5D Library › 五节点循环网络动力学、黄经历表与多尺度振子扰动 · English
第七章 「十五日为一节」:《淮南子·天文训》的节点规则与平气、定气之争
为什么在这里读《淮南子》
二十四节气最早的完整记载之一见于西汉的《淮南子·天文训》。它给出的是一套按日计数的规则:从冬至起,每十五日为一节。本书第一章与第六章已经算出了真实的节气时刻;本章把这套古老的计数规则与真实时刻对照,看它的误差有多大、误差从哪里来,并把这一对照放进中国历法史上平气与定气的长期并存中去读。
原文
注
版本与核对。 两段字句均据何宁《淮南子集释》的通行文字引录,已据通行点校本核对一致。第二段中「報德之維」指东北方的「维」(斗柄所指的四隅之一)。
十五日一节:计数规则与真实节气
第一段说太阳每日行一度,十五日为一节;第二段说从冬至到立春是四十六日——比三个十五日多出一日,多出的一日放在斗柄转过东北「维」之处。第十五段程序比较三种计数规则与真实节气:R1 每节十五日;R2 每节十五日,并在四个「维」(立春、立夏、立秋、立冬)各加一日——原文只给出了立春这一处,其余三处是本书的对称假定;R3 把回归年平分为二十四份。
推导
# Chapter 7. p15: Fifteen-day nodes in the Huainanzi, against the true terms.
# "Tianwen" counts the 24 terms in steps of fifteen days from the winter
# solstice and puts Spring begins 46 days after it (one extra day at the
# north-east corner). Three day-count rules are compared with the true
# terms (low-precision theory of p03), as days after the solstice, for
# two skies: 2026-27, and 104-103 BCE, near the Han reform. The Sun is
# fastest at perigee, whose longitude moves about 1.7 deg a century, so
# the pattern of errors depends on the epoch. (The solar theory is
# extrapolated 21 centuries back; its error there is not checked here.)
# Rules:
# R1 15 k days
# R2 15 k days plus one day at each of the four corners (Spring,
# Summer, Autumn, Winter begins). Only the Spring-begins day is in
# the text quoted in Chapter 7; the other three corners are this
# program's symmetric assumption.
# R3 mean terms, 15.2184 k days (tropical year / 24)
import math
def sun_lon(jd):
t = (jd - 2451545.0) / 36525
l0 = 280.46646 + 36000.76983 * t + 0.0003032 * t * t
m = math.radians(357.52911 + 35999.05029 * t - 0.0001537 * t * t)
c = ((1.914602 - 0.004817 * t - 0.000014 * t * t) * math.sin(m)
+ (0.019993 - 0.000101 * t) * math.sin(2 * m)
+ 0.000289 * math.sin(3 * m))
om = math.radians(125.04 - 1934.136 * t)
return (l0 + c - 0.00569 - 0.00478 * math.sin(om)) % 360
def when(target, jd0):
lo = jd0
while not ((sun_lon(lo) - target) % 360 > 180
and (sun_lon(lo + 1) - target) % 360 < 180):
lo += 1
hi = lo + 1
for _ in range(60):
mid = (lo + hi) / 2
if (sun_lon(mid) - target) % 360 < 180:
hi = mid
else:
lo = mid
return (lo + hi) / 2
def table(jd0):
ws = when(270.0, jd0)
tt = [ws]
for k in range(1, 25):
tt.append(when((270 + 15 * k) % 360, tt[-1] + 10))
return [t - ws for t in tt]
if __name__ == "__main__":
rules = {"R1": [15 * k for k in range(25)],
"R2": [15 * k + sum(k >= c for c in (3, 9, 15, 21))
for k in range(25)],
"R3": [365.2422 / 24 * k for k in range(25)]}
for label, jd0 in (("2026-27", 2461380.5), ("104-103 BCE", 1683595.5)):
true = table(jd0)
print("{}: days after the winter solstice".format(label))
print(" term (k) true R1 R2 R3")
for k, name in ((3, "Spring begins"), (6, "Spring equinox"),
(12, "Summer solstice"), (18, "Autumn equinox")):
print(" {:<15} {:2d} {:6.2f} {:3d} {:3d} {:6.2f}".format(
name, k, true[k], rules["R1"][k], rules["R2"][k],
rules["R3"][k]))
print(" rule largest |error|, 24 terms left over at year end")
for key in ("R1", "R2", "R3"):
errs = [abs(r - t) for r, t in zip(rules[key][:24], true[:24])]
left = round(true[24] - rules[key][24], 2) + 0.0
print(" {} {:5.2f} days {:+6.2f}".format(
key, max(errs), left))
print()
print("Reading: plain fifteen-day steps fall five and a quarter days")
print("short by the year's end; a day at each corner leaves only the")
print("quarter-plus-one to place. But in neither sky does the true")
print("Spring begins fall at 46 days - it comes before day 45 in both.")
print("The corner day balances the count over the year; it is not a")
print("correction for the Sun's uneven speed, which no equal-step rule")
print("captures: even equal division of the year misses the true")
print("terms by more than two days.")
# 2026-27: days after the winter solstice
# term (k) true R1 R2 R3
# Spring begins 3 44.20 45 46 45.66
# Spring equinox 6 88.98 90 91 91.31
# Summer solstice 12 181.72 180 182 182.62
# Autumn equinox 18 275.38 270 273 273.93
# rule largest |error|, 24 terms left over at year end
# R1 5.78 days +5.24
# R2 2.78 days +1.24
# R3 2.39 days +0.00
#
# 104-103 BCE: days after the winter solstice
# term (k) true R1 R2 R3
# Spring begins 3 44.56 45 46 45.66
# Spring equinox 6 90.21 90 91 91.31
# Summer solstice 12 184.22 180 182 182.62
# Autumn equinox 18 276.59 270 273 273.93
# rule largest |error|, 24 terms left over at year end
# R1 6.59 days +5.24
# R2 3.59 days +1.24
# R3 2.81 days +0.00
#
# Reading: plain fifteen-day steps fall five and a quarter days
# short by the year's end; a day at each corner leaves only the
# quarter-plus-one to place. But in neither sky does the true
# Spring begins fall at 46 days - it comes before day 45 in both.
# The corner day balances the count over the year; it is not a
# correction for the Sun's uneven speed, which no equal-step rule
# captures: even equal division of the year misses the true
# terms by more than two days.按 2026 至 2027 年的天象,单纯的十五日一节到年终少了 5.24 日,最大误差 5.78 日;四维各加一日后只剩 1.24 日需要安放,最大误差降到 2.78 日(汉代天象下为 3.59 日);把一年平分(平气)也还差 2.39 日。
问题在于第二段的四十六日。按 2026 年的天象,真实的立春在冬至后 44.20 日;按公元前 104 年的天象,在 44.56 日。两个时代的真实立春都早于第 45 日,第 46 日并没有更接近真实的立春。程序的结论是:多出的那一日是为了让全年的计数接近一年,而不是为了修正太阳运动的不均匀;任何等步长的规则都捕捉不到后者,平分一年也会差两天以上。
注
两个时代的天空不同。 近地点的黄经每世纪移动约 1.7 度,所以太阳何时快、何时慢,汉代与今天不同;本章因此同时给出两个时代的结果。程序所用的低精度太阳理论被外推了二十一个世纪,它在那里的误差本书没有核对。结论(第 46 日不是对真实立春的修正)在两个时代都成立,但汉代的具体数字只能作为近似。
平气与定气:计算与民用的长期并存
第六章说明,平气与定气最多相差两天多。中国历法史上,两者并存了一千多年:隋代刘焯在《皇极历》中提出按太阳实际运动划分节气的「定气」,唐代一行的《大衍历》在推算太阳位置时以定气为节点(第六章的不等间距插值);但民用历书中注记节气,长期仍用平气,直到清初的时宪历才正式改用定气 (Needham and Wang 1959; 陈美东 2003)。
把这段历史与第一章并读,可以看到同一个结构:一边是天文上的真实时刻,一边是社会上的计时约定。定气对应「太阳真正到了哪里」,平气对应「日子怎样均匀地数」。两者的差别——第六章 p12 算出的最多 2.386 天——不是错误,而是两种目的之间的距离。
假说 · 尚未验证
作者关于文本的假说(尚未验证): 《淮南子·天文训》的十五日一节加四维各一日,是一套以斗柄方位为参照的计数方案,其设计目标是全年日数闭合,而不是逐节逼近太阳的实际位置;因此,若原文对立夏、立秋、立冬也有类似的「四十六日」表述,它们与真实节气的偏差应与立春同样不具系统性的改进。什么结果会证明它错:原文对其余三维的日数安排显示出与太阳运动不均匀相符的系统差异(例如冬半年与夏半年不同)。
注
类比的界限。 本章比较的是计数规则与天象的数值差异。它不评价《淮南子》的宇宙论,不把斗柄指向、律吕与节气的对应当作物理关系,也不主张古人有现代意义上的「误差分析」。
小结
《淮南子·天文训》以十五日为一节、立春在冬至后四十六日。与真实节气相比,单纯十五日一节到年终少 5.24 日,四维各加一日后最大误差降到 2.78 日(汉代天象下 3.59 日),但第 46 日在今天与汉代的天空中都不比第 45 日更接近真实的立春:多出的一日服务于全年计数的闭合,而非太阳运动的不均匀。平气与定气在中国历法中并存千余年,直到时宪历才统一为定气;这与第一章钟表时间和太阳时间的分歧是同一种结构。
参考文献
Fan, Y. (2026). Five-Node Cyclic Network Dynamics, Solar-Longitude Ephemeris and Multi-Scale Oscillator Perturbations: Rock-Paper-Scissors-Lizard-Spock Lotka-Volterra Graphs, Equation-of-Time Keplerian Mechanics and Jacobian Sensitivity Bounds [五节点循环网络动力学、黄经历表与多尺度振子扰动] (Version 1.0.0). 5D Library. https://doi.org/10.5281/zenodo.24090168 · Licensed under Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0).