--- title: "EPIQUANTI : Partie logiciel" date: 2021-10-05 14:00 categories: [tronc commun S9, EPIQUANTI] tags: [tronc commun, EPIQUANTI, S9] math: true description: Partie logiciel --- Lien de la [note Hackmd](https://hackmd.io/@lemasymasa/SJRoAnFEt) # Registers | **n bits register** | **n qubits register** | | ---------------------------------------- | --------------------------------------------- | | $\color{red}{2^n\text{ possible states } \textbf{once at a time}}$ | $\color{green}{ 2^n \text{possible states }\textbf{linearly superposed}}$ | | evaluable | partially evaluable | | independant copies | no copies | | individually erasable | non individualy erasable | | non destructive readout | value changed after readout | | deterministic | probabilistic | # Gates ## Classical logic gates ![](https://i.imgur.com/D2uCweD.png) Irreversible gates: - NAND - NOR - AND - OR *Quelle est leur consequence ?* > Comme on perd un bit, on a une **perte d'energie** > Decouverte par *Rolf Landauer* :::success Des gens travaillent aujourd'hui pour creer une informatique classique sans perte d'energie ::: ## Quantum gates :::info Matrix based reversible **unitary transformations** ::: ![](https://i.imgur.com/QddSjHl.png) - NOT: rotation $X$ - Rotation $Y$ $$ \begin{bmatrix} 0&-i\\ i&0 \end{bmatrix} $$ - Pauli-Z: rotation $Z$ - Hadamard: superposition ### Porte CNOT :::info On va changer la valeur d'un qubit en fonction d'un autre ![](https://i.imgur.com/mQz42Js.png) ::: *Mathematiquement, a quoi ca ressemble ?* $$ \begin{bmatrix} 1&0&0&0\\ 0&1&0&0\\ 0&0&0&1\\ 0&0&1&0 \end{bmatrix} $$ ![](https://i.imgur.com/2BSkpZE.png) *Si on intrique des qubits a des portes a 2 qubits, est-ce que ca reste ?* > Oui ## C2NOT ![](https://i.imgur.com/9vIlFMQ.png) $$ \begin{bmatrix} 1&0&0&0&0&0&0&0\\ 0&1&0&0&0&0&0&0\\ 0&0&1&0&0&0&0&0\\ 0&0&0&1&0&0&0&0\\ 0&0&0&0&1&0&0&0\\ 0&0&0&0&0&1&0&0\\ 0&0&0&0&0&0&0&1\\ 0&0&0&0&0&0&1&0 \end{bmatrix} $$ ## SWAP ![](https://i.imgur.com/cAxqf6h.png) $$ \begin{bmatrix} 1&0&0&0\\ 0&0&1&0\\ 0&1&0&0\\ 0&0&0&1 \end{bmatrix} $$ ## Fredkin :::info Conditional SWAP ::: ![](https://i.imgur.com/FIvsp0Y.png) $$ \begin{bmatrix} 1&0&0&0&0&0&0&0\\ 0&1&0&0&0&0&0&0\\ 0&0&1&0&0&0&0&0\\ 0&0&0&1&0&0&0&0\\ 0&0&0&0&1&0&0&0\\ 0&0&0&0&0&0&1&0\\ 0&0&0&0&0&1&0&0\\ 0&0&0&0&0&0&0&1\\ \end{bmatrix} $$ ## Single qubit operations visualization ![](https://i.imgur.com/fW0h0fw.gif) ## CNOT gate effect $$ \begin{matrix} \color{blue}{\text{control qubit}} &&\color{blue}{\text{tensor product of control and target qubits before CNOT}}\\ \alpha_1\vert0\rangle &&\alpha_1\alpha_1\vert00\rangle+\alpha_1\beta_2\vert01\rangle + \alpha_2\beta_1\vert10\rangle+\beta_1\beta_2\vert11\rangle\\ \bigotimes&\Rightarrow&\text{CNOT}\\ \alpha_2\vert0\rangle+\beta_2\vert1\rangle&&\alpha_1\alpha_1\vert00\rangle+\alpha_1\beta_2\vert01\rangle + \alpha_2\beta_1\vert11\rangle+\beta_1\beta_2\vert10\rangle\\ \color{blue}{\text{target qubit}}&&\color{blue}{\text{control and target qubits state after CNOT}}\\ \color{blue}{\text{control qubit is }\vert0\rangle}\\ \alpha_1=1&&\alpha_2\vert00\rangle+\beta_2\vert01\rangle\\ &\Rightarrow&\text{CNOT}\\ \beta_1=0&&\alpha_2\vert00\rangle+\beta_2\vert01\rangle\\ \end{matrix} $$ :::info CNOT is not changing the qubit ::: ## The EPR pair entanglemet building block Put control qubit into superposition state, then future gates act on 2 states simultaneously $$ \frac{\vert0\rangle+\vert1\rangle}{\sqrt 2} $$ ![](https://i.imgur.com/U6hQUlf.png) $$\biggr\}\frac{\vert00\rangle+\vert11\rangle}{\sqrt{2}}$$ Subsenquently, flipping a qubit in an entangled state modifies all of tis components ## Control-U gate On prend une porte U qui est une porte arbitraire ![](https://i.imgur.com/PYYoF2m.png) $$ \begin{bmatrix} 1&\dots&\dots&\dots\\ \dots&1&\dots&\dots\\ \dots&\dots&U_{11}&U_{12}\\ \dots&\dots&U_{21}&U_{22} \end{bmatrix} $$ # Qubit lifecycle - Initialization - $\vert0\rangle$ - Hadamard gate - $\frac{\vert0\rangle + \vert1\rangle}{\sqrt{2}}$ - Other gate - aubit vector turning around in Bloch sphere - Measurement - Measurement returns $\vert 0\rangle$ qith a probability $\alpha^2$ depending on the qubit state, then qubit state becomes $\vert0\rangle$ - Measurement returns $\vert1\rangle$ with a probability $\beta^2$ # Universal gates sets :::info **Jeu de portes universel** Jeu de portes *simples* qu'on peut combiner pour recreer toutes les transformations unitaires ::: > Ex: CNOT peut etre recree avec HZH > Three CNOT gates: one SWAP gate :::danger **Universal quantum computing** requires a T gate ($\frac{\pi}{4}$ rotation) ::: ## Getting confused with phase rotations - One round = $2\pi$ - $S=$ one quarter round $=\frac{\pi}{2}$ - $T=$ one eight roung ## Solovay-Kitaev theorem :::info **Theorem** Any desired gate can be approximated by a sequence of gates from an universal gates set. A quantum circuit of $m$ constant-qubit gates can be approximated to $\varepsilon$ error by a quantum circuit of $O(m\log^c(\frac{m}{\varepsilon}))$ gates from a desired finite universal gate set with $c=3,97$ For example, creating a $R_{15}$ gate requires $127$ H/Z/T gates ::: ## In other words :::success On veut appliquer a $n$ qubits n'importe quelle operation generique $U$, on enchaine une serie de transformations unitaires. ::: # $SU(2^n)$ - Space of unitaries on $n$ qubits :::info Espace contenant toutes les transformations ::: ![](https://i.imgur.com/NTCjIQu.png) # On reversibility :::info **All quantum gates are mathematically reversible**, this is a property of the matrix linear transformations ::: :::danger We could theortically run an algorithm and rewinf it entirely to return to the initial state, which could help recover port of the energy spent in the system ::: On a practical basis: - The gates are not physically and thermodynamically reversible due to some irreversible processes like micro-wave generations and DACs (digital analog converters) - The whole digital process taking place before micro-wave generation and after their readout conversion back to digital could be implemented in classical adiabatic\thermodynamically reversible fashion - Currently being investigated at Sandia Labs, Wisconsin University and with SeeQC ![](https://i.imgur.com/onkL3T5.png) # Inputs and outputs ![](https://i.imgur.com/1SeTmqH.png) ## Probabilistic or deterministic readouts ? :::info A single qubit measurement is probabilistic, ie: a qubit registered after a Hadamard gate applied to all qubits is a simple random numbers generator ::: On a practical basis: - the algorithm is executed many times, up to 8000 for IBM Q Experience - an average of qubits results is computed, producing a real number - the averahed result is theoratically deterministic - modulo the error generated by noise and decoherence # Basis, pure and mixed states ![](https://i.imgur.com/rHcKh3T.png) ## Examples ![](https://i.imgur.com/z70JJHi.png) > *Normalement vous avez rien compris* > [name=Olivier Ezratty] [time=Tue, Oct 5, 2021 3:55 PM] [color=#907bf7] ![](https://i.imgur.com/TzGs5Aw.png) :::success L'origine aleatoire du photon provient de la physique classique et non quantique ::: ## Single qubit mixed state ![](https://i.imgur.com/Mhd7O1E.png) # Toying with density matrices ![](https://i.imgur.com/fEpnGpS.png) # Qubits measurement :::info **Measurement** is using a collection ${M_m}$ of operators acting on the measured system state space $\vert\psi\rangle$, with probability of $m$ being: $$ p(m)=\langle\psi\vert M_m^✝M+m\vert\psi\rangle $$ ::: System state after measurement becomes: $$ \frac{M_m\vert\psi\rangle}{\sqrt{\langle\psi\vert M_m^✝M+m\vert\psi\rangle}} $$ with: $$ \sum_mM_m^✝M+m=1 $$ ## Various qubits measurement methods ![](https://i.imgur.com/ucEpkHD.png) # Computing semantics summary ![](https://i.imgur.com/6F6xAjC.png) # 5 DiVienzo criteria (IBM, 2000) ![](https://i.imgur.com/SuKkNyW.png) ## Main qubit types ![](https://i.imgur.com/5LfK2lw.png) # From lab to packaged computers Les ordinateurs quantiques actuels d'IBM: ![](https://i.imgur.com/ourihch.png) L'ordinateur version commerciale: ![](https://i.imgur.com/TGDouPp.png) > Il y a un cube derriere qui contient l'ordinateur IBM pense atteindre $1000$ qubits d'ici 2 ans, mais ca a pas trop l'air possible car au-dessus de $28$ qubits il y a une enorme perte de qualite. ## Inside a typical quantum computer ![](https://i.imgur.com/JLlYQaX.png) En resume: 4 composantes Avec des atomes froids, on n'aurait pas des compresseurs mais des pompes a ultra-vide. ## Chez Google ![](https://i.imgur.com/DJlYR8A.jpg) *Pourquoi les fils tournent ?* > Pour passer plus de temps dans le froid ? :::success Systeme de **dilatation thermique** du au changement de temperature hardcore - Refroidit: contracte - Rechauffement: dilate ::: *Pourquoi plusieurs etages ?* > On est a $300K$ a l'exterieur, on veut minimiser plusieurs poches > Chaque etage = une temperature > Chaque disque a une taille plus petite en descendant les etages, pour faire passer le moins de chaleur possible > Chaque etage est isole de celui au-dessus > Les fils sont des attenuateurs de puissance mais ils generent de la chaleur :::success C'est l'isolation thermique ::: ## Quantum computer architecture ![](https://i.imgur.com/iy4vWcY.png) ## Physical layout example ![](https://i.imgur.com/TND8OMC.png) ![](https://i.imgur.com/DzcqICU.png) # Error correction :::danger Each quantum gate and readout generate significant errors ::: Coming form decoherence generated by: - flip, phase and leakage error - calibration errors - thermal noise - electric and magnetic noise - gravity - radioactivty - vacuum quantum fluctuations - cosmical rays :::warning It accumulates with the number of quantum gates and qubits ::: ![](https://i.imgur.com/43XcjSf.png) ## QEC zoo ![](https://i.imgur.com/YVbY7p2.png)