We present a decomposition of the real time evolution operator $e^{-i T H}$of any local Hamiltonian $H$ on lattices $Lambda subseteq mathbb Z^D$ intolocal unitaries based on Lieb-Robinson bounds. Combining this with recentquantum simulation algorithms for real time evolution, we find that theresulting quantum simulation algorithm has gate count $mathcal O( T n~mathrm{polylog} (T n/epsilon))$ and depth $mathcal O( T~mathrm{polylog}(Tn/epsilon))$, where $n$ is the space volume or the numberof qubits, $T$ is the time of evolution, and $epsilon$ is the accuracy of thesimulation in operator norm. In contrast to this, the previous best quantumalgorithms have gate count $mathcal O(Tn^{2} ~mathrm{polylog} (Tn/epsilon))$. Our approach readily generalizes to time-dependent Hamiltoniansas well, and yields an algorithm with similar gate count for any piecewiseslowly varying time-dependent bounded local Hamiltonian. Finally, we also provea matching lower bound on the gate count of such a simulation, showing that anyquantum algorithm that can simulate a piecewise time-independent bounded localHamiltonian in one dimension requires $Omega(Tn / mathrm{polylog}(Tn) )$gates in the worst case. In the appendix, we prove a Lieb-Robinson boundtailored to Hamiltonians with small commutators between local terms. Unlikeprevious Lieb-Robinson bounds, our version gives zero Lieb-Robinson velocity inthe limit of commuting Hamiltonians. This improves the performance of ouralgorithm when the Hamiltonian is close to commuting.
展开▼
机译:我们展示了实时演化运营商$ e ^ { - - i t h} $的分解,任何当地的汉密尔顿人$ h $ on paters $ lambda subseteq mathbb z ^ d $ INTOLOCOLERIES基于LIEB-Robinson边界。与recentquantum模拟算法实时进化结合这一点,我们发现,theresulting量子仿真算法有门数$ mathcal O(T N〜 mathrm {polylog}(T N / 小量))$和深度$ mathcal O( t〜 mathrm {polylog}(tn / epsilon))$,其中$ n $是空间卷或qubits的number,$ t $是演变的时间,$ epsilon $是运营商中的显影的准确性规范。与此相反,以前的最好的quantumalgorithms有门数$ mathcal O(Tn的^ {2}〜 mathrm {polylog}(TN / 小量))$。我们的方法很容易推广到时间依赖的HamiltoniansAs,并产生一种具有类似栅极计数的算法,对于任何分段不同的时间依赖于局部哈密顿人。最后,我们还普遍匹配这种模拟的栅极计数的下限,表明可以在一个维度中模拟分段时间独立的Localhamiltonian的任何算法需要$ omega(tn / mathrm {polylog}(tn)) $盖茨在最坏的情况下。在附录中,我们证明了一个位于当地条款之间的小型换向器的垃圾罗宾逊。与Previous Lieb-Robinson界,我们的版本给出了Zero BieB-robinson速度,距通勤Hamiltonians的限制。当Hamiltonian接近通勤时,这提高了Ourgorithm的性能。
展开▼