771824033003565655563583
11:
-13.24130784630380179056940600835851472701711140456386055969313869618629622740519407209100767439034056
12:
-14.13335803157540377486364486431917656099870954505061342483981812867626243789939299681135147577235248
Definition
Let $\operatorname{Gi}(x)$ be the Scorer function $\operatorname{Gi}(x)=\pi^{-1}\int_0^\infty\sin(xt+t^3/3)\,dt$, with the integral interpreted as an oscillatory improper integral [1]. This table gives its signed real zeros $g_n$, indexed by $0>g_1>g_2>\cdots$ for positive integers $n$.
Parameters
$n$
—   zero index ($n\geq1$)
Formulas
(1)
$\operatorname{Gi}^{\prime\prime}(x)-x\operatorname{Gi}(x)=-1/\pi$ [1].
(2)
With $\Gamma$ denoting the Gamma function, $\operatorname{Gi}(0)=1/(3^{7/6}\Gamma(2/3))$ and $\operatorname{Gi}^{\prime}(0)=1/(3^{5/6}\Gamma(1/3))$ [3].
(3)
For the Airy function $\operatorname{Bi}$ and the companion Scorer function $\operatorname{Hi}$, $\operatorname{Bi}(g_n)=\operatorname{Hi}(g_n)$, since $\operatorname{Gi}(x)+\operatorname{Hi}(x)=\operatorname{Bi}(x)$ [4].
Comments
(4)
The real zeros of $\operatorname{Gi}$ are all negative and simple, and there are infinitely many of them [2].
(5)
The normalization is that of DLMF §9.12 [1]. Derivatives are with respect to $x$, and real powers of $3$ are positive.
Programs
(P1)
Python
import mpmath as mp

with mp.workdps(130):
    g1 = mp.findroot(mp.scorergi, (-1, 0))
    print(mp.nstr(g1, 100))
Links
Similar tables
Values of the Scorer function Gi —   values of the function whose real zeros are tabulated here
Zeros of the Airy function of the second kind Bi —   zeros of the homogeneous Airy solution related to Gi by (3)
Values of the Scorer function Hi —   values of the companion Scorer function in (3)
Data properties
Entries are of type: real number
Table is complete: no (it holds the first twelve negative real zeros, $1\leq n\leq12$, a consecutive reference selection beginning with the zero nearest the origin)
How they were obtained:

The values are written to 100 significant digits. The generator uses Arb balls at 700 bits and a 900-term Maclaurin polynomial fixed by (1) and (2). Its coefficients satisfy $c_2=-1/(2\pi)$ and $c_{k+3}=c_k/((k+3)(k+2))$. Each residue class modulo three has a separate geometric bound for its omitted tail. The bounds for the function and its first two derivatives on $|x|\leq16$ are respectively less than $7\cdot10^{-432}$, $4\cdot10^{-430}$ and $3\cdot10^{-428}$; they are added to the ball radii.

more

The generator certifies exactly twelve roots on $[-227/16,0]$. It covers this interval by 227 adjacent segments of length $1/16$. Local function and derivative enclosures follow from the differential equation and a Gronwall bound. On 206 segments the function excludes zero. On the remaining 21 segments the derivative has a fixed nonzero sign: nine have equal endpoint signs and twelve have opposite endpoint signs. Thus each of the twelve latter segments contains exactly one root and no root is omitted. Within each such segment, an mpmath candidate is enclosed by exact rational endpoints of separation $2\cdot10^{-120}$. Ball evaluations certify opposite function signs at these endpoints and a nonzero derivative throughout the bracket. Each bracket lies strictly inside the interval represented by its stored decimal.

Independent certified quadrature of the Airy representation [5], using $\int_0^\infty\operatorname{Ai}(t)\,dt=1/3$ [6], also establishes the endpoint signs for every root. The quadrature controls integrate $t^2$ and the derivative of the Airy function, with known nonzero answers. The attached checks compare every root with mpmath [7] at 150 and 200 decimal digits, test the initial values, differential equation and sum identity, and repeat the endpoint checks on the intervals read from the database. The mpmath comparisons are numerical checks separate from the rigorous certificates.