Research on Evolution of Stellar Systems with Highly Parallelized Special-purpose Computer for Gravitational Many-body Problems

1990年，我们在一块钢丝缠绕的木板上制作了GRPE-1A。它是万有引力多体问题专用计算机的原型。然后，我们与富士施乐株式会社电子成像和设备研究实验室密切合作，在一块定制的LSI上设计了一款处理器芯片，该芯片容纳了GRPE-1A的所有逻辑。这个大规模集成电路的尺寸是毫米，包含113,000个晶体管。利用该处理器芯片，我们于1991年构建了一台并行机GRAPE-3：我们在一块VME板上布线了24个处理器芯片，并通过适当设计的接口将两块板与主机工作站连接起来。该机器在1991年进行了调整，以达到14Gflopps的峰值性能和10Gflopps的持续速度，以解决20万个重力物体的问题。就持续速度而言，它比1991年任何商用的超级计算机都要快。我们利用GRAPE-3对恒星系统的演化进行了各种模拟，发现了…更重要的是，星系的碰撞和合并在星系及其星系团的演化中扮演着重要的角色。在下文中，我们总结了新的发现：1)当两个具有中心大质量黑洞的星系相互碰撞时，由于轨道中心附近的场星，黑洞受到动力摩擦。与其说它们失去能量，不如说它们失去了角动量。轨道的偏心率接近统一，黑洞可能会迅速合并。与此同时，场恒星获得能量并变热。这样的图像很好地解释了核心半径和残余物总质量之间的关系。2)在星系碰撞过程中，引力场发生剧烈变化。结果，恒星在相空间中变得完全混合。另一方面，在能量空间中，混合只在它们碰撞的初始阶段进行得很快，而在之后进行得更慢。3)经过这样的过程后，每个星系碰撞前核心(或晕)中的恒星倾向于停留在合并的核心(或晕)中。因此，在巨型椭圆星系中观察到的色梯度不能证明其起源的合并假说。4)星系合并是通过与晕中的恒星一起驱逐角动量而有效进行的。因此，星系合并可以被认为是频繁的现象，并可能在星系和宇宙的演化和历史中发挥主导作用。较少

In 1990 we constructed GRAPE-1A on a wire-wrapped board. It is a prototype of a special-purpose computer for gravitational many-body problems. In close collaboration with Electronic Imaging and Device Research Laboratory of Fuji XEROX Co., Ltd. we then designed a processor chip on one customized LSI which accommodated all the logics of GRAPE-1A. This LSI is 8x8 mm in size and contains 113, 000 transistors. Using this processor chip we constructed a parallelized machine GRAPE-3 in 1991 : We wired 24 processor chips on a VME board, and connected two boards with the host workstation via appropriately designed interface. The machine was tuned in 1991 to reach the peak performance of 14 Gflops, and the sustained speed of 10 Gflops in solving a gravitational 200, 000body problem. It is faster than the any commercially available supercomputer in 1991 so far as the sustained speed is concerned. Using GRAPE-3 we have done a variety of simulations on evolution of stellar systems, and have found … More that collision and resultant merging of galaxies play important roles in evolution of galaxies and their clusters. In what follows we summarize the new findings.1) When two galaxies with central massive black holes collide each other, the black holes suffer from dynamical friction due to the field stars near the apocenter of the orbit. They loose rather their angular momentum than their energy. The eccentricity of the orbit approaches to unity and the black holes may merge quickly. At the same time the field stars aquire energy and become hot. Such picture nicely interprets the relation between the core radius and the total mass of the remnants.2) During the collision of galaxies the gravitational field changes violently. As a result the stars become throughly mixed in the phase space. On the other hand, in the energy space the mixing proceeds quickly only in the initial phases of their collision and much more slowly afterward. Therefore, it is not appropriate to call such process as the "violent" relaxation.3) After such process, the stars in the core (or halo) of each galaxy before the collision tend to stay in the core (or halo) of the merger. Therefore, the color gradient observed in giant elliptical galaxies can not testify against the merger hypothesis of their origin.4) Merging of galaxies proceeds effectively by expelling the angular momentum together with the stars in the halo. Thus, the merging of galaxies may be regarded to be frequent phenomena and may play one of the leading roles in the evolution and history of galaxies and the universe. Less