Specify the positive spectrum

For integer d≥2d\geq2, consider one eigenvalue per positive integer index, with no additional multiplicity:

Pd(n)=S(n,d)=(n+d−1d)=n(n+1)⋯(n+d−1)d!,Zd(s)=∑n=1∞Pd(n)−s. P_d(n)=S(n,d)=\binom{n+d-1}{d} =\frac{n(n+1)\cdots(n+d-1)}{d!}, \qquad Z_d(s)=\sum_{n=1}^{\infty}P_d(n)^{-s}.

Comparison with n−dsn^{-ds} proves absolute convergence for Re⁡s>1/d\operatorname{Re}s>1/d. All eigenvalues in this definition are dimensionless. The integer dd labels the polynomial sequence; interpreting an actual spatial operator requires its eigenvalues and multiplicities. The associated positive heat sum is developed in Simplex heat trace.

Continuation by subtracting the large-index terms

Set

Ad=d(d−1)2,bn=∏j=0d−1(1+jn). A_d=\frac{d(d-1)}2, \qquad b_n=\prod_{j=0}^{d-1}\left(1+\frac jn\right).

Uniformly for ss in a compact set, bn−s=1−sAd/n+O(n−2)b_n^{-s}=1-sA_d/n+O(n^{-2}). Therefore

Rd(s)=∑n=1∞n−ds[bn−s−1+sAdn] R_d(s)=\sum_{n=1}^{\infty} n^{-ds} \left[b_n^{-s}-1+\frac{sA_d}{n}\right]

converges locally uniformly and defines a holomorphic function for Re⁡s>−1/d\operatorname{Re}s>-1/d. Initially in the convergent half-plane, then by continuation,

Zd(s)=(d!)s[ζ(ds)−sAdζ(ds+1)+Rd(s)]. \boxed{Z_d(s)=(d!)^s \left[\zeta(ds)-sA_d\zeta(ds+1)+R_d(s)\right].}

Here ζ\zeta is the Riemann zeta function. The remainder bound follows by expanding log⁡bn=Ad/n+O(n−2)\log b_n=A_d/n+O(n^{-2}) and its exponential; the tail is O(n−dRe⁡s−2)O(n^{-d\operatorname{Re}s-2}). This also justifies local termwise differentiation.

In this half-plane, the pole at s=1/ds=1/d has residue (d!)1/d/d(d!)^{1/d}/d. At s=0s=0, the apparent pole in the second term cancels against its factor ss.

Value and derivative at zero

Use ζ(0)=−1/2\zeta(0)=-1/2 and ζ(1+ds)=1/(ds)+γ+O(s)\zeta(1+ds)=1/(ds)+\gamma+O(s), where γ\gamma is Euler's constant. Since Rd(0)=0R_d(0)=0,

Zd(0)=−12−Add=−d2. \boxed{Z_d(0)=-\frac12-\frac{A_d}{d}=-\frac d2.}

The derivative of the remainder is an ordinarily convergent series:

Rd′(0)=∑n≥1[−log⁡bn+Adn]=∑j=0d−1log⁡Γ(1+j)+γAd. R_d'(0)=\sum_{n\geq1} \left[-\log b_n+\frac{A_d}{n}\right] =\sum_{j=0}^{d-1}\log\Gamma(1+j)+\gamma A_d.

To obtain the last equality, take the logarithm of the gamma-function product:

log⁡Γ(1+j)=−γj+∑n≥1[jn−log⁡(1+jn)]. \log\Gamma(1+j)=-\gamma j+ \sum_{n\geq1}\left[\frac jn-\log\left(1+\frac jn\right)\right].

Only finitely many jj are summed. Differentiating the continuation formula then cancels the γAd\gamma A_d terms and gives

Zd′(0)=−d2log⁡(d!)+dζ′(0)+∑j=0d−1log⁡(j!)=−d2log⁡(2πd!)+∑j=0d−1log⁡(j!). \begin{aligned} Z_d'(0) &=-\frac d2\log(d!)+d\zeta'(0) +\sum_{j=0}^{d-1}\log(j!)\\ &=\boxed{-\frac d2\log(2\pi d!) +\sum_{j=0}^{d-1}\log(j!)}. \end{aligned}

The Riemann values used here, including ζ′(0)=−12log⁡(2π)\zeta'(0)=-\tfrac12\log(2\pi), are recorded in DLMF §25.6.

Regularized product and scale

Define the zeta determinant by det⁡ζPd=exp⁡[−Zd′(0)]\det_\zeta P_d=\exp[-Z_d'(0)]. The result is

det⁡ζPd=(2πd!)d/2∏j=0d−1j!. \boxed{\det\nolimits_\zeta P_d =\frac{(2\pi d!)^{d/2}}{\prod_{j=0}^{d-1}j!}.}

This regularized product has the following simple benchmarks:

Spectrum, one eigenvalue per n≥1n\geq1 Value at zero Zeta determinant
nn −12-\tfrac12 2π\sqrt{2\pi}
n(n+1)/2n(n+1)/2 −1-1 4π4\pi
n(n+1)(n+2)/6n(n+1)(n+2)/6 −32-\tfrac32 (12π)3/2/2(12\pi)^{3/2}/2

The first row follows directly from Z1=ζZ_1=\zeta. If the physical eigenvalue is μPd(n)\mu P_d(n), choose a fixed comparison unit E∗E_* and set r=μ/E∗>0r=\mu/E_*>0. Then ZrPd(s)=r−sZd(s)Z_{rP_d}(s)=r^{-s}Z_d(s) and

log⁡det⁡ζ(rPd)=log⁡det⁡ζPd−d2log⁡r. \log\det\nolimits_\zeta(rP_d) =\log\det\nolimits_\zeta P_d-\frac d2\log r.

Thus the normalization convention remains explicit. Zd(0)Z_d(0) is an analytically continued spectral quantity, not a count of spatial directions. A gravitational or vacuum-energy use additionally needs a specified physical operator, action, regularization prescription, and scale matching.

spectral_analysis_checks.py checks convergent sums, remainder bounds, and independent centered-polynomial continuations for d=2,3d=2,3, including numerical derivatives at zero and scale changes. The general formulas follow from the analytic derivation above.