界面已本地化;技能介绍与说明文档按所选语言由 AI 自动翻译,代码、命令、路径和专业术语保留原文,以原作者文档为准。
探索技能
设计与创意 · SKILL.md

skill-147

skill-147

【设计与创意】A skill for visualizing quantum states and dynamics using QuTiP's advanced plotting capabilities. Perfect for researchers who want to create intuitive visual representations of quantum systems and their behavior over time.

独立 GitHub 快照 · 自动分类 · 以原作者文档为准

技能说明

GitHub

name: skill-147 description: "A skill for visualizing quantum states and dynamics using QuTiP's advanced plotting capabilities. Perfect for researchers who want to create intuitive visual representations of quantum systems and their behavior over time."

QuTiP Visualization Tools

Overview

QuTiP provides powerful visualization tools to help researchers and educators illustrate quantum phenomena. This skill focuses on generating graphs and plots for quantum states, operators, and dynamics.

Installation

uv pip install qutip

Quick Start

from qutip import *
import numpy as np

# Create a quantum state
psi = basis(2, 0)  # Ground state |0⟩

# Visualize the state on the Bloch sphere
plot_bloch(psi)  # Interactive Bloch sphere

Core Capabilities

1. Quantum State Visualization

Visualize quantum states on the Bloch sphere:

# Generate a superposition state
psi = (basis(2, 0) + basis(2, 1)).unit()  # |0⟩ + |1⟩

# Plot the state on the Bloch sphere
plot_bloch(psi)

2. Dynamics of Quantum Systems

Plot time evolution of observables:

H = sigmax()  # Hamiltonian
psi0 = basis(2, 0)  # Initial state
tlist = np.linspace(0, 10, 100)
result = sesolve(H, psi0, tlist, e_ops=[sigmaz()])

# Plot the expectation value over time
plt.plot(tlist, result.expect[0])
plt.xlabel('Time')
plt.ylabel('⟨σz⟩')
plt.title('Time Evolution of ⟨σz⟩')
plt.show()

3. Advanced Visualization Techniques

Use advanced techniques to visualize correlation functions and more:

# Visualize Wigner function for a coherent state
alpha = 1.0
rho_coherent = coherent_dm(10, alpha)
plot_wigner(rho_coherent)
plt.title('Wigner Function of Coherent State')
plt.show()

4. Custom Plots

Create custom plots for specific needs:

# Custom plot function
def custom_plot(result, title):
    plt.figure()
    plt.plot(result.times, result.expect[0])
    plt.title(title)
    plt.xlabel('Time')
    plt.ylabel('Expectation Value')
    plt.grid()
    plt.show()

custom_plot(result, 'Custom Expectation Value Plot')

Conclusion

The visualization tools offered by QuTiP enhance the understanding of quantum mechanics by providing intuitive graphical representations. This skill empowers researchers to communicate their findings effectively.

我的收藏