Institut Pluridisciplinaire Hubert Curien, Physique Théorique, Université de Strasbourg, F-67037 Strasbourg, France.
Department of Physics and Astronomy, McMaster University, Hamilton, Canada L8S4M1.
Phys Rev E. 2017 May;95(5-1):052108. doi: 10.1103/PhysRevE.95.052108. Epub 2017 May 5.
In this paper, we discuss P(n), the number of ways a given integer n may be written as a sum of primes. In particular, an asymptotic form P_{as}(n) valid for n→∞ is obtained analytically using standard techniques of quantum statistical mechanics. First, the bosonic partition function of primes, or the generating function of unrestricted prime partitions in number theory, is constructed. Next, the density of states is obtained using the saddle-point method for Laplace inversion of the partition function in the limit of large n. This gives directly the asymptotic number of prime partitions P_{as}(n). The leading term in the asymptotic expression grows exponentially as sqrt[n/ln(n)] and agrees with previous estimates. We calculate the next-to-leading-order term in the exponent, proportional to ln[ln(n)]/ln(n), and we show that an earlier result in the literature for its coefficient is incorrect. Furthermore, we also calculate the next higher-order correction, proportional to 1/ln(n) and given in Eq. (43), which so far has not been available in the literature. Finally, we compare our analytical results with the exact numerical values of P(n) up to n∼8×10^{6}. For the highest values, the remaining error between the exact P(n) and our P_{as}(n) is only about half of that obtained with the leading-order approximation. But we also show that, unlike for other types of partitions, the asymptotic limit for the prime partitions is still quite far from being reached even for n∼10^{7}.
在本文中,我们讨论了 P(n),即给定整数 n 可以表示为质数和的方式的数量。特别是,使用量子统计力学的标准技术从解析上获得了 n→∞时有效的渐近形式 P_{as}(n)。首先,构造了质数的玻色配分函数,或者数论中无限制的质数分区的生成函数。接下来,使用鞍点方法获得了在大 n 极限下配分函数的拉普拉斯反演的态密度。这直接给出了渐近质数分区数 P_{as}(n)。渐近表达式中的主导项按 sqrt[n/ln(n)] 指数增长,与以前的估计一致。我们计算了指数中的次主导项,与 ln[ln(n)]/ln(n) 成正比,并且我们表明文献中的其系数的早期结果是不正确的。此外,我们还计算了下一个更高阶的修正项,与 1/ln(n) 成正比,并在 Eq. (43) 中给出,这迄今为止在文献中尚未提供。最后,我们将我们的分析结果与 P(n) 的精确数值进行了比较,直到 n∼8×10^{6}。对于最高的值,精确的 P(n)与我们的 P_{as}(n) 之间的剩余误差仅约为使用主导阶近似值获得的误差的一半。但我们也表明,与其他类型的分区不同,即使对于 n∼10^{7},质数分区的渐近极限仍然相当远。