Simulations for how a quantum computer functions.
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
haskell Migration Dec 13, 2016
python Finished. Dec 14, 2016
.gitignore Updates Dec 8, 2016
LICENSE Initial commit Dec 1, 2016
README.md Finished. Dec 14, 2016

README.md

Quantum Computer Simulation

Note: In order to use the code in this repo, you will need to have Haskell-Stack installed. You can install it using these instructions https://docs.haskellstack.org/en/stable/install_and_upgrade/. You need Python3 installed.

This repository holds files necessary to run simulation of quantum computation using 5 fundamental quantum logic gates (X, Y, Z, Hadamard and CNOT). There are almost two implementations of basic quantum computation, one in Python and one in Haskell. Currently the Haskell version is having difficulties and is not functional.

One of the main reasons for implementing quantum computation simulation in Python/Haskell is to more deeply understand how a quantum computer operates. By implementing a simulation for quantum computation, one can better understand where a quantum computer can be better than a classical computer.

How a Quantum Computer Works

A quantum computer is similar to a classical computer in many ways. For instance, in classical computing you have the bit (a 1 or a 0) which is used to represent all data. Quantum computers use a similar concept, known as quantum bit, or qubit for short. Classical computers are built using a series of logic gates, and that is still true for quantum computers, although the types of logic gates are slightly different.

Lastly, both classical computers and quantum computers solve computational problems. A quantum computer can simulate a classical computer, and vis versa (which we do here). There are many applications for quantum computing, including solving problems that classical computer would take many thousands of years to do. However, there are many use-cases that a classical computer is more suited for than a quantum computer such as databases, where saving and reading memory is a common operation.


Implementation Details

The following sections go through the different aspects of basic quantum computation, as well as how they are represented in the simulation code in this repository.

The Quantum bit

Instead of being a 1 or a 0, a qubit is the linear combination of the |0> and |1> basis vectors (|0> is Ket notation for vectors, which is the usual notation when talking about quantum computers). We will not concern ourselves with the physical implementation of a qubit (there are many different methods).

For instance, a typical qubit can be represented as a|0> + b|1> where |a^2| + |b^2| = 1. There is also the notion of global phase, which adds in another layer of complexity to the qubit, but for this short intro to basic quantum computation we will not go there. However, since the Y, Z and Hadamard gates change the global phase, it is worth mentioning.

Qubit in the Code

Representing the qubit in a quantum computation simulator is one of the trickiest aspects of designing a simulator. There are two ways that I considered modeling the qubit. The first was to represent each qubit as a vector of length 2, with values as the coefficients of the basis vectors |0> and |1>. This means it is trivially easy to operate on a single qubit, because you can just apply a matrix multiplication on it.

For example, in Python, it can be represented as a 1D matrix with numpy:

import numpy as np
class Qubit:
    __init__(self):
        qubit = np.matrix('1 0')

In Haskell, it can be represented as a vector using HMatrix:

import Numerical.LinearAlgebra
type Qubit = Vector (Complex Double)

createQubit :: Qubit
createQubit = fromList [ 1.0 :+ 0, 0 :+ 0]

This method of representation allows for very easy single-qubit operations. Since every quantum gate can be represented as a unitary matrix, you basically have the quantum gate operations as mere matrix multiplications. Easy.

The downside is that two-qubit operations are now impossible, or at least non-trivial to implement. Since the CNOT gate causes for qubits to become entangled, you must represent this entanglement in some way. For example, in order to create the Bell State (a two-qubit system of |00> + |11> all over sqrt(2) to keep magnitudes correct) one needs to allow the qubits to become entangled. This turns out to be non-trivial if you implement the qubits as discrete objects because there is no good way to show entanglement.

Thus, as many online quantum simulators do it, the other method is to simply keep track of all states the qubits can be in. For a system of n qubits, this means keeping track of 2^n qubits. This is the way the qubits are represented in the simulators in this repository.

For example, in Python, they are represented as an array:

class Register:
    def __init__(self, size):
        self.size = size
        self.qubits = [ 0+0j for x in range(2**size) ] 
        self.qubits[0] = 1+0j

Note that they are now called a "Register."

In Haskell, I couldn't quite figure out a clean way of doing this, but I gave an attempt.

data Register = Register { nStates :: Int, qubits :: [Qubit] }

newRegister :: Int -> Register
newRegister n = Register { nStates = n, qubits = [blankQubit (x-1) | x <- [1..2^n] ] }
                where blankQubit 0 = Qubit { probability = 1.0, qubit = 0 }
                      blankQubit y = Qubit { probability = 0.0, qubit = y }

This new way of representing the qubit states comes with its own drawbacks. It is now trivial to represent the entanglement of two qubits, but it becomes more difficult to apply single qubit operations. One can no longer just do simple matrix multiplication and know that everything will work.


Quantum Logic Gate

A quantum logic gate is similar to a classical logic gate. It is the method for changing the states of qubits. There are a few subtle differences that end up having some dramatic consequences.

One of the first differences is that a quantum gate is always reversible. Thus, a 2-input quantum gate has 2-outputs. Since every quantum gate can be represented as a matrix, every quantum gate is a special type of matrix called a unitary matrix.

The second difference is that a quantum gate always preserves the magnitude of the qubit(s) it operates on. Technically this is also true for a classical gate, as they can only switch a bit from 0 to 1, which if we consider them orthogonal vectors would always have a magnitude of 1. However, a quantum gate can change a qubit into a superposition of the |0> and |1> states. One of the main rules (in order to follow the laws of quantum physics) is that any superposition of states must always have a magnitude of 1. Thus, a|0> + b|1> would have a^2 + b^2 = 1.

There are quite a few quantum gates. In theory, there are actually an infinite amount of them. For the purposes of our simulators, we will only discuss 5 of them.

The first four gates are single-qubit gates, which means they operate on only 1 qubit.

The X Gate (or Sauli-X gate, or Pauli-X gate) is the basic "not" gate for a qubit. It is represented by the matrix [ 0 1 ; 1 0 ] and will swap |0> for a |1>, and |1> for a |0>.

The Y Gate (Sauli-Y/Pauli-Y gate) is a slight variation on the X gate. It is represented by the matrix [ 0 -i ; i 0 ]. This gate is essentially the X gate with a change of phase.

The Z Gate (Sauli-Z/Pauli-Z gate) is simply a phase-shift gate and is represented by the matrix [ 1 0 ; 0 -1 ]. It takes the |1> state and phase-shifts it by pi radians. Since we did not go into what the phase has to do with quantum computation, we do not need to go into how this gate is useful.

The Hadamard gate is a fun gate. It is represented by the matrix 1/sqrt(2) * [ 1 1 ; 1 -1 ]. What this gate does is transform |0> into (|0> + |1>) / sqrt(2) and |1> into (|0> - |1>) / sqrt(2). This gate basically takes a qubit that is not doing quantum-fun and makes it into the superposition of two states.

The last gate we will discuss and implement is the CNOT gate, or controlled-Not gate. This is the only 2-qubit gate, which takes in two qubits and outputs two qubits. It is quite easily described, as it applies the X-Gate on the second qubit IF the first qubit is a |1>.


Phew that was a lot. Hopefully you came out of that with some understanding of the 5 main quantum gates. The interesting stuff comes when you start linking the gates together to do nifty things.

Quantum Logic Gates in the Code

Depending on how the qubits are represented in code, the quantum gates can be very trivial to implement or a complete pain. By representing the qubit states the way I did in Python, there are a bit of a pain to implement. Since I couldn't quite figure out how to implement the gates in Haskell, I will not go over that here.

Also, since many of the gates share very similar traits, I will only go over the X Gate. The rest of the gates can be found in python/quantum.py.

Here is the full implementation of the X gate:

def X(self, bit):
    newReg = [0 for x in range(2**self.size)]
    for i, val in enumerate(self.qubits):
        newIdx = i ^ (1 << bit)
        newReg[newIdx] = val

    self.qubits = newReg

Note that this is implemented as a method of the Register class we defined above

The basic mode of operation is to set up a new "Register" or list of all qubit states (I try to keep things functional where possible) and then populate it with the swapped around coefficients.

For instance, let us consider a Register of length 2 (thus 2 qubits, and 4 unique states). We represent that in our code as an array of length four. The values stored in each index of the array are the coefficients for that index's quantum state.

The |00> state, where both qubits are of state |0> is thus in the 0th index of the array. At the start of the program, this would have a value of 1, since both qubits start out as |0>. Thus our Register (or array) looks like [ 1, 0, 0, 0 ].

Applying the X Gate to the first qubit in this register would flip the first qubit to a |1>. Thus, the |01> state would have a value of 1 and the |00> state would have a value of 0. Our new Register looks like [ 0, 1, 0, 0 ].

For ease of reading, the program will write the binary representation of the combined states as simply the decimal equivalent. Thus, the |10> state is just |2>.

Linking It All Together

Much like a classical computer, in order for a quantum computer to do useful computations, multiple gates have to be strung together and connected. Thankfully, this is as easy as with a classical computer. You simply link the output of one gate into the input of another. There are nifty things you can do with a quantum computer that you can't do with a classical computer. This makes sense, or else why would people even care about quantum computers?

One of the most fundamental things is called the Bell State. This is essentially a 2-qubit system which has the qubits entangled with each other. One example of a bell state is the |00> + |11> state for two qubits. (Note the 1/sqrt(2) prefactor is dropped as is usual in most writings).

The easiest way to construct a Bell State is to take two qubits with values of |0>. Take the first qubit and put it through a Hadamard gate, and then apply the cnot gate to the two qubits, with the first qubit controlling the gate.

Linking It Together In The Code

The easiest way to construct a Quantum Computer using the simulator in Python is to go into the python/main.py file. On a fresh clone of the repository there is code for constructing the bell state using a 2 qubit system.

import quantum as Q

base = Q.Register(2)
base.hadamard(1)
base.cnot(1, 0)
print(base)

You can run the program using python main.py and it will print out the bell state. The results will look something like:

$ python main.py
['(0.7071067811865475+0j) |0>', '(0.7071067811865475+0j) |3>']

To interpret the results, notice that it is an array of qubit states with coefficients in front of them. For this case, there is a 1/sqrt(2) coefficient in front of the |00> state and the |3> state (or the |11> state). This is the most basic Bell State we discussed above.

You can now go forth and try to implement algorithms using this simulator, or submit pull-requests to implement additional quantum gates.

FAQ

  • Why do this in Haskell Python?

Initially I wanted to this entirely in Haskell. The main reason is because I thought it would be easy to implement and verify in Haskell. However, I had a ton of trouble figuring out a clean way to represent the qubit states and apply gates to them. You can check out the Haskell Simulator in the Resources section, because someone else managed to get it to work.

I decided to just use Python for speed of development. Python is very quick to just whip up some code, although difficult to make production-ready code. Considering this was a learning experience, it was the perfect tool.

  • What are the use cases of this repository?

The main use case is as a reference for others learning about Quantum computing and interested in designing a quantum computer simulator. There are a lot of very complex simulators out there that are very difficult to grok because they have many optimizations in order to make them run more quickly. A quick glance through my code and you will notice that it is not not at all optimized, making it very easy to understand.

  • Where to go from here?

I would like to try implementing the quantum computer simulation in Haskell again. Figuring out a very succinct way to represent the qubit states would allow for a very simple-to-parse quantum computer simulator in Haskell.

Resources